WATER RESEARCH

A Journal of the International Water Association

Editor-in-Chief Guang-Hao Chen Yoshihiko Matsui
Hong Kong University of Science Hokkaido University
Mark van Loosdrecht & Technology Japan
Delft University of Technology China
and KWR Watercycle Research Armand Maul
Tom Curtis Université de Metz
The Netherlands University of Newcastle upon Tyne
E-mail: M.C.M.vanLoosdrecht@tudelft.nl France
UK
Francis de los Reyes III Max Maurer
Editors North Carolina State University Eawag
Jean-Claude Block USA Switzerland
Université H. Poincaré, Nancy I Ana Deletic How Yong Ng
France Monash University National University of Singapore
David Dixon USA Singapore
University of Melbourne Rob Eldridge Satoshi Okabe
Australia Glen Waverley Hokkaido University
Hiroaki Furumai Australia Japan
University of Tokyo Monica B. Emelko
Jong Moon Park
Japan University of Waterloo
Canada Pohang University of Science & Technology
Xiaodi Hao Korea
Beijing University of Civil Engineering and Rosina Girones
University of Barcelona Susan Richardson
Architecture
Spain US Environmental Protection Agency
China
Stephen Gray USA
Gregory Korshin Victoria University Miguel Salgot
University of Washington Australia
USA University of Barcelona
Kate Grudpan Spain
Anna Ledin Chiang Mai University
Formas David Sedlak
Thailand
Sweden University of California, Berkeley
Edwin Herricks
USA
Eberhard Morgenroth University of Illinois at Urbana-Champaign
ETH Zürich and Eawag, USA Jean-Philippe Steyer
Swiss Federal Institute of Aquatic Peter Hillis LBE-INRA
Science and Technology AECOM France
Switzerland Australia Masahiro Takahashi
Wolfgang Rauch Hong-Ying Hu Hokkaido University
University of Innsbruck Tsinghua University Japan
Austria China Nathalie Tufenkji
Maria Reis Bruce Jefferson McGill University
Universidade Nova de Lisboa/FCT Cranfield University Canada
Portugal UK
Kai Udert
Ulf Jeppsson
Hang-Shik Shin Eawag
Lund University
Korea Advanced Institute of Science Switzerland
Sweden
and Technology Vayalam Venugopalan
Korea Sergey Kalyuzhnyi
RUSNANO BARC Facilities
Thomas Ternes Russian Federation India
Bundesanstalt für Gewässerkunde
Jae-Hong Kim Eduardo von Sperling
Germany
Georgia Institute of Technology Federal University of Minas Gerais
Stefan Wuertz USA Brazil
University of California, Davis Günter Langergraber
USA Julie Zilles
University of Natural Resources and
University of Illinois at Urbana-Champaign
Applied Life Sciences, Vienna (BOKU)
Associate Editors Austria USA
Andrew Baker Shang-Lien Lo Anastasios Zouboulis
University of New South Wales National Taiwan University Aristotle University of Thessaloniki
Australia Taiwan Greece
Damien Batstone Dionisis Mantzavinos
University of Queensland Technical University of Crete Editorial Office
Australia Greece E-mail: wr-eo@elsevier.com

Publication information: Water Research (ISSN 0043-1354). For 2011, volume 45 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: For enquiries relating to the submission of articles (including electronic submission) please visit this journal’s homepage at http://www.elsevier.com/
locate/watres. Contact details for questions arising after acceptance of an article, especially those relating to proofs, will be provided by the publisher. You can track
accepted articles at http://www.elsevier.com/trackarticle. You can also check our Author FAQs at http://www.elsevier.com/authorFAQ and/or contact Customer Support
via http://support.elsevier.com.
Orders, claims, and journal enquiries: Please contact the Elsevier Customer Service Department nearest you:
St. Louis: Elsevier Customer Service Department, 3251 Riverport Lane, Maryland Heights, MO 63043, USA; phone: (877) 8397126 [toll free within the USA]; (+1) (314) 4478878
[outside the USA]; fax: (+1) (314) 4478077; e-mail: JournalCustomerService-usa@elsevier.com
Oxford: Elsevier Customer Service Department, The Boulevard, Langford Lane, Kidlington OX5 1GB, UK; phone: (+44) (1865) 843434; fax: (+44) (1865) 843970; e-mail:
JournalsCustomerServiceEMEA@elsevier.com
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: JournalsCustomerServiceJapan@elsevier.com
Singapore: Elsevier Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519; phone: (+65) 63490222; fax: (+65) 67331510; e-mail:
JournalsCustomerServiceAPAC@elsevier.com

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: members@IWAhq.org.uk; website: http://www.IWAhq.org.uk).
Registered Charity (England) No. 289269. Individual membership is available from £30 upwards.
For details contact IWA.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Model selection, identification and validation in anaerobic

digestion: A review

Andres Donoso-Bravo a,*, Johan Mailier a, Cristina Martin b, Jorge Rodrı́guez c,d,
César Arturo Aceves-Lara e,f,g, Alain Vande Wouwer a
a
Automatic Control Laboratory, University of Mons 31 Boulevard Dolez, B-7000 Mons, Belgium
b
modelEAU, Département de génie civil et genie des eaux, Université Laval, 1065 av. de la Médecine, Québec (QC) G1V 0A6, Canada
c
Department of Chemical Engineering, University of Santiago de Compostela, Spain
d
Masdar Institute of Science and Technology, Abu Dhabi, United Arab Emirates
e
Université de Toulouse, INSA, UPS, INP, LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France
f
INRA, UMR792, Ingénierie des Systèmes Biologiques et des Procédés, F-31400 Toulouse, France
g
CNRS, UMR5504, F-31400 Toulouse, France

article info abstract

Article history: Anaerobic digestion enables waste (water) treatment and energy production in the form of
Received 13 June 2011 biogas. The successful implementation of this process has lead to an increasing interest
Received in revised form worldwide. However, anaerobic digestion is a complex biological process, where hundreds
26 August 2011 of microbial populations are involved, and whose start-up and operation are delicate
Accepted 29 August 2011 issues. In order to better understand the process dynamics and to optimize the operating
Available online 3 September 2011 conditions, the availability of dynamic models is of paramount importance. Such models
have to be inferred from prior knowledge and experimental data collected from real plants.
Keywords: Modeling and parameter identification are vast subjects, offering a realm of approaches
Anaerobic digestion and methods, which can be difficult to fully understand by scientists and engineers
Modeling dedicated to the plant operation and improvements. This review article discusses existing
Identification modeling frameworks and methodologies for parameter estimation and model validation
Kinetic parameters in the field of anaerobic digestion processes. The point of view is pragmatic, intentionally
Sensitivity analysis focusing on simple but efficient methods.
ª 2011 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5348
2. Modeling procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5348
3. Mathematical models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5349
4. Experimental information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5350
4.1. Available measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5350
4.2. Experimentation mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5352
4.2.1. Batch operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5352
4.2.2. Continuous operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5353
4.2.3. Fed-batch operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5353

* Corresponding author. Tel./fax: þ32 065 374. 130.

E-mail address: andres.donosobravo@umons.ac.be (A. Donoso-Bravo).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.059
5348 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4

5. Structural adequacy and parameter identifiability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5353

6. Methods for parameters estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5354
6.1. Cost function selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5354
6.2. Optimization techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5355
6.2.1. Local methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5355
6.2.2. Global methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5356
6.3. What about optimization constraints? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5356
6.4. Alternative methods: The Bayesian inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5356
6.5. Some considerations for parameter estimation in AD models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5357
6.5.1. Steady states analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5357
6.5.2. Mass continuity (conservation laws) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5357
6.5.3. Initial conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5358
7. Parameter uncertainty estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5358
7.1. Error covariance matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5358
7.2. Confidence intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5358
7.3. Joint posterior distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5359
8. Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5359
8.1. Direct validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5359
8.2. Cross validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5359
9. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5359
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5360
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5360

1. Introduction understanding of the process or to predict the behavior of the

system. This issue has been addressed, in a more general
Anaerobic Digestion (AD) is a chain of interconnected biolog- context of environmental models, by Jakeman et al. (2006),
ical reactions, where the organic matter (in the form of from a in-depth theoretical point of view by Walter and
carbohydrates, proteins, lipids or more complex compounds), Pronzato (1997) and specifically for wastewater treatment
is transformed into methane, carbon dioxide and anaerobic processes by Dochain and Vanrolleghem (2001).
biomass, in an oxygen-free environment. This biological Several mathematical models of anaerobic digestion have
process is used to simultaneously treat waste and wastewater been proposed in the last two decades and a variety of
and to produce biogas. AD is now considered as a consolidated methods have been used for parameter estimation and model
technology with more than 2200 high-rate reactors imple- validation. AD process is characterized by its high complexity
mented worldwide (Van Lier, 2008). In Europe’s case, between and non-linearity and by the difficulty to collect large
1995 and 2010, the number of plants installed increased from amounts of informative experimental data for modeling
15 to 200, which implies an installation capacity rise of nearly purposes. One of the consequences of the latter is variety of
6,000,000 tons per year (from 200,000 to 6,000,000 tons per approaches to modeling and parameter identification is the
year) (de Baere et al., 2010). Moreover, the number of AD important variability in values reported for the kinetic
reactors is expected to increase due to both climate change parameters, even when the same operational and environ-
awareness and the significant boost in the use of renewable mental conditions have been evaluated.
energy. The main characteristics of the process, such as This paper presents an overview of the main procedures
reactor design issues, microbial aspects, inhibition that can be used for developing and assessing dynamic
phenomena, and of course, the strategic advantages of this models of the anaerobic digestion process. It is structured
process, are well described in the literature (Appels et al., 2008; according to a step-by-step approach of the modeling task,
Chen et al., 2008; Ward et al., 2008). Likewise, a thorough i.e., from model selection up to model validation.
description of the role of the microorganisms in the bioenergy
production, with a special emphasis in the anaerobic diges-
tion process, can be found in Rittmann (2008).
Mathematical models enable the representation of the
2. Modeling procedure
main aspects of a biological system. They improve the
understanding of the system, the formulation and validation
The whole modeling process (selecting a model structure,
of some hypothesis, the prediction of the system’s behavior
identifying the parameter values, and planning the experi-
under different conditions, reducing, consequently, the
mental measurements) should be in coherence with the
experimental information requirements, costs, risk and time.
objective pursued. In general, the three most common objec-
The proper evaluation and application of mathematical
tives of using a model are: understanding the system’s
models in bioprocesses must follow several stages if the final
behavior and interaction of components; quantitatively
goal of the approach is to generate useful tools to improve the
expressing or verifying our hypothesis and predicting the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
behavior of the system in the future or under other similar
circumstances.
Adequate model structures should be chosen according to
four principles (Spriet, 1985): (i) simplicity, the model should
be as simple as possible; (ii) causality, the model should
represent the most relevant causeeeffect relationships; (iii)
identifiability, the values of the unknown parameters should
be identifiable from the available data; and (iv) predictive
capability, the model should remain valid under future or
alternative reasonable conditions.
As stated by Flotats et al. (2003) model identification and
parameter estimation have not been given the same attention
in anaerobic digestion processes, as it does with activated
sludge systems in which considerable efforts have been
rightfully devoted (Ossenbruggen and Stevens, 1996; Weijers,
2003; Liwarska-Bizukojc and Biernacki, 2010). Fig. 1 shows
a schematic view of the parameter estimation and model
validation procedure. At first, it is of course very important to
define the purpose of the modeling exercise. An explicative
(mechanistic) model intended for process investigation and
hydraulic/chemical/biological analysis will likely include
a detailed description of specific mechanisms and
phenomena, which would probably be irrelevant for a global
dynamic analysis or the design of controllers, for instance
(Jakeman et al., 2006). Therefore, the level of details of the
description has to be selected with care depending on the
targeted application of the model (physical/chemical/biolog-
ical investigation, process design, dynamic simulation, opti-
mization, control, supervision).
Once an appropriate model structure has been selected
(usually a system of non-linear differential equations
including a number of unknown or uncertain parameters),
a simulator can be implemented using a platform of choice (a
programming language such as Fortran or C, or an environ-
ment such as Matlab or its open-source counterparts Octave
and Scilab). Local and global parametric sensitivity analysis
can then be used to assess, on the one hand, the most influ-
ential parameters, and on the other hand, the parameters
Fig. 1 e Parameter estimation procedure.
5349
with weaker influences on the measured outputs (at least in
the scenario under consideration), possibly involved in
correlation with other parameters. This first analysis can lead
to a model reformulation or simplification, eliminating
correlated parameters, and it can also lead to the reformula-
tion of a new experimental design and data collection
methods in order to have more informative data in relation
with the parameters to be estimated. Next, the experimental
data must be examined in terms of potential errors (outliers,
missing data, etc.) and the deviation between the model
prediction and the measured outputs. In this last part, special
attention has to be paid to the selected cost function. Finally,
the model has to be evaluated with regard to the experimental
data used so far (direct validation), as well as fresh data
(unseen in the identification process e cross validation). These
last two steps, if unsuccessful, can lead back to model refor-
mulation and/or experiment design and data collection.
3. Mathematical models
Mathematical modeling of the anaerobic digestion process
was motivated by the need for efficient operation of anaerobic
systems in the early 70’s. The first models were relatively
simple due to the limited knowledge about the process.
Experimental investigation, further system analysis and the
increase in computing capacity lead to the development of
much more detailed models in recent years. As it is not the
goal of this review to list the available models in anaerobic
digestion, a brief overview is given in the next paragraphs.
The first modeling approaches focused on describing the
limiting step of the process, considering that anaerobic
digestion is a multistep process where one slower step
controls the global rate (Hill and Barth, 1977). Such limiting
step can, however, be different under different operating
conditions (Speece, 1996). Some authors considered meth-
anogenesis as the limiting step or the conversion of fatty acids
into biogas or the hydrolysis of suspended solids (Eastman
and Ferguson, 1981). These series of models were simple and
easy to use but were unable to adequately describe the process
performance, especially under transient conditions.
A second generation of models considered the concentration
of volatile fatty acids as the key parameter, incorporating
acidogenesis and acetogenesis separately (Hill, 1982). The
hydrogen partial pressure, as a key regulatory parameter influ-
encing the redox potential in the liquid phase and more bacterial
groups, with differentiated acetoclastic and hydrogenotroph
methanogens, was included in several models (Costello et al.,
1991; Ruzicka, 1996). The redox potential (as NADH/
NAD þ ratio) is a function of the hydrogen partial pressure and
determines the VFA production in this family of models.
Further microbiological studies led to another generation
of models (Angelidaki et al., 1993, 1999; Siegrist et al., 1993;
Vavilin et al., 1994, 1995; Kalyuzhnyi and Davlyatshina, 1997;
Kalyuzhnyi, 1997; Kalyuzhnyi and Fedorovich, 1998; Batstone
et al., 2000; Tartakovsky et al., 2002; Keshtkar et al., 2003;
Haag et al., 2003). These models incorporated additional
processes and species, more detailed kinetics with inhibition
and considered different substrates.
5350 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
As a response to the need for a generic model of anaerobic
digestion, the IWA Task Group for Mathematical Modeling of
Anaerobic Digestion Processes developed the generic Anaer-
obic Digestion Model No.1 (ADM1) (Batstone et al., 2002) in
order to reach a common basis for further model development
and validation studies with comparable results. The ADM1
model describes the dynamics of 24 species and includes 19
bioconversion processes. The latter, makes ADM1 a model
with a large number of parameters. In view of its general
purpose, the ADM1 neglects some processes and species,
which are related to more specific applications, in order to
avoid extreme complexity. Still, the large number of param-
eters and identifiability difficulties are the major drawbacks of
ADM1, as well as, some structural weaknesses (Kleerebezem
and Van Loosdrecht, 2004).
Many applications based on the ADM1 have been pub-
lished in recent years. Some authors applied the model to
stirred tank systems while others considered distributed
parameter systems (Batstone et al., 2004a, 2005). Extensions
have been developed to incorporate processes that were
absent in the original model. Also, reports of the applications
of the ADM1 to particular types of wastewater have been
published (Batstone and Keller, 2003; Fedorovich et al., 2003;
Batstone et al., 2004b; Fezzani and Ben Cheikh, 2008; Fezzani
and Ben Cheikh, 2009; Derbal et al., 2009; Gali et al., 2009;
Lee et al., 2009; Ramirez et al., 2009; Ozkan-Yucel and Gok-
cay, 2010). The framework provided by the ADM1 is useful
especially for process design and dynamic simulation. Due to
its fixed stoichiometry approach, its applicability would,
however, require important structural modifications for some
processes. Implications of structural changes in some
processes of the ADM1 toward a variable stoichiometry
structure have been recently analyzed (Rodriguez et al., 2006).
Efforts have also been directed in recent years to simplify this
model (Siegrist et al., 2002; Rodriguez et al., 2008). In order to
ease the application of the ADM1 some methodologies have
been developed (Zaher et al., 2004; Kleerebezem and Van
Loosdrecht, 2006), as well as, some structural simplifications
of the model under certain conditions (Bernard et al., 2006).
Among the simplified models of the AD process, the one
developed by Bernard et al. (2001) has been used in different
applications. This model considers two-reactions (acido-
genesis and methanogenesis) and has been widely applied for
control purposes, for AD process optimization (Dalmau et al.,
2010) and for mathematical analysis (Dimitrova and
Krastanov 2009; Rincon et al., 2009; Sbarciog et al., 2010).
However, only few applications with data from lab- or full-
scale plants have been reported (Donoso-Bravo et al., 2009a;
Lopez and Borzacconi, 2009).
Several reviews of the existing models have been pub-
lished in the last two decades. Husain (1998) made a brief
review of steady state and dynamic models of the kinetics of
anaerobic digestion. Later on, Gavala et al. (2003) presented
a comprehensive review, describing from the simplest to the
most complex models. Following these general reviews, more
specific studies appeared in response to the increasing
number of available models. For instance, Tomei et al. (2009)
focused on models developed for the anaerobic treatment of
sewage sludge. Likewise, Batstone (2006) addressed anaerobic
digestion modeling in the framework of domestic sewage
systems. Other studies have focused on the type of reactor,
instead. For instance, Saravanan and Sreekrishnan (2006)
described the different available models for UASB (Up-flow
Anaerobic Sludge Blanket), AF (Anaerobic Filter) and EGSB
(Expanded Granular Sludge Blanket) reactors, in which the
biomass is attached to either a support or forming granules.
In Table 1, a summary of different studies where modeling
and optimization have been performed is shown.
4. Experimental information
4.1. Available measurements
Accurate and reliable measurements of key variables of the
process are very important. These data will be used for model
identification and validation therefore they should contain
the most relevant information at the lowest possible cost of
monitoring. Counting on several measured variables will
increase the possible parameters that can be reliably esti-
mated; however, and particularly in the case of large and
complex models as ADM1, identifying all the parameters and
coefficients is not feasible mainly due to the extreme difficulty
of separately identifying specific biomass concentrations from
the maximum specific uptake rates. Methanogenic applica-
tions, however, might require only very accurate identifica-
tion of a limited number of key parameters to provide good
results due to the dynamics of AD in which acidogenesis and
acetogenesis are much faster than methanogenesis and
hydrolysis (Rodriguez et al., 2006).
For any AD model in general, a proper structural identifi-
cation (discussed in Section 5), for instance using sensitivity
analysis methods together with knowledge of the process
dynamics, can play a key role in the success of the optimization
process to select the most relevant parameters. In general, two
types of data can be considered, those that come from off-line
or on-line measurements. Off-line sensors are those in which
the sample is taken usually manually (low sample frequency)
and analyzed by an operator; the data will be available after
hours or days. On-line measurements are attached to the
process and the analysis is automatic. The analysis is per-
formed under a high frequency, so data produced by these
sensors are considered continuous compared to processes that
use a time scale. These above-mentioned aspects must be
carefully taken into account since the quality of experimental
data, in terms of measurement error and sampling frequency,
will have a substantial influence on the parameter estimation
of the model (Guisasola et al., 2006).
To characterize the substrates and intermediates charac-
terization, the organic matter content (i.e. all the organic
compounds present in the solution) is usually calculated
through the chemical oxygen demand (COD), which is the
most common (off-line) measured variable. This measure,
and in a lesser extent, along with the total organic carbon
(TOC) analysis has been also used for this purpose. In order to
recognize the dynamic of specific variables, off-line tests are
usually used for volatile fatty acids (VFAs) and for the main
macromolecular compounds such as carbohydrates, proteins
and lipids. New on-line sensors have been developed which
are starting to be used more often (Molina et al., 2009; Boe
Table 1 e Summary and brief description of the studies found in literature about modeling and kinetic parameters identification in AD systems (IC: Initial conditions, KP:
Kinetic parameters, YC: Yield coefficients, PCC: physico-chemical constants, CF: Conversion factors).
Reference Model Estimates Measurements Estimation method Uncertainty

Batch
Batstone et al. (2009) ADM1 2 KP (hydrolysis) Biogasb Gradient search technique Confidence region
Lopez and Model for complex substratesa 7 KP Methaneb Multiple shooting Monte Carlo
Borzacconi (2010)
Palatsi et al. (2010) ADM1 3 IC, 4 KP Methane, Acetic, Non-linear weighted square minimization n.d.
butyric, propionic
acidc
Noykova and 3-reaction model 2 KP, 2 YC Biogasb Hooke and Jeeves optimization method n.d.
Gyllenberg (2000) Least-square
Muller et al. (2002) Monod and Non-competitive 1 IC, 2 KP, 1 YC Biogasb Non-linear weighted square minimization Monte Carlo
model
Lokshina et al. (2001) Monod and Haldane Equations 1 ratio (IC/YC), 1 YC, 3 KP Methanec Non-linear regression with the Marquardte Covariance matrix-FIM
Levenberg algorithm
Flotats et al. (2003) ADM1 3 KP, 2 YC Acetate, propionate, Combination of random direct search and Covariance matrix-FIM

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
valerate, methanec gradient methods
Continuous
Batstone et al. (2009) ADM1 2 KP (hydrolysis) Biogas,b VSSc Gradient search technique Confidence region
Bernard et al. (2001) 2-reaction model 4 KP, 6 YC, 1 PCC Methane,b Carbon Linearization at different steady states n.m.
dioxide,c COD,
VFA, Z, ICc
Haag et al. (2003) 3-reaction model 26 IC, 8 KP, 4 YC, 6 CF CODt, CODs, TOC, Directed search method followed by a Covariance matrix-FIM
MiS, TDE, ODE, VFAc gradient-based method and confidence interval
Batstone et al. (2003) ADM1 6 KP VFA, Biogas, pH, Secant method Confidence region
methane contentb
Lopez and Borzacconi 2-reaction model 6 YC Biogas,b COD, VFA, Least-squares criterion n.d.
(2009) gas compositionc
Ghaniyari-Benis 1-reaction model 1 KP CODc Non-linear regression using least-squares n.m.
et al. (2010) criterion
Bhunia and Ghangrekar Monod, Grau-2nd order and 3 KP, 1 YC, 1 CF COD, VSSc Linearization at different steady states n.d.
(2008) Haldane equations
Kalfas et al. (2006) ADM1 2 KP, 2 YC (meso- TSS, VSS, COD, VFA, Secant method using unweighted Confidence region and
thermophilic conditions) BIogas, gas least-square criterion linear confidence intervals
composition, pHc
Koch et al. (2010) ADM1 5 KP, 1 CF Biogas,b gas Evaluation of the modified Nash-Sutcliffe n.d.
composition, NH4, coefficient
NKT, VFA,
alkalinity, TSc
Initial rate
Donoso-Bravo 1 reaction model 2 KP Methanec Non-linear regression using Covariance-FIM
et al. (2011) (Monod Kinetic) least-squares criterion
Donoso-Bravo 1st order, Monod, Haldane 5 KP Carbohydrates, Non-linear regression using n.d.
et al. (2009b) equations. VSS, VFAc least-squares criterion

n.d. not determined, n.m. determined but not mentioned the used method.
a Proposed by Angelidaki et al. (1993).

5351
b on-line.
c off-line.
5352 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
et al., 2010; Ward et al., 2011; Jacobi et al., 2011). Specific
species concentration of the anaerobic biomass (state vari-
ables are unknown variables due to difficulties in performing
measurements of the concentration of each population.
Molecular biological techniques have been executed for these
purposes (Sanchez et al., 1994; Pobeheim et al., 2010));
however, these are costly and usually only qualitative infor-
mation can be drawn. This issue may trigger some identifi-
cation problems since some parameters cannot be determine
independently (Bernard et al., 2001; Noykova et al., 2002), and
may cause some inaccurate model predictions (Batstone et al.,
2004b). The volatile suspended solid content (VSS/l), which is
normally known, has been used to approximately estimate
the total biomass concentration, i.e., the sum of all the pop-
ulation involved (Bernard et al., 2001; Lopez and Borzacconi,
2009). In any case, this measure has not been used for
parameter fit but for model validation. A typical model
application, in this context, has been the use of software
sensors to estimate the concentration of each population
(Bernard et al., 2000; Lopez and Borzacconi, 2009). In regards to
the product of the reaction, the sum of all gas compounds
(formed during the process) is the most common on-line
performed measurements and consequently used widely in
modeling applications. The biogas measurement has been
used as the only measurement for parameters estimation in
plenty of articles. Independent measurement of the different
gases (CO2, H2 and H2S) is generally done off-line. Depending
on the model, the biogas production can be considered as
a state variable (Batstone, 2006; Keshtkar et al., 2003) or as
a dependent variable (Bernard et al., 2001). The biogas cumu-
lative volume is normally used in the case of using data from
batch test and biogas flow rate in the case of continuous
system. Biogas flow rate contains more information than
cumulative biogas volume and the latest can be derived from
the previous, by numerical integration. Instantaneous gas
flow rate is typically however more costly and for many
applications cumulative volume can also be used to derive the
instantaneous flow rate by derivation but the accuracy will be
affected by the frequency of sampling points available of
accumulated volume. Among other measurements, for
instance, the pH is a variable which is easily measured by on-
line sensor. In most of the model is used as an input in the
form of an inhibition function (Angelidaki et al., 1999;
Batstone, 2006; Haag et al., 2003; Keshtkar et al., 2003), but
also can be found as an output of the model to be used for
validation (Bernard et al., 2001). So far, no models have used
pH for parameter identification; firstly, because it is used as an
input and secondly, because pH may present a low sensitivity
in well-buffered systems.
4.2. Experimentation mode
Understanding the nature of the experimental data is a crucial
point when making good use of the modeling procedure. This
section analyses the different experimental conditions in
which anaerobic tests are carried out. Two main issues have
to be considered: 1. culture history, and 2. the selected oper-
ation mode of the anaerobic digestion process.
The culture or inoculum history, involves the specific
characteristics of the anaerobic biomass used for the assay,
which in the case of anaerobic digestion systems is normally
a mixed culture. The crucial factor is the manner in which the
culture has been developed since it determines which species
are predominantly present which also influences the physio-
logical state. It is clearly different if the anaerobic inoculum
comes from a continuous reactor (where organisms with
higher affinity enzymes are favored), if it comes from a batch
reactor, or if the inoculum has been exposed to specific
conditions for some time (microorganisms have the ability to
change their macromolecular composition as a result of
physiological adaptation). Usually, the culture history cannot
be easily modified and in most cases, the tests are simply
carried out with the available inoculum. However, it is of
prime importance to document the specific conditions of the
inoculum culture since the results are likely to be valid only
under these experimental culture conditions.
The operation mode of the assay has a paramount influ-
ence on the information content of the collected data, and
thus, on the quality of the estimated parameters. The variety
of operational conditions explain the variability of the re-
ported parameter values (Grady et al., 1996), and implicitly
shows that either the studies have considered too limited
experimental data information or have not properly applied
the parameter estimation procedure (otherwise, more refer-
ence parameter sets would have been published and vali-
dated). Batch assays are commonly used in AD for kinetic
parameter determination, even though other types of opera-
tion modes have also been employed for these purposes. This
situation is rather unfortunate since it has been demonstrated
that the parameters of a simple Monod law cannot be
uniquely determined from a batch experiment (Baltes et al.,
1994). It is therefore very unlikely that the complex kinetics
of AD could be determined from batch tests only, which will
become clearer after the next section’s explanation on
parameter sensitivity analysis.
4.2.1. Batch operation
Batch operation can be defined as a biological process in
which there is no interchange of mass with the environment,
i.e. there are no input or output flow (except for the gas
stream). All substrates and nutrients are added at the
beginning of the reaction cycle. Several advantages have
been stated with regards to the use of batch tests (often
known as the biochemical methane potential test, BMP) for
kinetic parameter determination in AD, such as: (1) the
possibility to easily record the time evolution of several
variables (2) the relatively short time span as compared to
continuous operations and (3) simplicity and popularity as
reflected in a wide acceptance (Noykova et al., 2002; Flotats
et al., 2003; Batstone et al., 2009; Lopez and Borzacconi,
2010). However, the main drawback of batch tests, stems
from the lack of input excitation (since the only input is the
initial condition) resulting in a lack of parameter sensitivity
(Lokshina et al., 2001). This can be partly alleviated using
different sets of initial conditions (Flotats et al., 2003, 2006)
and determining a proper range of substrate/biomass (S/X )
ratio (Grady et al., 1996). The S/X factor, whose inverse is also
used in some studies (X/S: inoculumesubstrate ratio or IRS,
(Raposo et al., 2009), may influence the correlation between
some parameters, for instance, mm and Ks of the Monod-
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
equation, and thus, affect the results of parameter identifi-
cation. Despite its importance, the value of S/X is seldom
mentioned in the studies where batch tests are used for
parameter estimation.
Whereas BMP tests are widely used for parameter estima-
tion, most full-scale reactors operate in continuous or semi
continuous conditions which are drastically different condi-
tions. This explains the proposal to use alternative procedures
such as the initial rate reaction measurement, which has been
widely applied in enzymatic processes. It is used to obtain
either production or consumption rates of a specific
compound involved in a reaction on a short time span (Illanes,
2008). Donoso-Bravo et al. (2009b) used the initial rate
measurement of the substrate degradation (starch, glucose
and VFAs) to evaluate the influence of the temperature on the
main reactions of the anaerobic digestion, just as Flotats et al.
(2003) estimated kinetic parameters of anaerobic degradation
of gelatin using the same technique. According to Donoso-
Bravo et al. (2011) this method can be an interesting alterna-
tive to classical batch tests, as it allows to alleviate inhibitory
effects of byproducts or substrate limitation (which are more
likely to occur in batch than in continuous operation). The
main drawback of this technique is the lack of research and
validation since only a few studies have used initial rate tests
in AD applications.
4.2.2. Continuous operation
In this operation, spent medium or digestate is continuously
replaced with an equal volume of fresh medium (substrate
solution) and therefore a continuous discharge of biomass
also occurs. Continuous systems also offer a proper platform
for kinetic analysis, as long as a series of experiments at
different dilution rates (D) are carried out. Overall, it is a more
time-consuming method than a batch test and, therefore, the
kinetic parameter calculation is usually performed with data
from continuous anaerobic reactors which are already oper-
ating. For instance, more than 3 months of a pilot up-flow
fixed-bed anaerobic reactor operation were required by
Bernard et al. (2001) or 1.5 year for full-scale anaerobic digester
by Batstone et al. (2009) in order to perform parameter esti-
mation. A less time consuming and appealing alternative is to
evaluate the dynamic response of a continuous reactor after
specific substrate pulses (Batstone et al., 2003; Kalfas et al.,
2006). This method allows the estimation of the kinetic
parameters of specific compounds since the pulses provide
some decoupling of the biological phenomena, and in turn
lower parameter correlation, and thus better identifiability.
The pulse amplitude has to be selected so that the substrate
concentration crosses the affinity-saturation constant
(Batstone et al., 2003). The main drawback of this pulse-based
methodology is that it cannot be applied at full-scale and it
may be expensive at laboratory scale. New approaches in
parameters estimation in this field try to combine data from
both continuous and batch experiments (Girault et al., 2011).
4.2.3. Fed-batch operation
In this process, substrate and nutrients are added continu-
ously or intermittently into the reactor, without an output
stream from it, so that the volume of the reaction media
increases during the cycle of operation. Fed-batch operations
5353
are scarcely used in AD, neither at full-scale nor at a pilot/
bench/lab-scale. Hence, only a few studies have used these
systems for kinetic parameter determination. Rodrigues et al.
(2003) evaluated the behavior of a fed-batch reactor by fitting
some apparent parameters with a simple global model of AD.
Likewise, Redzwan and Banks (2004) used a simplified model
in order to assess the kinetic of methane production in a fed-
batch reactor. Effective volume of a fed-batch reactor always
increases, whereas it remains constant in batch or continuous
operation; this may make the mathematical analysis of the
system more complicated. In addition, the measurement of
the biogas flow has to be corrected to take this volume change
into account. These issues may represent significant draw-
backs for the use of fed-batch operation in kinetic parameter
determination.
5. Structural adequacy and parameter
identifiability
After the formulation of the modeling objectives and data
collection, a double question arises:
a) Is the selected model structure able to fit the data? To
achieve this objective, the model has to include the
necessary degrees of freedom, but not too many as there is
a risk of overparametrization. This risk is linked to the
other side of the question.
b) Once a model structure is selected, is it possible to deter-
mine a unique optimal set of parameters based on the
experimental data at hand?
This double-sided analysis can lead to model simplifica-
tions or, on the contrary, to the introduction of additional
terms or equations, and in turn to the elimination or the
introduction of some parameters.
In general, AD models are mechanistic models that
synthesize extensive scientific research work dedicated to
understand most of the physical, chemical and biological
mechanisms of the processes involved. As consequence, most
of the model parameters have some physical meaning and
generally some default values are available (Batstone et al.,
2002). The identifiability problem is then a delicate issue
where the modeler should calibrate only those parameters
necessary to explain the observed mechanisms without
“overfitting” the data, i.e.: an “overcalibrated” model would
reproduce the experimental data pretty well but would lose
predictive or exploration capability (Reichert, 2010).
These structural and parametric identifiability questions
are relatively seldom addressed in the reported AD modeling
studies. Taylor Series Expansion is the most reported tech-
nique. It consists of calculating the successive derivatives of
the output function with respect to the unknown model
parameters, so as to obtain a system of independent equa-
tions in these parameters. Flotats et al. (2003) applied this
approach with four of the state variables of ADM1 (Acetate,
propionate, valerate and methane) so as to design an exper-
iment for the identification of parameters related to the
anaerobic degradation of valerate and the initial biomass
concentrations. Noykova et al. (2002), using a more simplified
5354 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
model, evaluated identifiability with respect to three
unknown parameters based on the measurement of the
biogas production. In this case, the vector of parameters was
reduced in order to decrease the computational requirement
of this method. In fact, this method is mostly applicable to
simple models, whereas it leads to complex systems of
equations in more general cases (Muller et al., 2002).
Another approach to the study of identifiability is based
on local and global sensitivity analysis. The goal of sensitivity
analysis is to explore the change in model output resulting
from a change in model parameters (kinetic or stoichiometric
coefficients, parameters, input conditions, initial conditions,
etc.) (Sin et al., 2011). Most of the sensitivity analysis tech-
niques encountered in literature, for AD process, are of local
nature and use a differential analysis of outputs with respect
to parameters (eq. (1)):
vyj yj ðqi Þ  yj ðqi þ Dqi Þ
¼ (1)
vqi Dqi
Where yj corresponds to the jth output and qi to the ith
parameter. Examples of local sensitivity analysis in AD
modeling can be found at Tartakovsky et al. (2008) and
Noykova and Gyllenberg (2000). The main drawback of this
method is that it is based on the linearization of the model
equations at a given set of parameter values and therefore it
only describes local model behavior at this point.
Other authors attempt to get a more global picture of the
sensitivity by varying the parameters (one at a time) and
aggregating the relative difference observed in the outputs
(Vavilin et al., 2003), either by integrating the errors in time
(Bernard et al., 2001), or by estimating a weighted sum of them
(Wichern et al., 2009; Lin and Wu, 2011). However, none of
these methods are able to detect correlations among the
parameters and aggregating the errors can lead to erroneous
conclusions when compensations between negative and
positive terms occur.
An alternative definition of sensitivity analysis is the so
called “global sensitivity analysis (GSA)” and relates to
uncertainty analysis. It can be viewed as an analysis of
variance (ANOVA) problem (Sobol, 2001; Helton and Davis,
2003; Saltelli et al., 2006). Hence the output variance is
decomposed into fractions which are attributed to the single
model inputs. Examples of such sensitivity analysis methods
include Morris Screening (Morris, 1991), the spectral infor-
mation of measurements to characterize parameter interac-
tions (Tarantola et al., 2006), linear regression of Monte Carlo
outputs (Helton and Davis, 2003) and variance decomposition
(Saltelli et al., 2008). Global sensitivity analysis has recently
been applied to biological models of plant cell cultures
(Mailier et al., 2011) and to activated sludge systems (Sin
et al., 2011) and it offers promising perspectives in AD
models.
6. Methods for parameters estimation
Unfortunately, AD models are not universal enough and some
parameters need to be estimated for each particular case
study. Traditionally, they have been calibrated by a trial and
error approach. However, this method is very time consuming
and does not provide any information about the uncertainty
associated to the parameter values nor any guarantee about
its uniqueness. The selection of the objective function (usually
also called cost function), which can play a crucial role in the
result of the optimization, will be reviewed in the first part of
this section and will then follow with the most used tech-
niques for parameter estimation.
6.1. Cost function selection
In order to find the best-fit of a model to given experimental
data, an appropriate criterion for the optimal solution of the
model parameter vector must be selected. Several cost func-
tions have been used for parameter identification in AD
models mostly in the form of output-error criteria, i.e.,
measuring the deviation between the model and real system
outputs. The type of selected function may influence how the
optimization procedure acts and how it adjusts the parame-
ters (Batstone et al., 2003).
The most popular cost function is the sum of least squares
(OLS, eq. (2)) (Noykova and Gyllenberg, 2000; Bhunia and
Ghangrekar, 2008; Batstone et al., 2009; Donoso-Bravo et al.,
2010; Lopez and Borzacconi, 2010), where it is implicitly
assumed that the standard deviation of the measurement
errors, which can be known or unknown, is constant. In eq. (2),
J is the objective function, nexp are the collected measure-
ments, nsim are the model-predicted outputs, q represents the
parameters to be determined (which can include the stoichi-
ometry, the kinetic parameters and also the unknown initial
conditions of some experiments) and N is the number of
measurements. Minimizing the cost function has been
acknowledged as an important issue for prediction purposes
or process stability (Batstone et al., 2003).
X
N
 2
JðqÞ ¼ min vexp ðtÞ  vsim ðt; qÞ (2)
t¼1
When the measurement errors do not have a constant
standard deviation, then it is generally required to introduce
weighting factors wt into eq. (2), leading to a weighted least-
square criterion
X
N
 2
JðqÞ ¼ min wt vexp ðtÞ  vsim ðt; qÞ (3a)
t¼1
in scalar form, or more generally, when vectors of measure-
ments are considered,
X
N
   
JðqÞ ¼ min vexp ðtÞ  vsim ðt; qÞ W vexp ðtÞ  vsim ðt; qÞ (3b)
t¼1
where W is a N  N weighting matrix to be selected.
If the measurement errors are white and normally
distributed, i.e. ε w N(0,Q), then the best choice of the
weighting matrix W, in a maximum likelihood sense, is the
inverse of the covariance matrix of the measurement noise,
i.e. W ¼ Q1.
If these assumptions do not hold or a deterministic
approach is preferred, some other weighting could be used. A
variety of weightings have been used in published studies
(Smith et al., 1998; Lokshina et al., 2001; Noykova et al., 2002;
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
Flotats et al., 2003; Palatsi et al., 2010). For instance, the
weighting factors have been estimated by calculating a local
slope of the output variation (Lokshina et al., 2001), by the
difference between maximum and minimum values (Palatsi
et al., 2010) or by using the deviation with respect to the
mean (Flotats et al., 2003).
A frequent situation corresponds to constant, possibly
unknown, relative errors. In this case, eq. (3a) can simply be
formulated as:
N 
X 
vexp ðtÞ  vsim ðt; qÞ 2
JðqÞ ¼ min
t¼1
vexp ðtÞ
or, noting that a constant absolute error on a logarithm, is
equivalent to a constant relative error on its argument:
X
N
   2
JðqÞ ¼ min ln vexp ðtÞ  ln ðvsim ðt; qÞÞ
t¼1
as it has been used successfully in several studies, e.g.
(Batstone et al., 2003; Haag et al., 2003; Vande Wouwer et al.,
2006)
When a cost function has been formulated, a numerical
procedure has to be used to minimize it with respect to the
unknown parameters.
6.2. Optimization techniques
In order to avoid the tedious trial and error approach, several
algorithms have been developed. They are search techniques
that numerically approach the optimum parameter values by
optimizing an objective function. These algorithms can be
divided into local algorithms and global algorithms.
6.2.1. Local methods
The vast majority of optimization methods are local in nature,
i.e. they assume the convexity of the cost function. When this
condition is not fulfilled, and a local method is applied, there is
a high risk that the algorithm will get trapped into a local
minimum. To alleviate this problem, it is recommended to
start the search from several randomly selected initial
parameter values so as to explore the parameter space. This
procedure, called multi-start strategy (Kocsis and György,
2009), allows the assessment of the problem multimodality
and, for instance by drawing a histogram of the frequency of
occurrences of the different minima, to determine the global
minimum and its basin of attraction (i.e., the ball-sized region
containing the initial guesses leading to the global optimum).
As initialization is of paramount importance, it is advised
to decompose a complex optimization problem into several
simpler ones, whenever possible, and to use the solution of
these intermediate problems as initial guesses for the next
step. This type of procedure has been successfully applied in
the identification of bioprocess models by (Hulhoven et al.,
2005).
Among local methods, one basically distinguishes
gradient-based methods, which make use of the first- and, in
some cases, the second-order derivatives of the cost function,
and gradient-free methods which do not require the cost
function differentiability such as the direct-search methods.
Another important feature of the optimization problem is the
5355
presence of equal and inequal constraints. These methods are
briefly described in the sequel with respect to their application
to the estimation of parameters in AD models.
6.2.1.1. Simple unconstrained optimization: steepest descent,
GausseNewton and LevenbergeMarquardt methods. The
steepest descent is an iterative method which uses first-order
information to move downhill in the gradient direction. A
large number of iterations are often required to achieve
convergence. GausseNewton method is particularly well
suited to the minimization of sum of squares (and therefore to
(3c)
the least-squares approach) and avoids the costly evaluation
of the Hessian (second-order information) by building an
approximation based on the Jacobian while preserving the
quadratic convergence of the original method of Newton.
The LevenbergeMarquard method (LMA) (Marquardt, 1963)
(3d) blends the two previous methods, steepest descent and
GausseNewton. LMA usually starts using a steepest descent
method and progressively becomes a GausseNewton method
as it gets closer to the optimum. This way, the algorithm is
more robust than GausseNewton but achieves better
convergence than steepest descent. LMA has been commonly
applied to parameter identification in AD models for the
treatment of livestock manure (Garcia-Ochoa et al., 1999), raw
industrial wine distillery vinasses (Aceves-Lara et al., 2005;
Martin et al., 2002), baker’s yeast effluents (Deveci and Ciftci,
2001), low temperature acetoclastic methanogenesis
(Lokshina et al., 2001) and animal wastes from calf farms
(Simeonov 1999). Aceves-Lara et al. (2005) combined this
algorithm with an asymptotic observer to evaluate the
parameters kinetics.
6.2.1.2. Non-linear constrained optimization: sequential
quadratic programming. Sequential Quadratic Programming
(SQP) is one of the most successful methods for the numerical
solution of constrained non-linear optimization problems.
SQP (Nocedal and Wright, 2006) is an iterative method which
solves, for each iteration, a quadratic problem (QP), i.e.,
a quadratic approximation of the objective function subject to
a linearization of the constraints. If the problem is uncon-
strained, then the method reduces to the Newton method. For
solving the QP problem under inequality constraints, a variety
of methods are commonly used, including among others
interior point, active set. SQP has been used for parameter
estimation in AD models (Sales-Cruz and Gani, 2004; Aceves-
Lara et al., 2005).
6.2.1.3. Multiple shooting. In the previous methods,
a sequential optimization approach has always been
assumed, i.e., the optimization algorithm repeatedly evalu-
ates the cost function by a call to a time integrator which
numerically solves the dynamic equations of the process
model (for instance, an LMA algorithm evaluates the cost
function through the solution of the model differential equa-
tions e which depends on the current values of the model
parameters e using a Runge-Kutta method).
There is another family of methods which rather use
a simultaneous approach, i.e., discretize the model differen-
tial equations and uses the resulting set of algebraic equations
as constraints to the optimization algorithm. Multiple
5356 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
shooting (Ascher et al., 1995) discretizes the time span into
time intervals (ti, tiþ1) and new optimization variables are
introduced which correspond to initial conditions of the state
variables on each interval. A set of boundary conditions
ensures the continuity of the solution. As time integration is
performed over short time intervals, the numerical stability
property of the algorithm is improved. In addition, state
constraints can be easily incorporated. Recently, multiple
shooting has been used for parameter estimation in AD
(Muller et al., 2002; Lopez and Borzacconi, 2010).
6.2.1.4. Direct-search methods. Direct-search methods (Lewis
et al., 2000) are derivative-free methods, which do not even
require numerical function values since the relative rank of
objective values is sufficient. Several classes of methods exist,
such as pattern search methods, simplex methods, and
methods with adaptive sets of search directions. These
methods date back to the 60’s and have since been replaced by
more sophisticated techniques. However, they still have an
undisputed popularity due to their simplicity of use and good
performance in practical use. In engineering applications, the
simplex methods have always been in wide use.
The basic idea of simplex search is to construct a non-
degenerate simplex in the parameter space and use the
simplex to drive the search (a simplex is a set of n þ 1 points in
the n dimensional space, e.g. a triangle in 2D; a non-
degenerate simplex is one for which any point in the
domain of the search can be constructed by taking linear
combinations of the edges adjacent to any given vertex).
Not only does the simplex provide a frugal design for
sampling the space, it has the added feature that if one
replaces a vertex by reflecting it through the centroid of the
opposite face, then the result is also a simplex. It means that
one can proceed parsimoniously, reflecting one vertex at
a time, in the search for an optimizer.
The simplex algorithm is usually less sensitive to local
minima than the gradient-based methods, such as the Lev-
enbergeMarquardt method. However, the convergence is
usually slower and closer to the optimum, and the algorithm
of course does not provide any sensitivity information (Jaco-
bian) that could be used as a byproduct to estimate the Fisher
Information Matrix; as will be introduced in the sequel.
The simplex algorithm has been widely applied to param-
eter estimation in AD models (Mosche and Jordening, 1999;
Simeonov 1999; Ruel et al., 2002; Haag et al., 2003; Guisasola
et al., 2009; Lopez and Borzacconi, 2010)
6.2.2. Global methods
Non-linear parameter identification problems are often char-
acterized by the presence of various local minima. Global
optimization is aimed at finding the best solution to these
kinds of problems. This is a very active research area and two
main families of methods have emerged: deterministic algo-
rithms, including for instance grid search and branch and
bound, and stochastic algorithms including for instance
simulated annealing, tabu search, genetic algorithms, differ-
ential evolution, ant colony optimization and particle swarm
optimization.
Among all these methods, Simulated Annealing (SA),
Genetic Algorithms (GA), Particle Swarm Optimization (PSO)
have been quite popular in engineering applications. SA is
a probabilistic algorithm, whose initial idea comes from
annealing in metallurgy, a technique involving heating and
controlled cooling of a material to increase the size of its
crystals and reduce their defects. The objective function to be
minimized is analogous to the internal energy of the system.
GA belong to the larger class of evolutionary algorithms (EA),
which generate solutions to optimization problems using
techniques inspired by natural evolution, such as inheritance,
mutation, selection, and crossover. PSO are search algorithms
based on the simulation of the animal social behavior in
a group.
The advantages of these algorithms is that they do not
require the objective function to be differentiable as in classic
local (gradient-based) optimization algorithms, which make
few assumptions about the problem to be solved, and can
explore a large space of candidate solutions. Although, these
algorithms produce better solutions they do not guarantee
that an optimal solution will ever be found at the price of large
amounts of computation.
These algorithms have found applications in the identifi-
cation of AD models, for instance Simulated Annealing (Haag
et al., 2003), Genetic Algorithms (Jeong et al., 2005; Abu Qdais
et al., 2010; Wichern et al., 2009), and Particle Swarm Optimi-
zation (Wolf et al., 2008).
6.3. What about optimization constraints?
Constraints in the estimated parameter values are usually
employed in the optimization process as long as they are
allowed by the selected technique. Simple and logic
constraints are the most used ones, such as: positive values
and within certain reasonable ranges (i.e., parameters
between some minimum and maximum values that the
experimenter could determine from past experience). Never-
theless, sometimes linear and non-linear constraints can be
useful during parameters estimation since they enable the
inclusion of conservations laws (i.e. yields) and avoid some
mathematical uncertainties linked to the model’s structure
(i.e. pH and gas transfers). In other cases, it is practical to
estimate separately some parameters, e.g. volumetric coeffi-
cient of mass transfer (kLa), with iterative linear approxima-
tions in order to simplify the estimation task (Batstone, 1999).
On the other hand, when using sophisticated non-linear
constrained algorithms is not necessary, and simple methods
such as the simplex can be used, some transformation of the
parameter space can be done. For instance, positivity
constraints can be imposed using a logarithmic trans-
formation, i.e. if q ¼ ln (l) is a positive unknown parameter,
then the optimization algorithm explore the full real param-
eter space (positive and negative) for l. This technique has
been applied for instance in Vande Wouwer et al. (2006).
6.4. Alternative methods: The Bayesian inference
The Bayesian approach opens a new calibration framework
because it abandons the idea of believing that the model
parameters are fixed and not known, and treats them as
probabilistic (or random) variables having a probability
density function. This function is called joint posterior
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
distribution of parameters and defines subjective beliefs of
parameter values by summarizing the state of knowledge
about the system performance (Omlin and Reichert, 1999).
Equation (4) shows a very general expression of the Bayes
formula:
pðyjqÞ  pðqÞ pðyjqÞ  pðqÞ
pðqjyÞ ¼ ¼Z fpðyjqÞ  pðqÞ
pðyÞ
pðyjqÞ  pðqÞdq
q a 2-reaction model in the AD of wine distillery wastewater.
where q is the model’s parameters vector and y represents
experimental data.
The Bayes theorem states that the posterior probability
function, p(qjy), is proportional to the multiplication of the
prior beliefs, p(q), and the likelihood function of observations,
p( yjq). The prior probability distribution, p(q), expresses
modeler prior beliefs about possible parameter values while
the posterior, p(qjy), expresses the posterior beliefs after
having evaluated the model residuals. The likelihood function
p( yjq) plays a very important role in Bayes’ formula because it
is the function through which the data y modifies prior
knowledge of q. Finally, the probability of the observations,
p( y), is the expected value of the likelihood function over the
parameter space, and acts as a normalizing constant.
Qian et al. (2003) explained that the most important limi-
tation of using Bayesian methods for scientific inference was
that analytical solutions of the posterior distributions are
available for fairly limited combinations of model forms and
probability distributions (such as the linear model leading to
a normal distribution of the residuals). For most non-linear
models, or models with a large number of parameters to be
calibrated, the estimation of the posterior likelihood becomes
intractable. Fortunately, advent of fast and inexpensive
computing has promoted the implementation of numerical
techniques. Particularly important are the Markov Chain
Monte Carlo (MCMC) techniques that construct a Markov
chain which asymptotically converges into the posterior
distribution.
The Bayesian techniques are especially advisable in the
case of poor parameter identifiability because subjective prior
knowledge about possible parameter values can be used.
Omlin and Reichert (1999) stress that in environmental
modeling the use of complex model structures and limited
experimental data makes the Bayesian techniques very
important. An interesting application of Bayesian calibration
of an AD model has been recently proposed by Martin et al.
(2011). In this example, a digester model (De Gracia et al.,
2009) able to work under anaerobic and aerobic conditions is
calibrated by using a complete so-called Integrated Monte
Carlo Methodology (Martin and Ayesa, 2010).
6.5. Some considerations for parameter estimation in
AD models
The complexity and particularities of AD processes give rise to
some considerations when trying to calibrate a model.
6.5.1. Steady states analysis
When the selected model is simple and some convenient
assumptions are considered (mainly that the system has
5357
reached a steady state condition) some mathematical modi-
fications can be done in order to get rid of the ODE and to
express the outputs of the systems as a functions of other
variables. Once these expressions are obtained, simple linear
regression may be used to estimate the model parameters.
Bernard et al. (2001) used this approach to draw some
(4) expressions in order to estimate several kinetic parameters,
stoichiometric coefficients and one mass transfer constant of
Likewise, Simeonov et al. (1996) employed steady state anal-
ysis to estimate some kinetic parameters of a 3-reaction
model in the AD of different type of animal waste at lab-
scale. Both studies evaluated the models with 7 steady
states varying either the dilution rate or the organic matter
concentration in the influent. The main drawback of this
method is that reaching steady state conditions requires
a highly controlled reactor, which is quite difficult to achieve
in a full-scale reactor because of the flow and concentration
disturbances.
In this context, Bhunia and Ghangrekar (2008) assessed the
application of the linearized form of the non-linear expression
drawn from the steady states analysis with three simple
models (Monod, Haldane and second-order functions) in the
treatment of synthetic sucrose-based wastewater using lab-
scale UASB reactors. Overall, non-linear optimization
showed better performance than linearization.
6.5.2. Mass continuity (conservation laws)
The problem of mass continuity in the context of AD has been
analyzed (Banks et al., 2011; Ekama et al., 2007; De Gracia et al.,
2006; Huete et al., 2006). Tracking the mass trajectories of the
elemental compounds (C, H, O, N and P) can uncover hidden
processes when studying experimental data or identify inad-
equate model structures when analyzing model results. Con-
cerning the latter objective, De Gracia et al. (2006) proposed
a mass and charge conservation check methodology to verify
the consistency of AD models. On the other hand, the equa-
tions of the mass continuity can represent by themselves
biochemical transformation models. This is the case of Zaher
et al. (2009) where a simple AD model is presented to study the
microbial activity in the treatment of dairy manure.
The study of the energy balance is also crucial when
analyzing the process’s performance. Banks et al. (2011) pre-
sented a complete energy study to assess the performance of
the anaerobic digestion of source-segregated domestic food
waste. In the modeling field, De Gracia et al. (2009) incorpo-
rated the energy balance equation to model the digestion of
sludge generated in wastewater treatment plants. Lübken
et al. (2007) proposed a modified version of ADM1 model to
simulate energy production in the digestion of cattle manure
and renewable energy crops. A thermodynamic analysis of the
acidogenic reaction was performed by Bastidas-Oyanedel
et al. (2008), where was demonstrated that the energy trans-
fer efficiency is influenced by some operational conditions,
such as pH and hydraulic retention time.
The mass continuity equations have also been used to
characterize input conditions in terms of substrate charac-
teristics (Huete et al., 2006). In the case of sludge produced by
wastewater treatment plants, and when trying to characterize
it in terms of the ADM1 model, Huete et al. (2006) points out
5358 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
that “the biggest uncertainty comes from the elemental
composition of the organic composites and the inert soluble
and particulate components, since the elemental mass frac-
tions depend on the specific case under study”. When
approaching the same problem, Ekama et al. (2007) found that
the non-biodegradable particulate organics, including both (a)
those originated in influent wastewater, and (b) those gener-
ated in the activated sludge endogenous process, are also non-
biodegradable under anaerobic conditions.
Grau et al. (2007a) proposed a new model building philos-
ophy called plant wide modeling which indirectly addresses
the characterization problem of sludge. It proposes to
generate a list of non-redundant model components for the
biochemical processes and to define them in terms of C, H, N,
O, P, charge and other possible elements. This extensive
description allows for a straightforward relationship between
the model components and the most common analytical
measurements carried out in the wastewater sludge. Using
this approach De Gracia et al. (2011) has presented a new tool
to characterize the wastewater sludge in terms of the model
components. This estimation is carried out by minimising
a cost function (Grau et al., 2007b) and taking into account the
physical and chemical restrictions due to the components’
elemental composition.
6.5.3. Initial conditions
Establishing the initial conditions of the state variables of the
selected model, which in many articles is merely omitted, is
one the first issues that have to be defined before the opti-
mization process itself. As previously mentioned, several
experimental methods for determining substrate character-
ization may be used, thus knowing the initial conditions for
the different organic compounds considered in the model
should not de considered a big issue. However, setting the
initial concentration of all the microbial populations of the
model, especially in large model such as ADM1, is quite
complicated considering the difficulty of its experimental
estimation.
On the other hand, the type of operation mode has an
influence in how accurate the initial conditions has to be
established. In batch tests, the initial conditions are the sole
inputs of the system, hence the initial values of the state
variables will exert great influence in the model behavior. By
contrast, in continuous systems the initial condition had
a negligible effect especially in the case of long-term operation
evaluations, as long as an initial simulation period is not taken
into account, allowing convergence (Batstone et al., 2003). In
other cases, initial values of the microbial population have
been either arbitrarily fixed (Knobel and Lewis, 2002; Noykova
et al., 2002; Nopharatana et al., 2003; Ozkan-Yucel and Gokcay,
2010) or estimated through a prior simulation evaluation
(Batstone et al., 2004b). A typical strategy to achieve this is to
run, for a very long time, a steady state simulation of a similar
system to that from where the sludge comes from and use the
biomass relative composition from the results of such simu-
lation and the total biomass from an available measurement
in the real system to be simulated. In some cases, the initial
concentration of each population has been considered as an
unknown value, thus it has been estimated in the optimiza-
tion process by using adequate experimental information
(Haag et al., 2003; Flotats et al., 2006; Palatsi et al., 2010);
however, some important identification problems have been
encountered in other studies (Flotats et al., 2003; Jeong et al.,
2005). Regardless, estimating the initial conditions of the
biomass composition is the most recommended choice,
especially in the case of batch test, since besides being valu-
able information, it would not be appropriate to force the best-
fit by fixing these initial values.
7. Parameter uncertainty estimation
If the model has passed direct validation, and whenever
possible cross validation (this later step can be made difficult
by the scarcity of experimental data), it is interesting to
analyze further the accuracy of the model parameters, and to
provide confidence intervals for the parameters and in turn,
for the model prediction.
7.1. Error covariance matrix
The Cramer-Rao bound (Walter and Pronzato, 1997), which
corresponds to the inverse of the Fisher Information Matrix
(FIM), provides an optimistic estimate of the parameter error
covariance matrix, i.e., information on the parameter stan-
dard deviation and correlation. The FIM can be computed
using the output parameter sensitivity matrix and the inverse
of the covariance matrix of measurement noise. In AD
systems, FIM has been evaluated, for instance in Noykova
et al. (2002), Flotats et al. (2003) and Haag et al. (2003). In the
case of simple models, the parametric sensitivity can be
calculated analytically as in Lokshina et al. (2001). For more
complex models, a numerical procedure is preferred, e.g. finite
differences. When using a gradient-based method, e.g. Lev-
enbergeMarquardt, it is usually possible to extract the sensi-
tivities at the optimum (as a byproduct of the algorithm
computation).
It is important to keep in mind that the inverse of FIM just
provides a (maybe too) optimistic estimate of the parameter
error covariance matrix, as has been pointed out in the case of
large and non-linear biochemical systems by Schenkendorf
et al. (2009), Joshi et al. (2006) and Lopez and Borzacconi (2010).
7.2. Confidence intervals
The covariance matrix of the parameter errors, as evaluated in
the previous subsection, together with a model linearization,
allows the computation of confidence intervals for the model
prediction. These intervals are of course only approximations,
whose quality depends on the model non-linearity in the
parameters, but they provide a first view of the model’s
uncertainty.
More accurate, but also more computationally demanding,
approaches to the estimation of confidence intervals have
been proposed (Dochain and Vanrolleghem, 2001). Confidence
regions have been obtained by Batstone et al. (2003), Batstone
et al. (2004b), Kalfas et al. (2006) and Batstone et al. (2009), in
the case of the estimation of kinetic parameters or stoichio-
metric coefficients of ADM1 for the description of the AD of
several substrates.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
7.3. Joint posterior distribution
In the Bayesian approach the model parameters are defined as
random variables, and therefore, the uncertainty of the
parameters is defined by the joint posterior distribution. In
this case, the posterior distribution of parameters is generally
derived by a Markov Chain Monte Carlo technique, i.e., using
a numerical implementation. This is a great advantage in the
case of dynamic non-linear models because: first, the cali-
bration method is independent of the model structure; and
second, the uncertainty of the parameters is estimated by the
distribution of their most feasible values, without making any
assumptions about the model structure.
8. Model validation
Once a set of parameters has been obtained, it is necessary to
question the predictive quality of the resulting model and to
assess the parameter accuracy. This will determine the
confidence behind the model, and tell the modeler if he needs
to revise the model’s identification. The overall procedure,
called model validation, consists of several steps.
8.1. Direct validation
The first test is to check whether the model is able to repro-
duce the experimental data that has been used for parameter
identification. Otherwise, there is obviously something wrong
in the identification procedure (see details in Section 2) and it
has to be adjusted and repeated.
There are different ways of checking the model’s
adequacy. One of the best, even if it cannot be cast into
mathematical formulas, is the visual inspection: the model
has to follow well the data evolution while smoothing off the
noise (a model that tends to reproduce noise is over-
parametrized and will fail later on in cross-validation tests).
Model performance by visual inspection is widely used in AD
system, and has even been the only applied method in many
cases.
On a more mathematical basis, a good test is based on
residuals analysis. If the model is predicting the data well, the
residual can be directly related to the measurement error.
Different types of information may be drawn from the resid-
uals, such as the determination coefficient (R2), an estimation
of the variance of the data, analysis of randomness, etc.
Actually, the determination coefficient has been the sole tool
used to evaluate the model fit in several studies (Redzwan and
Banks, 2004; Flotats et al., 2006; Palatsi et al., 2010). This
parameter was also used by Flotats et al. (2003) to evaluate the
model fit quality as part of a more detailed model analysis.
Other statistical tests have been used to compare several
models, such as the Fisher test (Aceves-Lara et al., 2005), the
sum of normalized errors (Bhunia and Ghangrekar, 2008) or
the determination coefficient (Donoso-Bravo et al., 2010).
Nevertheless, few studies have estimated the variance of the
experimental data through the residuals (Haag et al., 2003).
Analysis of randomness in the residual (another method to
evaluate the fit quality) is seldom performed, even though it
5359
may represent a good criterion to determine autocorrelation
or the need to change the formulation of the cost function.
A particular case is the study developed by Barampouti
et al. (2005) who used this type of statistical analysis to
generate an empirical model to predict the biogas production
and to evaluate goodness-of-fit with the simulated data.
8.2. Cross validation
Direct validation is a necessary condition, but by no means
a sufficient condition to accept a model as being one that can
reproduce the behavior of the system under consideration. It
may well be that the model fits the data that has been used for
identification adequately, but performs poorly with new
unseen data. To this end, enough data must be available and
divided these into two subsets, one for parameter identifica-
tion (and afterward direct validation), and the other for cross
validation. Notice that cross validation can in turn imply the
identification of the initial conditions of the experiments
under consideration (i.e., the initial conditions of the unseen
experiments are sometimes unknown, or at least, know but
no well, and have to be estimated before checking that the
model with the previously identified parameters fits well the
new data). This procedure has been applied to check the AD
model validity, especially when complex models, such as
ADM1, are considered (Ozkan-Yucel and Gokcay, 2010;
Fezzani and Ben Cheikh, 2009; Fezzani and Ben Cheikh,
2008; Tartakovsky et al., 2008; Siegrist et al., 2002; Lübken
et al., 2007). In the same context, short calibration steps may
also be performed regularly during the validation of the model
(Batstone et al., 2009; Bernard et al., 2001), in order to take the
possible variations of the substrate characteristics as well as
the changes in the anaerobic population into account espe-
cially when long-term operation data are used.
9. Conclusion
Anaerobic digestion is a very complex process involving
various bacterial populations and substrates. With the prog-
resses in instrumentation and in computer science, the
development of mathematical models, predicting the
dynamic process behavior has attracted considerable atten-
tion in the last two decades. ADM1 is undoubtedly one of the
milestones of this research era. However, modeling is always
a goal-driven exercise, and many alternative models have
been proposed in the literature, depending on the aim, e.g.,
process understanding, dynamic simulation, optimization, or
control.
Models contain unknown parameters, e.g., initial condi-
tions, stoichiometry, and kinetic parameters which have to be
estimated from experimental data. Parameter identification is
a delicate task due the potentially large number of parameters
and the scarcity of informative experimental data. This review
attempts to summarize the efforts that have been accom-
plished in the parameter estimation of models of anaerobic
digestion processes and highlights the critical steps of the
identification procedure. In general, the literature shows
a lack of systematic and clear procedure for modeling AD
processes. Also, sets of parameter values are reported without
5360 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
thorough analysis of the model’s validity and parameter
accuracy, which makes it difficult to exploit all of the pub-
lished information.
This situation will certainly improve with more awareness
of these important issues. As a general recommendation, the
development of benchmarks, and the availability of data bases,
as open resources on the internet, would certainly speed up
these developments and consolidate knowledge in the field.
Acknowledgments
This paper presents research results of the Belgian Network
DYSCO (Dynamical Systems, Control, and Optimization),
funded by the Interuniversity Attraction Poles Programme,
initiated by the Belgian State, Science Policy Office. The
scientific responsibility rests with its author(s). This study is
also supported by a grant from Belspo (Belgian Science Policy)
through its Postdoc fellowships to non-EU researchers
program. Johan Mailier is a research fellow supported by the
FNRS (Belgian National Fund for Scientific Research).
references
Abu Qdais, H., Hani, K.B., Shatnawi, N., 2010. Modeling and
optimization of biogas production from a waste digester using
artificial neural network and genetic algorithm. Resources
Conservation and Recycling 54 (6), 359e363.
Aceves-Lara, C.A., Aguilar-Garnica, E., Alcaraz-Gonzalez, V.,
Gonzalez-Reynoso, O., Steyer, J.P., Dominguez-Beltran, J.L.,
Gonzalez-Alvarez, V., 2005. Kinetic parameters estimation in
an anaerobic digestion process using successive quadratic
programming. Water Science and Technology 52 (1e2),
419e426.
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1993. A mathematical
model for dynamic simulation of anaerobic digestion of
complex substrates: focusing on ammonia inhibition.
Biotechnology and Bioengineering 42 (2), 159e166.
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1999. A comprehensive
model of anaerobic bioconversion of complex substrates to
biogas. Biotechnology and Bioengineering 63 (3), 363e372.
Appels, L., Baeyens, J., Degreve, J., Dewil, R., 2008. Principles and
potential of the anaerobic digestion of waste-activated sludge.
Progress in Energy and Combustion Science 34 (6), 755e781.
Ascher, U.M., Mattheij, R.M.M., Russell, R.D., 1995. Numerical
Solution of Boundary Value Problems for Ordinary Differential
Equations. Classics in Applied Mathematics. SIAM,
Philadelphia, PA.
Baltes, M., Schneider, R., Sturm, C., Reuss, M., 1994. Optimal
experimental design for parameter estimation in unstructured
growth models. Biotechnology Progress 10 (5), 480e488.
Banks, C.J., Chesshire, M., Heaven, S., Arnold, R., 2011. Anaerobic
digestion of source-segregated domestic food waste:
performance assessment by mass and energy balance.
Bioresource Technology 102 (2), 612e620.
Barampouti, E.M.P., Mai, S.T., Vlyssides, A.G., 2005. Dynamic
modeling of biogas production in an UASB reactor for potato
processing wastewater treatment. Chemical Engineering
Journal 106 (1), 53e58.
Bastidas-Oyanedel, J.R., Aceves-Lara, C.A., Ruiz-Filippi, G.,
Steyer, J.P., 2008. Thermodynamic analysis of energy transfer
in acidogenic cultures. Engineering in Life Sciences 8 (5),
487e498.
Batstone, D., 1999. High Rate Anaerobic Treatment of Complex
Wastewater. PhD thesis, Department of Chemical
Engineering, University of Queensland.
Batstone, D., Keller, J., Newell, R., Newland, M., 2000. Modelling
anaerobic degradation of complex wastewater. I: model
development. Bioresource Technology 75, 67e74.
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V.,
Pavlostathis, S.G., Rozzi, A., Sanders, W.T., Siegrist, H.,
Vavilin, V.A., 2002. Anaerobic Digestion Model No. 1. (ADM1).
Iwa Task Group, IWA Task Group for Mathematical Modelling
of Anaerobic Digestion Processes, IWA Publishing, London.
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), 195e204.
Batstone, D.J., Keller, J., 2003. Industrial applications of the IWA
anaerobic digestion model No. 1 (ADM1). Water Science and
Technology 47 (12), 199e206.
Batstone, D., Keller, J., Bonnet, B., 2004a. The influence of substrate
kinetics on the microbial community structure in granular
anaerobic biomass. Water Research 38 (6), 1390e1404.
Batstone, D.J., Torrijos, M., Ruiz, C., Schmidt, J.E., 2004b. Use of an
anaerobic sequencing batch reactor for parameter estimation
in modelling of anaerobic digestion. Water Science and
Technology 50 (10), 295e303.
Batstone, D., Hernandez, J.L.A., Schmidt, J.E., 2005. Hydraulics of
laboratory and full-scale upflow anaerobic sludge blanket
(UASB) reactors. Biotechnology and Bioengineering 91 (3),
387e391.
Batstone, D.J., 2006. Mathematical modelling of anaerobic
reactors treating domestic wastewater: rational criteria for
model use. Re-views in Environmental Science and
Biotechnology 5 (1), 57e71.
Batstone, D.J., Tait, S., Starrenburg, D., 2009. Estimation of
hydrolysis parameters in full-scale anerobic digesters.
Biotechnology and Bioengineering 102 (5), 1513e1520.
Bernard, O., Zakaria, H., Dochain, D., 2000. Software sensors to
monitor the dynamics of microbial communities: application
to anaerobic digestion. Acta Biotheoretica 48, 197e205.
Bernard, O., Hadj-Sadok, Z., Dochain, D., Genovesi, A., Steyer, J.P.,
2001. Dynamical model development and parameter
identification for an anaerobic wastewater treatment process.
Biotechnology and Bioengineering 75 (4), 424e438.
Bernard, O., Chachuat, B., Helias, A., Rodriguez-, J., 2006. Can we
assess the model complexity for a bioprocess: theory and
example of the anaerobic digestion process. Water Science
and Technology 53 (1), 85e92.
Bhunia, P., Ghangrekar, M.M., 2008. Analysis, evaluation, and
optimization of kinetic parameters for performance appraisal
and design of UASB reactors. Bioresource Technology 99 (7),
2132e2140.
Boe, K., Batstone, D.J., Steyer, J.P., Angelidaki, I., 2010. State
indicators for monitoring the anaerobic digestion process.
Water Research 44 (20), 5973e5980.
Chen, Y., Cheng, J.J., Creamer, K.S., 2008. Inhibition of anaerobic
digestion process: a review. Bioresource Technology 99 (10),
4044e4064.
Costello, D.J., Greenfield, P.F., Lee, P.L., 1991. Dynamic modeling of
A single-Stage high-rate anaerobic reactor .1. Model
derivation. Water Research 25 (7), 847e858.
Dalmau, J., Comas, J., Rodriguez-Roda, I., Pagilla, K., Steyer, J.P.,
2010. Model development and simulation for predicting risk of
foaming in anaerobic digestion systems. Bioresource
Technology 101 (12), 4306e4314.
de Baere, L. A., Mattheeuws, B., Velghe, F., 2010. State of the art of
anaerobic digestion in Europe. 12th World Congress on
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
Anaerobic Digestion (AD12), Guadalajara, Mexico, October 31st
to November 4th, 2010.
De Gracia, M., Sancho, L., Garcı́a-Heras, J.L., Vanrolleghem, P.,
Ayesa, E., 2006. Mass and charge conservation check in
dynamic models: application to the new ADM1 model. Water
Science and Technology 53 (1), 225e240.
De Gracia, M., Grau, P., Huete, E., Gómez, J., Garcı́a-Heras, J.L.,
Ayesa, E., 2009. New generic mathematical model for WWTP
sludge digesters operating under aerobic and anaerobic
conditions: model building and experimental verification.
Water Research 43 (18), 4626e4642.
De Gracia, M., Huete, E., Beltrán, S., Grau, P., Ayesa, E., 2011.
Automatic characterisation of primary, secondary and mixed
sludge inflow in terms of the mathematical generalised sludge
digester model. Water Science and Technology 64 (3), 557e567.
Derbal, K., Bencheikh-Lehocine, M., Cecchi, F., Meniai, A.-H.,
Pavan, P., 2009. Application of the IWA ADM1 model to
simulate anaerobic co-digestion of organic waste with waste
activated sludge in mesophilic condition. Bioresource
Technology 100, 1539e1543.
Deveci, N., Ciftci, G., 2001. A mathematical model for the
anaerobic treatment of baker’s yeast effluents. Waste
Management 21 (1), 99e103.
Dimitrova, N., Krastanov, M., 2009. Nonlinear stabilizing control
of an uncertain bioprocess model. International Journal of
Applied Mathematics and Computer Science 19 (3), 441e454.
Dochain, D., Vanrolleghem, P., 2001. Dynamical Modelling &
Estimation in Wastewater Treatment Processes. IWA
Publishing, London.
Donoso-Bravo, A., Carballa, M., Ruiz-Filippi, G., Chamy, R., 2009a.
Treatment of low strength sewage with high suspended
organic matter content in an anaerobic sequencing batch
reactor and modelling application. Electronic Journal of
Biotechnology 12 (3), 13e14.
Donoso-Bravo, A., Retamal, C., Carballa, M., Ruiz-Filippi, G.,
Chamy, R., 2009b. Influence of temperature on hydrolysis,
acidogenesis and methanogenesis in anaerobic digestion:
parameter identification and modeling application. Water
Science and Technology 60 (1), 9e17.
Donoso-Bravo, A., Perez-Elvira, S., Fdz-Polanco, F., 2010.
Application of simplified models for anaerobic
biodegradability tests. Evaluation of pre-treatment processes.
Chemical Engineering Journal 160, 607e614.
Donoso-Bravo, A., Garcia, G., Perez-Elvira, S.I., Fdz-Polanco, F.,
2011. Initial rate technique as a procedure to predict the
anaerobic digester operation. Biochemical Engineering Journal
53, 275e280.
Eastman, J.A., Ferguson, J.F., 1981. Solubilization of particulate
organic-carbon during the acid phase of anaerobic digestion.
Journal of Water Pollution Control Federation 53 (3), 352e366.
Ekama, G.A., Sötemann, S.W., Wentzel, M.C., 2007.
Biodegradability of activated sludge organics under anaerobic
conditions. Water Research 41 (1), 244e252.
Fedorovich, V., Lens, P., Kalyuzhnyi, S., 2003. Extension of
anaerobic digestion model No. 1 with processes of sulfate
reduction. Applied Biochemistry and Biotechnology 109 (1e3),
33e45.
Fezzani, B., Ben Cheikh, R., 2008. Implementation of IWA
anaerobic digestion model No. 1 (ADM1) for simulating the
thermophilic anaerobic co-digestion of olive mill wastewater
with olive mill solid waste in a semi-continuous tubular
digester. Chemical Engineering Journal 141 (1e3), 75e88.
Fezzani, B., Ben Cheikh, R., 2009. Extension of the anaerobic
digestion model No. 1 (ADM1) to include phenolic
compounds biodegradation processes for the simulation of
anaerobic co-digestion of olive mill wastes at thermophilic
temperature. Journal of Hazardous Materials 162 (2e3),
1563e1570.
5361
Flotats, X., Ahring, B.K., Angelidaki, I., 2003. Parameter
identification of thermophilic anaerobic degradation of
valerate. Applied Biochemistry and Biotechnology 109 (1e3),
47e62.
Flotats, X., Palatsi, J., Ahring, B.K., Angelidaki, I., 2006.
Identifiability study of the proteins degradation model, based
on ADM1, using simultaneous batch experiments. Water
Science and Technology 54 (4), 31e39.
Gali, A., Benabdallah, T., Astals, S., Mata-Alvarez, J., 2009.
Modified version of ADM1 model for agro-waste application.
Bioresource Technology 100, 2783e2790.
Garcia-Ochoa, F., Santos, V.E., Naval, L., Guardiola, E., Lopez, B.,
1999. Kinetic model for anaerobic digestion of livestock
manure. Enzyme and Microbial Technology 25 (1e2), 55e60.
Gavala, H.N., Angelidaki, I., Ahring, B.K., 2003. Kinetics and
modeling of anaerobic digestion process. Advances in
Biochemical Engineering/biotechnology 81, 57e93.
Ghaniyari-Benis, S., Martin, A., Borja, R., 2010. Kinetic modelling
and performance prediction of a hybrid anaerobic baffled
reactor treating synthetic wastewater at mesophilic
temperature. Process Biochemistry 45 (10), 1616e1623.
Girault, R., Rousseau, P., Steyer, J., Bernet, N., Beline, F., 2011.
Combination of batch experiments with continuous reactor
data for ADM1 calibration: application to anaerobic digestion
of pig slurry. Water Science and Technology 63 (11),
2575e2582.
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), 742e748.
Grau, P., de Gracia, M., Vanrolleghem, P.A., Ayesa, E., 2007a. A
new plant-wide modelling methodology for WWTPs. Water
Research 41 (19), 4357e4372.
Grau, P., Beltrán, S., de Gracia, M., Ayesa, E., 2007b. New
mathematical procedure for the automatic estimation of
influent characteristics in WWTPs. Water Science and
Technology 56 (8), 95e106.
Guisasola, A., Baeza, J.A., Carrera, J., Sin, G., Vanrolleghem, P.A.,
Lafuente, J., 2006. The influence of experimental data quality
and quantity on parameter estimation accuracy. Education for
Chemical Engineers 1, 139e145.
Guisasola, A., Sharma, K.R., Keller, J., Yuan, Z.Q., 2009.
Development of a model for assessing methane formation in
rising main sewers. Water Research 43 (11), 2874e2884.
Haag, J.E., Vande Wouwer, A., Queinnec, I., 2003. Macroscopic
modelling and identification of an anaerobic waste treatment
process. Chemical Engineering Science 58, 4307e4316.
Helton, J.C., Davis, F.J., 2003. Latin hypercube sampling and the
propagation of uncertainty in analyses of complex systems.
Reliability Engineering and System Safety 81, 23e69.
Hill, D.T., Barth, C., 1977. A dynamical model for simulation of
animal waste digestion. Journal of Water Pollution Control
Association 10, 2129e2143.
Hill, D.T., 1982. A comprehensive dynamic model for animal
waste methanogenesis. Transactions of the ASAE 25 (5),
1374e1380.
Huete, E., de Gracia, M., Ayesa, E., Garcia-Heras, J.L., 2006. ADM1-
based methodology for the characterisation of the influent
sludge in anaerobic reactors. Water Science and Technology
54 (4), 157e166.
Hulhoven, X., Vande Wouwer, A., Bogaerts, P., 2005. On
a systematic procedure for the predetermination of
macroscopic reaction schemes. Journal of Bioprocess and
Biosystems Engineering 27, 283e291.
Husain, A., 1998. Mathematical models of the kinetics of
anaerobic digestion e a selected review. Biomass and
Bioenergy 14 (5e6), 561e571.
Illanes, A., 2008. Enzyme Biocataysis. Principles and Applications.
Springer Science, Valparaiso.
5362 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
Jacobi, H.F., Moschner, C.R., Hartung, E., 2011. Use of near infrared
spectroscopy in online-monitoring of feeding substrate
quality in anaerobic digestion. Bioresource Technology 102 (7),
4688e4696.
Jakeman, A.J., Letcher, R.A., Norton, J.P., 2006. Ten iterative steps
in development and evaluation of environmental models.
Environmental Modelling and Software 21 (5), 602e614.
Jeong, H.S., Suh, C.W., Lim, J.L., Lee, S.H., Shin, H.S., 2005. Analysis
and application of ADM1 for anaerobic methane production.
Bioprocess and Biosystems Engineering 27 (2), 81e89.
Joshi, M., Seidel-Morgenstern, A., Kremling, A., 2006. Exploiting
the bootstrap method for quantifying parameter confidence
intervals in dynamical systems. Metabolic Engineering 8 (5),
447e455.
Kalfas, H., Skiadas, I.V., Gavala, H.N., Stamatelatou, K.,
Lyberatos, G., 2006. Application of ADM1 for the simulation of
anaerobic digestion of olive pulp under mesophilic and
thermophilic conditions. Water Science and Technology 54 (4),
149e156.
Kalyuzhnyi, S.V., 1997. Batch anaerobic digestion of glucose and
its mathematical modeling .2. Description, verification and
application of model. Bioresource Technology 59 (2e3),
249e258.
Kalyuzhnyi, S.V., Davlyatshina, M.A., 1997. Batch anaerobic
digestion of glucose and its mathematical modeling .1. Kinetic
investigations. Bioresource Technology 59 (1), 73e80.
Kalyuzhnyi, S.V., Fedorovich, V.V., 1998. Mathematical modelling
of competition between sulphate reduction and
methanogenesis in anaerobic reactors. Bioresource
Technology 65 (3), 227e242.
Keshtkar, A., Meyssami, B., Abolhamd, G., Ghaforian, H., Khalagi
Asadi, M., 2003. Mathematical modeling of non-ideal mixing
continuous flow reactors for anaerobic digestion of cattle
manure. Bioresource Technology 87 (1), 113e124.
Kleerebezem, R., Van Loosdrecht, M.C.M., 2004. Criticizing some
concepts of ADM1. 10th IWA World Congress Anaerobic
Digestion, Montreal, Canada.
Kleerebezem, R., Van Loosdrecht, M.C.M., 2006. Waste
characterization for implementabon in ADM1. Water Science
and Technology 54 (4), 167e174.
Knobel, A.N., Lewis, A.E., 2002. A mathematical model of a high
sulphate wastewater anaerobic treatment system. Water
Research 36 (1), 257e265.
Koch, K., Lubken, M., Gehring, T., Wichern, M., Horn, H., 2010.
Biogas from grass silageemeasurements and modeling with
ADM1. Bioresource Technology 101 (21), 8158e8165.
Kocsis, L., György, A., 2009. Efficient Multi-start Strategies for
Local Search Algorithms. Lecture Notes in Computer Science,
5781/2009 doi:10.1007/978-3-642-04180-8_63705e720.
Lee, M.Y., Suh, C.W., Ahn, Y.T., Shin, H.S., 2009. Variation of
ADM1 by using temperature-phased anaerobic digestion
(TPAD) operation. Bioresource Technology 100 (11),
2816e2822.
Lewis, R.M., Torczon, V., Trosset, M.W., 2000. Direct-Search
Methods: Then and Now. ICASE Report No. 2000e26.
Lin, Y.H., Wu, C.L., 2011. Sensitivity analysis of phenol
degradation with sulfate reduction under anaerobic
conditions. Environmental Modeling & Assessment 16 (2),
213e225.
Liwarska-Bizukojc, E., Biernacki, R., 2010. Identification of the
most sensitive parameters in the activated sludge model
implemented in BioWin software. Bioresource Technology 101
(19), 7278e7285.
Lokshina, L., Vavilin, V., Kettunen, H., Rintala, J., Holliger, C.,
Nozhevnikova, A., 2001. Evaluation of kinetic coefficients
using integrated Monod and Haldane models for low-
temperature acetoclastic methanogenesis. Water Research 35
(12), 2913e2922.
Lopez, I., Borzacconi, L., 2009. Modelling a full scale UASB reactor
using a COD global balance approach and state observers.
Chemical Engineering Journal 146, 1e5.
Lopez, I., Borzacconi, L., 2010. Modelling of slaughterhouse solid
waste anaerobic digestion: determination of parameters and
continuous reactor simulation. Waste Management 30,
1813e1821.
Lübken, M., Wichern, M., Schlattmann, M., Gronauer, A., Horn, H.,
2007. Modelling the energy balance of an anaerobic digester
fed with cattle manure and renewable energy crops. Water
Research 41 (18), 4085e4096.
Mailier, J., Delmotte, A., Cloutier, M., Jolicoeur, M., Wouwer, A.V.,
2011. Parametric sensitivity analysis and reduction of
a detailed nutritional model of plant cell cultures.
Biotechnology and Bioengineering 108 (5), 1108e1118.
Marquardt, D.W., 1963. An algorithm for least squares estimation
of nonlinear parameters. Journal of the Society for Industrial
and Applied Mathematics 11, 431e441.
Martin, C., Ayesa, E., 2010. An integrated Monte Carlo
methodology for the calibration of water quality models.
Ecological Modelling 221, 2656e2667.
Martin, C., de Gracia, M., Ayesa, E., 2011. Bayesian Calibration of
the Disintegration process in WWTP Sludge Digesters. 8th
IWA Symposium on Systems Analysis and Integrated
Assessment (WaterMatex), June 20e22, San Sebastián, Spain.
Martin, M.A., Raposo, F., Borja, R., Martin, A., 2002. Kinetic study
of the anaerobic digestion of vinasse pretreated with ozone,
ozone plus ultraviolet light, and ozone plus ultraviolet light in
the presence of titanium dioxide. Process Biochemistry 37 (7),
699e706.
Molina, F., Castellano, M., Garcia, C., Roca, E., Lema, J., 2009.
Selection of variables for on-line monitoring, diagnosis, and
control of anaerobic digestion processes. Water Science and
Technology 60 (3), 615e622.
Morris, M.D., 1991. Factorial sampling plans for preliminary
computational experiments. Technometrics 33 (2), 161e174.
Mosche, M., Jordening, H.J., 1999. Comparison of different models
of substrate and product inhibition in anaerobic digestion.
Water Research 33 (11), 2545e2554.
Muller, T.G., Noykova, N., Gyllenberg, M., Timmer, J., 2002.
Parameter identification in dynamical models of anaerobic
waste water treatment. Mathematical Biosciences 177, 147e160.
Nocedal, J., Wright, S., 2006. Numerical Optimization. Springer
Verlag.
Nopharatana, A., Pullammanappallil, P.C., Clarke, W.P., 2003. A
dynamic mathematical model for sequential leach bed
anaerobic digestion of organic fraction of municipal solid
waste. Biochemical Engineering Journal 13 (1), 21e33.
Noykova, N., Gyllenberg, M., 2000. Sensitivity analysis and
parameter estimation in a model of anaerobic wastewater
treatment processes with substrate inhibition. Bioprocess
Engineering 23, 343e349.
Noykova, N., Muller, T.G., Gyllenberg, M., Timmer, J., 2002.
Quantitative analyses of anaerobic wastewater treatment
processes: identifiability and parameter estimation.
Biotechnology and Bioengineering 78 (1), 89e103.
Omlin, M., Reichert, P., 1999. A comparison of techniques for the
estimation of model prediction uncertainty. Ecological
Modelling 115 (1), 45e59.
Ossenbruggen, P.J., Stevens, D.K., 1996. Mechanistic model
identification for nitrification in activated sludge.
Environmetrics 7 (6), 579e600.
Ozkan-Yucel, U.G., Gokcay, C.F., 2010. Application of ADM1 model
to a full-scale anaerobic digester under dynamic organic
loading conditions. Environmental Technology 31 (6),
633e640.
Palatsi, J., Illa, J., Prenafeta-Boldu, F.X., Laureni, M., Fernandez, B.,
Angelidaki, I., Flotats, X., 2010. Long-chain fatty acids
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
inhibition and adaptation process in anaerobic thermophilic
digestion: batch tests, microbial community structure and
mathematical modelling. Bioresource Technology 101 (7),
2243e2251.
Pobeheim, H., Munk, B., Mueller, H., Berg, G., Guebitz, G., 2010.
Characterization of an anaerobic population digesting a model
substrate for maize in the presence of trace metals.
Chemosphere 80 (8), 829e836.
Qian, S.S., Stow, C.A., Borsuk, M.E., 2003. On Monte Carlo methods
for bayesian inference. Ecological Modelling 159 (2e3), 269.
Ramirez, I., Mottet, A., Carrere, H., Deleris, S., Vedrenne, F.,
Steyer, J.P., 2009. Modified ADM1 disintegration/hydrolysis
structures for modeling batch thermophilic anaerobic
digestion of thermally pretreated waste activated sludge.
Water Research 43, 3479e3492.
Raposo, F., Borja, R., Martı́n, M.A., Martı́n, A., de la Rubia, M.A.,
Rincón, B., 2009. Influence of inoculumesubstrate ratio on the
anaerobic digestion of sunflower oil cake in batch mode:
process stability and kinetic evaluation. Chemical Engineering
Journal 149, 70e77.
Reichert, P., 2010. Environmental Systems Analysis Lecture Notes
Summer Course, June 21e25, Eawag Dübendorf, Switzerland.
Redzwan, G., Banks, Ch, 2004. The use of a specific function to
estimate maximum methane production in a batch-fed
anaerobic reactor. Journal of Chemical Technology and
Biotechnology 79, 1174e1178.
Rincon, A., Angulo, F., Olivar, G., 2009. Control of an anaerobic
digester through normal form of fold bifurcation. Journal of
Process Control 19 (8), 1355e1367.
Rittmann, B.E., 2008. Opportunities for renewable bioenergy using
microorganisms. Biotechnology and Bioengineering 100 (2),
203e212.
Rodrigues, J.A.D., Ratusznei, S.M., Zaiat, M., 2003. Fed-batch and
batch operating mode analysis of a stirred anaerobic
sequencing reactor with self-immobilized biomass treating
low-strength wastewater. Journal of Environmental
Management 69 (2), 193e200.
Rodriguez, J., Lema, J.M., Van Loosdrecht, M.C.M., Kleerebezem, R.
, 2006. Variable stoichiometry with thermodynamic control in
ADM1. Water Science and Technology 54 (4), 101e110.
Rodriguez, J., Roca, E., Lema, J.M., Bernard, O., 2008.
Determination of the adequate minimum model complexity
required in anaerobic bioprocesses using experimental data.
Journal of Chemical Technology and Biotechnology 83 (12),
1694e1702.
Ruel, S.M., Comeau, Y., Ginestet, P., Heduit, A., 2002. Modeling
acidogenic and sulfate-reducing processes for the
determination of fermentable fractions in wastewater.
Biotechnology and Bioengineering 80 (5), 525e536.
Ruzicka, M., 1996. The effect of hydrogen on acidogenic glucose
cleavage. Water Research 30 (10), 2447e2451.
Sales-Cruz, M., Gani, R., 2004. Aspects of modelling and model
identification for bioprocesses through a computer-aided
modelling system. In: Barbosa-Póvoa, Ana Paula,
Matos, Henrique (Eds.), Computer Aided Chemical Engineering
European Symposium on Computer-Aided Process Engineering-
14, 37th European Symposium of the Working Party on
Computer-Aided Process Engineering. Elsevier, pp. 1123e1128.
Saltelli, A., Ratto, M., Tarantola, S., Campolongo, F., 2006.
Sensitivity analysis practices: strategies for model-based
inference. Reliability Engineering and System Safety 91,
1109e1125.
Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J.,
Gatelli, D., Saisana, M., Tarantola, S., 2008. Global Sensitivity
Analysis. The Primer. John Wiley & Sons, West Sussex,
England.
Sanchez, J.M., Arijo, S., Muñoz, M.A., Morinigo, M.A., Borrego, J.J.,
1994. Microbial colonization of different support materials
5363
used to enhance the methanogenic process. Applied
Microbiology and Biotechnology 41 (4), 480e486.
Saravanan, V., Sreekrishnan, T.R., 2006. Modelling anaerobic
biofilm reactors e a review. Journal of Environmental
Management 81 (1), 1e18.
Sbarciog, M., Loccufier, M., Noldus, E., 2010. Determination of
appropriate operating strategies for anaerobic digestion
systems. Biochemical Engineering Journal 51 (3), 180e188.
Schenkendorf, R., Kremling, A., Mangold, M., 2009. Optimal
experimental design with the sigma point method. IET
Systems Biology 3 (1), 10e23.
Siegrist, H., Renggli, D., Gujer, W., 1993. Mathematical-Modeling
of anaerobic mesophilic sewage-sludge treatment. Water
Science and Technology 27 (2), 25e36.
Siegrist, H., Vogt, D., Garcia-Heras, J., Gujer, W., 2002.
Mathematical model for meso- and thermophilic anaerobic
sewage sludge digestion. Environmental Science and
Technology 36 (5), 1113e1123.
Simeonov, I., Momchev, V., Grancharov, D., 1996. Dynamic
modeling of mesophilic anaerobic digestion of animal waste.
Water Research 30 (5), 1087e1094.
Simeonov, I.S., 1999. Mathematical modeling and parameters
estimation of anaerobic fermentation processes. Bioprocess
Engineering 21 (4), 377e381.
Sin, G., Gernaey, K.V., Neumann, M.B., Van Loosdrecht, M.C.M.,
Gujer, W., 2011. Global sensitivity analysis in wastewater
treatment plant model applications: prioritizing sources of
uncertainty. Water Research 45 (2), 639e651.
Sobol, I.M., 2001. Global sensitivity indices for nonlinear
mathematical models and their Monte Carlo estimates.
Mathematics and Computers in Simulation 55 (1e3),
271e280.
Smith, L.H., McCarty, P.L., Kitanidis, P.K., 1998. Spreadsheet
method for evaluation of biochemical reaction rate
coefficients and their uncertainties by weighted nonlinear
least-squares analysis of the integrated monod equation.
Applied and Environmental Microbiology 64 (6), 2044e2050.
Speece, R.E., 1996. Anaerobic Biotechnology: For Industrial
Wastewater. Archae Press.
Spriet, J.A., 1985. Structure characterization e an overview. In:
Barker, H.A., Young, P.C. (Eds.), Identification and System
Parameter Estimation 1985 e Proceedings of the seventh IFAC/
IFORS Symposium. Pergamon Press, Oxford, pp. 749e756.
Tarantola, S., Gatelli, D., Mara, T.A., 2006. Random balance
designs for the estimation of first order global sensitivity
indices. Reliability Engineering and System Safety 91 (6),
717e727.
Tartakovsky, B., Morel, E., Steyer, J.P., Guiot, S.R., 2002.
Application of a variable structure model in observation and
control of an anaerobic digestor. Biotechnology Progress 18 (4),
898e903.
Tartakovsky, B., Mu, S.J., Zeng, Y., Lou, S.T., Guiot, S.R., Wu, P.,
2008. Anaerobic digestion model no. 1-based distributed
parameter model of an anaerobic reactor: II. Model validation.
Bioresource Technology 99 (9), 3676e3684.
Tomei, M.C., Braguglia, C.M., Cento, G., Mininni, G., 2009.
Modeling of anaerobic digestion of sludge. Critical Reviews
in Environmental Science and Technology 39 (12),
1003e1051.
Vande Wouwer, A., Renotte, C., Queinnec, I., Bogaerts, P., 2006.
Transient analysis of a wastewater treatment biofilter-
distributed parameter modeling and state estimation.
Mathematical and Computer Modelling of Dynamical Systems
12, 423e440.
Van Lier, J.B., 2008. High-rate anaerobic wastewater treatment:
diversifying from end-of-the-pipe treatment to resource-
oriented conversion techniques. Water Science and
Technology 57 (8), 1137e1148.
5364 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 4 7 e5 3 6 4
Vavilin, V.A., Vasiliev, V.B., Ponomarev, A.V., Rytow, S.V., 1994.
Simulation-model methane as a tool for effective biogas
production during anaerobic conversion of complex organic-
matter. Bioresource Technology 48 (1), 1e8.
Vavilin, V.A., Rytow, S.V., Lokshina, L.Y., 1995. Modelling
hydrogen partial pressure change as a result of competition
between the butyric and propionic groups of acidogenic
bacteria. Bioresource Technology 54 (2), 171e177.
Vavilin, V., Rytov, S.V., Pavlostathis, S.G., Jokela, J., Rintala, J.,
2003. A distributed model of solid waste anaerobic digestion:
sensitivity analysis. Water Science and Technology 48 (4),
147e154.
Walter, E., Pronzato, L., 1997. Identication of Parametric Models
from Experimental Data. Springer-Verla, Berlin.
Ward, A.J., Hobbs, P.J., Holliman, P.J., Jones, D.L., 2008.
Optimisation of the anaerobic digestion of agricultural
resources. Bioresource Technology 99 (17), 7928e7940.
Ward, A.J., Hobbs, P.J., Holliman, P.J., Jones, D.L., 2011. Evaluation
of near infrared spectroscopy and software sensor methods
for determination of total alkalinity in anaerobic digesters.
Bioresource Technology 102 (5), 4083e4090.
Weijers, S.R., 2003. Model reduction of activated sludge model no.
1 and bioprocess models for identification and control.
Biotechnology for the environment: wastewater treatment
and modeling. Waste Gas Handling 3C, 11e63.
Wichern, M., Gehring, T., Fischer, K., Andrade, D., Lubken, M.,
Koch, K., Gronauer, A., Horn, H., 2009. Monofermentation of
grass silage under mesophilic conditions: measurements and
mathematical modeling with ADM 1. Bioresource Technology
100 (4), 1675e1681.
Wolf, C., McLoone, S., Bongards, M., 2008. Plant optimization
using genetic algorithms and particle swarm optimization. In:
Signals and Systems Conference, 208. (ISSC 2008). IET Irish.
Zaher, U., Vanrolleghem, P., Rodriguez, J., Franco, A., 2004.
Conceptual approach for ADM1 applications. In: Ujang, Z.,
Henze, M. (Eds.), Environmental Biotechnology: Advancement
in Water and Wastewater Applications in the Tropics. IWA
Publishing.
Zaher, U., Pandey, P., Chen, S., 2009. A simple elemental
continuity based model application to study the anaerobic
microbial activity for the treatment of dairy manure. Applied
Mathematical Modelling 33 (9), 3553e3564.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Evaluation of a laboratory-scale bioreactive in situ sediment

cap for the treatment of organic contaminants

David W. Himmelheber a,*, Kurt D. Pennell b, Joseph B. Hughes a,c

a
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
b
Department of Civil and Environmental Engineering, Tufts University, Medford, MA, USA
c
School of Material Science and Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA

article info abstract

Article history: The development of bioreactive sediment caps, in which microorganisms capable of
Received 13 December 2010 contaminant transformation are placed within an in situ cap, provides a potential remedial
Received in revised form design that can sustainably treat sediment and groundwater contaminants. The goal of
7 April 2011 this study was to evaluate the ability and limitations of a mixed, anaerobic dechlorinating
Accepted 17 June 2011 consortium to treat chlorinated ethenes within a sand-based cap. Results of batch exper-
Available online 30 June 2011 iments demonstrate that a tetrachloroethene (PCE)-to-ethene mixed consortium was able
to completely dechlorinate dissolved-phase PCE to ethene when supplied only with sedi-
Keywords: ment porewater obtained from a sediment column. To simulate a bioreactive cap,
Sediment remediation laboratory-scale sand columns inoculated with the mixed culture were placed in series
In situ capping with an upflow sediment column and directly supplied sediment effluent and dissolved-
Microbial processes phase chlorinated ethenes. The mixed consortium was not able to sustain dechlorina-
Bioremediation tion activity at a retention time of 0.5 days without delivery of amendments to the
sediment effluent, evidenced by the loss of cis-1,2-dichloroethene (cis-DCE) dechlorination
to vinyl chloride. When soluble electron donor was supplied to the sediment effluent,
complete dechlorination of cis-DCE to ethene was observed at retention times of 0.5 days,
suggesting that sediment effluent lacked sufficient electron donor to maintain active
dechlorination within the sediment cap. Introduction of elevated contaminant concen-
trations also limited biotransformation performance of the dechlorinating consortium
within the cap. These findings indicate that in situ bioreactive capping can be a feasible
remedial approach, provided that residence times are adequate and that appropriate levels
of electron donor and contaminant exist within the cap.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction (Reible et al., 2003). Clean sand has traditionally been

The management and remediation of contaminated aquatic
sediments pose major technical and economic challenges.
Treatment of contaminated sediment sites with in situ
caps has become an established practice that can provide
advantages over alternative methods in certain settings
employed as capping material, and remains a large compo-
nent of many field-scale capping applications. Sand-based
caps have the potential to delay contaminant breakthrough
when diffusive transport dominates (Go et al., 2009; Thoma
et al., 1993), but eventual contaminant breakthrough
remains a source of concern. Additionally, traditional sand

* Corresponding author. Geosyntec Consultants, 10220 Old Columbia Road, Suite A, Columbia, MD 21046, USA. Tel.: þ1 410 381 4333; fax:
þ1 410 381 4499.
E-mail address: dhimmelheber@geosyntec.com (D.W. Himmelheber).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.06.022
5366 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4
caps are less effective at sites where groundwater seepage or
mobile contaminants (i.e., low Koc) are present (Go et al., 2009).
Research studies have focused on in situ sequestration
(Cho et al., 2007; Zimmerman et al., 2004), in situ trans-
formation (Krumins et al., 2009; Lowry and Johnson, 2004), and
the development of active caps which incorporate reactive
and/or sorptive constituents designed to reduce contaminant
and bioavailability (Choi et al., 2009; Hyun et al., 2006; Jacobs
and Förstner, 1999; McDonough et al., 2007; Murphy et al.,
2006; Reible et al., 2007). Ideally, active caps eliminate
the risk of contaminant breakthrough into the overlying
water column, and can potentially be implemented at sedi-
ment sites with groundwater seeps and relatively mobile
contaminants.
The employment of physicochemical-based active caps
appears promising, but possible limitations (e.g., high mate-
rial costs, sorption and reaction capacities) have stimulated
the consideration of in situ bioreactive caps, in which
contaminant biotransformations are designed to occur within
the cap matrix to produce environmentally-acceptable reac-
tion products. Enhanced in situ bioremediation, through
biostimulation and bioaugmentation, has proven to be
a successful groundwater remediation technology for
a diverse range of contaminants (Löffler and Edwards, 2006).
Adaptation of these principles to subaqueous sediment
remediation has not been demonstrated, prompting the
recent identification of in situ bioremediation as a priority
research and development need (SERDP/ESTCP, 2008).
Biologically-based active caps have the potential to maintain
reactivity over long periods of time and could serve as
a sustainable remedial option if microorganisms capable of
biotransformation are present and necessary metabolic
requirements are met.
Previous studies that investigated the activity of microbial
populations within a sediment cap demonstrated that micro-
organisms indigenous to underlying sediment, including
organisms capable of contaminant biotransformation, are
able to colonize the overlying cap and possibly participate
in contaminant bioattenuation processes (Himmelheber
et al., 2009). Bioaugmentation of microorganisms within
a cap, as opposed to intrinsic colonization (defined here as the
natural redistribution of microorganisms native to the sedi-
ment into the cap matrix), could provide enhanced degrada-
tion capacity and minimize the potential for contaminant
release to benthic and aqueous receptors. Such a bio-
augmentation strategy was recently evaluated by the US
Geological Survey (USGS) as a means to reductively dechlori-
nate a mixture of chlorinated ethenes, ethanes, and methanes
present in a groundwater seep discharging into a tidal wetland
(Majcher et al., 2007). A mixed, anaerobic culture was enriched
from the site (Lorah et al., 2008) and incorporated into an
organic-based matrix that was placed at the sediment-water
interface. This bioreactive mat successfully treated the chlo-
rinated contaminants prior to discharge (Majcher et al., 2009).
Although the bioreactive mat was constructed on the banks of
a tidal wetland (i.e., not completely subaqueous) and the
design is not immediately suitable for submergence
(e.g., buoyancy restrictions, delivery of bioaugmentation
culture), the success of the approach supports the concept of
bioreactive capping as an in situ remedial technique.
The USGS bioreactive mat was designed in part because the
chlorinated organics present in the groundwater were
undergoing only partial dechlorination in the sediment prior
to discharge, a phenomenon commonly reported at sediment
sites (Abe et al., 2009; Conant et al., 2004; Hamonts et al., 2009;
Himmelheber et al., 2007; Lendvay et al., 1998; Lorah and
Voytek, 2004; Majcher et al., 2007). Additionally, recent
studies have demonstrated that anaerobic conditions develop
within sediment caps subject to diffusive and upflow condi-
tions (i.e., groundwater seeps) (Himmelheber et al., 2008,
2009). It is therefore expected that contaminated ground-
water seeps will carry partially-degraded contaminants into
the overlying anaerobic cap, thereby providing an opportunity
for treatment by reductive biotransformations.
Detailed assessment of bioreactive in situ sediment caps
has not been previously undertaken and little is currently
known about the feasibility of bioreactive caps, particularly
their limitations and maintenance requirements. The objec-
tive of this work was to establish an actively dechlorinating
microbial consortium within a simulated overlying cap and to
determine how contaminant mass flux and electron donor
amendments influenced bioreactive cap performance. More
specifically, the bioreactive cap experiments were designed to
determine whether or not amendments are necessary to
sustain complete reductive dechlorination by an active
microbial community. Chlorinated ethenes were utilized as
the contaminants due to their frequent occurrence as
groundwater contaminants, their presence in groundwater
seeps, and their greater mobility relative to other sediment
contaminants (e.g., chlorinated benzenes, polychlorinated
biphenyls). Batch reactor and bioaugmented column studies
were conducted to assess bioreactive cap performance over
a range of electron donor and contaminant conditions.
2. Materials and methods
2.1. Chemicals
PCE (99þ%, SigmaeAldrich, St. Louis, MO), TCE (99.5%,
SigmaeAldrich), cis-DCE (97%, Acros Organics, Morris Plains,
NJ), trans-DCE (99.7%, Acros Organics), and 1,1-DCE (99.9%,
Acros Organics) were obtained in neat liquid form. Vinyl
chloride (8%/N2 balance), ethene (99.5%), ethane (99.5%), and
methane (99%) were obtained from Matheson Tri-Gas (Parsip-
pany, NJ). Sodium bicarbonate, potassium chloride, magne-
sium chloride, and calcium chloride were used in the
preparation of simulated groundwater and were purchased
from Fisher Scientific (Pittsburgh, PA). Sodium lactate syrup
(60% vol/vol, Fisher Scientific) was used during the preparation
of stock lactate solutions. Potassium bromide, calcium sulfate,
and potassium phosphate were purchased from Fisher Scien-
tific and used for IC standards and non-reactive tracer studies.
2.2. Batch reactors
Batch reactors were established in triplicate and consisted of
Anacostia River (Washington, D.C., USA) sediment porewater,
dissolved-phase PCE, and a mixed PCE-to-ethene dechlori-
nating consortium. A PCE-to-ethene dechlorinating mixed
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4
consortia referred to as OW served as the inoculum. The OW
consortia, which is capable of complete reductive dechlori-
nation of PCE to ethene, has been described previously
(Daprato et al., 2006). The OW culture has been found to
contain multiple dechlorinating microorganisms, including
Dehalococcoides species, and known reductive dehalogenases
including tceA, vcrA, and bvcA (Daprato et al., 2006).
Three 25 mL aliquots of OW culture were transferred to
70 mL serum bottles pre-capped with Teflon-faced butyl
septa and sparged with N2 gas for 15 min to remove oxygen
from the empty bottles. The collected OW aliquots were then
sparged with N2 gas for 15 min in attempt to remove residual
chlorinated ethenes, methanol, and volatile fatty acids from
the batch reactors. Sediment effluent was collected from
a sediment column that was supplied only with simulated
groundwater and dissolved-phase PCE (Himmelheber et al.,
2007). The composition of simulated groundwater was
slightly modified from that described by Dries et al. (2004) and
consisted of 3.5 mM NaHCO3, 0.1 mM KCl, 0.25 mM MgCl2,
0.75 mM CaCl2, and resazurin as a redox indicator. Sediment
effluent was collected under anoxic conditions and 25 mL of
effluent were added directly to batch reactors containing the
OW consortium. Dissolved-phase PCE was obtained from
a saturated stock solution containing neat PCE in contact with
anaerobic, sterilized simulated groundwater. Stock PCE
concentrations were quantified immediately prior to injection
into the batch reactors. Approximately 16 mmol of dissolved-
phase PCE was added to each microcosm using a 10 mL Ham-
ilton glass syringe. All reactors were wrapped in foil and
incubated at 20 C on an orbital shaker operated at 150 rpm.
Chlorinated ethenes, ethene, ethane, and methane concen-
trations were determined from headspace samples of the
microcosms.
2.3. Bioreactive cap operation
Two one-dimensional (1-D) columns (designated herein as
Bioreactive Cap A and Bioreactive Cap B) were constructed
using 2.5 cm inside diameter (I.D.) glass chromatography
columns  30 cm in length (Spectrum Chromatography,
Houston, TX) and equipped with custom-built stainless steel
end plates (Dutton & Hall, Atlanta, GA). A 2.5 cm diameter disc
of 80 mesh stainless steel (Small Parts, Inc., Miami Lakes, FL)
was placed on the column end plates to retain sand grains
within the column. A fabricated glass reservoir (15 mL) fitted
with a stopcock was placed at the column effluent to allow for
aqueous effluent sampling. The columns were packed with
ASTM C-33 grade concrete sand (U.S. Silica, Mauricetown, NJ).
This particular sand was selected because it is representative
of the solids used for submerged sediment caps and was
utilized in the Anacostia River Capping Demonstration Project
(Reible, 2005). An elemental analysis of the sand was per-
formed at the University of Georgia Laboratory for Environ-
mental Analysis (see Supplementary Information, Table S.1).
The dry, autoclaved sand was packed into the bioreactive
columns under aerobic conditions in 5-cm increments with
vibration along the outside wall of the column. Three pore
volumes of N2-sparged, autoclaved simulated groundwater
were flushed through the columns to check for leakage and
to ensure anaerobic conditions. The columns were then
5367
inoculated by flushing the columns with three pore volumes
of the OW culture suspension.
Following inoculation, the two sand cap columns were
wrapped in foil to avoid exposure to light then connected in
series with an upflow column packed with Anacostia River
sediment as depicted in Fig. 1. The sediment column effluent,
which was provided only with simulated groundwater and
dissolved-phase PCE, served as the influent for the bioreactive
sand columns. Therefore, the influent for the sand columns
consisted of sediment effluent and a mixture of partial PCE-
dechlorination products, similar to the conditions that
would be anticipated in a submerged sediment capping
scenario subject to a PCE-contaminated groundwater seep.
Table 1 provides a summary of experimental conditions
employed for each bioreactive sand column. Chlorinated
ethene and ethene concentrations in the effluent of Bio-
reactive Caps A and B were normalized on a molar basis to
total chlorinated ethenes and ethene eluted per sample to
reduce scatter in concentration data and to monitor product
distribution.
2.3.1. Bioreactive Cap A
Bioreactive Cap A was designed to assess the ability of sedi-
ment effluent to maintain an external dechlorinating
community in a cap, simulating a bioreactive cap inoculated
with a mixed dechlorinating consortia and operating under
reducing conditions. Prior to inoculating Bioreactive Cap A,
a tracer test was conducted with a pulse injection of
100 mg L1 (1.25 mM) bromide obtained from an autoclaved,
sparged stock solution of potassium bromide in simulated
groundwater. A total of 1.2 pore volumes were flushed
through the column, collected with a fraction collector, and
analyzed via ion chromatography. Three pore volumes of
simulated groundwater were then flushed through the
column following the tracer test to remove residual bromide
prior to inoculation. A 200 mL aliquot of aqueous OW culture
was obtained for inoculation and stored in a 160 mL ser-
um bottle that had previously been capped with a Teflon-
faced butyl septum and sparged with N2 for 15 min to
remove oxygen. The 200 mL aliquot was tested for its
dechlorination ability in batch conditions by spiking with PCE
and methanol. After successfully dechlorinating PCE to
ethene (Supplementary Information, Fig. S.1A), 1.5 pore
volumes of the OW culture were supplied to the column at
a flow rate of 2.2 mL h1 (1-day residence time). Following
a 24-h attachment period during which there was no flow,
Bioreactive Cap A was connected in series with the sediment
column from 67 to 83 sediment pore volumes. The
unamended sediment column effluent served as the influent
for the duration of the Bioreactive Cap A experiment.
2.3.2. Bioreactive Cap B
The Bioreactive Cap B experiment was designed to simulate
a dechlorinating bioreactive cap operating under reducing
conditions, but differed from Cap A in that the influent for this
experiment was supplied at various flow rates and periodi-
cally spiked with amendments. Thus, Bioreactive Cap B
demonstrates the impact of contaminant influx and the
presence of reducing equivalents on the capacity of sediment
column effluent to maintain an external dechlorinating
5368 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4

Gas purge following

sample collection

Effluent Samples Effluent Samples

Sand
incoluated
with PCF,

Upflow

Upflow
Anacostia
30 cm Sediment 30 cm to ethene
mixed
consortia

(C) Amendments and

Flow Rate Control to
Bioreactive Cap B
(A)

Anoxic Simulated (B) Captured Sediment

Groundwater + Effluent for
Dissolved PCE Bioreactive Cap B

Fig. 1 e Conceptual schematic of laboratory simulation of a bioreactive sand cap placed in series with an anaerobic sediment
bed subject to a PCE-contaminated groundwater seep. (A) Sediment effluent was directly supplied to Bioreactive Cap A. (B)
Sediment effluent was initially captured via a syringe for Bioreactive Cap B, spiked with amendments, then (C) supplied to
Bioreactive Cap B at select flow rates.

community. An aliquot of OW culture was retrieved and
sparged with nitrogen prior to inoculation as described for
Bioreactive Cap A. The aliquot of OW culture again demon-
strated the ability to completely dechlorinate PCE to ethene in
batch culture (Supplementary Information, Fig. S.1B). A total
of 1.7 pore volumes of OW culture was then supplied to the
column at a flow rate of 2.6 mL h1 (1-day residence time),
followed by a no-flow attachment period of one day. Unlike
Bioreactive Cap A, Bioreactive Cap B was not immediately
connected to the sediment column effluent, but rather
positive-control experiments were conducted to ensure the
inoculated column could completely dechlorinate cis-DCE to
ethene when provided DCB-1 media, Wolin vitamins, and
5 mM lactate as an electron donor and carbon source.
Following this demonstration of complete dechlorination
in the cap under optimal conditions (Supplementary
Information, Fig. S.2), one pore volume of anaerobic simu-
lated groundwater was flushed through the column to remove
these constituents from the system prior to the introduction
of sediment effluent. Sediment column effluent was obtained
from 146 to 180 sediment pore volumes to serve as Bioreactive
Cap B influent.
For Bioreactive Cap B, sediment column effluent was
captured under anoxic conditions by connecting an empty,
gas-tight syringe to sediment effluent tubing and allowing the
aqueous flow to gradually fill the syringe at the same rate of
sediment column influent (5.5 mL h1). Once the effluent
syringe had been filled, it was immediately transferred to
a separate syringe pump and introduced into the sand column
as the influent. This method allowed for manipulation of flow
rates within the sand column and for addition of electron
donor and acceptor to the influent prior to connection with
the sand column. The electron donor used for this study was
lactate, which was obtained from a 100 mM stock solution in
autoclaved, sparged simulated groundwater. Lactate was
supplied to the bioreactive sand column (Cap B) at a concen-
tration of 5 mM from 0 to 13.3 pore volumes (Table 1).
The experimental conditions employed for Bioreactive Cap
B were designed to gradually decrease aqueous residence
times, as well as electron donor concentrations, to determine
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4 5369

100 ion chromatograph with a Dionex AG4A IonPac guard

Table 1 e Summary of experimental conditions for sand
column experiments. column and Dionex AS4A IonPac column at a flow rate of
1.5 mL/min and an ED40 electrochemical detector.
Parameter Bioreactive Bioreactive
Cap A Cap B

Pore volume (PV)a (mL) 62.72 61.82

Porosity (n)a (cm3 void 0.41 0.41 3. Results and discussion
(cm3 total)1)
Connected in series to 67.0e83.2 146.0e180.0
sediment column
3.1. Batch reactors
(sediment pore volumes)
Experimental flow rate (Q) 5.46 1.29; 2.58; 5.46 The OW culture successfully dechlorinated PCE to ethene
(mL h1) when provided only sediment effluent and dissolved-phase
Porewater velocity (v) 62.67 (25.88) 14.99; 29.98; 63.59 PCE (Fig. 2). Complete PCE dechlorination to ethene was ach-
(cm day1) (Darcy velocity (6.10); (12.20); ieved after 19 days of incubation. Chlorinated ethene mass
(cm day1)) (25.88)
balance was within 10% for each time point except day 12,
Peclet number (Pe)b 80.5 N/Ac
(dimensionless)
when chloroethene mole totals were 124% of the initial
Alterations to influent None Addition of dissolved-phase PCE introduced to the batch reactors. This
Lactate discrepancy arose because one of the triplicate reactors
Addition of cis- recorded unusually high concentrations of VC, despite
DCE balanced ethene concentrations at the end of the experiment.
Decrease of flow
This is reflected in the relatively high standard deviation of VC
rate
at day 12. Duplicate analysis at the same time point yielded
Influent chloroethene 16.19  11.06d 0e3.44 PV:
concentration (mM total 200  42d similar results. Regardless of this isolated analytical discrep-
chlorinated ethenes) 3.44 PV to end: ancy, the presence of VC and ethene indicates that dechlori-
34  3.6d nating species within the OW consortium, specifically
Dehalococcoides, remained active for at least one dechlori-
a Estimated from mass difference between dry and wet packed
columns. nation cycle when provided only sediment effluent and a dis-
b Obtained with the CFITM3 breakthrough curve fitting program solved-phase electron acceptor (PCE). Thus, the sources of
under equilibrium constraints. carbon, electron donor, and micronutrients were provided by
c Tracer test not performed. the sediment effluent or from microbial biomass (Adamson
d Average  one standard deviation. and Newell, 2009). Methane concentrations rose steadily
during the dechlorination of PCE (Fig. 2), indicating that
limitations on dechlorination (Table 1). The influent flow rate methanogenic populations were also able to remain active
for Bioreactive Cap B was increased step-wise from 1.3 mL h1 when provided only sediment effluent. These data suggest
(2-day retention time), to 2.6 mL h1 (1-day retention time) to that dechlorinating species within a bioreactive cap inocu-
5.5 mL h1 (0.47-day retention time). From 0 to 3.4 sand pore lated with a methanogenic mixed consortia may have to
volumes, additional cis-DCE was provided to the influent to
ensure chlorinated ethenes were present due to complete 20 120
Methane
dechlorination of PCE to ethene in the sediment column
PCE 100
effluent prior to connecting Bioreactive Cap B. cis-DCE was
Total Chloroethene (µmoles)

16 TCE
chosen assuming partial, intrinsic PCE dechlorination would DCE

Total Methane (mM)

VC 80
occur in sediment beds, based on prior research findings ETH
12
(Himmelheber et al., 2007). The cis-DCE was obtained from
60
a saturated stock solution of cis-DCE (i.e., NAPL present) in
8
autoclaved, sparged simulated groundwater and supplied to
40
the influent at a concentration of 200  42 mM. After 3.4 pore
volumes, however, the only source of chlorinated ethenes to 4
20
Bioreactive Cap B was the sediment effluent. Lactate (5 mM)
was provided from 0 to 13.3 pore volumes, at which point it 0 0
was removed from the influent and no electron donor was 0 3 6 9 12 15 18
provided for the remainder of the experiment. Time (days)

Fig. 2 e Batch microcosm results of OW culture provided

2.4. Analytical methods
PCE, TCE, DCE isomers, VC, ethene, ethane, and methane
concentrations were determined from the headspace of 5 mL
aqueous effluent samples, which were analyzed using an
Agilent 6890 gas chromatograph (GC) equipped with a flame
ionization detector (FID), as described previously (Carr and
Hughes, 1998). Bromide was measured using a Dionex DX-
only PCE and sediment effluent. Chlorinated ethenes are
reported as the sum of aqueous and gas phases within the
microcosms. Error bars represent one standard deviation
calculated from triplicate reactors. The shaded background
area corresponds to methane production (mM) at each time
point and is referenced to the right vertical axis. Total
methane is calculated as the sum of aqueous and gas
phase methane.
5370 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4
compete with methanogens for electron donors, which could
result in reduced dechlorination efficiency over time. Metha-
nogens and other microbial populations indigenous to the
sediment are also expected to populate the cap material
(Himmelheber et al., 2009) and may, therefore, compete for
electron donor and other nutrients.
3.2. Bioreactive Cap A
The pore volume of Bioreactive Cap A was estimated to be
62.7 mL from differences between wet and dry column mass
and assuming complete water saturation (Table 1). The non-
reactive tracer test conducted at the onset of Bioreactive Cap
A operation yielded a symmetrical breakthrough curve,
indicative of the absence of immobile regions of water
(see Supplementary Information, Fig. 2). The measured tracer
BTC, expressed as the relative concentration versus
number of dimensionless pore volumes applied, was fit to
an analytical solution of the one-dimensional advective-
dispersive reactive (ADR) transport equation using the CXTFIT
model (van Genuchten, 1981). As anticipated, the fitted retar-
dation factor (RF) obtained from the tracer BTC was approxi-
mately equal to 1.0, indicating no detectable interactions
between the solid phase and tracer during transport through
the sand column. The fitted Peclet number (Pe) was approxi-
mately 81, yielding a hydrodynamic dispersion coefficient (DH)
of 2.8  108 m2 s1 and a hydrodynamic dispersivity (aD) of
0.37 cm. These data are consistent with values reported for
similar water-saturated columns packed with graded sands,
and indicate that advective flow and transport through
the column was normal and not subject to physical non-
equilibrium.
Chlorinated ethene effluent product distributions,
normalized to moles of chlorinated ethenes and ethene
eluted, are shown for Bioreactive Cap A (Fig. 3B). The applied
influent flow rate of 5.5 mL h1 corresponded to a column
residence time of 0.47 days. When Bioreactive Cap A was
connected in series to the sediment column, cis-DCE was the
predominant chlorinated ethene present in influent solution.
The bioreactive sand column was initially able to dechlorinate
cis-DCE to VC, but ethene was not detected (Fig. 3B). This
dechlorination activity disappeared prior to 5 pore volumes,
and eventually only 5% of the cis-DCE was dechlorinated to
VC, indicating that Dehalococcoides activity was impaired.
Methane data collected during the Bioreactive Cap A
experiment reveal that microbes other than dechlorinators
also lost activity, suggesting microbial impairment in the
system as a whole and not just for the dechlorinating pop-
ulation (Fig. 3C). Based on data presented in Fig. 3B, the
sediment column effluent was not able to sustain the dech-
lorinating consortium OW without additional amendments.
Data were not collected to determine if non-contaminant
stressors (e.g., ammonia) were present in the sediment,
which could suppress microbial activity. However, previous
research (Himmelheber et al., 2007) has demonstrated that
microorganisms, specifically Dehalococcoides strains, can be
stimulated in the Anacostia sediment with the addition of
electron donor, suggesting that non-contaminant stressors
were not a major concern in the system. Previous research
(Himmelheber et al., 2007) has also demonstrated that
microbial activity in the sediment column was limited by
electron donor availability. It was therefore hypothesized that
the levels of electron donor eluting from the sediment column
effluent prevented the dechlorinating community in the sand
cap from maintaining sufficient activity to achieve complete
reductive dechlorination of the cis-DCE introduced to the sand
cap column. A second possibility is that the relatively high
flow rates through the sand cap column did not provide
sufficient contact time between the contaminants and the
dechlorinating community to achieve complete reductive
dechlorination.
3.3. Bioreactive Cap B
To address the hypotheses raised above, the second sand
column, Bioreactive Cap B, was operated at three different
flow rates with and without the addition of lactate as an
electron donor and cis-DCE as an electron acceptor (Table 1).
The experimental conditions associated with Bioreactive Cap
B are presented in Fig. 4A, while normalized chloroethene
product distributions are shown in Fig. 4B. The pore volume
for Bioreactive Cap B was estimated to be 61.8 mL from
differences between wet and dry column mass and assuming
complete water saturation (Table 1). Prior to supplying Bio-
reactive Cap B with sediment effluent, the inoculated column
was able to completely dechlorinate cis-DCE to ethene when
provided electron donor, carbon sources, vitamins, and
reduced media; confirming the ability of the OW culture to
achieve complete dechlorination within the column
(Supplementary Information, Fig. S.2). After applying one pore
volume of simulated groundwater, sediment effluent was
supplied to Bioreactive Cap B, indicated as pore volume 0 in
Fig. 4A and B. The influent for Bioreactive Cap B was the
sediment column effluent from 146 to 180 sediment pore
volumes, which contained a mixture of cis-DCE, VC, and
ethene. The influent solution provided to Bioreactive Cap B
was initially augmented with cis-DCE to yield a total influent
chloroethene concentration of approximately 200 mM and
5 mM lactate, operated at a residence time of 2 days (flow
rate ¼ 1.29 mL h1) (Table 1). Incomplete dechlorination was
observed during this period, with a mix of VC and ethene in
the sand column effluent. From 3.4 to 5.7 pore volumes, only
lactate was provided to the influent porewater (i.e., no cis-DCE
was added) and the sediment effluent served as the sole
source of chlorinated ethenes (ca. 34 mM). The sand column
successfully achieved complete reductive dechlorination of
the applied chlorinated ethenes to ethene during this period,
demonstrating that with lactate addition and a residence time
of 2 days the sand cap could detoxify the flux of chlorinated
ethenes exiting the sediment column. These data, coupled
with the lack of complete dechlorination during the previous
condition (0e3.4 sand pore volumes) when additional cis-DCE
was provided to the influent, suggests that high chloroethene
concentrations entering the sand column limited the extent of
dechlorination.
The results obtained from Bioreactive Cap B indicate that
electron donor concentrations and contaminant residence
times within the cap can impact dechlorination activity.
Complete dechlorination was observed between 8.0 and 13.3
pore volumes when lactate was provided to the sand column
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4 5371

A Bioreactive Cap Experimental Conditions

Sand Cap connected in series with mud effluent
No lactate provided
R = 0.47 days

0 5 10 15 20 25 30 35
B Bioreactive Sand Cap Effluent Product Distribution

Total Chloroethenes + Ethenes Eluted

Product Distribution Normalized to 1.0 1.0

0.8 0.8 DCE

VC
ETH
0.6 0.6 Ethane

0.4 0.4

0.2 0.2

0.0 0.0
0 5 10 15 20 25 30 35
Pore Volumes Eluted from Cap
C Bioreactive Sand Cap Effluent Methane Production
4 4
Cummulative Methane (mg/L)

3 3

2 2

1 1

0 0
0 5 10 15 20 25 30 35
Pore Volumes Eluted from Cap

Fig. 3 e AeB. (A) Operating conditions for Bioreactive Cap A. The effluent of the sediment column served as the influent of
the sand column and was not amended with exogenous electron donors, electron acceptors, carbon sources, minerals, nor
vitamins. (B) Effluent product distribution of Bioreactive Cap A inoculated with a PCE-to-ethene dechlorinating mixed
consortia and connected in series with sediment column effluent between 68 and 83 sediment pore volumes. (C) Cumulative
aqueous methane concentration in samples collected from Bioreactive Cap A effluent.

despite relatively fast flow rates (0.47 residence time). When 3.4. Implications for capping
lactate was removed from the influent at 13.3 pore volumes,
however, a mixture of chlorinated ethenes was observed in The combined results from the batch study and the two sand
the effluent, indicating the importance of exogenous reducing columns suggest that the sediment effluent alone could not
equivalents to the sand column. Delivery of external electron sustain complete dechlorination in a bioreactive cap over the
donor is a common technique used to stimulate and enhance range of residence times (0.5e2 days) examined in this study.
reductive dechlorination in groundwater aquifers (Anderson Batch results showed complete dechlorination occurred after
et al., 2003; Haas and Trego, 2001; Lendvay et al., 2003; 19 days when the OW consortium was provided only sediment
Scow and Hicks, 2005) and may be necessary for bioreactive effluent, much longer than the 2 day retention times utilized
caps employing anaerobic biotransformations. Contaminant for these column studies. However, at sites where diffusive
mass entering the cap also dictated performance, as noted conditions exist, or where groundwater seepage rates are
above, since incomplete dechlorination was observed when significantly slower than those employed here, complete
additional cis-DCE was supplemented into the influent dechlorination could be achieved. For instance, at the USGS
(0e3.4 PV) while complete dechlorination was observed when biomat pilot test described by Majcher et al. (2009), a bio-
the cap was only treating sediment effluent (5e10 PV). reactive layer successfully dechlorinated a range of
5372 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4

Bioreactive Cap Experimental Conditions

A

5 mM Lactate

5 mM Lactate
Rt = 0.47 day
DCE to influent

5 mM Lactate

Rt = 1 day
5 mM Lactate

Rt = 2 days
R = 2 days
No Lactate
R = 0.47 day

0 5 10 15 20 25 30 35

B
Bioreactive Sand Cap Effluent Product Distribution
Total Chloroethenes + Ethene Eluted 1.0 1.0
Product Distribution Normalized to

0.8 0.8

0.6 0.6

0.4 0.4

DCE
0.2 0.2 VC
ETH
Ethane
0.0 0.0
0 5 10 15 20 25 30 35
Pore Volumes Eluted from Cap
C Bioreactive Sand Cap Effluent Cummulative Methane

80 80
Cummulative Methane (mg/L)

60 60

40 40

20 20

0 0
0 5 10 15 20 25 30 35
Pore Volumes Eluted from Cap
Fig. 4 e AeB. (A) Operating conditions for Bioreactive Cap B. The effluent of the sediment column served as the influent of
the Bioreactive Cap and was not amended with minerals nor vitamins. Exogenous electron donor (lactate), carbon sources
(lactate), and electron acceptor (cis-DCE) were added where indicated. (B) Effluent product distribution of Bioreactive Cap B
inoculated with a PCE-to-ethene mixed dechlorinating consortia and connected in series with sediment column effluent
between 146 and 185 sediment pore volumes. (C) Cumulative aqueous methane concentration in samples collected from
Bioreactive Cap B effluent.

chlorinated aliphatics at a site where average hydraulic resi- complete reductive dechlorination of dissolved chlorinated
dence times in the reactive mat were assumed to be 8e14 ethenes to ethene.
days. Thissystem also included an organic layer composed of  Incorporation of electron donor was required to stimulate
a mixture of peat, compost, and chitin to provide long-term and sustain long-term contaminant biotransformations in
electron donor. a bioreactive cap under the conditions tested. At sites with
lower seepage velocities, allowing for greater residence time
4. Conclusions in the cap, complete dechlorination without electron donor
may be possible.
Based on the results presented herein,  The need for electron donor delivery in bioactive design
could support greater cell numbers of degrading pop-
 Engineered controls may be needed to maintain microbial ulations, resulting in greater degradation rates and possible
dechlorination activity, reduce contaminant flux, or increase deployment at sites with reasonably high contaminant flux
contaminant residence time for bioreactive caps to achieve (e.g, high concentrations, high flow rates). This is supported
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4
by data in this study during Bioreactive Cap B, where the
addition of electron donor, albeit at relatively high concen-
trations, supported complete dechlorination under relative
short residence times (1e2 days).
 Careful attention should be provided to accurately charac-
terize seepage rates and contaminant concentrations at
sites where contaminated groundwater seeps are present.
In summary, this study examined the conditions governing
the implementation of novel subaqueous bioreactive in situ
caps. Experimental results suggest that the process is feasible
provided that sufficient electron donor and contaminant mass
fluxes exist in the bioactive cap.
Acknowledgments
Funding for this research was provided by the Hazardous
Substance Research Center-South and Southwest, the
National Institute of Environmental Health Sciences, and
a fellowship to D.W.H from the Georgia Institute of Technology.
Appendix. Supplementary information
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.06.022.
references
Abe, Y., Aravena, R., Zopfi, J., Parker, B., Hunkeler, D., 2009.
Evaluating the fate of chlorinated ethenes in streambed
sediments by combining stable isotope, geochemical and
microbial methods. Journal of Contaminant Hydrology 107
(1e2), 10e21.
Adamson, D.T., Newell, C.J., 2009. Support of source zone
bioremediation through endogenous biomass decay and
electron donor recycling. Bioremediation Journal 13 (1), 29e40.
Anderson, R.T., Vrionis, H.A., Ortiz-Bernad, I., Resch, C.T., Long, P.E.,
Dayvault, R., Karp, K., Marutzky, S., Metzler, D.R., Peacock, A.,
White, D.C., Lowe, M., Lovley, D.R., 2003. Stimulating the in situ
activity of Geobacter species to remove uranium from the
groundwater of a uranium-contaminated aquifer. Applied and
Environmental Microbiology 69 (10), 5884e5891.
Carr, C.S., Hughes, J.B., 1998. Enrichment of high rate PCE
dechlorination and comparative study of lactate, methanol,
and hydrogen as electron donors to sustain activity.
Environmental Science & Technology 32 (12), 1817e1824.
Cho, Y.-M., Smithenry, D.W., Ghosh, U., Kennedy, A.J.,
Millwarde, R.N., Bridges, T.S., Luthy, R.G., 2007. Field methods
for amending marine sediment with activated carbon and
assessing treatment effectiveness. Marine Environmental
Research 64 (5), 541e555.
Choi, H., Agarwal, S., Al-Abed, S.R., 2009. Adsorption and
simultaneous dechlorination of PCBs by GAC impregnated
with ZVI/Pd bimetallic particles: mechanistic aspects and
reactive capping barrier concept. Environmental Science &
Technology 43 (2), 488e493.
Conant, B., Cherry, J.A., Gillham, R.W., 2004. A PCE groundwater
plume discharging to a river: influence of the streambed and
5373
near-river zone on contaminant distributions. Journal of
Contaminant Hydrology 73 (1e4), 249e279.
Daprato, R.C., Löffler, F.E., Hughes, J.B., 2006. Comparative
analysis of three tetrachloroethene to ethene halorespiring
consortia suggests functional redundancy. Environmental
Science & Technology 41 (7), 2261e2269.
Dries, J., Bastiaens, L., Springael, D., Agathos, S.N., Diels, L., 2004.
Competition for sorption and degradation of chlorinated
ethenes in batch zero-valent iron systems. Environmental
Science & Technology 38 (10), 2879e2884.
Go, J.L., Lampert, D.J., Stegemann, J.A., Reible, D.D., 2009.
Predicting contaminant fate and transport in sediment caps:
mathematical modeling approaches. Applied Geochemistry.
24 (7), 1347e1353.
Haas, J.E., Trego, D.A., 2001. A field application of hydrogen-
releasing compound (HRC) for the enhanced bioremediation
of methyl tertiary butyl ether (MTBE). Soil & Sediment
Contamination 10 (5), 555e575.
Hamonts, K., Kuhn, T., Maesen, M., Bronders, J., Lookman, R.,
Kalka, H., Diels, L., Meckenstock, R.U., Springael, D.D.,
Dejonghe, W., 2009. Factors determining the attenuation of
chlorinated aliphatic hydrocarbons in eutrophic river
sediment impacted by discharging polluted groundwater.
Environmental Science & Technology 43 (14), 5270e5275.
Himmelheber, D.W., Pennell, K.D., Hughes, J.B., 2007. Natural
attenuation processes during in situ capping. Environmental
Science & Technology 41 (15), 5306e5313.
Himmelheber, D.W., Taillefert, M., Pennell, K.D., Hughes, J.B., 2008.
Spatial and temporal evolution of biogeochemical processes
following in situ capping of contaminated sediments.
Environmental Science & Technology 42, 4113e4120.
Himmelheber, D.W., Thomas, S.H., Löffler, F.E., Taillefert, M.T.,
Hughes, J.B., 2009. Microbial colonization of an in situ sediment
cap and correlation to stratified redox zones. Environmental
Science & Technology 43, 66e74.
Hyun, S., Jafvert, C.T., Lee, L.S., Rao, P.S.C., 2006. Laboratory
studies to characterize the efficacy of sand capping a coal tar-
contaminated sediment. Chemosphere 63 (10), 1621e1631.
Jacobs, P.H., Förstner, U., 1999. Concept of subaqueous capping of
contaminated sediments with active barrier systems (ABS)
using natural and modified zeolites. Water Research 33 (9),
2083e2087.
Krumins, V., Park, J.-W., Son, E.-K., Rodenburg, L.A., Kerkhof, L.J.,
Häggblom, M.M., Fennell, D.E., 2009. PCB dechlorination
enhancement in Anacostia River sediment microcosms.
Water Research 43 (18), 4549e4558.
Lendvay, J.M., Dean, S.M., Adriaens, P., 1998. Temporal and
spatial trends in biogeochemical conditions at a groundwater-
surface water interface: implications for natural
bioattenuation. Environmental Science & Technology 32 (22),
3472e3478.
Lendvay, J.M., Löffler, F.E., Dollhopf, M., Aiello, M.R., Daniels, G.,
Fathepure, B.Z., Gebhard, M., Heine, R., Helton, R., Shi, J.,
Krajmalnik-Brown, R., Major, C.L., Barcelona, M.J.,
Petrovskis, E., Hickey, R., Tiedje, J.M., Adriaens, P., 2003.
Bioreactive barriers: a comparison of bioaugmentation and
biostimulation for chlorinated solvent remediation.
Environmental Science & Technology 37 (7), 1422e1431.
Löffler, F.E., Edwards, E.A., 2006. Harnessing microbial activities
for environmental cleanup. Current Opinion in Biotechnology
17 (3), 274e284.
Lorah, M.M., Voytek, M.A., 2004. Degradation of 1,1,2,2-
tetrachloroethane and accumulation of vinyl chloride in wetland
sediment microcosms and in situ porewater: biogeochemical
controls and associations with microbial communities. Journal
of Contaminant Hydrology 70 (1e2), 117e145.
Lorah, M.M., Majcher, E.H., Jones, E.J., Voytek, M.A., 2008.
Microbial Consortia Development and Microcosm and Column
5374 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 6 5 e5 3 7 4
Experiments for Enhanced Bioremediation of Chlorinated
Volatile Organic Compounds, West Branch Canal Creek
Wetland Area, Aberdeen Proving Ground, Maryland. USGS
Scientific Investigations Report. 2007-5165, pp. 1e79.
Lowry, G.V., Johnson, K.M., 2004. Congener specific PCB
dechlorination by microscale and nanoscale zero-valent iron
in a methanol/water solution. Environmental Science &
Technology 38 (19), 5208e5216.
Majcher, E.H., Lorah, M.M., Phelan, D.J., McGinty, A.L., 2009.
Design and performance of an enhanced bioremediation pilot
test in a tidal wetland seep, West Branch Canal Creek,
Aberdeen Proving Ground, Maryland. USGS Scientific
Investigations Report 2009-5112, 1e70.
Majcher, E.H., Phelan, D.J., Lorah, M.M., McGinty, A.L., 2007.
Characterization of Preferential Ground-water Seepage from
a Chlorinated Hydrocarbon-contaminated Aquifer to West
Branch Canal Creek, Aberdeen Proving Ground, Maryland, 2002-
2004. USGS Scientific Investigations Report. 2006-5233, pp. 1e193.
McDonough, K.M., Murphy, P., Olsta, J., Zhu, Y.W., Reible, D.,
Lowry, G.V., 2007. Development and placement of a sorbent-
amended thin layer sediment cap in the Anacostia River. Soil
& Sediment Contamination 16 (3), 313e322.
Murphy, P., Marquette, A., Reible, D., Lowry, G.V., 2006. Predicting
the performance of activated carbon-, coke-, and soil-
amended thin layer sediment caps. Journal of Environmental
Engineering e ASCE 132 (7), 787e794.
Reible, D., 2005. Personal communication.
Reible, D., Hayes, D., Lue-Hing, C., Patterson, J., Bhowmik, N.,
Johnson, M., Teal, J., 2003. Comparison of the long-term risks
of removal and in situ management of contaminated
sediments in the Fox River. Soil & Sediment Contamination 12
(3), 325e344.
Reible, D.D., Lampert, D., Constant, D.W., Mutch, R.D., Zhu, Y.,
2007. Active capping demonstration in the Anacostia River,
Washington, DC. Remediation 17 (1), 39e53.
Scow, K.M., Hicks, K.A., 2005. Natural attenuation and enhanced
bioremediation of organic contaminants in groundwater.
Current Opinion in Biotechnology 16 (3), 246e253.
SERDP and ESTCP, 2008. Expert Panel Workshop on Research and
Development Needs for Understanding and Assessing the
Bioavailability of Contaminants in Soils and Sediments.
Thoma, G.J., Reible, D.D., Valsaraj, K.T., Thibodeaux, L.J.,
1993. Efficiency of capping contaminated sediments in situ.
2. Mathematics of diffusion adsorption in the capping
layer. Environmental Science & Technology 27 (12),
2412e2419.
van Genuchten, M., 1981. Non-equilibrium Transport Parameters
from Miscible Displacement Experiments. Research Report
119, U.S. Salinity Lab, USDA, pp. 1e94.
Zimmerman, J.R., Ghosh, U., Millward, R.N., Bridges, T.S.,
Luthy, R.G., 2004. Addition of carbon sorbents to reduce PCB
and PAH bioavailability in marine sediments:
physicochemical tests. Environmental Science & Technology
38 (20), 5458e5464.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Preparation of ion exchanger layered electrodes for advanced

membrane capacitive deionization (MCDI)

Ju-Young Lee, Seok-Jun Seo, Sung-Hyun Yun, Seung-Hyeon Moon*

School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), 261 Cheomdan-gwagiro, Buk-gu,
Gwangju 500-712, Republic of Korea

article info abstract

Article history: A noble electrode for capacitive deionization (CDI) was prepared by embedding ion
Received 18 February 2011 exchanger onto the surface of a carbon electrode to practice membrane capacitive deion-
Received in revised form ization (MCDI). Bromomethylated poly (2, 6-dimethyl-1, 4-phenylene oxide) (BPPO) was
30 April 2011 sprayed on carbon cloth followed by sulfonation and amination to form cation exchange
Accepted 22 June 2011 and anion exchange layers, respectively. The ion exchange layers were examined by
Available online 3 July 2011 Scanning electron microscopy (SEM) and Fourier transform infrared spectrometer (FT-IR).
The SEM image showed that the woven carbon cloth was well coated and connected with
Keywords: BPPO. The FT-IR spectrum revealed that sulfonic and amine functional groups were
Capacitive deionization (CDI) attached on the cationexchange and anionexchange electrodes, respectively. The advan-
BPPO tages of the developed carbon electrodes have been successively demonstrated in a batch
Carbon cloth and a continuous mode CDI operations without ion exchange membranes for salt removal
Spraying using 100 mg/L NaCl solution.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction et al., 2010). Hereby utilization of the adsorption ability of an

Recently, capacitive deionization (CDI) has been used as
a water treatment technology due to the simple principle
and low operating potential, without the need for chemicals
(Ito et al., 2007). Therefore, it is considered as an
environmentally-friendly and economical system (Foo and
Hameed, 2009). A CDI system operation consists of adsorp-
tion and desorption periods for obtaining purified water and
concentrated water, respectively (Lee et al., 2010). When an
electric potential is applied to CDI cells, charged ions in
contaminant water are adsorbed onto the surface of charged
electrodes, and formed an electric double layer due to the
charged electrode and adsorbed ions, producing purified
water. After the adsorption of ions, the saturated electrode
undergoes regeneration by desorption of the adsorbed ions
under zero electrical potential or reversed electric field (Seo
electrode is the key parameter for the CDI operation. In order
to maintain acceptable operation efficiencies, the complete
adsorption and desorption of charged ions should be accom-
plished within appropriate periods. Practically, however,
when a potential is applied to a CDI cell, counter ions are
attracted onto the electrode surface, simultaneously co-ions
expelled from the counter electrode (Kim and Choi, 2010). It
leads to a higher energy consumption and a lower operation
efficiency due to mobility of unwanted ions.
To avoid this phenomenon, a membrane-CDI (MCDI) is
employed with the help of ion selective membranes in the CDI
cell. A MCDI has two types of ion exchange membranes, i.e.
anion exchange and cation exchange membranes (AEM &
CEM, respectively). The AEM and CEM are positioned in front
of the positively and negatively charged electrodes, respec-
tively (Lee et al., 2006). The ion exchange membrane has the

* Corresponding author. Fax: þ82 62 715 2434.

E-mail address: shmoon@gist.ac.kr (S.-H. Moon).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.06.028
5376 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0
ability to selectively permeate ions, i.e., a CEM permit the
passage of cations only, while an AEM allow the passage of
anions only. The selectivity of ion exchange membranes
prevent reverse adsorption and prohibit the mobility of
unwanted ions. However, a MCDI requires strong physical
pressure for smooth contact between the membrane and
electrode material. Also the diffusion layer on the membrane
surface will become thick when the concentration of
contaminants in the feed water is low (Dlugolecki et al., 2010).
Accordingly, this phenomenon induces a decreased mobility
of wanted ions and a high interfacial resistance of the
membrane.
In this study to solve the problems of contact resistance
and the diffusion layer, an advanced-MCDI (A-MCDI) was
developed by adhering the ion exchanger onto the carbon
electrode surface as a thin layer, which reduces the contact
resistance between the ion exchanger and electrode of
a MCDI. Therefore, an A-MCDI is expected to exhibit a high
removal efficiency and a low current consumption compared
to a conventional MCDI.
2. Experimental
2.1. Materials
Bromomethylated poly (2, 6-dimethyl-1, 4-phenylene oxide)
(BPPO) was used as the base polymer, and N-methyl-2-
pyrrolidone (NMP, Fluka, Japan) was purchased as the
solvent to dissolve BPPO. Carbon cloth (Kuraray, Japan),
3 cm  8.5 cm in size, was employed as the carbon electrode
material. The BPPO embedded electrode was sulfonated using
sulfuric acid (H2SO4 99%, DC chemical, Korea), while amina-
tion was performed using trimethylamine (TMA, 25wt % in
D.I.water, Aldrich). In the salt removal test, sodium chloride
(NaCl, Dongyang Chemical, Korea) was used to prepare the
feed water.
2.2. Preparation of the ion exchanger embedded
electrodes
Embedding the ion exchanger onto the electrode surface was
carried out in two steps; coating the carbon electrode with
a base polymer and then attaching functional groups onto the
polymer.
BPPO (donated by laboratory of fundamental membranes
in USTC) slurry was prepared by mixing 1 g of BPPO with 5 ml
of NMP at room temperature for 1 day. The BPPO slurry was
then sprayed onto the surface of one side of the carbon cloth
using an air brush. In order to evaporate the NMP, the elec-
trode was dried in an oven at 40  C for 12 h, forming the BPPO
coated electrodes for sulfonation and amination.
The sulfonation of the BPPO embedded electrode was
conducted using four different solution concentrations, i.e. 99,
80, 50 and 30% sulfuric acid solutions. The BPPO embedded
carbon cloth was initially immersed in 99% sulfuric acid
solution for 20 min, and subsequently moved to each of the 80,
50 and 30% sulfuric acid solutions for 1 min. The embedded
electrode was finally maintained in distilled water (Liu et al.,
2006).
The amination of the BPPO embedded electrode was con-
ducted by immersion of the electrode in 25% TMA for 20 min.
The embedded electrode was maintained in distilled water
(Tang et al., 2005).
2.3. Characterization of the embedded electrodes
The surface morphology of the BPPO embedded electrodes
were characterized using Scanning electron microscopy (SEM,
JEOL, Japan). Both the bare and embedded carbon electrodes
were scanned to compare the morphologies before and after
coating with the BPPO slurry, at magnifications of 2,000 and
10,000 times.
The sulfonated and aminated electrodes were examined
using Fourier transform infrared spectrometry (FT-IR, 460
Plus, Jasco Japan) within the wavelength range
400e4000 cm1. FT-IR is used to confirm the functional groups
via the bond vibration and stretching energies between atoms.
The electrode with the base polymer, sulfonated and ami-
nated electrodes were analyzed to compare the peaks corre-
sponding the functionalization.
2.4. Salt removal test
2.4.1. Comparison of the CDI, MCDI and A-MCDI
To compare the salt removal efficiency of the A-MCDI,
experiments were performed using 3 cm  8.5 cm CDI, MCDI
and A-MCDI unit cells, with simultaneous recording of the ion
conductivity and current in a batch system (Broséus et al.,
2009). Fifty mililiters of feed water was continuously circu-
lated by a pump (Masterflex Cole-parmer, USA). The initial
conductivity of feed water, 100 mg/L sodium chloride, was
190 mS/cm. The variation in the ion conductivity was
measured every 30 s in the reservoir using a TDS conductivity
meter (OAKTON, Japan). A potential of 1.8 V was applied to the
CDI cells using an Agilent 6613C (Agilent, USA) power supply.
Feed water was continuously circulated at a flow rate of
4 ml/min, which was determined with respect to the surface
area of the electrode.
Fig. 1 shows schematic diagrams of the assembled CDI,
MCDI and A-MCDI cells. In the structure of the CDI cell, the
electrodes were located between the current collectors;
whereas, in case of the MCDI cell, the positively and negatively
charged electrodes were located behind the anion exchange
(AMX, Tokuyama, Japan) and cation exchange membranes
(CMX, Tokuyama, Japan), respectively. In the A-MCDI cell, the
sulfonated BPPO embedded electrode was positioned in front
of negatively charged current collector, while the aminated
BPPO embedded electrode was positioned in front of positively
charged current collector. The arrangement of the ion
exchangers in A-MCDI is the same as MCDI, because even
though the ion exchangers have the ion exchange capacity for
the selected ions, the system still requires electrical potential
as a driving force between the cathode and anode. The three
systems were compared in terms of their removal efficiencies
and energy consumptions.
2.4.2. Cyclic testing of the A-MCDI
Cyclic testing of the A-MCDI was performed to demonstrate its
ability of repeated operation in a continuous mode (Dai et al.,
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0 5377

Fig. 1 e Schematics of the CDI, MCDI and A-MCDI cell structures.

2005). The experiment was carried out under the same
conditions as the previous salt removal test and subjected to
repeated periods of adsorption and desorption. The adsorp-
tion and desorption periods were determined to be 5 and
5 min, respectively, to allow sufficient time for adsorption and
desorption.
3. Results and discussion
3.1. Morphologies of the electrodes
electrode surface is observed to have clean surface, with only
a tiny amount of dust, whereas, in case of the ion exchanger
embedded electrode, the polymer is filled within the carbon
cloth and interconnects the individual fibers. In a higher
magnification, the bare electrode shows porous carbon cloth
surface. However, the ion exchanger embedded electrode
shows the pores blocked by the base polymer. Consequently,
it was found that the ion exchanger layer was well formed on
the surface of the carbon cloth, so that the ion-conducting
surface of the embedded electrode may provide good ion
selectivity and enhanced conductivity.

Fig. 2 shows the SEM images of the surfaces of the bare elec-
trode and ion exchanger embedded electrodes. The left and
right sides of the figure are the bare carbon cloth and ion
exchanger embedded electrode surfaces, respectively. Each
sample was magnified by 2,000 and 10,000 times, respectively.
Originally the woven carbon cloth material was compli-
catedly interlinked (Ahn et al., 2007). The bare carbon
3.2. FT-IR analyses
To impart ion exchange capabilities, the sulfonic and amine
functional groups should be attached to the base polymer. The
sulfonic and amine groups have the abilities to transport
cations and anions, respectively in accordance with the
Grotthuss mechanism (Agmon, 1995) and vehicle mechanism

Fig. 2 e SEM images of (a) bare carbon cloth surface at 2,000 times, (b) embedded carbon electrode surface at 2,000 times, (c)
bare carbon cloth surface at 10,000 times, (d) embedded carbon cloth surface at 10,000 times.
5378 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0
Fig. 3 e FT-IR analyses of the untreated, sulfonated and aminated electrode surfaces.
(Schuster et al., 2008). In order to examine the sulfonic and
amine groups on the BPPO surface, three types of electrodes
prepared with the base polymer, sulfonated and aminated
polymers were analyzed using FT-IR.
Fig. 3 shows the existence of functional groups, indicated
by the unique peaks in the sulfonated and aminated BPPO.
When the sulfonated BPPO was compared with the untreated
BPPO, the sulfonic group peak appeared around
1165e1120 cm1 (Panicker et al., 2006). Similarly, when the
aminated BPPO was compared with the untreated BPPO, the
amine group peak appeared around 850e750 cm1 (Volkov
et al., 1980). The results show that the functional groups
were successfully attached to the BPPO.
3.3. Salt removal test
3.3.1. Comparison of the CDI, MCDI and A-MCDI
CDI, MCDI, and A-MCDI tests were performed to observe the
initial adsorption capability of each system. The results
Fig. 4 e Variations in the ion conductivity of the CDI,
membrane-CDI (MCDI) and advanced-MCDI (A-MCDI)
during their operation.
obtained for the CDI, MCDI and A-MCDI were compared in
terms of salt removal and current consumption, as shown in
Figs. 4 and 5, respectively, at an applied potential of 1.8 V and
flow rate of 4 ml/min over a 30 min period.
As shown in Fig. 4, the CDI and A-MCDI exhibited higher
degrees of removal than the MCDI. It means that the removal
efficiency of MCDI was lower than the CDI and A-MCDI at the
same voltage, because the MCDI suffered problems, such as
contact resistance and membrane resistance.
Fig. 5 shows that the MCDI and A-MCDI allowed lower
currents than that of the CDI. This phenomenon is due to the
fact that the current depends on the total ionic flux. In other
words, the ion selectivity increases the ion mobility resistance
at a constant voltage; therefore, a high resistance will
decrease the current in accordance with Ohm’s law (I ¼ V/R)
(Kim and Choi, 2010).
The operation parameters, the ion removal efficiency and
power consumption of each system, are listed in Table 1.
These results were obtained based on the measured voltages
and currents at 30 min after starting the operation.
Fig. 5 e Operating currents of the CDI, MCDI and A-MCDI.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0 5379

concentration of the charged ions in the effluent was reduced.

Table 1 e Numerical results of the ion removal
efficiencies, currents and power consumptions during When the saturated electrode underwent regeneration with
the operation of each system. zero electrical potential, the adsorbed ions were desorbed.
Accordingly, the charged ions in the effluent increased again.
CDI MCDI A-MCDI
The results showed that the repeated adsorption and
Applied electric potential 1.8 V 1.8 V 1.8 V desorption occurred in a regular pattern of a CDI system. This
Initial conductivity 190.8 mS/cm 185.1 mS/cm 188.9 mS/cm
implies that the A-MCDI system would have stable perfor-
Conductivity at 30 min 39.9 mS/cm 168 mS/cm 31.5 mS/cm
mance over repeated operation; thus, demonstrated the
Removal efficiency 79.1% 9.23% 83.4%
Current at 30 min 235.7 mA 2.575 mA 25.64 mA practical applicability as a water treatment system.
Power consumption 212.3 mWh 2.318 mWh 23.07 mWh

The variation in the feed water conductivity was measured 4. Conclusion

to get the removal efficiency, h which was determined using
the following equation (Dermentzis and Ouzounis, 2008):
Ci  C
hð%Þ ¼  100
Ci
where Ci is the initial conductivity [mS/cm] of the feed water
and C is the conductivity of the diluted feed water [mS/cm] at
30 min after starting the operation. According to Eq. (1), the
removal efficiencies of the CDI, MCDI and A-MCDI were
approximately 79, 9.2 and 83%, respectively. Obviously, the
removal efficiency of A-MCDI was the highest.
The power consumption of each system was calculated
based on the current measured at the same applied potential
according to the following equation (Masiuk, 1999):
E ¼ VIt
An A-MCDI has been developed by introducing an ion
exchanger layer on electrode surface for a CDI system. The
electrodes were prepared by coating a base polymer on the
(1)
carbon cloth followed by functionalization of sulfonic and
amine groups on the polymer structure. The noble electrode
enables to overcome the drawbacks of a membrane-CDI
system by reducing the interfacial resistance between the
ion exchanger layer and the carbon electrode. Practically the
A-MCDI is operated without ion exchange membranes while
the system performs better than a conventional MCDI in
terms of salt removal efficiency. This research may contribute
significantly in application of CDI in various water treatment
systems such as desalination, hardness removal, and drinking
water treatment.
(2)

where E is the power consumption [mWh], V the applied

potential [V], I the current [A] and t the time [h]. As a result, the
power consumptions of the CDI, MCDI and A-MCDI were
approximately 210, 2.3 and 23 mWh, respectively. Considering
the removal efficiency and energy consumption, the A-MCDI
showed better performance among the systems tested.
3.3.2. Cyclic testing of the A-MCDI
A cyclic test was performed for a continuous operation of A-
MCDI. Fig. 6 shows the variation of the ion conductivity in the
effluent stream.
During the initial adsorption period, charged ions in feed
water were adsorbed onto the electrode; therefore, the
Fig. 6 e Continuous mode operation of the A-MCDI system
by repeated cycles of adsorption and desorption.
Acknowledgment
This research was supported by a grant (07seaheroB02-02-01)
from the Plant Technology Advancement Program, funded by
the Ministry of Land, Transport and Maritime Affairs.
references
Agmon, N., 1995. The Grotthuss mechanism. Chemical Physics
Letters 244, 456e462.
Ahn, H.-J., Lee, J.-H., Jeong, Y., Lee, J.-H., Chi, C.-S., Oh, H.-J., 2007.
Nanostructured carbon cloth electrode for desalination from
aqueous solutions. Materials Science and Engineering: A
449e451, 841e845.
Broséus, R., Cigana, J., Barbeau, B., Daines-Martinez, C., Suty, H.,
2009. Removal of total dissolved solids, nitrates and
ammonium ions from drinking water using charge-barrier
capacitive deionisation. Desalination 249, 217e223.
Dai, K., Shi, L., Fang, J., Zhang, D., Yu, B., 2005. NaCl adsorption in
multi-walled carbon nanotubes. Materials Letters 59, 1989e1992.
Dermentzis, K., Ouzounis, K., 2008. Continuous capacitive
deionization-electrodialysis reversal through electrostatic
shielding for desalination and deionization of water.
Electrochimica Acta 53, 7123e7130.
Dlugolecki, P., Anet, B., Metz, S.J., Nijmeijer, K., Wessling, M., 2010.
Transport limitations in ion exchange membranes at low salt
concentrations. Journal of Membrane Science 346, 163e171.
Foo, K.Y., Hameed, B.H., 2009. A short review of activated carbon
assisted electrosorption process: an overview, current stage
and future prospects. Journal of Hazardous Materials 170,
552e559.
5380 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 7 5 e5 3 8 0
Ito, E., Mozia, S., Okuda, M., Nakano, T., Toyoda, M., Inagaki, M.,
2007. Nanoporous carbons from cypress II. Application to
electric double layer capacitors. New Carbon Materials 22,
321e326.
Kim, Y.-J., Choi, J.-H., 2010. Enhanced desalination efficiency in
capacitive deionization with an ion-selective membrane.
Separation and Purification Technology 71, 70e75.
Lee, J.-B., Park, K.-K., Eum, H.-M., Lee, C.-W., 2006. Desalination of
a thermal power plant wastewater by membrane capacitive
deionization. Desalination 196, 125e134.
Lee, J.Y., Seo, S.J., Park, J.W., Moon, S.H., 2010. A study on the cell
structure for capacitive deionization system. Korean Chemical
Engineering Research 48, 791e794.
Liu, J., Xu, T., Han, X., Fu, Y., 2006. Synthesis and characterizations
of a novel zwitterionic hybrid copolymer containing both
sulfonic and carboxylic groups via sulfonation and zwitterionic
process. European Polymer Journal 42, 2755e2764.
Masiuk, S., 1999. Power consumption measurements in a liquid
vessel that is mixed using a vibratory agitator. Chemical
Engineering Journal 75, 161e165.
Panicker, C.Y., Varghese, H.T., Philip, D., Nogueira, H.I.S., 2006.
FT-IR, FT-Raman and SERS spectra of pyridine-3-sulfonic acid.
Spectrochimica Acta Part A: Molecular and Biomolecular
Spectroscopy 64, 744e747.
Schuster, M., Kreuer, K.-D., Steininger, H., Maier, J., 2008. Proton
conductivity and diffusion study of molten phosphonic acid
H3PO3. Solid State Ionics 179, 523e528.
Seo, S.-J., Jeon, H., Lee, J.K., Kim, G.-Y., Park, D., Nojima, H., Lee, J.,
Moon, S.-H., 2010. Investigation on removal of hardness ions
by capacitive deionization (CDI) for water softening
applications. Water Research 44, 2267e2275.
Tang, B., Xu, T., Gong, M., Yang, W., 2005. A novel positively
charged asymmetry membranes from poly(2,6-dimethyl-1,4-
phenylene oxide) by benzyl bromination and in situ
amination: membrane preparation and characterization.
Journal of Membrane Science 248, 119e125.
Volkov, A., Tourillon, G., Lacaze, P.-C., Dubois, J.-E., 1980.
Electrochemical polymerization of aromatic amines: IR, XPS
and PMT study of thin film formation on a Pt electrode. Journal
of Electroanalytical Chemistry 115, 279e291.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Electrochemical sulfide oxidation from domestic wastewater

using mixed metal-coated titanium electrodes

Ilje Pikaar, René A. Rozendal, Zhiguo Yuan, Jürg Keller, Korneel Rabaey*
The University of Queensland, Advanced Water Management Centre (AWMC), Brisbane, QLD 4072, Australia

article info abstract

Article history: Hydrogen sulfide generation is a major issue in sewer management. A novel method based
Received 8 April 2011 on electrochemical sulfide oxidation was recently shown to be highly effective for sulfide
Received in revised form removal from synthetic and real sewage. Here, we compare the performance of five
11 July 2011 different mixed metal oxide (MMO) coated titanium electrode materials for the electro-
Accepted 25 July 2011 chemical removal of sulfide from domestic wastewater. All electrode materials performed
Available online 6 August 2011 similarly in terms of sulfide removal, removing 78
5%, 77
1%, 85
4%, 84
1%, and
83
2% at a current density of 10 mA/cm2 using Ta/Ir, Ru/Ir, Pt/Ir, SnO2 and PbO2,
Keywords: respectively. Elevated chloride concentrations, often observed in coastal areas, did not
Electrochemical oxidation entail any significant difference in performance. Independent of the electrode material
Oxygen generation used, sulfide oxidation by in situ generated oxygen was the predominant reaction mech-
Sulfide oxidation anism. Passivation of the electrode surface by deposition of elemental sulfur did not occur.
Sewer However, scaling was observed in the cathode compartment. This study shows that all the
MMO coated titanium electrode materials studied are suitable anodic materials for sulfide
removal from wastewater. Ta/Ir and Pt/Ir coated titanium electrodes seem the most suit-
able electrodes since they possess the lowest overpotential for oxygen evolution, are stable
at low chloride concentration and are already used in full scale applications.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction Recently, we described the electrochemical oxidation of

Hydrogen sulfide is ubiquitously present in industrial and
domestic waste streams, and is of special concern in sewer
systems since it is responsible for odor issues in urban areas,
is toxic to sewer workers, and is the main cause for sewer pipe
corrosion (Zhang et al., 2008). Repair and/or replacement of
corroded sewer pipes results in considerable costs (Sydney
et al., 1996; Vincke et al., 2002; Kaempfer and Berndt, 1998),
and therefore measures for mitigating sulfide production and
emission are required. Current strategies to prevent sewer
corrosion come with substantial costs due to both chemical
consumption and system maintenance (Zhang et al., 2008;
Hvitved-Jacobsen, 2001).
aqueous sulfide using Ta/Ir coated titanium electrodes from
domestic wastewater (Pikaar et al., 2011). At the used current
densities of 5 mA/cm2 sulfide could be oxidized, producing
elemental sulfur, thiosulfate and sulfate as the final products.
Indirect oxidation of sulfide with in situ generated oxygen
rather than the direct oxidation of sulfide at the electrode was
shown to be the predominant reaction mechanism due to the
low sulfide concentrations (i.e. w10 mg/L) normally observed
in sewers. In comparison to sulfide control with conventional
oxygen injection, the method does not require any transport
and storage of oxygen, whereas in situ generated oxygen is
expected to have a much higher efficiency due to the fine
dispersion (i.e. 1e30 mm) of the generated oxygen (Chen, 2004).

* Corresponding author. Tel.: þ61 7 3365 7519; fax: þ61 7 3365 4726.
E-mail address: k.rabaey@awmc.uq.edu.au (K. Rabaey).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.07.033
5382 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8
Electrode materials possess different selectivity and cata-
lytic activities, and may hence produce different intermedi-
ates such as oxygen, chlorine/hypochlorous acid, hydroxyl
radicals and other reactive oxygen species (Chen, 2004). These
intermediates may have a different product spectrum for
sulfide oxidation, and may increase/decrease overall oxida-
tion efficiency. In this study, the feasibility of anodic sulfide
oxidation in domestic wastewater using titanium electrodes
coated with 5 different types of electrocatalyst materials,
namely Ta/Ir, Pt/Ir, Ru/Ir, PbO2 and SnO2, was investigated.
Ta/Ir coated titanium electrodes are known for their low
overpotential for oxygen and their stability. They are therefore
commonly used as oxygen evolving electrodes (Chen, 2004).
Ru/Ir electrodes find widespread use in the chloro-alkali
industry for the in situ oxidation of chloride to chlorine/
hypochlorite (Feng and Li, 2003; Takasu et al., 2010). Both Ru/Ir
and Pt/Ir electrodes also possess a low overpotential for
oxygen evolution, and therefore especially at low chloride
concentrations oxygen evolution may become a predominant
reaction mechanism. It is important to note that these elec-
trodes have a low overpotential for the production of chlorine
from chloride. In coastal areas (e.g., Queensland, Australia)
the chloride concentrations in domestic wastewater can be
a high as >1 g/L due to marine infiltrations into the sewer
system (Taylor and Gardner, 2007). Hence, in situ generation
of hypochlorous acid/hypochlorite may become an important
reaction mechanism for sulfide oxidation from domestic
wastewater. Hypochlorous acid/hypochlorite can oxidize
sulfide to elemental sulfur at pH  7.5, values normally
observed in sewer systems. Disadvantage of in situ chlorine
generation is the possible formation of toxic organochlorine
derivatives (Sun et al., 2009), which needs to be prevented.
Contrarily to the three aforementioned materials, PbO2 and
SnO2 have a high overpotential for oxygen evolution. Just like
boron doped diamond (BDD), the electrocatalysts PbO2 and
SnO2 are known to generate hydroxyl radicals from the oxida-
tion of water (Panizza and Cerisola, 2008; Zhu et al., 2008;
Panizza et al., 2008). However, contrarily to BDD, PbO2 and
SnO2 are made of inexpensive materials which are readily
available in practical mesh geometries and at scale, and have
a low electrical resistivity. This makes them suitable materials
for applications on large scale (i.e. industrial applications and
sewer systems).
Considering the above, the main aims of this study are to
assess the impact of electrode coating on electrochemical
sulfide oxidation in wastewater, to assess the differences in
energy requirement for the different electrode materials and
to assess the impact of chloride concentrations on the sulfide
oxidation process.
2. Materials and methods
2.1. Electrochemical cell and operation
Fig. 1 gives a schematic diagram of the electrochemical cell. The
two-chambered electrochemical cell consisted of two parallel
Perspex frames (internal dimensions 20  4.8  1.2 cm) sepa-
rated by a cation exchange membrane (Ultrex CM17000,
Membranes International Inc., USA) to create an anode and
cathode compartment each with a volume of about 100 mL. In
the anode chamber, mesh shaped Ta/Ir (TaO2/IrO2: 0.35/0.65),
Ru/Ir (RuO2/IrO2: 0.70/0.30), Pt/Ir (PtO2/IrO2: 0.70/0.30), PbO2 and
SnO2 coated titanium electrodes (diameter: 240 mm; thickness:
1 mm; specific surface area: 1.0 cm2/cm2) were used (Magneto
Anodes BV, The Netherlands). Stainless steel fine mesh (24 cm2)
with a stainless steel current collector (6 mm mesh size,
0.8 mm wire connected via a 6 mm stainless steel rod) was used
as electrode material in the cathode chamber. Both the anode
and cathode had a projected electrode surface area of 24 cm2. In
all experiments, an Ag/AgCl (RE-5B, Bio Analytical, USA) was
used as the reference electrode (þ197 mV versus SHE).
The anode liquid medium was constantly recirculated over
a 5 L vessel, allowing a total anode liquid volume of 5 L. The
influent flow rate through the anode chamber was maintained
at 3.6 L/h using a peristaltic pump (Watson Marlow, UK). An
additional recirculation flow in the anode chamber, which
was kept at 22 L/h using a peristaltic pump (Watson Marlow,
UK), to obtain a good mixing rate in the anode chamber was
used. PVC tubing with an internal diameter of 8 mm was used
for the feeding and recirculation lines. The off-gas coming
from the external buffer vessel was captured in a gas collec-
tion bag (TKC Tedlar bags, Air-Met Scientific Pty Ltd, USA).
An external buffer flask of 2 L was used in the recirculation
of the cathode chamber. A 0.10 M NaCl solution in the cathode
chamber was used in all experiments. The recirculation flow
of the cathode solution was kept at 22 L/h using a peristaltic
pump (Watson Marlow, UK). The anode liquid medium,
domestic wastewater, was collected weekly from a local
pumping station and stored at 4  C. Prior to use, 5 L of the
domestic wastewater was heated up to ambient temperatures
(24.3
0.5  C) and fed to the influent buffer tank. pH in the
influent was measured using a pH probe (Ionode Pty Ltd., AU)
and maintained at 7.5 during the experiment through a PLC
controlled dosage of a 0.5 M NaOH solution. The wastewater
was continuously recirculated. This fed batch system was
used to minimize the amount of domestic wastewater needed
(continuous feeding would have required significant amounts
of wastewater per day). A concentrated sodium sulfide
(Na2S$9H2O) stock solution (
5.5 g/L sulfide-S), prepared
according to Dutta et al. (2008), was continuously supplied to
the incoming line of the anode chamber via a syringe pump
(NE-1600, New Era Pump Systems, Inc., USA) at a dosing rate of
w36 mg sulfide-S/h, which was sufficient to give an anode
influent concentration of w10.0 mg S/L by assuming that the
recirculated wastewater contained no sulfide.
2.2. Measurements and calculations
Galvanostatic measurements and control were performed
using a Wenking potentiostat/galvanostat (KP07, Bank Elek-
tronik, gmbH, Germany). The anode, cathode potentials and the
current were recorded every 60 s using an Agilent 34970A data
acquisition unit. The amount of sulfide dosed to the reactor can
be expressed as a charge quantity expressed in Coulomb:
Q ¼ nFcadded =M (1)
With F the Faraday constant (96,485 C/mol, n the number of
electrons involved (i.e. 8 electrons for the oxidation of 1 mol
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8 5383

13

Power
supply

14 14
7
12
15 10
2

+
H
+
Na 11
+ 8
3 K

1
5 6 9

Sulfide solution
4

Fig. 1 e Schematic overview experimental setup. 1. Influent buffer (5 L); 2. Influent anodic compartment; 3. Recirculation
flow anodic compartment; 4. Sulfide feeding line; 5. Anode compartment; 6. Cathode compartment; 7. Effluent anode
compartment; 8. Influent cathode compartment; 9. Cathode buffer (2 L); 10. Effluent cathode compartment; 11. Cathode
water-lock; 12. Cathode vent gas (H2); 13. Potentiostat/galvanostat; 14. Sampling points; 15. Anode vent gas (H2S, O2, CO2).

sulfide to sulfate), c the amount of sulfide added (g) and M the
molar weight of sulfide (i.e. 32 g/mol). Hence, by determining
the coulombic efficiency (CE) based on the conversion of
sulfide to sulfate the coulombic efficiency can be calculated as
follows:
(Cyberscan PC 300, Eutech Instruments). The produced gas (i.e.
in situ oxygen generation) was analyzed using a Gas Chro-
matography (Shimadzu, Japan).

nFcremoved =M cremoved 2.4. Experimental procedures

CE ¼ ¼ (2)
nFcadded =M cadded
2.3. Chemical analyses
Sulfide, sulfite, thiosulfate and sulfate concentrations were
measured by Ion Chromatography (IC), using a Dionex 2010i
system, according to Keller-Lehmann et al. (2006). Samples
were immediately filtered using a 0.22 mm syringe filter (Mil-
lipore, USA) and preserved in previously prepared Sulfide
Antioxidant Buffer (SAOB) solution prior to ion chromatog-
raphy analysis. SAOB solution was prepared using Helium
purged MilliQ (18MU) water, 3.2 g/L NaOH and 2.8 g/L a-
ascorbic acid. After preparation, the solution was kept refrig-
erated, shielded from light and not used beyond 24 h.
Elemental sulfur was assumed to be the difference between
the total sulfide added and the soluble sulfur species (i.e.
sulfide, sulfite, thiosulfate and sulfate).
COD (range 25e1500 mg/L) and free chlorine (range
0e4 mg/L) concentrations were determined by means of
cuvette tests (Merck, Germany). COD concentrations were
corrected for the soluble sulfur species in the solution. The
conductivity was measured using a hand-held meter
Experiments were divided into 2 different sets. The first set of
experiments compared the performance of five different
mixed metal-coated titanium electrode materials (Ta/Ir, Ru/Ir,
Pt/Ir, PbO2, SnO2,). Key focal points were (a) the kinetics of the
sulfide oxidation, (b) possible oxidation of organics (i.e. COD),
(c) the amount of excess in situ generated oxygen and (d) the
required energy input (i.e., based on the obtained cell poten-
tial) of the different electrode materials. The second set of
experiments investigated the influence of chloride concen-
trations on (a) the kinetics of the sulfide oxidation, (b) possible
oxidation of organics (i.e. COD) and (c) the required energy
input (i.e. obtained cell potential) using Ru/Ir and Ta/Ir coated
titanium as electrode material.
Each time, the performance was assessed during 6-h
experimental runs using galvanostatic control at a fixed
current density of 10 mA/cm2. This current density level was
enough to oxidize all sulfide added to the system to sulfate. All
experiments were performed in triplicate. Prior to each
experiment, the headspace of the influent buffer vessel was
flushed with nitrogen for at least 5 min to obtain a headspace
that consisted of 100% nitrogen. In this way, any excess in situ
5384 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8

generated oxygen could be measured using Gas Chromatog-
raphy after every experiment.
3. Results and discussion
3.1. Influence of electrode material on sulfide oxidation
The influence of the electrode material on the kinetics of
anodic sulfide oxidation was investigated during 6-h experi-
ments at a fixed current density of 10 mA/cm2. During the
experimental runs, sulfide concentrations increased to
approximately 7e10 mg/L due to the recirculatory mode of
operation. The typical profiles of the sulfide concentration are
presented in the supplementary information S1. In Table 1 an
overview of the results using Ta/Ir, Ru/Ir, Pt/Ir, PbO2 and SnO2
as electrode material is presented. The obtained removal and
Coulombic efficiencies using Ta/Ir, Ru/Ir, Pt/Ir, PbO2 and SnO2
were 78
5, 77
1, 85
4, 83
2 and 84
1%, respectively.
This is equal to sulfide removal rates of 10.8
0.2, 11.7
1.1,
12.4
0.4, 12.4
0.4 and 12.9
0.8 g S m2 1
electrode surface h . The
obtained Coulombic efficiencies were calculated based on the
oxidation of sulfide to sulfate (see Equation (2)).
The values observed for the sulfide removal rates
expressed in mg S/L wastewater h1 7.5
0.3 (Ta/Ir) to 7.8
0.6
(SnO2) are in agreement with the chemical sulfide oxidation
rates found under high dissolved oxygen concentrations in
domestic wastewater (Sharma and Yuan, 2010). Furthermore,
GC analysis of the produced gas at the end of every experi-
ment showed that in all experiments similar amount of excess
oxygen was generated and transferred to the headspace. The
obtained excess oxygen generation using Ta/Ir, Ru/Ir, Pt/Ir,
PbO2 and SnO2 was 65
13, 60
17, 69
8, 66
17 and
47
8 mg (Table 1). In a previous study, we showed that direct
sulfide oxidation at the electrode surface was negligible under
the given operational conditions (Pikaar et al., 2011). If reactive
oxygen species were primarily responsible for sulfide oxida-
tion, we would have expected sulfate being the primary
product of oxidation, contrarily to the results (see Table 1). If
oxygen would be the predominant reaction mechanism,
a mixture of sulfur species would be expected. The results
suggest that in situ oxygen generation is significant and could
be primarily responsible for the sulfide oxidation observed.
The materials used are known to have different over-
potentials for oxygen evolution. This however does not
necessarily mean that they will produce different amounts of
oxygen. The formation of reactive oxygen species such as OH
radicals is intermediate products during the oxidation of
water to oxygen. In the first step, adsorbed OH radicals are
formed on the electrode surface. In the second step, the
adsorbed OH interacts with the oxygen already present in the
oxide anode to form physisorbed or chemisorbed ‘active
oxygen’. In absence of any oxidizable pollutant this ‘active
oxygen’ subsequently produces oxygen (Comninellis, 1994).
Thus, the efficiency of the formation of reactive oxygen
species does not only depend on the electrode material but
also on the concentration of the pollutant and its reactivity
toward oxidation. Especially at low pollutant concentrations,
low Coulombic efficiencies for the formation of reactive
oxygen species have been observed (Martinez-Huitle and
Brillas, 2009), which means that high Coulombic efficiencies
for oxygen evolution can be expected. Considering the low
sulfide (i.e. w10 mg/L) and organics concentrations (i.e.
380
140 mg/L) during the experiments, this could explain the
similar sulfide removal and excess oxygen production rates
for all electrode materials.
It should be noted that in this study a laboratory reactor
was used rather than a real rising main system. Hence, our

Table 1 e Sulfide oxidation (n [ 3) from domestic wastewater using MMO coated titanium electrodes at a current density of
10 mA/cm2.
Parameter Unit Ta/Ir Ru/Ir Pt/Ir SnO2 PbO2
a
Coulombic efficiency % 78
5 77
1 85
4 84
1 83
2
Removal efficiency % 78
5 77
1 85
4 84
1 83
2
Removal rate g S m2
electrode surface h
1
10.8
0.2 11.7
1.1 12.4
0.4 12.9
0.8 12.4
0.1
Removal rate mg S L1 h1 7.5
0.3 7.7
0.5 7.7
0.2 7.8
0.6 7.8
0.7
Total S added (mg) mg 199
9 219
21 210
18 220
12 214
6
Final sulfide conc. mg/L 7.9
2.6 10.0
1.2 6.2
2.3 6.6
0.6 6.9
1.8
S0 produced mg 100
9 97
4 108
19 136
25 101
36
S2O2
3 produced mg 63
9 64
13 52
6 43
10 50
8
SO2
3 produced mg 0.9
0.2 1.5
0.5 1.1
0.7 0.8
0.5 1.1
0.7
SO2
4 produced mg 10
5 14
15 34
16 28
8 30
17
COD removed mg 199 256 204 274 117
COD removal rate mg COD h1 43 33 34 46 26
O2 produced mg 60
17 65
13 69
8 47
8 66
17

Temperature C 24.5
0.5 24.1
0.4 24.6
0.4 25.0
0.6 24.3
0.4
Conductivity mS/cm 1.11
0.01 1.15
0.01 1.17
0.06 1.16
0.04 1.09
0.09
Chloride concentration mg/L 114
9 117
9 115
9 109
10 114
6
pHb e 7.5 7.5 7.5 7.5 7.5
Average anode potential V 1.41
0.08 1.42
0.06 1.53
0.12 2.2
0.11 1.8
0.04
Average cell voltage V 5.3
0.4 5.2
0.4 5.0
0.2 9.0
0.4 5.0
0.2

a Coulombic efficiency is based on the oxidation of sulfide to sulfate.

b pH is controlled online.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8
system differed from a real sewer system as such that in our
system (a) a headspace was present and (b) in situ generated
oxygen was dissolved into only 5 L. Hence, we produced much
more oxygen per unit wastewater. The above caused that
oxygen was transferred to the headspace instead of remaining
in the water phase.
The anode potentials remained constant during the course
of all the experiments. The anode potentials for Ta/Ir, Ru/Ir
and Pt/Ir (low overpotential for oxygen evolution) were
1.41
0.08, 1.42
0.06 and 1.53
0.12 V vs. SHE, whereas the
anode potentials for PbO2 and SnO2 were 1.80
0.04 and
2.22
0.11 V vs. SHE, respectively. The overall cell voltage for
all electrode materials, except for SnO2 (9.0
0.4 V), was
approximately 5 V (Table 1). As yet, we have no explanation
for the unexpectedly high cell voltage for SnO2, the use of this
electrode apparently caused a higher ohmic resistance or
higher cathodic overpotential.
The mechanism of the generation of oxygen active species
is very complex and can involve the generation of several
radical species containing oxygen and/or halogen atom (e.g.
OC C C
2 , HO2 , HClO ) other than hydroxyl radicals which can
occur at lower potentials. Hence, reactive oxygen species
other than hydroxyl radicals might have been formed but did
not have a significant impact on the sulfide removal process.
Fig. 2 shows the electron distribution among the different
electron sinks (i.e. sulfur, thiosulfate, sulfite, sulfate, excess
oxygen and COD (i.e. organics)) during the oxidation of sulfide
using the different electrode materials. A significant part of
the charge supplied to the system was used for the oxidation
of organics. The charge used for the oxidation of organics was
46%, 60%, 47%, 64% and 36% for Ta/Ir, Ru/Ir, Pt/Ir, PbO2 and
SnO2, respectively.
Concentrations of the different dissolved sulfur species (i.e.
sulfide, sulfite, thiosulfate and sulfate) indicate that 43
2%,
100
electron distribution (%)
80
60
40
20
0
Ta/Ir
Fig. 2 e Electron distribution (%) among the different electron sinks (i.e. sulfur, thiosulfate, sulfite, sulfate, excess oxygen
and organics) during the oxidation of sulfide using Ta/Ir, Ru/Ir, Pt/Ir, PbO2 and SnO2 electrodes at a current density of 10 mA/
cm2 (n [ 1).
5385
39
7%, 47
8%, 42
11% and 57
10% of the total sulfide
added was oxidized to elemental sulfur using Ta/Ir, Ru/Ir, Pt/
Ir, PbO2 and SnO2, respectively. This is further supported by
the reasonably small gap in the electron balances of the
experiments (see Fig. 2). No elemental sulfur was visually
observed on the electrode surface. This suggests that the
elemental sulfur was formed in the bulk by means of indirect
oxidation with oxygen.
In summary, the results highlight that under the applied
operational conditions the sulfide removal process in terms of
removal efficiency, excess oxygen generation, overall cell
potential (except SnO2) and electron distribution is not
significantly influenced by the electrode material used.
3.2. Influence of chloride concentration on sulfide
oxidation
The impact of elevated chloride concentration was investi-
gated using Ta/Ir and Ru/Ir coated titanium electrodes; elec-
trodes with a low and high reported catalytic activity toward
chlorine generation, respectively. Table 2 shows that similar
sulfide removal efficiencies and removal rates as in the
experiments at low chloride concentrations were obtained
(i.e. 7.7
0.5 versus 7.0
1.0 and 7.5
0.3 versus 7.7
0.3 mg
S/L h). Fig. 3 shows the electron distribution among the
different electron sinks (i.e. sulfur, thiosulfate, sulfite, sulfate,
excess oxygen and organics) during the oxidation of sulfide
using of Ta/Ir and Ru/Ir electrodes, respectively. Similar to the
experiments at low chloride concentrations a large part of the
charge supplied to the system was used for the oxidation of
organics (i.e., 50
10, 50
8% for Ta/Ir and Ru/Ir, respectively).
Ru/Ir coated titanium electrodes are well-known for their
low overpotential for chlorine generation. Hence, at high
chloride concentrations in situ chlorine generation may have
organics
excess oxygen
sulfate
sulfite
thiosulfate
sulfur
Ru/Ir Pt/Ir SnO2 PbO2
5386 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8

an impact on the sulfide oxidation process and might also 70

affect the removal of organics. The maximum chlorine

generation under the mass transfer limiting conditions at 60

a planar electrode (under optimum mixing conditions) can be

elecetron distribution (%)

50
described according to:
h i

JL ¼ nFkm Cl (3) 40

where JL is the mass transfer limited current density (mA/ 30

cm2), n is the number of electrons involved (i.e. 1 for the

oxidation of chloride to chlorine), F the Faraday constant 20

(96,485.3 C/mol), km the mass transport rate coefficient (m/s)

and Cl the chloride concentration (mg/L). Assuming a km of 10

0.8  105 m/s (Szpyrkowicz et al., 2005) and an average chlo-

0
ride concentration of 114 mg/L (i.e. the chloride concentration sulfur thiosulfate sulfite sulfate oxygen COD
present in the sewage wastewater used) the mass transfer
limited current is only 0.25 mA/cm2. In addition, a migration Fig. 3 e Electron distribution (in %) among the different
current due to the migration in the chloride transport can be electron sinks (i.e. sulfur, thiosulfate, sulfite, sulfate,
expected as result of the low conductivity of the wastewater excess oxygen and COD (i.e. organics)) during the
(Bergmann and Koparal, 2005). However, at the applied oxidation of sulfide using Ta/Ir and Ru/Ir, electrodes at
current density (10 mA/cm2), the chloride concentration and a current density of 10 mA/cm2 at high chloride
the applied electrical field the relative importance of the concentrations (n [ 3).
migration current density is small. The applied current
density in all experiments is 10 mA/cm2 and thus sulfide
oxidation by means of in situ chlorine production is expected
to be small. Hence, it is expected that at low chloride
chlorine generation can play a significant role in the anodic
concentrations Ru/Ir coated titanium electrodes mainly will
sulfide removal process. Depending on the pH, sulfide can be
result in the in situ generation of oxygen and hence similar
oxidized by chlorine either to elemental sulfur or sulfate. At
kinetics for the anodic sulfide oxidation are expected.
pH values  7.5, sulfide is oxidized by chlorine to elemental
However, at high chloride concentrations up to 1100 mg/L,
sulfur (Chwirka and Satchell, 1990). Thus, under the condi-
which are often observed in sewers in many coastal areas,
tions normally observed in sewer systems (i.e. pH  7.5) the
addition of chlorine would result in the oxidation of sulfide
(HS) to elemental sulfur. The amount of sulfide dosed in the
experiments was equivalent to 2.5 mA/cm2 (when oxidized to
Table 2 e Sulfide oxidation (n [ 3) from domestic
sulfur), whereas the mass transport limited current for the
wastewater at elevated chloride concentrations using Ta/
Ir and Ru/Ir coated titanium electrodes at a current generation of chlorine at chloride concentrations of 1100 mg/L
density of 10 mA/cm2. is approximately 2.4 mA/cm2.
Thus, under elevated chloride levels sufficient chlorine
Parameter Unit Ta/Ir Ru/Ir
could have been produced to oxidize the sulfide added while
Coulombic % 87
4 86
5 this might not have been the case at low chloride concentra-
efficiencya
tions. The results showed that at elevated chloride concen-
Removal rate g S m2
electrode surface h
1
12.9
1.0 12.2
1.2
trations similar sulfide and organic removal rates were
Removal rate mg S L1 h1 7.7
0.3 7.0
1.0
Total S added (mg) mg 213
8 203
8 obtained (Table 2). Moreover, similar electron distribution
Final sulfide conc. mg/L 5.4
1.3 5.7
1.8 among the electron sinks was observed. Hence, it appears that
S0 produced mg 139
0 103
20 elevated chloride concentrations did not entail any significant
S2O2
3 produced mg 29
14 33
5 differences in sulfide removal rate as well as in the organic
SO2
3 produced mg 0.1
0.1 0.3
0.1 removal. Analysis of the free chlorine concentration revealed
SO2
4 produced mg 20
12 39
14
that in all experiments no free chlorine or other reactive
COD removed mg 213
16 216
33
oxygen species (which are also detected with the used
COD removal rate mg COD h1 36
6 37
7
O2 produced mg 77
19 65
10 method) were present. It cannot be excluded that chlorine was
Temperature 
C 24.0
0.5 24
0.4 formed and instantly reacted with the sulfide and/or organics
Conductivity mS/cm 3.73
0.05 3.73
0.05 present. This would be in line with several studies earlier
Chloride mg/L 1117
37 1119
25 reported (Chen, 2004; Bergmann and Koparal, 2005).
concentration
pH e 7.5 7.5
3.3. Implications for practice
Average anode V 1.51
0.18a 1.43
0.03
potential
Average cell V 4.21
0.9a 4.64
0.2 The results could indicate that sulfide removal by in situ
voltage generated oxygen was the predominant reaction mechanism,
independent of the electrode material used. Other reaction
a n ¼ 2.
mechanisms including the formation of radical species
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8
containing oxygen and/or halogen atom (e.g. OC C
2 , HO2 ,
HClOC) cannot be excluded but did not have a significant
impact on the sulfide removal process. In perfectly mixed
conditions, such as in a rotating disc setup, the different
mechanisms for the electrodes would be observed to a higher
extent. However, in the reactors here, which are more
amenable to wastewater treatment, limitations exist on the
mixing intensity.
Oxygen injection is presently considered as an attractive
option for sulfide abatement in sewer systems. It is less
expensive than most other chemicals, and can target rising
mains where the SRB activity is the highest (Hvitved-Jacobsen,
2001). However, transport and storage of pure oxygen carries
serious safety issues and precise control of dosing is not
straightforward. By generating oxygen in situ the requirement
for transport and storage are avoided, thus mitigating safety
concerns. Other advantages of in situ oxygen generation
compared to traditional methods for oxygen supply are the
fine dispersion, high controllability and the ease to monitor.
The disadvantage is the cost of the oxygen per unit weight.
All electrode materials performed similarly in terms of
sulfide removal. Therefore, Ta/Ir, Ru/Ir and Pt/Ir coated tita-
nium electrodes seem the most suitable electrodes since they
posses the lowest overpotential for oxygen evolution and they
are already used in full scale applications. However, the life
time of Ru/Ir electrodes for the oxygen evolution reaction,
which is the predominant reaction at low chloride concen-
trations, is low (Hine et al., 1979). Hence, Ta/Ir and Pt/Ir coated
titanium electrodes appear the most suitable electrodes for
sulfide oxidation from domestic wastewater in sewer systems.
Based on a cell potential of 5 V, a Coulombic efficiency of
95% (for oxygen generation) and a cost of $0.06 per kWh, the
estimated delivery cost is $1.06 per kg, relative to a delivery
cost of $0.54e0.82 per kg for standard oxygen purchases (de
Haas et al., 2008). However, standard oxygen injection in
sewer systems has low efficiency (i.e. w20 to 40%) (de Haas
et al., 2008) since oxygen is often dosed in an inefficient way
(i.e., coarse bubbles) which results in a significant loss of
undissolved gas from air in gas release valves downstream.
The latter is avoided when oxygen is generated in situ due to
the high transfer efficiency and fine dispersion of in situ
generated oxygen. The deployment of a sulfide or dissolved
oxygen sensor before or after the electrochemical system
respectively would also allow further fine-tuning of the
oxygen dosing. While entailing a higher cost, sustainable
electricity solutions such as photovoltaic power would allow
total independence of the dosing system of transport or utility
requirements. An in-depth life cycle analysis would provide
more insight into the sustainability of the existing dosing
methods.
In all experiments excess oxygen was produced, which in
a practical situation would be used for chemical oxidation or
by the biofilm in the sewer pipes for biological sulfide oxida-
tion or the removal of organic matter. Nonetheless, in this
study, the impact of electrode material and chloride concen-
tration on the kinetics of sulfide oxidation from real domestic
wastewater was successfully investigated.
Over the course of the experiments, none of the electrode
materials deteriorated or demonstrated changing potential
over time. Earlier studies at lower current densities indicated
5387
electrode fouling with sulfur (Dutta et al., 2009; Ateya et al.,
2003; Rabaey et al., 2006), it appears that at the higher
current densities used here sulfur either flakes off or is further
oxidized to its soluble forms. However, in realistic conditions
of the sewer, other forms of fouling such as ragging, particle
settling and scaling may occur. Long term trials on site are
presently underway to investigate these possible impacts,
which were not observed here.
Scaling of the membrane and the electrode surface, on the
other hand, was observed in the cathode chamber. This is
caused by transport of bivalent cations such as calcium
through the cation exchange membrane causing precipitation
of inorganics such as calcium hydroxide or calcium carbonate
(Jeremiasse et al., 2010). To overcome problems with scaling
the cathode needs to be cleaned periodically. This can be done
either by chemical cleaning (i.e. the addition of citric or
hydrochloric acid) or by periodic switching of the polarity of the
electrodes (anode becomes cathode/cathode becomes anode),
the latter option is attractive from an operational perspective
but does restrict the types of electrodes that can be used.
4. Conclusions
In this study, we investigated the kinetics of sulfide oxidation
from real domestic wastewater using five different types of
mixed metal-coated titanium electrodes (Ta/Ir, Ru/Ir, Pt/Ir,
SnO2 and PbO2) at a current density of 10 mA/cm2. The
obtained sulfide removal efficiencies were not significantly
influenced by the electrode material used. The results indicate
that, independent of electrode material used, sulfide was
removed by means of chemical oxidation with in situ gener-
ated oxygen. Ta/Ir and Pt/Ir appear the most suitable elec-
trodes since they have a low overpotential for oxygen
evolution and are known for their stability even at low chlo-
ride concentrations. The obtained sulfide removal rates were
in the same order of chemical rates found under high oxygen
concentrations whereas in all experiments excess oxygen
was produced. Elevated chloride concentrations did not entail
any significant difference in sulfide removal rate. Analysis
revealed that no free chlorine was present even at high chlo-
ride concentrations.
Acknowledgments
Ilje Pikaar, René Rozendal and Korneel Rabaey thank the
University of Queensland for scholarship (University of
Queensland Research Scholarship) and fellowship support (RR
UQ Postdoctoral Fellowship). KR is supported by the Australian
Research Council (APD, DP0879245). This work was funded by
the Australian Research Council (ARC Linkage project:
LP0882016 “Optimal Management of Corrosion and Odor Prob-
lems in Sewer Systems”). The authors also want to acknowl-
edge Dr. Beatrice Keller-Lehmann and Ms. Susan Cooke for
their helpful collaboration with the chemical analyses.
5388 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 1 e5 3 8 8
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.07.033.
references
Ateya, B.G., AlKharafi, F.M., Al-Azab, A.S., 2003. Electrodeposition
of sulfur from sulfide contaminated brines. Electrochemical
and Solid-State Letters 6 (9), C137eC140.
Bergmann, M.E.H., Koparal, A.S., 2005. Studies on electrochemical
disinfectant production using anodes containing RuO2.
Journal of Applied Electrochemistry 35 (12), 1321e1329.
Chen, G., 2004. Electrochemical technologies in wastewater
treatment. Separation and Purification Technology 38 (1), 11e41.
Chwirka, J.D., Satchell, T.T., 1990. A 1990 guide for treating
hydrogen sulfide in sewers. Water Engineering & Management
137, 32e35.
Comninellis, C., 1994. Electrocatalysis in the electrochemical
conversion/combustion of organic pollutants for waste water
treatment. Electrochimica Acta 39, 1857.
de Haas, D.W., Corrie, S., O’Halloran, K., Keller, J., Yuan, Z., 2008.
Odour control by chemical dosing: a review. Journal of the
Australian Water Association 35 (02), 138e143.
Dutta, P.K., Rabaey, K., Yuan, Z., Keller, J., 2008. Spontaneous
electrochemical removal of aqueous sulfide. Water Research
42 (20), 4965e4975.
Dutta, P.K., Rozendal, R.A., Yuan, Z., Rabaey, K., Keller, J., 2009.
Electrochemical regeneration of sulfur loaded electrodes.
Electrochemistry Communications 11 (7), 1437e1440.
Feng, Y.J., Li, X.Y., 2003. Electro-catalytic oxidation of phenol on
several metal-oxide electrodes in aqueous solution. Water
Research 37 (10), 2399e2407.
Hine, F., Yasuda, M., Noda, T., Yoshida, T., Okuda, J., 1979.
Electrochemical behavior of the oxide-coated metal anodes.
Journal of the Electrochemical Society 126 (9), 1439e1445.
Hvitved-Jacobsen, T., 2001. Sewer processes: microbial and
chemical process engineering of sewer networks. xi þ 237 pp.
Jeremiasse, A.W., Hamelers, H.V.M., Buisman, C.J.N., 2010.
Microbial electrolysis cell with a microbial biocathode.
Bioelectrochemistry 78 (1), 39e43.
Kaempfer, W., Berndt, M., 1998. Polymer modified mortar with
high resistance to acid to corrosion by biogenic sulfuric acid.
In: Proceedings of the IX ICPIC Congress, Bologna, Italy, pp.
681e687.
Keller-Lehmann, B., Corrie, S., Ravn, R., Yuan, Z., Keller, J., 2006.
Preservation and simultaneous analysis of relevant soluble
sulfur species in sewage samples. In: 2nd International IWA
Conference on Sewer Operation and Maintenance, Vienna,
Austria.
Martinez-Huitle, C.A., Brillas, E., 2009. Decontamination of
wastewaters containing synthetic organic dyes by
electrochemical methods: a general review. Applied Catalysis
B: Environmental 87 (3e4), 105e145.
Panizza, M., Cerisola, G., 2008. Electrochemical degradation of
methyl red using BDD and PbO2 anodes. Industrial and
Engineering Chemistry Research 47 (18), 6816e6820.
Panizza, M., Kapalka, A., Comninellis, C., 2008. Oxidation of
organic pollutants on BDD anodes using modulated current
electrolysis. Electrochimica Acta 53 (5), 2289e2295.
Pikaar, I., Rozendal, R.A., Yuan, Z., Keller, J., Rabaey, K., 2011.
Electrochemical sulfide removal from synthetic and real
domestic wastewater at high current densities. Water
Research 45 (6), 2281e2289.
Rabaey, K., Van de Sompel, K., Maignien, L., Boon, N.,
Aelterman, P., Clauwaert, P., De Schamphelaire, L., Pham, H.T.
, Vermeulen, J., Verhaege, M., Lens, P., Verstraete, W., 2006.
Microbial fuel cells for sulfide removal. Environmental Science
& Technology 40 (17), 5218e5224.
Sharma, K., Yuan, Z., 2010. Kinetics of chemical sulfide oxidation
under high dissolved oxygen levels. Submitted for Oral
Presentation, 6th International Conference on Sewer
Processes and Networks, 7e10 November 2010.
Sun, Y.-X., Wu, Q.-Y., Hu, H.-Y., Tian, J., 2009. Effects of operating
conditions on THMs and HAAs formation during wastewater
chlorination. Journal of Hazardous Materials 168 (2e3),
1290e1295.
Sydney, R., Esfandi, E., Surapaneni, S., 1996. Control concrete
sewer corrosion via the crown spray process. Water
Environment Research 68, 338e347.
Szpyrkowicz, L., Kaul, S.N., Neti, R.N., Satyanarayan, S., 2005.
Influence of anode material on electrochemical oxidation for
the treatment of tannery wastewater. Water Research 39 (8),
1601e1613.
Takasu, Y., Sugimoto, W., Nishiki, Y., Nakamatsu, S., 2010.
Structural analyses of RuO2eTiO2/Ti and IrO2eRuO2eTiO2/Ti
anodes used in industrial chlor-alkali membrane processes.
Journal of Applied Electrochemistry, 1e7.
Taylor, B., Gardner, T., 2007. Southeast Queensland recycled
water aspects and soil impacts, Sunshine Coast,
Australia.
Vincke, E., Wanseele, E.V., Monteny, J., Beeldens, A., Belie, N.D.,
Taerwe, L., Gemert, D.V., Verstraete, W., 2002. Influence of
polymer addition on biogenic sulfuric acid attack of concrete.
International Biodeterioration & Biodegradation 49 (4),
283e292.
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 Research 42 (1e2), 1e12.
Zhu, X., Tong, M., Shi, S., Zhao, H., Ni, J., 2008. Essential
explanation of the strong mineralization performance of
boron-doped diamond electrodes. Environmental Science and
Technology 42 (13), 4914e4920.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Biological iron oxidation by Gallionella spp. in drinking water

production under fully aerated conditions

W.W.J.M. de Vet a,c,d,*, I.J.T. Dinkla b,1, L.C. Rietveld d, M.C.M. van Loosdrecht c,a
a
Oasen Drinking Water Company, PO Box 122, 2800 AC Gouda, The Netherlands
b
Bioclear bv. Rozenburglaan 13, 9727 DL Groningen, The Netherlands
c
Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands
d
Department of Water Management, Delft University of Technology, Stevinweg 1, 2628 CN Delft, The Netherlands

article info abstract

Article history: Iron oxidation under neutral conditions (pH 6.5e8) may be a homo- or heterogeneous
Received 7 December 2010 chemically- or a biologically-mediated process. The chemical oxidation is supposed to
Received in revised form outpace the biological process under slightly alkaline conditions (pH 7e8). The iron
8 July 2011 oxidation kinetics and growth of Gallionella spp. e obligatory chemolithotrophic iron
Accepted 25 July 2011 oxidizers e were assessed in natural, organic carbon-containing water, in continuous lab-
Available online 29 July 2011 scale reactors and full-scale groundwater trickling filters in the Netherlands. From Gallio-
nella cell numbers determined by qPCR, balances were made for all systems. The homo-
Keywords: geneous chemical iron oxidation occurred in accordance with the literature, but was
qPCR retarded by a low water temperature (13
C). The contribution of the heterogeneous
Gallionella spp. chemical oxidation was, despite the presence of freshly formed iron oxyhydroxides, much
Groundwater trickling filtration lower than in previous studies in ultrapure water. This could be caused by the adsorption
Biological and chemical of natural organic matter (NOM) on the iron oxide surfaces. In the oxygen-saturated
iron oxidation natural water with a pH ranging from 6.5 to 7.7, Gallionella spp. grew uninhibited and
biological iron oxidation was an important, and probably the dominant, process. Gallionella
growth was not even inhibited in a full-scale filter after plate aeration. From this we
conclude that Gallionella spp. can grow under neutral pH and fully aerated conditions when
the chemical iron oxidation is retarded by low water temperature and inhibition of the
autocatalytic iron oxidation.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
The existence and relevance of iron-oxidizing bacteria (FeOB)
in drinking water treatment has been well established from
the very beginning of central water supply. Berger and Berger
(1928) mentioned that only five years after start-up, all Berlin
Water Works were forced to switch from groundwater to
surface water in 1882 due to a so-called ‘Eisenkalamität’ (iron
calamity). Biological essays demonstrated that Crenothrix
polyspora and probably also Leptothrix ochracea caused

Abbreviations: FeOB, iron-oxidizing bacteria; NOM, natural organic matter; qPCR, (quantitative) real-time polymerase chain reaction;
WTP, water treatment plant.
* Corresponding author. Oasen Drinking Water Company, PO Box 122, 2800 AC Gouda, The Netherlands. Tel.: þ31 610927947; fax: þ31
152782355.
E-mail addresses: W.W.J.M.deVet@tudelft.nl (W.W.J.M. de Vet), dinkla@bioclear.nl (I.J.T. Dinkla), L.C.Rietveld@tudelft.nl (L.C. Riet-
veld), M.C.M.vanLoosdrecht@tudelft.nl (M.C.M. van Loosdrecht).
1
Tel.: þ31 505718455; fax: þ31 505717920.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.07.028
5390 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
pollution of the unfiltered, distributed water with ‘ocher-
yellow, dirty brownish up to coffee brown flocky deposits’
(Ibid.). Similar problems with iron-containing groundwater
occurred in Rotterdam, The Netherlands (de Vries, 1890).
Iron-containing groundwater could only be used for drinking
water production once properly working de-ironing filters
were developed.
Since the establishment of de-ironing filters, there has
been an ongoing discussion with regard to the importance of
chemical versus biological iron oxidation (Czekalla et al., 1985;
Sharma et al., 2005, and references therein). In drinking water
filters in northern Germany, at least four species of FeOB have
been reported (Gallionella sp., L. ochracea, Toxothrix trichogenes
and an unknown bacterium), next to five species of
manganese-oxidizing bacteria. From these observations it
was concluded that iron and manganese removal was
a bacterial process (Czekalla et al., 1985). Søgaard et al. (2000)
studied precipitates from backwash sludge from three water
treatment plants (WTP) in Denmark. They suggested low
oxygen content of the raw water, poor aeration and relatively
low pH as the determining prerequisites for biological iron
oxidation; however, they did not provide consistent data from
the WTPs to substantiate these presumptions. The presence
of ferrous iron in combination with low dissolved oxygen and/
or slightly acidic pH is also regarded by other researchers as
prerequisites for growth of FeOB (Hallbeck and Pedersen, 1990;
Emerson and Floyd, 2005).
The distinction between heterogeneous chemical and
biological iron oxidation is, however, hard to make (Sharma
et al., 2005; Tekerlekopoulou and Vayenas, 2008). Only in
some cases, the distinguishable characteristic forms of iron
deposits e like the twisted stalks formed by Gallionella spp. e
indicate biological action, but in many other cases, particulate
amorphous iron oxyhydroxides, very similar to chemical
precipitates, are shown to be of biological origin as well
(Emerson and Weiss, 2004).
In recent studies the catalysis of iron oxidation by excreted
RedOx-enzymes like flavins (Degrémont, 2007) or exopol-
ymers (Søgaard et al., 2000) has been reported, however this
chemical process does not yield energy for bacterial growth.
The chemo-lithotrophy of some FeOB is still under dispute
(Spring and Kämpfer, 2005). Gallionella spp. are, however,
generally regarded as strictly lithotrophic, unable to catabo-
lize organic matter (Lütters-Czekalla, 1990), so the growth of
Gallionella spp. can be seen as direct proof of biological iron
oxidation. For this reason, this paper focuses on Gallionella
spp., even though other FeOB such as Leptothrix spp. were
found to be growing in the studied systems as well (data not
shown).
New molecular techniques provide powerful tools to
assess and quantify the role of FeOB in full-scale treatment
systems. For this paper, the kinetics of iron oxidation and the
growth of the iron-oxidizing Gallionella bacteria were assessed
in continuous lab- and full-scale reactors and trickling filters.
The results of these studies were used to discuss the compe-
tition of biological iron oxidation with chemical iron oxidation
at different pH’s in groundwater filtration. We hypothesize
that Gallionella spp. can also grow under fully aerated and
slightly alkaline pH conditions when chemical iron oxidation
is retarded.
2. Methods and materials
2.1. Lab-scale experiments
The oxidation and removal of iron were investigated in two
lab-scale setups at WTP Lekkerkerk of the Oasen drinking
water company in The Netherlands. The lab-scale research
consisted of oxidation column and filtration column experi-
ments, which are described separately in the next two
sections. Both experimental setups were fed with drinking
water locally produced from riverbank groundwater. This
water is moderately hard (Ca2þ w2 mM), well buffered (HCO 3
w3.0 mM), has a constant temperature of 13
C, a pH of
7.8  0.1. The dissolved oxygen content of the feed water was
9.9  0.6 mg L1 (74 data points in the period April
2008eAugust 2009 by potentiometric measurement in accor-
dance with NEN-ISO 5814, NEN, 1993). Ferrous iron (FeS-
O47H20, Merck 103965 5000) was added to the feed water of
all but the reference filter columns in a concentration of
3.3 mg L1 Fe, resembling the groundwater quality at WTP
Lekkerkerk. A nutrient solution, containing phosphorus (0.6 mM
PO4-P), nitrogen (3.8 mM NH4-N) and trace elements (Zn, Co, Cu
and Mo), was added to prevent bacterial growth limitation. All
water and chemical flows were controlled by tube pumps and
all flow rates were checked weekly by mass measurements.
All columns had an internal diameter of 0.089 m resulting in
a water velocity of about 2.2 m h1, similar to the full-scale
filters. While the desired flow rate was 14.0 L h1, the real-
ized flow rates for the oxidation and filter columns were
13.9  0.4 and 13.5  0.8 L h1, respectively. The flow direction
was upwards for the oxidation columns and downwards for
the filter columns. Additional information on the columns’
setup, including schemes and pictures, is given in the
Supplementary Material A.
2.1.1. Oxidation columns’ setup
The chemical iron oxidation strongly depends on pH (Sung,
1980; Tamura et al., 1976), it is therefore supposed that, with
a decreasing pH, biological oxidation might outcompete
chemical oxidation processes. In order to determine which
rates of chemical oxidation still allow simultaneous biological
oxidation by Gallionella spp., different pH conditions allowing
different rates of chemical oxidation were applied. The influ-
ence of feed water pH on the oxidation rate of the ferrous iron
and the growth of Gallionella spp. in the natural water of WTP
Lekkerkerk was studied in six oxidation columns. With an
overflow level of 0.58 m, the residence time calculated from
mass balances was 16 min. Determination of the residence time
by NaCl spiking and corresponding conductivity measure-
ments (not presented) showed no short-circuiting, indicating
mainly plug flow conditions in the oxidation columns.
To set the pH, HCl or NaOH was added to the column
influent. The required doses of HCl and NaOH were deter-
mined in triplicate by titration of the feeding drinking water
and checked by offline potentiometric pH measurement in
conformity with the Dutch NEN-ISO 10523 protocol (NEN,
2009) of the columns’ influent, also in triplicate (Figure B.1 of
Supplementary Material B). Although the titrations and
control measurements were executed at 20
C, operational pH
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
values at 13
C will not have differed much because of the
good buffering of the feed water. The deviation of the control
measurements at the more extreme pH values was probably
caused by gas exchange during sampling and offline
measurements and calcium carbonate precipitation during
storage. As the values determined by the titrations best rep-
resented the actual system during the experiments, these pH
values will be used in the results section, with the uncertainty
range calculated from mass measurements of the acid and
base dosing. The realized iron dose determined by mass
balances and total iron analysis (see Section ‘Iron analyses’)
was 3.3  0.7 mg L1 Fe and is shown per column in Figure B.2
of the Supplementary Material B. In total, 58  4 g Fe was
dosed over 7 weeks.
The oxidation columns were run for seven weeks from July
8 to August 26, 2009. The concentrations of ferrous and ferric
iron in the column effluents were determined weekly. The
Gallionella spp. cell numbers were determined after three and
seven weeks in the column influents and effluents and after
seven weeks in the accumulated sludge in the oxidation
columns. A picture of the oxidation columns’ setup at the end
of the experiment is given as Figure A.2 in the Supplementary
Material A.
2.1.2. Filter columns’ setup
To model groundwater trickling filtration, iron removal was
studied in a lab-scale filter columns’ setup, consisting of seven
trickling filter columns in duplicate, in total 14 columns. All
columns were filled with standard filter sand (1.7e2.5 mm). The
pH of the feed water was lowered by HCl to resemble the
groundwater before filtration (7.25  0.15). One column was
used as a reference with no iron removal. In three columns,
ferrous iron was dosed just before the filter top; in three other
columns, ferrous iron passed a pre-oxidation column before
filtration. These columns are referred to as “without pre-
oxidation” and “with pre-oxidation", respectively. All trickling
filter columns had forced ventilation which raised the pH in the
filter effluents to 7.67  0.07. Each filter column was automati-
cally backwashed every 24 h with a fixed volume of 30 L drinking
water and under expansion (fluidization) of the filter bed.
The filtration columns were run for six months from April 2
to October 8, 2008. The Gallionella spp. cell numbers were
determined after six months in the influents and effluents, the
backwash water and the filter material of seven columns (one
of each duplicate).
2.2. Full-scale groundwater trickling filters
The growth of Gallionella spp. and their role in iron oxidation
was verified in three full-scale trickling filters at two Oasen
WTPs. All three full-scale filters treated moderately hard and
well-buffered anoxic groundwater. The filters were back-
washed automatically after a filter runtime of 48 h, to prevent
clogging by removal of inorganic precipitates and excess
biomass. At WTP Lekkerkerk, the filter material of a trickling
filter was externally washed and the filter performance and
growth of Gallionella spp. were monitored extensively for nine
months after restart of the filters from December 12, 2007 to
September 19, 2008. At WTP De Hooge Boom, the growth of
Gallionella spp. in two groundwater trickling filters was
5391
assessed in a quick scan on March 1, 2010. In one of these
filters, the anoxic groundwater was sprayed directly on the
trickling filter, while in the other the groundwater was
intensively aerated on a plate aerator prior to spraying on top
of the trickling filter. Both plate aeration and trickling filtration
raised the dissolved oxygen content to nearly a saturated level
and the pH by the stripping of carbon dioxide (see Table 1). All
iron present in the groundwater was virtually completely
removed in the filters (>95%). Table 1 gives an overview of the
groundwater and filtrate qualities as well as the characteris-
tics of the three studied filters.
2.3. DNA extraction
Samples for detection and identification of Gallionella spp.
were taken in sterilized glass bottles at different points in the
Oasen WTP Lekkerkerk. All samples were stored at 4
C. Gal-
lionella spp. cell numbers were determined by qPCR.
Groundwater, influent and effluent water and backwash
water samples were filtered over 0.2 mm polycarbonate
membranes to concentrate the cells prior to DNA extraction.
DNA was extracted from a volume of 100e150 ml water per
sample. The filter was subsequently subjected to DNA
extraction by bead beating in a Fast DNA spin kit for soil (MP
Biomedicals, Soton, Ohio, United States). The DNA was puri-
fied using a silica-based column and eluted in 100 ml TE. DNA
from approximately 10 g of filter sand was extracted as
described by de Vet et al. (2009). In all cases, an internal
control was used to determine the extraction efficiency.
2.4. Quantification of Gallionella spp.
In order to quantify the number of Gallionella spp. cells in the
systems, a specific PCR was developed to detect these bacteria,
including the Gallionella spp. sequence that was previously
found in the drinking water filters (Ibid.). PCR primers were
developed for the detection of the 16S rRNA gene from Gal-
lionella spp. using ARB software. One forward primer, GAL-
FER0218-F 50 -GCTTTCGGAGTGGCCGATA-30 -, and one reverse
primer, GALFER1408-R 50 - CAGATTCCACTCCCATGGTG -30
were designed. Amplification was performed by initial dena-
turation for 3 min at 94
C, followed by 35 cycles of amplifi-
cation (30 s denaturation at 94
C; 30 s annealing at 62
C; 1 min
elongation at 72
C), and 5 min at 72
C to complete elongation.
Quantification was based on a comparison of the sample Ct
value to the Ct value of a calibration curve using standard
numbers of 16S rDNA fragments of Gallionella (see
Supplementary Material C). It was assumed that Gallionella
cells contain 1 16S gene copy per cell. An internal control was
added to all samples to correct for the efficiency of the PCR
reaction. The specificity of the qPCR method was checked
through the construction and sequencing of clone libraries of
the PCR products from filtrate and backwash water samples of
the WTP Lekkerkerk full-scale filter (de Vet, 2011).
From the qPCR enumeration results, balances for the lab-
scale columns and the full-scale filters were calculated. To
assess the role of biological iron oxidation, the following
assumption was made: when the cell numbers entering and
leaving the filters are constant in time, no net accumulation
occurs and the net washout measured by qPCR balances the
5392 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8

Table 1 e Groundwater and filtrate quality and filter characteristics of the full-scale trickling filters at the Oasen WTPs.
WTP Lekkerkerk WTP De Hooge Boom

Direct trickling Direct trickling Plate aeration and

filtration filtration trickling filtration

Filter sand d10-d90, mm 1.7e2.5 2.0e3.15

Filter bed area, m2 18.0 28.0
Average production flow, m3 h1 37 64
Filter runtime, h 48 48
Groundwater quality Average  St. Dev. Average  St. Dev.
Jan.eSept. 2008 Jan. 2008eMar. 2010
Temperature,
C 11.6  0.3 11.5  0.2
pH, 7.33  0.04 7.10  0.05
HCO 3 , mg L
1
216  7 387  12
Total iron, mg L1 5.5  0.6 8.5  0.8
TOC, mg L1 C 2.2  0.1 8.3  0.2
Effluent plate aerator
pH, e e 7.7a
O2, mg L1 e e 10.1a
Filtrate quality
pH, 7.69  0.10 7.8a 7.80  0.03
O2, mg L1 9.7  0.6 (10 data points) 9.7a 10.2  0.4
(3 data points)
Total iron removed per 9.6  1.1 26.1  2.3
filter runtime, kg Fe

a Indicative local measurement on March 1, 2010.

growth of Gallionella spp., during one filter run. For every water
flow entering or leaving a system, the totalized values for the
cell numbers were calculated by multiplying the measured
concentration with the flow rate and duration of the phase.
For the full-scale filter at WTP Lekkerkerk, the groundwater
was sampled in duplicate; the filtrate in duplicate five and
nine months after external washing with four samples per
filter runtime of 48 h; the backwash water was sampled three,
six and nine months after external washing; control backwash
samples were taken in quintuplet 15e16 months after
external washing.
At WTP De Hooge Boom, the influent and backwash water of
each filter and the effluent of the plate aerator were sampled
once; the filtrate water of each filter was sampled twice, at the
beginning and at the end of the filter runtime of 48 h.
The filtrate of the filter columns was sampled two hours
after backwash. Filter column sand samples were taken from
the top half of the bed during expansion backwashing.
2.5. Iron analyses
Samples for iron analysis were taken directly into acid con-
taining bottles to set the pH below 2. Nitric and hydrochloric
acid were used to stabilize the samples for total and ferrous
iron analysis, respectively. All samples were stored cool and
analyzed within 24 h after sampling. Total iron in water
samples was determined by inductively coupled plasma mass
spectrometry (ICP-MS). Ferrous iron was determined by the
1,10-phenanthroline method according to the Dutch NEN 6482
protocol, based on Standard methods (1975). The iron
concentration in the sludge was measured by atomic emission
spectroscopy after sample destruction in a microwave. The
mass of the filter coating was determined by measurements of
the dry mass before and after acidification with 4 M
hydrochloric acid and oxalic acid. The iron concentration in
the decanted acid solution was measured by ICP-MS.
3. Results
3.1. pH effect on growth of Gallionella and iron
oxidation rate in oxidation columns
The pH dependency of Gallionella growth and iron oxidation
was examined in six lab-scale oxidation columns.
During the first week after start-up of the experiment, the
degree of oxidation was lower than the average for the following
six weeks for all oxidation columns except the one with pH 8.25
(Fig. 1). Apart from the first week, the oxidation degree of iron in
the columns’ effluent e calculated from the ferrous iron
concentrations in filter effluent (by 1,10-phenanthroline
method; Figure B.3 of Supplementary Material B) and added
iron concentrations (from mass balances) e was constant in
time. During the first week after start-up, virtually no iron
oxyhydroxides or FeOB had formed in the columns yet, and the
measured iron oxidation was accounted for mainly by the
homogeneous chemical process. After the start-up period, both
iron oxyhydroxides and FeOB accumulated in the columns and
influenced the oxidation kinetics.
The numbers of Gallionella cells determined by qPCR in the
influents and effluents of the columns after 3 and 7 weeks and
the total iron and Gallionella cell numbers accumulated as sludge
in the oxidation columns after 7 weeks are shown in Fig. 2.
The cell concentrations in the influent, effluent and sludge
show that Gallionella grew and accumulated in all oxidation
columns. Statistical analysis (Supplementary Material D)
shows that Gallionella grew equally fast in all columns with
a pH between 7.0 and 7.73, and slightly (but not significantly)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
Fig. 1 e Oxidation degree of iron after 16 min’ passage
through oxidation columns; green solid bars, averages and
standard deviations for 6e7 data points for whole period
except first week after start-up, calculated from the ferrous
iron concentrations in filter effluent (by 1,10-
phenanthroline method) and total iron concentrations
(from mass balances); blue striped bars, oxidation degree
of iron during first week after start-up; calculated with
[OHL]0.6, calculated with [OHL]2 (see Discussion section
for explanation). (For interpretation of the references to
colour in this figure legend, the reader is referred to the
web version of this article.)
faster at pH 6.5. The increasing rate of chemical iron oxidation
did not inhibit the growth of Gallionella up to a pH of 7.73. Only
in the column with pH 8.25 was Gallionella growth significantly
slower. The total iron accumulated in the oxidation columns
had a maximum at pH 7.00, and was for all columns between
1.8 and 3.6 g (3e6  102 mol). This was between 3 and 6% of
the iron loading. The majority of the iron, therefore, was
Fig. 2 e Gallionella spp. concentrations (left axis, in cells
mLL1) in the influent and the effluents of the oxidation
columns operated at different pH values after 3 weeks
(yellow solid bars) and after 7 weeks (green striped bars);
total Gallionella spp. numbers (red dotted bars on left axis,
in cells) and total iron (white bars on right axis, in g Fe)
accumulated in the oxidation column after 7 weeks; error
bars show uncertainty of qPCR method (between 0.5*N and
2*N). (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of
this article.)
5393
present in the effluent of the columns in dissolved ferrous or
colloidal ferric form.
Growth of Gallionella spp. was confirmed based on the
morphology of the deposits. Fig. 3 shows a phase contrast
picture of deposits in the oxidation column with pH 7.73.
In the oxidation column experiment, the specific growth
rate m of Gallionella spp. can be approached by Equation (1).
   
Xt;column þ Q  Dt  xt;effluent
ln
Xt;column
ðDtY0Þm ¼ (1)
Dt
where, Xt,column ¼ total cells in column (cells), Q ¼ flow rate
(m3 h1), [x]t,effluent ¼ cell concentration washed out of column
(cells m3), t ¼ experimental time (h).
At the end of the oxidation column experiment, m was
0.08  0.06 h1, corresponding to a doubling time of 8.4 h on
average. Hallbeck and Pedersen (1990) found a generation time
of 8.3 h in vitro at the optimal temperature of 20
C. This is
comparable to the growth rate observed in our experiments at
13
C, which suggests slightly more favorable growth condi-
tions in situ.
3.2. Effect of pre-oxidation on Gallionella growth in
trickling filters
The effect of pre-oxidation on the number of Gallionella spp. in
trickling filters was studied in the combined oxidation and
filtration column experiment. The pre-oxidation caused an
oxidation degree of 29  6% before the water entered the
trickling filters. At the applied pH (7.25  0.15) this corre-
sponded with the oxidation degree measured in the oxidation
column experiments (Fig. 1). The total numbers of Gallionella
spp. for the water flows cumulated over one filter runtime
(24 h) and for the filter beds at the end of the test period of 6
months are shown in Fig. 4. This figure clearly shows the
growth of Gallionella spp. in all filter columns spiked with
ferrous iron, but none in the reference column. No significant
difference was found in the water samples from filter columns
without and with pre-oxidation and only marginally more
Fig. 3 e Phase contrast microscopic picture (4003
magnification) of a sludge sample from the bottom of the
oxidation column with pH set to 7.73 after three weeks of
operation.
5394 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8

Fig. 4 e Iron deposition in filter coating and cumulative numbers of Gallionella spp. by qPCR determined for a reference
column, three columns without and three with pre-oxidation in the filter columns’ setup at the end of 6-months’ trial;
numbers in water flows (solid and striped bars) cumulated over one filter run of 24 h, yellow solid bar in feed water, green
vertically striped bar in filtrate water; and blue horizontally striped bars backwash water; total present in filter material
(dotted bars); iron deposition in filter coating ( after 91 days, after 187 days, on the right axis); error bars show uncertainty
of qPCR method for the outflow measurements of the reference and standard deviation for the other measurements. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Gallionella spp. in the filter samples from filter columns
without pre-oxidation (shown in Supplementary Material D).
Fig. 4 also shows iron deposition in the filter coating. After 91
days, the amount of iron deposited in the filter coatings was
comparable regardless of pre-oxidation or not, and on average
60% of the loaded iron (Supplemental Material C) was encap-
sulated in the filter coating. After 187 days, however, the amount
of iron in the filter coatings of columns with pre-oxidation had
not increased, while it had in the columns without pre-
oxidation. This suggests that the growth of attached FeOB in
the filters may enhance the formation of iron coating.
In the filter column experiments, the absence of pre-
oxidation resulted, on average, in slightly higher cell
numbers attached to the filter material, but due to error
margins it is not possible to judge if this is significant.
Although this is consistent with the higher ferrous iron
loading in the columns without pre-oxidation, there was no
significant difference in Gallionella spp. numbers in the water
flows from columns with and without pre-oxidation. With the
approach according to Equation (1), m was calculated as
0.01  0.005 and 0.03  0.01 h1 for the filter columns without
and with pre-oxidation, respectively (with equal distribution
of the cells washed out during backwash periods over the filter
runtime). This suggests that the Gallionella cells in the
columns without pre-oxidation were better attached, more
encapsulated in the iron oxyhydroxide filter coating, and less
active than in the columns with pre-oxidation.
3.3. Full-scale groundwater trickling filters
In order to determine the potential role of biological oxidation
in the groundwater trickling filters, the abundance and growth
of the iron-oxidizing Gallionella species were assessed in the
three full-scale filters by qPCR. The balances for Gallionella spp.
in duplicate calculated over one filter run of 48 h from the
qPCR cell numbers, water flows, and time are shown in Fig. 5.
Repeated measurements over the trial period of nine months
at WTP Lekkerkerk (see Supplemental Material C) showed no
trend in cell numbers, indicating a stable population.
The measurements show that significant numbers of Gal-
lionella cells were found in all three full-scale filters despite the
fact that these filters were very well aerated and the oxygen
content of the filtrate water was close to saturation level. This
condition is usually associated with chemical iron oxidation
(Sharma et al., 2005). The cell numbers leaving the filter
through the filtrate and backwash water were much higher
than in the groundwater feeding the filter. This indicates
a strong growth of Gallionella spp. in these full-scale trickling
filters. The plate aeration prior to the filtration at WTP De Hooge
Boom did not inhibit the growth of Gallionella spp. in the filter,
despite the oxygen saturation and elevated pH of the effluent
water. Gallionella spp. started to grow in the plate aerator, were
filtered off in the trickling filter and continued growing there.
4. Discussion
4.1. Chemical versus biological iron oxidation in
groundwater filtration
Iron oxidation under aerobic, neutral pH conditions may be
homogeneous, heterogeneous or biologically mediated. At the
start of the oxidation column experiment, a negligible amount
of iron oxyhydroxides and FeOB was present, and the iron
oxidation was predominantly homogeneous. The general
kinetic equation for homogeneous iron oxidation is given by
Equation (2):
 n  m
dFe=dt ¼ k  OH PO2  Fe2þ (2)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8 5395

Fig. 5 e Gallionella spp. balances for the WTP Lekkerkerk trickling filter (multiple measurements, graph A, red solid bars) and
for the WTP De Hooge Boom trickling filters (singular measurements, graph B, green square bars, trickling filtration after plate
aeration; blue line bars, direct trickling filtration); cumulative cell numbers inoculated directly from groundwater or via plate
aerator, washed out to effluent and to backwash are calculated for one filter runtime of 48 h; Influent filter values are for
effluent plate aerator if present and equal to groundwater for the other filters; error bars show standard deviation (graph A)
and uncertainty of qPCR method (graph B, between 0.5*N en 2*N). (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

where m ¼ 1 and n ¼ 2, as found by Sung (1980); only for the first
week measurements did the model reasonably match the
measured data (blue squares in Fig. 1; 13
C, ionic strength
0.01 M) with a rate constant of k ¼ 4$1012 M2 atm1 min1. This
is approximately 10 times lower than the rate reported by Sung
(Ibid.) in water with similar salinity but at 25
C. The tempera-
ture difference between 25
C and 13
C, explains this difference
for the larger part. The lower temperature reduces the rate
constant by a factor of 7, not because of changes in the activa-
tion energy (almost zero), but by the decline in Kw and thus [OH]
activity (Stumm and Lee, 1961). This strongly indicates that the
iron oxidation was mostly a homogeneous chemical reaction in
the first week of the oxidation column experiment.
The measurements during the rest of the experimental
period can only be fitted to the model by reducing the order of
[OH], i.e. n, to 0.6 (green triangles in Fig. 1). Tamura et al.
(1976) found that rate of heterogeneous chemical iron
oxygenation was proportional to the first order of the recip-
rocal [Hþ]. The general kinetic equation for heterogeneous
chemical iron oxidation is given by Equation (3):
 
  cannot be deduced from our experiments, but is probably
dFe= ¼ k1 þ k2  Fe3þ  Fe2þ (3)
dt
where
 2
k1 ¼ khom  OH PO2 Homogeneous Oxidation rate Constant;
(3a)
 1
k2 ¼ ks;O  ½O2   K  Hþ Heterogenous oxidation rate constant;
(3b)
ks;O ¼ 4380 M1 min
1
Surface rate ; (3c)
K ¼ 104:85 Adsorption constant of Fe2þ on FeOOH; (3d)
The iron sludge that accumulated at the end of the experi-
mental period of seven weeks in the oxidation columns was
assumed to be equally distributed in the oxidation columns,
leading to a ferric iron concentration of 9e18 mM (see Fig. 2;
sludge volume per column was 4.0  0.1 L). Under these
conditions, the heterogeneous oxidation rate constant k2 Fe3þ
according to Equation (3) would be in the range 1e35 min1
(for pH 6.5 up to 8.25, respectively). As this means an oxidation
half-life (t½) of less than 1 min, it would implicate a nearly
complete chemical oxidation of iron after the average resi-
dence time of 16 min in all the oxidation columns, which was
not the case in our experiments.
The reason for this reduced heterogeneous oxidation rate
related to the composition of the natural water. In many
studies on chemical iron oxidation (Stumm and Lee, 1961;
Sung, 1980; Tamura et al., 1976) the oxidation rates were
determined with ultrapure water. Some studies determined
the effects of natural organic matter (NOM) on the chemical
iron oxidation. Davison and Seed (1983) and Liang et al. (1993)
5396 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
found that the rate constant for homogeneous iron oxidation
in natural freshwater under oxygen-saturated conditions was
comparable to the one in synthetic water, as we did. Other
researchers found a significant effect of NOM complexation
on iron removal and oxidation, but that effect could be either
accelerating (Ninh Pham et al., 2004) or inhibiting (Theis and
Singer, 1974). All these studies were confined to homoge-
neous chemical iron oxidation.
Sung (1980) stated that catalytic iron oxidation was only
noticeable at pH 7 and above because of the slow surface
formation at a lower pH. Our research with natural water
indicates that the heterogeneous iron oxidation rate was
strongly reduced even at a higher pH. This reduced hetero-
geneous chemical iron oxidation rate may be caused by
surface complexation of inorganic and (natural) organic
compounds in water. Complexation of inorganic ions had
little influence on the adsorption capacity of ferrous iron
(Sharma, 2001). Tipping (1981) showed that the surface charge
and adsorption capacity of iron oxyhydroxides could be
influenced by the complexation of humic substances. Gallio-
nella growth may have contributed to the surface complexa-
tion and stabilization of the iron (oxy)hydroxides by the
excretion of polysaccharides, comparable to the stalk forma-
tion described by Chan et al. (2011).
Analysis of the growth of Gallionella spp. by qPCR demon-
strates the significance of bacterial iron oxidation in the full-
scale filter and laboratory filter columns. The direct enumer-
ation of Gallionella spp. by this method combined with the
biomass yield on iron oxidation makes it possible to quantify
the share of biological iron oxidation. The maximum biomass
yield reported in the literature is low (0.006 g DW g1 Fe
oxidized (Lütters and Hanert, 1989) and 0.013 g DW g1 Fe
oxidized (Neubauer et al., 2002). Thermodynamically,
a maximum theoretical yield of 0.012 g DW g1 Fe can be
expected, based on the anabolic reaction energy of
3500 kJ C mol1 biomass (Heijnen and Van Dijken, 1992) and
the catabolic reaction energy (Hanselmann, 1991): Fe2þ þ ¼
O2 þ 1½ H2O / FeOOH þ 2Hþ with ΔGr ¼ 83.8 kJ mol1 Fe at
pH 7.73 and 1 mM Fe2þ.
During one filter runtime, 0.9 g of iron was removed in the
laboratory filter system (Supplementary Material B) and
9.6  1.1 kg in the full-scale filter at WTP Lekkerkerk. During
one runtime, in total 4.7  2.9  1011 and 3.4  2.9  1015
Gallionella cells were washed out of these filters, respectively.
When biomass accumulation, maintenance and decay were
not considered, the observed yield was 5.1  3.3  1011 and
3.6  3.4  1011 Gallionella cells g1 Fe oxidized, respectively.
Table 2 e Overview of iron conversion, net Gallionella cells washout and calculated yield for filter column experiment and
full-scale trickling filter at WTP Lekkerkerk.
Parameters Unit
Iron removed per filter runtime g Fe
Gallionella cells washed out per filter runtime cells
Cell yield cells g1 Fe
Biomass yielda g DW g1 Fe
a Calculated with 1.2  1013 g DW cell1, mean cell volume 0.4 mm3, Hallbeck and Pedersen (1991).
This assumes that the iron oxidation was completely biolog-
ical and exclusively by Gallionella spp.. These maximum Gal-
lionella cell yields can be related to dry weight (DW) by using
the cell dimensions (mean volume of 0.4 mm3) determined by
Hallbeck and Pedersen (1991). With a specific cell DW of
1.2  1013 g, the yield equals 0.062  0.040 and 0.043  0.041 g
DW g1 Fe oxidized, for the filter columns and the full-scale
filter, respectively (Table 2).
Although the standard deviations are large, the average
yield was higher than reported in the literature and the
theoretical maximum. The high cell yields found suggest that
biological iron oxidation by Gallionella spp. played a dominant
role in both the full-scale filter and in the filter columns.
4.2. Growth conditions of Gallionella spp.
The results reported in this manuscript show that Gallionella
spp. may grow under broader conditions than generally
assumed. No growth inhibition was found in the natural water
under fully aerated conditions and at a pH ranging from 6.5 to
7.73. This finding contrasts with the general perception of
Gallionella being strictly microaerophilic (Emerson, 2000). The
oxygen concentration at the microsites where Gallionella grew
in our experiments may have been reduced compared to the
bulk water. We did not measure the oxygen concentration in
those microsites. When iron oxidation is the dominant
oxygen-consuming process, however, only a limited reduction
in oxygen concentration may be expected from the bulk water
into microsites, based on the stoichiometry of the iron
oxidation, the measured oxygen and ferrous iron concentra-
tions in the bulk water and the diffusion coefficients for both
compounds from the literature (Broecker and Peng, 1974; Li
and Gregory, 1974).
According to Degrémont (2007), biological iron oxidation
will only prevail under conditions where physico-chemical
iron oxidation is not possible: oxygen concentration between
0.2 and 0.5 mg L1, pH 6.3, oxidation reduction potential
þ100 mV and rH2 between 14 and 20 (whereas rH2 ¼ log
( pH2) ¼ Eh/0.0296 V þ 2 pH). Under rH2 of 14, the biological
oxidation should be inhibited, while over 20, the bacteria
would lose the competition with the physico-chemical iron
precipitation. At pH 7.73, the upper limit of rH2 indicates
a maximum redox potential of 135 mV and an oxidation
degree of less than 98%.
It was stated that the boundaries are not strictly defined
and can shift e.g. by chelation. Hanert (2006) listed the broad
array of the environments where Gallionella spp. have been
Filter columns Full-scale trickling filter
0.92  0.04 9.6  1.1  103
4.7  2.9  1011 3.4  2.9  1015
5.1  3.3  1011 3.6  3.4  1011
0.062  0.040 0.043  0.041
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
found and concluded that the stability of ferrous iron in
combination with oxygen is crucial for their existence, more
than mere pH or Eh. This paper substantiates this claim by
showing the growth of Gallionella spp. on ferrous iron under
fully aerated and slightly alkaline circumstances, when the
chemical iron oxidation is slow.
In the oxidation column experiment with a fixed pH
ranging from 6.5 to 7.73 and oxygen-saturated natural water,
the initial iron oxidation was homogeneous with rates
consistent with the literature. After the start-up period, FeOB
and iron oxyhydroxydes accumulated in the columns but the
oxidation rate increased less than theoretically expected from
heterogeneous chemical oxidation. Heterogeneous chemical
iron oxidation may be seriously hampered in natural water
compared to synthetic water by complexation of natural
organic matter on iron oxyhydroxide surfaces. The specific
growth of Gallionella spp. was in accordance with the values
found in culture experiments. The comparable Gallionella cell
growth and the increase in iron oxidation degree indicate that,
for pH ranging from 6.5 to 7.73, the increased iron oxidation
rate had to be attributed to the growth and activity of FeOB,
rather than to chemical catalysis. Yield calculations for the
biological iron oxidation by Gallionella spp. in lab- and full-
scale trickling filters, indicate that the dominant iron oxida-
tion mechanism in groundwater filtration is biological under
wider process conditions (pH and oxygen content) than
previously thought.
5. Conclusions
 The quantitative PCR approach targeting the 16S rRNA of Gal-
lionella spp. was successfully used to determine the significance
of biological versus chemical oxidation in full-scale ground-
water trickling filters and lab-scale column experiments.
 Gallionella spp. grew in fully aerated full-scale groundwater
trickling filters and lab-scale oxidation columns and trick-
ling filters at neutral pH (up to pH 7.7) and at a moderate
temperature of 13
C.
 Biological oxidation by Gallionella spp. was the dominant
process for iron oxidation in this type of groundwater, and
heterogeneous chemical iron oxidation in natural water was
substantially reduced, compared to experimental results
from the literature for synthetic water.
Acknowledgements
The authors gratefully acknowledge the contribution of Peter
Dijkstra for assistance in the full-scale research, Petra Lafeber
for support in the column experiments and Sabine Doddema
and Paul van der Wielen for the DNA-isolation for the qPCR.
Appendix. Supplementary data
Supplementary data related to this article can be found online
at doi:10.1016/j.watres.2011.07.028.
5397
references
Berger, H., Berger, E., 1928. Biologie der Trink- und
Brauchwasseranlagen. Verlag von Gustav Fischer, Jena. Berlin-
Dahlem, Germany.
Broecker, W.S., Peng, T.H., 1974. Gas exchange rates between air
and sea. Tellus 26 (1e2), 21e35.
Chan, C.S., Fakra, S.C., Emerson, D., Fleming, E.J., Edwards, K.J.,
2011. Lithotrophic iron-oxidizing bacteria produce organic
stalks to control mineral growth: implications for biosignature
formation. ISME Journal 5 (4), 717e727.
Czekalla, C., Mevius, W., Hanert, H., 1985. Quantitative removal of
iron and manganese by microorganisms in rapid sand filters
(in situ investigations). Water Supply 3 (1), 111e123.
Davison, W., Seed, G., 1983. The kinetics of the oxidation of
ferrous iron in synthetic and natural waters. Geochimica et
Cosmochimica Acta 47 (1), 67e79.
Degrémont (Ed.), 2007. Water Treatment Handbook. Lavoisier SAS.
Emerson, D., Floyd, M.M., 2005. Enrichment and isolation of iron-
oxidizing bacteria at neutral pH. Methods in Enzymology 397,
112e123.
Emerson, D., Weiss, J.V., 2004. Bacterial iron oxidation in
circumneutral freshwater habitats: findings from the field and
the laboratory. Geomicrobiology Journal 21 (6), 405e414.
Emerson, D., 2000. Environmental microbe-metal interactions. In:
Lovley, D.R. (Ed.). ASM, Washington, pp. 31e52.
Hallbeck, L., Pedersen, K., 1990. Culture parameters regulating
stalk formation and growth rate of Gallionella ferruginea.
Journal of General Microbiology 136 (9), 1675e1680.
Hallbeck, L., Pedersen, K., 1991. Autotrophic and mixotrophic
growth of Gallionella ferruginea. Journal of General Microbiology
137 (11), 2657e2661.
Hanert, H.H., 2006. The Prokaryotes: a Handbook on the Biology of
Bacteria. Springer, pp. 990e995.
Hanselmann, K.W., 1991. Microbial energetics applied to waste
repositories. Experientia 47 (7), 645e687.
Heijnen, J.J., Van Dijken, J.P., 1992. In search of thermodynamic
description of biomass yields for the chemotrophic growth of
microorganisms. Biotechnology and Bioengineering 39 (8),
833e858.
Li, Y.-H., Gregory, S., 1974. Diffusion of ions in sea water and in
deep-sea sediments. Geochimica et Cosmochimica Acta 38 (5),
703e714.
Liang, L., Andrew McNabb, J., Paulk, J.M., Gu, B., McCarthy, J.F.,
1993. Kinetics off Fe(II) oxygenation at low partial pressure of
oxygen in the presence of natural organic matter.
Environmental Science and Technology 27 (9), 1864e1870.
Lütters, S., Hanert, H.H., 1989. The ultrastructure of
chemolithoautotrophic Gallionella ferruginea and Thiobacillus
ferrooxidans as revealed by chemical fixation and freeze-
etching. Archives of Microbiology 151 (3), 245e251.
Lütters-Czekalla, S., 1990. Lithoautotrophic growth of the iron
bacterium Gallionella ferruginea with thiosulfate or sulfide
as energy source. Archives of Microbiology 154 (5),
417e421.
Nederlands Normalisatie instituut (NEN), 1993. NEN-ISO 5814:
1993 en, Water e Bepaling van het gehalte aan opgeloste
zuurstof e Elektrochemische methode. http://www.nen.nl/
web/Normshop/Norm/NENISO-58141993-en.htm.
Nederlands Normalisatie instituut (NEN), 2009. NEN-ISO 10523:
2008 en, Water e Bepaling van de pH. http://www.nen.nl/web/
Normshop/Norm/NENISO-105232008-en.htm.
Neubauer, S.C., Emerson, D., Megonigal, J.P., 2002. Life at the
energetic edge: kinetics of circumneutral iron oxidation by
lithotrophic iron-oxidizing bacteria isolated from the wetland-
plant rhizosphere. Applied and Environmental Microbiology
68 (8), 3988e3995.
5398 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 8 9 e5 3 9 8
Ninh Pham, A., Rose, A.L., Feitz, A.J., Waite, T.D., 2004. The Effect
of Dissolved Natural Organic Matter on the Rate of Removal of
Ferrous Iron in Fresh Waters, pp. 213e219.
Sharma, S.K., Petrusevski, B., Schippers, J.C., 2005. Biological iron
removal from groundwater: a review. Journal of Water Supply:
Research and Technology e AQUA 54 (4), 239e247.
Sharma, S.K. (2001) Adsorptive iron removal from groundwater.
Dissertation Wageningen University/International Institute
for Infrastructural, Hydraulic and Environmental Engineering,
Delft, the Netherlands.
Søgaard, E.G., Medenwaldt, R., Abraham-Peskir, J.V., 2000.
Conditions and rates of biotic and abiotic iron precipitation in
selected Danish freshwater plants and microscopic analysis of
precipitate morphology. Water Research 34 (10), 2675e2682.
Spring, S., Kämpfer, P., 2005. Bergey’s Manual of Systematic
Bacteriology, pp. 740e746.
Standard methods for the Examination of Water and Waste
Water, fourteenth ed., 1975 American Public Health
Association, Washington, pp. 208e213.
Stumm, W., Lee, G.F., 1961. Oxygenation of Ferrous Iron, pp.
143e146.
Sung, W., 1980. Kinetics and product of ferrous iron oxygenation
in aqueous systems. Environmental Science and Technology
14 (5), 561e568.
Tamura, H., Goto, K., Nagayama, M., 1976. Effect of ferric
hydroxide on the oxygenation of ferrous ions in neutral
solutions. Corrosion Science 16 (4), 197e207.
Tekerlekopoulou, A.G., Vayenas, D.V., 2008. Simultaneous
biological removal of ammonia, iron and manganese from
potable water using a trickling filter. Biochemical Engineering
Journal 39 (1), 215e220.
Theis, T.L., Singer, P.C., 1974. Complexation of iron(II) by organic
matter and its effect on iron(II) oxygenation. Environmental
Science and Technology 8 (6), 569e573.
Tipping, E., 1981. The adsorption of aquatic humic substances by
iron oxides. Geochimica et Cosmochimica Acta 45 (2), 191e199.
de Vet, W.W.J.M., Dinkla, I.J.T., Muyzer, G., Rietveld, L.C., van
Loosdrecht, M.C.M., 2009. Molecular characterization of
microbial populations in groundwater sources and sand filters
for drinking water production. Water Research 43 (1), 182e194.
de Vet, W.W.J.M. (2011). Biological drinking water treatment of
anaerobic groundwater in trickling filters. Dissertation.
Technical University Delft, The Netherlands. http://repository.
tudelft.nl.
de Vries, H., 1890. Die Pflanzen und Thiere in den dunklen
Räumen der Rotterdamer Wasserleitung; Bericht über die
biologischen Untersuchungen der Crenothrix-Commission zu
Rotterdam, vom Jahre 1887. Verlag von Gustav Fischer, Jena.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Temporal variability of pharmaceuticals and illicit drugs

in wastewater and the effects of a major sporting event

Daniel Gerrity a,b,*, Rebecca A. Trenholm b, Shane A. Snyder b,c

a
Trussell Technologies, Inc., 6540 Lusk Blvd., Suite C274, San Diego, CA 92121, United States
b
Applied Research and Development Center, Southern Nevada Water Authority, River Mountain Water Treatment Facility, P.O. Box 99954,
Las Vegas, NV 89193-9954, United States
c
Department of Chemical and Environmental Engineering, University of Arizona, 1133 E. James E. Rogers Way, Harshbarger 108, Tucson,
AZ 85721-0011, United States

article info abstract

Article history: Diurnal variations in wastewater flows are common phenomena related to peak water use
Received 23 April 2011 periods. However, few studies have examined high-resolution temporal variability in trace
Received in revised form organic contaminant (TOrC) concentrations and loadings. Even fewer have assessed the
23 June 2011 impacts of a special event or holiday. This study characterizes the temporal variability
Accepted 17 July 2011 associated with a major sporting event using flow data and corresponding mass loadings of
Available online 23 July 2011 a suite of prescription pharmaceuticals, potential endocrine disrupting compounds (EDCs),
and illicit drugs. Wastewater influent and finished effluent samples were collected during
Keywords: the National Football League’s Super Bowl, which is a significant weekend for tourism in
Pharmaceutical the study area. Data from a baseline weekend is also provided to illustrate flows and TOrC
Endocrine disrupting loadings during “normal” operational conditions. Some compounds exhibited interesting
compound (EDC) temporal variations (e.g., atenolol), and several compounds demonstrated different loading
Illicit drug profiles during the Super Bowl and baseline weekends (e.g., the primary cocaine metabolite
Temporal variation benzoylecgonine). Interestingly, the influent mass loadings of prescription pharmaceuti-
Loading cals were generally similar in magnitude to those of the illicit drugs and their metabolites.
Reuse However, conventional wastewater treatment was more effective in removing the illicit
Wastewater drugs and their metabolites. Total influent and effluent mass loadings are also provided to
summarize treatment efficacy and environmental discharges.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction aquatic ecosystems (Snyder et al., 2001, 2004; Lange et al.,

Pharmaceuticals and personal care products (PPCPs) and
endocrine disrupting compounds (EDCs) are often considered
“emerging contaminants,” but researchers have been aware of
their presence in water for decades. However, the occurrence
of PPCPs and EDCs in water did not become a mainstream
research topic until the late 1990s and early 2000s. The spike in
scientific interest stemmed from demonstrated impacts on
2009), potential human health effects (Snyder et al., 2008;
Schriks et al., 2010; Stanford et al., 2010), and increased
media coverage (Donn et al., 2008), which ultimately led to
increased public awareness. This increased interest was
coupled with the development of extremely sensitive analyt-
ical methods such as liquid chromatographyetandem mass
spectrometry (LCeMS/MS) that allowed researchers to
approach parts-per-quadrillion (sub-ng/L) detection limits for

* Corresponding author. Trussell Technologies, Inc., 6540 Lusk Blvd., Suite C274, San Diego, CA 92121, United States. Tel.: þ1 858 458 1030;
fax: þ1 626 486 0571.
E-mail address: dan.gerrity@trusselltech.com (D. Gerrity).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.07.020
5400 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
a variety of trace organic contaminants (TOrCs) (Ternes et al.,
2002; Snyder et al., 2003; Vanderford and Snyder, 2006; Postigo
et al., 2011). Each of these factors led to more thorough
scientific investigations into the presence, fate, and transport
of TOrCs in natural and engineered systems.
With respect to organic compounds intended for human
consumption, contamination of water supplies stems from
their release during manufacturing, excretion after personal
use, and public disposal of unused quantities (Daughton and
Ternes, 1999). Considering that each of these routes directly
impacts wastewater, it is reasonable to assume that dis-
charged wastewater is a major source of these contaminants
in environmental waters. In the past, wastewater treatment
trains were generally not designed for TOrC removal.
However, the prevalence of indirect potable reuse, whether
“planned” or “unplanned”, and demonstrated impacts on
aquatic ecosystems now justify some consideration of TOrCs
in the design process. In fact, expansion and optimization of
wastewater treatment processes may be the most efficient
strategy to mitigate the potential effects of these contami-
nants (Nelson et al., 2011). To aid in this effort, one design
factor that must be studied in greater detail is the temporal
variability of TOrC occurrence in wastewater.
Diurnal variations in wastewater flows are common
phenomena related to peak water use periods (Nelson et al.,
2011). Sewers and wastewater treatment plants must be
designed to account for the maximum and minimum flows
and associated loadings each day (Ort et al., 2010). However,
few studies have examined temporal fluctuations in TOrC
concentrations and mass loadings (Joss et al., 2005; Takao
et al., 2008; Nelson et al., 2011; Plosz et al., 2010; Postigo
et al., 2011). In the existing studies, the temporal resolution
was typically limited to sampling intervals of 8 h or longer.
However, one recent study evaluated finished effluent
samples for a suite of TOrCs on an hourly basis (Nelson et al.,
2011). These high-resolution finished effluent samples char-
acterize the temporal variability of environmental discharges,
but they do not indicate the temporal variability of the
influent mass loadings due to attenuation during treatment.
Another recent study indicated that influent TOrC
concentrations may vary on extremely short time scalesd-
even as short as 2 mindand this may bias many of the recent
wastewater monitoring studies (Ort et al., 2010). The authors
emphasized that influent wastewater is “composed of
a number of intermittently discharged, individual wastewater
packets from household appliances, industries, or subcatch-
ments” (Ort et al., 2010). The authors demonstrated that such
temporal variation exists by monitoring TOrCs at time scales
that could capture a single toilet flush. As noted in their study,
such resolution is often limited by the costly, labor-intensive
analytical methods necessary to detect trace concentrations
of organic contaminants in wastewater. In fact, the authors
indicated that only two previous studies had reported TOrC
concentrations with sufficient temporal resolution (Ort et al.,
2005; Ort and Gujer, 2006).
The extent of temporal variation is dependent on the
characteristics of the target compounds and the size of the
service area for a particular wastewater treatment plant,
thereby accounting for the number of toilet flushes containing
the compounds of interest (Ort et al., 2010). In a small
catchment, contrast media used for magnetic resonance
imaging will demonstrate much more temporal variability
than compounds with more widespread human consumption
(Joss et al., 2005; Ort et al., 2010; Nelson et al., 2011), such as
non-steroidal anti-inflammatory drugs (Ternes, 1998; Joss
et al., 2005). In particular, X-ray contrast media are more
prevalent on weekdays when most scheduled medical
appointments occur (Nelson et al., 2011). Ort et al. (2010) also
emphasized that sampling uncertainty will even be a factor
for large systems. Therefore, additional studies are necessary
to characterize the temporal variation that is generally lost in
large composite samples.
Days of the week, seasons, and even special events may
warrant design or operational considerations given their
potential for unusual flow patterns and contaminant loadings.
Nelson et al. (2011) reported significant concentration spikes
for the insect repellant N,N-diethyl-meta-toluamide (DEET)
during the warmer months when mosquitoes are most prev-
alent. With respect to special events, a recent article docu-
mented spikes in wastewater flows caused by different stages
of a major auto race in Speedway, Indiana, USA (Enfinger and
Stevens, 2011). Another study evaluated days of the week,
seasons, and winter holidays for their effects on illicit drug
concentrations in Spanish surface water (Huerta-Fontela
et al., 2008). The authors observed higher illicit drug concen-
trations, including amphetamine-type stimulants, cocaine,
and cocaine metabolites, on weekends compared to week-
days, and the authors also observed the highest concentra-
tions in the winter, particularly after the Christmas and New
Year holidays (Huerta-Fontela et al., 2008). These weekly
fluctuations and holiday-specific spikes are supported by
other studies of illicit drug use in Canada and Spain (Metcalfe
et al., 2010; Postigo et al., 2011).
The current study addresses some of these issues, including
high-resolution temporal variability and the effects of a special
event. This study presents flow data and corresponding
influent and effluent mass loadings of a suite of prescription
pharmaceuticals, potential EDCs, and illicit drugs at a waste-
water treatment plant in a major metropolitan area in the
United States (U.S.). Samples were collected during the
National Football League’s Super Bowl weekend in addition to
a baseline weekend to compare flows and mass loadings during
“special event” versus “normal” operational conditions.
Although the game was not held in the study area, the Super
Bowl causes a tremendous spike in tourism and associated
wastewater flows. The additional wastewater flows pose
potential operational issues for local wastewater treatment
plants, including fluctuations in the loadings of prescription
and illicit drugs. This study characterizes these issues and
provides further evidence of the importance of sample collec-
tion strategies in accurately characterizing TOrC concentra-
tions in wastewater, as emphasized in Ort et al. (2010).
2. Materials and methods
2.1. Sampling location
Samples were collected from a municipal wastewater treat-
ment plant with an average daily flow of approximately
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
380,000 m3/day (100 million gallons per day (MGD)). The service
area of the wastewater treatment plant is approximately 505
km2 (195 mi2) with a total population of approximately 1
million people. More than 99% of the flow is delivered to the
treatment plant by 3206 km (1992 mi) of gravity sewer lines,
while the remaining portion is delivered by continuously
operated lift stations and 64 km (40 mi) of pressurized lines. The
flow rates at the lift stations remain relatively constant
throughout the day.
The principal treatment train consists of bar screens;
grit removal; primary clarification with ferric chloride
addition; activated sludge with full nitrification (NH3,eff
<0.1 mg-N/L), partial denitrification, and biological phos-
phorus removal (TPeff < 100 mg/L); secondary clarification;
dual-media filtration with alum addition; and UV disinfec-
tion (UV dose ¼ 40 mJ/cm2). The activated sludge process is
generally operated with a solids retention time (SRT) of 7
days. Approximately 50% of the secondary effluent is
diverted to an advanced water treatment plant, which
includes coagulation/flocculation, sedimentation, dual-
media filtration, and UV disinfection or chlorination
(depending on the final use). A simplified treatment sche-
matic for the principal treatment train is provided in Fig. S1
(Supplementary data).
The chlorine-disinfected effluent from the advanced
water treatment plant, which constitutes an extremely small
portion of the overall flow, is used for reclaimed water
applications (e.g., golf course irrigation). The UV-disinfected
effluent from the principal treatment train (z50% of
the flow) and the advanced water treatment plant (z50% of
the flow) is discharged into a local wash and ultimately
a drinking water reservoir. Since the reservoir is the imme-
diate drinking water source for the study area and millions of
people downstream, this facility and the other local waste-
water treatment plants are contributors to indirect potable
reuse.
2.2. Sample collection and preservation
Thirty-minute, composite samples were obtained by diverting
a constant, continuous side stream of primary clarifier effluent.
Although the sampling was not flow-proportional, the relative
standard deviation (RSD) of the flow rate over the 30-min
sampling periods was less than 5%. These wastewater
“influent” samples were collected over 12-h periods on
February 7th/8th, 2010 (Super Bowl), and March 7th/8th, 2010
(baseline), for a total of 24 samples per sample event. The
samples were collected from 3:00 PM Sunday afternoon to
3:00 AM Monday morning to capture any wastewater effects
related to Super Bowl viewing (all times are local and based on
Pacific Standard Time (PST)). The 12-h time period was inten-
ded to account for the duration of pregame, Super Bowl, and
postgame activities in addition to potential lag times in the
sewer system. March 7th/8th was selected as a baseline
weekend since there were no large events occurring in the area.
In order to assess treatment efficacy, corresponding finished
effluent samples from the principal treatment train were also
collected during the Super Bowl and baseline weekends, but
the sampling times were adjusted to account for the hydraulic
retention time (HRT) of the wastewater treatment plant (z9 h).
5401
Finished effluent samples were collected from 12:00 AM
Monday morning to 12:00 PM Monday afternoon.
All samples were collected with refrigerated composite
samplers in amber glass bottles preserved with 1 g/L of sodium
azide as a biocide and 50 mg/L of ascorbic acid to quench any
potential oxidant residual. Samples were stored at 4  C before
laboratory filtration with 0.7-mm GF/F filters (Whatman, Pis-
cataway, NJ, USA), on-line solid phase extraction (for PPCPs/
EDCs), and analysis. Prior to analysis, a separate aliquot of
sample was prepared for each analytical method, spiked with
an appropriate stock solution of isotopically labeled standards,
and placed into 2-mL autosampler vials.
2.3. Target compounds
The following pharmaceuticals and potential EDCs were
monitored during the study: atenolol, atrazine, carbamazepine,
DEET, meprobamate, phenytoin, primidone, sulfamethoxazole,
tris-(2-chloroethyl)-phosphate (TCEP), and trimethoprim.
Except for atrazine, which was purchased from ChemService
(West Chester, PA, USA), all of these compounds were obtained
from SigmaeAldrich (St. Louis, MO, USA). Meprobamate-d3,
sulfamethoxazole-d4, and trimethoprim-d9 were obtained from
Toronto Research Chemicals (Ontario, Canada). Phenytoin-d10,
atrazine-d5, and atenolol-d7 were obtained from C/D/N Isotopes
(Pointe-Claire, Canada). DEET-d6, carbamazepine-d10, and pri-
midone-d5 were obtained from Cambridge Isotope Laboratories
(Andover, MA, USA). TCEP-d12 was synthesized by Isotec (St.
Louis, MO, USA). All concentrated stock solutions were
prepared in methanol and stored at 20  C. Working stock
solutions were prepared frequently in either reagent water or
methanol and stored at 4  C. All solvents were trace analysis
grade from Burdick and Jackson (Muskegon, MI). Reagent water
was obtained using a Milli-Q Ultrapure Water Purification
System (Millipore, Bedford, MA, USA).
The following illicit drugs and metabolites were monitored
during the study: methamphetamine and its metabolite
amphetamine; cocaine and its metabolites ecgonine, ecgonine
methyl ester, benzoylecgonine (BZE), and norcocaine; 3,4-
methylenedioxymethamphetamine (MDMA) and its metabo-
lite 3,4-methylenedioxyamphetamine (MDA); heroin and its
metabolites 6-acetylmorphine and morphine; and D-9-
tetrahydrocannabinol (THC) and its metabolite 11-hydroxy-
D-9-THC. All illicit drug standards and isotopically labeled
standards were obtained from Cerilliant (Austin, TX, USA) as
individual concentrated stock solutions in either methanol or
acetonitrile.
2.4. Analysis of pharmaceuticals and potential EDCs
Extraction and analysis for the pharmaceuticals and potential
EDCs were performed using on-line solid phase extraction
and liquid chromatographyetandem mass spectrometry
(SPEeLCeMS/MS). Following laboratory filtration with 0.7-mm
GF/F filters, the extraction was performed using a Symbiosis
(Spark Holland, Emmen, the Netherlands) automated on-line
solid phase extractor and a 4000 QTRAP triple quadrupole-
linear ion trap hybrid mass spectrometer (ABSCIEX, Foster
City, CA, USA). Full method descriptions and parameters,
including details for the on-line SPE method, have been
5402 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
described previously (Trenholm et al., 2009). Briefly, for each
extraction, a 1-mL sample volume was loaded onto a condi-
tioned HLB cartridge. The analytes were eluted from the
cartridge with 100 mL of methanol directly into the LC mobile
phase just prior to the LC column. Separation was performed
on a 150  4.6-mm Luna C18(2) column with a 5-mm particle
size (Phenomenex, Torrance, CA, USA) and a mobile phase
consisting of 5 mM ammonium acetate in DI water (A) and
methanol (B) gradient. The flow rate was maintained at
a constant 800 mL/min. All samples were analyzed using
positive electrospray ionization (ESI) and tandem mass spec-
trometry with multiple reaction monitoring (MRM). Two MS/
MS transitions were used for each compound for quantitation
and confirmation. Quantitation was performed using isotope
dilution, which involves the correction of target compound
concentrations based on the recovery of spiked isotopically
labeled standards (Vanderford and Snyder, 2006). The target
compounds, method reporting limits (MRLs), and MRM tran-
sitions are listed in Table S1 (Supplementary data). Fig. S2
(Supplementary data) is an example chromatogram of a 100-
ng/L calibration standard; for clarity, isotopically labeled
standards and confirmation transitions were not shown.
2.5. Analysis of illicit drugs and metabolites
The illicit drugs and their major metabolites were analyzed by
liquid chromatographyetandem mass spectrometry (LCeMS/
MS) using a CTC Autosampler (CTC Analysis, Zwingen,
Switzerland), an Agilent 1100 LC Binary Pump (Palo Alto, CA,
USA), and an ABSCIEX 4000 QTRAP mass spectrometer.
Following laboratory filtration with 0.7-mm GF/F filters,
samples were analyzed using direct injection without solid
phase extraction. All analytes were monitored using positive
ESI with MRM. A 100-mL sample loop (sample volume) was
used for each injection. Separation was performed on
a 150  4.6-mm Allure Biphenyl column with a 5-mm particle
size (Restek, Bellefonte, PA, USA) at a temperature of 40  C. A
0.1% formic acid in reagent water solution (A) and methanol
(B) gradient was used for LC mobile phases at a constant
800 mL/min flow rate. Full method details (with slight modifi-
cations) are described elsewhere (Trenholm and Snyder,
2011). The target compounds, MRLs, and MRM transitions
used for quantitation and confirmation are listed in Table S2
(Supplementary data). Fig. S3 (Supplementary data) is an
example chromatogram of a 500-ng/L calibration standard.
2.6. Mass loading rates and environmental discharge
calculations
Mass loading rates provide a valuable indication of the treat-
ment needs and environmental discharges from wastewater
treatment plants due to their incorporation of both flow rate
and contaminant concentrations. This study presents
temporal variations in mass loading rates (g/day) for influent
and effluent samples from the study site. The mass loading
rates were calculated as follows, where Ci is the 30-min
composite concentration for each sample and Qi is the
average flow rate over each 30-min sampling period:
Mass loading rateðg=dayÞ ¼ Ci  Qi
The effluent mass loading rates were actually estimates
because effluent samples were only collected from the prin-
cipal treatment train (z50% of the overall flow). Effluent
concentrations were not determined for the advanced water
treatment plant (z50% of the overall flow). However, coagu-
lation/flocculation and sedimentation are the only additional
processes at the advanced water treatment plant when
considering the water discharged to the wash (i.e., UV-
disinfected effluent). Coagulation/flocculation and sedimen-
tation have been shown to be highly ineffective for TOrC
removal so it is reasonable to assume the TOrC concentrations
in both effluents were similar (Westerhoff et al., 2005; Snyder
et al., 2007). As a result, the concentrations in the principal
treatment train effluent were extended to the entire flow to
provide an estimate of the effluent mass loading rate, or
environmental discharge rate. Total influent and effluent
mass loadings over the 12-h sampling period were calculated
as follows (with appropriate unit conversions for time):
X
24
 
Total mass loadingðgÞ ¼ ðMass loading rateÞi 30 min (2)
i¼1
3. Results and discussion
3.1. Wastewater flow data
Flow data from January 31 to February 1 (baseline), February 7
to 8 (Super Bowl), and March 7 to 8 (baseline) 2010 are provided
in Fig. 1 to illustrate the effects of Super Bowl tourism on
wastewater influent flow rates. Flow data from January 25 to
26 (baseline), February 1 to 2 (Super Bowl), and February 8 to 9
(baseline) 2009 are also provided to determine whether flow
phenomena are consistent between years. The graphs indi-
cate that the flow patterns for both baseline weekends (i.e.,
before and after the Super Bowl) were relatively consistent
between 2009 and 2010. The baseline flows fluctuated between
95 and 120 MGD between 3:00 PM and 11:00 PM and then
dropped to 60 MGD between 11:00 PM and 3:00 AM. In both
years, the flows on Super Bowl Sunday were consistently
higher (maximum differential of approximately 10 MGD)
between 3:00 PM and 5:00 PM, but they were also consistently
lower (maximum differential of approximately 10 MGD)
between 7:00 PM and 9:00 PM, as indicated by the boxes in
Fig. 1. Therefore, restroom use likely peaked immediately
before the game started and then reached a minimum
immediately after the game ended. Outside of these particular
time periods, the Super Bowl and baseline flow data were
relatively similar.
3.2. Pharmaceuticals and potential EDCs
3.2.1. Wastewater influent
Fig. 2 illustrates the influent mass loading rates for the target
pharmaceuticals and potential EDCs over the sampling
period. Atrazine was not detected in any of the samples at
a MRL of 10 ng/L so it was excluded from the results. Average
loading rates, standard deviations, and RSDs are also provided
in Table 1. Relative to the influent concentrations (Fig. S4 and
(1)
Table S3; Supplementary data), the influent mass loading
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1 5403

A 2009
140

120

Flow Rate (MGD) 100

80

60

40

20

2: M
3: M
M
1: M

2: M
12 AM

1: M
4: M

6: M
5: M

6: M

PM

PM

8: M
9: M
7: M

PM

10 P M
3: M

4: M

PM

10 PM
11 PM

11 PM
12 PM

A
A

A
A
A
P

P
P

P
P

P
P

P
00

00

00

30
30
00

30
00

00
30

30

30

00
30

0
0
0

30

00

30
00
00
0

0
:0

:3
:0
:3

:0

:3
7:

8:

9:
3:

5:

Time
Baseline Before Super Bowl Super Bowl Baseline After Super Bowl

2010
B 140

120

100
Flow Rate (MGD)

80

60

40

20

0
1: M

2: M

2: M
3: M
M
12 AM

1: M
8: M

9: M

9: M
10 PM
3: M

4: M

5: M

5: M

6: M

7: M

8: M
4: M

6: M

7: M

10 PM

12 PM
11 PM

11 PM

A
A
A
P

P
P

P
P

P
P

P
30

00

30

00

30
00

30

00

30
00

30

00

30

00

0
0

00

30

00

30
00
0
0
:0

:3

:0

:3
:0

:3
3:

Time
Baseline Before Super Bowl Super Bowl Baseline After Super Bowl

Fig. 1 e Influent wastewater flow data for the study site during the (A) 2009 and (B) 2010 Super Bowls. The boxes represent
flow periods during which the Super Bowl appears to have a discernable effect in comparison to the baseline weekends.

rates were slightly higher at the beginning of the sampling
period and slightly lower at the end of the sampling period due
to the change in flow rate. The steady decrease at the end of
the day was expected due to decreased water use (i.e., toilet
flushing) while many people were sleeping.
There was a distinct difference in the magnitudes of the
influent mass loading rates for the various compounds, and
some compounds experienced more dramatic fluctuations
than others on an absolute basis. In fact, the influent mass
loading rate for atenolol nearly doubled several times over the
two sampling weekends. DEET, primidone, sulfamethoxazole,
TCEP, and trimethoprim also varied greatly throughout the
sampling period. With respect to magnitude, the target
compounds could be classified as follows: atenolol was always
present at the highest influent mass loading rate (400e1100 g/
day); sulfamethoxazole, meprobamate, and trimethoprim
were grouped together in the middle (200e600 g/day); and
carbamazepine, phenytoin, primidone, DEET, and TCEP were
present at the lowest influent mass loading rates (0e200 g/
day). However, there were several instances where the
influent mass loading rates for phenytoin, DEET, and TCEP
spiked above 200 g/day. Furthermore, some compounds
spiked at similar times (e.g., atenolol and primidone during
the Super Bowl; atenolol, TCEP, and phenytoin during the
baseline weekend), but the trends were not entirely
consistent.
Nelson et al. (2011) reported their temporal variability in
terms of RSD. The RSDs for the current study are provided in
Table 1, and it is important to note that the temporal fluctu-
ation in flow rate accounted for 13e15% of the total variability
in the mass loading rates. The RSDs actually contradict the
visual observations in that the compounds with the highest
5404 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1

Fig. 2 e Influent loading rates of pharmaceuticals and EDCs during the (A) Super Bowl and (B) baseline weekends in 2010.

influent mass loading rates appeared to show the greatest
variability (e.g., atenolol and sulfamethoxazole), but it was
actually the compounds with the lowest influent mass loading
rates that had the highest RSDs (e.g., primidone, phenytoin,
and DEET). This also occurred in the Nelson et al. (2011) study
as the authors noted that higher baseline concentrations were
often linked to lower RSDs despite significant changes in
concentration. Therefore, RSD may be more appropriate for
TOrCs present at similar concentrations, while standard
deviation may be more appropriate for TOrCs with a wide
range of concentrations.
With respect to the special event effect, there was not
a distinct difference in influent mass loading rates for the
Super Bowl and baseline weekends. This is apparent in the
time series plots and the average influent mass loading rates
over the 12-h sampling period (Table 1). As shown in Fig. S8C
(Supplementary data), the influent mass loading rate for ate-
nolol was slightly elevated prior to 10:00 PM in the Super Bowl
samples, while the baseline weekend demonstrated higher
influent mass loading rates after 10:00 PM. Although atenolol
spiked several times during the Super Bowl period, there was
a similar spike for both weekends near 10:00 PM, which was
followed by a steady decline thereafter. The trends for the
other compounds were not as dramatic.
For compounds that are prescribed or administered by
medical personnel, the total wastewater loadings were
generally not affected by the Super Bowl, but the game may
have affected dosing schedules and their arrival times at the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1 5405

Table 1 e Summary of flow rate and mass loadings during sampling periods.
Compound Super Bowl Baseline
d
Influent Effluent Influent Effluentd

Average RSDb Total Average RSD Total Average RSD Total Average RSD Total
loading loading loading loading loading loading loading loading
rate (g/day)a (g)c rate (g/day) (g) rate (g/day) (g) rate (g/day) (g)

Pharmaceuticals and potential EDCs

Atenolol 713  247 35% 349 42  25 59% 21 700  152 22% 340 48  22 46% 23
Carbamazepine 38  16 43% 18 60  10 17% 29 37  12 33% 18 74  9 13% 36
DEET 63  52 82% 30 57  9 17% 28 71  38 53% 35 71  16 22% 35
Meprobamate 313  62 20% 153 139  39 28% 68 301  59 19% 146 94  21 22% 46
Phenytoin 30  10 33% 15 47  10 22% 23 37  57 156% 18 42  7 18% 20
Primidone 68  54 79% 33 51  9 18% 25 42  18 43% 20 112  21 19% 55
Sulfamethoxazole 348  68 19% 170 455  101 22% 222 414  90 22% 202 504  91 18% 246
TCEP <94e N/Ae <45e 157  42f 27% 77 <128e N/Ae <62e 118  24f 21% 57
Trimethoprim 248  54 22% 121 45  19 43% 22 262  45 17% 128 23  7 31% 11

Illicit drugs and metabolites (indented and italicized)

Methamphetamine 806132 16% 393 <18e N/Ae <9e 930  154 17% 451 <10e N/Ae <5e
Amphetamine 114  17 14% 56 <10e N/Ae <5e 125  23 18% 61 <10e N/Ae <5e
Cocaine 294  70 24% 142 <4e N/Ae <2e 295  47 16% 143 <4e N/Ae <2e
Ecgonine 271  52 19% 132 <19e N/Ae <9e 266  40 15% 129 <19e N/Ae <9e
Ecgonine methyl ester 161  27 16% 78 <10e N/Ae <5e 135  23 17% 66 <10e N/Ae <5e
Benzoylecgonine 718  142 20% 349 <10e N/Ae <5e 494  82 17% 240 <10e N/Ae <5e
Norcocaine 72 34% 3 <4e N/Ae <2e 72 28% 3 <4e N/Ae <2e
MDMA 106  30 28% 51 30  12 40% 15 97  43 44% 47 23  8 37% 11
MDA 17  3 17% 8 <19e N/Ae <9e 18  5 29% 9 <19e N/Ae <9e
Heroin <10e N/Ae <5e <10e N/Ae <5e <10e N/Ae <5e <10e N/Ae <5e
Morphine 231  112 48% 113 <19e N/Ae <9e 269  61 23% 131 <19e N/Ae <9e

Average RSD Total Average RSD Total Average RSD Total Average RSD Total
flow volume Flow volume flow volume flow volume
(MGD)a (MG) (MGD) (MG) (MGD) (MG) (MGD) (MG)
Flow rate 101  15 15% 49 101  15 15% 49 103  13 13% 50 103  13 13% 50

a Average of 24 sampling periods  one standard deviation.

b RSD ¼ relative standard deviation ¼ (one standard deviation)/(average).
c Total over 12-h sampling period.
d Estimated loading rates and loadings (see Section 2.5 for full explanation).
e Some or all samples were <MRL.
f All samples >MRL in effluent.

wastewater treatment plant. In particular, atenolol appears to 3.2.2. Wastewater effluent

be a relatively consistent outlier in terms of magnitude and
variability, which may be partially attributable to dosing
schedules. In contrast to the other prescribed pharmaceuti-
cals, which are administered in 2e4 doses throughout the day,
atenolol is administered in a single dose. Therefore, the large
atenolol spikes may have been attributable to initial metabo-
lism of the daily dose, while the other loads were related to
excretion of residual concentrations throughout the day.
Furthermore, some prescribed pharmaceuticals, including
atenolol, primidone, and phenytoin, spiked at similar times,
which might be indicative of large point source discharges
(e.g., hospitals). The spikes in TCEP might also be linked to
hospital wastewater (Stapleton et al., 2011) or industrial
manufacturing, but the available literature is insufficient to
develop explanations with confidence. Similar to Nelson et al.
(2011), this study was only intended to characterize the
temporal variability of certain compounds so the exact
reason(s) for the variability is unclear. A study focused on
pharmacokinetics and consumer behavior would be neces-
sary to develop more definitive explanations.
Fig. 3 provides estimated (see Section 2.5) effluent mass loading
rates for the target compounds during the Super Bowl and
baseline weekends. Average loading rates, standard devia-
tions, and RSDs are also provided in Table 1. Fig. S5 and Table
S3 (Supplementary data) provide actual effluent concentra-
tions. With the exception of sulfamethoxazole, Fig. S5 and
Fig. 3 indicate that nearly all of the target TOrCs were present
at less than 300 ng/L in the finished effluent, which corre-
sponds to an estimated effluent mass loading rate of <150 g/
day for each compound. The aforementioned discrepancy
between RSD and absolute variability was also apparent in the
effluent samples, particularly for sulfamethoxazole. Consid-
ering the special event effect, the Super Bowl effluent mass
loading rates were slightly higher for TCEP and meprobamate,
while the baseline effluent mass loading rates were slightly
higher for primidone and sulfamethoxazole.
Whether due to mixing or treatment, the treatment train
seemed to reduce the temporal variability for many of the
target compounds. However, the concentrations of some
compounds, particularly sulfamethoxazole, actually increased
5406 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1

Fig. 3 e Estimated effluent loading rates of pharmaceuticals and EDCs during the (A) Super Bowl and (B) baseline weekends in
2010.

through the treatment plant. For the more biologically recal-
citrant compounds, that may have been partially attributable
to sampling error. In addition, the removal of sulfamethoxazole
during secondary treatment is reported to be highly variable
(Suarez et al., 2010), and some studies indicate that sulfame-
thoxazole metabolites may cleave conjugated functional
groups during biological treatment and subsequently be
detected as the parent compound (Joss et al., 2005; Radjenovic
et al., 2009). This may provide a partial explanation for the
high effluent concentrations for sulfamethoxazole, which was
clearly an outlier compared to the other target compounds.
However, the steady decline in sulfamethoxazole concentra-
tions coupled with the decreasing flow rate resulted in a steep
decline in effluent mass loading rates over the sampling period.
With respect to TOrC treatment efficacy, the study site
currently employs one treatment processdactivated sludg-
edcapable of achieving substantial reductions in the
concentrations of certain compounds. This was apparent in
the effluent concentrations and mass loading rates as the
recalcitrant compounds (e.g., sulfamethoxazole, carbamaze-
pine, phenytoin, primidone, DEET, and TCEP) (Stevens-
Garmon et al., 2011, submitted for publication) remained
relatively stable through the treatment train, whereas the
compounds susceptible to either biotransformation or sorp-
tion (e.g., atenolol and trimethoprim) (Stevens-Garmon et al.,
2011, submitted for publication) experienced high removals.
The generally recalcitrant compound meprobamate (Stevens-
Garmon et al., 2011, submitted for publication) experienced
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1 5407

relatively high removals, but this particular compound still
had one of the higher effluent concentrations.
Therefore, biologically amenable compounds like atenolol
and trimethoprim are useful indicators of secondary treat-
ment efficacy, whereas recalcitrant compounds like mepro-
bamate, primidone, DEET, sulfamethoxazole, and
carbamazepine may be useful indicators of anthropogenic
contamination in the environment. The most appropriate
indicators of anthropogenic contamination may differ
between sites depending on each facility’s unit treatment
processes. If some form of oxidation is employed, the list of
environmental indicators would have to be altered (e.g.,
destruction of sulfamethoxazole with chlorination and car-
bamazepine with ozonation). Primidone (Guo and Krasner,
2009) and meprobamate appear to be conservative indicators
in nearly all applications, while compounds like DEET might
be affected by geographic and seasonal variability due to
climate and usage patterns.
Treatment efficacy, environmental discharges, and the
associated indicator concept are reflected in Table 1, which
provides total influent and effluent mass loadings over the
12-h sampling periods. A total influent loading is not
provided for TCEP because some samples were <MRL. All
other pharmaceuticals and potential EDCs had reportable
concentrations in the influent and effluent over the entire
sampling period so total loading estimates could be devel-
oped. As mentioned earlier, different unit treatment
processes, particularly those related to disinfection, will have

Fig. 4 e Influent loading rates of illicit drugs and their metabolites during the (A) Super Bowl and (B) baseline weekends in
2010.
5408 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1

a significant effect on effluent mass loadings or environ-
mental discharges. For example, the effluent mass loading
for sulfamethoxazole and trimethoprim could be reduced
dramatically with chlorination. The effluent mass loadings of
sulfamethoxazole, carbamazepine, trimethoprim, and, to
a lesser extent, atenolol, DEET, meprobamate, phenytoin, and
primidone could be reduced with ozonation. Cost effective
mitigation strategies for TCEP are still lacking.
3.3. Illicit drugs and metabolites
3.3.1. Wastewater influent
Fig. 4 illustrates the influent mass loading rate for each of the
illicit drugs and metabolites over the sampling period.
Average loading rates, standard deviations, and RSDs are also
provided in Table 1. Heroin, acetylmorphine, THC, and
hydroxy-THC were <MRL (i.e., 25, 25, 100, and 100 ng/L,
respectively) in all samples so they were excluded from the
results. According to Postigo et al. (2011), these particular
compounds and metabolites are characterized by extremely
low excretion rates (<3%) relative to the other illicit drugs.
Relative to the influent concentrations (Fig. S6 and Table S3;
Supplementary data), the influent mass loading rates were
slightly higher at the beginning of the sampling period and
slightly lower at the end of the sampling period due to the
change in flow rate. With the exception of methamphetamine,
BZE, and isolated spikes of morphine and cocaine, the illicit
drugs and metabolites were present at influent mass loading

Fig. 5 e Estimated effluent loading rates of illicit drugs and their metabolites during the (A) Super Bowl and (B) baseline
weekends in 2010.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
rates less than 400 g/day. The higher influent mass loading
rates for methamphetamine and BZE (400e1,100 g/day) may
be partially attributable to their higher excretion rates (43%
and 45%, respectively) (Postigo et al., 2011). Methamphet-
amine was similar to the pharmaceutical atenolol in that it
was the most prevalent compound (600e1200 g/day) and
experienced the most dramatic fluctuations in concentration
and influent mass loading rate.
It is also important to note that some metabolites are
consumed directly or are metabolites of other compounds,
specifically in the case of morphine (Postigo et al., 2011).
Despite being a major metabolite of heroin, morphine can also
be administered directly in medical facilities for pain mitiga-
tion. However, the spikes in morphine did not always coincide
with the dramatic spikes in atenolol, primidone, and
phenytoin, which may negate the aforementioned hospital
discharge theory. However, it is not possible to differentiate
illicit and prescribed morphine use in the current study.
Similar to the pharmaceuticals and potential EDCs, there
was no discernable difference between the two weekends for
most of the illicit drugs and metabolites. Of the 24 pharma-
ceuticals, potential EDCs, illicit drugs, and metabolites moni-
tored during this study, only the cocaine metabolite BZE
experienced a conspicuous difference in influent loading rates
between the two weekends. Influent mass loading rates for
BZE during the Super Bowl ranged from 500 to 1100 g/day with
an average of 718 g/day, whereas the influent mass loading
rates for the baseline weekend ranged from only 300 to 700 g/
day with an average of 494 g/day. Although not as apparent as
the difference for BZE, influent mass loading rates for meth-
amphetamine were slightly elevated in the baseline weekend
as compared to the Super Bowl. The trends for BZE and
methamphetamine are illustrated in Fig. S8A and B
(Supplementary data), respectively. A larger sample size
would be necessary to reach any definitive conclusions, but
the data suggest that drug abuse may shift from metham-
phetamine to cocaine during the Super Bowl.
3.3.2. Wastewater effluent
Fig. 5 provides estimated (see Section 2.5) effluent mass loading
rates for the illicit drugs and metabolites during the Super
Bowl and baseline weekends. Average loading rates, standard
deviations, and RSDs are also provided in Table 1. Fig. S7 and
Table S3 (Supplemental data) provide actual effluent concen-
trations. Methamphetamine and MDMA were the only illicit
drugs detected in the effluent samples, but methamphet-
amine was only present in a subset of those samples. The
Super Bowl effluent samples contained slightly higher
concentrations of methamphetamine and MDMA in compar-
ison to the baseline weekend, and both compounds demon-
strated a relatively constant decline in both effluent mass
loading rates and concentrations over the sampling period.
Methamphetamine and MDMA were detected at maximum
concentrations of 86 and 118 ng/L, respectively, in the Super
Bowl samples and 32 and 83 ng/L, respectively, in the baseline
samples. The corresponding maximum effluent mass loading
rates were 41 and 56 g/day, respectively, for the Super Bowl
samples and 14 and 36 g/day, respectively, for the baseline
samples. With respect to treatment efficacy, conventional
wastewater treatment was highly effective in removing the
5409
illicit drugs and metabolites based on the number of
compounds present at concentrations <MRL. Methamphet-
amine and MDMA were slightly more resistant to conven-
tional wastewater treatment, which may be related to their
similar structures (e.g., branched amines). Metcalfe et al.
(2010) reported similar removal trends for illicit drugs, and
their study also indicated that methamphetamine and MDMA
were slightly more resistant to secondary treatment than the
other target compounds.
Treatment efficacy and environmental discharges are
reflected in Table 1, which provides total influent and effluent
mass loadings over the 12-h sampling periods. Total influent
loadings for heroin and total effluent loadings for all of the
compounds except MDMA are not provided because some or
all of the samples were <MRL. Due to the extremely low
effluent loads to the environment, this set of target
compounds does not provide any suitable candidates for
indicators of anthropogenic contamination.
4. Conclusion
Temporal variations in wastewater flows are common
phenomena, but the corresponding effects on TOrCs are not
entirely understood. The intent of this study was to evaluate
high-resolution temporal variability in TOrC concentrations
and mass loading rates during a baseline weekend and
a special event. In support of the available literature, this
study suggests that temporal variability can be a significant
factor for TOrC mitigation efforts in that treatment designs
may be based on composite samples that do not accurately
characterize the discrete nature of wastewater matrices. In
other words, unit processes that target a specific effluent
concentration may consistently fail to achieve their goals if
the designs are based on composite samples where significant
spikes are attenuated.
With respect to the current study, the Super Bowl flows
were consistently higher than the baseline weekends during
the early part of the sampling period but consistently lower
during the middle part of the sampling period. Although there
were interesting temporal variations for some compounds,
particularly atenolol, the data did not indicate any significant
effect of the Super Bowl on the loadings or loading rates of
many of the compounds. The unique loading profile for ate-
nolol may have been affected by its once-daily dosing
schedule compared to the other prescription pharmaceuticals
that are dosed multiple times throughout the day. Based on
the BZE data, limited evidence suggests that cocaine use was
elevated during Super Bowl weekend as compared to the
baseline, whereas methamphetamine use was slightly lower.
It is also interesting to note that the influent loadings of
prescription pharmaceuticals were generally similar in
magnitude to those of the illicit drugs and metabolites.
However, conventional wastewater treatment was more
effective in removing the illicit drugs and metabolites targeted
in this study.
In order to more effectively evaluate the effects of holidays,
seasons, special events, and other unusual circumstances,
a more comprehensive database of “normal” temporal vari-
ability must be developed for a variety of compounds. These
5410 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
expanded studies should determine whether “normal”
temporal variability is consistent between sites for certain
compounds. These studies should also identify strategies to
predict temporal variability. For example, two studies sug-
gested that influent nitrogen or ammonia spikes might coin-
cide with spikes in prescription pharmaceuticals (Joss et al.,
2005; Nelson et al., 2011). Future studies should evaluate this
correlation as a potential modeling tool and determine
whether the correlation is suitable for compounds that are
directly related to consumption and excretion, such as
prescription pharmaceuticals, and compounds from other
sources, such as TCEP. Similarly, future studies should
determine whether over-the-counter medications with more
widespread use exhibit similar temporal variability to
prescription pharmaceuticals. Finally, the current data set
indicates that some compounds show definitive spikes
throughout the day, but additional studies are needed to link
these spikes to particular aspects of human behavior.
Acknowledgments
The authors would like to thank members of the Applied
Research and Development Center at the Southern Nevada
Water Authority, including Josephine Chu, Shannon Fergu-
son, Jasmine Koster, Roxanne Phillips, Janie Holady, Yongrui
Tan, and Brett Vanderford, for all of their efforts during this
study. The authors would also like to thank personnel at the
study site for their assistance with experimental design,
scheduling, and sampling efforts. Finally, the authors would
like to thank Dr. Christoph Ort from Eawag for reviewing the
study and providing valuable comments.
Appendix. Supplementary data
Supplementary data associated with this article can be found
in the online version, at doi:10.1016/j.watres.2011.07.020.
references
Daughton, C.G., Ternes, T.A., 1999. Pharmaceuticals and personal
care products in the environment: agents of subtle change?
Environ. Health Perspect. 107, 907e938.
Donn, J., Mendoza, M., Pritchard, J., 2008. Pharmaceuticals
Lurking in U.S. Drinking Water. Associated Press.
Enfinger, K.L., Stevens, P.L., 2011. Pit stop. Water Environ.
Technol. 23, 88.
Guo, Y.C., Krasner, S.W., 2009. Occurrence of primidone,
carbamazepine, caffeine, and precursors for N-
nitrosodimethylamine in drinking water sources impacted by
wastewater. J. Am. Water Resour. As. 45, 58e67.
Huerta-Fontela, M., Galceran, M.T., Ventura, F., 2008. Stimulatory
drugs of abuse in surface waters and their removal in
a conventional drinking water treatment plant. Environ. Sci.
Technol. 42, 6809e6816.
Joss, A., Keller, E., Alder, A.C., Gobel, A., McArdell, C.S., Ternes, T.,
Siegrist, H., 2005. Removal of pharmaceuticals and fragrances
in biological wastewater treatment. Water Res. 39, 3139e3152.
Lange, A., Paull, G.C., Coe, T.S., Katsu, Y., Urushitani, H., Iguchi, T.,
Tyler, C.R., 2009. Sexual reprogramming and estrogenic
sensitization in wild fish exposed to ethinylestradiol. Environ.
Sci. Technol. 43, 1219e1225.
Metcalfe, C., Tindale, K., Li, H., Rodayan, A., Yargeau, V., 2010.
Illicit drugs in Canadian municipal wastewater and estimates
of community drug use. Environ. Pollut 158, 3179e3185.
Nelson, E.D., Do, H., Lewis, R.S., Carr, S.A., 2011. Diurnal
variability of pharmaceutical, personal care product, estrogen
and alkylphenol concentrations in effluent from a tertiary
wastewater treatment facility. Environ. Sci. Technol. 45,
1228e1234.
Ort, C., Gujer, W., 2006. Sampling for representative micropollutant
loads in sewer systems. Water Sci. Technol. 54, 169e176.
Ort, C., Lawrence, M.G., Reungoat, J., Mueller, J.F., 2010. Sampling
for PPCPs in wastewater systems: comparison of different
sampling modes and optimization strategies. Environ. Sci.
Technol. 44, 6289e6296.
Ort, C., Schaffner, C., Giger, W., Gujer, W., 2005. Modeling
stochastic load variations in sewer systems. Water Sci.
Technol. 52, 113e122.
Plosz, B.G., Leknes, H., Liltved, H., Thomas, K.V., 2010. Diurnal
variations in the occurrence and the fate of hormones and
antibiotics in activated sludge wastewater treatment in Oslo,
Norway. Sci. Total Environ. 408, 1915e1924.
Postigo, C., Lopez de Alda, M., Barcelo, D., 2011. Evaluation of
drugs of abuse and trends in a prison through wastewater
analysis. Environ. Int. 37, 49e55.
Radjenovic, J., Petrovic, M., Barcelo, D., 2009. Fate and distribution
of pharmaceuticals in wastewater and sewage sludge of the
conventional activated sludge (CAS) and advanced membrane
bioreactor (MBR) treatment. Water Res. 43, 831e841.
Schriks, M., Heringa, M.B., van der Kooi, M.M.E., de Voogt, P., van
Wezel, A.P., 2010. Toxicological relevance of emerging
contaminants for drinking water quality. Water Res. 44, 461e476.
Snyder, E.M., Snyder, S.A., Kelly, K.L., Gross, T.S., Villeneuve, D.L.,
Fitzgerald, S.D., Villalobos, S.A., Giesy, J.P., 2004. Reproductive
responses of common carp (Cyprinus carpio) exposed in cages
to influent of the Las Vegas Wash in Lake Mead, Nevada, from
late winter to early spring. Environ. Sci. Technol. 38,
6385e6395.
Snyder, S., Vanderford, B., Pearson, R., Quinones, O., Yoon, Y.,
2003. Analytical methods used to measure endocrine
disrupting compounds in water. Pract. Periodical of Haz.,
Toxic, and Radioactive Waste Mgmt. 7, 224e234.
Snyder, S.A., Trenholm, R.A., Snyder, E.M., Bruce, G.M., Pleus, R.C.
, Hemming, J.D.C., 2008. Toxicological Relevance of EDCs and
Pharmaceuticals in Drinking Water. American Water Works
Association Research Foundation, IWA Publishing.
Snyder, S.A., Villeneuve, D.L., Snyder, E.M., Giesy, J.P., 2001.
Identification and quantification of estrogen receptor agonists
in wastewater effluents. Environ. Sci. Technol. 35, 3620e3625.
Snyder, S.A., Wert, E.C., Lei, H., Westerhoff, P., Yoon, Y., 2007.
Removal of EDCs and Pharmaceuticals in Drinking and Reuse
Treatment Processes. American Water Works Association
Research Foundation, IWA Publishing.
Stanford, B.D., Snyder, S.A., Trenholm, R.A., Holaday, J.C.,
Vanderford, B.J., 2010. Estrogenic activity of US drinking
waters: a relative exposure comparison. JAWWA 102, 55e65.
Stapleton, H.M., Klosterhaus, S., Keller, A., Ferguson, P.L., van
Bergen, S., Cooper, E., Webster, T.F., Blum, A., 2011.
Identification of flame retardants in polyurethane foam
collected from baby products. Environ. Sci. Technol. 45,
5323e5331.
Stevens-Garmon, J., Drewes, J., Khan, S., McDonald, J., Dickenson,
E. Biotransformation of emerging trace organic compounds by
wastewater activated sludge. Bioresour. Technol., submitted
for publication.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 3 9 9 e5 4 1 1
Stevens-Garmon, J., Drewes, J.E., Khan, S.J., McDonald, J.A.,
Dickenson, E.R., 2011. Sorption of emerging trace organic
compounds onto wastewater sludge solids. Water Res. 45,
3417e3426.
Suarez, S., Lema, J.M., Omil, F., 2010. Removal of pharmaceutical
and personal care products (PPCPs) under nitrifying and
denitrifying conditions. Water Res. 44, 3214e3224.
Takao, Y., Shimazu, M., Fukuda, M., Ishibashi, H., Nagae, M.,
Kohra, S., Tabira, Y., Ishibashi, Y., Arizono, K., 2008. Seasonal
and diurnal fluctuations in the concentrations of
pharmaceuticals and personal care products (PPCPs) in
residential sewage water. J. Health Sci. 54, 240e243.
Ternes, T.A., 1998. Occurrence of drugs in German sewage
treatment plants and rivers. Water Res. 32, 3245e3260.
Ternes, T.A., Meisenheimer, M., McDowell, D., Sacher, F.,
Brauch, H.J., Haist-Gulde, B., Preuss, G., Wilme, U., Zulei-
Seibert, N., 2002. Removal of pharmaceuticals during drinking
water treatment. Environ. Sci. Technol. 36, 3855e3863.
5411
Trenholm, R.A., Snyder, S.A., 2011. Analysis of illicit drugs in
water using direct-injection liquid chromatographyetandem
mass spectrometry. In: Castiglione, S., Zuccato, E., Fanelli, R.
(Eds.), Illicit Drugs in the Environment: Occurrence, Analysis,
and Fate Using Mass Spectrometry. John Wiley & Sons, Inc,
Hoboken.
Trenholm, R.A., Vanderford, B.J., Snyder, S.A., 2009. On-line solid
phase extraction LC-MS/MS analysis of pharmaceutical
indicators in water: A green alternative to conventional
methods. Talanta 79, 1425e1432.
Vanderford, B.J., Snyder, S.A., 2006. Analysis of
pharmaceuticals in water by isotope dilution liquid
chromatography/tandem mass spectrometry. Environ. Sci.
Technol. 40, 7312e7320.
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, 6649e6663.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fluid shear influences on the performance of hydraulic

flocculation systems

Ian C. Tse, Karen Swetland, Monroe L. Weber-Shirk*, Leonard W. Lion

School of Civil and Environmental Engineering, Cornell University, Ithaca, NY 14853, USA

article info abstract

Article history: Gravity driven hydraulic flocculators that operate in the absence of reliable electric power
Received 26 May 2011 are better suited to meet the water treatment needs of green communities, resource-poor
Received in revised form communities, and developing countries than conventional mechanical flocculators.
7 July 2011 However, current understanding regarding the proper design and operation of hydraulic
Accepted 29 July 2011 flocculation systems is insufficient. Of particular interest is the optimal fluid shear level
Available online 23 August 2011 needed to produce low turbidity water. A hydraulic tube flocculator was used to study how
fluid shear levels affect the settling properties of a flocculated alum-kaolin suspension. A
Keywords: Flocculation Residual Turbidity Analyzer (FReTA) was used to quantitatively compare the
Hydraulic flocculation sedimentation velocity distributions and the post-sedimentation residual turbidities of the
Fluid shear flocculated suspensions to see how they were affected by varying fluid shear, G, and
Sedimentation velocity hydraulic residence time, q, while holding collision potential, Gq, constant. Results show
Velocity gradient that floc breakup occurred at all velocity gradients evaluated. High floc settling velocities
were correlated with low residual turbidities, both of which were optimized at low fluid
shear levels and long fluid residence times. This study shows that, for hydraulic floccu-
lation systems under the conditions described in this paper, low turbidity water is
produced when fluid shear is kept at a minimum. Use of the product Gq for design of
laminar flow tube flocculators is insufficient if residual turbidity is used as the metric for
performance. At any Gq within the range tested in this study, best performance is obtained
when G is small and q is long.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction resources and a lack of reliable electric power prevent the

Surface water resources such as rivers and lakes contain
considerable amounts of colloidal matter which cannot easily
be removed because the parameters determining separation
performance such as particle size, concentration and surface
properties are often unfavorable for aggregation and sedi-
mentation. Sustainable treatment technologies that are both
economical and robust are needed in many resource-poor
communities where turbid surface waters are often not
treated prior to consumption, and where limited financial
implementation of conventional water treatment plants.
AguaClara, a program at Cornell University, has collaborated
with the NGO Agua Para el Pueblo of Honduras to design and
implement gravity powered water treatment plants in rural
communities and small cities. The process train used in these
plants includes a sequence of flocculation, up-flow sedimen-
tation, and chlorination (additional information about Agua-
Clara can be found at http://aguaclara.cee.cornell.edu). In
AguaClara treatment plants, conventional mechanical floc-
culators are replaced with hydraulic flocculators in which the

* Corresponding author. Tel.: þ1 607 255 8445; fax: þ1 607 255 9004.
E-mail address: mw24@cornell.edu (M.L. Weber-Shirk).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.07.040
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8
water is agitated by being routed around baffles rather than
mixed by motor driven paddles.
The goal of any flocculation process is to transform sus-
pended colloidal particles into flocs that can be removed by
sedimentation. The design of sedimentation tanks is dictated
by the settling velocity of the flocs; as floc capture requires that
the fluid residence time in a sedimentation tank or in lamellar
plate or tube settlers be greater than the time required for flocs
to settle to a surface. Therefore, one design goal of flocculators
is to produce flocs with sufficiently high sedimentation
velocities. Unfortunately, guidelines for proper design and
operation of hydraulic flocculators are incomplete. Conven-
tional designs utilize the product of the average velocity
gradient (G) and hydraulic residence time (q) as a measure of
the extent of flocculation in a reactor. In principle, any
combination of G and q that gives the same product should
work equally well (Ives, 1981). The appropriate fluid shear
levels (measured as the average energy dissipation rate, ε)
required at different points along a flocculator that will
produce the best flocs are not well understood. It is expected
that a high energy dissipation rate will increase the collision
frequency and hence create large floc aggregates quickly; but
on the other hand, a high energy dissipation rate will breakup
large flocs. In addition, higher ε may form denser flocs
(Gregory, 1998). This study evaluated the effect of ε and q on the
sedimentation velocity and residual turbidity of the resulting
floc suspension.
2. Theoretical considerations
Colloidal particles present in natural waters generally have
negatively charged surfaces causing inter-particle repulsion
that inhibits aggregation into larger particles which can be
removed by gravity. Coagulants are normally added to
enhance the kinetics for particle aggregation into flocs. When
a coagulant such as aluminum sulfate (alum) is added to
water, soluble positively charged hydrolysis species are
formed that adsorb onto the colloids. In addition, precipitation
of Al(OH)3(am) can occur on colloid surfaces and this solid
phase is positively charged at circumneutral pH values. The
neutralization of negative particle surface charge that ensues
after coagulant addition is typically reflected in rapid forma-
tion of colloid aggregates or flocs.
Contact of colloidal sized particles is primarily facilitated by
diffusion (known as perikinetic flocculation). Fully destabilized
particles aggregate as soon as they come into contact with one
another (Serra et al., 2008). As flocs grow, diffusion is replaced
by differential fluid velocities as the dominant particle to
particle transport mechanism in orthokinetic flocculation. It
has been shown that the frequency of particle collisions in
orthokinetic flocculation is related to the magnitude of the
energy dissipation rate, ε (Ives, 1981; Cleasby, 1984). As flocs
grow larger, they become more susceptible to breakup. Even-
tually the particle size distribution can reach a pseudo-steady
state during which breakup balances aggregation (Spicer and
Pratsinis, 1996).
In hydraulic flocculators, differential fluid velocities are
generated by the flow of water around baffled channels. Most
hydraulic flocculators have staggered baffles which form
5413
S-shaped channels that may wind horizontally or vertically.
The magnitude of the energy dissipation rate (and thus the
magnitude of fluid shear) can be controlled by adjusting either
the flow rate through the flocculator or the spacing between
the baffles. Since water treatment plants are designed to
operate within a target range of flow rates determined by the
needs of the communities they serve, varying the spacing
between baffles is the primary method of controlling the
magnitude of fluid shear in the flocculator. Hydraulic floccu-
lators have narrow, long flow passages and thus approach
plug flow. Floc growth occurs as the suspension moves
through the flocculator, and thus the extent of flocculation at
any position is a function of the local energy dissipation rate,
ε, and the hydraulic residence time, q, required to reach that
position.
In laminar flow, the collision potential, a measure of the
ability of the reactor to produce collisions, can be quantified by
the product of the mean velocity gradient (G) and q (Ives, 1981).
As noted below, the mean velocity gradient is related to ε. In
a comparison of two laminar flow flocculators where one has
a higher G value but shorter q, the collision potential should be
equal as long as the dimensionless Gq terms are identical. The
goal of this study was to evaluate the relative importance of G
and q in hydraulic flocculation.
3. Materials and methods
Experiments were conducted using an apparatus comprised of
synthetic raw water and coagulant metering systems, a coiled
tube hydraulic flocculator, and a flocculation residual turbidity
analyzer (FReTA) (see Fig. 1). Tse et al. (2011) provide a complete
description of the experimental apparatus and methods; only
flocculator length and flow rate have been changed for the data
presented here. The synthetic raw water (SRW) metering
system consisted of a concentrated stock suspension of
kaolinite clay (R.T. Vanderbilt Co., Inc., Norwalk, CT) mixed
with tap water to produce a feedback-regulated constant
turbidity raw water source. Tap water characteristics are: total
hardness z 150 mg/L as CaCO3, total alkalinity z 113 mg/L as
CaCO3, pH z 7.7 and dissolved organic carbon z 1.9 mg/L
(Bolton Point Municipal Water System, 2009). The concen-
trated stock and the SRW feedstock were each stirred by
a variable speed electric mixer to ensure homogeneous
suspensions. A float valve regulated the flow of temperature
controlled (25  C) tap water into the SRW tank to maintain
a constant water level. For all of the experiments performed in
this study, the SRW was maintained at a constant turbidity of
50  5 NTU, which corresponded to a clay concentration of
approximately 50 mg/L. Technical grade aluminum sulfate
(Al2(SO4)312-16H2O) from Fisher Scientific was used as the
coagulant for all experiments. The alum stock was prepared
with distilled water at a concentration of 200 mg/L as Al. Based
on initial settling experiments performed with a tube floccu-
lator at a G ¼ 40 s1 and Gq ¼ 19700,an alum dose of 3.1 mg/L Al
was determined by to be optimal for a SRW with a turbidity of
50 NTU. Both the SRW and the alum were metered with Cole
Parmer MasterFlex L/S digital computer controlled peristaltic
pumps.
5414 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8
clay alum
synthetic
raw water
NTU
turbidimeter
(feedback loop ) solenoid
valves
backwash tube flocculator
effluent
pressure sensor
(measure head loss)
Fig. 1 e Schematic of the experimental assembly.
SRW (turbidity ¼ 50 NTU) and the alum stock were
combined to give an alum concentration of 3.1 mg/L Al. The
mixture was passed through a rapid mix unit comprised of
a 120 cm segment of 4.3 mm (0.17”) ID tubing coiled around
a cylinder with an outer diameter of 5 cm to ensure thorough
mixing of the SRW and the alum. The mixed flow entered
a 28 m flocculator divided into 6 equal segments of 466 cm
each. The velocity gradient was varied from 40 to 250 s1by
varying the volumetric flow rate for each of the six lengths;
while the upper end of this range exceeds G values typically
used for flocculation, it was chosen to ensure that breakup by
floc shear would be observed. Gq values remained within
a non-overlapping range at each of the flocculator lengths,
because as G would increase q would decrease. In initial
experiments, a minimum G value of 40 s1 was shown to be
needed to keep lose flocs in suspension within the flocculator.
As Owen et al., (2008) note, flocculation is frequently
studied in batch reactors with offline size measurements for
aggregation processes, resulting in poor control over reaction
time and questionable size measurements. A tube flocculator
was used because it can be idealized as a high Peclet number
reactor much like a hydraulic flocculator and also because the
average velocity gradient (G) in laminar tube flow is well
defined (Equation (1)) (Gregory, 1981).
8Q
Gs ¼
3pr3
where: Q is the volumetric flow rate and r is the inner radius of
the tube.
The tube flocculator consisted of a 28 m segment of 9.5 mm
(3/8”)inner diameter transparent plastic tubing wrapped in
FReTA
peristaltic actuated
pumps ball valve
NTU infrared
turbidimeter
tight coil
rapid mix
settling column
effluent
discharge
solenoid
valve
high-pressure
backwash
a figure eight shape around two 11 cm outer diameter parallel
cylinders for structural support. The length of tubing was
chosen based on Camp and Stein’s (1943) recommendation
that a Gq of 20,000 is typically needed for sufficient floccula-
tion. The diameter of the tubing was chosen to be large rela-
tive to floc diameter and to reduce the fraction of the overall
residence time spent in the settling column. The settling
column in the turbidimeter has a 1” diameter and a length of
12”. A large diameter tube for flocculation requires a high flow
rate, which keeps the residence time in the settling column
low. A differential pressure sensor was attached at each end of
the tube flocculator to monitor head loss.
The flocculator was coiled to reduce the effects of sedi-
mentation. In laminar flows there are no turbulent eddies to
resuspend flocs that settle on the bottom surface of the floc-
culator tube. In coiled tubes fluid inertia causes secondary
flow patterns that consist of two vortical cells with the line of
symmetry being the radius of curvature of the coil (Berger
et al., 1983). The secondary flows helped resuspend flocs
because the secondary flows have a velocity component that
moves flocs away from the bottom of the tube. A simple
helical coil was observed to concentrate the flocs into two
zones that correspond to the two circulating cells. Therefore,
the coils were configured into a tight figure eight pattern to
disrupt the two circulating cells and cause particles to move
(1) throughout the cross-section of the tube.
The secondary circulation caused by coiling acted to
increase the magnitude of the average velocity gradient inside
the tube. Tse et al. (2011) provide a derivation of the velocity
gradient in a coiled tube, the result of which is shown in
Equation (2).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8 5415

1 normalized turbidity, and G(a) is the gamma function. The

4 2
Gc ¼ Gs 1 þ 0:033ðlogðDeÞÞ (2) mean Vs is equal to the product ab and the variance is equal to
the product ab2. A Gamma function was selected because it
where: the subscripts c and s refer to coiled and straight tubes,
pffiffiffiffiffiffiffiffiffi better accounted for skewed distributions when observed,
respectively. De is the Dean number: e ¼ r=Rc Red , r is the
however it is also capable of describing a log normal
inner radius of the tube, Rc is the radius of curvature,
Red ¼ Ud=v, U is the average axial velocity, d is the inner
diameter of the tube, and n is the kinematic viscosity of the A 50
fluid. Raw Turbidity (L axis)
Normalized & smoothed (R axis)
The flocculation residual turbidity analyzer (FReTA)
40
described by Tse et al. (2011)was used to measure both the

Normalized Turbidity
Turbidity (NTU)
sedimentation rate and the post-sedimentation residual 1

turbidity of the effluent from the coiled tube flocculator. 30

FReTA is capable of optically measuring floc sedimentation

velocities (Vs) and residual turbidity and does so without 20
altering or damaging the structure of the flocs in suspension. 0.5

FReTA consists of three primary components: an HF Scientific

10
MicroTOL 2 infrared turbidimeter, a transparent glass settling
column, and an electrically actuated ball valve. The valve is
situated atop the glass column that has been inserted through 0 0
0 500 1000 1500
the measurement chamber of the infrared turbidimeter. For
Settling Time (sec)
this study, a distance of 16 cm separated the bottom of the
closed ball valve and the middle of the 5 mm zone monitored B Normalized & Smoothed
by the turbidimeter. Any floc contained within this length of
Modified Gamma CDF
the settling column that settled past the measurement area
1
and crossed the beam of infrared light had its turbidity
detected.
Normalized Turbidity

4. Data analysis 0.5

Turbidity data was collected over 30 min at a 1 Hz sampling

frequency for each flocculated suspension. Turbidity fluctu-
ations were observed when large flocs (high sedimentation
velocities) moved past the measurement area and refracted 0
0.01 0.1 1 10 100
more light into the light sensor. Cheng et al. (2010) confirm the
Settling Time (sec)
correlation between turbidity standard deviation and floc
size. The large fluctuations were often problematic for data C Modified gamma PDF (L axis)
fitting routines and required smoothing. The data was first 0.8
Modified gamma CDF (R axis)
Normalized Turbidity/(mm/s)

1
averaged over 9 s intervals and then a filter that reported the

Normalized Turbidity
moving median value over a set of 5 averaged 9 s intervals
was used to smooth the data. This smoothing technique acted 0.6

to exclude extreme fluctuations while preserving the shape

characteristics of the Vs and particle size distributions. The
0.4
data was normalized to range between 0 and 1 by dividing 0.5

turbidity values by the initial turbidity of the settling period.

Normalized turbidity, recorded as a function of time, was 0.2
converted into a settling velocity, Vs, by dividing the length of
the settling column (16 cm) by the elapsed time. The resulting
normalized turbidity vs. Vs curves were then fit with the 0 0
0.01 0.1 1 10 100
following modified gamma distribution’s cumulative distri- Vs (mm/s)
bution function (CDF) in order to determine the mean and
variance of the data: Fig. 2 e A: raw time series turbidity data obtained from
FReTA and the same data set after being normalized and
Zx median smoothed. B: normalized turbidity-Vs curve on
ex=b
F0ðx; a; b; gÞ ¼ ð1  gÞ xa1 dx þ g (3)
ab GðaÞ a semi-log graph, and cumulative distribution function
0
fitted to the data using Equation (3). C: modified cumulative
where: the independent variable is a base 10 logarithm of Vs, distribution function and probability distribution function
a and b are parameters of the distribution that are fitted, g is for the modified Gamma distribution, where the data has
an offset parameter that accounts for the non-zero residual mean Vs [ 0.94 mm/s.
5416 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8

distribution that is often attributed to particle suspensions

Table 1 e Gq range for each flocculator length used in this
(Aksoy, 2000). Curve fitting was performed using the genfit study.
function in Mathsoft’s MathCAD 14.0. Genfit is capable of
Length 2796 cm 2330 cm 1864 cm 1398 cm 932 cm
fitting a user defined equation to a set of data points using
LevenbergeMarquardt method for minimization. High Gq 26,000 21,600 17,300 13,000 8700
Low Gq 19,700 16,400 13,100 9800 6600

5. Results
Fig. 2A shows an example of a typical data set obtained using
FReTA before and after smoothing and normalization. Fig. 2B
shows the smoothed and normalized data transformed into
normalized turbidity as a function of settling velocity and the
modified gamma distribution fit of this data. Fig. 2C shows the
cumulative distribution function (CDF) and the resulting
probability density function (PDF) of settling velocities
obtained from the fitted gamma distribution to the data
shown in Fig. 2A. The area under each PDF curve in Fig. 3 is
unity as expected; however, this is visually obscured by the
semi-log plot used to display the PDFs.
The sedimentation velocity distributions and residual
turbidities of flocs formed inside several tube flocculators (see
Table 1) were measured with FReTA. For each of the tube
flocculator lengths tested, the mean sedimentation velocities
were obtained from the PDFs of the gamma distribution fits.
Fig. 3 shows an exemplary set of PDFs obtained from the
modified gamma distribution fits of experiments performed in
a flocculator with Gq ¼ 19,000.
Fig. 4 shows the family of Vs values obtained over the range
of Gq values tested as a function of G. Each curve in Fig. 4
represents a set of flocculation experiments performed
under conditions of almost identical collision potential (or Gq).
If Gq accurately predicted the extent of flocculation and if floc
breakup were not significant, the curves in Fig. 4 should not
vary with G, but only be affected by the magnitude of the
product Gq. However, each of the plots in Fig. 4 has
a significant negative slope indicating that flocs were able to
grow to a larger size at lower G values as evidenced by the
larger Vs. These results suggest that floc breakup was
a significant factor in limiting floc size for each of the five tube
flocculator lengths and for all of the velocity gradients that
were tested. Results from Fig. 4 exhibit similarities to the
findings of Serra et al. (2008), who showed that shear-induced
breakup limited the size of latex flocs formed in various types
of reactors when G was greater than 30 s1.
As expected, floc sedimentation velocities generally
increased with Gq (see Fig. 4), indicating that many of the flocs
continued to grow when given more time to flocculate.
However, floc size did not increase when additional floccula-
tion time was provided for G > 100 s1 at Gq  15,000, as shown
by the convergence of the curves for Gq equal to 15,000, 19,000
and 23,000. Shear-induced breakup prevented further growth
at high G values. Floc size was independent of Gq for G > 200 s1
and Gq > 11,000 suggesting that for these conditions floc
growth was limited by floc breakup. A similar convergence was
observed for G > 100 s1 for Gq  15,000. The curve corre-
sponding to Gq ¼ 7500 did not converge with any of the higher
Gq curves, indicating that flocs produced under those condi-
tions had not yet reached a breakup limited size. Nevertheless,
floc breakup may have retarded floc growth at Gq ¼ 7500.
Fig. 5 shows the observed residual turbidity at a sedimen-
tation capture velocity of 0.09 mm/s as a function of G. Data
sets correspond to experiments with nearly constant Gq.
Residual turbidity increased with higher fluid shear. Thus,
increased fluid shear not only decreased the average size and
sedimentation velocity of flocs as shown above, but it resulted

G = 40 /s
G = 52 /s 3
G = 103 /s G = 23000
Normalized turbidity/(mm/s)

2
G = 175 /s G = 19000
G = 237 /s G = 15000
G = 11000
Mean Vs (mm/s)

2 G = 7500

0
0.01 0.1 1 10
Vs (mm/s)
0
0 100 200 300
Fig. 3 e Modified gamma distribution PDFs of floc
G (1/s)
sedimentation velocities from experiments performed
using a tube flocculator with an average Gq of 19,000. Mean Fig. 4 e Mean sedimentation velocities plotted vs. average
and variance statistics for each data set were calculated velocity gradients for various flocculator lengths (listed
from similar PDF curves. with their mean Gq values).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8 5417

100
G = 23000
Residual Turbidity (% of initial turbidity)

G = 23000

Residual Turbidity (% of initial turbidity)

30 G = 19000 G = 19000
G = 15000 G = 15000
G = 11000 G = 11000
G = 7500 G = 7500
20

10

10

0 1
0 100 200 300 1 10 100
G (1/sec)
(mW/kg)
Fig. 5 e Residual turbidity corresponding to a capture
Fig. 7 e Residual turbidity after a 30 min settling period
velocity of 0.09 mm/s (16 cm height in 30 min of
(given as percentage of the turbidity at the start of each
sedimentation and given as percentage of the turbidity at
settling period) vs. energy dissipation rate for various
the start of each settling period) vs. average velocity
flocculator lengths (listed with their mean Gq values).
gradient for various flocculator lengths (listed with their
mean Gq values).

rffiffiffi
in higher residual turbidities as well. The observed relation- ε
G¼ (4)
v
ships could be explained by two distinct hypotheses: (1) the
breakup of flocs due to fluid shear released smaller floc frag- where n is the kinematic viscosity.
ments with poor settling properties, or (2) larger flocs that are Consistent with prior reports, a power law relationship
created under low fluid shear may be more effective in between sedimentation velocity and the energy dissipation
sweeping up smaller particles. rate is evident for hydraulic flocculators after this analysis.
Fig. 5 also shows that use of the product Gq for design of The curves corresponding to the three largest flocculator
laminar flow tube flocculators is insufficient if residual lengths are superimposed on top of one another in Fig. 6 at
turbidity is used as the metric for performance. Over the range ε  3 m W/kg. Thus, increasing Gq appears to have no effect on
of Gq studied, best performance is obtained when G is small mean floc sedimentation velocity for a given ε once breakup
and q is long. from shear limits the steady state floc size.
The velocity gradient (G) term used in describing the Residual turbidity data from Fig. 5 are plotted against ε in
magnitude of fluid shear in laminar flow tube flow was con- Fig. 7. Best results for residual turbidity require a combination
verted into energy dissipation rate (ε) using Equation (4) and of low ε and long q. The more efficient removal of colloids at
the resulting data is shown in Figs. 6 and 7. low ε could be due to the creation of larger flocs that have
a higher sedimentation velocity which would increase their
10 ability to collide with other colloids. Larger flocs also have
G = 23000 a more porous structure as manifested by a lower fractal
G = 19000 dimension (Li et al., 2007), which could make them effective in
G = 15000 filtering other colloids.
G = 11000 In Figs. 4e7 it is apparent that the Gq ¼ 7500 plot is unique
G = 7500 since it does not converge to the same values at high energy
Vs (mm/s)

dissipation rates or high velocity gradients. The difference

1 may be that, at Gq of 7500, floc growth was affected but not yet
completely limited by shear-induced breakup. The negative
slope in Fig. 4 suggests that breakup was occurring. Floc
growth was also still occurring at all measured velocity
gradients because larger flocs appeared by a Gq of 11,000.

0.1
1 10 100
6. Conclusions
(mW/kg)

Fig. 6 e Mean sedimentation velocities are plotted against The results shown above demonstrate that increased collision
the energy dissipation rate. potential (increasing Gq from 11,000 to 23,000) did not increase
5418 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 2 e5 4 1 8
mean particle settling velocities when the velocity gradient
was greater than 200 s1. Increased collision potential
(increasing Gq from 11,000 to 23,000) also did not reduce the
residual turbidity when the velocity gradient was at 250 s1.
Both quickly-settling flocs and low turbidity water are
best produced when the energy dissipation rates are low. If
the best way to produce low turbidity water is to produce
very large flocs, baffle spacing and the flow conduits between
the flocculator and the sedimentation tank must be designed
to ensure they do not breakup large flocs. Since large flocs
settle very quickly, it is difficult to reduce the energy dissi-
pation rate sufficiently to reduce their breakup while still
maintaining flow velocities high enough to prevent sedi-
mentation in the flow conduits. Thus, the flocculator to
sedimentation tank transition may ultimately set a practical
limit to the goal of creating quickly-settling flocs in
flocculation.
The data presented in this report were obtained under
conditions of laminar tube flow. In practice, most hydraulic
flocculators operate under conditions of turbulent flow. It is
reasonable to expect that turbulent eddies promote orthoki-
netic flocculation and floc breakup in a manner that is similar
to the effect of fluid shear in laminar flow. While similar trends
are expected with respect to the effect of energy dissipation
rate and residence time on floc characteristics under turbulent
flow conditions, the temporal and spatial variability in the
energy dissipation rate in turbulent flow is expected to influ-
ence the specific relationships that are observed.
Acknowledgments
This work was made possible through the generous support of
the Sanjuan Foundation. Partial support was also obtained
from the National Science Foundation under Grant CBET-
0604566. Additional information on AguaClara can be found
at: http://aguaclara.cee.cornell.edu
references
Aksoy, H., 2000. Use of gamma distribution in hydrological
analysis. Turkish Journal of Engineering and Environmental
Science 24 (2000), 419e428.
Berger, S.A., Talbot, L., Yao, L., 1983. Flow in curved pipes. Annual
Review of Fluid Mechanics 15, 461e512.
Bolton Point Municipal Water System, 2009. City of Ithaca Water
System, Cornell University Water System. Drinking Water
Quality Report.
Camp, T.R., Stein, P.C., 1943. Velocity gradients and internal work
in fluid motion. Journal of the Boston Society of Civil Engineers
30, 219.
Cheng, W., Chang, J., Chen, P., Ruey Yu, R., Huang, Y., Hsieh, Y.,
2010. Monitoring floc formation to achieve optimal
flocculation in water treatment plants. Environmental
Engineering Science 28 (6), 523e530.
Cleasby, J., 1984. Is velocity gradient a valid turbulent flocculation
parameter? Journal of Environmental Engineering 110 (5),
875e897.
Gregory, J., 1998. The role of floc density in solid-liquid
separation,". Filtration and Separation 35 (4), 366e371.
Gregory, J., 1981. Flocculation in laminar tube flow. Chemical
Engineering Science 36 (11), 1789e1796.
Ives, K., 1981. Orthokinetic flocculation. In: Svarovsky, L. (Ed.),
Solid-Liquid Separation. Butterworths, London, pp. 86e119.
Li, T., Zhu, Z., Wang, D., Yao, C., Tang, H., 2007. The strength and
fractal dimension characteristics of alumekaolin flocs.
International Journal of Mineral Processing 82 (1), 23e29.
Owen, A.T., Fawell, P.D., Swift, J.D., Labbett, D.M., Benn, F.A.,
Farrow, J.B., 2008. Using turbulent pipe flow to study the factors
affecting polymer-bridging flocculation of mineral systems.
International Journal of Mineral Processing 87 (3e4), 90e99.
Serra, T., Colomer, J., Logan, B.E., 2008. Efficiency of different
shear devices on flocculation. Water Research 42 (4e5), 1113.
Spicer, P.T., Pratsinis, S.E., 1996. Shear-induced flocculation: the
evolution of floc structure and the shape of the size distribution
at steady state,". Water Research 30 (5), 1049e1056.
Tse, I.C., Weber-Shirk, M.L., Lion, L.W., 2011. Method for
quantitative analysis of flocculation performance. Water
Research 45 (10), 3075e3084.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Quantifying quagga mussel veliger abundance and

distribution in Copper Basin Reservoir (California) using
acoustic backscatter

Michael A. Anderson a,*, William D. Taylor b

a
Department of Environmental Sciences, University of California, Riverside, CA 92521, USA
b
Metropolitan Water District of Southern California, 700 Moreno Ave, La Verne, CA 91750, USA

article info abstract

Article history: Quagga mussels (Dreissena bugensis) have been linked to oligotrophication of lakes, alter-
Received 14 May 2011 ation of aquatic food webs, and fouling of infrastructure associated with water supply and
Received in revised form power generation, causing potentially billions of dollars in direct and indirect damages.
29 July 2011 Understanding their abundance and distribution is key in slowing their advance, assessing
Accepted 1 August 2011 their potential impacts, and evaluating effectiveness of control strategies. Volume back-
Available online 27 August 2011 scatter strength (Sv) measurements at 201- and 430-kHz were compared with quagga
mussel veliger and zooplankton abundances determined from samples collected using
Keywords: a Wisconsin closing net from the Copper Basin Reservoir on the Colorado River Aqueduct.
Quagga mussel The plankton within the lower portion of the water column (>18 m depth) was strongly
Veliger dominated by D-shaped quagga mussel veligers, comprising up to 95e99% of the
Distribution community, and allowed direct empirical measurement of their mean backscattering
Hydroacoustics cross-section. The upper 0e18 m of the water column contained a smaller relative
Backscatter proportion of veligers based upon net sampling. The difference in mean volume back-
scatter strength at these two frequencies was found to decrease with decreasing
zooplankton abundance (r2 ¼ 0.94), allowing for correction of Sv due to the contribution of
zooplankton and the determination of veliger abundance in the reservoir. Hydroacoustic
measurements revealed veligers were often present at high abundances (up to 100e200 ind
L1) in a thin 1e2 m layer at the thermocline, with considerable patchiness in their
distribution observed along a 700 m transect on the reservoir. Under suitable conditions,
hydroacoustic measurements can rapidly provide detailed information on the abundance
and distribution of quagga mussel veligers over large areas with high horizontal and
vertical resolution.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction power generation, causing hundreds of millions of dollars of

Dreissenid mussels have been linked to oligotrophication of
Lake Michigan and Lake Huron (Evans et al., 2011), alteration
of aquatic food webs (MacIsaac et al., 1995; Wong et al., 2003),
and fouling of infrastructure associated with water supply and
economic costs to these facilities through 2004 (Connelly et al.,
2007). Economic loss in the U.S. due to the invasion by quagga
and zebra mussels is thought to be as high as $1B per year
(Pimentel et al., 2005). As a result, considerable interest exists
in the U.S., Canada and elsewhere concerning the distribution,

* Corresponding author. Tel.: þ1 951 827 3757; fax: þ1 951 827 3993.
E-mail address: michael.anderson@ucr.edu (M.A. Anderson).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.018
5420 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7

fate and control of quagga and zebra mussels. Quagga mussels the distribution and densities of zooplankton not possible
(Dreissena bugensis) have recently invaded the western U.S., from net sampling (Hembre and Megard, 2003), although such
being first identified in Lake Mead in January 2007 (LaBounty methods do not provide direct information about species
and Roefer, 2007), and have quickly spread throughout the composition within a diverse community. Acoustic back-
lower Colorado River, Southern California, Arizona and scatter by zooplankton is strongly dependent upon acoustic
Nevada (Wong et al., 2010). Their occurrence in the Colorado frequency, organism size, shape, and other factors, with
River system thus threaten a complex and vast conveyance stronger backscatter generally found at higher frequencies.
system that delivers water to 7 states and has a hydroelectric Multifrequency backscatter measurements can, under some
generating capacity exceeding 4 GW. circumstances, provide information about mean size(s) and
Quagga and zebra mussels are r-strategists, with life- type(s) of zooplankton, however (Greenlaw, 1979; Greene
history characteristics promoting rapid population growth et al., 1989; Mitson et al., 1996).
(Vanderploeg et al., 2002). They accomplish rapid population Hydroacoustic measurements were conducted at Copper
growth through broadcast spawning of very large numbers of Basin Reservoir to test the ability of the technique to (i) provide
eggs and sperms; a single female can release up to a million remotely-sensed population estimates of quagga mussel
eggs to the water column that can then be externally fertilized veligers, (ii) provide fine-scale spatial information about their
to form trochophore larvae (Ackerman et al., 1994). The distribution, and (iii) compare day and night distributions of
trochophore larvae quickly develop to the planktonic larval veligers (and other zooplankton) in the reservoir.
veliger stage that secrete an unadorned D-shaped CaCO3 shell
that are generally 70e160 mm in height. Within 7e9 days post-
fertilization, a second more ornamented shell is secreted from
the mantle tissue that has a more pronounced umbonal 2. Materials and methods
region near the hinges and is round or clam-like in profile.
This umbonal veliger is somewhat larger in size (120e280 mm) 2.1. Field sampling and measurements
and is the last free-swimming veliger stage routinely found in
the plankton, although a developmental bottleneck signifi- Coupled net sampling and volume backscattering strength
cantly limits the number D-shaped dreissenid veligers that measurements were conducted on August 10-11, 2009 at
develop into umbonal veligers (Schneider et al., 2003). Thus Copper Basin Reservoir (Fig. 1). Copper Basin Reservoir is
the D-shaped veligers are the dominant larval form of quagga a 172 ha reservoir on the Colorado River Aqueduct and located
and zebra mussels present in invaded lakes. near the California-Arizona border (34.2876 N, 114.2331 W)
Sampling using a Wisconsin closing net revealed a strong (Fig. 1). The reservoir has a mean depth of 17.5 m, high Secchi
vertical gradient in quagga mussel veliger abundance in
Copper Basin Reservoir on the Colorado River Aqueduct near
Lake Havasu and the California-Arizona border (Reid et al.,
2010). Veligers dominated the plankton community in the
lake, with very high abundances present near the thermocline
(Reid et al., 2010). Sampling was restricted to 3 locations on the
lake, with samples collected at 3-6 discrete depth intervals,
however, thus providing information on a coarse vertical and
lateral scale.
The abundance and distribution of zooplankton in lakes
vary widely in space and time, with patchy distributions and
large gradients in populations over very short length scales
(Hembre and Megard, 2003; Rinke et al., 2009), marked diel
migration for many species (Masson et al., 2001), and fine-
scale heterogeneities within communities due to fish preda-
tion and other factors (Benoit-Bird, 2009a). Much of the
detailed information about the distribution of zooplankton
has been determined through use of hydroacoustic
measurements (Holliday and Pieper, 1995; Hembre and
Megard, 2003). Sound is reflected off objects within the water
column based upon their size, shape, and density and
soundspeed contrasts with water. As a result, hydroacoustic
measurements are commonly used in fishery assessments
(Simmonds and MacLennan, 2006; Godlewska et al., 2009),
although acoustic backscatter is also used to quantify abun-
dance and distribution of zooplankton (Hembre and Megard,
2003), larval aquatic insects (Kubecka et al., 2000), and Fig. 1 e Study site: Copper Basin Reservoir on the Colorado
submerged aquatic vegetation (Winfield et al., 2007). River Aqueduct near Lake Havasu, Bathymetric map
Volume backscattering strength measured using hydro- showing sampling site for net and hydroacoustic
acoustic methods has provided detailed information about measurements, and survey transect.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7
depth (9.4 m), low mean chlorophyll concentration (5 mg L1)
and short hydraulic detention time (about 6e10 d depending
upon flow) (Reid et al., 2010). The Colorado River Aqueduct is
operated by the Metropolitan Water District of Southern Cal-
ifornia and is one of the primary sources of drinking water for
18M residents in Southern California.
Following Reid et al. (2010), a 0.3 m diameter, 63 mm mesh
Wisconsin closing net was used to collect 6 m-interval
samples from the surface to 36 m depth at the deepest loca-
tion on the lake (Zmaxw38 m) (Fig. 1). Sampling was conducted
at approximately 9:30 p.m. on August 10th (about 1 h after
sunset), 12 h later the following morning (about 9:30 a.m.,
August 11th), and later that afternoon at approximately 1:30
p.m. Each depth interval represented a 424 L sample of water;
total plankton (zooplankton and veligers) retained on the net
were quantitatively transferred by rinsing with 70% ethanol
into 125-mL wide mouth polypropylene bottles. Samples were
stored on ice until returned to the laboratory.
Hydroacoustic measurements were made immediately
prior to net sampling with a BioSonics DT-X Echosounder
multiplexed to a 430-kHz 7 single-beam transducer with an
orientation sensor and a 201-kHz 6.6 split-beam transducer
(Table 1). Both transducers were mounted to a common swivel
mount secured to the bow of a 17-ft Boston Whaler and low-
ered 0.6 m below the water surface. A JRC 212W real-time
differential global-positioning satellite (DGPS) receiver was
mounted directly over the transducers and recorded
differentially-corrected positions every 1-s. Data were
acquired using BioSonics VisualAcquisition software on a Dell
ATG laptop at a rate of 5 pps and 0.4 ms pulse duration (Table
1). The 430- and 201-kHz transducers were calibrated prior to
data collection using 17 and 33 mm tungsten-carbide cali-
bration spheres, respectively, with known target strengths.
Temperature, pH, conductivity and dissolved oxygen (DO)
profiles were measured using a Hydrolab DataSonde4a and
Surveyor 4 to quantify thermal structure, availability of DO,
and basic chemical properties in the water column. Temper-
ature, conductivity and pH values were also used in sound
speed and range (depth) calculations.
2.2. Laboratory measurements and data analyses
Veliger enumeration was conducted using a Nikon E600
compound microscope fitted with cross-polarization and
Whipple grid micrometer. One-mL samples were removed
from gently inverted, mixed samples of known volume and
placed on a gridded Sedgewick-Rafter counting cell. The
sample was first analyzed for veligers under cross-
polarization, and then inspected under phase-contrast or
Table 1 e Transducer configurations used in this study.
Property DTX-201 DTX-420
Frequency (kHz) 201 430
Beam angle ( ) 6.6 7.0
Source level (dB/mPa) 221.3 220.0
Receive sensitivity (dBC/mPa) 63.6 62.9
Pulse length (ms) 0.4 0.4
Pings per second (pps) 5 5
5421
brightfield for identification of other zooplankton. Size was
determined on at least 10 randomly selected individuals from
each taxa or group when available.
Volume backscattering strength (Sv) was determined from
echograms using echointegration with a 20logR time-varied
gain after manual removal of fish echoes using Sonar5Pro
(Balk and Lindem, 2007). Echogram analyses were conducted
by averaging across layers corresponding to net-sampled
depth intervals, except for the 0e6 m depth interval, where
the acoustically-sampled depth interval was restricted to
1.5e6 m depth as a result of the transducer depth and near-
field effects.
Volume backscatter strength is a function of the abun-
dance of scatterers and their backscattering cross-section
(Hembre and Megard, 2003). That is,
X
n
Sv ¼ 10log sbs;i Ni (1)
i¼1
where sbs,i is the backscattering cross-section of species i and
N is the abundance of species i (Benoit-Bird, 2009b). Back-
scattering cross-sections have been empirically measured
(Hembre and Megard, 2003) or theoretically calculated from
scattering models (e.g., Greenlaw, 1979; Stanton, 1989) for
a limited number of zooplankton. The target strength (TS, in
dB) of a scatterer can in turn be calculated from sbs as (Hembre
and Megard, 2003):
TS ¼ 10logsbs (2)
3. Results
Total plankton (zooplankton and veligers) exhibited a strong
vertical gradient in abundance during the 9:30 p.m. sampling
on August 10, 2009 (Fig. 2a). Quagga mussel veligers were
numerically the dominant plankton below 6 m depth and
comprised 70e99% of the total invertebrate plankton recov-
ered in the deeper net samples, while the surface (0e6 m)
interval exhibited low abundances of all taxa (Fig. 2a). The
non-veliger zooplankton community was quite modest in
density in the reservoir, reflecting both low productivity
(chlorophyll concentrations approximately 5 mg L1) (Reid
et al., 2010), and the hydraulics of the system. That is,
Copper Basin Reservoir has a short hydraulic detention (6 d
when operated at the full flow capacity of the Colorado River
Aqueduct), with water pumped up about 100 m from Gene
Wash Reservoir through the pumps, surge chambers and
penstocks of the Gene Pumping Plant. Inspection of
zooplankton (chiefly Bosmina, small adult copepods and
nauplii) at the inlet and outfall of the pumps indicated >90%
mortality associated with pumping (unpubl. data). The short
hydraulic residence coupled with the lack of viable
zooplankton delivered with inflows is thought to be a chief
reason for the low abundance of zooplankton in the reservoir.
In contrast, mortality to veligers, as determined from ciliar or
internal organelle motion, was low (<10% mortality) (unpubl.
data).
Mean volumetric backscatter strength measured at 201-
kHz (Sv201) followed trends in overall plankton abundance
(Fig. 2b). Sv201 of about 101 dB was measured in the surface
5422 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7

a Concentration (ind L-1) b Sv201 (dB re 1 m-1) c Sv430 (dB re 1 m-1)

0 20 40 60 80 -102 -100 -98 -96 -89 -88 -87 -86 -85
0

Veligers
Bosmina
6 Cyclopoid
Calanoid
Nauplii
Rotifers
12
Depth (m)

18

24

30

36

Fig. 2 e Results from measurements made at approximately 9:00 p.m. on August 10, 2009: a) zooplankton and quagga
mussel veliger abundances from net sampling, b) volume backscatter strength at 201-kHz, and c) volume backscatter
strength at 430-kHz.

layer and corresponded to low overall plankton abundance, a general increase in Sv201 with depth and veliger abundance
while Sv201 increased to a maximum value ofe96.9 dB at the to a maximum value of 96.5 dB in the depth interval imme-
24e30 m depth interval (Fig. 2b) that corresponded to a total diately above the thermocline (Fig. 4b). Backscatter strength at
plankton abundance of 67 ind L1 (Fig. 2a). Veligers comprised 430-kHz yielded a somewhat different distribution with depth
>97% of the plankton community at that depth. Abundance than found the previous evening, however, with Sv430
and Sv201 both declined below that depth at the deepest decreasing from a maximum value of 85.6 dB in the surface
sampled layer in the profile. The mean volumetric backscatter layer to 87.6 dB in the 12e18 m depth range (Fig. 4c). The
strength at 430-kHz (Sv430) was uniformly high at approxi- Sv430 values did not follow veliger abundance quite as well as
mately 87.5 dB over the 12e30 m depth range, and lower both Sv201, and reflected stronger contributions from zooplankton.
near the surface (0e12 m) and lowest depth interval (30e36 m) These results suggest that Sv at 201-kHz offers some potential
(Fig. 2c). utility in estimating veliger abundance, although significant
Temperature profile measurements at that time revealed overestimates would be expected for surface samples or
the presence of stratification, with a deep thermocline located elsewhere where veligers comprise a minority of the total
at approximately 30 m (Fig. 3a). Dissolved oxygen (DO) plankton community.
concentrations were high throughout the epilimnion, with
reductions across the metalimnion and hypolimnion (Fig. 3b). a Temperature (oC) b Dissolved O2 (mg L-1)
Sampling was repeated the following morning to inspect 16 20 24 28 0 2 4 6 8 10
vertical distribution at that time and further test use of the 0
hydroacoustic method to probe veliger distribution in the lake.
Veligers once again dominated the plankton community 6
deeper in the water column, with maximum abundance (65
ind L1) again present from 24 to 30 m depth (Fig. 4a). The
overall trend in veliger abundance with depth was broadly 12
similar to that found 12 h earlier (Fig. 2a), although higher
Depth (m)

numbers of zooplankton were present in the upper water

18
column at the sampling site in the morning than found the
previous night (Fig. 4a). The relative composition of the
plankton community was broadly similar in the upper 12 m 24
for the two sampling times, with about 10% Bosmina (about
300 mm in length) and 20% small (300e400 mm) adult copepods,
although the morning sampling did have a larger fraction of 30
nauplii (100e150 mm) and smaller proportion of veligers.
Veligers in all samples were predominantly D-shaped, with
36
lengths generally 80e90 mm.
Volume backscatter strength measurements over these Fig. 3 e Vertical profiles of a) temperature and b) dissolved
same depth intervals yielded similar trends at 201-kHz, with oxygen at Copper Basin Reservoir on August 11, 2009.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7
Determination of veliger abundance from Sv requires
information about their backscattering cross-section (sbs) (eq
(1)). Values for sbs for the veligers and zooplankton in
Copper Basin Reservoir are not known, although the very high
abundance of veligers relative to other plankton, especially
near the thermocline (Figs. 2a and 4a), suggest that they would
dominate the backscatter at those depths. The mean back-
scattering cross-section for quagga mussel veligers was thus
calculated from Sv for those depth intervals where they were
present at high concentrations and comprised 95e99% of the
total plankton populations to minimize contributions to
backscatter from other zooplankton. Rearranging eq (1) and
solving for sbs, we calculated a value of 3.66  0.73  1015 m2.
The empirically-derived sbs value was further used to
calculate, from eq (2), a target strength of 144.4  0.8 dB for
a D-shaped quagga mussel veliger at 201-kHz. In a similar way,
we calculate at 430-kHz the sbs and TS to be
4.16  1.28  1014 m2 and e133.9 dB  1.3, respectively.
Using this information, we can estimate veliger concen-
trations for the deeper parts of the water column where veli-
gers dominated the plankton community, although
contributions to backscatter from zooplankton would result in
significant overpredictions of veliger abundance in the upper
water column (e.g., Fig. 4a). Since the acoustical response of an
organism is a complex function of its size, shape, density and
soundspeed contrasts with water, and other factors, as well as
the frequency of the soundwave (Stanton, 1989), differential
frequency response has been used in some studies to resolve
different scatterers (Mitson et al., 1996; Godlewska et al., 2009).
The presence of several different groups of zooplankton make
it impossible in this study to resolve the contribution to Sv for
each group of zooplankton, although it is relevant to note that
the difference in volume backscatter strength at 430- and 201-
kHz (Sv430e201) varied linearly with log zooplankton abun-
dance (dashed line, R2 ¼ 0.94) over the 0e18 m depth interval
for the 2 sampling events (Fig. 5, solid symbols). Strong
a Concentration (ind L-1)
0 20 40 60
0
Veligers
Bosmina
6 Cyclopoid
Calanoid
Nauplii
Rotifers
12
Depth (m)
18
24
30
36
Fig. 4 e Results from measurements made at approximately 9:00 a.m. on August 11, 2009: a) zooplankton and quagga
mussel veliger abundances from net sampling, b) volume backscatter strength at 201-kHz, and c) volume backscatter
strength at 430-kHz.
5423
deviation from this linear trend was found at lower depths
when zooplankton abundances were low and veligers domi-
nated the total plankton community (Fig. 5, open symbols).
The empirical regression line from Fig. 5 was thus used to
correct Sv201 for the contribution from (non-veliger)
zooplankton. That is, the log zooplankton concentration in
each depth interval was estimated from the difference in
mean volume backscatter strength (Sv430e201) and, based upon
an ensemble-average backscattering cross-section calculated
for zooplankton from Sv measured at 201-kHz
(1.26  1014 m2), used to correct Sv201 for the contributions
from zooplankton. The mean backscattering coefficient for
veligers at 201-kHz (3.66  1015 m2) was then used to estimate
the veliger concentrations from corrected Sv201 values over
the entire depth range (Fig. 6). Fig. 6c shows predicted and
observed veliger concentrations for measurements collected
from Copper Basin Reservoir at approximately 1:30 p.m. on
August 11, 2009 (this data set was not included in any of the
prior figures or calculations). One sees generally fair agree-
ment between predicted concentrations (Fig. 6, dashed line)
and observed concentrations (Fig. 6, solid symbols) of veligers.
This approach yielded an average absolute error in predicted
veliger abundance of 9.9 ind L1 (mean relative error of 37.6%
for veliger concentrations greater than 5 ind L1). Relative
error was smaller when veliger concentrations were high and
zooplankton abundances were low (e.g., mean relative error
dropped to 16.8% when veliger concentrations exceeded 20
ind L1 and zooplankton abundances were less than 2 ind L1).
The net sampling results shown in Fig. 6 were necessarily
integrated over finite depth intervals and thus potentially
mask finer-scale heterogeneities within the water column.
Moreover, even with the relatively large net used (0.3 m
diameter opening) the net samples a very small cross-
sectional area of the water column. Hydroacoustic measure-
ments can provide far greater vertical resolution and rapidly
sample a much greater area and volume of the lake. For
b Sv201 (dB re 1 m-1) c Sv430 (dB re 1 m-1)
80 -102 -100 -98 -96 -89 -88 -87 -86 -85
5424 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7

2 veliger concentrations with depth in three adjacent profiles

(Fig. 7). Similar to findings from net sampling made earlier in
log Zooplankton Concentration (ind L-1)

the day (Fig. 4a), the highest abundances of zooplankton were

1.5
1
0.5
0
-0.5
-1
10
Sv430-201 (dB re 1 m-1)
Fig. 5 e Log zooplankton abundance (excluding veligers)
vs. the mean difference in volume backscatter strength at
430- and 201-kHz (Sv430e201).
example, dual frequency echograms from a short 40 m section
of survey collected about 2 p.m. on August 11, 2009 were
binned into 1 m  1 m cells and Sv values corrected for the
contribution of fish by removal of echoes with greater than
75 dB target strength using the cross-filter detector for noise
reduction in Sonar5. The difference in mean volume back-
scatter strength (Sv430e201) was used to estimate the abun-
dance of zooplankton in the water column, while Sv201 values
after correction for zooplankton were used with the veliger
scattering cross-section at this acoustic frequency to estimate
present in the upper 12 m of the water column, with generally
very few zooplankton below this depth (Fig. 7). Veliger abun-
dances were very low in the uppermost 6 m of water and
present at a maximum concentration at 30 m (Fig. 7), which
coincides with the depth of the thermocline (Fig. 3a).
The hydroacoustic measurements reveal that this layer of
veligers is quite thin, on the order of 1e2 m thick. Some lateral
variability was also seen in the three adjacent profiles (each
representing the average of 5 pings) (Fig. 7), although as
a result of beam spreading, the ensonified volumes increas-
Depth 0-18 m ingly overlap with depth. Lateral and vertical heterogeneities
Depth 18-36 m in veliger abundances from a 700 m transect from Copper
Basin Reservoir on August 10, 2009 (Fig. 1) can be seen in Fig. 8.
Veliger concentrations generally increased with depth, with
12 14 16
highest values frequently near the thermocline, although
substantial variability was present at that depth as well as
elsewhere. Consistently high concentrations were found
where the thermocline intercepts the bottom sediments
(about 300 m on transect, Fig. 8); this suggests that internal
waves and local resuspension may play some role in high
observed veliger concentrations there. Previous studies have
shown that settlement rates of quagga mussel veligers were
extremely low below the thermocline (Mueting et al., 2010)
and water velocity is critical in determining quagga mussel
veliger settlement (Chen et al., 2011).Very high concentrations
were also present near the surface extending down to about
10 m depth at approximately 420 m from the start of the
transect, and also above the lake bottom at approximately
670 m on the transect that may represent local spawning
inputs, aggregation by mixing, settling, transport or other
processes (Fig. 8). Veliger abundance was thus found to be
strongly heterogeneous in both lateral and vertical directions.

Veliger Concentration (ind L-1)

0 20 40 60 80 0 20 40 60 80 0 20 40 60 80
0
a b c
6

12
Depth (m)

18

24

30

36

Fig. 6 e Measured (solid symbols) and predicted (dashed lines) veliger concentrations in Copper Basin Reservoir on: a)
August 10, 2009, 9 p.m.; b) August 11, 2009, 9 a.m.; and c) August 11, 2009, 1:30 p.m.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7 5425

Veliger Concentration (ind L-1)

0 50 100 150 200 0 50 100 150 200 0 50 100 150 200
0
a b c
6

Zooplankton
(lower x-axis)
12
Depth (m)

18

24
Veligers
(upper x-axis)
30

36
0 10 20 30 40 50 0 10 20 30 40 50 0 10 20 30 40 50
Zooplankton Concentration (ind L-1)

Fig. 7 e Vertical distribution of a) zooplankton and b) quagga mussel veligers estimated from volume backscatter strength
measurements on Copper Basin Reservoir made on August 11, 2009, 2:00 p.m. Each panel represents the average of 5
consecutive pings on a transect at a survey speed of 1 m sL1.

or more frequency measurements are necessary when size

4. Discussion
Volume backscatter strength has been used in a number of
previous studies to quantify the abundance and distribution
of zooplankton in lakes (e.g., Hembre and Megard, 2003;
Holbrook et al., 2006). Single-frequency measurements can
provide reasonable estimates under certain conditions where
a single known size/target strength organism dominates the
acoustic backscatter (Holliday and Pieper, 1995), although two
and abundance are not known (Mitson et al., 1996; Lavery
et al., 2007). Multiple frequencies can also help separate fish
from other scatterers (Jurvelius et al., 2008). Empirical
measurements of backscattering cross-section have been
made (e.g., Hembre and Megard, 2003), as well as scattering
models developed that include size, shape and density and
sound speed contrasts with water (Stanton et al., 1994; Lavery
et al., 2007).

Fig. 8 e Distribution of veligers determined from corrected volume backscatter strength measurements across a 700 m
transect on Copper Basin Reservoir on August 10, 2009, 10 p.m.
5426 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7
This study represents the first using hydroacoustic
methods to infer the abundance and distribution of quagga (or
zebra) mussel veligers in an invaded lake. Veligers are smaller
than most other zooplankton that have been studied and as
a result are relatively weak scatters of sound. The effective
diameter (d) of a typical D-shaped veliger is estimated to be
about 60 mm, based upon the volume calculated for a scalene
ellipsoid that is 85 mm  65 mm x 40 mm in size. As a result,
veligers are Rayleigh scatterers at all typical echosounder
frequencies, and thus scatter sound at an intensity that is
proportional to (d/l)4. The wavelengths at the 2 frequencies
used in this investigation (430-kHz and 201-kHz) are 3.49 and
7.46 mm, respectively; as a result, we expect the veligers to
scatter sound at an intensity that is about 20 greater at 430-
kHz than at 201-kHz, although the empirically-derived back-
scattering cross-section was about 10 greater. Nonlinear
frequency response in backscatter has been seen in several
other studies and is thought to result from resonance or other
factors (Benoit-Bird, 2009b). Perhaps more interesting, the
ensemble-average backscattering cross-section for the
somewhat larger zooplankton present in the reservoir that
included adult and nauplii copepods and Bosmina, was infer-
red to be only 3e4 greater than that of D-shaped veligers at
201-kHz (1.24  1014 vs. 3.66  1015 m2, respectively). At
effective diameters near 100e200 mm for the copepods and
Bosmina present, one would expect backscattering cross-
sections 10e100 greater. The dense CaCO3 shell of the
quagga mussel veligers provide a significant density contrast
with water, and is thus thought to help make veligers
comparatively strong scatterers of sound for their size.
Stanton et al. (1994) previously noted that the hard aragonite
shell of marine gastropods was responsible for their much
higher echo energy per unit biomass when compared with
fluid-like zooplankton.
The difference in mean volume backscatter strength was
found to be strongly correlated with log zooplankton
concentration in the upper (0e18 m) water column (r2 ¼ 0.94),
although a multiple linear regression of log zooplankton
concentration with Sv201 and Sv430 yielded a similarly strong
correlation (r2 ¼ 0.96). The response in Sv across these 2
frequencies thus provided a convenient way to estimate
zooplankton abundance in this reservoir and also correct for
their contributions to overall acoustic backscatter. Changes in
difference in mean backscatter strength were also found to
track changes in composition of micronekton (Benoit-Bird,
2009b).
Our findings support the previous observations of Reid
et al. (2010) that quagga mussel veligers are often most
abundant at the thermocline, with the hydroacoustic
measurements reported here indicating their accumulation to
high concentrations within a thin layer that can be 1e2 m
thick (Fig. 7). At the same time, strong patchiness is present
(Fig. 8), as previously found in high resolution measurements
of other zooplankton communities (e.g., Hembre and Megard,
2003). The patchiness is presumed to reflect spatially and
temporally heterogeneous spawning inputs, turbulent hori-
zontal and vertical transport, and complex ecological factors.
This patchiness is also thought to account for some of the
deviations between measured and predicted veliger concen-
trations shown in Fig. 6. That is, the 0.3 m diameter Wisconsin
net necessarily sampled a different volume of water than the
acoustic beam owing to the time required to complete the
plankton tows, spreading of the acoustic beam, and related
factors (Hembre and Megard, 2003).
Additional work is nonetheless needed to better under-
stand the acoustical properties of quagga mussel veligers and
measurement of their abundance and distribution using
hydroacoustic techniques. Copper Basin Reservoir provided
favorable conditions for acoustical measurements owing to
the very large abundance of veligers in the lower portion of the
water column with few other scatterers present (often
comprising 95e99% of the total plankton there). The small size
and low overall abundances of zooplankton also provided
favorable conditions for acoustical measurements, with
a relatively simple way to correct for their contribution to
overall Sv. Additional studies are planned for other reservoirs
on the Colorado River Aqueduct where veliger abundances are
somewhat lower, and populations of other larger zooplankton
(e.g., Daphnia) are much higher. Measurements that include
volume backscatter strength at a third frequency will further
define the acoustical response of quagga mussel veligers.
Notwithstanding, results of this study indicate that
hydroacoustic measurements can offer a way to rapidly esti-
mate veliger abundance and distribution over large areas with
high horizontal and vertical resolution in invaded lakes and
reservoirs. Such information can increase our understanding
of their ecology, relationships to other zooplankton, effects on
food webs, and potential impacts to infrastructure associated
with water supply and power generation in invaded systems
such as the Colorado River system in the western USA.
5. Conclusions
1. Volume backscatter strengths measured at 430- and 201-
kHz were related to abundances of quagga mussel veli-
gers and zooplankton in Copper Basin Reservoir.
2. The difference in mean volume backscatter strength at 430-
and 201-kHz was strongly correlated with log zooplankton
abundance and allowed for correction of volume back-
scatter strength and determination of veliger abundance in
the water column.
3. Hydroacoustic measurements indicate that high abun-
dances of veligers were often present within a thin 1e2 m
layer at the thermocline.
4. Veliger distribution was nonetheless strongly heteroge-
neous in both the horizontal and vertical directions, and
thought to reflect spatially and temporally heterogeneous
spawning and other inputs, and complex transport, life-
history and settling processes.
Acknowledgments
This research was conducted under agreement no. 95835
between the University of California-Riverside and the
Metropolitan Water District of Southern California. Special
thanks to Michelle Tobin for her assistance in the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 1 9 e5 4 2 7
identification and enumeration of zooplankton in the net
samples. A warm thanks also for Jim Nafsey and the staff at
Gene Camp for their assistance and hospitality.
references
Ackerman, J.D., Sim, B., Nichols, S.J., Claudi, R., 1994. A review of
the early life history of zebra mussels (Dreissena polymorpha):
comparisons with marine bivalves. Canadian Journal of
Zoology 72, 1169e1179.
Balk, H., Lindem, T., 2007. Sonar4 and Sonar5-Pro post processing
systems operator manual version 5.9.8. Lindem Data
Acquisition, Oslo, 455 pp.
Benoit-Bird, K.J., 2009a. Dynamic 3-dimensional structure of thin
zooplankton layers is impacted by foraging by fish. Marine
Ecology Progress Series 396, 61e76.
Benoit-Bird, K.J., 2009b. The effects of scattering-layer
composition, animal size, and numerical density on the
frequency response of volume backscatter. ICES Journal of
Marine Science 66, 582e593.
Chen, D., Gerstenberger, S.L., Mueting, S.A., Wong, W.H., 2011.
Environmental factors affecting settlement of quagga mussel
(Dreissena bugensis) veligers in Lake Mead, Nevada-Arizona,
USA. Aquatic Invasions 6, 149e156.
Connelly, N.A., O’Neill, C.R., Knuth, B.A., Brown, T.L., 2007.
Economic impacts of zebra mussels on drinking water
treatment and electric power generation facilities.
Environmental Management 40, 105e112.
Evans, M.A., Fahnenstiel, G., Scavia, D., 2011. Incidental
oligotrophication of North American Great lakes.
Environmental Science and Technology 45, 3297e3303.
Godlewska, M., Colon, M., Doroszczyk, L., Dlugoszewski, B.,
Verges, C., Guillard, J., 2009. Hydroacoustic measurements at
two frequencies: 70 and 120 kHz e consequences for fish stock
estimation. Fisheries Research 96, 11e16.
Greene, C.H., Wiebe, P.H., Burczynski, J., 1989. Analyzing
zooplankton size distributions using high-frequency sound.
Limnology and Oceanography 34, 129e139.
Greenlaw, C.F., 1979. Acoustical estimation of zooplankton
populations. Limnology and Oceanography 24, 226e242.
Hembre, L.K., Megard, R.O., 2003. Seasonal and diel patchiness of
a Daphnia population: an acoustic analysis. Limnology and
Oceanography 48, 2221e2233.
Holbrook, B.V., Hrabik, T.R., Branstrator, D.K., Yule, D.L.,
Stockwell, J.D., 2006. Hydroacoustical estimation of
zooplankton biomass at two shoal complexes in the Apostle
Islands region of Lake Superior. Journal of Great Lakes
Research 32, 680e696.
Holliday, D.V., Pieper, R.E., 1995. Bioacoustical oceanography at
high frequencies. ICES Journal of Marine Science 52, 279e296.
Jurvelius, J., Knudsen, F.R., Balk, H., Marjomaki, T.J., Peltonen, H.,
Taskinen, J., Tuomaala, A., Viljanen, M., 2008. Echo-sounding
can discriminate between fish and macroinvertebrates in
fresh water. Freshwater Biology 53, 912e923.
Kubecka, J., Frouzova, J., Cech, M., Peterka, J., Ketelaars, H.A.M.,
Wagenwoort, A.J., Papacek, M., 2000. Hydroacoustic
assessment of pelagic stages of freshwater insects. Aquatic
Living Resources 13, 361e366.
LaBounty, J.F., Roefer, P., 2007. Quagga mussels invade Lake Mead.
Lake Line 27, 17e22.
Lavery, A.C., Wiebe, P.H., Stanton, T.K., Lawson, G.L., Benfield, M.
C., Copley, N., 2007. Determining dominant scatterers of
5427
sound in mixed zooplankton populations. Journal of the
Acoustical Society of America 122, 3304e3326.
MacIsaac, H.J., Lonnee, C.J., Leach, J.H., 1995. Suppression of
microzooplankton by zebra mussels. Importance of mussel
size. Freshwater Biology 34, 379e387.
Masson, S., Angeli, N., Guillard, J., Pinel-Alloul, B., 2001. Diel
vertical and horizontal distribution of crustacean zooplankton
and young of the year fish in a sub-alpine lake: an approach
based upon high frequency sampling. Journal of Plankton
Research 23, 1041e1060.
Mitson, R.B., Simard, Y., Goss, C., 1996. Use of a two-frequency
algorithm to determine size and abundance of plankton in
three widely spaced locations. ICES Journal of Marine Science
53, 209e215.
Mueting, S.A., Gerstenberger, S.L., Wong, W.H., 2010. An
evaluation of artificial substrates for monitoring quagga
mussel (Dreissena bugensis) in Lake Mead, NV-AZ. Lake and
Reservoir Management 26, 283e292.
Pimentel, D., Zuniga, R., Morrison, D., 2005. Update on the
environmental and economic costs associated with alien-
invasive species in the United States. Ecological Economics 52,
273e288.
Reid, N.J., Anderson, M.A., Taylor, W.D., 2010. Distribution of
quagga mussel veligers, Dreissena bugensis, in the reservoirs
of the Colorado River Aqueduct. Lake and Reservoir
Management 26, 328e335.
Rinke, K., Huber, A.M.R., Kempke, S., Eder, M., Wolf, T., Probst, W.
N., Rothhaupt, K.O., 2009. Lake-wide distribution of
temperature, phytoplankton, zooplankton, and fish in the
pelagic zone of a large lake. Limnology and Oceanography 54,
1306e1322.
Schneider, D.W., Stoeckel, J.A., Rehmann, C.R., Blodgett, K.D.,
Sparks, R.E., Padilla, D.K., 2003. A developmental bottleneck in
dispersing larvae: implications for spatial population
dynamics. Ecology Letters 6, 352e360.
Simmonds, E.J., MacLennan, D.N., 2006. Fisheries Acoustics:
Theory and Practice, second ed.. Blackwell, Oxford, 437 pp.
Stanton, T.K., 1989. Simple approximate formulas for
backscattering of sound by spherical and elongated bodies.
Journal of the Acoustical Society of America 86, 1499e1510.
Stanton, T.K., Weibe, P.H., Chu, D., Benfield, M.C., Scanlon, L.,
Martin, L., Eastwood, R.L., 1994. On acoustic estimates of
zooplankton biomass. ICES Journal of Marine Science 51,
505e512.
Vanderploeg, H.A., Nalepa, T.F., Jude, D.J., Mills, E.L., Holeck, K.T.,
Liebig, J.R., Grigorovich, I.A., Ojaveer, H., 2002. Dispersal and
emerging ecological impacts of Ponto-Caspian species in the
Laurentian Great lakes. Canadian Journal of Fisheries and
Aquatic Sciences 59, 1209e1228.
Winfield, I.J., Onoufriou, C., O’Connel, M.J., Godlewska, M.,
Ward, R.M., Brown, A.F., Yallop, M.L., 2007. Assessment in two
shallow lakes of a hydroacoustic system for surveying aquatic
macrophytes. Hydrobiology 584, 111e119.
Wong, W.H., Levinton, J.S., Twining, B.S., Fisher, N., 2003.
Assimilation of micro- and mesozooplankton by zebra
mussels: a demonstration of the food web link between
zooplankton and benthic suspension feeders. Limnology and
Oceanography 48, 308e312.
Wong, W.H., Tietjen, T., Gerstenberger, S., Holdren, G.C.,
Mueting, S., Loomis, E., Roefer, P., Moore, B., Turner, K.,
Hannoun, I., 2010. Potential ecological consequences of
invasion of the quagga mussel (Dreissena bugensis) into Lake
Mead, NevadaeArizona. Lake Reservoir Management 26,
306e315.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Seawater pretreatment for reverse osmosis: Chemistry,

contaminants, and coagulation

James K. Edzwald a,*, Johannes Haarhoff b

a
Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, MA 01003-9293, USA
b
Department of Civil Engineering Science, University of Johannesburg, Box 524, Auckland Park, 2006 Johannesburg, South Africa

article info abstract

Article history: The paper addresses the effects of salinity and temperature on the chemistry of
Received 8 May 2011 important parameters affecting coagulation pretreatment including the ion product of
Received in revised form water, acid-base chemistry, dissolved metal speciation, and precipitation reactions for
11 July 2011 aluminum and iron coagulants. The ion product of seawater is greater than for fresh-
Accepted 1 August 2011 waters and affects chemical hydrolysis and metal-hydroxide solubility reactions. Inor-
Available online 16 August 2011 ganic carbon is the main cause of seawater alkalinity and buffer intensity but borate
BðOHÞ1
4 also contributes. Buffer intensity is an important parameter in assessing coag-
Keywords: ulation pH adjustment. Mineral particles are relatively unstable in seawater from elec-
Algae trical double layer compression, and when present these particles are easily coagulated.
Seawater chemistry Algal-particle stability is affected by steric effects and algal motility. Dissolved natural
Desalination organic matter from algae and humic substances causes fouling of RO membranes and
Coagulation pretreatment removal is essential. Aluminum coagulants are not recommended, and not
Membrane fouling used, because they are too soluble in seawater. Ferric coagulants are preferred and used.
Particles The equilibrium solubility of Fe with amorphous ferric hydroxide in seawater is low over
Pretreatment a wide range of pH and temperature conditions. Ferric chloride dosing guidelines are
Reverse osmosis presented for various raw seawater quality characteristics. The effect of pH on coagulant
dose and the role of buffer intensity are addressed. A dual coagulation strategy is rec-
ommended for treating seawater with moderate to high concentrations of algae or
seawater with humic matter. This involves a low and constant dose with high charge-
density cationic polymers using Fe as the main coagulant where it is varied in
response to raw water quality changes.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction separation. The particle separation processes can consist of (1)

It is essential to have pretreatment prior to reverse osmosis
(RO) desalination (Voutchkov, 2010a, 2010b). Pretreatment
serves to remove contaminants that can foul RO membranes,
and the integration of pretreatment with RO makes for more
efficient plant performance with respect to water quality and
energy usage. There are numerous pretreatment configura-
tions. All involve coagulation, flocculation, and particle
direct filtration with granular media, (2) sedimentation and
granular media filtration, (3) sedimentation and low-pressure
membrane filtration, (4) dissolved air flotation (DAF) and
granular media filtration, or (5) DAF and low-pressure
membrane filtration. Many desalination plants are faced
with removing algae in pretreatment, and some plants face
the problem of dealing with harmful algae (red and brown-
tide algae).

* Corresponding author. 4 Hillcrest Dr., Potsdam, NY 13676, USA. Tel.: þ1 315 261 4186.
E-mail addresses: Edzwald@ecs.umass.edu (J.K. Edzwald), jhaarhoff@uj.ac.za (J. Haarhoff).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.014
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0 5429

Seawater is high in total dissolved solids (TDS) and often pretreatment as summarized in Table 1. Some of these are
expressed as salinity (S ) in ppt on a mass basis (or & for ppt). discussed next while others are discussed under Section 3 on
The salinity of the open ocean is about 35 ppt (g/kg) but varies contaminants.
from about 31 to 45 ppt. Higher salinity seawater is found in
the Mediterranean Sea, Red Sea, and the Persian Gulf, while
2.1. pH and the dissociation of water
lower salinities are found near the mouths of rivers. High
temperatures of 35  C, and even slightly greater, occur for
The pH of seawater is mainly between 8.1 and 8.3 but varies
seawater in the Middle East.
from about 7.5 to 8.3 (Table 1). It affects dissolved chemical
This paper does not address flocculation and particle
speciation, other chemical reactions, coagulation, other
separation processes. Rather, it addresses the effects of
pretreatment processes, and RO desalination. Because of the
seawater chemistry and contaminants on coagulation of
high ionic strength of seawater, the dissociation of water or
seawater. The purposes of the paper follow: first, to summa-
ion product (Kw) is affected and differs from that for fresh-
rize the effects of seawater salinity (ionic strength) and
waters. Eq. (2) in Table 2 presents the ion product ðKsw w Þ as
temperature on important chemical parameters: water ion
a function of salinity and temperature.
product, boron, silica, inorganic carbon, alkalinity, buffer 13.21
For 298.15 K (25  C) and S of 35 ppt, the Ksww is 10 in
intensity, and the chemistry of metal ferric (Fe) and aluminum 14
contrast to 10 for freshwater. The Kw values are 1013.83
sw
(Al) coagulants; second, to examine important seawater
and 1012.84 at 10  C and 35  C, respectively, and greater than
contaminants that affect coagulation such as turbidity, algae,
those for freshwaters of 1014.53 and 1013.67 at 10 and 35  C.
and TOC; and finally, to examine the potential use of various
This effect on the ion product affects, in particular, dissolved
coagulants and why coagulation with Fe coagulants is
metal-hydrolysis and metal-hydroxide precipitation reactions
preferred. In the latter part of the paper, ferric dosages and
such as occur in coagulation processes (covered in Section 4).
optimum pH conditions are addressed for various seawater
Here, it is demonstrated with the simple illustration. For
quality characteristics. Dissolved chemical speciation and
coagulation processes in water treatment, we use pH as
solubility plots for Fe in seawater are presented for S of 35 ppt
a control-operating variable; however, it is the concentration
and two temperatures: 10  C to represent temperate regions of
of [OH] that affects the chemical reactions cited above. For
the world and 35  C to represent summer seawater conditions
example, precipitation of ferric hydroxide in coagulation in
for the Middle East and other warm-water regions.
seawater carried out at pH 7 and 35  C, [OH] is 105.84
compared to 106.67 for freshwater (correspondingly at 10  C in
seawater, [OH] is 106.83 compared to 107.53 for freshwater).
2. Chemistry of seawater In summary, [OH] is much greater for seawater than what
occurs for freshwaters for the same pH conditions.
Ionic strength (I ) is the measure used to account for the
effects of the ionic content of seawater on chemical reactions.
2.2. Boron
It is defined by Eq. (1) where Ci is the ion concentration in mol/
kg and zi is the charge of the ion.
The chemistry of boron is of interest. First, it affects the
! alkalinity (Alk) and buffer intensity of seawater therefore
1 X
I¼ Ci z2i (1) affecting pH pretreatment control. Second, boron can have
2 i
detrimental health effects (developmental and reproductive)
Seawater has an average ionic strength of 0.7 mol/kg as reported by Post et al. (2010).
compared to freshwaters at about 5  104 to 102 (expressed In seawater the average total boron (BT) concentration is
as mol/L). Brackish and estuarine waters have I starting at 4.45  103 g/kg, and it is composed of boric acid (H3BO3) and
about 102 increasing to just less than seawater. borate ðBðOHÞ 4 Þ. The speciation depends on pH and is
There are several water quality parameters that are accounted for through its dissociation (acidity) constant. The
important in considering seawater chemistry and boron acidity constant ðKsw B Þ, as a function of salinity and

Table 1 e Some important seawater quality parameters.

Ion Concentration/Value References or Comments

pH 8.1e8.3 (mainly) 7.5e8.3 (range) Eby (2004)

CT, mol/kg 2.2  103 Stumm and Morgan (1996) Eby (2004)
Alk, mg/L as CaCO3 110e135 See text
SiT as SiO2, mg/L 0.12e6 See text
Turbidity, NTU 0.5 to 2 (dry weather conditions) 0.1 to 100 (range) Freeman (2009) Voutchkov (2010a; 2010b)
TOC, mg/L 2e5 (typically, can be greater e see text) Freeman (2009) Voutchkov (2010a)
UV254, cm1 0.01 or less (typically, can be greater e see text) See text
SUVA, m1 per mg/L <2 Little humic matter
>2 Some humic matter
Algae Counts, #/mL <103 Low concentrations
>103e106 Moderate to bloom conditions
5430 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0

Table 2 e Key equilibrium constants as a function of salinity and temperature.

Constants Equations References

13847:26
w ¼148:9802   23:6521 ln T
Water: ion lnKsw Millero (1995), Stumm
product  T  and Morgan (1996)
(2)
118:67
þ  5:977 þ þ 1:0495 ln T S0:5  0:01615 ðSÞ
T

1  
Boron: acidity B ¼
lnKsw 8966:902890:51 S0:5 77:942ðSÞ Millero (1995)
T       
þ1:726 S1:5 0:0993 S2 þ 148:0248þ137:194 S0:5
   0:5 
þ1:62247ðSÞ þlnT 24:434425:085 S
    
0:2474ðSÞ þ 0:053105 S0:5 T (3)

8904:2
Si ¼ 117:40 
ln Ksw  19:334 ln T
Silica: acidity Millero (1995)
 T   
458:79 0:5 188:74
þ 3:5913  I þ 1:5998 þ I
 T  T
12:1652 2
þ 0:07871  I (4)
T

2329:1378
1 ¼3:17537 
ln Ksw  1:597015 lnðTÞ
Inorganic Millero (1995), Stumm
carbon:  T  and Morgan (1996)
(8)
5:79495
þ 0:210502  ðSÞ0:5 þ0:0872208ðSÞ  0:00684651ðSÞ1:5
T

: 3403:8782
2 ¼  8:19754 
ln Ksw  0:352253 lnðTÞ
 T  (9)
25:9316
þ 0:088885  ðSÞ0:5 þ0:1106658ðSÞ  0:00840155ðSÞ1:5
T

Notation: KSW SW SW SW
w , KB , KSi , K1 and KSW
2 : are the water ion product constant, boron acidity constant, silica acidity constant and inorganic carbon
acidity constants for seawater; S is the salinity in &, T is the absolute temperature (T ), and I is the ionic strength.

temperature, is given by Eq. (3) in Table 2. For S of 35 ppt, the
 
pKsw
B is 8.76 at 10 C and 8.47 at 35 C. The fractions of boric acid
and borate as a function of pH are presented in Fig. 1. Boron
removal by RO depends on pH and is favored when the anion
ðBðOHÞ 4 Þ dominates. Therefore, some RO plants practice
enhanced removal of boron by increasing the pH to achieve
treated water 0.5 mg/L (Voutchkov, 2010a), which has been
a WHO (January 2011) provisional guideline. From Fig. 1 the pH
should be adjusted above pH 8.5 for warm waters (e.g., 35  C),
and above pH 8.8 for lower temperature waters (e.g., 10  C).
2.3. Silica
Dissolved Si is of interest because it can theoretically affect
Alk and because many solids (SiO2 and aluminum silicates) are
affected by its solubility. Furthermore, Si is incorporated and
released by algae, particularly diatoms. The concentration of
dissolved Si thus varies in seawater from about 0.12 to 6 mg/L
(2  106 to 1.0  104 mol/L) as SiO2.
Like boron, it is a weak acid and occurs in acid (H4SiO4)
and base (H3SiO 4 ) forms. Its acidity constant ðKSi Þ, as
sw
a function of ionic strength and temperature, is given by Eq.
(4) in Table 2. For average seawater I of 0.7 mol/kg, the pKsw
Si is
9.64 at 10  C and 9.21 at 35  C. Fig. 2 shows the distribution of
dissolved Si as a function of pH. H4SiO4 dominates in
seawater unless in pretreatment you increase the pH above
9.2 for warm waters (35  C) and above 9.6 for lower temper-
ature waters (10  C).
2.4. Inorganic carbon
Inorganic carbon is important in several chemical and
biochemical reactions in seawater including CO2 exchange
with the atmosphere, solubility reactions with carbonate
minerals, algal growth and respiration, and bacterial decay of
organic matter. Inorganic carbon is the main contributor to
alkalinity and buffer intensity, both of which are discussed
below.
Total inorganic carbon (CT) is defined by Eq. (5) where
½H2 CO3  is the sum of dissolved carbon dioxide: [CO2(aq)] and
[H2CO3]. ½H2 CO3  is mostly [CO2(aq)] (see Stumm and Morgan,
1996). The equilibrium concentration of H2 CO3 is set by Hen-
ry’s law. For seawater in equilibrium with atmospheric CO2(g)
at 380 ppm yields ½H2 CO3  concentrations of 1.67  105 M and
0.80  105 M for 10 and 35  C, respectively.
Surface seawater CT reported in Eby (2004) is about
2.3  103 mol/kg or 2.36  103 M assuming a seawater
density of 1025 kg/m3. This agrees closely with Stumm and
Morgan (1996) who report an average CT of 2  103 M. In
this paper, a CT of 2  103 M is used in calculations.
     2 
CT ¼ H2 CO3 þ HCO
3 þ CO3 (5)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0 5431

pKBSW = 8.76 Recognizing that [OH] is small for seawater unless you
1.0 add strong base and increase the pH to above 9, we can
H3BO3
neglect it. [Hþ] is very small and can be neglected. Other
B(OH)4-
Fraction of Species

0.8 bases in seawater are small, compared to inorganic carbon

and borate, and can be neglected. These are: (1) orthophos-
0.6 phate ðH2 PO 2 3
4 ; HPO4 ; PO4 Þ; (2) base form of dissolved Si

(H3 SO4 is small as discussed above for pH <w9); (3) base
forms of humic and fulvic acids, small except at the mouths
0.4
of some rivers; and (4) others such as NH3, HS, and HF.
10oC, S 35 ppt
Neglecting all these minor bases, we obtain Eq. (11a). This
0.2 equation can be rewritten in a practical and useful form for
seawater as Eq. (11b).
0.0    2  h i
1.0 pKBSW = 8.47 Alk ¼ þ HCO 3 þ 2 CO3 þ BðOHÞ14 (11a)
H3BO3
8  þ   þ 2 !1 9
Fraction of Species

0.8 B(OH)4-  þ !1

< H KSW H H =
Alk ¼ 1þ SW þ 2
þ2 1þ SW þ SW SW C
: K1 ½H K2 K1 K2 ; T
0.6
!
KSW
þ B
 þ
 BT (11b)
0.4 o B þ H
KSW
35 C, S 35 ppt
0.2 From earlier, CT is about 2  103 M. BT averages
4.45  103 g/kg or 4.12  104 mol/kg. For consistent units, the
BT molar concentration assuming a seawater density of
0.0
4 6 8 10 1025 kg/m3 is 4.2  104 M. Some HCO 2
3 and CO3 are com-
plexed by metals in seawater and would not contribute to
pH
alkalinity. However the fraction of HCO 3 complexed is small
Fig. 1 e Fraction of H3BO3 and BðOHÞ1L
4 as a function of pH and since it is the main contributor of alkalinity for inorganic
for seawater S& of 35: top figure at 10  C, bottom figure at carbon and other bases were neglected, complexation is
35  C. ignored here. Using the above CT and BT values and appropriate
equilibrium constants found from Eqs. (3), (8) and (9) in Table 2,
we can use Eq. (11b) to calculate alkalinities for seawater at pH
of 8.1e8.3. They are: (1) 2.2e2.3 meq/L (110e115 mg/L as CaCO3)
at 10  C where inorganic carbon accounts for 95 to 97 percent
Inorganic carbon participates in two acid-base reactions and boron the remainder; and (2) 2.5e2.7 meq/L (125e135 mg/L
shown by Eqs. (6) and (7). as CaCO3) at 35  C where inorganic carbon accounts for 93 to 95
percent and boron the remainder.
H2 CO3 #HCO
3 þH
þ
(6) Equation (11b) is the theoretical basis for Alk, but for seawater
water quality and pretreatment monitoring it is not normally
HCO 2
3 #CO3 þ H
þ
(7) used. To do so requires measurements of CT, BT, pH, and
temperature. Rather Alk is measured directly by titrating
The acidity constants ðKsw 1 and K2 Þ as a function of salinity
sw a sample with strong acid to reach a pH end-point of about 4.5.
and temperature are presented in Eqs. (8) and (9) in Table 2. Hence, it is a capacity measurement and rightly called the acid
The first acidity constant ðKsw 1 Þ for seawater salinity of 35 ppt,
neutralizing capacity. Alkalinity is used as a water quality
is 105.99 at 10  C and 105.76 at 35  C. These differ greatly from parameter as an indicator of a water’s resistance to pH decrease
freshwater (I approaching 0 M), which are 106.46 and 106.34. from acid addition, but it is a capacity measurement. To prop-
The second acidity constant ðKsw 2 Þ for seawater salinity of
erly assess the resistance for a specific change in pH from acid or
35 ppt, is 109.18 at 10  C and 108.75 at 35  C compared to base addition, one should consider the buffer intensity.
freshwater constants of 1010.49 and 1010.25.

2.6. Buffer intensity

2.5. Alkalinity

Buffer intensity (b) is more useful than alkalinity in evaluating

Alkalinity (Alk) is the acid neutralizing capacity of water and is
changes in seawater pH over specified pH changes of interest.
due to the presence of bases in seawater that can accept a
b is the resistance to a unit change in pH for a unit addition of
proton from acid addition. Alk for seawater is defined by Eq.
strong acid (Ca) or strong base (Cb), and it is defined by Eq. (12)
(10).
(Stumm and Morgan, 1996). It is the inverse of the slope of
   2  h i     alkalinity titration curves, thus a negative sign for strong acid
Alk ¼ þ HCO
3 þ2 CO3 þ BðOHÞ
4 þ OH

þotherbases Hþ
addition and a decrease in Alk and positive sign if base is
(10) added increasing Alk.
5432 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0

1.0 pKSiSW = 9.64 a

H4SiO4
Fraction of Species

0.8
1-
H3SiO4
0.6
o
10 C, S 35 ppt
0.4

0.2

0.0

1.0 pKSiSW = 9.21

H4SiO4
Fraction of Specis

0.8 H3SiO41-
b
0.6
35oC, S 35 ppt
0.4

0.2

0.0
2 4 6 8 10 12

pH
1L
Fig. 2 e Fraction of H4SiO4 and H3 SiO4 as a function of pH
for seawater S& of 35: top figure at 10  C, bottom figure at
35  C.

Fig. 3 e (a) Buffer intensity of seawater as a function of pH

DCa DAlk DCb DAlk
b¼ ¼ ¼ ¼ (12)
DpH DpH DpH DpH
Applying this to Eq. (11b), we can obtain Eq. (13) for b as
a function of pH for seawater (Eby, 2004) for a closed system,
meaning the dissolved carbon dioxide concentration is not
forced to be in equilibrium with carbon dioxide in the atmo-
spheric gas phase. This is applicable because dissolved carbon
dioxide in seawater is usually out of equilibrium with the
atmosphere, and because pretreatment-detention times are
short so with pH changes equilibrium with atmospheric
carbon dioxide is not established.
The causes of buffer intensity are identified below the
terms in Eq. (13). Buffer intensity is due to borate, inorganic
carbon, OH at high pH, and Hþ at low pH.
for temperatures at 10 and 35  C. (b) Individual buffer
intensity terms at 35  C (Conditions: CT at 2.0 3 10L3 M,
BT at 4.2 3 10L4 M).
percent at 10  C. Fig. 3(a) shows the total buffer intensity, which
is instructive. If we wish to decrease the pH of seawater in
coagulation, then small amounts of acid either from Fe dosing or
from a strong acid (HCl or H2SO4) are required to reach target pH
conditions near 7 because of the relatively low b. To decrease the
pH to 5.5 to 6.5, the buffer intensity increases significantly
requiring larger amounts of Fe or strong acid. The effect of buffer
intensity on coagulation is discussed further in Section 4.

(13)
Fig. 3 presents buffer intensity as a function of pH for
seawater at 10 and 35  C. Fig. 3(b) shows that the primary
contributor to buffer intensity is inorganic carbon. Boron
contributes to buffer intensity for pH >7, and for seawater pH at
about 8 boron contributes about 18 percent at 35  C and 12
3. Contaminants
The presentation below focuses on the commonly occurring
contaminants that must be removed in pretreatment to
prevent fouling of RO membranes and to enhance the opera-
tional efficiency of RO. These contaminants are (1) mineral
particles, (2) algae, and (3) TOC and natural organic matter
(NOM). The occurrence and concentrations of these contam-
inants depend on the location of the desalination plants. Most
plants are in coastal areas and the waters in these areas are
subject to variation in water quality. Other contaminants that
may occur are oil and grease (O&G) and hydrocarbons, and
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
these can be removed via coagulation and separation and are
discussed briefly.
3.1. Particles
3.1.1. Turbidity and silt density index
The presence of particles is normally measured as turbidity.
Seawater turbidity is usually fairly low as indicated in Table 1,
but can be greater at some coastal locations and when algal
blooms occur. Particles can cause colloidal/particle fouling of
RO membranes in which a cake of particles form reducing the
permeate flux. Algae can also cause biofouling e see Section
3.2.2. One pretreatment criterion for the water applied to RO
membranes is the turbidity of the water following pretreat-
ment. A turbidity goal of 0.1 NTU is desired while some
membrane companies accept a higher turbidity such as 0.5
NTU (Freeman, 2009; Voutchkov, 2010a, 2010b). Another
widely used pretreatment criterion of the acceptability of
water applied to RO membranes is the SDI (Silt Density Index).
It is a measure of colloidal/particle fouling. The SDI is a labo-
ratory filtration test using a 0.45 mm membrane filter at
a pressure of 207 kPa (30 psi) using three time intervals to
collect filtered water (ASTM, 2007). Samples that contain
greater colloid concentrations foul the membrane filter and
take longer times to filter; hence higher SDI values. Guidelines
for RO feedwater are: SDI >5 is not acceptable, SDI <2 is good
quality, and SDI maximum of 3e4 (Freeman, 2009; Duranceau
and Taylor, 2010; Voutchkov, 2010a, 2010b).
3.1.2. Mineral particles
These particles can be (1) clays (aluminum silicates) carried by
rivers and direct runoff to coastal waters; (2) other aluminum-
silicate minerals carried by rivers and runoff or produced in
the ocean by precipitation processes; (3) SiO2(s) (sand) parti-
cles of small sizes, present in the ocean and coastal areas, and
(4) CaCO3(s) solids produced in the ocean by precipitation.
Mineral particles are stable (tendency not to flocculate) in
freshwaters. The primary cause of particle stability is the
negative charge on particle surfaces. Other possible causes are
hydration and steric effects, but surface charge is the primary
one. As particles approach each other, a repulsive force occurs
from electrical double-layer interaction from the particle-
surface charge and an attractive force occurs from van der
Waals forces at close particleeparticle separation distances. In
freshwaters, the electrical-double layer thickness and repulsive
interaction extends far enough from the particle surfaces that
there is a net repulsive force causing stability. In seawater, the
particle zeta potential is reduced as illustrated by clay particle
electrophoretic particle mobility (EPM) data for freshwater
versus seawater shown in Fig. 4. In seawater the electrical
double layer is compressed toward the particle surface allowing
for clay particles to approach close enough to each other that the
attractive van der Waals forces can dominate and clays floccu-
late. This occurs for mineral particles in seawater, and in estu-
aries where mineral particles carried by rivers flocculate and
undergo deposition at the mouths of rivers (Edzwald et al., 1974).
Based on surface charge, mineral particles carried by rivers
and runoff into the sea should flocculate and settle and
therefore be at relatively low concentration at the intakes to
desalination plants, but some mineral particles can be present
5433
2
Fresh Water
Fresh Water
Seawater
Seawater
Seawater
Seawater
EPM x 108 (m/s per V/m)
1
0
anophagefferens
Auerococcus
Karenia brevis
-1
-2
Kaolinite
-3
Montmorillonite
-4
Fig. 4 e Electrophoretic mobility (EPM) for two algae in
seawater (Karenia brevis and Auerococcus anophagefferens)
and two clays (montmorillonite and kaolinite) in
freshwater and seawater (data from Sengco (2001).
in which particle stability is attributed to adsorbed NOM and
hydration effects at particle surfaces. Other mineral particles
can be present because they are produced in seawater by
precipitation of CaCO3 or aluminum silicates. In all cases,
however, the stability due to charge should be low because of
the high ionic strength of seawater and these particles are
easily treated by coagulation.
3.1.3. Algae
Algae are often a major problem for desalination plants and
must be removed by pretreatment. For all RO installations,
algae are present at low to moderate concentrations (Table 1),
but many plants are located in areas subject to algal blooms
where concentrations may greatly exceed 103 cells per mL,
especially those plants located in warm, shallow seas. Almost
all groups of phytoplankton-type algae (floating or suspended)
are found in seawater. Unlike mineral particles, algae remain
suspended while undergoing growth and do not flocculate
until they are in the later stages of their growth phase. With
the high ionic strength of seawater, the algal-particle stability
is not due to surface charge, although algae do carry a negative
charge with low negative zeta potentials or EPM as shown for
two algae in Fig. 4. Algal-particle stability can be due to (1)
steric effects, (2) algal motility, and (3) hydration effects. Steric
effects are attributed to their surface structure that prevents
aggregation. Motility or swimming refers to the fact that many
algae have flagella (such as dinoflagellates and others) or cilia
(hair-like structure around the exterior of the cells) that
provide a means of swimming and steric repulsive interac-
tion. Hydration effects mean the presence of cell-surface
groups with bound water inhibiting flocculation. Two exten-
sive reviews of algae in freshwaters by Henerdson et al. (2008)
and Ghernaout et al. (2010) cover algal properties, dissolved
organic matter excreted by algae, and coagulation of algae.
In many coastal areas and in warm-shallow seas, one can
get harmful algal blooms often referred to as red tides and
brown tides depending on the pigments associated with the
algae. Some of these harmful algae can also bloom in colder
waters. Without proper pretreatment, the algae and associ-
ated toxins can shut down desalination plants. Many dino-
flagellates are associated with red tide; for example, Karenia
5434 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0

brevis and Noctiluca scintillans (Fig. 5). Also shown in Fig. 5 is
brown-tide algae, Auerococcus anophagefferens (a chryso-
phyte). Some algae are large (e.g., dinoflagellates of 10s to
100s of mm), while other algae are small, roughly 1 to several
mm (e.g., diatoms, green algae, cyanobacteria, and chryso-
phytes). Removal of algae by pretreatment is essential in
preventing algae from accumulating on RO membranes
causing biofouling (see Section 3.2.2). Algae can be coagu-
lated effectively with good separation by DAF clarification
(Edzwald, 2010) and granular media filtration prior to RO
membranes.
3.2. Natural organic matter
The concentrations and types of NOM present in seawater are
highly dependent on the water plant intake location. Higher
levels are found at the mouths of river carrying NOM and in
waters subject to algal blooms. This discussion begins with
collective and surrogate parameters for NOM and then is fol-
lowed by consideration of types of NOM.
3.2.1. Collective and surrogate parameters
TOC is a collective parameter, and it is the most widely used
measurement to characterize the concentration of organic
matter. TOC is the sum of particulate organic carbon (POC)
and dissolved organic carbon (DOC). In the absence of algal
blooms, TOC is approximately the DOC. The TOC of seawater
used at desalination plants is typically 2e5 mg/L (see Table 1),
but it can be greater when algal blooms occur or for seawater
affected by aquatic humic matter from rivers e for example,
the Tampa Bay (Florida, United States) facility has raw water
TOC averaging 6.2 mg/L with a range of 4.1e12 mg/L
(Schneider, 2011).
A useful surrogate parameter for DOC (or TOC
when w DOC) is the UV absorbance at 254 nm (UV254) - see
Edzwald et al. (1985), Edzwald and Tobiason (2010). Quite often
UV254 is low for seawater at 0.01 cm1 or less (see Table 1);
consequently, in measuring UV254 for raw seawater and
especially across the treatment plant, spectrophotometer
cells with a path-length greater than the typical 1 cm should
be used to increase accuracy. Where seawater is affected by

Fig. 5 e Photos of examples of algae that occur in seawater (a) Karenia brevis, a dinoflagellate, size 25 mm, associated with red
tide (Source: Florida Fish and Wildlife Conservation Commission, http://research.myfwc.com/images/articles/23559/23559_
5513.jpg&imgrefurl); (b) Auerococcus anophagefferens a chrysophyte, size 2e3 mm, associated with brown tide, (Source: U.S.
Department of Energy Joint Genome Institute, http://www.jgi.doe.gov/); (c) Noctiluca scintillans, a dinoflagellate, size
200e2000 mm, associated with red tide (Source:Australian Government, Department of the Environment, Water, Heritage,
and the Arts, http://www.tafi.org.au/zooplankton/imagekey/dinophyta/) (all accessed 12 January 2011).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
the presence of aquatic humic matter and by algal blooms,
raw seawater will have higher UV254 than 0.01 cm1.
3.2.2. Types of organic matter and biofouling
A variety of types of organic matter can be present in seawater.
For seawater that contains aquatic humic matter, the NOM is
a mixture of aquatic humic and fulvic acids (especially, fulvic
acids). Aquatic humic matter has properties of fairly high
molecular weight (100se1000s) and while it carries a negative
charge it has hydrophobic character, is reactive and can be
removed in pretreatment by coagulation and separation of the
precipitated solids. Other NOM can come from the decay of
plant and algal matter. In addition algae, while undergoing
growth and respiration, impart soluble organic matter to the
water called extracellular organic matter (EOM). The algae can
also impart intracellular organic matter (IOM). EOM and IOM
are referred to in some literature as algogenic organic matter
(AOM). The compounds composing AOM consist of acids,
proteins, simple sugars, anionic polymers, negatively charged
and neutral polysaccharides. The negatively charged
compounds and some neutral compounds exert a coagulant
demand as discussed further in Section 4.2.2. Another concern
is that, without good pretreatment, algae such as those in Fig. 5
that are transported to RO membranes undergo cell lysis, due
to the transmembrane pressure, releasing IOM. The
compounds released are soluble, biodegradable, and a major
concern regarding biofouling of RO membranes. Measure-
ments for the presence of these compounds are not common,
but fluorescence type measurements may have promise (Para
et al., 2010; Baghoth et al., 2011).
Ultraviolet light at 254 nm is absorbed by a variety of
organic molecules, but especially organic compounds with an
aromatic structure. Aquatic humic matter has this structural
feature so it absorbs more light per unit concentration of DOC
(called the specific UV absorbance or SUVA) than other types of
NOM such as hydrophilic acids, hydrophobic bases (e.g.,
proteins and aromatic amines), and hydrophobic neutrals (e.g.,
aldehydes) (Edzwald and Van Benschoten, 1990; Edzwald, 1993;
Bose and Reckhow, 1998; Edzwald and Tobiason, 2010). SUVA is
a simple method for characterizing the nature of NOM.
Applying the concept to seawater, the following guidelines are
provided: (1) SUVA of about 4 or greater indicates the NOM is
dominated by aquatic humic matter, (2) SUVA of 2e4 indicates
the NOM is composed of a mixture of aquatic humic matter,
EOM and IOM from algae, and (3) SUVA <2 indicates the NOM is
composed largely of algal derived EOM and IOM.
Coagulation of organic matter is an important pretreat-
ment step to prevent fouling of RO membranes. Voutchkov
(2010a; 2010b) reports that if the TOC is reduced to 0.5 mg/L
or less then biofouling is unlikely, while seawater with TOC
above 2 mg/L is most likely to cause biofouling. TOC of 0.5 mg/
L is difficult to achieve without treatment by granular acti-
vated carbon or nanofiltration. In assessing good coagulation
pretreatment, for seawater containing several mg/L of TOC,
a guideline of 1e2 mg/L is recommended.
3.3. Other contaminants
Removal of oil and grease (O&G) and oil-based hydrocarbons
in pretreatment is essential to prevent fouling of RO
5435
membranes. The sources of O&G can be from wastewater
discharge or from storm drains in the vicinity of desalination
plant intakes. Voutchkov (2010a; 2010b) reports that
O&G > 0.02 mg/L can cause membrane fouling. Removal of
O&G can be achieved by coagulation followed by separation
through sedimentation or more effectively by DAF.
Hydrocarbons from ship spills or seawater oil wells must
also be removed through pretreatment. The hydrocarbons are
incorporated into the O&G measurement and can be removed
as indicated above. Depending on the magnitude of the oil
spill, removal of the oil by ferric chloride coagulation followed
by DAF may be feasible.
4. Coagulation
Coagulation is a chemical pretreatment step. It concerns
treating seawater so that particles in the supply and particles
produced within pretreatment through precipitation
processes (e.g., metal hydroxides from coagulation) are
destabilized (flocculate readily) for downstream flocculation
and particle removal processes. Metal coagulants (Al and Fe)
are used commonly to treat freshwaters under dose and pH
conditions in which they precipitate as aluminum or ferric
hydroxides. This precipitation process produces new particles
that can incorporate the raw water particles into flocs, called
sweep-floc coagulation in water treatment. Coagulation can
also remove NOM by a phase change converting the dissolved
NOM into particles directly by precipitation or by adsorption
onto particles produced by the coagulant. Coagulation is an
essential pretreatment process because it affects all down-
stream pretreatment processes. It affects flocculation (parti-
cles do not aggregate well without coagulation), DAF (bubble
attachment to flocs and removal of floc-bubble aggregates),
sedimentation, and granular media filtration.
Ferric salts, particularly ferric chloride, are the most
common coagulants used in the pretreatment of seawater, but
there have been numerous studies of the potential use of
other coagulants. In this section, the chemistry of several
coagulants (Table 3) is examined, and it is explained why
ferric coagulants are preferred for seawater pretreatment.
Table 3 e Potential coagulants for treating seawater.
Coagulants
Alum and PACls
Concerns about precipitative scaling
Ferric Salts
Widely used, especially ferric chloride
Organic Cationic Polymerelctrolytes
Low Molecular Weight, High Charge-Density Cationic
Some RO membrane manufacturers allow; case made in text for
their use in dual-coagulant strategy with ferric salts
Flocculant, Flotation, and Filter Aids
High Molecular Weight Nonionic and Anionic Polymers
Not primary coagulants: possible use as flocculant,
flotation, and filter aids Concerns about fouling RO membranes
5436
4.1. Aluminum coagulants
While alum and polyaluminum chlorides (PACls) have been
studied in laboratory and pilot-scale work for RO pretreatment
of seawater, they are not used in full-scale plants. The primary
reason is because of the relatively high solubility of Al. The
residual soluble Al would carry over to the RO membrane
leading to concentration and precipitative scaling.
The problem is demonstrated by showing solubility
curves for amorphous Al(OH)3(am), which would be formed
in coagulation. Four soluble Al species are considered in
equilibrium with Al(OH)3(am) for seawater ionic strength:
Al3þ, Al(OH)2þ, AlðOHÞþ 
2 , and AlðOHÞ4 . The solubility product
constant was taken from Van Benschoten and Edzwald
(1990) and the hydrolysis constants were taken from
Nordstrom and May (1996) for infinitely dilute solutions and
adjusted for temperature. They were then adjusted to
account for the ionic strength of seawater as explained in
Section 4.2 for ferric coagulants. For the solubility constants
of Al(OH)3(am), pKSWs1 of 12.27 and 10.83 were used for 10 and
35  C, respectively, as defined by the stoichiometry in Eq.
(14).
AlðOHÞ3 ðamÞ þ 3Hþ #Al
3þ
þ 3H2 O
Fig. 6 shows the solubility curves. Al is more soluble in
seawater compared to freshwater because of the ionic
strength of seawater. Comparing the two temperatures, Al is
more soluble at 35  C above about pH 6 but less soluble below
pH 6. When using Al coagulants, optimum pH conditions
occur near the pH of minimum solubility at Al doses that
minimize turbidity and NOM. This also produces the
maximum amount of precipitated solids for sweep-floc
coagulation and minimizes residual soluble Al. The pH of
minimum solubility at 10  C is pH 6.8 yielding a soluble Al of
about 106 mol/L (27 mg/L). At 35  C, the pH of minimum
solubility is lower at pH 6 and the soluble Al is much greater
at about 106 mol/L (270 mg/L). To practice alum coagulation
for seawater near the pH of minimum solubility, high alum
dosages are required to overcome the buffer intensity (see
Fig. 3(a)) or the addition of strong acid with alum would be
0
-1
Al(OH) (am)
-2 precipitation for dosing above curves
-3
Log Al (M)
-4
-5
-6
Solubility
Minimum
Solubility
Minimum
-7
-8
-9
4 5 6 7
pH
3
Fig. 6 e Solubility of Al(OH) (am) in seawater as a function
of pH at 10 and 35  C (pH of minimum solubility shown by
arrows).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
required. The pH of raw seawater is generally near 8, if one
were to practice alum coagulation at pH 7.5 to 8, the soluble
Al would be much greater than at the pH of minimum
solubility.
PACls act as coagulants by charge neutralization and by
precipitation as an aluminum hydroxide solid. These modes of
coagulation are highly dependent on dose and pH. If used at pH
>7, precipitation of an aluminum hydroxide solid occurs leaving
a high soluble Al concentration at the tenths to 1 mg/L or greater
depending on water temperature (Pernitsky and Edzwald, 2003).
If used at pH <7, soluble Al can be greater.
In short, Al is too soluble when using Al coagulants and
would be carried to the RO membranes where it can concen-
trate producing aluminum hydroxide and aluminum silicate
solids causing precipitative scaling. The best choice of a metal
coagulant is ferric salts.
4.2. Ferric coagulants
Ferric coagulants are the best choice for seawater coagulation
(Table 3). Ferric chloride is commonly used, but there has been
some consideration of the use of other ferric salts such as
ferric sulfate and FeClSO4 as well the production of Fe by
(14) electrocoagulation employing Fe electrodes (Timmes et al.,
2009). Ferric chloride is very insoluble leaving little residual
dissolved Fe in the water after pretreatment (shown below)
and thus avoiding precipitative scaling problems. We discuss
next the dissolved Fe speciation, the solubility of ferric
hydroxide, and effects of seawater buffer intensity on reach-
ing target pH coagulation conditions. The presentation below
considers coagulation with ferric chloride, but the principles
also apply to ferric sulfate.
4.2.1. Dissolved Fe speciation and ferric hydroxide solubility
Thermodynamic data (Drever, 1997; Stumm and Morgan,
1996) were used to calculate equilibrium constants at 10 and
35  C for infinitely dilute solution. The equilibrium constants
were then adjusted for seawater ionic strength conditions
using activity coefficients for the ions: monovalent 0.7, diva-
lent 0.25, and trivalent 0.05 (Stumm and Morgan, 1996). The
calculated seawater equilibrium constants are presented in
Table 4.
Figs. 7(a) and 8(a) present the fractions of dissolved Fe
species as a function of pH for 10 and 35  C, and Figs. 7(b) and
8(b) present solubility curves for amorphous ferric hydroxide.
The dissolved Fe species is discussed first with emphasis on
pH 7.5 to 8 representing raw seawater or coagulation with
Al ( g/L)
little pH adjustment, and then for pH 6 to 7 representing
2700
coagulation with pH adjustment. For pH 7.5 to 8 and 10  C as
270
shown in Fig. 7(a), there would be fairly high fractions (80e97
27
percent) of positively charged Fe as FeðOHÞþ 2 available for
charge neutralization coagulation reactions; however, as
10 Co
35oC
shown in Fig. 8(a) there would be much less (10e55 percent)
for the same pH range for seawater at 35  C. For pH 6 to 7 and
8 9 10 10  C (Fig. 7(a)), the FeðOHÞþ 2 fraction is maximized near 99
percent. At 35  C (Fig. 8(a)), the FeðOHÞþ2 fraction is still maxi-
mized near 99 percent at pH of about 6, but decreases to about
90 percent as the pH increases to 7. In general, then the frac-
tion of positively charge Fe varies with pH and temperature
with less of it with increasing pH.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0 5437

These are equilibrium concentrations and assume no

Table 4 e Ferric chemistry and equilibrium constants for
seawater.
Equations
FeðOHÞ3 ðamÞ þ 3Hþ #Fe3þ þ 3H2 O
2þ
Fe 3þ
þ H2 O#FeðOHÞ þH
Fe3þ þ 2H2 O #FeðOHÞþ
2 þ 2H
þ
Fe3þ þ 4H2 O #FeðOHÞ
4 þ 4H
þ
The solubility curves for 10 and 35  C are presented in
Fig. 7(b) and Fig. 8(b), If we coagulate with the addition of Fe at
dosages of 1 mg/L (1000 mg/L) or greater (not accounting for
coagulant demand by complexation of NOM, to be discussed
in Section 4.2.2), then initially the system is oversaturated
with Fe and precipitation of amorphous ferric hydroxide
occurs, which is called sweep-floc coagulation. The soluble Fe
concentration is reduced and at equilibrium drops to the
concentrations shown on the curves. Fe is very insoluble over
a wide range of pH conditions making it a better coagulant
than Al salts because the residual soluble concentrations
would be very low preventing precipitative scaling of RO
membranes. For pH of 6e8 at 10 and 35  C (Fig. 7(b) and
Fig. 8(b)), the residual soluble Fe is much less than 1 mg/L.
complexation reactions with anions that can increase soluble
Fe concentrations, but these effects should be small and in
10  C 35  C
practice the soluble Fe is expected to be higher but still low.
(15) pKSW
S1 4.95 3.63
4.2.2. Ferric coagulation dosing
þ pbSW
Ferric chloride doses range typically from 1 to 10 mg/L as Fe
(16) 11 3.15 2.49
depending on (1) raw water quality characteristics, (2) pH of
coagulation, and (3) whether a strong acid (e.g. HCl or H2SO4) is
(17) pbSW
12 6.84 6.32
used to adjust the pH of coagulation.
For raw seawater of good quality, say low to moderate
(18) pbSW
14 23.38 21.39 turbidity (<10 NTU), low algae concentrations (<103 cells/mL),
low TOC (3 mg/L), low UV254 (<0.03 cm1) and low SUVA
(<1.5 m1 per mg/L), then ferric chloride doses should be low
to moderate (several mg/L or less depending on how pH is
adjusted). Because TOC is low and there is little organic matter
of aromatic character (low SUVA) with negative charge, then
there would little coagulant demand for positively charged Fe.
Thus, coagulation can be carried out at relatively high pH
values with some temperature dependence as discussed
above with Figs. 7 and 8. For warm waters (say 20e35  C) or
even greater, coagulation pH of 7e7.5 is recommended. This
may be achieved with low Fe doses (few mg/L or less) espe-
cially if strong acid is used to achieve the target pH (see the
b curve for 35  C in Fig. 3(a)). For lower temperature seawater
(say <20  C), then coagulation pH conditions of 7.5e8 should
be effective. Fe doses should be low (few mg/L or less) and
addition of strong acid should not be necessary assuming raw
seawater pH is about 8 e see b curve for 10  C Fig. 3(a) the fairly
low buffer intensity at pH 7.5 to 8.

a 1.0
a 1.0
Fe(OH)2+
Fraction of Dissolved Species

Fe(OH)2+
Fe(OH)4- Fe(OH)4-
Fraction of Dissolved Species

0.8 Fe3+ 0.8

Fe3+

0.6
0.6

0.4 0.4
o
Fe(OH)2+ 10 C Fe(OH)2+ 35oC
0.2 0.2

0.0 0.0
b b
-2 -2

-4 Fe(OH)3(am) 5600 -4 Fe(OH)3(am) 5600

Fe ( g/L)

Fe ( g/L)

precipitation for dosing above curve

Log Fe (M)

precipitation for dosing above curve

Log Fe (M)

-6 56 -6 56

-8 0.56 -8 0.56

-10 0.0056 -10 0.0056

10oC 35oC
-12 -12

-14 -14
2 4 6 8 10 12 2 4 6 8 10 12

pH pH

Fig. 7 e (a) Fraction of dissolved Fe species as a function of Fig. 8 e (a) Fraction of dissolved Fe species as a function of
pH for seawater at 10  C, (b) solubility diagram for ferric pH for seawater at 35  C, (b) solubility diagram for ferric
hydroxide for seawater at 10  C. hydroxide for seawater at 35  C.
5438 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
Raw seawater with higher turbidity, if caused by minerals,
should not have much effect on ferric chloride doses and
coagulation pH as long as algae concentrations, TOC, UV254,
and SUVA are low as presented above.
The effect of algae concentration on coagulation is
important. With moderate algae concentrations and algal
blooms (>103 to 106 cells/mL), turbidity and TOC will increase
affecting ferric chloride dosing. If algal blooms occur
increasing the background TOC by 2 mg/L or more, then the
algae and especially the dissolved organic matter associated
with the algae create a coagulant demand that must be
satisfied by adding sufficient Fe in coagulation to first complex
with the NOM compounds and then to precipitate ferric
hydroxide under sweep-floc conditions at the pH conditions
specified below. The concept is to maximize the fraction of
dissolved cationic Fe species that would be available when the
ferric coagulant is added to react with negatively charged
dissolved organic matter and solids. For warm waters (say
20e35  C) or even greater, coagulation pH of 6e6.5 is recom-
mended to maximize FeðOHÞþ 2 (Fig. 8(a)). The buffer intensity
increases below pH 7 (see the b curve for 35  C in Fig. 3(a)) so
ferric chloride doses could be high (several mg/L to say 10) to
achieve pH 6 to 6.5 without the use of strong acid. For lower
temperature seawater (say <20  C), then coagulation pH
conditions of 6.5e7 should be effective to maximize the
availability of FeðOHÞþ2 (Fig. 7(a)). Because the buffer intensity
for lower temperature seawater increases below pH 7.5, like
for the warm water case ferric chloride doses could be high
(several mg/L to say 10) to achieve pH 6.5 to 7 without the use
of strong acid. Coagulation followed by downstream floccu-
lation and particle separation thus removes the particles
originally in the raw water, the particles produced by precip-
itation, and some NOM that was complexed by positively
charged Fe and adsorbed to flocs.
Another raw water quality characteristic that affects coag-
ulant dose and optimum coagulation pH conditions is the
presence of humic substances. SUVA is a good and practical
indicator of whether the DOC of seawater contains humic
substances or not, and it is discussed in Section 3.2.2. Seawater
with aquatic humic matter (SUVA >2) requires ferric coagula-
tion at certain pH conditions because of the fact there is little
positively charged Fe available to react with negatively charged
humic matter that creates a coagulant demand unless the pH
is 6e6.5 for warm waters (Fig. 8(a)) or about pH 6.5 or slightly
greater for cooler waters (Fig. 7(a)). Dosages of 5e10 mg/L may
be needed depending on the raw seawater TOC and whether
strong acid is used to adjust pH. In a laboratory jar-test study at
24e25  C, Duan et al. (2002) found that ferric chloride at
5e10 mg/L as Fe with pre-adjustment to pH 6.1 to 6.2 (final pH
of 5.7e6.1 after addition of ferric chloride) produced good
removals (85e90 percent UV254) of humic acid from seawater.
For seawater in which NOM from algae or humic substances
is exerting a coagulant demand, then a dual-coagulant strategy
should be considered in which both a low molecular weight
high charge-density cationic polymer (Table 3) and ferric
coagulant are used. The cationic polymer provides positive
charge to satisfy partially the negative charge associated with
particles and more importantly the dissolved NOM. How to do
this properly and avoid the possibility of overdosing the
cationic polymer is addressed next.
4.3. Organic polymers
Polymers are classified into two categories as summarized in
Table 3: (1) low molecular weight, high charge-density cationic
polymers, and (2) high molecular weight polymers. The former
are primary coagulants that accomplish coagulation of nega-
tively charged particles by adsorption on particle surfaces
through charge neutralization. The cationic polymers can also
react with a small fraction of dissolved negatively charged
humic substances and lead to charge neutralization and
precipitation of a solid phase, cationic polymer-humate
precipitate. However, the degree of reaction and removal of
dissolved organic matter is low compared to metal coagulants.
Some RO suppliers allow the use of cationic polymers, while
others do not because they are concerned about overdosing of
the cationic polymers leading to carry-over to RO membranes
where they would adsorb and foul the membrane. However, as
is discussed below the cationic polymers would not be used as
the sole coagulant but used at low dosages in a dual-coagulant
strategy with ferric coagulant.
High molecular weight nonionic and anionic polymers are
not primary coagulants. They have several other potential
uses. One is a flocculant aid in which they are added after
coagulation for the purposes of increasing floc sizes by
bridging flocs together and of strengthening flocs. This appli-
cation applies to plants with sedimentation where large flocs
are desired. Another use is as a flotation aid for plants with
DAF. Their function is to strengthen flocs and to retain floated
sludge. A third use is as a filter aid for granular media filtra-
tion. Here they aid attachment of flocs to filter grains.
Voutchkov (2010a) reports that these anionic and nonionic
polymers are an option in RO pretreatment; however, some
RO manufacturers do not favor their use. If overdosing occurs
and granular media filtration is included in pretreatment,
then the filters would perform poorly often in the form of
plugging near the surface producing high head loss. Over-
dosing can occur where poor filter performance is not always
so obvious and control of the polymer dose can be problem-
atic. There has been some use of these type polymers as floc-
aids at dosages at tenths of a mg/L. If these polymers reach RO
membranes, then they adsorb strongly on the membranes
producing scaling and fouling. Their use should be avoided,
but if used then use with caution.
High charge-density cationic polymers are frequently used
in a dual-coagulant strategy for treating freshwaters. Their
positive charge can neutralize the negative charge of particles
and can satisfy some of the negative charge associated with
aquatic humic matter and algae thus reducing metal coagu-
lant dosages. Their use in seawater coagulation to comple-
ment ferric coagulation can be advantageous, especially
because seawater pH is fairly high limiting the fraction of
positively charged Fe species available for charge neutraliza-
tion e Figs. 7(a) and 8(a). Some RO membrane manufacturers
allow their use as stated earlier, while others may not because
of concern of overdosing and fouling negatively charged RO
membranes.
This concern is without sound basis and cationic polymers
should be considered in a dual-coagulant strategy with ferric
chloride. Overdosing of cationic polymers in a dual-coagulant
strategy is almost impossible. In a dual-coagulant strategy the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
cationic polymer dosage is held constant at a low concentra-
tion (typically a mg/L or less on a liquid product basis) where
overdosing cannot occur because ferric chloride serves as the
primary coagulant. With changes in raw water quality, the
ferric chloride dosage is adjusted not the cationic polymer.
One can use a streaming-current monitor to insure no over-
dosing of the cationic polymer. Bench-scale and pilot-scale
studies can easily evaluate the low and constant cationic
polymer dose that is beneficial in conjunction with Fe dosing.
This type of dual-coagulant strategy has been reported in use
for the Al Jubail desalination plant in Saudi Arabia (Baig and Al
Kutbi, 1998). Acid is added to control pH at about 6.5, ferric
chloride dose is low (about 1 mg/L as Fe), and the cationic
polymer dosage is 0.2e0.4 mg/L.
5. Summary
Chemical equilibrium constants are presented as a function of
salinity (S ) and temperature for the ion product of water and
for important seawater quality parameters. The latter include:
the acid-base chemistry of boron, silica, and inorganic carbon.
The salinity of seawater shifts pKSW s to lower values (higher
equilibrium constants (KSW s)) than for freshwaters making
chemical speciation in seawater much different. For the
Middle East and tropical areas, water temperatures can
be quite high (w35  C). High water temperatures yield lower
pKSW s. While alkalinity is often used as an indicator of buff-
ering from addition of acids, it is a capacity measurement.
Buffer intensity (b) is a better measure of resistance to speci-
fied changes in pH and is discussed with regard to pH control
and adjustment in coagulation. Inorganic carbon is the main
cause of seawater alkalinity and buffer intensity, but borate
ðBðOHÞ 4 Þ also contributes.
Commonly occurring contaminants for seawater are mineral
particles, algae, and dissolved organic matter that should be
removed by pretreatment coagulation and particle separation to
minimize RO membrane fouling. The concentrations of these
contaminants vary and are influenced by river inputs and algal
blooms. To avoid fouling from mineral particles, turbidity
following pretreatment should be low (<0.5 NTU with goal of
<0.1 NTU) and the SDI should be low (<2). Dissolved natural
organic matter and algae can cause biofouling. A collective
measure of organic matter is TOC, while UV254 is a good surro-
gate measure of DOC. The nature of DOC can be characterized by
SUVA where SUVA >2 indicates the presence of humic matter.
Guidelines for good quality water following coagulation
pretreatment for raw seawater with moderate to high TOC are:
TOC 1e2 mg/L, UV254 < 0.01 cm1, and SUVA<2. Algae are
a particular problem because if carried to RO units, they will
undergo lysis due to the transmembrane pressure releasing
intracellular organic matter. Algae in seawater may occur at
relatively low concentration of <103 cells/mL or at moderate to
bloom concentrations (<103 to 106 cells/mL). Coagulation
pretreatment and particle separation should maximize
removals of algae with removals of 99 percent or greater.
Aluminum based coagulants are too soluble for use in
pretreatment of seawater. The residual dissolved Al can carry to
RO membranes and concentrate causing precipitative scaling.
Ferric coagulants are the best choice and they are widely used.
5439
The solubility of Fe in seawater is low over a wide range of pH
and temperature conditions for equilibrium conditions with
amorphous ferric hydroxide. Ferric chloride doses range typi-
cally from 1 to 10 mg/L as Fe depending on (1) raw water quality
characteristics, (2) pH of coagulation, and (3) whether a strong
acid (e.g., HCl or H2SO4) is used to adjust the pH of coagulation.
For raw water containing mineral particle turbidity and low
concentrations of algae and TOC, ferric chloride is effective at
relatively low dosages and at pH conditions of 7e7.5 for warm
waters and 7.5 to 8 for cooler waters. Algae at high concentra-
tions produce AOM and some of these soluble organic
compounds create a coagulant demand. Thus, seawater with
high algae can require higher Fe doses and lower coagulation
pH. For these waters, the optimum pH conditions are 6e6.5 for
warm waters and 6.5 to 7 for lower temperatures. For waters
influenced by humic matter (SUVA>2), then the coagulant doses
and pH conditions are controlled by the humic matter (exert
a coagulant demand). Fe dose depends on the TOC concentra-
tion and increases with TOC. The optimum pH conditions are
6e6.5 for warm waters and about 6.5 or slightly greater for lower
temperatures. A dual-coagulant strategy is recommended for
treating seawater containing moderate to high concentrations
of algae or seawater containing humic matter. This strategy
involves low and constant dosing (1 mg/L or less) with high
charge-density cationic polymers and Fe as the main coagulant
where it is varied in response to raw water quality changes. The
positive charge from the polymer can neutralize the negative
charge of particles and can satisfy some of the coagulant
demand from the aquatic humic matter or from the algae and
AOM, thus reducing metal coagulant dosages.
references
ASTM D4189-07, 2007. Standard Test Method for Silt Density
Index (SDI) of Water. doi:10.1520/D4189-07. www.astm.org,
Conshohocken, PA.
Baghoth, S.A., Sharma, S.K., Amy, G.L., 2011. Tracking natural
organic matter (NOM) in a drinking water treatment plant
using fluorescence excitation-emission matrices and
PARAFAC. Water Research 45 (2), 797e809.
Baig, M.B., Al Kutbi, A.A., 1998. Design features of a 20 migd SWRO
desalination plant, Al Jubial, Saudi Arabia. Desalination 118,
5e12.
Bose, P., Reckhow, D.A., 1998. Adsorption of organic matter on
preformed aluminum hydroxide flocs. Journal of
Environmental Engineering 124 (9), 803e811.
Drever, J.I., 1997. The Geochemistry of Natural Waters, third ed.
Prentice Hall, Englewood Cliffs, NJ.
Duan, J., Graham, N.J.D., Wilson, F., 2002. Coagulation of humic
acid by ferric chloride in saline (marine) water conditions.
Water Science and Technology 47 (1), 41e48.
Duranceau, S.J., Taylor, J.S., 2011. Membranes, chapter 11 in water
quality and treatment. In: . Edzwald, J.K. (Ed.), A Handbook on
Drinking Water, sixth ed. McGraw-Hill, New York.
Eby, G.N., 2004. Principles of Environmental Geochemistry.
Thomson/Brooks Cole, Pacific Grove, CA.
Edzwald, J.K., 1993. Coagulation in drinking water treatment:
particles, organics, and coagulants. Water Science and
Technology 27 (11), 21e35.
Edzwald, J.K., 2010. Dissolved air flotation and me. Water
Research 44 (7), 2077e2106.
5440 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 2 8 e5 4 4 0
Edzwald, J.K., Tobiason, J.E., 2011. Chemical principles, source
water composition, and watershed protection, chapter 3 in
water quality and treatment. In: Edzwald, J.K. (Ed.), A
Handbook on Drinking Water, sixth ed. McGraw-Hill, New
York.
Edzwald, J.K., Van Benschoten, J.E., 1990. In: Hahn, H.H., Klute, R.
(Eds.), Aluminum Coagulation of Natural Organic Matter,
Chapter in Chemical Water and Wastewater Treatment.
Springer-Verlag, New York.
Edzwald, J.K., Upchurch, J.B., O’Melia, C.R., 1974. Coagulation in
estuaries. Environmental Science and Technology 8 (1), 58e63.
Edzwald, J.K., Becker, W.C., Wattier, K.L., 1985. Surrogate
parameters for monitoring organic matter and THM
Precursors. Journal of the American Water Works Association
77 (4), 122e131.
Freeman, S., 2009. Seawater Pretreatment Operations
Presentation at the workshop On Coagulation for Seawater and
Reuse applications, AWWA, water quality Technology
Conference, November 15, 2009, Seattle, WA.
Ghernaout, B., Ghernaout, D., Saiba, A., 2010. Algae and
cyanotoxins removal by coagulation/flocculation: a review.
Desalination and Water Treatment 20 (1e3), 133e143.
Henerdson, R., Parsons, S.A., Jefferson, B., 2008. The impact of
algal properties and pre-oxidation on solideliquid separation
of algae. Water Research 42, 1827e1845.
Millero, F.J., 1995. Thermodynamics of the carbon dioxide system
in the ocean. Geochimica et Cosmochimica Acta 59 (4), 661e677.
Nordstrom, D.K., May, H.M., 1996. Aqueous equilibrium data for
mononuclear aluminum species. In: Sposito, G. (Ed.), The
Environmental Chemistry of Aluminum. CRC Press, Boca
Raton, pp. 39e80.
Para, J., Coble, P.G., Tedetti, M., Sempéré, R., 2010. Fluorescence
and absorption properties of chromophoric dissolved organic
matter (CDOM in coastal surface waters of the Northwestern
Mediterranean Sea (Bay of Marseilles, France). Biogeosciences
Discussions 7 (4), 5675e5718.
Pernitsky, D.J., Edzwald, J.K., 2003. Solubility of polyaluminum
coagulants. Journal of Water Supply: Research and
Technology e Aqua 52 (6), 395e406.
Post, G.B., Atherholt, T.B., Cohn, P.D., 2011. Health and aesthetic
aspects of drinking water, chapter 2 in water quality and
treatment. In: Edzwald, J.K. (Ed.), A Handbook on Drinking
Water, sixth ed. McGraw-Hill, New York.
Schneider, O., 2011. Personal Communication. American Water,
Voorhees, NJ.
Sengco, M.R., 2001. The Aggregation of clay minerals and marine
Microalgal cells: Physicochemical Theory and Implications for
controlling harmful algal blooms. Doctoral dissertation,
Massachusetts Institute of Technology and Woods Hole
Oceanographic Institution.
Stumm, W., Morgan, J.J., 1996. Aquatic Chemistry, third ed. Wiley-
Interscience, NY.
Timmes, T.C., Kim, H., Dempsey, B.A., 2009. Electrocoagulation
pretreatment of seawater prior to ultrafiltration: bench-scale
applications for military water purification systems.
Desalination 249, 895e901.
Van Benschoten, J.E., Edzwald, J.K., 1990. Chemical aspects of
coagulation using aluminium salts. I. Hydrolytic reactions of
alum and polyaluminium chloride. Water Research. 24 (12),
1519e1526.
Voutchkov, N., 2010a. Seawater Pretreatment. Water Treatment
Academy, Bangkok, Thailand.
Voutchkov, N., 2010b. Considerations for selection of seawater
filtration pretreatment system. Desalination 261, 354e364.
WHO (accessed January 2011) www.who.int/water_sanitation_
health/dwq/en.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Partitioning of dissolved organic matter-bound mercury

between a hydrophobic surface and polysulfide-rubber
polymer

Eun-Ah Kim, Richard G. Luthy*

Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305-4020, USA

article info abstract

Article history: This study investigated the role of dissolved organic matter on mercury partitioning
Received 23 April 2011 between a hydrophobic surface (polyethylene, PE) and a reduced sulfur-rich surface (pol-
Received in revised form ysulfide rubber, PSR). Comparative sorption studies employed polyethylene and poly-
23 June 2011 ethylene coated with PSR for reactions with DOM-bound mercuric ions. These studies
Accepted 3 August 2011 revealed that PSR enhanced the Hg-DOM removal from water when DOM was Suwannee
Available online 11 August 2011 River natural organic matter (NOM), fulvic acid (FA), or humic acid (HA), while the same
amount of 1,3-propanedithiol-bound mercuric ion was removed by both PE and PSR-PE.
Keywords: The differences for Hg-DOM removal efficiencies between PE and PSR-PE varied depend-
Dissolved organic matter ing on which DOM was bound to mercuric ion as suggested by the PE/water and PSR-PE/
Polysulfide-rubber polymer water partition coefficients for mercury. The surface concentrations of mercury on PE
Mercury and PSR-PE with the same DOM measured by x-ray photoelectron spectroscopy were
Addition reaction similar, which indicated the comparable amounts of immobilized mercury on PE and PSR-
Adsorption reaction PE being exposed to the aqueous phase. With these observations, two major pathways for
Mercury complexation the immobilization reactions between PSR-PE and Hg-DOM were examined: 1) adsorption
of Hg-DOM on PE by hydrophobic interactions between DOM and PE, and 2) addition
reaction of Hg-DOM onto PSR by a complexation reaction between Hg and PSR. The percent
contribution of each pathway was derived from a mass balance and the ratios among
aqueous mercury, PE-bound Hg-DOM, and PSR-bound Hg-DOM concentrations. The results
indicate strong binding of mercuric ion with both dissolved organic matter and PSR poly-
mer. The FT-IR examination of Hg-preloaded-PSR-PEs after the reaction with DOM
corroborated a strong interaction between mercuric ion and 1,3-propanedithiol compared
to Hg-HA, Hg-FA, or Hg-NOM interactions.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction (Schaefer and Morel, 2009), a small organic molecule with

Methylmercury is a potent neurotoxin and has high bio-
accumulation factors (Driscoll et al., 2007). A recent study
showed an enhanced mercury methylation rate with Geobacter
sulfurreducens when mercuric ion is bound to cysteine
a thiol group. This finding implies enhanced methylation of
mercury in the presence of certain organic molecules in
sediments, which may depend on the strains of the mercury
methylating bacteria. Therefore, it is crucial to have an accu-
rate estimation of the amount of mercury that could be

* Corresponding author.
E-mail address: luthy@stanford.edu (R.G. Luthy).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.003
5442 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8
transformed into a source for methylation when there are
perturbations of the sediment organic carbon speciation due
to erosion, runoff, changes of pH/redox potentials, or sudden
input of sediment organic matter.
Since the mobile mercury is more readily bioavailable for
mercury methylation compared with solid-bound mercury
species (Benoit et al., 2001a,b; Skyllberg et al., 2009), the
mobile portion of mercury is a better proxy for methylmer-
cury concentration than the total mercury concentration.
Mercury mobilization is related the presence of other metal
binding ligands such as dissolved organic matter (DOM) and
organic or inorganic thiols/sulfides (Ravichandran, 2004;
Skyllberg, 2008). The latter would include DOM with reduced
sulfur functional groups that strongly bind mercuric ion and
prevent the mercury from precipitating as mercuric sulfide.
Dissolution of Hg-S(s) by DOM (Ravichandran et al., 1998;
Waples et al., 2005; Slowey, 2010) indicates a strong interac-
tion between DOM and Hg, and critical roles of DOM in
mercury mobilization.
The implications of strong Hg-DOM interaction are yet
unclear with respect to mercury partitioning in the presence
of a strong binding ligand for mercuric ion on a solid surface.
For example, with polysulfide-rubber (PSR) polymer, the DOM
may compete with PSR for mercury binding, or PSR may
simply provide additional binding sites for DOM-bound
mercuric ion without competition. These interactions may
be examined by preparing polysulfide-rubber polymer-coated
polyethylene (PSR-PE) and conducting competitive sorption
studies with various forms of DOM. Assuming Hg-DOM
interaction mainly comprises a HgeS bond, immobilization
of the Hg-DOM species on the reduced sulfur-rich sites can
occur via multiple HgeS bond formation (Hesterberg et al.,
2001). In this case, the Hg ion would be encapsulated in DOM
and the PSR polymer simultaneously. An exchange of
mercuric ion between DOM and PSR is also possible when
DOM-bound mercuric ion is transferred to PSR, subsequently
reaching a new equilibrium between Hg-DOM and Hg-PSR. In
summary, the possible reaction pathways between PSR-PE
and Hg-DOM can be classified as 1) adsorption of Hg-DOM
on PE via hydrophobic interaction giving a Hg-DOM-PE
species, and 2) additional bond formation between PSR and
Hg-DOM giving a Hg-PSR species.
The purpose of this study is to assess the role of dissolved
organic matter on mercury partitioning between a hydro-
phobic surface (i.e., polyethylene) and a reduced sulfur-rich
surface (i.e., polysulfide-rubber polymer). Better under-
standing of the partitioning behavior of Hg-DOM with PSR-PE
or PE is expected to provide clues to delineate the Hg-DOM
immobilization processes on PSR-PE. The results indicate
that both PE and PSR participate in Hg-DOM removal from
water, and mercury complexation with PSR polymer (6.7% of
the total PSR-PE surface area) and DOM contributes greatly to
the overall immobilization reaction on PSR-PE in the presence
of Suwannee River natural organic matter, fulvic acid, and
humic acid. Depending on the relative affinities of a hydro-
phobic surface and a polysulfide-rich surface for DOM-bound
mercuric ion, the mercury removal efficiency of a multi-
functional sorbent, for instance, PSR-coated activated carbon
(Kim et al., 2011), can be optimized by adjusting the coverage
of PSR polymer on activated carbon.
2. Materials and methods
2.1. Polymer coating on polyethylene strips
Polyethylene (PE) strips were pre-cleaned with methylenech-
loride, methanol, and DI-water consecutively for one day at
each step. The PE was dried in a convection oven at 60
C for
4 h and cut into 2 cm  2 cm (18  0.5 mg) pieces. The
polysulfide-rubber polymer was synthesized following the
procedure described by Kalaee et al. (2009). Condensation
polymerization between sodium tetrasulfide and 1,2-
dichloroethane, using methytributyllammonium chloride as
a phase transfer catalyst, produced a yellowish elastic solid in
water. One hundred mg of the PSR polymer in 40 mL toluene
was refluxed until the polymer block was completely dis-
solved. The solution was cooled to room temperature before it
was used for coating PE strips. A piece of PE was dipped into
the PSR solution for less than 1 min and taken out for an
immediate drying in air for 10 s. The PSR-PE was dried again
under vacuum for 1 h.
The sulfur content of the polymer-coated polyethylene
strip was determined in duplicate by elemental analysis
(Atlantic Microlab, GA).
2.2. Surface reactions of Hg-DOM on PE or PSR-PE
Suwannee River natural organic matter (NOM), and fulvic
acid (FA) were obtained from the International Humic
Substance Society (IHSS). Humic acid (HA) and 1,3-
propanedithiol were purchased from Sigma Aldrich and
Alfa Aesar. Because the source and the isolation method for
Sigma Aldrich HA are different from those for IHSS HA, the
differences between FA and HA reported in this paper do not
necessarily represent any FA and HA properties in a specific
natural system. Suwannee River NOM, Suwannee River fulvic
acid (FA) and Sigma Aldrich humic acid (HA) were dissolved in
250 mL borosilicate glass bottles to make 10 mg DOM L1
solutions. Suwannee River NOM and FA dissolved in water
readily, but the HA solution was sonicated until its complete
dissolution. The solutions were filtered through 0.45 mm
polyvinylidene fluoride (PVDF) membrane filters to remove
particulate matter. An aqueous solution of 1,3-propanedithiol
was freshly prepared by dissolving 1,3-propanedithiol in 1 M
NaOH and diluted 400 fold to make 5.41 mg L1 solution. HgCl2
stock solution in 1 M HCl was added to each solution of dis-
solved organic matter (DOM) to make 10 ppb (50 nM) Hg
solutions. Because Suwannee River NOM, FA, and HA have
0.65 wt%, 0.44 wt% and 0.96 wt% sulfur respectively, accord-
ing to the elemental analysis results reported by IHSS and
Pitois et al. (2008), 10 mg L1 DOM solutions have 1.4e3.0 mM
sulfur, which is an excess amount for mercury binding in
50 nM Hg solutions. The pH was adjusted to 7 with 0.01M
potassium monophosphate buffer. The mixtures of mercuric
ion and DOM were shaken for one week to allow sufficient
reaction time for forming Hg-DOM. One piece of PE or PSR-PE
was in contact with 40 mL of these solutions for 4 weeks. The
PE or the PSR-PE strip was taken out, washed with DI-water,
and gently pressed on Kimwipes to remove water on a PE or
a PSR-PE strip. PSR-PEs after the reaction with Hg-DOM were
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8 5443

washed with MilliQ water followed by drying in vacuum

for 1 h. The remaining aqueous solution was preserved by 3. Results and discussion
adding 400 mL of BrCl.
3.1. Synthesis of PSR-PE and proposed reaction
pathways
2.3. DOM contact with Hg pre-loaded PSR-PE
The acronyms, definitions, and units for the terms used in this
HgCl2 stock solution in 1 M HCl was used to make 10 ppm Hg paper are summarized in Table 1. The synthesis procedure for
solution in 0.01M potassium monophosphate buffer to PSR-PEs involves a solution casting of PE strips with PSR
maintain the pH at 7. One piece of PSR-PE was in contact with polymer solution in toluene, and drying in vacuum to produce
40 mL of 10 ppm Hg solutions for 1 week. The PSR-PE strips a thin layer of PSR polymer over the PE surface as depicted in
were taken out, washed with DI-water, and gently pressed on Fig. 1. The surface areas of PSR and PE are assumed to be
Kimwipes to remove water on the PSR-PE strip. The Hg- proportional to the atomic counts of the constituting atoms (C
loaded PSR-PE was then in contact with 40 mL of 10 mg L1 and S for PSR, and C for PE) based on the XPS analysis of PSR-
NOM, FA, HA, or 5.41 mg L1 1,3-propanedithiol for 4 weeks. PE before the reactions with DOM-bound Hg. The PSR is
The PSR-PE strips were taken out after the reactions, washed constituted of the repeating eC2S4- segments, and with the
with DI-water, and gently pressed on Kimwipes to remove average atomic % of sulfur on PSR-PE at 4.0 atom%, the
water on the PE strips. The PSR-PE strips were dried for 1 h in contribution of PSR to the total carbon atomic count is
vacuum and analyzed by Fourier transform infra-red spec- approximately 2.0 atom%. Accounting for the different atomic
troscopy (FT-IR) and x-ray photoelectron spectroscopy (XPS). radii of sulfur and carbon, 0.109 nm and 0.091 nm respectively,
The remaining aqueous solution was preserved by adding the average fractions of PSR (fPSR) and PE (fPE) on a PSR-PE strip
400 mL of BrCl to the reaction vessels. were determined as fPSR ¼ 0.067, and fPE ¼ 0.933. The possible
reactions of DOM-bound mercuric ion with a PE strip or a PSR-
PE strip are illustrated in Fig. 2. Each Hg-DOM species (NOM,
2.4. Total mercury analysis FA, HA, or 1,3-propanedithiol) was reacted with PE or PSR-PE
separately for comparison. As illustrated schematically in
Four mL of preserved duplicate samples was filtered through Fig. 2, Hg-PSR and Hg-DOM-PE are proposed as the products of
0.45 mm polyvinylidene fluoride (PVDF) membrane filter. The the two possible reaction pathways after the DOM-bound
filtrate was diluted to ensure mercury concentrations were mercuric ion encounters PSR-PE. These reactions represent
within the detection range of 0.5e400 ng L1. The total the sorption of Hg-DOM to the surface via direct Hg-sulfur
mercury concentrations were measured by Tekran 2600 cold interactions (Hg-PSR) or indirectly by DOM-PE hydrophobic
vapor atomic fluorescent spectrometry (CVAFS) following the interactions (Hg-DOM-PE).
US EPA (Environmental Protection Agency) method 1631 The reaction represented by Hg-PSR differs from that for
revision E. Hg-DOM-PE in terms of where a mercuric ion is situated.
Mercuric ion binds both with DOM and polysulfide in Hg-PSR,
after which DOM may be detached from mercuric ion.
2.5. X-ray photoelectron spectroscopy (XPS)
However, in the case of the Hg-DOM-PE pathway, which
utilizes DOM hydrophobic interaction with polyethylene,
XPS techniques were used to determine the surface concen-
trations of sulfur and mercury on PSR-PE, PE, or Hg-preloaded
PSR-PE before and after the reactions with Hg-DOM or DOM.
Three spots for each sample were analyzed to account for
Table 1 e Summary of the definitions and units for the
non-uniform distributions of sulfur and mercury atoms on
terms that describe the reaction of mercury species with
the PSR-PE surface. PHI 5000 Versa-Probe scanning XPS
polyethylene (PE) and polysulfide-rubber coating on
microprobe with Al Ka x-ray radiation (1486 eV) was used polyethylene (PSR-PE).
under high vacuum condition (below 105 Pa). Charging
Acronym Definition Unit
effects by the poor surface conductivity were minimized by
applying 10 eV argon ions. Analytical sample size for the [Hg-DOM] total aqueous mercury concentration mg Hg L1
[Hg-PSR] mercury associated with PSR per unit mg Hg m2
survey scans was 1  1 mm. An averaged spectrum from five
area of PSR-PE
survey scans over 0e1000 eV was obtained with a resolution
[Hg-DOM-PE] mercury associated with PE via DOM mg Hg m2
of 1 eV. per unit area of PSR-PE
[Hg]sorbed total mass of mercury sorbed per unit mg Hg m2
area of PSR-PE
2.6. FT-IR analysis of PSR-PE Hgtot total mass of mercury in the water mg Hg
and PSR-PE
Far infra-red (FIR) spectra of Hg-pre-loaded PSR-PE (Hg-PSR- Vaq volume of water L
PE) after the reaction with DOM were obtained with Bruker SAPE PE area on PSR-PE m2
SAPSR PSR area on PSR-PE m2
Vertex 70 FT-IR spectrometer using a deuterated triglycine
SAPSR-PE area of PSR-PE m2
sulfate (DTGS) detector. A piece of PSR-PE was placed where
fPSR SAPSR/SAPSR-PE e
the IR beam passes through perpendicularly. Forty scans were fPE SAPE/SAPSR-PE e
averaged for each spectrum.
5444 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8

a Synthesis ofpolysulfide-rubber-coated polyethylene (PSR-PE)

2cm X 2cm X 51µm

b Reactions with dissolved organic matter (DOM)-bound mercury

Fig. 1 e a) Synthesis of polysulfide-rubber-coated polyethylene (PSR-PE) with a piece of PE and PSR polymer solution in
toluene, showing a schematic distribution of sulfur atoms on PSR-PE, b) reactions between Hg-DOM (NOM, FA, HA, and 1,3-
propanedithiol) and PE or PSR-PE. The mercury ion is not necessarily situated at the exterior of NOM, FA or HA.

mercuric ion does not form a chemical bond with polysulfide.
The total aqueous mercury concentration measured after
the reaction between PSR-PE and Hg-DOM is defined as [Hg-
DOM]. Due to the 1000-fold mass ratio of DOM to Hg
and the high stabilization constants for the association
reactions between mercuric ion and DOM (Benoit et al.,
2001a,b; Khwaja et al., 2006; Miller et al., 2009; Dong et al.,
2010), DOM-bound mercuric ion would be the major constit-
uent of total mercury concentration in the aqueous phase.
[Hg]sorbed can be defined by using a mass balance with the
total aqueous mercury concentrations before and after the
reaction with DOM-bound mercuric ion (equation (1), (2)).
the reaction vessel (glass vial) because BrCl, a strong oxidant,
was directly spiked to the vial after PSR-PE strip was removed
so that BrCl would oxidize and extract any trace mercury from
the glass wall and the lid. However, such mercury partitioning
on the glass vial was negligible because the mercuric ion loss
by the same glass vial with the same or higher DOM concen-
tration was not significant according to the previous experi-
mental data not reported here. Hg-DOM-PE.
3.2. Surface reaction of DOM-bound mercuric ion on
PSR-PE

Hgtot ¼ Vaq ,½Hg  DOM þ SAPSRPE ,½Hgsorbed


½Hgsorbed ¼ ½Hg  PSR þ Hg  DOM  PE
By our experimental methods, [Hg-DOM] includes the
amount of aqueous phase mercury and any mercury loss by
(1) In order to estimate overall Hg-DOM removal efficiencies by
PSR-PE, dissolved mercury concentrations after the reactions
were measured and compared with the Hg-DOM solutions
(2)
without any sorbents. Because dissolved humic substances,
especially those with reduced sulfur groups are the main
competitors for mercuric ion, we tested dissolved natural
organic matter, fulvic acid, humic acid, and 1,3-
propanedithiol (PDT) as representatives of reactive constitu-
ents in sediment pore water. The results are depicted in Fig. 3,
showing a strong partitioning of PDT-bound mercuric ion as
well as HgCl2 by PSR-PE. The partition coefficient K1 is defined
as the ratio of the total mercury concentration on PSR-PE
surface to the total aqueous mercury concentration (equa-
tion (3)).

½Hgsorbed ½Hgsorbed Hgtot  Vaq,½Hg  DOM


Fig. 2 e Possible interactions between DOM-bound K1 ¼ h ¼ Lm2
½Hgaqueous ½Hg  DOM SAPSRPE ,½Hg  DOM
mercury (NOM, FA, HA, or 1,3-propanedithiol) and PSR-PE,
(3)
which results in either PSR-bound Hg (Hg-PSR) via an
addition reaction as illustrated by the first pathway, or PE- Thus, the overall removal processes incorporated in K1
bound Hg-DOM (Hg-DOM-PE) via hydrophobic adsorption represent the combination of the two proposed reaction
reaction on the PE surface as illustrated by the second pathways. As shown in Fig. 3, the partition coefficients vary
pathway. from 13 to 115 L m2.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8
150
100
K1[L m-2]
50
0
NONE NOM FA HA
Additives (DOM)
Fig. 3 e Partition coefficients (K1) of mercury between PSR-
PE and water as results from the reactions between Hg-
DOM and PSR-PE. These K1 values correspond to 21e70%
removal of the initial Hg-DOM in 40 mL vials with 8 cm2
PSR-PE having 6.7% polysulfide rubber.
3.3. Reactions between PE strip and DOM-bound
mercuric ion
Because the complexation of mercuric ion with DOM itself
transforms ionic mercury into a more hydrophobic species,
namely Hg-DOM, the PE surface can provide sorption sites for
Hg-DOM via a favorable DOM-PE hydrophobic interaction, i.e.,
the second pathway in Fig. 2. The PE/water partition coeffi-
cient with Hg-DOM gives a good criterion to estimate how
much Hg-DOM sorption on PE alone contributes to the overall
reaction between PSR-PE with Hg-DOM. A hydrophobic
partition coefficient assigned for this reaction equals the ratio
of [Hg-DOM-PE] to the total aqueous mercury species (equa-
tion (4))
½Hg  DOM  PE fPE ½Hg  DOM  PEPEonly
K2 h ¼ (4)
½Hg  DOM ½Hg  DOM
where Hg-DOM-PE denotes PSR-PE-bound Hg-DOM due to
a favorable hydrophobic interaction between DOM and PE. As
shown in Fig. 4 and 1,3-propanedithiol-bound mercuric ion
has a high affinity for PE, which indicates a significant
encapsulation of mercuric ion with a hydrophobic bidentate
ligand, and a large contribution of the adsorption reaction to
the overall reaction between PSR-PE and Hg-PDT. In contrast,
the mercuric ion complexed with Suwannee River natural
organic matter, fulvic acid, or humic acid as well as HgCl2
does not exhibit high removal efficiency by PE alone (Fig. 4)
compared with that by PSR-PE (Fig. 3). These differences
suggest the overall reactions between PSR-PE and Hg-DOM
(NOM, FA, or HA) comprise other reactions than the hydro-
phobic adsorption reactions between DOM and the PE itself.
3.4. Comparison on the atomic % of mercury on PSR-PE
and PE surfaces after the reactions with Hg-DOM
The surface concentrations of mercury on PSR-PE and PE
were measured after the reaction with Hg-DOM. The XPS
5445
200
150
K2 [L m-2 ]
100
50
0
PDT NONE NOM FA HA PDT
Additives (DOM)
Fig. 4 e Partition coefficients (K2) of mercury ion between
PE alone and water after the reaction with Hg-DOM. In
comparison to PSR-PE/water partitioning of Hg-DOM
shown in Fig. 3, these data indicate relatively weak
hydrophobic interactions between Hg-DOM and PE except
for 1,3-propanedithiol. The columns represent the average
values of the triplicates, and the error bars correspond to
the standard deviations. These K2 values correspond to
5e72% removal of the initial DOM-bound Hg.
analyses were used to estimate the proportion of the easily
accessible (i.e., not significantly covered by organic matter)
mercury among the total immobilized mercury on PSR-PE.
Because XPS is a surface sensitive technique that measures
the atomic compositions in the top w10 nm layer, any
deposition of organic substances over mercuric ion will shield
the escaping electron and reduce the signal. Therefore, we
can qualitatively estimate the extent of the shielding effect
from DOM or PSR by comparing the total immobilized
mercury concentrations on PSR-PE or PE as exhibited in Fig. 3
or Fig. 4 and the XPS results as shown in Fig. 5. Whereas the
Hg-DOM removal efficiency of PSR-PE (Table S1) ranges from
96% to 531% of the same Hg-DOM removal by PE, the esti-
mated surface concentrations of mercury on PSR-PE and PE
(Fig. 5) after the reactions with the same Hg-DOM species are
not significantly different with each other. This indicates that
Hg-DOM removal by PSR-PE involves partial covering of Hg
ions with DOM, or migration of mercuric ion into the inner
PSR layer.
3.5. Overall Hg-DOM immobilization reaction pathways
with PSR-PE
Table 2 summarizes two partition coefficients obtained from
the two sets of surface reactions, one with PSR-PE and Hg-
DOM, and the other with PE and Hg-DOM. Since K1 and K2
formulate two equations for three unknowns, the two classes
of sorbed mercury concentrations, [Hg-PSR] and [Hg-DOM-PE],
can be expressed as functions of the aqueous mercury
concentration, [Hg-DOM]. With these constants and the
following reaction model equations (5)e(7),
f add : PSR  PE þ Hg  DOM/Hg  PSR (5)
5446 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8

0.5
0.45
0.4
0.35

Atomic %
0.3
0.25
0.2 Hg on PE
0.15 Hg on PSR-PE
0.1
0.05
0
NOM FA HA PDT
Additives (DOM)

Fig. 5 e X-ray photoelectron spectroscopic analysis of the surface concentrations (atomic %) of mercury on PE or PSR-PE after
reaction with Hg-DOM. The columns represent the average values of the triplicates, and the error bars correspond to the
standard deviations.

f ads : PSR  PE þ Hg  DOM/Hg  DOM  PE (6) f add þ f ads ¼ 1 (10)

The definitions of the reaction constants and the solutions

Overall reaction:
for fadd and fads are summarized in Table 3. The results shown
PSR  PE þ Hg  DOM/f add Hg  PSR þ f ads Hg  DOM  PE (7) in Table 4 reveal that the reaction of Hg-DOM on PSR-PE is
mainly by the addition reaction, i.e., via pathway 1, with
we can calculate how much each reaction pathway contrib- complexation of mercuric ion with PSR, except for the case
utes to the overall reaction. By definition, a contribution factor with 1,3-propanedithiol. The hydrophobic partitioning reac-
(fadd, or fads) means the fractions of addition (via pathway 1) or tion has the highest importance in the PDT-mediated mercury
adsorption (via pathway 2) reaction to the overall reaction, sorption on PSR-PE.
which can be defined as a ratio of each corresponding reaction The dependence of the major reaction pathway on DOM
product concentration to the total surface mercury concen- may stem from the Hg-DOM binding strength or the bulkiness
tration (equations (8)e(10)). of Hg-DOM, which varies with DOM. Suwannee River NOM,
FA, and HA solutions have excess amounts of sulfur
½Hg  PSR (1.4e3.0 mM) for mercury binding in 50 nM Hg solutions, and
f add ¼ (8) the proportion of reduced sulfur groups in river DOM ranges
½Hgsorbed
from 13 to 36% of the total sulfur (Ravichandran, 2004).
Therefore, the amount of reduced sulfur atoms in the DOM
½Hg  DOM  PE solutions would not affect the overall affinity of DOM for
f ads ¼ (9)
½Hgsorbed mercuric ion. Instead, the interaction of mercuric ion with

Table 2 e Summary of the partition coefficients K1 and K2.

DOM K1 (errora) K2 (error) Table 3 e Summary of the partition coefficients and the
[L m2] [L m2] solutions of the contribution factors, fadd, and fads,
expressed in terms of the defined constants, K1, and K2.a
None 86.6 (4.3) 6.61 (0.66)
Definitions of K1, K2 Solutions of fadd, fads
Suwannee River 13.2 (6.4) 4.15 (0.50)
natural organic matter ½Hgsorbed ½Hg  PSR ðK1  K2 Þ
hK1 fadd ¼
 ¼ ¼ 1  fads
Suwannee River fulvic acid 38.2 (17.7) 4.15 (0.53) ½Hg  DOM Hg sorbed K1
Humic acid 14.8 (2.2) 2.71 (0.89)
1,3-propanedithiol 115 (5.4) 122 (32.1) ½Hg  DOM  PE ½Hg  DOM  PE K2
hK2 fads ¼ ¼
½Hg  DOM ½Hgsorbed K1
a The errors in K1 and K2 are
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
   ffi a K1 is derived from the reactions of Hg-DOM with PSR-PE; K1 is the
error in½Hgsorbed 2 error in½Hg  DOM 2
K1 þ , ratio of the total surface mercury concentration to the total
½Hgsorbed ½Hg  DOM
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi aqueous mercury concentration. K2 is a partition coefficient of Hg-
 2  2
error in½Hg  PE error in½Hg  DOM DOM between water and PE obtained from the Hg-DOM adsorption
K2 þ respectively
½Hg  PE ½Hg  DOM study with the PE strips; K2 from the Hg-DOM reaction with pure PE
(Valcárcel, 2000). is multiplied by fPE to reflect the PE surface area on PSR-PE.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8 5447

3.6. Changes in the chemical bonding between PSR and

Table 4 e Relative significance of the addition via
pathway 1 (fadd), and the hydrophobic adsorption via
pathway 2 (fads) reactions for Hg-DOM partitioning to PSR-
PE having 6.7% polysulfide.
DOM fadd (error ) a
None 0.924 (0.008)
Suwannee River 0.686 (0.157)
natural organic matter
Suwannee River fulvic acid 0.891 (0.052)
Humic acid 0.817 (0.066)
1,3-propanedithiol 0 (0.283)
a The error in fadd ¼ 1 e fads equals to the corresponding error in
fads.
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
error in K1 error in K2
b The error in fads ¼ fads $ ð Þ2 þð
K1
2000).
c The algebraic value for fads ¼ K2/K1 is 1.06. It was approximated to
the maximum fads value, 1.0.
a reduced sulfur group in a DOM molecule and other adjacent
reactive binding sites in DOM would determine Hg-DOM
binding strength. The different DOM molecule sizes can
also explain the difference in fadd because the limited
PSR surface area (0.536 cm2) can only accommodate
4.5  10105.6  1016 mol of Hg-DOM out of 2  109 mol of
the total Hg-DOM when Hg-DOM is assumed to be a sphere
with diameter 0.5e450 nm, where 0.5 nm corresponds to the
approximate size of one 1,3-propanedithiol molecule, and
450 nm corresponds to the filter pore size that the NOM, FA
and HA molecules passed in preparation of the DOM solu-
tions. Therefore, when the surface area is the limiting factor,
a smaller Hg-DOM molecule would be able to react more
extensively with PSR regardless of the Hg-DOM bond strength,
and fadd may not provide information about the relative
binding strength of Hg-DOM among the tested DOM. In this
case, an analogous experiment with excess surface area of
PSR for accommodating Hg-DOM may reveal the relative
Hg-DOM bond strength.
mercuric ion in the presence of DOM
The reaction between Hg-preloaded PSR-PE with DOM
fads (error ) b provides information about the Hg-PSR bond strength, as DOM
can dissolve Hg ions from PSR. Mercuric chloride was pre-
0.076 (0.008)
loaded on PSR-PE, which produced Hg-PSR-PE. This surface
0.314 (0.157)
was then reacted with DOM for one month. As the addition
0.109 (0.052) reaction proceeds, the PSR-PE surface would be covered with
0.183 (0.066) DOM, and Hg-DOM could be dissolved into the water
1.0c (0.283) depending on the Hg-DOM binding strength. It is also possible
to have a subsequent migration of DOM-bound mercuric ion
toward the hydrophobic PE surface. Our data on the aqueous
mercury concentration after the reaction with each DOM
Þ2 (Valcárcel,
K2 (Table S2) reveal that the total amount of the immobilized
mercury remaining on PSR-PE is close to that without DOM,
which indicates an insignificant amount of remobilization of
PSR-PE-bound mercuric ion by DOM.
One way to assess the amount of Hg-PSR bond breakage
or weakening by DOM is to analyze FT-IR spectra around
350e390 cm1 for the characteristic IR peaks for Hg-S (Al-
Jeboori et al., 2010). As shown in Fig. 6, without any DOM,
the major peak for Hg-S appears at 354 cm1, which is lower
than the emerging peaks at around 376 cm1 in the presence
of DOM. This change indicates that new Hg-S bonds are
formed with DOM and existing bonds between polysulfide
and mercuric ion are weakened or broken. The extent of the
change is most prominent with 1,3-propanedithiol, which
has two highly reactive thiol groups per molecule to
compete with PSR for mercuric ion. According to the
decrease in the absorbance intensities at 354 cm1 with
respect to those at 376 cm1 as shown in Fig. 6, the extent of
perturbation to Hg-PSR bond is comparable among NOM, FA,
and HA. A diminished binding strength of Hg-PSR does not
necessarily imply a lower stability of the immobilized
mercuric ion unless Hg-DOM formation results in its trans-
port into the water. The favorable interaction between DOM
and hydrophobic PE surface or fouling of Hg-PSR by the DOM

0.35

0.3 Hg-PSR-PE-buffer

0.25 Hg-PSR-PE-fulvic acid

Absorbance

0.2
Hg-PSR-PE-humic acid
0.15
Hg-PSR-PE-NOM
0.1
Hg-PSR-PE-
0.05 propanedithiol

0
320 340 360 380 400 420

Wave number (cm-1)

Fig. 6 e FT-IR spectra (far-infrared region) of Hg-PSR-PE after the reaction with various DOM for one month. Hg-PSR-PE-
buffer represents a control group that has no dissolved organic matter.
5448 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 1 e5 4 4 8
could compensate for any diminishment of Hg-PSR bonding
and keep the immobilized mercury from being released into
the solution.
4. Conclusion
A comparison of Hg-DOM interactions with a hydrophobic
surface (PE) and a sulfur-rich surface (PSR-PE) showed the
stronger interactions of Hg-DOM with PSR-PE. A greater
amount of Hg-DOM was removed by PSR-PE than PE when
NOM, FA, or HA was present in the solution, during which
DOM accumulation over mercury or migration of mercuric ion
into the inner PSR layer occurred. This implies a significant
contribution of the PSR-mediated reaction to the overall Hg-
DOM immobilization reaction. The organic compound, 1,3-
propanedithiol, was examined as a strong competing ligand
for mercuric ion, and 1,3-propanedithiol-bound mercuric ion
was most effectively removed via an adsorption reaction
pathway. The changes in the relative peak intensities at
354 cm1 and 376 cm1 in the FT-IR spectra of HgCl2-preloaded
PSR-PE after the reaction with DOM suggest that partial
breakage or weakening of PSR-Hg bonds took place with
additional complexation with DOM. The magnitude of the
change is most prominent with 1,3-propanedithiol. Since both
PSR and PE participate in Hg-DOM removal from water, it is
beneficial to develop a multi-functional sorbent that has high
affinities for both DOM and mercuric ions in order to achieve
high mercury removal efficiencies in sediments. Depending
on the fadd and fads values for the Hg-DOM of concern, different
ratios of hydrophobic/reduced-sulfur-rich sorbent can be
determined for an optimum Hg-DOM removal efficiency.
Acknowledgments
The authors acknowledge the National Institute of Environ-
mental Health Sciences (NIEHS), grant number R01 ES016143-02,
for the financial support of this study.
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.003.
references
Al-Jeboori, M.J., Al-Tawel, H.H., Ahmad, R.M., 2010. New metal
complexes N2S2 tetradentate ligands: synthesis and spectral
studies. Inorganica Chim. Acta 363 (6), 1301e1305.
Benoit, J.M., Gilmour, C.C., Mason, R.P., 2001a. The influence of
sulfide on solid-phase mercury bioavailability for methylation
by pure cultures of Desulfobulbus propionicus (1pr3). Environ.
Sci. Technol. 35 (1), 127e132.
Benoit, J.M., Mason, R.P., Gilmour, C.C., Aiken, G.R., 2001b.
Constants for mercury binding by dissolved organic matter
isolated from the Florida Everglades. Geochimica et
Cosmochimica Acta 65 (24), 4445e4451.
Dong, W., Liang, L., Brooks, S., Southworth, G., Gu, B., 2010. Roles
of dissolved organic matter in the speciation of mercury and
methylmercury in a contaminated ecosystem in Oak Ridge,
Tennessee. Environ. Chem. 7 (1), 94e102.
Driscoll, C.T., Lambert, K.F., Kamman, N., Holsen, T., Han, Y.-J.,
Chen, C., Coodale, W., Butler, T., Clair, T., Munson, R., 2007.
Mercury matters-linking mercury science and public policy in
the Northeastern United States. Sci. Links Publ. 1 (3), 11.
Hesterberg, D., Chou, J.W., Hutchison, K.J., Sayers, D.E., 2001.
Bonding of Hg(II) to reduced organic sulfur in humic acid as
affected by S/Hg ratio. Environ. Sci. Technol. 35 (13),
2741e2745.
Kalaee, M.R., Famili, M.H.N., Mahdavi, H., 2009. Synthesis and
characterization of polysulfide rubber using phase transfer
catalyst. Macromol. Symp. 277 (1), 81e86.
Khwaja, A.R., Bloom, P.R., Brezonik, P.L., 2006. Binding constants
of divalent mercury (Hg2þ) in soil humic acids and soil organic
matter. Environ. Sci. Technol. 40 (3), 844e849.
Kim, E.A., Seyfferth, A.L., Fendorf, S., Luthy, R.G., 2011.
Immobilization of Hg(II) in water with polysulfide-rubber
(PSR) polymer-coated activated carbon. Water Res. 45 (2),
453e460.
Miller, C.L., Southworth, G., Brooks, S., Liang, L., Gu, B., 2009.
Kinetic controls on the complexation between mercury and
dissolved organic matter in a contaminated environment.
Environ. Sci. Technol. 43 (22), 8548e8553.
Pitois, A., Abrahamsen, L.G., Ivanov, P.I., Bryan, N.D., 2008. Humic
acid sorption onto a quartz sand surface: a kinetic study and
insight into fractionation. J. Colloidal Interface Sci. 325 (1),
93e100.
Ravichandran, M., 2004. Interaction between mercury and
dissolved organic matter e a review. Chemosphere 55 (3),
319e331.
Ravichandran, M., Aiken, G.R., Reddy, M.M., Ryan, J.N., 1998.
Enhanced dissolution of cinnabar (mercuric sulfide) by
dissolved organic matter isolated from the Florida Everglades.
Environ. Sci. Technol. 32 (21), 3305e3311.
Schaefer, J.K., Morel, F.M.M., 2009. High methylation rates of
mercury bound to cysteine by Geobacter sulfurreducens. Nat.
Geosci. 2, 123e126. Feb.
Skyllberg, U., 2008. Competition among thiols and inorganic
sulfides and polysulfides for Hg and MeHg in wetland oils and
sediments under suboxic conditions: illumination of
controversies and implications for MeHg net production. J.
Geophys. Res. 113, G00C03.
Skyllberg, U., Westin, M.B., Meili, M., Björn, E., 2009. Elevated
concentrations of methylmercury in streams after forest
clear-cut: a consequence of mobilization from soil or new
methylation? Environ. Sci. Technol. 43 (22), 8535e8541.
Slowey, A.J., 2010. Rate of formation and dissolution of mercury
sulfide nanoparticles: the dual role of natural organic matter.
Geochim. Cosmochim. Acta 74 (16), 4693e4708.
Valcárcel, M., 2000. Principles of analytical chemistry, a textbook:
2.5 Basic analytical properties. Springer-Verlag, Berlin
Heidelberg.
Waples, J.S., Nagy, K.L., Aiken, G.R., Ryan, J.N., 2005. Dissolution of
cinnabar (HgS) in the presence of natural organic matter.
Geochim. Cosmochim. Acta 69 (6), 1575e1588.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Seawater quality and microbial communities at a desalination

plant marine outfall. A field study at the Israeli Mediterranean
coast

Dror Drami a,b, Yosef Z. Yacobi c, Noga Stambler d, Nurit Kress a,*
a
Israel Oceanographic and Limnological Research, National Institute of Oceanography, P.O. Box 8030, Tel Shikmona, Haifa 31080, Israel
b
The Porter School of Environmental Studies, Tel Aviv University, Israel
c
Israel Oceanographic and Limnological Research, Kinneret Limnological Laboratory, Israel
d
Mina & Everard Goodman Faculty of Life Sciences, Bar Ilan University, Israel

article info abstract

Article history: Global desalination quadrupled in the last 15 years and the relative importance of seawater
Received 10 April 2011 desalination by reverse osmosis (SWRO) increased as well. While the technological aspects
Received in revised form of SWRO plants are extensively described, studies on the environmental impact of brine
2 August 2011 discharge are lacking, in particular in situ marine environmental studies. The Ashqelon
Accepted 5 August 2011 SWRO plant (333,000 m3 d
1 freshwater) discharges brine and backwash of the pre-
Available online 11 August 2011 treatment filters (containing ferric hydroxide coagulant) at the seashore, next to the
cooling waters of a power plant. At the time of this study brine and cooling waters were
Keywords: discharged continuously and the backwash discharge was pulsed, with a frequency
Seawater desalination dependent on water quality at the intake. The effects of the discharges on water quality
Reverse osmosis and neritic microbial community were identified, quantified and attributed to the different
Environmental impact discharges. The mixed brine-cooling waters discharge increased salinity and temperature
Brine discharge at the outfall, were positively buoyant, and dispersed at the surface up to 1340 m south of
the outfall. Nutrient concentrations were higher at the outfall while phytoplankton
densities were lower. Chlorophyll-a and picophytoplankton cell numbers were negatively
correlated with salinity, but more significantly with temperature probably as a result of
thermal pollution. The discharge of the pulsed backwash increased turbidity, suspended
particulate matter and particulate iron and decreased phytoplankton growth efficiency at
the outfall, effects that declined with distance from the outfall. The discharges clearly
reduced primary production but we could not attribute the effect to a specific component
of the discharge. Bacterial production was also affected but differently in the three surveys.
The combined and possible synergistic effects of SWRO desalination along the Israeli
shoreline should be taken into account when the three existing plants and additional ones
are expected to produce 2 Mm3 d
1 freshwater by 2020.
ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: þ972 4 8515202; fax: þ972 4 8511911.

E-mail addresses: nuritblu@gmail.com, nurit@ocean.org.il (N. Kress).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.005
5450 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
1. Introduction
Desalination to provide potable water is increasing world-
wide. The total global production capacity increased from 17.3
to 68 million m3day
1 (Mm3 d
1) from 1994 to 2009 and is
projected to reach 130 Mm3 d
1 by 2016 (Water-world, 22nd-
GWI/IDA; Yermiyahu et al., 2007). The relative contribution
of desalination by Reverse Osmosis (RO) of brackish and
seawater (SW) stands at 44% (Greenlee et al., 2009). A combi-
nation of large plants, modern membrane technology and
improved energy systems have reduced the cost of SWRO
desalinated water by ca. 3 times in the last 15 years (Campbell
and Jones, 2005; Service, 2006; Tal, 2006; Fritzmann et al.,
2007). Therefore, the relative importance of SWRO is pro-
jected to increase globally. In Israel, three SWRO plants
produce ca. 700,000 m3 d
1 freshwater, 17% of the total potable
water resources in the country. By 2020, SWRO is planned to
supply more than 30% of Israel’s freshwater needs (Dreizin
et al., 2008).
The RO process uses high pressure to force water mole-
cules through a semi-permeable membrane that retains the
salts (up to a maximum of 50% conversion factor for SWRO),
producing freshwater and brine. However, the process is not
just a simple separation between water and salts. Chemicals
are used at the various stages of the desalination process and
may include: a) coagulants in the pre-treatment stage (iron or
aluminum salts, polymers); b) biocides (such as chlorine) and
neutralizers (sodium sulfite); c) antiscalants to prevent fouling
of the membranes (polyphosphates, polyphosphonates, pol-
yacrylic acid, polymaleic acid); d) cleaning solutions for RO
membranes (acidic and alkaline solutions and detergents);
and e) pH and hardness adjustors for the product water (lime)
(NRC, 2008; UNEP, 2008). These chemicals are disposed off
mostly with the brine. Until recently, the extensive literature
on desalination did not address the environmental impacts
associated with the process but focused on plant planning,
site selection and construction, operational aspects, techno-
logical improvements, and energetic cost. It was assumed that
when properly engineered and constructed, brine discharge is
environmentally safe (Ahmed et al., 2001; Campbell and Jones,
2005; Alpert et al., 2007; Lattemann and Hopner, 2008; Safrai
and Zask, 2008; Shannon et al., 2008).
Environmental impacts are now addressed in the literature
but mostly as a theoretical analysis that include entrainment
and impingement of organisms at the intake; and at the
outfall, increased salinity and stratification, reduced vertical
mixing, decrease in oxygen concentration, increased
turbidity, euthrophication, decreased or increased produc-
tion, toxicity, mitigation techniques (UNEP/MAP/MEDPOL,
2003; Fritzmann et al., 2007; NRC, 2008; UNEP, 2008; Khan
et al., 2009; Lattemann, 2010). A large proportion of the pub-
lished work is descriptive and provides little quantitative data.
The number of published articles with actual measurements
of effects in situ or in lab experiments is small and limited in
scope (Roberts et al., 2010). Most of the publications empha-
size the effects of salinity on the benthic communities and
those are site and organism specific (Fernández-Torquemada
et al., 2005; Raventos et al., 2006; Gacia et al., 2007; Sánchez-
Lizaso et al., 2008).
In this study we aim to show the effects of the discharges
of Ashqelon’s (Israel) SWRO plant on seawater quality and on
the marine microbial community at the discharge site. To the
best of our knowledge this is the first paper to present the
effects of desalination effluents on the neritic marine envi-
ronment. The motivation of this research was an unexpected
red plume observed at the outfall site, attributed to the
pulsed discharge of ferric hydroxide used as a coagulant in
the pre-treatment stage. This notable esthetic effect moti-
vated the regulator (Safrai and Zask, 2008) to initiate this
research despite claims that iron discharge has no ecological
consequences.
2. Study site
The Ashqelon SWRO plant is located at the southern Medi-
terranean coast of Israel (Fig. 1). It started to operate in 2005
and now produces (2011) 330,000 m3 d
1 of freshwater, one
of the largest SWRO plants in the world. The detailed plan-
ning and operation of the plant were described previously
(Kronenberg, 2004; Sauvet-Goichon, 2007) as were some of
the possible adverse environmental aspects (Einav et al.,
2002; Einav and Lokiec, 2003; Safrai and Zask, 2008). At the
time of this study (2008e2009) the average seawater intake,
conversion factor and salinity were: 737,000 m3 d
1, 43.5%.
and 75.3, respectively. The brine (18,000 m3 h
1), containing
chemicals used in the process, is discharged at the shore-
line, next to four discharge points of cooling waters
(308,000 m3 h
1) of a power plant adjacent to the SWRO plant
(Fig. 1). The water temperature at the discharge point was
higher by up to 10  C compared to the temperature at the
intake (Glazer, 2009, 2010b). At the time of this study,
backwash of the sand filters, containing iron hydroxide used
as coagulant in the pre-treatment step, was discharged in
pulses (up to 6500 m3 h
1), with a frequency dependent on
the seawater quality at the intake. As a result, a “red plume”
formed during the pulsed discharges (Safrai and Zask, 2008;
UNEP, 2008). In 2007, 535 ton of iron was discharged at the
outfall, decreasing to 270 and 175 ton in 2008 and 2009,
respectively (I. Safrai, personal communication). The
concentration of iron in the brine ranged from 200 to
1100 mg L
1 in 2008e2009 (Glazer, 2010b). Since May 2010, the
backwash is mixed and discharged continuously with the
brine, in an attempt to mitigate the esthetics effect of the
“red plume”. Additional chemicals used in the process and
discharged with the brine during 2009 were: poly-
phosphonate anti scalant (34 t as P), HCl (15 t) and NaOH
(20 t) for cleaning the RO membranes neutralized to NaCl,
Sodium bisulfite (NaHSO3 e 70 t) used for membrane pres-
ervation and wash water of the CaCO3 reactor used for
alkalinity and pH adjustment of the desalinated water (I.
Safrai, personal communication). It should be noted that
brine from well amelioration RO treatment is discharged
with the SWRO brine. At the time of this study the discharge
rate was 130 m3 h
1 and the concentration of nitrate and
silicic acid high (47 mg L
1 NO3eN and 62 mg L
1 Si(OH)4eSi)
(I. Safrai, Ministry of Environmental Protection (MEP),
personal communication).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2 5451

Fig. 1 e Location of Ashqelon at the Mediterranean coast of Israel and schematic map of the study site. The Ashqelon
seawater desalination plant (A) is adjacent to a power plant (B). The brines are discharged at the shoreline through one
outfall (2) that is located next to cooling waters discharges (1). Line 3 represents the coal unloading quay. Sampling was
performed along the broken line 4 (see Table 1) and at a background station (W), 650 m from the outfall. The water intake for
the desalination plant (SWRO) and the cooling waters (CW) are also depicted.

3. Methods
3.1. Sampling rationale
This study took place while the backwash discharge was
pulsed creating a transient signal (the iron containing back-
wash) on top of a constant influence (brine from the SWRO
and well amelioration plants and cooling waters from the
power plant). Therefore, the sampling was designed to
differentiate among these environmental signals, looking at
time-dependent gradients from backwash discharge and
distance-dependent gradients relative to the outfall. In the
field, the backwash discharge was easily distinguished by the
red tinge of the seawater that served as a marker of the
dispersion path. This was confirmed by the temperature and
salinity of the samples. Seawater samples were collected near
the outfall before and during the backwash discharge
(Stations O and OR, respectively), during dispersion (Stations
DX, where X is the sampling order and increased with time
from the backwash discharge or with distance from the
outfall) and at a background station located west to the outfall
(W). The sampling scheme differentiated between samples
under brine and cooling water influence only (Stations O),
under brine, cooling waters and backwash discharge influence
(OR and DX) and background (W) (Table 1). The best sampling
scheme was executed at the end of winter when the same
geographical position was occupied, before, during and at the
wake of the backwash discharge (Owi, ORwi, D1wi).
3.2. Sampling
Three surveys were conducted on board the R/V “Dolphin”
from the Marine and Coastal Environment Division (MCED) of
the MEP. The surveys took place in Spring (sp)-April 15th, 2008,
Summer (su)-August 19th, 2008, and April 2nd, 2009. Although
the last survey took place at the beginning of April, we
considered it as end of winter (wi) because from mid February
to the end of March 2009 winter storms delayed the estab-
lishment of spring conditions. Sampling stations are
described in Table 1 and Fig. 1.
In situ depth profiles of temperature, salinity, turbidity and
dissolved oxygen were measured with a Yellow Spring
Instruments YSI 6000 probe. Seawater for the analysis of
nutrients, particulate matter, chlorophyll-a (chl-a), and
microscopic community (picophytoplankton cell numbers,
primary and bacterial production) were collected with a FLO-
JET pump, whose inlet was attached to the YSI. Five replicate
seawater samples for nutrient determinations were collected
into 15 ml acid-washed plastic scintillation vials, brought
refrigerated to the laboratory (within 4 h from collection) were
they were immediately frozen and kept frozen until labora-
tory analysis. Samples for chl-a and suspended particulate
matter (SPM) determination were collected into plastic bottles,
kept refrigerated and in the dark and brought to the labora-
tory. At the same day, samples for chl-a determination were
filtered through GF/F filters that were folded, wrapped in
aluminum paper, and frozen. Filters were kept frozen for 1
week until laboratory analyses. Samples for SPM determina-
tion were filtered into pre-weighted 0.45 mm Nuclepore
membrane filters, washed with MilliQ water and frozen.
Duplicate samples of 1.8 cm3 for picophytoplankton
enumeration were fixed immediately upon collection with
20 mL of 25% glutaraldehyde and transported to the laboratory
in the dark. There, the samples were immediately frozen in
liquid nitrogen and stored at
80  C until analysis within 2
months. Samples for the measurement of primary and
bacterial production were collected into plastic bottles, kept
5452 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2

filtered. The filters were digested with a mixture of hydro-

Table 1 e Station’s position and variables measured in
discrete water samples in spring (sp, 15 April 2008), fluoric acid and aqua-regia (ASTM, 1983). The concentrations
summer (su, 15 August 2008) and at the end of winter (wi, of Fe and Al in the digest were determined by Atomic
2 April 2009). Station names with d as a suffix indicate Absorption Spectrometry using a Varian SpectrAA 220 FS
samples collected 0.5 m above the bottom, and the spectrometer. Quality control and quality assurance of the
sampling depth is indicated by *. All other samples were results were performed with standard reference materials
collected at the surface. Background stations are given in (NIST e Estuarine Sediment 1646 and NRCC MESS-3) that were
bold.
digested and analyzed in the same manner as the samples.
Station Distance Temp Salinity SPM NO

3 PO3

4 Si(OH)4 Chl-a was determined fluorometrically after overnight
(*sampling þ NO
2
extraction with 95% acetone, in the dark at 4  C (Holm-Hansen
depth)a
et al., 1965). The pigment extract was subsequently measured
m 
C mg L
1 mM following acidification by 1 N HCl to estimate chl-
Osp 230 23.53 40.11 3.05 0.68 0.08 3.47 a degradation products. Picophytoplankton was enumerated
ORsp 230 24.47 40.23 5.98 0.59 0.06 2.98 by flow cytometry. Prior to the analysis, samples were fast
D1sp 410 25.95 40.55 7.82 1.77 0.09 3.71 thawed at 37  C, and excited in a flow cytometer e FACScan
D2sp 550 27.6 40.51 2.71 0.96 0.09 3.11 Becton Dickinson, fitted with an Argon laser (488 nm).
D2dsp 3.9* 20.18 39.91 2.9 0.08 0.05 2.77 0.93 mm beads (Polysciences) served as standards (Marie et al.,
D3sp 1000 24.81 40.33 1.56 0.33 0.05 3.02
2005; Stambler, 2006). Taxonomic discrimination was based
D3dsp 4.2* 20.43 39.92 3.76 0.08 0.07 2.77
on: cell side-scatter e a proxy of cell volume; forward scatter e
Wsp 650 w 23.43 40.13 3.06 0.31 0.06 2.84
Wdsp 5.6* 19.89 39.91 6.15 0.08 0.07 2.61 a proxy of cell size; and orange and red fluorescence of
O(1)su 310 37.01 40.38 7.85 1.76 0.11 3.72 phycoerythrin and of chl-a (585 nm and 630 nm, respectively).
ORsu 300 34.19 40.5 10.9 2.21 0.11 3.84 Primary production (PP) was measured with the 14C technique
D1su 190 34.86 40.93 14.8 2.75 0.10 4.15 (Steemann-Nielsen, 1952) in duplicated 50 cm3 subsamples. A
D2su 400 35.26 40.46 13.2 1.82 0.13 3.81 spike of approximately 3*105 Bq of [14C] bicarbonate was added
D2dsu 4.1* 29.9 39.5 6.73 0.35 0.12 2.97
to each bottle. The bottles were incubated in the laboratory, at
D3su 590 33.7 40.11 8.13 1.33 0.13 3.52
D3dsu 2.4* 29.48 39.35 10.97 0.08 0.09 2.61
20  C and provided with light flux of 60 mmol photon m
2 s
1.
D4su 1340 33.39 39.84 7.74 0.34 0.04 2.5 After incubation, of approximately 3 h, the samples were
O(2)su 240 31.26 39.58 6.84 1.14 0.10 2.57 filtered onto poly-acetate 25 mm 0.45 mm membrane filters
Wsu 650 w 30.96 39.56 5.93 0.29 0.10 2.65 under light vacuum (about 100 mg Hg), rinsed with filtered
O(2)wi 330 22.26 39.72 3.09 0.14 0.09 0.33 seawater and left overnight in the presence of HCl vapor to
Owi 280 26.35 41.14 3.63 5.75 0.16 4.20
eliminate any remaining traces of inorganic 14C. Control
ORwi 280 26.23 40.84 28.8 4.61 0.13 2.20
samples poisoned by Lugol’s solution at zero time were run in
D1wi 280 26.34 41.16 3.67 2.18 0.14 0.89
D2wi 410 26.2 40.91 4.39 1.92 0.14 0.89 each experimental series to compensate for non-biological
D3wi 630 26.12 40.36 5.48 1.22 0.14 0.72 absorption to filters. The total added 14C was checked for
D4wi 1340 19.01 39.77 4.40 0.06 0.09 0.12 each sampling series by counting 0.1 cm3 portions withdrawn
D4dwi 2.7* 19.03 39.77 4.6 0.14 0.08 0.17 directly from each of the incubated bottles. Total radioactivity
D5wi 1810 19.15 39.77 4.43 0.13 0.08 0.20 in the particulate fraction retained on the filters was deter-
D5dwi 2.8* 19.12 39.76 6.27 0.07 0.09 0.21
mined by liquid scintillation with quench correction. The
Wwi 650 w 22.11 40.01 4.90 0.97 0.10 0.64
average difference between duplicates was w9%. Assimilation
Wdwi 5.3* 18.83 39.77 4.63 0.07 0.08 0.17
number (AN) was calculated by normalizing PP with chl-
SPM-suspended particulate matter. a concentration. AN is an important ecological indicator of
a South south west of the outfall. Distance of the W stations are
growth efficiency and/or shift in phytoplankton composition
toward the west.
(Karl et al., 1995). Bacterial production (BPP) was determined
by the modified (Smith and Azam, 1993) leucine uptake
method (Kirchman et al., 1985; Simon and Azam, 1989). Zero
refrigerated and in the dark, brought to the laboratory in 4 h
time controls were run for all samples. Leucine uptake was
and then processed immediately.
converted to carbon uptake using the conversion factors of
Simon and Azam (1989) with an isotope dilution factor of 2.
3.3. Laboratory analysis Regression analysis was performed with the Addinsoft XLStat
software under the assumption of 95% confidence level.
Nutrients (nitrate þ nitrite, phosphate and silicic acid) were
determined using a segmented flow SANplus SYSTEM from
Skalar (Kress and Herut, 2001). Samples were thawed at the 4. Results
day of analysis. The precision of nitrate þ nitrite, phosphate
and silicic acid measurements was 0.02, 0.003 and 0.06 mM, The brine and cooling waters mixed at the nearshore,
respectively. The limit of detection (2 times the standard following their discharge at the shoreline. The pulsed back-
deviation of the blank) was 0.075 mM for nitrate þ nitrite, wash discharge tinted the water red and showed that the
0.008 mM for phosphate and 0.03 mM for silicic acid. The SPM discharges dispersed toward the southesouth west during the
filters were dried by lyophylization and re-weighed and the three surveys. The mixing was very heterogeneous close to
concentration calculated from the difference and the volume the outfall, as visualized by patches of water with different
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
coloration. The frequency of the discharge in the spring and at
the end of winter was every 40e60 min, while in the summer
the frequency was every 20e30 min. The higher backwash
discharge frequency in the summer probably caused the area
to be chronically affected by it as will be shown hereafter.
4.1. Temperature and salinity
Temperature was seasonally dependent, as expected (Table 1,
Fig. 2a). At the background stations, surficial temperatures
were 23.43, 30.96 and 19.01  C in sp, su and wi, respectively
and decreased with increased water depth to 19.89, 29.93 and
18.83  C in sp, su and wi, respectively. The affected stations in
all surveys were warmer than the background. In spring and
summer all stations were stratified. In spring, maximum
stratification was at D2sp (DT ¼ 7.3  C) and minimal at Wsp
Fig. 2 e Representative depth profiles of temperature (A)
and salinity (B) during the three surveys: spring (squares),
summer (triangles) and end of winter (circles). Background
stations are drawn with solid lines and filled symbols and
affected stations with dashed lines and open symbols. The
water column was stratified in the spring and the summer,
and the temperature and salinity at the upper layers were
higher at the affected stations. At the end of winter the
water column was mixed and the affected station more
saline and warmer than background. The dispersion
stations (exemplified by D3wi) were stratified.
5453
(DT ¼ 3.2  C). In the summer, maximum stratification was at
D2su (DT ¼ 4.8  C). The warmest station was O(1)su (37  C)
while the western station, Wsu, was the coldest (30.96  C) and
mixed down to 3 m. At the end of winter most of the stations
were mixed. There was a clear separation between the colder
background (D4wi and D5wi with 19.5  C) and the affected
stations (26.2  C). The dispersion station D3wi was highly
stratified (DT ¼ 6.7  C). The western station, Wwi, that served
as background in the previous surveys, was stratified as well
indicating influence of the discharges down to 2 m depth,
where temperature reached background.
Salinity was independent of sampling season, ranging
from 39.32 to 41.54 (39.9e40.64, 39.32e40.93 and 39.55e41.54,
in sp, su and wi, respectively) (Table 1, Fig. 2b). The water
column was stratified with respect to salinity, similarly to the
thermal stratification. There was a significant positive
increasing correlation between temperature and salinity in
spring and summer (Fig. 3). The correlation at the end of the
winter is similar to that of the spring except for stations Owi,
ORwi, D2wi where salinity varied while temperature
remained essentially constant. Although it is assumed that
the coefficient of turbulent (eddy) mixing for temperature and
salinity are equal, here, because the brine and cooling waters
are discharged from different outfalls with different discharge
rates (cooling waters ca. 10 times more than brine), there may
be an initial different stratification of the discharges that
make the mixing of the warm waters with coastal waters
different from that of the brine. Heterogeneity of the dispersal
can be shown by the differences in salinity and temperature at
outfall stations O(2)su/O(1)su and Owi/O(2)wi located 50 m
apart (Table 1).
During the three surveys the discharge was positively
buoyant and dispersed close to the surface, traced up to
1340 m south of the outfall. Similar dispersion patterns were
Fig. 3 e Temperature versus salinity during the three
surveys: spring (squares), summer (triangles) and end of
winter (circles). The more saline and warmer waters were
found closer to the outfall. Linear regressions were
significant for spring and summer (R2 [ 0.902 and 0.804,
respectively, p < 0.0001). At the end of winter some of the
points were similar to the spring while close to the outfall
salinity varied at constant temperature, probably due to
non-homogeneous mixing.
5454 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
documented during bi-annual monitoring studies performed
by the Israel Electric Corporation (IEC) at the discharge site
(Glazer, 2009, 2010b). The maximal differences in temperature
between affected and background stations were similar at the
three surveys: 7.7, 7.8 and 7.3  C in sp, su and wi, respectively.
However, at the surface only, the maximal difference was
measured at the end of winter (4.2, 6.0 and 7.3  C in sp, su and
wi, respectively). The differences in salinity between the
affected stations and background were 0.74, 1.61 and 1.84 in
sp, su and wi, respectively, and at the surface 0.38, 0.90 and
0.79 in sp, su and wi, respectively. The IEC monitoring in
2008e2009 found similar differences in temperature (7.4  C
and 6.2  C in spring and winter 2008) and in salinity (2.08 and
1.89 in spring and winter 2008) between the stations at the
vicinity of the outfall and the reference station located ca.
2.5 km south of the outfall.
4.2. Turbidity and suspended particulate matter (SPM)
Seawater turbidity without the backwash discharge was
similar at all stations at the three surveys and ranged from
<0.1 to 1.5 NTU. Backwash discharge increased water turbidity
rapidly following its discharge as exemplified by the time
series measurement at the end of winter (Fig. 4). At the three
surveys, turbidity returned to background values at ca. 500 m
from the outfall or after 20 min from the discharge at the
outfall.
SPM ranged from 1.56 to 7.82 mg L
1 in spring, from 5.93 to
14.8 mg L
1 in summer and from 3.09 to 6.27 mg L
1 at the end
of winter, with a very high concentration, 28.8 mg L
1,
measured at station ORwi (Table 1). The concentrations
measured at the deep samples at the same station were lower
than the surficial samples. In the summer, SPM concentra-
tions were the highest among the surveys, probably caused by
the high frequency of the backwash discharge. In the spring
and summer the SPM concentration doubled immediately
following the backwash discharge (stations ORsp and ORsu)
and increased 8 times at the end of winter.
Fig. 4 e Seawater turbidity at the outfall increased steeply
following the backwash discharge and returned to
background values within 20 min. The horizontal line
represents the time the discharge reached the outfall
station.
As expected, turbidity increased with increased SPM
concentration. During spring and end of winter the relations
were similar while in the summer more SPM was present at
the same turbidity and the data more scattered.
4.3. Dissolved oxygen and nutrients
The seawater was saturated or close to saturation with dis-
solved oxygen during the three surveys, and not influenced by
the discharges. Therefore, dissolved oxygen concentrations
will not be further discussed.
The background concentrations of the nutrients
(nitrate þ nitrite, phosphate, silicic acid) were similar during
the three surveys, except for silicic acid, that was about 2
times lower at the end of winter. The concentrations
measured at the deep samples at the same station were
similar or lower than the surficial samples (Table 1). Nutrient
concentrations were higher at the outfall prior and after the
backwash discharge (O and OR) and at dispersion stations,
compared to background, more noticeable in the
nitrate þ nitrite concentrations. While in spring and summer
the concentrations were higher at the affected stations, and
similar to background in the others, at the end of the winter
concentrations decreased as a function of distance from the
outfall and time from the discharge.
4.4. Particulate iron (Fe)
Iron in particulate matter ranged from to 35e1611 mg L
1,
similar in the three surveys (Fig. 5). As Fe is supplied both with
the coagulant and by natural processes we normalized the
concentrations to aluminum (Al) concentrations as done for
the sediments in the area (Goldsmith et al., 2001). The rela-
tionship between Fe and Al was similar during the three
Fig. 5 e Iron versus aluminum concentration in suspended
particulate matter during the three surveys (spring
(squares), summer (triangles) and end of winter (circles))
shows six samples with higher than natural iron
concentration. Those were found at the stations at the
outfall with backwash presence and at the first dispersion
station. Regression line solid (R2 [ 0.751, p < 0.0001), 95%
confidence and prediction intervals in dotted and dashed
lines, respectively.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2 5455

surveys, indicating a common source and no seasonal varia-

tion. Fe concentrations were higher than natural at stations
ORsp, D1sp, D2sp, ORsu, D1su and ORwi, namely, the outfall
stations with the backwash discharge and the first and second
dispersion stations (Fig. 5, Table 1).

4.5. Chlorophyll-a and picophytoplankton cells

Chl-a concentrations during the three surveys were similar

and ranged from 0.26 to 0.94 mg m
3 in spring, from 0.42 to
0.74 mg m
3 in summer and from 0.23 to 1.03 mg m
3 at the
end of winter (Supplementary material 1). In the spring and in
particular at the end of winter there was a clear separation
into two groups of stations: the affected stations group with
low chl-a and higher temperature and salinity and the back-
ground group, with high chl-a and lower temperature and
salinity. The average chl-a concentrations at the affected, low
chl-a group were 0.32  0.08 and 0.29  0.06 mg m
3 in spring
and end of winter, respectively. These concentrations were
significantly lower (more than 60%) than the background
average concentrations of 0.80  0.11, and 0.89  0.04 mg m
3
in spring and end of winter, respectively. It should be noted
that we did not find detectable concentrations of degraded
chl-a (data not shown), indicating that the phytoplankton
cells at the discharge site were probably intact. In the summer,
the concentrations were more dispersed and the concentra-
tion at the affected stations (0.45  0.03 mg m
3) 32% lower
than background (0.66  0.06 mg m
3). There was a significant
linearly decreasing correlation between chl-a and tempera-
ture at each survey, and also between chl-a and salinity, but
with lower correlation coefficients than with temperature
(Fig. 6). Fig. 6 e Chlorophyll-a concentrations decreased with
The total number of picophytoplankton cells in spring increased salinity (A) (R2 [ 0.497, 0.694 and 0.666 in sp, su
(1.83e10.2  107 cells L
1) and summer (3.74e15.5 and wi, respectively) and more significantly with increased
 107 cells L
1) were similar (Supplementary material 1) as temperature (B) ( R2 [ 0.594, 0.794 and 0.718 in sp, su and
well as the relative contribution of the groups (Fig. 7). Syn- wi, respectively) during the three surveys.
echococcus comprised of an average 86 and 93% of the
total cells in spring and summer, respectively. At the end
of winter, the number of cells was 10 times lower suggesting utilization by larger cells, possible diatoms. The
(1.15e8.81  106 cells L
1) with a different relative contribu- two pico-eukaryotes groups contributed up to 97% to the total
tion: pico-eukaryotes dominated (50% of the total cells), Syn- biomass during the winter (calculation not shown).
echococcus and Prochlorococcus contributed 18% each to the total
and an unexpected 4th group comprised of 13% of the total
cells. This 4th group belonged probably to the pico- 4.6. Primary production (PP) and bacterial production
eukaryotes, slightly larger than the “classical” group. During (BPP) rates
each survey the total cell numbers were the lowest at the
affected stations lower by 3.8, 2 and 6 times from background The average primary production rates (PP) at the non affected
in sp, su and wi, respectively. stations during the three surveys were similar: 4.41  1.4,
In the three surveys, cell numbers decreased with 4.79  0.85 and 5.53  0.37 mgC m
3 h
1 the sp, su and wi,
increased temperature and salinity and were positively respectively. PP was significantly lower at the affected
correlated to chl-a concentration (Fig. 8). At the end of winter stations: 0.60  0.56, 1.84  0.67 and 1.43  0.31 mgC m
3 h
1 in
the stations could be grouped into two: higher and lower cell sp, su and wi, respectively. During each survey, PP was
numbers, similar to chl-a. Moreover, even though the cell negatively correlated with temperature, similar in the spring
numbers were 10 times lower at the end of winter the chl- and at the end of winter and slightly different in the summer.
a concentration was similar to the other surveys. That could PP was also negatively correlated with salinity; however, the
be due to photo-acclimation of the cells, increasing chl-a per correlation was significant only when all data were pooled. PP
cell and/or the presence of cells larger than 10 mm, not was positively correlated with chl-a concentrations, showing
enumerated by flow cytometry. The latter is indicated by the a fairly uniform relationship in all surveys (Fig. 9).
large contribution of the pico-eukaryotes and the low The ranges of assimilation number (AN ¼ PP/Chl-a) in the
concentration of Si(OH)4 in seawater at the end of winter spring and at end of winter were similar (3.19e6.53 (median
5456 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2

100

80

Percentage
60

40

20

O i

D wi

W i
D wi
D wi
D wi
D 4wi

D 5wi

W wi
i
D su

D 3su
D 2su

D 4su

O wi
O )su

O )su

D i
D su

D su

Wu
D sp
D 2sp

D 3sp

W su
O dsu
D sp

Wp

D u
D sp
O sp

W sp
O dsp

)w

dw
Rw
s
s

Rs

1
2
3
1

4d

5d
2d

3d

4d
1

2d

3d
R
O

(1

(2
(2
Station
Syn Pro Pico-euk 4th

Fig. 7 e Relative picophytoplankton distribution at the different stations during the three surveys show that Synechococcus
dominated in spring and summer while at the end of winter the pico-eukaryotes constituted 50% of the total cells,
Synechococcus and Prochlorococcus contributed 18% each to the total and an unexpected 4th group comprised of 13% of the
total cells. In addition, the relative distribution at the affected stations was different from background at the end of winter.

4.88) and 2.19e5.69 (median 4.97)) mgC (mg Chl)
1 h
1. AN was
lower in the summer (1.07e3.77 (median 2.46) with a high
value of 6.79 mgC (mg Chl)
1 h
1 at station D2dsu), showing
that the area in the summer was chronically affected probably
due to the high frequency of the backwash discharge. AN was
negatively correlated to particulate Fe concentration, indi-
cating that the backwash caused a decrease in phytoplankton
growth efficiency (Fig. 10).
The average rates of bacterial production (BPP) at the non
affected stations were similar in summer and at the end of
winter (0.58  0.25 and 0.46  0.05 mgC m
3 h
1, respectively)
and lower in the spring (0.19  0.10 mgC m
3 h
1). The rates
were much lower at the affected stations in spring and end of
winter (0.03  0.03 mgC m
3 h
1 during both surveys)
compared to background (Supplementary material 1).
However, in the summer BPP at the affected stations was
higher than background with an average rate of
1.64  0.73 mgC m
3 h
1. BPP decreased at surface at the
dispersion stations D2su and D3su (0.79 mgC m
3 h
1) and at
the western station Wsu.
The relationship between BPP and temperature and
salinity was inconsistent, being positive in the summer, and
negative in the spring and at the end of winter (Fig. 11).
5. Discussion
5.1. Cause e effect identification and quantification
Seawater at the discharge site was warmer and more saline
than the background due to the constant discharge of brine
and cooling waters, while the pulsed backwash discharge
added an additional pressure to the already impacted envi-
ronment. The mixed discharge was positively buoyant, and
dispersed at the surface, mainly toward the south. This is in

Fig. 8 e Chlorophyll-a increased with picophytoplankton

cells numbers during the three surveys (R2 [ 0.647, 0.422 Fig. 9 e Primary production increased with increasing
and 0.809 in sp, su and wi, respectively) despite the 10 chlorophyll-a similarly during the three surveys
times lower number of cells at the end of winter. (R2 [ 0.820, p < 0.0001).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
Fig. 10 e The assimilation number (a proxy of
phytoplankton growth efficiency) decreased with
increased particulate iron, a marker of the backwash
presence, during the three surveys (R2 [ 0.585, p < 0.0001.
Two extreme points in the summer were neglected).
Fig. 11 e Bacterial production increased with decreased
salinity in spring and at the end of winter (R2 [ 0.558) and
was not correlated to salinity in the summer due to the
high rates measured at the stations located close to the
outfall (A). Bacterial production increased with increased
temperature in the summer (R2 [ 0.519), decreased with
temperature at the end of winter (R2 [ 0.838) and was not
correlated to temperature in the spring (B).
5457
agreement with the findings of monitoring studies at the site
(Glazer, 2009, 2010b) but in contrast to the dominant northward
current direction at the inner shelf (26 m depth) region
(Rosentraub and Brenner, 2007). The difference is probably due
to the localized conditions at the outfall, with high cooling
water discharge rate and the presence of the breakwater north
to the outfalls (Fig. 1). As the brine dispersed at the surface we
will concentrate the discussion on the upper layer and identify
and quantify the effects based on the background values. To do
so, we redefined the background stations as those with
minimal temperature and salinity at each survey. In the spring
and summer the background stations were the western ones
as planned (Wps and Wsu, Table 1). At the end of winter, the
western station (Wwi) was affected by the discharges therefore
the southernmost station (D5wi) was used as the background
(Table 1). In addition, we calculated the fractional change of
a parameter (F ¼ value at a sampling point divided by the
background value), essentially an enrichment (F > 1) or
depletion factor (0 < F < 1) from background as a mean to
quantify and compare among the effects (Supplementary
material 2). The effects were classified into three groups: 1)
Possibly attributed to brine and cooling waters discharge; 2)
Possibly attributed to backwash discharge, and 3) Inconclusive.
In the first group, brine and cooling waters affected salinity,
temperature, nutrients, chl-a, and picophytoplankton cell
numbers. The salinity and temperature increases were obvi-
ously attributed to brine and cooling waters, respectively. The
maximal enrichment factor of temperature (F ¼ 1.39) was
larger than for salinity (F ¼ 1.04) probably due to the 10 times
larger discharge of cooling waters. This may also explain why
temperature was a more accurate proxy of the dispersion in
this study. The enrichment factors for both the temperature
and salinity were larger at the end of the winter than at the
other two surveys. Similar enrichment factors were also
calculated with the IEC monitoring data (Glazer, 2009, 2010b)
showing that background conditions were reached during this
study. For example, in the spring of 2008 the maximal salinity
measured was 41.50 compared to the background of 38.42
(F ¼ 1.05) and the maximal temperature was 30.4  C compared
to the background of 22.3  C (F ¼ 1.32). Numerical modeling of
the dispersion run prior to the operation of the plant simu-
lated a larger effect of the brine, with 49 salinity at 100 m from
the shore (Sladkevich and Kit, 2004). Lower salinities were
measured in this study, however the stations were located at
least 250 m from the discharge point.
Nutrient concentrations were higher at the outfall and
decreased with increased distance. As for salinity and
temperature, the effect was more evident at the end of winter,
with the better sampling scheme. Maximal enrichment
factors for nitrate and silicic acid were F ¼ 96 and 35, respec-
tively, at the outfall station (Owi) and decreased with
increased time from the discharge and distance from the
outfall (Supplementary material 2). Maximal enrichment
factor in spring and summer occurred at the D1 dispersion
stations, with F ¼ 5.7 and 9.4 for nitrate in the spring and the
summer, respectively and F ¼ 1.31 and 1.6 for silicic acid.
Nitrate and silicic acid were introduced with the brines but
they originate mostly from the well amelioration plant and
not at the desalination plant. Therefore, based on the
discharge rate, the pulsed backwash would dilute the
5458 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
concentrations and decrease them by 20%. And indeed, the
enrichment factor for nitrate decreased by 20% at ORwi,
following the backwash discharge, but more (48%) for silicic
acid. Based on the concentration of nitrate in the well
amelioration brines (47 mg L
1) and the discharge rates we
would expect the concentration at the outfall to be 15.5 mM if
mixing only with the SWRO brines or 1.45 mM if mixing with
the brine and cooling waters. However, the maximal
concentration detected was 5.75 mM at the end of the winter
(Table 1). That shows that at our sampling point, ca. 250 m
from the brine discharge point, the brines were mixed with
just one fifth of the cooling waters. As the cooling waters are
discharged through 4 separate outfalls this is reasonable,
although mixing was lower than expected.
Chl-a in general was lower at the affected stations during
the three surveys, with maximal depletion factors of 0.59, 0.71
and 0.26 in sp, su and wi, respectively (Supplementary
material 1 and 2). The phytoplankton near the outfall did not
benefit from the increased concentrations of nutrients and the
chl-a and PP rates were lower than in the unaffected, nutrient
poor, stations. The effect was larger at the end of the winter,
followed by the summer, while in the spring the effect was
more variable. The concentrations decreased with increasing
salinity but more significantly with increasing temperature
(Fig. 6). There were two exceptions. In the spring, chl-
a concentration decreased only after the backwash
discharge and in the winter, the concentration increased
following the backwash discharge and decreased drastically at
the start of the dispersion (Supplementary material 1).
However, thermal pollution is known to cause a decrease in
chl-a concentrations and change primary production and
community structure (Langford, 1990; Langford and John,
2001). The effects may vary and are season-dependent (Choi
et al., 2002; Chuang et al., 2009). Therefore, we hypothesize
that the decrease in chl-a in this study was a result of
increased water temperature and not salinity. This hypothesis
is supported by the results of monitoring studies at the brine
only outfall of the SWRO plant “Via Maris” off Palmahim
(Israel, Fig. 1), located at ca. 10 m depth and 1 km from the
coast which showed no effect on chl-a concentrations (Kress
et al., 2010). Picophytoplankton cell numbers were lower at
affected stations and increased with decreasing temperature
and salinity. Moreover, the depletion factors were larger than
for chl-a, which represents all components of the phyto-
plankton, with F ¼ 0.27, 0.49 and 0.13 in the spring, summer
and end of winter, respectively (Supplementary material 2).
The relative distribution of the 4 picoplankton groups in spring
and summer were similar at all stations. This may be due to
the fact that more than 85% were Synechococcus and small
changes may have been undetectable. At the end of winter,
when the relative distribution of the groups was more
balanced, the affected stations had a lower contribution of
pico-eukaryotes and a higher contribution of Synechococcus
and Prochlorococcus cell numbers than at the background
stations. Temperature affects size distribution and metabolic
rates of phytoplankton (Finkel et al., 2010) therefore we
assume that was the case in this study as well.
In the second group, discharge of the pulsed backwash
increased turbidity, SPM, particulate Fe and reduced AN.
Turbidity and SPM concentrations were similar to the
background at the outfall (O) stations and increased following
the backwash discharge. Maximal enrichment factors were
found in the summer, followed by the spring and the lowest
found at the end of winter, except for a very high enrichment
factor at station ORwi (Supplementary material 2). The
maximal enrichment factors were not found at the same
surveys as the first group and may be connected to the
frequency of the backwash discharge. During the summer, the
backwash was discharged every 20e30 min, probably causing
a constant influence. During the spring and at the end of
winter, the frequency was lower (every 40e60 min), while in
the winter the desalination plant was working only at half
capacity. Particulate Fe was highest at the backwash affected
stations, with no gradual decrease with distance from the
outfall or time from backwash discharge. AN was negatively
correlated with particulate Fe concentration in seawater
(Fig. 10), a proxy for coagulant presence. It is reasonable to
assume that the presence of the coagulant decreased the
growth efficiency of the phytoplankton. As with turbidity and
SPM, the depletion factors were lower in the summer and
spring and less affected at the end of winter.
In the third group, the rates of primary and bacterial
production could not be definitely attributed to one of
discharges. PP seemed to follow chl-a and picophytoplankton
population, and it may be that the decrease could be also
attributed to thermal pollution. The most affected survey was
the end of winter, with maximal depletion factors of 0.21
(Supplementary material 2). The correlation of BPP with
temperature and salinity varied: negative in the spring and at
the end of winter and positive in the summer. Maximal
enrichment factor in the summer was 11.9 while depletion
was much higher at the end of winter than in the spring (down
to 0.02, Supplementary material 2). The highest rates were
measured in the summer in the vicinity of the outfall
(Supplementary material 1), while at the other stations the
rates were similar to those found in the other surveys at
similar salinities (Fig. 11).
5.2. Comparison to other areas along the coast and to
environmental quality guidelines
The area of study is located at the shoreline of the ultra-
oligotrophic Eastern Mediterranean that has exceptionally
low nutrient concentrations, chl-a, PP and BPP and the
phytoplankton community is dominated by the pico and nano
fractions (Krom et al., 1991; Yacobi et al., 1995; Zohary and
Robarts, 1998; Kress and Herut, 2001; Pitta et al., 2005; Psarra
et al., 2005; Yogev et al., 2011). However, the neritic waters
are characterized by somewhat higher nutrient and chl-
a concentrations, higher PP and a greater abundance of
larger size phytoplankton (Azov, 1986; Berman et al., 1986;
Kimor et al., 1987; Gitelson et al., 1996; Herut et al., 2000),
and particularly in estuaries of coastal streams (Herut et al.
2010) and at locations influenced by anthropogenic sources.
Our study was performed in a nearshore location, and by
definition modified by anthropogenic activity, and its char-
acteristics manifest that fact. For example, surficial nitrate
and phosphate increased from <0.03 to <0.008 mM, respec-
tively at the open sea (Krom et al., 2005), through 0.02 and
0.03 mM in the summer at 120 m bottom depth (Herut et al.,
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
2000), to 0.31 and 0.05 mM at the spring background station in
this study. Chl-a at the same stations increased from 0.018,
0.03 and 0.44 mg m
3. A gradual decrease in chl-a was shown
along a westward transect (0.5e22 km distance from the
shore) from 0.36 to 0.06 mg m
3 in another study, while SPM
changed from 3 to 0.9 mg L
1 (Gitelson et al., 1996). Primary
production and bacterial production in the open sea in the
spring were 0.091 and 0.043 mg C m
3 h
1 (Pitta et al., 2005;
Psarra et al., 2005) while in this study at the background
station in the spring the rates were higher: 1.68 and
0.07 mg C m
3 h
1 for PP and BPP, respectively. The concen-
tration of chl-a and the PP and BPP rates at the affected
stations in this study resembled the lower values found at the
open sea stations. However, the lower values in this study
were accompanied by high concentrations of nutrients, SPM
and higher turbidity, contrary to those found in the open sea.
A different comparison can be made to data collected
during monitoring studies conducted along the Mediterra-
nean coast of Israel: The National monitoring program (Herut
et al., 2010) performed in the summer, monitoring at the Via
Maris SWRO brine discharge site off Palmahim (Kress et al.,
2010) and monitoring at the outfall of power stations per-
formed by the Israel Electric Corp (IEC) (Glazer, 2010a,b). These
studies concentrate on water quality and benthic community
while the biological component of the water column is small.
In general, salinity, temperature, nutrients, chl-a and partic-
ulate Fe concentrations measured during this study at the
background stations represented the non polluted waters
along the Mediterranean coast. Temperatures higher than
background, by up to 6  C, are found at the outfall of three
additional power plants located along the coast, similar to the
increase found in this study. Regular monitoring performed at
those plants did not show a correlation between seawater
temperature and chl-a concentrations. However, at one
instance (Glazer, 2010a) the sampling points were not located
at the maximal temperatures, so the sampling may have
missed the affected areas. Moreover, using only the data for
the surficial samples collected during 2009 at the Ashqelon
power plant (Glazer, 2010b), chl-a decreased with increased
temperature similar to the findings of this study. Salinities
measured during this study were higher than those measured
along the coast, except for the area of the SWRO Via Maris
plant (Kress et al., 2010) where values up to 43.5 were
measured near the bottom. This plant is smaller
(123,000 m3 d
1) than the Ashqelon’s SWRO Plant and is not
located next to a power plant so the brines are discharged
through a marine outfall. Chl-a concentrations were not
affected by the brine discharge and were natural for the area
and seasonally dependent. Much higher concentrations of
nutrients and chl-a were measured at the mouth of coastal
streams (Herut et al., 2010) and at other polluted areas along
the coast then at the effected stations during this study.
We attempted to evaluate the environmental significance
of the findings of this study based on available marine envi-
ronmental quality guidelines (Bricker et al., 1999; ANZECC,
2000; MCED, 2002; Buchman, 2008). Most guidelines do not
address salinity. However, increased salinity may have an
adverse environmental effect, as recorded, for instance, in the
case of Posedonia oceanica seagrass meadows off the Mediter-
ranean coast of Spain (Sánchez-Lizaso et al., 2008). Most
5459
guidelines do not address also temperature, iron, PP and BPP as
well. Chl-a is usually given an upper limit concentration to
prevent eutrophication, in contrast to the decrease in
concentrations detected in this study. On the other hand,
guidelines for turbidity, suspended particulate matter (SPM)
concentration, water color, total nitrogen and phosphorus are
usually present in marine water environmental quality
guidelines. In Israel, the increase in turbidity should not be
higher than 10% of the seasonal average, the increase in SPM
should not exceed the seasonal average by more than
10 mg L
1 and the color of the water should not be affected
outside the mixing zone (MCED, 2002). Turbidity exceeded the
Israeli guidelines at the end of the winter and SPM exceeded
the guidelines at station ORwi, but the effect of water color was
more extensive and well outside the mixing zone. We think
that the Israeli allowable upper limits for total nitrogen and
total phosphorus, 71 and 3.2 mM, are too high for the area, and
therefore used NOAA’s criteria for oligotrophic areas of 7.1 and
0.32 mM, respectively (Bricker et al., 1999) and assumed that
nitrate and phosphate were about half of the total N and P (3.5
and 0.16 mM) (N. Kress, unpublished results). The ANZECC
guidelines are stricter for nitrate (0.6 mM) and similar for
phosphate (0.17 mM). Nitrate concentrations higher than half of
NOAA’s criteria were measured at stations Owi, ORwi, while at
3 samples in the spring, 6 in the summer and 5 at the end of the
winter, the concentrations were higher than the ANZECC
guideline limits (Table 1). The concentrations of phosphate
were lower than the guideline limits at all samples except at
Owi, that was similar to the guideline. The water quality
criteria for British Columbia for iron is 300 mg L
1 (cited in
Buchman, 2008). In this study, we measured concentrations up
to 1611 mg L
1, with 9 samples above 300 mg L
1, mostly in the
summer (Fig. 5) when the discharge frequency was the highest
(every 20e30 min). In summary, even though the stations in
this study were located outside the mixing zone, some of the
measured parameters exceed accepted environmental guide-
lines, indicating possible effects at the discharge site. The
infringement of the water color guideline gave the regulator
the means to request improvements in brine management,
because of the lack of relevant legislation (Drami, 2011). The
study we present herein indicates that such a simple criterion
as color, which ostensibly is only of an esthetic value, is affil-
iated with a change in water components that may potentially
affect microbial life in the marine environment.
5.3. Conclusions and future outlook
This research was prompted by the adverse esthetic effect of red
backwash discharge on the marine environment. The first steps
taken by the Ministry of Environmental Protection (MEP) were to
require a reduction of iron utilization and discharge, followed
by a requirement to stop the pulsed backwash discharge, mix it
with the brine and discharge it continuously. Since May 2010,
the backwash at the Ashqelon plant is mixed and discharged
continuously with the brine. Moreover, The Hadera desalina-
tion plant, that started operating on March 2010, discharges the
backwash continuously, while future SWRO desalination plants
will be required to dispose of the coagulant on land. In addition,
assessment of the environmental and economic implications of
positioning the SW intake and brine outfall at different
5460 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
distances from the shore are now required at the planning
stages of future SWRO plants.
We were able to show that in addition to the esthetic effect
(increased turbidity and SPM) the backwash affected the
phytoplankton growth efficiency. Moreover, increased
salinity, increased temperature or both together affected
water quality and the microbial community at the disposal
site. It is still not clear if the mentioned changes have
a detectable impact beyond the immediate vicinity of the
discharge location, however, the very existence of adverse
environmental impact call for further specific research in situ
and in controlled laboratory experiments. Marine pollution
was not taken into account when the government decided to
increase SWRO desalination from ca. 700,000 m3day
1 fresh-
water today to ca. 2 Mm3day
1 by 2020 (Dreizin et al., 2008).
Three SWRO desalination plants are operational now at the
200 km long Mediterranean coast of Israel, two more passed
the tender/biding stage while others are at the planning stage.
While we approve the steps taken to reduce the environ-
mental impact of desalination, the combined and possible
synergistic effects of the plants at the Israeli shoreline as well
as those at neighboring countries were not considered. While
SW desalination is a viable option for freshwater production,
we caution the decision makers to continue to study and
understand the risks to the marine environment.
Acknowledgments
We are grateful to Itzhak Kodovizky (Kodo) and Galia Pasternak
of the Marine and Coastal Environmental Division of the
Ministry of Environmental Protection for the cruises. We thank
Yaron Gertner, Semion Kaganovsky, Benayahu Sulimani and
Lora Izraelov for field and laboratory assistance. The research
was funded by Israel’s Ministry of Environmental Protection
(Marine and Coastal Environmental Division) and Israel’s
Nature and Parks Authority. The thorough comments by three
reviewers helped improve this manuscript and are greatly
appreciated. This paper was written during N. Kress sabbatical
at UC Santa Cruz (with A. Paytan) and MBARI (with K. Johnson).
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.005.
references
Ahmed, M., Shayya, W.H., Hoey, D., Al-Handaly, J., 2001. Brine
disposal from reverse osmosis desalination plants in Oman
and the United Arab Emirates. Desalination 133 (2), 135e147.
Alpert, H., Borrowman, C., Haddad, B., 2007. Evaluating
Environmental Impacts of Desalination in California. Center
for Integrated Water Research. <http://ciwr.ucsc.edu/
desalplanning/workshops.html>.
ANZECC, 2000. Australian and New Zealand Guidelines for Fresh
and Marine Water Quality. http://www.mfe.govt.nz/
publications/water/anzecc-water-quality-guide-02/anzecc-
water-quality-guide-02-pdfs.html.
ASTM, 1983. American Society for Testing and Materials
Designation -D 3683-78. Standard Test Method for Trace
Elements in Coal and Coke Ash by Atomic Absorption, pp.
472e475.
Azov, Y., 1986. Seasonal patterns of phytoplankton productivity
and abundance in nearshore oligotrophic waters of the Levant
Basin (Mediterranean). Journal of Plankton Research 8, 41e53.
Berman, T., Azov, Y., Schneller, A., Walline, P.D., Townsend, A.R.,
1986. Extent, transparency, and phytoplankton distribution of
the neritic waters overlying the Israeli coastal shelf.
Oceanologica Acta 9, 439e447.
Bricker, S.B., Clement, C.G., Pirhalla, D.E., Orlando, S.P., Farrow, D.
R.G., 1999. National estuarine eutrophication assessment
effects of nutrient enrichment in the nation’s estuaries.
NOAA, National Ocean Service, Special Projects Office and the
National Centers for Coastal Ocean Science. Silver Spring, MD,
71 pp.
Buchman, M.F., 2008. NOAA Screening Quick Reference Tables.
NOAA OR&R Report 08e1. Office of Response and Restoration
Division, National Oceanic and Atmospheric Administration,
Seattle WA, 34 pp.
Campbell, R.L., Jones, A.T., 2005. Appropriate disposal of effluent
from coastal desalination facilities. Desalination 182 (1e3),
365e372.
Choi, D.H., Park, J.S., Hwang, C.Y., Huh, S.H., Cho, B.C., 2002.
Effects of thermal effluents from a power station on bacteria
and heterotrophic nanoflagellates in coastal waters. Marine
Ecology-Progress Series 229, 1e10.
Chuang, Y.L., Yang, H.H., Lin, H.J., 2009. Effects of a thermal
discharge from a nuclear power plant on phytoplankton and
periphyton in subtropical coastal waters. Journal of Sea
Research 61 (4), 197e205.
Drami, D., 2011. Ashkelon’s Desalination Plant and the Red Brine
Phenomenon: Chemo-oceanographic and Role of Scientific
Knowledge on Environmental Policy Making Concerning
Desalination. M.Sc Thesis. The Porter School of Environmental
Studies, Tel Aviv University, Israel.
Dreizin, Y., Tenne, A., Hoffman, D., 2008. Integrating large scale
seawater desalination plants within Israel’s water supply
system. Desalination 220 (1e3), 132e149.
Einav, R., Harussi, K., Perry, D., 2002. The footprint of the
desalination processes on the environment. Desalination 152
(1e3), 141e154.
Einav, R., Lokiec, F., 2003. Environmental aspects of a desalination
plant in Ashkelon. Desalination 156 (1e3), 79e85.
Fernández-Torquemada, Y., Sánchez-Lizaso, J.L., González-
Correa, J.M., 2005. Preliminary results of the monitoring of the
brine discharge produced by the SWRO desalination plant of
Alicante (SE Spain). Desalination 182 (1e3), 395e402.
Finkel, Z.V., Beardall, J., Flynn, K.J., Quigg, A., Rees, T.A.V.,
Raven, J.A., 2010. Phytoplankton in a changing world: cell size
and elemental stoichiometry. Journal of Plankton Research 32
(1), 119e137.
Fritzmann, C., Löwenberg, J., Wintgens, T., Melin, T., 2007. State-
of-the-art of reverse osmosis desalination. Desalination
216 (1e3), 1e76.
Gacia, E., Invers, O., Manzanera, M., Ballesteros, E., Romero, J.,
2007. Impact of the brine from a desalination plant on
a shallow seagrass (Posidonia oceanica) meadow. Estuarine,
Coastal and Shelf Science 72 (4), 579e590.
Gitelson, A., Karnieli, A., Goldman, N., Yacobi, Y.Z., Mayo, M.,
1996. Chlorophyll estimation in the Southeastern
Mediterranean using CZCS images: adaptation of an algorithm
and its validation. Journal of Marine Systems 9 (3e4), 283e290.
Glazer, A., 2009. Monitoring the Coastal and Marine Environment
at the Discharge Site of the Rutenberg Power Plant, VID
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
Desalination Plant and Mekorot’s Well Amelioration Plant,
Results from 2008. Israel Electric Corp. R-ELP 21e2009. 151 pp.
(In Hebrew).
Glazer, A., 2010a. Monitoring the Coastal and Marine
Environment at the Discharge Site of the Orot Rabin Power
Plant, Results from 2009. Israel Electric Corp. 115 pp. (In
Hebrew).
Glazer, A., 2010b. Monitoring the Coastal and Marine
Environment at the Discharge Site of the Rutenberg Power
Plant, VID Desalination Plant and Mekorot’s Well
Amelioration Plant, Results from 2009. Israel Electric Corp.
R-ELP 04e2010. 216 pp. (In Hebrew).
Goldsmith, S.L., Krom, M.D., Sandler, A., Herut, B., 2001.
Spatial trends in the chemical composition of sediments
on the continental shelf and slope off the Mediterranean
coast of Israel. Continental Shelf Research 21 (16e17),
1879e1900.
Greenlee, L.F., Lawler, D.F., Freeman, B.D., Marrot, B., Moulin, P.,
2009. Reverse osmosis desalination: water sources,
technology, and today’s challenges. Water Research 43 (9),
2317e2348.
Herut, B., Almogi-Labin, A., Jannink, N., Gertman, I., 2000. The
seasonal dynamics of nutrient and chlorophyll
a concentrations on the SE Mediterranean shelf-slope.
Oceanologica Acta 23 (7), 771e782.
Herut, B., Shefer, E., Gordon, N., Galil, B., Tibor, G., Tom, M.,
Rilov, G., 2010. Environmental Quality of Israel’s
Mediterranean Coastal Waters in 2009, IOLR Report H64/2010.
179 pp. (in Hebrew).
Holm-Hansen, O., Lorenzen, C.J., Holmes, R.W., Strickland, J.D.H.,
1965. Fluorometric determination of chlorophyll. Journal du
Conseil, Permanent International pour l’Exploration de la Mer
30, 3e15.
Karl, D.M., Letelier, R., Hebel, D., Tupas, L., Dore, J., Christian, J.,
Winn, C., 1995. Ecosystem changes in the North Pacific
subtropical gyre attributed to the 1991e92 El Nino. Nature 373
(6511), 230e234.
Khan, S.J., Murchland, D., Rhodes, M., Waite, T.D., 2009.
Management of concentrated waste streams from high-
pressure membrane water treatment systems. Critical
Reviews in Environmental Science and Technology 39 (5),
367e415.
Kimor, B., Berman, T., Schneller, A., 1987. Phytoplankton
assemblages in the deep chlorophyll maximum layers off the
Mediterranean coast of Israel. Journal of Plankton Research
9 (3), 433e443.
Kirchman, D., K’Nees, E., Hodson, R., 1985. Leucine incorporation
and its potential as a measure of protein synthesis by bacteria
in natural aquatic systems. Applied and Environmental
Microbiology 49 (3), 599e607.
Kress, N., Galil, B., Shoham-Frider, E., 2010. Marine Monitoring at
the Brine Disposal Site of the Via Maris Desalination Plant off
Palmahim, Results from May and September 2009. IOLR Report
H27/2010. 71 pp. (In Hebrew).
Kress, N., Herut, B., 2001. Spatial and seasonal evolution of
dissolved oxygen and nutrients in the Southern Levantine
Basin (Eastern Mediterranean Sea): chemical characterization
of the water masses and inferences on the N:P ratios. Deep-
Sea Research Part I-Oceanographic Research Papers 48 (11),
2347e2372.
Krom, M.D., Kress, N., Brenner, S., Gordon, L.I., 1991. Phosphorus
limitation of primary productivity in the eastern Mediterranean
sea. Limnology and Oceanography 36 (3), 424e432.
Krom, M.D., Woodward, E.M.S., Herut, B., Kress, N., Carbo, P.,
Mantoura, R.F.C., Spyres, G., Thingstad, T.F., Wassmann, P.,
Wexels-Riser, C., Kitidis, V., Law, C.S., Zodiatis, G., 2005.
Nutrient cycling in the south east Levantine basin of the
eastern Mediterranean: results from a phosphorus starved
5461
system. Deep Sea Research Part II: Topical Studies in
Oceanography 52 (22e23), 2879e2896.
Kronenberg, G., 2004. The largest SWRO plant in the world e
Ashkelon 100 million m3/y BOT project. Desalination 166,
457e463.
Langford, T.E.L., 1990. Ecological Effects of Thermal Discharges.
Elsevier.
Langford, T.E.L., John, H.S., 2001. Encyclopedia of Ocean Sciences,
pp. 2933e2940. Academic Press, Oxford.
Lattemann, S., 2010. Development of an Environmental Impact
Assessment and Decision Support System for Seawater
Desalination Plants. Ph.D. Dissertation, Delft University of
Technology, The Netherlands, 294 pp.
Lattemann, S., Hopner, T., 2008. Environmental impact and
impact assessment of seawater desalination. Desalination 220
(1e3), 1e15.
Marie, D., Simon, N., Vaulot, D., 2005. Phytoplankton cell
counting by flow cytometry. In: Andersen, R.A. (Ed.), Algal
Culturing Techniques. Phycological Society of America,
Burlington, pp. 1e15.
MCED, 2002. Seawater Quality Environmental Criteria for Israeli
Mediterranean Waters. Marine and Coastal Environmental
Division, Ministry of Environmental Protection (In Hebrew).
NRC, 2008. Desalination, a National Perspective National
Research Council of the National Academies. The National
Academies Press, Washington, D.C.
Pitta, P., Stambler, N., Tanaka, T., Zohary, T., Tselepides, A.,
Rassoulzadegan, F., 2005. Biological response to P addition in
the eastern Mediterranean sea. The microbial race against
time. Deep Sea Research Part II: Topical Studies in
Oceanography 52 (22e23), 2961e2974.
Psarra, S., Zohary, T., Krom, M.D., Mantoura, R.F.C.,
Polychronaki, T., Stambler, N., Tanaka, T., Tselepides, A.,
Thingstad, T.F., 2005. Phytoplankton response to a Lagrangian
phosphate addition in the Levantine Sea (Eastern
Mediterranean). Deep-Sea Research Part Ii-Topical Studies in
Oceanography 52 (22e23), 2944e2960.
Raventos, N., Macpherson, E., Garcı́a-Rubiés, A., 2006. Effect of
brine discharge from a desalination plant on macrobenthic
communities in the NW Mediterranean. Marine
Environmental Research 62 (1), 1e14.
Roberts, D.A., Johnston, E.L., Knott, N.A., 2010. Impacts of
desalination plant discharges on the marine environment:
a critical review of published studies. Water Research 44 (18),
5117e5128.
Rosentraub, Z., Brenner, S., 2007. Circulation over the
southeastern continental shelf and slope of the
Mediterranean Sea: direct current measurements, winds, and
numerical model simulations. Journal of Geophysical
Research 112 (C11), C11001.
Safrai, I., Zask, A., 2008. Reverse osmosis desalination plants e
marine environmentalist regulator point of view. Desalination
220 (1e3), 72e84.
Sánchez-Lizaso, J.L., Romero, J., Ruiz, J., Gacia, E., Buceta, J.L.,
Invers, O., Fernández Torquemada, Y., Mas, J., Ruiz-
Mateo, A., Manzanera, M., 2008. Salinity tolerance of the
Mediterranean seagrass Posidonia oceanica:
recommendations to minimize the impact of brine
discharges from desalination plants. Desalination 221 (1e3),
602e607.
Sauvet-Goichon, B., 2007. Ashkelon desalination plant e A
successful challenge. Desalination 203 (1e3), 75e81.
Service, R.F., 2006. Desalination freshens up. Science 313 (5790),
1088e1090.
Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G.,
Marinas, B.J., Mayes, A.M., 2008. Science and technology for
water purification in the coming decades. Nature 452 (7185),
301e310.
5462 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 4 9 e5 4 6 2
Simon, M., Azam, F., 1989. Protein content and protein synthesis
rates of planktonic marine bacteria. Marine Ecology Progress
Series 51, 201e213.
Sladkevich, M., Kit, E., 2004. VID Desalination Plant at Ashkelon.
Mathematical Model for Spreading of Brine Outlet at Sea
Report No. P.N. 612/04. Report No. P.N. 612/04.
Smith, S., Azam, F., 1993. A simple economical method for
measuring bacterial protein synthesis rates using 3H leucine.
Marine Microbial Food Webs 6, 107e114.
Stambler, N., 2006. Light and picophytoplankton in the Gulf of
Eilat (Aqaba). Journal of Geophysical Research 111 (C11),
C11009.
Steemann-Nielsen, E., 1952. The use of radioactive carbon (14C)
for measuring organic production in the sea. Journal des
Conseil International Pour Exploration de la Mer 18, 117e140.
Tal, A., 2006. Seeking sustainability: Israel’s evolving water
management strategy. Science 313 (5790), 1081e1084.
UNEP, 2008. Desalination Resource and Guidance Manual for
Environmental Impact Assessments. United Nations
Environment Programme, Regional Office for West Asia,
Manama, and World Health Organization, Regional Office for
the Eastern Mediterranean. Cairo Ed. S. Lattemann, 168 pp.
UNEP/MAP/MEDPOL, 2003. Sea Water Desalination in the
Mediterranean: Assessment and Guidelines. MAP Technical
Reports Series No. 139 UNEP/MAP Athens.
Water-world, 22nd-GWI/IDA. International Desalination
Association (IDA) and Global Water Intelligence (GWI). http://
www.waterworld.com/index.html.
Yacobi, Y.Z., Zohary, T., Kress, N., Hecht, A., Robarts, R.D.,
Waiser, M., Wood, A.M., 1995. Chlorophyll distribution
throughout the southeastern Mediterranean in relation to the
physical structure of the water mass. Journal of Marine
Systems 6 (3), 179e190.
Yermiyahu, U., Tal, A., Ben-Gal, A., Bar-Tal, A., Tarchitzky, J.,
Lahav, O., 2007. Rethinking desalinated water quality and
agriculture. Science 318 (5852), 920e921.
Yogev, T., Rahav, E., Bar-Zeev, E., Man-Aharonovich, D.,
Stambler, N., Kress, N., Béjà, O., Mulholland, M.R., Herut, B.,
Berman-Frank, I., 2011. Is dinitrogen fixation significant in the
Levantine Basin, East Mediterranean Sea? Environmental
Microbiology 13 (4), 854e871.
Zohary, T., Robarts, R.D., 1998. Experimental study of microbial P
limitation in the eastern Mediterranean. Limnology and
Oceanography 43 (3), 387e395.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Nitrate removal, communities of denitrifiers and adverse

effects in different carbon substrates for use in denitrification
beds

Sören Warneke a,*, Louis A. Schipper a, Michael G. Matiasek b, Kate M. Scow b,

Stewart Cameron c, Denise A. Bruesewitz d, Ian R. McDonald d
a
Department of Earth and Ocean Sciences, University of Waikato, Hamilton, New Zealand
b
Department of Land, Air and Water Resources, University of California Davis, Davis, United States
c
Wairakei Research Centre, GNS Science, Taupo, New Zealand
d
Department of Biological Sciences, University of Waikato, Hamilton, New Zealand

article info abstract

Article history: Denitrification beds are containers filled with wood by-products that serve as a carbon
Received 7 June 2011 and energy source to denitrifiers, which reduce nitrate (NO3
) from point source
Received in revised form discharges into non-reactive dinitrogen (N2) gas. This study investigates a range of
1 August 2011 alternative carbon sources and determines rates, mechanisms and factors controlling
Accepted 6 August 2011 NO3
removal, denitrifying bacterial community, and the adverse effects of these
Available online 17 August 2011 substrates. Experimental barrels (0.2 m3) filled with either maize cobs, wheat straw, green
waste, sawdust, pine woodchips or eucalyptus woodchips were incubated at 16.8  C or
Keywords: 27.1  C (outlet temperature), and received NO3
enriched water (14.38 mg N L
1 and
Denitrification 17.15 mg N L
1). After 2.5 years of incubation measurements were made of NO3
eN
Controlling factors removal rates, in vitro denitrification rates (DR), factors limiting denitrification (carbon
Bioreactor and nitrate availability, dissolved oxygen, temperature, pH, and concentrations of NO3
,
Denitrification genes nitrite and ammonia), copy number of nitrite reductase (nirS and nirK ) and nitrous oxide
nir reductase (nosZ ) genes, and greenhouse gas production (dissolved nitrous oxide (N2O)
and methane), and carbon (TOC) loss. Microbial denitrification was the main mechanism
for NO3
eN removal. NitrateeN removal rates ranged from 1.3 (pine woodchips) to 6.2 g N
m
3 d
1 (maize cobs), and were predominantly limited by C availability and temperature
(Q10 ¼ 1.2) when NO3
eN outlet concentrations remained above 1 mg L
1. The NO3
eN
removal rate did not depend directly on substrate type, but on the quantity of microbially
available carbon, which differed between carbon sources. The abundance of denitrifying
genes (nirS, nirK and nosZ ) was similar in replicate barrels under cold incubation, but
varied substantially under warm incubation, and between substrates. Warm incubation
enhanced growth of nirS containing bacteria and bacteria that lacked the nosZ gene,
potentially explaining the greater N2O emission in warmer environments. Maize cob
substrate had the highest NO3
eN removal rate, but adverse effects include TOC release,
dissolved N2O release and substantial carbon consumption by non-denitrifiers. Wood-
chips removed less than half of NO3
removed by maize cobs, but provided ideal condi-
tions for denitrifying bacteria, and adverse effects were not observed. Therefore we

* Corresponding author. Tel.: þ64 7 858 3700; fax: þ64 7 858 4964.
E-mail address: warni1@yahoo.com (S. Warneke).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.007
5464 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
recommend the combination of maize cobs and woodchips to enhance NO3
removal
while minimizing adverse effects in denitrification beds.
1. Introduction
Anthropogenic production of reactive nitrogen (N), through
the Haber Bosch process, cultivation of N-fixing crops, and
combustion of fossil fuels, contributes 45% of global N fixation
(Canfield et al., 2010). This human impact on the nitrogen cycle
leads to N enrichment of surface waters, with consequences
including eutrophication, hypoxia, harmful algae blooms and
habitat degradation in lakes, rivers and coastal zones, and an
increase in N2O emissions (Howarth et al., 2002; Rabalais, 2002;
Phoenix et al., 2006). Denitrification beds are a promising
approach to reduce reactive N release from point source
discharges into waterways. These denitrifying bioreactors are
containers filled with wood by-products, where the wood acts
as carbon and energy source for denitrifying microorganisms
(Schipper et al., 2010), which convert NO
3 to unreactive N gas
via microbial denitrification (Warneke et al., 2011b).
A wide range of carbon substrates have been trialled in
column studies to find appropriate media for bioreactors
(Volokita et al., 1996a,b; Soares and Abeliovich, 1998; Della
Rocca et al., 2005, 2006; Saliling et al., 2007; Gibert et al.,
2008; Cameron and Schipper, 2010). Nitrate removal rates in
column studies range from 3 g N m
3 d
1 (woodchips;
Cameron and Schipper, 2010) to 96 g N m
3 d
1 (rice husk;
Shao et al., 2008). The exceptionally high NO
3 removal rates of
many carbon substrates (e.g., rice husks, wheat straw, cotton)
were attributed to a large organic carbon release in the start-
up phase of the columns, and were not sustainable over
a longer time period (Cameron and Schipper, 2010). In a long-
term study, barrels filled with maize cobs removed 3e6.5
times more NO
3 eN than wood substrate, but also had higher
carbon leaching in the effluent (Cameron and Schipper, 2010).
Greenan et al. (2006) also reported that maize stalks produced
greater NO
3 removal than woodchips. However, little is
known about the mechanism responsible for NO
3 removal,
the controlling factors, denitrifying bacterial communities or
adverse effects, such as greenhouse gas release, when using
different carbon substrates than woodchips. Warneke et al.
(2011a, b) demonstrated that the mechanism responsible for
NO
3 removal in a full-scale woodchip bioreactor was micro-
bial denitrification, and the removal process was limited by
microbially available carbon and temperature. Smaller-scale
studies have also determined that microbial denitrification is
the dominant N removal mechanism, rather than dissimila-
tory NO


3 reduction to ammonium DNRA or NO3 immobiliza-
tion (Robertson, 2010; Greenan et al., 2006, 2009; Gibert et al.,
2008).
Greenhouse gas (GHG) production during denitrification
is an important issue to address when studying denitrifica-
tion beds. An in field woodchip bioreactor study by Warneke
et al. (2011a) yielded total N2O release of 4.3% of removed
NO
3 eN, whereas Greenan et al. (2009) reported negligible
release of dissolved N2O in a woodchip column study.
ª 2011 Elsevier Ltd. All rights reserved.
However, there have been no studies examining GHG
production in denitrification beds containing different
carbon sources.
So far, the population of denitrifying bacteria has not been
investigated in substrates for use in denitrification beds. The
abundance of denitrifying communities can be estimated by
quantifying the functional gene copy numbers for nitrite
reductase, nirS and nirK, and nitrous-oxide reductase, nosZ.
These denitrification genes express reductase enzymes
involved in denitrification. NirS expresses the cytochrome cd1-
containing nitrite reductase (which catalyses the reduction of
nitrite to nitric-oxide), nirK expresses the copper containing
nitrite reductase, and nosZ expresses nitrous oxide reductase
(which catalyses the reduction of N2O to N2) (Zumft, 1997;
Braker et al., 1998). The two different genes for nitrite reduc-
tase, nirS and nirK, have coevolved to produce two indepen-
dent pathways and no denitrifier is known to contain both
pathways (Philippot, 2002). Interestingly many denitrifying
organisms have been shown to reduce NO
3 only to nitrous
oxide (Cheneby et al., 1998, 2004) and some, such as Agro-
bacterium tumerfaciens C58 do not possess nitrous oxide
reductase (nosZ ) (Wood et al., 2001). Many studies have shown
that differences in the diversity and abundance of denitrifying
bacterial genes were correlated to a variety of physical and
chemical conditions; organic carbon in glacier foreland
(Kandeler et al., 2006), temperature in constructed wetlands
(Chon et al., 2010), water logging in rice paddy soils (Yoshida
et al., 2009), organic or conventional fertilizer in agricultural
soils (Dambreville et al., 2006; Enwall et al., 2005), native and
cultivated soils (Stres et al., 2004), soil pH in grassland soils
(Cuhel et al., 2010), nitrous-oxide emissions (Philippot et al.,
2009) and NO
3 concentration in woodlands with different
vegetation (Lindsay et al., 2010). However, the diversity and
abundance of denitrifying bacteria under consistent envi-
ronmental conditions (e.g., same temperature, NO
3 concen-
tration, DO concentration, flow rate), but with different carbon
substrates are poorly known.
This study followed a 2.5-year trial by Cameron and
Schipper (2010), where different C substrates were
compared for their ability to remove NO
3 from water at two
temperatures. The main objectives of the present study were
to determine the limiting factors and the microbial mecha-
nisms of the NO
3 removal for different C substrates such as
woodchips (Pine and Eucalyptus), sawdust, green waste,
maize cobs and wheat straw in these barrels. The abundance
of the denitrification functional genes nirS, nirK and nosZ were
compared across replicate barrels, different temperatures
and substrates. The factors affecting denitrifying communi-
ties were examined and whether NO
3 removal could be
predicted from the copy number of denitrification genes.
Adverse effects, including production of N2O and methane
(CH4), and total organic carbon (TOC) release, were also
determined to evaluate the benefit of the different C
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5 5465

substrates. These findings can be used to help select the 2.2. Solute concentrations and NO

3 removal rate
appropriate carbon substrate for denitrifying bioreactors
(denitrification beds and walls) to optimise NO

3 removal, Water was sampled from the inflow and outflow tubing of the
reduce GHG production, and maximize the lifetime of the cold and warm barrels. Samples were filtered through
bioreactor. disposable membrane filters (0.45 mm) and analysed for NO
3,
NHþ

4 and NO2 using a flow injection analyser (Lachat Instru-
ments; Loveland, USA) (APHA, 1992). TOC was determined
2. Materials and methods from unfiltered water samples using a Shimadzu TOC-5000
analyser (Shimadzu Corp.; Kyoto, Japan). Temperature and
2.1. Study site and substrate DO of the inlet and outlet of the barrels were measured with
an InLab 605 O2-Sensor (Mettler Toledo, Switzerland).
The design of the experimental setup was fully described in Nitrate removal rate was calculated as follows: NO
3 eN

Cameron and Schipper (2010). In this study, 24 experimental removal rate ¼ DNO

1

3 eN  FR  V , where DNO3 eN was the

barrels (0.2 m3) filled with six different carbon substrates and difference of inflow and outflow NO
3 eN concentration, FR

placed in a 7 m long shipping container were continuously was the flow rate of the NO
3 solution, and V was the volume of

loaded with a self-prepared NO
3 solution (in average the barrel.

about15.8 mg L
1; Table 1). Barrels were divided between
a cold treatment (16.8  C average outlet temperature) and
a warm treatment (27.1  C average outlet temperature), and
every carbon source had two replicate barrels at each
temperature. The selected substrates were: woodchips of
Pinus radiata (soft wood), woodchips of Eucalyptus “Red Duke”
(hardwood), sawdust (P. radiata), maize cobs, wheat straw and
green waste (shredded and chipped miscellaneous shrubbery
leaves and stems). The barrels had been loaded with NO

solution for 2.5 years before samples were taken for this
study.
2.3. Greenhouse gas production
Water from the inlet and outlet of the barrels was collected in
3.7 mL exetainers (Labco, UK) for analysis of dissolved N2O and
CH4 concentrations. The exetainers contained 0.2 mL H2SO4
(20%) to prevent further bacterial activity. After 12 h head-
space equilibrium at room temperature, headspace gas
3 samples were analysed for N2O and CH4 concentration using
a gas chromatograph equipped with an electron capture
detector and flame ionisation detector, respectively (Varian;

Table 1 e Solute concentrations and temperature at the inlet and outlet of barrels filled with different carbon substrates
under warm and cold incubation.
Barrela NO

3 eN NO

2 eN NHþ
4 eN pH Temp DO TC TOC
(cold line) (mg L
1) (mg L
1) (mg L
1) ( C) (mg L
1) (mg L
1) (mg L
1)

Inlet 14.4 0.023 <0.001 7.7 19.2 7.1 13.5 5.4

Outlet PW1 10 0.080 <0.001 6.9 18 1.9 18.3 6.8
Outlet PW2 10.5 0.179 <0.001 6.9 17.2 1.3 26.6 9.1
Outlet MC1 0.4 0.033 <0.001 6.2 17.2 1.1 100.4 70.2
Outlet MC2 0.1 0.003 <0.001 5.9 16.2 0.5 86.8 76.8
Outlet WS1 0.5 0.060 0.134 6.9 17.1 0.7 56.1 16.5
Outlet WS2 1 0.025 0.065 6.9 16.6 0.5 49.3 11
Outlet GW1 6.2 0.153 0.178 6.8 16.6 0.9 25.1 7
Outlet GW2 2.2 0.024 0.785 6.6 16.3 0.5 60 14.7
Outlet SD1 8.3 0.164 0.364 6.8 16.9 0.4 23.2 5.9
Outlet SD2 5.7 0.020 0.344 7.1 16.7 0.3 14.8 4.6
Outlet EW1 9.7 0.082 0.033 7.0 16.6 0.6 21.6 4.7
Outlet EW2 9.6 0.424 <0.001 7.0 16.3 0.5 29 7.4
Inlet 17.2 0.007 <0.001 8.3 36 5.9 14 6.0
Outlet PW1 12 0.234 0.027 7.6 26 1 19.5 5.8
Outlet PW2 11.2 0.171 <0.001 7.5 27.3 0.6 18.2 6.2
Outlet MC1 3.7 0.079 0.081 7.6 26.3 1 51.4 9.5
Outlet MC2 6.1 0.094 0.364 7.7 29 1 53.7 9.7
Outlet WS1 8.5 0.410 0.259 7.6 29.1 0.4 33.71 7.8
Outlet WS2 9.3 0.562 0.194 7.6 27 1.1 10.9 0.6
Outlet GW1 4.3 0.088 <0.001 7.5 27.4 0.9 52.1 8.3
Outlet GW2 7.2 0.234 <0.001 7.6 25 1.1 46.4 7.9
Outlet SD1 8.5 0.444 0.307 7.6 27.2 1.3 23.8 5.8
Outlet SD2 8.6 0.494 0.084 7.6 28.4 0.6 13.7 3.8
Outlet EW1 9.6 0.816 0.024 7.6 27.7 0.4 30 5.8
Outlet EW2 10.9 0.516 0.123 7.6 25.3 1.1 20.9 5.1

a PW1 and PW2, soft woodchips (pine); MC1 and MC2, maize cobs; WS1 and WS2, wheat straw; GW1 and GW2, green waste; SD1 and SD2,
sawdust; EW1 and EW2, hard woodchips (eucalyptus).
5466 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
Palo Alto, USA). Dissolved N2O and dissolved CH4 gas
concentrations were calculated after Weiss and Price (1980)
and Yamamoto et al. (1976) using the Bunsen coefficients.
For N2O analyses, the gas chromatograph was fitted with
a Hayesep D column (3.6 m  1/800  2.0 mm). The column oven
temperature was 80  C, the ECD detector temperature was
300  C, and the flow rate of the carrier gas (10% methane in
argon) was 40 mL min
1. For CH4 analyses, the GC was
equipped with a Hayesep Q column (80  1/800 SS; Q 80e100).
The column oven temperature was 90  C, the FID detector
temperature was 150  C, and the flow rate of the N2-carrier gas
was 30 mL min
1.
2.4. Denitrification rates
Denitrification rates (DR) of the different carbon substrates in
the barrels were determined using a modification of the
denitrifying enzyme activity (DEA) method of Tiedje et al.
(1989). Carbon substrate (600 g wet weight) from each barrel
was collected using a gloved hand from the centre of the barrel
and stored in plastic bags at 4  C. Rubber gloves were changed
after each sampling. Water samples (500 mL) from the outlet
of each barrel were stored in 1 L plastic bottles at 4  C. In the
laboratory, the substrate and water samples were equilibrated
to room temperature in a water bath. Carbon substrate (100 g
wet weight) and water (60 g) from each barrel were added to
four airtight bottles (600 mL). The headspace of the bottles was
flushed with N2 gas for 10 min prior to injection of 40 mL of
acetylene (10% of the headspace volume), to inhibit reduction
of N2O to N2. Each assay was amended with one of four
solutions (all 5 mL): i) glucose (8 g L
1; DR þ C), ii) potassium
NO

1

3 (4 g L ; DR þ N), iii) glucose and KNO3 (8 g L

1
and 4 g L
1
respectively; DR þ C/N), and iv) no amendment (DR), to iden-
tify whether DR was C and/or NO
3 limited. After bottles were
incubated at 27  C on a shaker table (100 rpm), headspace gas
samples were collected through a rubber septum after 30, 40,
50 and 60 min using a syringe. Gas samples were stored in
3.7 mL exetainers (Labco, UK) until analysis for N2O concen-
tration within 7 days via GC-ECD (see above).
2.5. DNA extraction
Carbon substrates (400 mL) were sampled from the centre of
each barrel, sealed in 500 mL airtight plastic containers and
stored at
24  C until frozen samples were vacuum freeze
dried. Several trial DNA extractions were performed on the 6
types of reactor bed material. It was determined that the corn
cobs, green waste and sawdust, performed best with the
FastDNA SPIN Kit for Soil (MP Biomedicals, Solon, OH)
whereas the bulkier samples, woodchips and wheat straw
performed better with the Mo Bio Ultra Clean Mega Prep Soil
DNA kit (Mo Bio Laboratories, Inc., Carlsbad, CA). The criteria
for selecting an extraction method was based on the amount
of DNA extracted per amount of material extracted and the
total number of 16S rRNA genes per gram dry material as
determined by quantitative PCR (data not shown). The corn
cobs, woodchips and wheat straw were reduced in size with
a sterile scalpel and or scissors, so that they could fit in the
initial extraction tube. The FastDNA SPIN Kit for Soil was
used to extract 0.05e0.1 g of corn cobs, 0.13e0.22 g green waste
and 0.1e0.14 g of sawdust as per manufacturer instructions.
The Ultra Clean Mega Prep Soil DNA kit was used to extract
2.27e2.65 g of pine woodchips, 0.45e0.69 g of wheat straw and
3.32e4.04 g of Eucalyptus woodchips as per manufacturer
instructions. All samples were extracted in duplicate. The
quantity of DNA extracted was quantified with a Qubit fluo-
rometer (Life Technologies, Carlsbad, CA).
2.6. Quantitative PCR
Thermal cycling, fluorescent data collection, and data analysis
were performed on an ABI Prism 7300 sequence detection
system (Life Technologies, Carlsbad, CA) according to the
manufacturer’s instructions using SYBR-green based detec-
tion. Initially, the DNA extractions for each sample type were
diluted from 20- to 1000-fold to determine the optimum DNA
concentration for QPCR. It was determined that a 200-fold
dilution was required for all samples to dilute past PCR
inhibitors that were coextracted (data not shown). QPCR
reactions for nirK, nirS and nosZ and 16S rRNA contain 5 uL of
template DNA, 0.5 mM of each forward and reverse primer
except nosZ which used 1.5 uM of primer, 12.5 mL of 2 SYBR
GreenER QPCR Super Mix (Life Technologies, Carlsbad, CA), in
a total volume of 25 mL. The primers (50 -30 ) used to detect the
nirK, nirS and nosZ and 16S rRNA genes are nirK876 (ATY GGC
GGV AYG GCG A) and nirK1040 (GCC TCG ATC AGR TTR TGG
TT) (Henry et al., 2005) nirSCd3aF (AAC GYS AAG GAR ACS GG)
and nirSR3cd (GAS TTC GGR TGS GTC TTS AYG AA) (Kandeler
et al., 2006), nosZ2F (CGC RAC GGC AAS AAG GTS MSS GT)
and nosZ2R (CAK RTG CAK SGC RTG GCA GAA) (Henry et al.,
2006), 341F (CCT ACG GGA GGC AGC AG) and 534R (ATT ACC
GCG GCT GCT GGC A, also referred to as 515R) 16S rRNA
primers (Lopez-Gutierrez et al., 2004) respectively. The
conditions for nirK and nirS real-time PCR are 10 min at 95  C
for enzyme activation; afterwards six touchdown cycles are
performed: 15 s at 95  C for denaturation, 30 s at 63  C for
annealing, and 30 s at 72  C for extension. The annealing
temperature is progressively decreased by 1  C down to 58  C.
Finally, a last cycle with an annealing temperature of 58  C is
repeated 40 times with the addition of a data acquisition step
of 30 s at 80  C after the extension phase. One last step of 95  C
for 15 s, 60  C for 30 s and 95  C for 30 s is added to obtain
a specific denaturation curve. The thermal cycling conditions
for nosZ are similar except for the annealing temperature,
which is 65  C for 30 s for the first 6 cycles and 60  C for 15 s for
the 40 cycles. 16S rRNA QPCR was performed with no touch-
down cycle, just one annealing temperature at 60  C for 30 s
and only 35 cycles instead of 40. Purity of amplified products
was checked by the observation of a single peak during the
dissociation analysis. Copy Numbers were determined by
using a standard curve obtained with serial plasmid dilutions
of a known amount of plasmid DNA containing a fragment of
the 16S rRNA gene, nirK, nirS and nosZ gene. Each DNA
extraction was analyzed for each gene in triplicate along with
three non-template controls. Denitrification gene copy
numbers are reported as copies per gram dry material and also
reported as normalized to 16S rRNA gene copies. The nitrite
reductase to nitrous oxide reductase ratio (Snir/nos) was
determined by summing the nir genes (nirS þ nirK ) and
dividing the sum by the nos genes and was used as an
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
indication of nitrous-oxide producing potential. To determine
if each environment selected more for nirS or nirK, the ratio of
nirS/nirK was also calculated. The authors acknowledge that
novel bacterial sequences were likely missed by the 16S rRNA
primers used in this study which might have resulted in an
underestimation of the community size in our soil, which
subsequently led to the calculation of higher relative abun-
dances of nirS functional genes.
2.7. Respirable C
Respirable C was measured, as an index of the availability of C
to microorganisms, using a modified alkali trap method of
Cheng and Coleman (1989). Carbon substrates (100 g wet
weight) and effluent (60 g) from each barrel were added to
airtight bottles (600 mL). Small beakers (30 mL) filled with
10 mL of 0.5 M KOH were placed into the jars to trap CO2. After
sealing the bottles, the headspaces were flushed with N2 gas
for 10 min and incubated at room temperature (22  C) for 4
days. After incubation, 5 mL of the CO2 trapping solution were
removed from the bottles and mixed with 10% BaCl2 solution
(10 mL) and phenolphthalein (pH indicator) in 100 mL flasks.
After back-titration of these solutions against the standard
0.1 M HCl to determine the amount of trapped CO2, respirable
carbon was expressed as CO2eC g
1 carbon substrate (dry
weight).
2.8. Statistical analysis
Similarities and differences of nitrite reductase gene copies
(Snir) per gram carbon substrate, nirS/nirK and nir/nosZ were
evaluated calculating the Wald confidence interval (95%) of
these gene copy numbers for each barrel (data not shown).
Associated errors of the results are reported as standard
errors.
3. Results
3.1. Solute concentrations
The average temperature of the cold and warm incubation
outlet was 16.8  C and 27.1  C respectively, and used as
calculation basis for determining the Q10. Q10 is the factor of
the reaction rate increase with every 10  C rise in temperature.
The inlet NO
3 eN concentration of the cold barrels was
14.4  0.6 mg L
1, and for the warm barrels was
17.2  1 mg L
1. The average flow rates of the cold and warm
barrels were 48.3  2.0 ml min
1 and 58.5  2.3 ml min
1,
respectively. NitriteeN concentrations in the outflow were
always below 0.2 mg L
1 for cold barrels and ranged from 0.08
to 0.82 mg L
1 for warm barrels (Table 1). In the cold incuba-
tions wheat straw, green waste and sawdust, and in the warm
incubations all the carbon substrates, except green waste,
released NHþ4 ranging from 0.03 to 0.79 mg L

1
(Table 1). All the
barrels showed a slight decrease in pH at the outflow (Table 1).
DO decreased from 7.1 mg L
1 (inlet concentration) to below
1.9 mg L
1 (outlet concentration) in cold barrels, and from
5.9 mg L
1 to below 1.3 mg L
1 in warm barrels (Table 1). TOC
was released in high concentrations from the cold incubated
5467
maize cobs (70.2 and 76.8 mg L
1) and wheat straw (16.5 and
11 mg L
1), and in the warm incubated maize cobs (9.5 and
9.7 mg L
1). However other carbon substrate barrels released
either low concentrations, or consumed, TOC (Table 1).
3.2. Nitrate removal and controlling factors of
denitrification
NitrateeN removal rates ranged from 1.3 (soft woodchip barrel
2) to 6.2 g N m
3 d
1 (maize cobs barrel 2), and were dependent
on temperature with a Q10 of 1.2  0.13 (Fig. 1). Maize cobs,
wheat straw and green waste showed the highest NO
3 eN
removal rates, ranging from 4.3 g N m
3 d
1 (green waste) to
5.7 g N m
3 d
1 (maize cobs) in cold barrels, and from 4.5 g N
m
3 d
1 (wheat straw) to 6.0 g N m
3 d
1 (maize cobs) in warm
barrels (Fig. 1).
NitrateeN removal increased linearly with the in vitro
denitrification rate DR þ C/N for cold and warm incubation
( y ¼ 0.16x þ 1.6; R2 ¼ 0.63; p ¼ 0.002 and y ¼ 0.24x þ 2.9;
R2 ¼ 0.65; p ¼ 0.001 respectively; where y ¼ NO
3 eN removal
rate in g N m
3 d
1 and x ¼ DR þ C/N in mg N h
1 g
1) (Fig. 2).
Furthermore, the NO
3 eN removal rate depended on the
available carbon content as shown in three ways. NitrateeN
removal rate was linearly correlated with respirable carbon
for both cold and warm incubated carbon substrates
( y ¼ 0.08x þ 1.6; R2 ¼ 0.82; p < 0.001 and y ¼ 0.15x þ 1.2;
R2 ¼ 0.62; p ¼ 0.002 respectively; where y ¼ NO
3 eN removal
rate in g N m
3 d
1 and x ¼ respirable carbon in mg C g
1 d
1)
(Fig. 3). In vitro measured DR could be enhanced with a glucose
amendment for all carbon substrates, except for maize cobs
and wheat straw in cold and warm barrels. DR in cold incu-
bated maize cobs and wheat straw were NO


3 limited (NO3 eN

1
concentration <1 mg L ) (Fig. 4; Table 1) and DR in warm
incubated maize cobs and wheat straw were not limited by
glucose or NO
3 , except for one warm barrel of maize cobs
(MC1), which was also limited by glucose (Fig. 4). Nitrate
amended DR (DR þ N) was also significantly correlated with
Fig. 1 e Nitrate removal rates for different carbon
substrates in cold (16.8  C) and warm (27.1  C) barrels. PW1
and PW2, soft woodchips (pine), replicates; MC1 and MC2,
maize cobs; WS1 and WS2, wheat straw; GW1 and GW2,
green waste; SD1 and SD2, sawdust; EW1 and EW2, hard
woodchips (eucalyptus).
5468 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
Fig. 2 e NitrateeN removal rate as a function of in vitro DR
amended with glucose and nitrate (DR D G/N) for cold and
warm incubated substrate. Linear regression statistics are
reported in text.
Fig. 3 e NitrateeN removal rate (A) and in vitro DR amended
with nitrate (DR D N) (B) as a function of respirable carbon
for cold and warm incubated substrate. Linear regression
statistics are reported in text.
Fig. 4 e In vitro denitrification rates (DR) at 27  C for
different carbon substrates in cold (A) and warm (B) barrels.
DR assays were amended with glucose (DR D C), NOL 3
(DR D N), glucose and NOL 3 (DR D C/N), and none amended
(DR). PW1 and PW2, soft woodchips (pine); MC1 and MC2,
maize cobs; WS1 and WS2, wheat straw; GW1 and GW2,
green waste; SD1 and SD2, sawdust; EW1 and EW2, hard
woodchips (eucalyptus).
respirable carbon for cold and warm incubations
( y ¼ 0.38x
1.3; R2 ¼ 0.70; p < 0.001 and y ¼ 0.37x e 4.3;
R2 ¼ 0.48; p ¼ 0.013 respectively; where y ¼ DR þ N in mg
N h
1 g
1 and x ¼ respirable carbon in mg C g
1 d
1) (Fig. 3B).
3.3. Copies of denitrification genes (nirS, nirK and
nosZ )
The abundance of nirS, nirK and nosZ ranged from
8.7  0.8  106 (pine woodchips) to 1.6  0.01  1010 (green
waste) copies of nirS g
1 dry substrate, 0.7  0.1  106 (pine
woodchips) to 6.8  0.1  109 (maize cobs) copies of nirK g
1 dry
substrate, and 1.2  0.1  106 (pine woodchips) to
9.0  0.2  109 (maize cobs) copies of nosZ g
1 dry substrate for
cold incubations (Table 2). Abundance of nirS, nirK and nosZ in
warm incubated substrate ranged from 2.0  0.1  107 (pine
woodchips) to 1.3  0.04  1011 (maize cobs) copies of nirS g
1
dry substrate, 7.4  0.4  106 (pine woodchips) to
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5 5469

Table 2 e Average copy number (3106) of denitrification genes (nirS, nirK and nosZ ) and 16S rRNA isolated from different
carbon substrates used in denitrifying barrels under cold and warm incubations.
Treatment Carbon nirS copies g
1 nirK copies g
1 nosZ copies g
1 16S rRNA copies g
1
substratea dry substrate dry substrate dry substrate dry substrate

Cold PW1 10.1  2.9 12  0.2 2.5  0.1 3.1  0.1

PW2 8.7  0.8 0.7  0.1 1.2  0.1 1.2  0.1
MC1 4718  299.2 4217  301.8 5760.1  783.8 96761.9  2649.3
MC2 7474.5  160.2 6815.9  147 8990.3  180.4 43138.8  2076.5
WS1 388.6  29.4 211.3  4.8 138.3  0.2 351.1  28.1
WS2 195  17.9 148.4  3.3 77.2  1.6 1.2  0.1
GW1 7296.5  368.3 2919  72.4 4183.6  109.3 22755.1  596.1
GW2 16163.2  137.3 5664.1  667.7 7168.5  143.8 22660.2  1595.2
SD1 3412.4  43.5 2227.5  185.9 1064.2  50.4 4745.9  253.5
SD2 3319.6  53.3 1998.5  126.2 927  25.2 3648.9  338.6
EW1 41.5  6.3 22.7  1.7 15.9  0.5 30.3  1.0
EW2 57.9  1.3 34  1.6 23.3  1.3 40.8  1.6
Warm PW1 29.2  4 22.8  0.6 14  1.1 23.9  1.2
PW2 19.9  1.1 7.4  0.4 6.5  0.3 59.6  1.2
MC1 70533.5  481.5 13860  504.3 19230.2  171.3 45035.9  1781.8
MC2 126400  3695.2 14926.2  719.7 18745.9  365 50774.6  4753.5
WS1 1596.6  135.9 393.6  33.5 200  9.9 3870.4  19.1
WS2 610.8  3.9 456.3 230.9  5 476.7  33.4
GW1 17237.9  488.2 5044.3  88.3 7415.2  182.7 35076.9  1550.5
GW2 13763.3  385.2 3957.4  158.3 2670.5  157.7 18592.3  919.1
SD1 2716.2  35.8 1224.2  48.2 696.8  18.5 2217.6  187.1
SD2 2967.5  98.4 1321.6  25.4 974.1  7.7 2710.5  61.5
EW1 67.1  4.4 18.7  0.4 14.8  0.6 29.6  0.8
EW2 72.1  3 37.5  2.4 25.1  0.8 263.8  18.8

a PW1 and PW2, soft woodchips (pine); MC1 and MC2, maize cobs; WS1 and WS2, wheat straw; GW1 and GW2, green waste; SD1 and SD2,
sawdust; EW1 and EW2, hard woodchips (eucalyptus).

1.5  0.07  1010 (maize cobs) copies of nirK g
1 dry substrate, In order to estimate how the abundance of the different
and 6.5  0.3  106 (pine woodchips) to 1.9  0.02  1010 (maize genes in the denitrifying pathway changed with respect to the
cobs) copies of nosZ g
1 dry substrate (Table 2). The NO
3 other steps in denitrification, the ratios of copies of nirS/nirK,
removal rate increases exponentially with the total copy and Snir/nosZ (nitrous oxide reductase) were determined
number of nitrite reductase genes (Snir) per gram substrate (Fig. 7). Increasing temperature increased the ratio of nirS/nirK,
and was significantly linearly correlated with the ln (Snir g
1 and Snir/nosZ, except for pine woodchips.
substrate) in cold and warm barrels ( y ¼ 0.45x
5.62; R2 ¼ 0.48;
p ¼ 0.012 and y ¼ 0.38x
3.68; R2 ¼ 0.73; p < 0.001 respectively;
where y ¼ NO
3 eN removal rate in g N m

3
d
1 and x ¼ ln

1
(copies Snir g substrate) (Fig. 5). Generally, the copies of Snir
were greater in warm than in cold barrels, except for sawdust.
A temperature increase of 10  C yielded 4-fold increases in Snir
(Fig. 6A).
The carbon substrates maize cobs and green waste had the
greatest bacterial population ranging from 18592.3  919.1  106
copies of 16S rRNA g
1 dry substrate (warm incubated green
waste) to 96761.9  2649.3  106 copies of 16S rRNA g
1 dry
substrate (cold incubated maize cobs), and the greatest Snir
per gram carbon substrate (Table 2). In contrast, nitrite reduc-
tase gene copies (Snir) normalized to total bacteria (16S rRNA
genes) of these substrates (maize cobs and green waste) were at
the lower end of the data generated in this study, ranging from
0.1  0.00 (cold incubated maize cobs) to 2.82  0.11 copies Snir
copies
1 16S rRNA g
1 dry substrate (warm incubated maize
cobs) (Fig. 6B). Cold incubated pine woodchips had the highest
Snir copy number normalized to total bacteria (7.29  0.58 and
7.89  0.07 copies Snir copies
1 16S rRNA g
1 dry substrate), Fig. 5 e NitrateeN removal rate as a function of total nitrite
followed by eucalyptus woodchips for cold incubations reductase gene (Snir) copies for cold and warm incubated
(Fig. 6B). substrates. Linear regression statistics are reported in text.
5470 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
Fig. 6 e Total number of nitrite reductase genes (Snir)
normalized per gram carbon substrate (A) and normalized
to total bacteria (16S rRNA) (B) of the different carbon
substrates used in the barrels under cold and warm
incubation. PW1 and PW2, pine woodchips; MC1 and MC2,
maize cobs; WS1 and WS2, wheat straw; GW1 and GW2,
green waste; SD1 and SD2, sawdust; EW1 and EW2,
eucalyptus woodchips. Error bars are one standard error
(n [ 3).
For cold incubations the ratios of nirS/nirK within the
same carbon substrate (replicates) were not different from
each other applying the Wald confidence interval (95%),
except for pine wood. The same was observed for the ratios of
Snir/nosZ within the same carbon substrate in cold barrels,
whereas in warm barrels differences in ratios of nirS/nirK, or
Snir/nosZ were shown for each carbon source, except for nirS/
nirK ratios of warm incubated green waste and sawdust
barrels (Fig. 7).
3.4. Greenhouse gases
The inlet concentrations of dissolved N2OeN were below
the detection limit (<1.1 mg L
1). Therefore the measured
dissolved N2OeN and CH4 concentrations in the outlet
water of the barrels are the net dissolved N2OeN release
from the barrels in the outlet water. The dissolved N2OeN
release from the cold barrels in the outlet ranged from
Fig. 7 e Ratios of gene copies of nirS/nirK (A) and total
nitrite reductase (Snir) to nitrous oxide reductase (nosZ ) (B).
PW1 and PW2, pine woodchips; MC1 and MC2, maize cobs;
WS1 and WS2, wheat straw; GW1 and GW2, green waste;
SD1 and SD2, sawdust; EW1 and EW2, eucalyptus
woodchips. Error bars are one standard error (n [ 3).
below detection limit (sawdust) to 214.5 mg L
1 (wheat
straw) and from the warm barrels from below detection
limit (sawdust) to 1472.5 mg L
1 (wheat straw). Wheat straw
was the largest source of N2O for both cold and warm
incubations, followed by green waste in warm incubations.
Warm wheat straw barrels released almost 10% of the
removed NO
3 eN as dissolved N2OeN in the outlet water. All
substrates at the warmer temperature released on average
about seven times more dissolved N2OeN in the outlet than
cold barrels (Fig. 8).
The inlet concentration of dissolved CH4 was 5.4 mg CH4 L
1
for cold and 16.8 mg CH4 L
1 for warm barrels. There was little
net dissolved CH4 release in the outlet of woodchips (hard and
soft wood) and sawdust (<40 mg L
1) detected. Wheat straw
released some dissolved CH4 in the outlet water at cold incu-
bations (139 mg L
1 and 1201 mg L
1) and maize cobs released
large amounts of dissolved CH4 at cold incubations
(10,600 mg L
1 and 7375 mg L
1) in the outlet of the barrels, but
less dissolved CH4 at warm incubation. Barrels of green waste
released dissolved CH4 in the outlet from cold and warm
barrels, with an average of 2970 mg L
1 and 3870 mg L
1,
respectively (Fig. 8).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
Fig. 8 e Dissolved nitrous oxide (A) and methane (B)
concentrations in the outlet water of different carbon
substrates in cold and warm barrels. PW1 and PW2, soft
woodchips (pine); MC1 and MC2, maize cobs; WS1 and
WS2, wheat straw; GW1 and GW2, green waste; SD1 and
SD2, sawdust; EW1 and EW2, hard woodchips (eucalyptus).
4. Discussion
In this study, several different carbon substrates (maize cobs,
wheat straw, green waste, sawdust, hardwood and softwood)
receiving NO


3 from a simulated household effluent (inlet NO3

1
concentration between 14 and 18 mg L ) were examined to
determine factors controlling NO
3 removal and the extent of
possible adverse effects. The denitrifying bacterial commu-
nities in the different barrels were also examined to deter-
mine whether microbial community structure could account
for differences in activity (NO
3 removal, dissolved GHG
concentrations). The experimental barrels had been operating
for 2.5 years prior to these measurements, thereby eliminating
short term study effects (i.e., high TOC release coupled with
high NO
3 removal rates), as have been described in other
column and barrel studies (Greenan et al., 2009; Cameron and
Schipper, 2010; Soares and Abeliovich, 1998). In our study
a single sampling was taken. However, we consider that
steady state had been reached in the microbial community,
which allow comparisons between substrates; e.g., Warneke
et al. (2011a) found only very small differences in dissolved
5471
N2O and CH4 concentrations along the length of a field-scale
woodchip bioreactor during a sampling period of one year.
4.1. Nitrate removal and microbial processes
The mean NO
3 eN removal rates of the experimental barrels
were less than the NO
3 eN removal rates reported by Cameron
and Schipper (2010) in the same experimental barrels for the
previous 2.5 years, and less than the reported rates of most
other column studies with alternative carbon substrates
(Gibert et al., 2008; Saliling et al., 2007; Greenan et al., 2006;
Della Rocca et al., 2005; Shao et al., 2008; Soares and
Abeliovich, 1998). These lower NO
3 removal rates were most
likely due to the age of the carbon material (>2.5 years in use)
and the 10-fold lower NO
3 eN inlet concentration than used by
Cameron and Schipper (2010). For example, in this study,
NO
3 eN removal rates of cold incubated maize cobs and wheat
straw were clearly limited by NO


3 eN concentrations (NO3 eN

1
outlet concentrations <1 mg L ; Table 1). Nitrate removal
rates of pine and eucalyptus woodchip and sawdust ranged
from 1.3 to 4.4 g N m
3 d
1 and were at the lower end of
removal rates determined for woodchip bioreactors in the
field (Schipper et al., 2010). Maize cobs, followed by wheat
straw and green waste, exhibited a higher NO
3 removal rate
than wood substrates in this study, as also reported by
Cameron and Schipper (2010) for the same experimental
system. However, the NO
3 eN removal rates for wood
substrates in this study were in the same range as the NO
3 eN
removal rates (3.9 g N m
3 d
1) measured by Greenan et al.
(2009) in a column study. Other column studies with wood-
chips showed NO
3 eN removal rates 2e10 times higher than
this study (Robertson, 2010; Saliling et al., 2007).
As expected, there was good evidence that the mechanism
for NO
3 eN removal in the substrates was most likely micro-
bial denitrification, because the measured in vitro DR þ C/N of
each experimental barrel were higher than many other NO
3
reducing ecosystems e.g., denitrification walls (Schipper et al.,
2005; Moorman et al., 2010), forested land-based wastewater
treatment system (Barton et al., 2000), riparian forest sites
(Groffmann et al., 1992), a natural wetland and a constructed
wetland (Duncan and Groffmann, 1994). Additionally, nitrite
reductase genes (nirS and nirK ), which are responsible for the
second step of denitrification, were on average more abun-
dant in this study (Table 2) than in constructed wetlands
(Chon et al., 2010), or rice fields (Yoshida et al., 2009).
Furthermore, the significant linear relationship of the increase
of NO
3 removal, and the increase of measured DR þ C/N,
indicated that microbial denitrification was responsible for
the NO
3 eN removal, regardless of the carbon substrate in the
experimental barrels and showed that the acetylene inhibi-
tion method was a good measure for comparative NO
3
removal estimations between C substrates (Fig. 2).
Although seven of the 12 cold barrels, and eight of the
warm barrels produced small amounts of NHþ 4 , neither
anammox or DRNA appeared to be significant contributors to
NO
þ
3 removal, because of the low NH4 eN concentration

1
(<0.8 mg L ) at the outlet. Both Gibert et al. (2008) and
Greenan et al. (2006) also suggested that DNRA is a minor
process involved in NO
3 removal (less than 5%).
5472 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
As NO
3 was depleted in the cold incubated maize cobs and
wheat straw barrels, methanogenic bacteria were able to
compete successfully with denitrifiers for carbon as suggested
by the high dissolved CH4 production of cold incubated maize
cobs and wheat straw barrels. Although NO
3 eN concentra-
tions were above 2 mg L
1 in the outlet of cold green waste
barrels and warm maize cobs and green waste barrels, we
observed dissolved CH4 production (Table 1, Fig. 8), which
suggests that methanogenes may occur even at relatively
moderate NO
3 concentrations. It is likely that once the
microbial consumption of NO
3 exceeded diffusion of NO3


within the carbon substrate, methanogenes could develop in
the interior of the substrate.
4.2. Factors controlling NO

3 removal
In general, denitrification is primarily controlled by carbon
availability, NO


3 , NO2 , sulphide, temperature, DO, and the
number of denitrifiers (Firestone and Davidson, 1989;
Seitzinger et al., 2006). In this study, carbon availability and
temperature were identified as the main factors limiting
nitrate removal in the experimental barrel systems, when
NO

1
3 concentrations were more than 1 mg L ; below this
concentration NO
3 limited denitrification.
The warm barrels removed more NO
3 than the cold
barrels, with a Q10 factor of 1.2  0.13 (Fig. 1). Cameron and
Schipper (2010) found a greater temperature dependence of
NO
3 removal (Q10 ¼ 1.6) in the same experimental system, but
these measurements were made with 10 times higher NO
3 wetlands or rice fields (Chon et al., 2010; Yoshida et al., 2009).
inlet concentrations, whereas in the present study NO
3 However, the abundance of denitrification genes in pine and
limited in some barrels the NO
3 removal. Studies of woodchip
bioreactors by Robertson et al. (2008), Elgood et al. (2010) and
Warneke et al. (2011a), also determined higher Q10s than in the
present study.
In most of the other experimental barrels, carbon amend-
ment (glucose) increased the denitrification activity (Fig. 4), as
reported by Warneke et al. (2011a) for a field-scale woodchip
bioreactor. Furthermore, NO
3 removal and the denitrification
rate (DR þ N; removing NO
3 limitation) were found to increase
linearly with the availability of carbon (measured as respirable
carbon, Fig. 3). Therefore, nitrate removal in the experimental
barrels was most likely limited by carbon availability, except
for cold maize cobs and cold wheat straw barrels. Nitrate
removal in cold maize cobs and cold wheat straw barrels was
limited by NO


3 likely due to low NO3 eN outlet concentrations

1
below 1 mg L (Fig. 4; Table 1). These findings confirm that in
anaerobic, NO
3 rich environments, carbon limits microbial
denitrification (Knowles, 1982; Reddy et al., 1982). This study
shows that respirable carbon measurements could also be
used to make comparative estimations of NO
3 removal in
carbon limited systems (Fig. 3).
In this study, the pH decreased slightly from inlet to outlet
as found in other studies (Van Driel et al., 2006; Robertson
et al., 2005; Robertson and Merkley, 2009), but was still in the
optimal range for denitrifiers (Bremner and Shaw, 1958;
Knowles, 1982). In contrast Warneke et al. (2011a) reported
an increase in pH along the length of a field-scale woodchip
bioreactor.
DO concentrations decreased from above 6 mg L
1 at the
inlet, to below 2 mg L
1 at the outlet. Robertson (2010) also
measured a similar decrease in DO in a woodchip column
study and that a substantial portion of microbially available
carbon was consumed by aerobic respiration. However, Gibert
et al. (2008) measured declines in DO from 4 to 1.2 mg L
1 in
the first 10 cm of a 90 cm long woodchip column. This fine-
scale work suggested that most of the substrate close to the
inlet served to provide anaerobic conditions for denitrifiers.
The NO
3 removal rate was significantly correlated to the
copy number of nitrite reductase genes (nirS and nirK ) (Fig. 5).
Furthermore the average nitrite reductase gene copies per
gram dry substrate increased 4-fold with a temperature
increase of 10  C (Fig. 6A), but the NO
3 eN removal rate
increased 1.2 times. This temperature dependence of deni-
trification genes corresponds with seasonal measurements of
nitrite reductase gene copies in wetlands (Chon et al., 2010).
The copies of 16S rRNA genes also increased with tempera-
ture, with the exception of the sawdust barrel (Table 2), so the
greater copy number of denitrification genes in the substrate
at higher temperature was also partially due to an increase in
bacterial biomass.
4.3. Denitrifying bacterial communities
Abundance of nirS, nirK and nosZ genes in maize cob, green
waste, sawdust and wheat straw ranged from 107 to 1011
copies g
1 dry substrate (Table 1, Fig. 6A), and these values
were on average greater than those measured in constructed
eucalyptus woodchips were slightly lower, but in the same
range as the wetland and rice field studies (Chon et al., 2010;
Yoshida et al., 2009). But woodchips, especially those from
cold incubations, showed the greatest abundance of nitrite
reductase genes as a proportion of total bacterial DNA (16S
rRNA), coupled with low 16S rRNA gene copies (Table 2,
Fig. 6B). Green waste and maize cobs, particularly cold incu-
bated maize cobs, had a low copy number of denitrification
genes as a proportion of total bacteria, and gave high 16S rRNA
gene copies (Table 2, Fig. 6B). Therefore, the bacterial
community in green waste and maize cob barrels had a low
ratio of denitrifying genes per copy number of 16S rRNA genes
even though green waste and maize cobs had on average more
denitrifiers per gram substrate than woodchips (Table 2,
Fig. 6). Consequently, a substantial proportion of carbon in
green waste and maize cob barrels was likely consumed by
non-denitrifying bacteria, fungi and/or yeasts, whereas
a greater proportion of C released from woodchips appeared
to be consumed by denitrifiers.
The ratios of nirS/nirK, and Snir/nosZ, were similar between
replicate barrels in cold incubations, except for pine wood
barrels (Fig. 7). In warm incubations, there was much greater
variation in replicates, and the ratios of nirS/nirK, and nir/nosZ,
varied significantly among carbon substrates (Fig. 7). There-
fore we assume that it was likely that the composition of
denitrifying bacteria in replicate barrels under cold incubation
was very similar, but in warm barrels the denitrifying pop-
ulation varied greatly between replicates. Furthermore it is
likely that the composition of denitrifier was also very distinct
in different carbon substrates, in both warm and cold barrels.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
At warm temperatures, the nirS/nirK ratio increased
(except in one pine woodchip barrel), suggesting that higher
temperature enhanced growth of nirS containing bacteria, or
did not encourage the growth of nirK containing bacteria
(Fig. 7). The nirS/nirK has been shown to be greater in unfer-
tilized soils, compared to those that were fertilized (Hallin
et al., 2009). The ratio also decreased with the presence of
cattle and increased with increasing nitrate, pH and soil
moisture (Philippot et al., 2009). Similar temperature depen-
dence was observed with the nitrite reductase/nitrous oxide
reductase gene ratio (Snir/nosZ ). The Snir/nosZ was signifi-
cantly higher in warm barrels than in cold barrels (Fig. 7). This
finding corresponded with the higher N2O concentrations in
warm barrels compared to cold barrels, and the observed
increase in N2O emission at higher temperatures in previous
studies (Warneke et al., 2011a; Teiter and Mander, 2005;
Johansson et al., 2003). High N2O fluxes have been shown to
correlate with a low ratio of nosZ/narG, where narG is the gene
responsible for nitrate reduction the first step in the denitri-
fication pathway (Philippot et al., 2009). Similarly, a high ratio
of N2O/N2O þ N2 has also been shown to correlate with the
Snir/nosZ ratio (Cuhel et al., 2010).
4.4. Evaluation of the different carbon substrates
Maize cobs, wheat straw and green waste barrels removed
more NO
3 than wood substrates. The dissolved N2OeN
production of maize cobs, green waste and wood-filled
barrels was moderate and the dissolved N2OeN outlet
concentrations ranged from 7 to 110 mg L
1 for cold barrels,
and from 207 to 566 mg L
1 for warm barrels. Wheat straw
produced on average about three times more dissolved N2O
(Fig. 8) than other carbon substrates. This corresponded with
the relatively high ratio of nitrite reductase gene copies to
P
nitrous oxide reductase gene copies ( nir/nosZ ) in the
wheat straw barrels (Fig. 7), which lead likely to more N2O
production than N2O consumption. The N2OeN release from
wheat straw in the effluent was almost 10% of the removed
NO
3 eN, which is also about three times greater than the
dissolved N2OeN release of a field-scale wood chip denitri-
fication bed (Warneke et al., 2011a). Only sawdust showed no
N2O release.
Maize cobs had the highest NO
3 removal rate and were
additionally limited by NO
3 concentration. Therefore,
a higher NO
3 removal rate could be expected for maize cobs if
it was loaded with more NO
3 as shown by Cameron and
Schipper (2010). However, maize cobs also released high
concentrations of TOC and dissolved CH4. It would be
expected that CH4 release from maize cobs in the outlet water
would decrease with a higher NO
3 concentration in inlet
water because denitrification would outcompete methano-
genesis. Additionally maize cobs had a low denitrifier/
bacteria ratio, which would probably yield substantial carbon
loss due to carbon consumption by non-denitrifiers, whereas
woodchips seemed to be an ideal substrate for denitrifying
bacteria. Furthermore, wood substrate showed moderate
NO
3 removal rates, with almost no adverse effects. As
demonstrated in previous studies (Warneke et al., 2011a;
Schipper et al., 2010; Robertson, 2010; Long et al., 2011)
5473
woodchips provide sustained NO
3 removal due to slow
decomposition of wood in the bioreactor.
5. Conclusions
This study suggested that microbial denitrification was the
main mechanism for nitrate removal for all carbon sources
tested, due to the high in vitro DR, the linear relationship
between NO
3 removal and in vitro DR þ C/N, high abundance
of nitrite reductase genes, and uniformly low NHþ 4
concentrations.
The denitrification process in the experimental barrels was
limited by carbon availability and temperature, except when
NO

1
3 eN outlet concentrations were below 1 mg L , when
NO
eN limitation occurred. The NO

eN removal rate was
3 3
dependent on the quantity of microbially available carbon,
which varied between carbon sources. Both the acetylene inhi-
bition method for measuring denitrification activity, and the
quantification of denitrification genes were good approaches for
determining comparative NO
3 removal in carbon limited
systems (Figs. 3 and 5). It would be useful to determine and
compare the slope of the linear regressions between NO
3
removal and Ln (Snir g
1 substrate) in different ecosystems to
estimate the nitrate removal rates only by the copy number of
nitrite reductase genes in similar ecosystems (Fig. 5).
Greatest dissolved N2O release in the outlet water was
detected for wheat straw and was about 10% of the removed
NO
3 eN, which was much greater than reported in previous
studies for wood substrates. Methanogenesis could compete
with denitrification when NO
3 eN concentrations were below
P
2 mg L
1 and nir/nosZ ratio was high.
Maize cobs had the highest NO
3 eN removal rate, but
released elevated amounts of TOC, and substantial carbon
consumption by non-denitrifiers was likely. Wood substrates
exhibited moderate and sustained NO
removal, and
3
appeared to be ideal for denitrifiers under anaerobic, high NO
3
conditions. Therefore it may be useful to combine maize cobs
with woodchips, to enhance C availability and increase the
denitrifying activity in the woodchip material. This approach
would possibly generate higher NO
3 eN removal rates than
woodchips alone, with only moderate adverse effects.
Furthermore, findings in this study suggest that increased
temperatures enhance the growth of nirS-containing and
nosZ-lacking bacteria, but further research is needed to
understand this effect.
Acknowledgments
This research was made possible by funding from Waikato-
Link Ltd. (New Zealand) and Hans-Sauer-Foundation
(Germany). Additional support was provided by grant number
5 P42 ES04699 from the National Institute of Environmental
Health Sciences, NIH. Its contents are solely the responsibility
of the authors and do not necessarily represent the official
views of the NIEHS, NIH. Many thanks to the technicians of
the Science and Engineering department of the University of
Waikato for their help.
5474 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
references
American Public Health Association, 1992. Standard Methods for
the Examination of Water and Wastewater, eighteenth ed.
USA.
Barton, L., Schipper, L.A., Smith, C.T., 2000. Denitrification
enzyme activity is limited by soil aeration in a wastewater-
irrigated forest soil. Biol. Fertil. Soils 32, 385e389.
Braker, G., Fesefeldt, A., Witzel, K.P., 1998. Development of PCR
primer systems for amplification of nitrite reductase genes
(nirK and nirS ) to detect denitrifying bacteria in environmental
samples. Appl. Environ. Microbiol. 64, 3769e3775.
Bremner, J.M., Shaw, K., 1958. Denitrification in soil: II. Factors
affecting denitrification. J. Agr. Sci. 51, 40e52.
Cameron, S.C., Schipper, L.A., 2010. Nitrate removal and
hydraulic performance of carbon substrates for potential use
in denitrification beds. Ecol. Eng. 36, 1588e1595.
Canfield, D.E., Glazer, A.N., Falkowski, P.G., 2010. The evolution
and future of earth’s nitrogen cycle. Science 330, 192e196.
Cheneby, D., Hartmann, A., Henault, C., Topp, E., Germon, J.C.,
1998. Diversity of denitrifying microflora and ability to reduce
N2O in two soils. Biol. Fertil. Soils 28, 19e26.
Cheneby, D., Perrez, S., Devroe, C., Hallet, S., Couton, Y.,
Bizouard, F., Iuretig, G., Germon, J.C., Philippot, L., 2004.
Denitrifying bacteria in bulk and maize-rhizospheric soil:
diversity and N2O-reducing abilities. Can. J. Microbiol. 50,
469e474.
Cheng, W., Coleman, D.C., 1989. A simple method for measuring
CO2 in a continuous airflow system: modifications to the
substrate-induced respiration technique. Soil Biol. Biochem.
21, 385e388.
Chon, K., Kim, K., Chang, N.I., Cho, J., 2010. Evaluating wastewater
stabilizing constructed wetland, through diversity and
abundance of the nitrite reductase gene nirS, with regard to
nitrogen control. Desalination 264, 201e205.
Cuhel, J., Simek, M., Laughlin, R.J., Bru, D., Cheneby, D.,
Watson, C.J., Philippot, L., 2010. Insights into the effect of soil
pH on N(2)O and N(2) emissions and denitrifier community
size and activity. Appl. Environ. Microbiol. 76, 1870e1878.
Dambreville, C., Hallet, S., Nguyen, C., Morvan, T., Germon, J.-C.,
Philippot, L., 2006. Structure and activity of the denitrifying
community in a maize-cropped field fertilized with
composted pig manure or ammonium nitrate. FEMS Microbiol.
Ecol. 56, 119e131.
Della Rocca, C., Belgiorno, V., Meric, S., 2005. Cotton-supported
heterotrophic denitrification of nitrate-rich drinking water
with sand filtration post treatment. Water SA 31, 1022e1028.
Della Rocca, C., Belgiorno, V., Meric, S., 2006. Heterotrophic/
autotrophic denitrification (HAD) of drinking water:
prospective use for permeable reactive barrier. Desalination
210, 194e204.
Duncan, C.P., Groffmann, P.M., 1994. Comparing microbial
parameters in natural and constructed wetlands. J. Environ.
Qual. 23, 298e305.
Elgood, Z., Robertson, W.D., Schiff, S.L., Elgood, R., 2010. Nitrate
removal and greenhouse gas production in a stream-bed
denitrifying bioreactor. Ecol. Eng. 36, 1575e1580.
Enwall, K., Philippot, L., Hallin, S., 2005. Activity and composition
of the denitrifying bacterial community respond differently to
long-term fertilization. Appl. Environ. Microbiol. 71,
8335e8343.
Firestone, M.K., Davidson, E.A., 1989. Microbiological basis of NO
and N2O production and consumption in soil. In: Exchange of
Trace Gases Between Terestrial Ecosystems and the
Atmosphere. John Wiley, USA, pp. 7e21.
Gibert, O., Pomierny, S., Rowe, I., Kalin, R.M., 2008. Selection of
organic substrates as potential reactive materials for use in
a denitrification permeable reactive barrier (PRB). Bioresour.
Technol. 99, 7587e7596.
Greenan, C.M., Moorman, T.B., Kaspar, T.C., Parkin, T.B.,
Jaynes, D.B., 2006. Comparing carbon substrates for
denitrification of subsurface drainage water. J. Environ. Qual.
35, 824e829.
Greenan, C.M., Moorman, T.B., Parkin, T.B., Kaspar, T.C.,
Jaynes, D.B., 2009. Denitrification in wood chip bioreactors at
different water flows. J. Environ. Qual. 38, 1664e1671.
Groffmann, P.M., Gold, A.J., Simmons, R.C., 1992. Nitrate
dynamics in riparian forests: microbial studies. J. Environ.
Qual. 21, 666e671.
Hallin, S., Jones, C.M., Schloter, M., Philippot, L., 2009.
Relationship between N-cycling communities and ecosystem
functioning in a 50-year-old fertilization experiment. ISME J. 3,
597e605.
Henry, S., Baudoin, E., Lopez-Gutierrez, J.C., Martin-Laurent, F.,
Brauman, A., Philippot, L., 2005. Quantification of denitrifying
bacteria in soils by nirK gene targeted real-time PCR (vol. 59,
pp. 327, 2004). J. Microbiol. Methods 61, 289e290.
Henry, S., Bru, D., Stres, B., Hallet, S., Philippot, L., 2006.
Quantitative detection of the nosZ gene, encoding nitrous
oxide reductase, and comparison of the abundances of 16S
rRNA, narG, nirK, and nosZ genes in soils. Appl. Environ.
Microbiol. 72, 5181e5189.
Howarth, R.W., Sharpley, A.W., Walker, D., 2002. Sources of
nutrient pollution to coastal waters in the United States:
implications for achieving coastal water quality goals.
Esturaries 25, 656e676.
Johansson, A.E., Kasimir Klemedtsson, A., Klemedtsson, L., 2003.
Nitrous oxide exchanges with the atmosphere of
a constructed wetland treating wastewater. Tellus 55B,
737e750.
Kandeler, E., Deiglmayr, K., Tscherko, D., Bru, D., Philippot, L.,
2006. Abundance of narG, nirS, nirK, and nosZ genes of
denitrifying bacteria during primary successions of a glacier
foreland. Appl. Environ. Microbiol. 72, 5957e5962.
Knowles, R., 1982. Denitrification. Microbiol. Rev. 46, 43e70.
Lindsay, A.E., Colloff, M.J., Gibb, N.L., Wakelin, S.A., 2010. The
abundance of microbial functional genes in grassy woodlands
is influenced more by soil nutrient enrichment than by recent
weed invasion or livestock exclusion. Appl. Environ. Microbiol.
76, 5547e5555.
Long, L.M., Schipper, L.A., Bruesewitz, D.A., 2011. Long-term
nitrate removal in a denitrification wall. Agr. Ecosyst. Environ.
140, 514e520.
Lopez-Gutierrez, J.C., Henry, S., Hallet, S., Martin-Laurent, F.,
Catroux, G., Philippot, L., 2004. Quantification of a novel group
of nitrate-reducing bacteria in the environment by real-time
PCR. J. Microbiol. Methods 57, 399e407.
Moorman, T.B., Parkin, T.B., Kaspar, T.C., Jaynes, D.B., 2010.
Denitrification activity, wood loss, and N2O emissions over 9
years from a wood chip bioreactor. Ecol. Eng. 36, 1567e1574.
Philippot, L., 2002. Denitrifying genes in bacterial and Archaeal
genomes. Biochim. Biophys. Acta 1577, 355e376.
Philippot, L., Cuhel, J., Saby, N.P.A., Cheneby, D., Chronakova, A.,
Bru, D., Arrouays, D., Martin-Laurent, F., Simek, M., 2009.
Mapping field-scale spatial patterns of size and activity of the
denitrifier community. Environ. Microbiol. 11, 1518e1526.
Phoenix, G.K., Hicks, W.K., Cinderby, S., Kuylenstierna, J.C.I.,
Stock, W.D., Dentener, F.J., Giller, K.E., Austin, A.T., Lefroy, R.
D.B., Gimeno, B.S., Ashmore, M.R., Ineson, P., 2006.
Atmospheric nitrogen deposition in world biodiversity
hotspots: the need for a greater global perspective in
assessing N deposition impacts. Glob. Change Biol. 12 (3),
470e476.
Rabalais, N.N., 2002. Nitrogen in aquatic ecosystems. Ambio 31
(2), 102e112.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 6 3 e5 4 7 5
Reddy, K.R., Rao, P.S.C., Jessup, R.E., 1982. The effect of Carbon
mineralization on denitrification kinetics in mineral and
organic soils. Soil Sci. Soc. Am. J. 46, 62e68.
Robertson, W.D., Merkley, L.C., 2009. In-stream bioreactor for
agricultural nitrate treatment. J. Environ. Qual. 38, 230e237.
Robertson, W.D., Ford, G.I., Lombardo, P.S., 2005. Wood-based
filter for nitrate removal in septic systems. Am. Soc. Agr. Eng.
48 (1), 121e128.
Robertson, W.D., Vogan, J.L., Lombardo, P.S., 2008. Nitrate removal
rates in a 15-year old permeable reactive barrier treating septic
system nitrate. Ground Water Monit. Remediat. 28, 65e72.
Robertson, W.D., 2010. Nitrate removal rates in woodchip media
of varying age. Ecol. Eng. 36, 1581e1587.
Saliling, W.J.B., Westerman, P.W., Losordo, T.M., 2007. Wood
chips and wheat straw as alternative biofilter media for
denitrification reactors treating aquaculture and other
wastewaters with high nitrate concentrations. Aquacult. Eng.
37, 222e233.
Schipper, L.A., Barkle, G.F., Vojvodic-Vukovic, M., 2005. Maximum
rates of nitrate removal in a denitrification wall. J. Environ.
Qual. 34, 1270e1276.
Schipper, L.A., Robertson, W.D., Gold, A.J., Jaynes, D.B., Cameron, S.
C., 2010. Denitrifying bioreactors e An approach for reducing
nitrate loads to receiving waters. Ecol. Eng. 36, 1532e1543.
Seitzinger, S., Harrison, J.A., Böhlke, J.K., Bouwman, A.F.,
Lowrance, R., Peterson, B., Tobias, C., Van Drecht, G., 2006.
Denitrification across landscapes and waterscapes:
a synthesis. Ecol. Appl. 16, 2064e2090.
Shao, L., Xu, Z.X., Jin, W., Yin, H.L., 2008. Rice husk as carbon
source and bilofilm carrier for water denitrification. Polish J.
Environ. Stud. 18 (4), 693e699.
Soares, M.I.M., Abeliovich, A., 1998. Wheatstraw as substrate for
water denitrification. Water Res. 32 (12), 3790e3794.
Stres, B., Mahne, I., Avgustin, G., Tiedje, J.M., 2004. Nitrous oxide
reductase (nosZ ) gene fragments differ between native and
cultivated Michigan soils. Appl. Environ. Microbiol. 70, 301e309.
5475
Teiter, S., Mander, U., 2005. Emissions of N2O, N2, CH4 and CO2
from constructed wetlands for wastewater treatment and
from riparian buffer zones. Ecol. Eng. 25, 528e541.
Tiedje, J.M., Simkins, S., Groffmann, P.M., 1989. Perspectives on
measurement of denitrification in the field including
recommended protocols for acetylene based methods. Plant
Soil 115, 261e284.
Van Driel, P.W., Robertson, W.D., Merkley, L.C., 2006.
Denitrification of agricultural drainage using wood based
reactors. Am. Soc. Agr. Biol. Eng. 49 (2), 565e573.
Volokita, M., Abeliovich, A., Soares, M.I.M., 1996a. Denitrification
of groundwater using cotton as energy source. Water Sci.
Technol. 34, 379e385.
Volokita, M., Belkin, S., Abeliovich, A., Soares, M.I.M., 1996b.
Biological denitrification of drinking water using newspaper.
Water Res. 30, 965e971.
Warneke, S., Schipper, L.A., Bruesewitz, D.A., McDonald, I.,
Cameron, S., 2011a. Rates, controls and potential adverse
effects of nitrate removal in a denitrification bed. Ecol. Eng. 37,
511e522.
Warneke, S., Schipper, L.A., Bruesewitz, D.A., Baisden, T.W.,
2011b. A comparison of different approaches for measuring
denitrification rates in a nitrate removing bioreactor. Water
Res.. doi:10.1016/j.watres.2011.05.027.
Weiss, R.F., Price, B.A., 1980. Nitrous oxide solubility in water and
seawater. Mar Chem. 8, 347e359.
Wood, D.W., et al., 2001. The genome of the natural genetic engineer
Agrobacterium tumefaciens C58. Science 294, 2317e2323.
Yamamoto, S., Alcauskas, J.B., Crozier, T.E., 1976. Solubility of
methane in distilled water and seawater. J. Chem. Eng. Data
21, 78e80.
Yoshida, M., Ishii, S., Otsuka, S., Senoo, K., 2009. Temporal shifts
in diversity and quantity of nirS and nirK in a rice paddy field
soil. Soil Biol. Biochem. 41, 2044e2051.
Zumft, W., 1997. Cell biology and molecular basis of
denitrification. Microbiol. Mol. Biol. Rev. 61, 533e616.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fine-scale bacterial community dynamics and the taxaetime

relationship within a full-scale activated sludge bioreactor

George F. Wells a, Hee-Deung Park a,1, Brad Eggleston b, Christopher A. Francis c,

Craig S. Criddle a,*
a
Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 94305, USA
b
Palo Alto Regional Water Quality Control Plant, 2501 Embarcadero Way, Palo Alto, CA 94303, USA
c
Department of Environmental Earth System Science, Stanford University, Stanford, CA 94305, USA

article info abstract

Article history:
Received 27 April 2011
Received in revised form
18 July 2011
Accepted 6 August 2011
Available online 16 August 2011
Keywords:
Activated sludge
Microbial community dynamics
Taxaetime relationship
T-RFLP
Multivariate statistics
16S rRNA
In activated sludge bioreactors, aerobic heterotrophic communities efficiently remove
organics, nutrients, toxic substances, and pathogens from wastewater, but the dynamics of
these communities are as yet poorly understood. A macroecology metric used to quantify
community shifts is the taxaetime relationship, a temporal analog of the speciesearea
curve. To determine whether this metric can be applied to full-scale bioreactors, activated
sludge samples were collected weekly over a one-year period at a local municipal waste-
water treatment plant. Bacterial community dynamics were evaluated by monitoring 16S
rRNA genes using Terminal Restriction Fragment Length Polymorphism (T-RFLP), corrob-
orated by clone libraries. Observed taxa richness increased with time according to a power
law model, as predicted by macroecological theory, with a power law exponent of
w ¼ 0.209. The results reveal strong long-term temporal dynamics during a period of stable
performance (BOD removal and nitrification). Community dynamics followed a gradual
succession away from initial conditions rather than periodicity around a mean “equilib-
rium”, with greater within-month then among-month community similarities. Changes in
community structure were significantly associated via multivariate statistical analyses
with dissolved oxygen, temperature, influent silver, biomass (MLSS), flow rate, and influent
nitrite, cadmium and chromium concentrations. Overall, our results suggest patterns of
bacterial community dynamics likely regulated in part by operational parameters and
provide evidence that the taxaetime relationship may be a fundamental ecological pattern
in macro- and microbial systems.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction promoting degradation of oxygen-depleting organics, trans-

formation of toxic substances, and removal of nutrients from
Biological wastewater treatment is, in essence, an endeavor in water; thus, a firm understanding of the microbial ecology of
microbial ecological engineering. The goal of this endeavor is wastewater treatment bioreactors is essential (Rittmann et al.,
to manage microbial communities for the good of society by 2006). Performance changes observed at the treatment plant

* Corresponding author. Environmental Engineering and Science Program, Department of Civil and Environmental Engineering, Yang &
Yamazaki Environment & Energy Bld., Rm. 151, 473 Via Ortega MC-4020, Stanford University, Stanford, CA 94305, USA. Tel.: þ1 650 723
9032; fax: þ1 650 725 3164.
E-mail address: criddle@stanford.edu (C.S. Criddle).
1
Present address: School of Civil, Environmental and Architectural Engineering, Korea University, Seoul 136-713, South Korea.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.006
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
scale (hundreds of cubic meters) are the emergent properties
of an incredibly diverse assembly of w1018 individual micro-
bial cells (1e2 cubic micrometers) (Curtis et al., 2003). Many of
the critical process failures (i.e. unacceptable reduction in
final effluent quality) that arise during biological wastewater
treatment are likely attributable to variations in the relative
abundance or activity of these cells (Graham and Smith, 2004).
Such variations in microbial community structure are thought
to be influenced by a combination of deterministic (reactor
design, environmental and operational variables) and
stochastic (probability of microbial dispersal into or out of
a reactor) properties (Curtis and Sloan, 2006). Critical ques-
tions important to rational design and operation of these
systems remain unanswered (Rittmann et al., 2006): What
community structures, and what range of community
dynamics, are optimal for different applications? What envi-
ronmental conditions trigger optimal community assembly
with its desired function? Are the conditions and outcomes
predictable, reproducible, and controllable? Understanding,
and eventually predicting, dynamics in community composi-
tion and its relationship to ecosystem function is one of the
key engineering problems that might be solved through
a more quantitative understanding of the microbial ecology of
wastewater treatment processes (Gentile et al., 2007; Graham
and Smith, 2004).
Removal of COD is catalyzed in biological wastewater treat-
ment plants (WWTPs) by activated sludge communities
composed of a diverse assemblage of chemoorganoheterotro-
phic microorganisms. Numerous studies in recent years have
characterized activated sludge microbial community structure
via culture-independent studies (Wagner et al., 2002; Xia et al.,
2010), and others have investigated succession and dynamics
in these communities, particularly within specific microbial
subpopulations such as nitrifiers (Wells et al., 2009),
phosphorus-accumulating organisms (Slater et al., 2010), deni-
trifiers (Gentile et al., 2007), and methanogens (Fernandez et al.,
1999). Many of these studies have occurred in lab-scale systems,
however, where selective pressures likely differ dramatically
from those in full-scale plants (Seviour and Nielsen, 2010).
Indeed, few studies have directly characterized long-term, fine-
scale microbial population dynamics in activated sludge from
full-scale WWTPs. Moreover, we lack a quantitative under-
standing of the relative influence of specific environmental or
operational drivers of community dynamics in full-scale
systems, and it is not clear whether these dynamics are
random or predictable.
Activated sludge bioreactors are excellent test beds for
fundamental questions in microbial ecology (Daims et al.,
2006). A longstanding area of inquiry in ecology as a whole
focuses on patterns of generation and maintenance of species
diversity through time and space. In macroecology, spatial
patterns in species diversity have long been recognized to
follow a SpecieseArea Relationship (SAR) of the form S ¼ cAz,
where S is species richness, A is the spatial scale of observa-
tion, c is an empirically derived constant, and z is a scaling
exponent that reflects species turnover. Accumulation of
microbial taxa richness with increasing spatial scales follow
a similar pattern (Horner-Devine et al., 2004). Conversely, the
speciesetime relationship (herein referred to as the tax-
aetime relationship [TTR]) has received comparatively little
5477
attention. First proposed by Preston (1960), the TTR describes
the increase in observed taxa richness, S, for increasing time
of observation, T, by a power law model, similar in form to the
SAR: S ¼ cTw, where the scaling exponent w is a reflection of
species turnover. The TTR has been documented in a rapidly
expanding literature in macroecology (White et al., 2006), and
has very recently been tested in a limited number of microbial
systems (Redford and Fierer, 2009; van der Gast et al., 2008).
The applicability of the power-law TTR to full-scale activated
sludge bioreactors, and a comparison of the associated
temporal scaling exponent (w) to previously reported values,
has not yet been explored.
The primary objectives of this study were thus two-fold: 1)
to monitor long-term, fine-scale temporal dynamics in the
overall bacterial community structure in activated sludge and
identify operational or environmental factors that most
significantly correlate with those dynamics; and 2) to assess
the applicability of a power law taxaetime relationship to
activated sludge microbial communities in full-scale biore-
actors. To this end, we collected weekly samples and
concurrently monitored 20 operational or environmental
variables over a 1-year period at a well-mixed, full-scale
municipal wastewater treatment bioreactor. We analyzed
shifts in bacterial community structure in this time series via
16S rRNA gene-based Terminal Restriction Fragment Length
Polymorphism (T-RFLP) corroborated by clone libraries. We
then employed multivariate statistical tools to identify
explanatory variables most significantly associated with these
dynamics, and we explored the relationship between time and
taxa in this critical application of environmental
biotechnology.
2. Materials and methods
2.1. Sample collection and site description
The Palo Alto Regional Water Quality Control Plant
(PARWQCP) is a WWTP located in Palo Alto, California
(37.2631 N, 122.0835 W) that treats primarily municipal
wastewater (100,000e170,000 m3 day1) for a population of
w225,000 people. Effluent is released to San Francisco Bay
after primary and secondary treatment (with nitrification),
tertiary treatment (dual media filtration), and disinfection.
Approximately 65% of total BOD removal is accomplished in
two trickling filters operated as roughing reactors; minimal
ammonia is removed in this unit. Immediately downstream of
the trickling filters are four parallel 6880 m3 aeration basins
comprising the activated sludge bioreactor, operated for
removal of remaining organics and for nitrification. The
average hydraulic residence time (HRT) in the activated sludge
bioreactor during the sampling period was 6.2 h, and the
solids residence time (SRT) was 21 days. The aeration basins
are well mixed with respect to microbial community struc-
ture, based on a baseline study in which biomass from three
locations within the bioreactor yielded essentially identical
microbial community composition via 16S rRNA T-RFLP (data
not shown). For a one-year period from February 2005 to
February 2006, a Sigma 900 refrigerated automatic sampler
(Hach, Loveland, CO) was used to obtain 24-h composite
5478 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
activated sludge samples (100 ml every 30 min) weekly from
the combined outlet of all four aeration basins. 1.5 ml
composite activated sludge samples were centrifuged onsite
for 3 min at 10,000 g, decanted, and stored at 20  C.
Concurrently, 20 operational and environmental variables
(influent flow rate; reactor temperature; reactor MLSS; dis-
solved oxygen (DO); influent and effluent COD, ammonium,
and nitrite; and primary influent chromium, nickel, copper,
zinc, silver, cadmium, lead, arsenic, selenium, and mercury)
were monitored, as described previously (Wells et al., 2009).
2.2. DNA extraction and PCR amplification of the 16S
rRNA gene
Activated sludge from 53 weekly samples (1 February 2005 to 2
February 2006) and a baseline sample from 29 September 2004
were washed using 1 ml TriseEDTA (TE) buffer (1 mM,
pH ¼ 7.0), concentrated at 10,000 g for 3 min, decanted, and
resuspended in 200 ml TE buffer. Duplicate genomic DNA
extractions from each sampling date were performed with the
MoBio Ultraclean Soil DNA Extraction Kit (Carlsbad, CA) as per
the manufacturer’s directions, with the exception that the
initial bead-beating step was performed with a Mini-
Beadbeater (Biospec Products, Bartlesville, OK) at 5000 rpm
for 1 min. Near full-length fragments (w1350 bp) of the 16S
rRNA gene were amplified using bacterial-specific primer set
8F (50 -AGAGTTTGATCCTGGCTCAG-30 ) and 1392R (50 -
ACGGGCGGTGTGTRC-30 ). For T-RFLP analyses, the forward
primer was labeled with the fluorophore 6-carboxyfluorescein
(FAM). Each 50 ml PCR mixture consisted of 0.25 mM of each
primer, 1 Fail-Safe PCR buffer F (Epicentre, Madison, WI), 1.25
U AmpliTaq LD Taq Polymerase (Applied Biosystems, Foster
City, CA), and 20e50 ng of genomic DNA. The PCR temperature
profile was as follows: 94  C for 5 min, then 35 cycles of 94  C
for 45 s, 55  C for 30 s, and 72  C for 90 s, followed by a final
extension at 72  C for 10 min. Amplicon presence and quality
was verified via 1.5% agarose gel electrophoresis.
2.3. Cloning, sequencing, and phylogenetic analyses
16S rRNA gene clone libraries were generated from four dates
of sampling (29 September 2004, 9 June 2005, 10 August 2005,
and 1 December 2005) to facilitate in silico choice of restriction
enzymes for T-RFLP and to provide a coarse baseline phylo-
genetic analysis of activated sludge microbial community
structure. Triplicate PCR products were pooled and purified
via the QIAEX II gel extraction kit (Qiagen, Valencia, CA) and
cloned using the TOPO TA Cloning Kit (Invitrogen, Carlsbad,
CA) or pGEM-T Easy Vector System with JM109 competent E.
Coli cells (Promega, Madison, WI), as per the manufacturer’s
instructions. Randomly picked clones were sequenced using
T7 and SP6 or M13 primers on ABI 3100 or 3730 automated
sequencers by MCLab (San Francisco, CA) or Elim Bio-
pharmaceuticals, Inc. (Hayward, CA), generating 172 near
full-length 16S rRNA gene sequences. Putative chimeric
sequences identified using BELLEROPHON (Huber et al., 2004)
were removed (47 total). Remaining 16S rRNA gene sequences
were aligned via the NAST utility (http://greengenes.lbl.gov/
NAST) (DeSantis et al., 2006) and inserted via parsimony to
an existing phylogenetic tree in ARB v5.1 (Ludwig et al., 2004)
containing all 236,469 sequences in the greengenes database
as of 18 November 2008. Following manual adjustment of the
alignment and removal of 15 sequences displaying low iden-
tity (<90%) to the closest related Genbank sequence, phylo-
genetic reconstruction via the neighbor-joining method was
performed in ARB for the 110 remaining sequences with the
Jukes-Cantor correction and Lane mask. A neighbor-joining
bootstrap analysis with 1000 replicates was also performed
in ARB under the same conditions. Consensus phylogenetic
assignments of 16S rRNA gene sequences were performed by
manual inspection of the phylogenetic tree and via the online
classifier at http://greengenes.lbl.gov/cgi-bin/nph-classify.cgi
by separately employing the RDP, NCBI, Hugenholtz, and
G2_Chip taxonomic nomenclatures. Nonchimeric 16S rRNA
gene sequences were submitted to Genbank and assigned
accession numbers HQ385507eHQ385631.
2.4. T-RFLP analyses
Temporal dynamics in activated sludge bacterial community
structure were monitored via Terminal Restriction Fragment
Length Polymorphism (T-RFLP) analyses targeting the 16S
rRNA gene. T-RFLP PCR conditions were identical to endpoint
PCR. Duplicate PCR products from each of 53 weeks of
sampling were pooled and purified using Montage PCR filter
units (Millipore, Billerica, MA). Purified products (400 ng/
digest) were digested in separate reactions with 0.25 U/ml of
RsaI or MspI restriction endonuclease (Promega, Madison, WI)
in buffer provided by the manufacturer for 3 h at 37  C,
providing two independent fingerprints of the bacterial
community structure for each time point. Digests were
repurified with Montage PCR filter units, loaded onto a capil-
lary electrophoresis machine (ABI 310 genetic analyzer, Foster
City, CA) with the MapMarker 1000 internal size standard
(BioVentures, Murfreesboro, TN), and analyzed at the Geno-
mics Technology Support Facility at Michigan State Univer-
sity. T-RF lengths (bps) and peak heights were extracted from
the resulting electropherograms using Genotyper v3.1 (ABI,
Foster City, CA) and standardized based on previously
described methods (Gentile et al., 2007). Briefly, total fluores-
cence signal was calculated as the sum of peak heights
between 50 and 1000 bp. Profiles were standardized to a total
sum of peak heights of 10,000 fluorescence units (FU). Peaks
with adjusted heights lower than the threshold of 50 FU were
removed, yielding a detection limit of 0.5% relative abun-
dance. The relative of abundance of each OTU (T-RF) was
calculated as the ratio of the peak height for that OTU to the
sum of peak heights for all T-RFs in a given profile and
expressed as a percentage.
2.5. Statistical analyses
Nonmetric Multidimensional Scaling (NMDS) analyses were
performed in PC-ORD v5.0 (MJM Software Design, Gleneden
Beach, OR). Pearson productemoment correlation coefficients
were calculated in R v2.5.1 (http://www.r-project.org/). Mantel
correlograms and Analyses of Similarity (ANOSIM) were per-
formed in PAST v2 (http://folk.uio.no/ohammer/past/).
Redundancy Analyses (RDA) were performed in Canoco for
Windows v4.53 (Plant Research International, Netherlands).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8 5479

Nonlinear regression analyses were performed in SPSS v19
(IBM, Somers, NY). A detailed discussion of statistical methods
is provided in the Supplementary Materials.
level (0.1 mg NO 
2 -N/L). Interestingly, while influent NO2 to
the activated sludge bioreactor was low relative to influent
NHþ 4 , it was highly variable, ranging from a minimum of
0.05 mg N/L to a maximum of 1.18 mg N/L. The observation of
low but significant and highly variable levels of NO 2 in the

3. Results influent to the PARWQCP bioreactor was unexpected, given

3.1. Bioreactor performance and operational conditions
Concurrent to weekly activated sludge sampling, we moni-
tored 20 operational or environmental variables that we
hypothesized may be significant drivers of temporal dynamics
in overall bacterial community structure. Variations in these
parameters are summarized in Table 1. Generally stable
influent flow rates led to maintenance of the hydraulic resi-
dence time within a narrow range (6.2  0.6 h). Reactor
temperature varied from a summer maximum of 25.4  C in
August to a winter minimum of 18.2  C in January. The
bioreactor was also operated with relatively high MLSS during
the winter of 2005, which led to a strong artificial correlation
between biomass and temperature in our time series. Reactor
performance was relatively stable across the sampling period.
The COD removal rate averaged 57  12%. However, 5 of 53
time points displayed COD removal rates less than 40%, with
a minimum of 22% on 24 January 2006. Moderate variation in
effluent COD was also observed, from a minimum of 18 mg/L
to a maximum of 67 mg/L. Nitrification performance was
significantly higher and more stable across the entire year of
sampling (95%  3% NHþ 4 removal rate), and accordingly,
effluent NHþ 4 was low (0.9  0.5 mg N/L). Secondary effluent
NO 2 was similarly maintained at a low and nearly constant
the lack of an explicitly designed nitrification function in the
upstream trickling filter and routinely negligible NO 2
concentrations in the raw sewage influent to the plant. While
aeration basin pH was not routinely monitored as part of this
study, 21 separate measurements between December 2004
and March 2005 were nearly constant (6.7  0.1). All 10 metals
monitored in the primary influent remained at ppb levels
throughout the sampling period, and ranged from
0.4  0.21 mg/l for cadmium to 136  27 mg/l for zinc.
3.2. Phylogenetic analysis of bioreactor bacterial
community structure
To phylogenetically characterize the bacterial community
structure in activated sludge at the PARWQCP and to inform
choice of restriction enzymes for T-RFLP analysis, nearly full-
length (w1350 bp) 16S rRNA genes from four time points were
cloned and sequenced. Phylogenetic inferences for these
sequences are summarized in Fig. 1, based on a neighbor-
joining phylogenetic tree (Fig. S1) and consensus classifica-
tions via the greengenes classification utility. All four dates of
sampling were dominated by sequences affiliated with the
Proteobacteria. 28% of pooled sequences clustered within the
Alphaproteobacteria, while 22% affiliated with the Betaproteo-
bacteria. Ten of the latter sequences (9% of total pooled

Table 1 e Operational conditions and bioreactor performance of the Palo Alto Regional Water Quality Control Plant. Unless
otherwise specified, influent and effluent refer to the entry and exit points to the activated sludge bioreactor.
Characteristics Rangea Average Standard
deviation

Wastewater inflow rate,b m3/day 82,000e181,000 105,000 11,500

Mixed liquor temperature,b  C 18.2e25.4 22.4 1.8
Biomass (mixed liquor suspended solids),c mg/L 2600e4900 3340 520
Dissolved Oxygen,d mg/L 3.08e4.50 3.88 0.37
Influent chemical oxygen demand (filtered),c mg/L 54e214 88.8 22.9
Effluent chemical oxygen demand (filtered),c mg/L 18.0e67.0 37.0 9.9
Influent ammonia (NH3 þ NHþ c
4 ), mg N/L 14.0e33.0 18.9 2.8
þ c
Effluent ammonia (NH3 þ NH4 ), mg N/L 0.20e2.10 0.84 0.52
Influent nitrite,c mg N/L 0.05e1.18 0.41 0.22
Effluent nitrite,c mg N/L 0.01e0.10 0.03 0.02
Primary Influent Silver,c mg/L 0.6e5.0 1.6 0.75
Primary Influent Arsenic,c mg/L 0.6e2.2 1.1 0.26
Primary Influent Cadmium,c mg/L 0.2e1.0 0.4 0.21
Primary Influent Chromium,c mg/L 1.8e9.0 4.4 1.40
Primary Influent Copper,c mg/L 43e110 66.6 15.2
Primary Influent Mercury,c mg/L 0.12e0.44 0.22 0.07
Primary Influent Nickel,c mg/L 5e12 7.2 1.37
Primary Influent Lead,c mg/L 2.6e9.0 5.0 1.53
Primary Influent Selenium,c mg/L 0.6e1.3 0.8 0.16
Primary Influent Zinc,c mg/L 84e220 135.9 26.7

a Range indicates minimum to maximum value.

b data based on daily measurements.
c data based on weekly measurements.
d data based on the 24-h average of measurements collected every 15 min in all four aeration basins.
5480 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8

Fig. 1 e Phylum or class-level (for Proteobacteria) phylogenetic affiliations for 110 near full-length (w1350 bp) 16S rRNA gene
sequences retrieved from four activated sludge sampling time points. The pie chart displays the percentage of clones from
the pooled libraries affiliated with each phylogenetic division, and the total number of clones affiliated with each division is
shown in parentheses.

sequences) affiliated with the Comamonadaceae, a family of
bacteria often reported in wastewater treatment systems
(Sadaie et al., 2007; Seviour and Nielsen, 2010). Eight Alphap-
roteobacterial sequences (7.3% of pooled sequences) were
affiliated with the Rhodobacterales order, and five putative
sphingomonad sequences (4.5% of pooled sequences) clus-
tered within the Sphingosinicella genus. The large majority
(7, or 6.3% of the total) of sequences within the Deltaproteo-
bacteria phyla affiliated with the order Myxococcales, which are
gliding bacteria commonly found in soils and activated sludge
that are thought to significantly impact biomass carbon
turnover due to their role as ‘micropredators’ (Lueders et al.,
2006). The aeration basins are operated for simultaneous
BOD removal and nitrification and, as expected, we recovered
sequence-based evidence for the latter process in this envi-
ronment. Six percent of sequences clustered within the
Nitrospira genus and can thus be considered putative nitrite-
oxidizing bacteria (NOB). Nitrospira have been found to be
the most abundant NOB in most nitrifying wastewater treat-
ment plants (Daims et al., 2006). No sequences affiliated with
the Nitrobacter genus, previously viewed as the dominant NOB
in activated sludge, were recovered in our clone libraries.
Moreover, strong bootstrap support indicated that two
sequences (1.8% of total sequences) in the betaproteobacterial
grouping clustered within the Nitrosomonas genus and are thus
putatively ammonia-oxidizing bacteria (AOB). Seven percent
of sequences affiliated with the Bacteroidetes, of which five
sequences (4.5%) clustered in the family Saprospiraceae.
Members of the Bacteroidetes phylum are thought to degrade
complex organic materials, and recent evidence suggests that
many Saprospiraceae are epiphytic protein-hydrolyzers
commonly observed attached to filamentous bacteria (Xia
et al., 2008). The Actinobacteria and deeply branching Plancto-
mycetes accounted for 6% and 5% of sequences, respectively,
and representatives from both phyla have been reported in
high numbers in activated sludge (Seviour and Nielsen, 2010).
Filamentous Chloroflexi, particularly those associated with the
class Anaerolineae as the majority of 9 (8.1%) Chloroflexi clones
did in this study, are also relatively cosmopolitan in waste-
water treatment processes and have been implicated in
sludge bulking events (Nielsen et al., 2009). The phyla Acid-
obacteria, Verrucomicrobia, and TM7 were each represented by
a single sequence (0.9% of the total) in our clone libraries.
3.3. Tracking bacterial community dynamics
Based on in-silico analyses of a baseline 16S rRNA gene clone
library from 29 September 2004 with 16 restriction enzymes,
we identified MspI and RsaI as optimal for resolving bacterial
community diversity and dynamics in PARWQCP activated
sludge via T-RFLP analyses. Previous analyses in the primary
literature (Engebretson and Moyer, 2003; Liu et al., 1997) and in-
silico digests of clone libraries from three separate sampling
dates during our year-long time series (9 June 2005, 5 August
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8 5481

2005, and 1 December 2005) corroborated our choice of
restriction enzymes. These two enzymes were employed in
parallel digests to provide an increased level of resolution. 318
total OTUs were observed, of which 192 associated with MspI
digestions and the remaining 126 with RsaI analyses (Fig. S2
and S3). Within the concatenated MspI and RsaI T-RFLP data-
sets, we observed an average of 102 OTUs (standard
deviation ¼ 8) in each time point. However, only 12 of these
were represented in all 53 time points examined, and less than
10% of OTUs were above the detection limit (>0.5% relative
abundance) in 50 or more weeks. Within any given time point,
the most abundant OTU (T-RF with highest peak height)
averaged 11.5% of the T-RFLP profile (sum of T-RF peak
heights). Interestingly, the 12 OTUs represented at all time
points were also among the most abundant OTUs we
measured; their average relative abundances across the time
series ranged from 1.6% to 6.2% of the total fluorescence signal.
Our results demonstrate strong variation in the relative
abundance of a highly diverse assemblage of ribotypes
throughout the 53 week sampling period.
To further explore temporal variability within the overall
bacterial community, T-RF patterns over the sampling period
were assessed via the indirect gradient ordination technique
Nonmetric Multidimensional Scaling (NMDS). Community
structures in 53 weekly samples from the PARWQCP aeration
basin are represented as colored points in the NMDS ordina-
tion (Fig. 2). The proximity of two points is an estimate of the
pairwise BrayeCurtis similarity index between the two
samples. The coefficient of determination (R2) between
distances in the original unreduced BrayeCurtis distance
matrix and Euclidean distances in ordination space was 0.76,
indicating that distances in ordination space are a reasonably
accurate representation of those in unreduced space. A stress
value of 18.4 provides additional verification that the reduced
NMDS ordinates preserve patterns in activated sludge bacterial
community dynamics. In general, samples collected in similar
time periods (e.g. the same month) cluster closely together in
the ordination. This ordination pattern suggests a gradual
succession within the overall bacterial community over time.
We constructed a Mantel correlogram to further charac-
terize this successional pattern and assess potential period-
icities in our time series (Fig. 3). The correlogram displays
average mean similarity (BrayeCurtis distance) between
bacterial community structure as a function of the lag time
between sampling events. A corresponding Mantel scalogram,
which color codes similarities between all pairs of points
along the time series, is available in the Supplementary
Materials (Fig. S4). A strong negative trend in community
similarity as a function of increasing lag time is evident, while
periodicity was not detected. This negative trend was
confirmed via correlation analysis (Spearman’s rank correla-
tion coefficient ¼ 0.998, p < 0.001). In agreement with the
NMDS analysis, this result provides evidence that, rather than
undergoing variation (periodicity) around a mean “equilib-
rium” community structure, activated sludge bacterial
community dynamics follow a gradual succession away from
initial conditions.
To test the hypothesis that within-month T-RFLP profile
similarities were greater than among-month similarities, an
analysis of similarity (ANOSIM) was conducted. Results are
summarized in Table 2. Global ANOSIM revealed strong and
significant variation in T-RFLP profiles as a function of month
of sampling ( p < 0.001). Moreover, 10 of 11 pairwise ANOSIM
tests demonstrated significantly higher within-group than

Fig. 2 e Nonmetric Multidimensional Scaling (NMDS) analysis calculated from normalized peak heights for T-RFLP
electropherograms for concatenated MspI and RsaI T-RFs. Circles and associated numbers indicate sampling points
analyzed in temporal order (1e4: February 2005, 5e9: March 2005, 10e13: April 2005, 14e17: May 2005, 18e22: June 2005,
23e26: July 2005, 27e30: August 2005, 31e35: September 2005, 36e39: October 2005, 40e43: November 2005, 44e48:
December 2005, 49e52: January 2006, 53: February 2006). Circles are color-coded according to month of sampling. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
5482 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8

one-year time series between relatively stable COD or NHþ 4

removal efficiencies and bacterial community dynamics. 64 of
66 ANOSIM tests for all pairwise between-month groupings
(not limited to consecutive months) indicated significant
separation in bacterial community structure at the p ¼ 0.1
level (data not shown). Our analysis reveals successive shifts
in bacterial community structure throughout the year of
sampling and demonstrates that bacterial community struc-
ture is dynamic during periods of relatively stable operation.

3.4. Correlations between operational parameters and

bacterial community structure

We interpreted NMDS ordination axes in terms of explanatory

variables by calculating Pearson productemoment correlation
coefficients between sample axis scores and 20 measured
operational or environmental variables (Table 3). DO concen-
trations displayed the strongest correlation to both axis 1 and
2. This correlation was positive and highly significant
Fig. 3 e Mantel correlogram displaying mean pairwise ( p  0.001) for both axes. Axis 2 was also significantly nega-
similarity (BrayeCurtis distance metric) between bacterial tively associated with influent silver concentrations, and
community structures (via T-RFLP profiles) as a function of significantly positively associated with influent arsenic and
lag time between activated sludge sampling events. reactor temperature.
To directly assess the influence of monitored operational
parameters on temporal dynamics in bacterial community
among-group similarity in T-RFLP profiles at the p ¼ 0.1 level structure, we employed the direct gradient ordination method
for consecutive months of sampling, although the strength of Redundancy Analysis (RDA). Axes in RDA ordinations are
this association was somewhat weaker than measured global constrained against input environmental variables. The final
variation. T-RFLP profiles from November and October 2005 RDA model is shown in Fig. 4, and results are summarized in
generated the sole pairwise assessment for sequential Tables S1 and S2. The model explained the majority of vari-
months of sampling that did not indicate significant separa- ance in the species-environment correlations (60.7%), but only
tion in community structure (R ¼ 0.157, p ¼ 0.267). Interest- encompassed a minority of the variance in species (T-RF) data
ingly, average COD removal efficiencies during these two (17.5%). Both axes displayed high specieseenvironment
months (65.5%  2.0% and 65.3%  7.6%, respectively) were correlation values, indicating strong correlations between
nearly identical and represented the two highest monthly T-RF patterns and operational data. Of 20 input explanatory
average removal efficiencies over this one-year time series. It variables, 7 were identified as significantly linked to bacterial
is tempting to speculate that overlap in bacterial community community variability at the p ¼ 0.05 level. These seven
structure between October and November is associated with
similarities in this performance metric, despite the lack of
a significant correlation via RDA or NMDS analyses over our
Table 3 e Significant Pearson productemoment
correlation coefficients between NMDS ordination axes
and 20 operational or environmental variables.
Table 2 e ANOSIM test for significant differences between Environmental or Operational Parameters
monthly groupings in activated sludge overall bacterial
Axis 1 Dissolved Oxygen (r ¼ 0.457, P ¼ 0.001)
community structure, as measured via concatenated
(R2 ¼ 0.413)
MspI and RsaI T-RFLP patterns.
Axis 2 Dissolved Oxygen (r ¼ 0.625, P ¼ 0.000001)
Scale or Comparison Sample P-Value (R2 ¼ 0.345) Influent Silver (r ¼ L0.440, P ¼ 0.001)
Statistic R Influent Arsenic (r ¼ 0.377, P ¼ 0.005)
Temperature (r ¼ 0.370, P ¼ 0.006)
Global, with months nested 0.742 <0.0001
February vs. March 0.256 0.045 Only parameters significantly correlated at an adjusted P-value of
March vs. April 0.205 0.077 0.05 after application of the Benjamini and Hochberg correction for
April vs. May 0.427 0.055 multiple comparisons are shown. Correlation coefficients and
May vs. June 0.375 0.025 associated P-values are indicated in parentheses. Parameters in
June vs. July 0.563 0.022 bold were significant at an adjusted P-value of 0.01 after application
July vs. August 0.521 0.058 of the Benjamini and Hochberg correction for multiple compari-
August vs. September 0.788 0.018 sons. R2 values beneath axis labels indicate the coefficient of
September vs. October 0.488 0.007 determination for correlations between ordination distances and
October vs. November 0.156 0.267 distances in the original unreduced space. The sum of coefficients
November vs. December 0.456 0.016 of determination for both axes (0.758) is the overall (cumulative)
December vs. January (2006) 0.408 0.008 correlation between ordination and unreduced distances.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8 5483

parameters are indicated in Fig. 4 by arrows. Arrow length Observed

Power Fit
indicates the strength of the relationship between that

300
parameter and bacterial community composition, and arrow
direction indicates the parameter’s approximate direction of

Cumulative Observed OTUs

increase. The strongest correlate was DO ( p < 0.005), in line

250
with Pearson correlation coefficients to NMDS axes. Also in
agreement with the NMDS analysis, reactor temperature and
primary influent silver concentration were identified as strong
correlates to variation in T-RFLP profiles. Biomass (MLSS)

200
levels, primary influent cadmium, flow rate, and bioreactor
influent nitrite concentrations rounded out the explanatory
variables included in the final RDA model, in order of addition

150
to the model.

3.5. A taxaetime relationship

100
To assess the applicability of a taxaetime relationship (TTR) to
temporal dynamics in the overall bacterial assemblage in 0 100 200 300

a full-scale activated sludge WWTP, we used nonlinear Time (days)

regression analysis to describe the cumulative observed OTUs Fig. 5 e Taxaetime relationship and least-squares
(S ) as a function of time of observation (T ) via a power law nonlinear regression power law fit between time and
model, S ¼ cTw. Here, the scaling exponent w can be consid- cumulative observed OTUs in the activated sludge
ered a measure of temporal turnover of microbial taxa, in the bioreactor.
same way that the exponent x of an ecological speciesearea
relationship is a measure of spatial turnover. We found that
the best-fit power law model displayed a strong (R2 ¼ 0.989)
and significant ( p < 0.001) fit to our molecular characteriza-
as measured by the proportion of variation in the data
tion of bacterial community dynamics (Fig. 5). The best-fit TTR
explained by each model (i.e. the R2 value).
exponent was w ¼ 0.209  0.007 (95% confidence interval) for
the concatenated MspI and RsaI T-RFLP profiles. Power law
models for MspI and RsaI datasets analyzed independently
yielded similar scaling exponents (w ¼ 0.213 and 0.203, 4. Discussion
respectively). The power-law fit to our data was superior to
eight other models (linear, logarithmic, inverse, quadratic, 16S rRNA gene clone libraries from four time points revealed
cubic, compound, S, growth, exponential, and logistic) tested, a highly diverse community, with sequences clustering within
0.6

Temperature 25 February 2005

24
26 March 2005
27 20 17
30 April 2005
34 29 28 May 2005
19
32 41 21 June 2005
23 22 15 14
31 18
July 2005
33 40 16
August 2005
Axis 2

35 37
13
38 IInfluent
fl September 2005
39 42
7 Ag
36 October 2005
6 12
43 44 Influent November 2005
NO 10 1
DO Influent 53 December 2005
45 46 Cd
5 4 11 January 2006
47 3 MLSS Flowrate
52 February 2006
50 2 9
8
49
48 51
-0.6

-0.8 0.8
Axis 1

Fig. 4 e RDA ordination calculated from normalized peak heights for T-RFLP electropherograms of concatenated MspI and
RsaI T-RFs. Variables that contributed significant improvement ( p < 0.05) to the explanatory power of the RDA models are
indicated by arrows in the ordination. Circles and associated numbers [ samples analyzed in temporal order; Ag [ silver;
Cd[Cadmium; MLSS [ Mixed Liquor Suspended Solids. Circles are color-coded according to month of sampling. (For
interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
5484 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
9 separate phyla. Xia et al. (2010) employed a 16S rRNA
microarray (PhyloChip) to characterize diversity in single time
points from five lab or full-scale bioreactors. In line with our
results, they found a high representation of Proteobacteria,
Actinobacteria, and Bacteroidetes, but also high abundance of
Firmicutes. In contrast, we observed no 16S rRNA gene
sequences showing high similarity to Firmicutes. This
discrepancy may be due to the fact that some of the reactors
sampled by Xia and colleagues contained anaerobic or anoxic
zones, thereby yielding niches for anaerobic Firmicutes such as
the Clostridia, whereas the bioreactor targeted in this study is
highly aerated. Nonetheless, the observed bacterial commu-
nity structure in this study largely mirrored previous obser-
vations in activated sludge based on 16S rRNA gene clone
library analyses (Seviour and Nielsen, 2010; Wagner et al.,
2002), suggesting that results of our investigation into
temporal dynamics and associated drivers of bacterial
community structure may be widely representative of
municipal activated sludge bioreactors. Further investigation
is warranted to test this hypothesis, particularly in additional
bioreactor process configurations, under varying environ-
mental conditions, and in different geographic locations.
As with many ecologically relevant methodologies, inter-
pretation of microbial community fingerprinting methods
requires caution and a healthy awareness of methodological
limitations. T-RFLP is subject to well-known biases associated
with PCR-based methods. However, all samples are subject to
the same biases, and it is thus valid to compare between
samples on a relative basis after standardization of profiles.
Indeed, T-RFLP has been shown to be a highly reproducible
and sensitive means of assessing fine-scale changes in
microbial community structure at a variety of temporal and
spatial scales (Schütte et al., 2008). While it is also possible in
some cases to track specific phylogenetic lineages in microbial
communities by linking peaks in T-RFLP profiles to in-silico
analyses of clone libraries, phylogenetic inference of T-RFs is
uncertain in activated sludge due to the highly complex
nature of this system. This is an acknowledged limitation of
T-RFLP (Marsh and Jared, 2005; Osborne et al., 2006) and other
microbial community fingerprinting techniques, and, in this
study, precludes identification of the exact phylogenetic scale
(e.g. family, genus, species) at which observed shifts in
community structure occur. However, our intention was to
characterize overall bacterial community dynamics in acti-
vated sludge at a full-scale WWTP and the relative influence
therein of a multitude of deterministic environmental drivers.
Employment of T-RFLP profiles over time for whole-
community ecological comparisons was fully sufficient for
this purpose.
Molecular signatures of the activated sludge bacterial
community revealed surprisingly strong temporal dynamics
during a period of relatively stable reactor performance. We
previously characterized ammonia-oxidizing bacterial pop-
ulation dynamics in the same full-scale bioreactor (Wells et al.,
2009), and found strong oscillations in several lineages during
a period of stable nitrification. Taken together, our combined
results support the argument that dynamic behavior is prob-
ably an innate property of biological systems, including engi-
neered bioreactors (Briones and Raskin, 2003). Steady-state
equilibrium-based concepts of stability most likely apply only
to aggregate community or ecosystem properties, and true
steady-state conditions may not exist in individual microbial
populations (Briones and Raskin, 2003; Huisman and Weissing,
2002). In fact, Fernandez et al. (1999) suggested that less stable
microbial community structure in lab-scale bioreactors may
actually be correlated with more stable aggregate ecosystem
function. Similar to our results, Nadarajah et al. (2007) and
Kaewpipat and Grady (2002) noted strong temporal bacterial
community dynamics in lab-scale reactors during periods of
stable operation. Among the few studies available assessing
long-term community dynamics in full-scale systems, Lapara
et al. (2002) noted moderate dynamics in microbial commu-
nity structure based on DGGE profiles in industrial wastewater
treatment bioreactors, particularly in response to changing
influent conditions, and Werker (2006) recorded significant
microbial community dynamics via microbial fatty acid anal-
yses during stable operation in a full-scale municipal WWTP
during stable operating conditions. In contrast, Smith et al.
(2003) reported relatively stable community structure based
on ribosomal intergenic spacer length polymorphism analyses
over a 55-day period in an industrial WWTP during a period of
stable operation. Interestingly, other recent studies have noted
strong associations between microbial community dynamics
and process performance. Gentile et al. (2007) reported that
nitrite appearance and nitrous oxide emissions were linked to
population dynamics in lab-scale denitrifying bioreactors, and
Zhang et al. (2005) characterized strong changes in microbial
community structure during a complete loss of phosphorus
removal functionality in a lab-scale EBPR reactor. The results
we present here in concert with these previous studies suggest
that activated sludge microbial communities exhibit a baseline
of moderate population dynamics not necessarily reflected in
system performance, with periodic dramatic shifts in
community structure that can impact functional stability.
Concurrent to weekly activated sludge sampling, we
monitored 20 operational and environmental variables that we
hypothesized could be deterministic drivers of overall bacte-
rial community dynamics. We identified these variables based
on previous reports in the primary literature, expected vari-
ability over the year-long sampling period, and importance to
process performance goals. DO emerged in both RDA (Fig. 4)
and NMDS (Table 3) analyses as the most important explana-
tory variable associated with bacterial community dynamics.
This finding dovetails with our previous assessment of the
influence of operational parameters on AOB community
structure in the PARWQCP (Wells et al., 2009). In activated
sludge systems, large variations in DO have been implicated in
the proliferation of bacteria involved in filamentous bulking
(Gaval and Pernelle, 2003). Not surprisingly, increases in
bacterial populations capable of anoxic or anaerobic metabo-
lisms, such as denitrifying Comamonadaceae, have also been
observed in activated sludge at low DO levels (<1 mg/L) (Sadaie
et al., 2007). High bulk DO levels (>3 mg/L) were maintained at
all times in the PARWQCP bioreactor during our year-long time
series. It is thus striking that variations in this parameter were
closely associated with bacterial community dynamics. It is
possible that the strong correlation between bulk DO and
bacterial community dynamics is linked to the presence of
anoxic microniches in activated sludge flocs. Schramm et al.
(1999) used O2 microsensors to verify the presence of anoxic
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
zones in large flocs from two separate activated sludge biore-
actors at 2 mg/L bulk DO concentration, although flocs from an
additional four bioreactors showed no evidence of anoxia. Li
and Bishop (2004) provided additional evidence for anoxic
zones within activated sludge flocs, albeit only at
DO  1.2 mg/L. In addition, biomass decay rates have previ-
ously been suggested to be a continuous increasing function of
DO concentration (Yuan and Blackall, 2002), and it is possible
that community dynamics associated with fluctuations in high
levels of DO are in fact linked to variations in these decay rates
and associated predation or grazing processes.
Reactor temperature was also strongly and significantly
associated with bacterial community dynamics via both RDA
and NMDS analyses. The impact of temperature on bioreactor
community structure has been noted previously in lab-scale
studies. Nadarajah et al. (2007) detailed strong divergence in
bacterial community structure based on DGGE profiles
following a temperature increase from 30  C to 45  C in lab-
scale reactors treating an industrial waste. Using culture-
based techniques, Alawi et al. (2009) demonstrated a temper-
ature-dependent population shift in NOB in activated sludge
incubated at 10  C, 17  C, and 28  C. Moreover, in a full-scale
municipal WWTP, Werker (2006) suggested that microbial
community dynamics were linked to seasonal temperature
variations between 12  C and 28  C. Interestingly, temperature
variation in the PARWQCP over our time series was moderate
(18.2e25.4  C) and lower than in studies discussed above. The
observed correlation to temperature may be a reflection of
seasonal periodicity in bacterial community structure. Further
investigation based on multi-year time series of activated
sludge samples is warranted to test this hypothesis.
Influent silver was the only metal of 10 assessed that was
significantly associated with bacterial community dynamics in
both RDA and NMDS analyses. Silver ions are a known highly
effective bacteriocide and have been observed to cause inhibi-
tion in activated sludge systems treating photoprocessing
wastewater, albeit at levels far exceeding those observed at the
PARWQCP (Pavlostathis and Maeng, 1998). We previously
documented a moderate association between AOB population
dynamics in activated sludge and low influent silver concen-
trations, similarly with no apparent impact on process perfor-
mance (Wells et al., 2009). While heavy metals (including
copper, zinc, and nickel) have previously been shown to strongly
impact microbial community structure in soils upon application
of sewage sludge (Baath et al., 1998), we are not aware of
previous reports implicating low levels of silver ions in activated
sludge microbial community dynamics. It is tempting to spec-
ulate that variations in both the overall bacterial community
and the AOB subpopulation are examples of community-level
adaptation in response to perturbations in influent character-
istics that facilitate maintenance of high effluent quality. This
hypothesis requires substantial follow-up testing.
Surprisingly, overall bacterial community dynamics were
also significantly linked in our RDA to influent nitrite levels to
the bioreactor, despite low levels (0.05e1.18 mg N/L) and low
and stable effluent nitrite. This observation dovetails with
previously documented associations in this bioreactor of
influent nitrite to AOB community dynamics (Wells et al., 2009)
and to variations in broad-scale taxonomic and functional gene
diversity based on high-density oligonucleotide microarray
5485
analyses (Wells, 2011). Under conditions observed in this study,
the concentration of free nitrous acid (FNA, the conjugate acid
of nitrite) is expected to be two orders of magnitude lower than
levels implicated in inhibition of AOB or phosphorus-
accumulating organisms (Anthonisen et al., 1976; Zhou et al.,
2007). It is thus unlikely that variable levels of FNA due to
fluctuations in nitrite significantly influenced the observed
population dynamics. Nitrite is generally present in municipal
wastewater at exceptionally low levels (Tchobanoglous et al.,
2002), and the observed fluctuations in this parameter are
unexpected. These fluctuations suggest that influent nitrite to
the aeration basins may be due to small levels of partial nitri-
fication in the upstream BOD-removal trickling filter. Indeed,
elsewhere we detail converging lines of evidence suggesting
that bioreactor influent nitrite may be a surrogate measure of
microbial immigration from the trickling filter to the down-
stream activated sludge bioreactor (Wells, 2011).
While bacterial community dynamics were also linked via
RDA to biomass levels and influent flow rate, both of these
parameters were significantly correlated to reactor tempera-
ture (r ¼ 0.65, p < 0.001 and r ¼ 0.31, p ¼ 0.02, respectively).
Their association to community dynamics may be an artifact
of this correlation. Nonetheless, it is feasible that both links
represent biologically relevant processes. SRT, which is
proportional to biomass concentration at constant HRT, has
previously been shown to correlate with shifts in microbial
community structure in lab-scale reactors (Saikaly et al.,
2005). Moreover, flow rate fluctuations may be associated
with varying levels of immigration of autochthonous micro-
organisms to activated sludge bioreactors, either from the
influent or from upstream reactors. We provide evidence
elsewhere that such immigration may be a significant driver
of activated sludge community dynamics (Wells, 2011).
While significant, our RDA explained a relatively low
percentage (17.5%) of the variance in T-RFLP profiles. This
result is not uncommon in ecological studies (see, for example
(Yannarell and Triplett, 2005).) Nonetheless, it is likely that
overall bacterial community dynamics in activated sludge are
mediated by additional factors beyond the parameters we
focused on, including stochastic influences, predation,
unmonitored inhibitory chemicals, and fluctuations in levels
of specific organic substrates (such as micropollutants) in the
plant influent. Indeed, Ofiteru et al. (2010) recently employed
a portion of the dataset presented here to show that both
neutral (random immigration and births/deaths) and niche-
based (deterministic) drivers were associated with pop-
ulation turnover in activated sludge. It also appears likely that
microbial community structure is shaped in part by phage
predation (Wu and Liu, 2009), protozoan grazing (Petropoulos
and Gilbride, 2005), and chaotic behavior (Graham et al., 2007).
Nonetheless, our results provide strong evidence that specific
deterministic environmental selection processes are signifi-
cantly linked to dynamics within the overall bacterial
community in full-scale activated sludge bioreactors.
We demonstrated that observed bacterial taxa richness in
activated sludge from a full-scale WWTP increased with the
length of time censused according to a power law model, as
predicted by macroecological theory, with a best-fit scaling
exponent of w ¼ 0.209. White et al. (2006) gathered 984 mac-
roecological community time series to evaluate the generality
5486 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
of the SpecieseTime Relationship (STR) across ecosystems and
taxonomic groups, and found that power law exponents typi-
cally ranged between 0.2 and 0.4. They further postulated that
this surprisingly narrow range of exponents suggests that the
STR is a fundamental ecological pattern, analogous to the well-
known SAR. Our observed scaling exponent falls in the lower
end of the range recorded by White and colleagues, suggesting
that species turnover in microbial and macrobial systems
follow similar ecological patterns. Our observed scaling expo-
nent is also well in line with the few previous studies that have
examined taxaetime relationships in microbial systems.
Redford and Fierer (2009) detailed a power law exponent of
w ¼ 0.312 for bacterial communities on leaf surfaces during one
growing season, based on bacterial 16S rRNA gene clone
libraries. van der Gast et al. (2008) employed lab-scale reactors
to test the impact of varying percentages of industrial and
municipal wastewater on turnover in microbial community
composition based on DGGE analyses, and observed
a pronounced decrease in power law exponent (from 0.512 to
0.162) as the percentage of industrial wastewater increased.
Interestingly, while the power law exponent we observed in
a full-scale bioreactor (w6900 m3) fell within this range, it was
well below the reported value for a lab-scale (5 L) reactor with
100% municipal wastewater as influent (w ¼ 0.512). It is
tempting to speculate that this discrepancy is associated with
the more than six order of magnitude variation in volume
between these reactors. Indeed, the exponent of the STR is
known in macroecology to be negatively correlated with the
spatial scale of observation (White et al., 2006), and van der
Gast (2008) suggests that this may be the case in microbial
systems as well. Moreover, Adler et al. (2005) suggested that the
SAR and STR are components of a unified speciesetimeearea
relationship. Additional research is warranted to establish
a firm understanding of how taxa turnover in space and time
are related in microbial systems. This question is of funda-
mental importance to the wider field of microbial ecology, and
has important implications for the management of complex
microbial communities in engineered systems to promote
adaption and resilience in the face of process perturbations
and variations in environmental parameters.
5. Conclusions
 To our knowledge, the year-long weekly time series described
in this report is the most detailed record of bacterial
community dynamics in full-scale bioreactors to date and the
first assessment of the applicability of a power-law taxaetime
relationship to a full-scale wastewater treatment plant.
 Our results reveal successive shifts in bacterial community
structure, with temporal dynamics significantly correlated
to temperature, dissolved oxygen, influent silver, and
influent nitrite concentrations. These correlations suggest
that community dynamics are likely regulateddat least in
partdby specific deterministic operational and environ-
mental parameters.
 Activated sludge bacterial community dynamics followed
a gradual succession away from initial conditions, with no
apparent impact on process performance.
 Accumulation of observed taxa over time followed a pow-
erelaw relationship, in line with predictions from macro-
ecological theory, with a power-law exponent of w ¼ 0.209.
Multi-year time series are needed to determine whether this
relationship extends over longer time periods.
 Taken together, our results provide a baseline assessment of
bacterial community temporal dynamics in a fully func-
tional activated sludge bioreactor and point to the utility of
integrating ecological theory with engineered systems. A
future challenge is to elucidate the relationship between
these temporal dynamics (i.e. the taxaetime powerelaw
exponent) and system stability and resilience, and to
employ this information to rationally design and manage
complex microbial communities.
Acknowledgments
We thank the PARWQCP staff for weekly sampling. This study
was funded by the Stanford Woods Institute for the Environ-
ment, NSF SGER Grant CBET-0630092 (to C.S.C. and C.A.F.), and
the PARWQCP. G.F.W. was supported by EPA STAR and NSF
Graduate Research Fellowships.
Appendix. Supplementary material
The supplementary data associated with this article can be
found in the online version at doi:10.1016/j.watres.2011.08.006.
references
Adler, P.B., White, E.P., Lauenroth, W.K., Kaufman, D.M.,
Rassweiler, A., et al., 2005. Evidence for a general
speciesetimeearea relationship. Ecology 86, 2032e2039.
Alawi, M., Off, S., Kaya, M., Spieck, E., 2009. Temperature
influences the population structure of nitrite-oxidizing
bacteria in activated sludge. Environmental Microbiology
Reports 1, 184e190.
Anthonisen, A.C., Loehr, R.C., Prakasam, T.B.S., Srinath, E.G.,
1976. Inhibition of nitrification by ammonia and nitrous acid.
Journal Water Pollution Control Federation 48, 835e852.
Baath, E., Diaz-Ravina, M., Frostegard, A., Campbell, C.D., 1998.
Effect of metal-rich sludge amendments on the soil microbial
community. Applied and Environmental Microbiology 64,
238e245.
Briones, A., Raskin, L., 2003. Diversity and dynamics of microbial
communities in engineered environments and their
implications for process stability. Current Opinion in
Biotechnology 14, 270e276.
Curtis, T., Sloan, W., 2006. Towards the design of diversity:
stochastic models for community assembly in wastewater
treatment plants. Water Science & Technology 54, 227e236.
Curtis, T.P., Head, I.M., Graham, D.W., 2003. Theoretical ecology
for engineering biology. Environmental Science & Technology
37, 64A.
Daims, H., Taylor, M.W., Wagner, M., 2006. Wastewater
treatment: a model system for microbial ecology. Trends in
Biotechnology 24, 483e489.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L.,
et al., 2006. Greengenes, a chimera-checked 16S rRNA gene
database and workbench compatible with ARB. Applied and
Environmental Microbiology 72, 5069e5072.
Engebretson, J.J., Moyer, C.L., 2003. Fidelity of select restriction
endonucleases in determining microbial diversity by
terminal-restriction fragment length polymorphism. Applied
and Environmental Microbiology 69, 4823e4829.
Fernandez, A., Huang, S.Y., Seston, S., Xing, J., Hickey, R., et al., 1999.
How stable is stable? Function versus community composition.
Applied and Environmental Microbiology 65, 3697e3704.
Gaval, G., Pernelle, J.-J., 2003. Impact of the repetition of oxygen
deficiencies on the filamentous bacteria proliferation in
activated sludge. Water Research 37, 1991e2000.
Gentile, M.E., Lynn Nyman, J., Criddle, C.S., 2007. Correlation of
patterns of denitrification instability in replicated bioreactor
communities with shifts in the relative abundance and the
denitrification patterns of specific populations. ISME Journal 1,
714e728.
Graham, D.W., Knapp, C.W., Van Vleck, E.S., Bloor, K., Lane, T.B.,
et al., 2007. Experimental demonstration of chaotic instability
in biological nitrification. ISME Journal 1, 385e393.
Graham, D.W., Smith, V.H., 2004. Designed ecosystem services:
application of ecological principles in wastewater treatment
engineering. Frontiers in Ecology and the Environment 2,
199e206.
Horner-Devine, M.C., Lage, M., Hughes, J.B., Bohannan, B.J.M.,
2004. A taxaearea relationship for bacteria. Nature 432,
750e753.
Huber, T., Faulkner, G., Hugenholtz, P., 2004. Bellerophon:
a program to detect chimeric sequences in multiple sequence
alignments. Bioinformatics 20, 2317e2319.
Huisman, J., Weissing, F.J., 2002. Oscillations and chaos generated
by competition for interactively essential resources. Ecological
Research 17, 175e181.
Kaewpipat, K., Grady, C.P.L., 2002. Microbial population dynamics
in laboratory-scale activated sludge reactors. Water Science
and Technology 46, 19e27.
LaPara, T.M., Nakatsu, C.H., Pantea, L.M., Alleman, J.E., 2002.
Stability of the bacterial communities supported by a seven-
stage biological process treating pharmaceutical wastewater
as revealed by PCR-DGGE. Water Research 36, 638e646.
Li, B., Bishop, P.L., 2004. Micro-profiles of activated sludge floc
determined using microelectrodes. Water Research 38,
1248e1258.
Liu, W.T., Marsh, T.L., Cheng, H., Forney, L.J., 1997.
Characterization of microbial diversity by determining
terminal restriction fragment length polymorphisms of genes
encoding 16S rRNA. Applied and Environmental Microbiology
63, 4516e4522.
Ludwig, W., Strunk, O., Westram, R., Richter, L., Meier, H., et al.,
2004. ARB: a software environment for sequence data. Nucleic
Acids Research 32, 1363e1371.
Lueders, T., Kindler, R., Miltner, A., Friedrich, M.W., Kaestner, M.,
2006. Identification of bacterial micropredators distinctively
active in a soil microbial food web. Applied and
Environmental Microbiology 72, 5342e5348.
Marsh, T.L., Jared, R.L., 2005. Culture-independent Microbial
Community Analysis with Terminal Restriction Fragment
Length Polymorphism, Methods in Enzymology. Academic
Press, pp. 308e329.
Nadarajah, N., Grant Allen, D., Fulthorpe, R.R., 2007. Effects of
transient temperature conditions on the divergence of
activated sludge bacterial community structure and function.
Water Research 41, 2563e2571.
Nielsen, P.H., Kragelund, C., Seviour, R.J., Nielsen, J.L., 2009.
Identity and ecophysiology of filamentous bacteria in
activated sludge. FEMS Microbiology Reviews 33, 969e998.
5487
Ofiteru, I.D., Lunn, M., Curtis, T.P., Wells, G.F., Criddle, C.S., et al.,
2010. Combined niche and neutral effects in a microbial
wastewater treatment community. Proceedings of the
National Academy of Sciences USA 107, 15345e15350.
Osborne, C.A., Rees, G.N., Bernstein, Y., Janssen, P.H., 2006. New
threshold and confidence estimates for terminal restriction
fragment length polymorphism analysis of complex bacterial
communities. Applied and Environmental Microbiology 72,
1270e1278.
Pavlostathis, S.G., Maeng, S.K., 1998. Aerobic biodegradation of
a silver-bearing photoprocessing wastewater. Environmental
Toxicology and Chemistry 17, 617e624.
Petropoulos, P., Gilbride, K.A., 2005. Nitrification in activated
sludge batch reactors is linked to protozoan grazing of the
bacterial population. Canadian Journal of Microbiology 51,
791e799.
Preston, F.W., 1960. Time and space and the variation of species.
Ecology 41, 612e627.
Redford, A., Fierer, N., 2009. Bacterial succession on the leaf
surface: a novel system for studying successional dynamics.
Microbial Ecology 58, 189e198.
Rittmann, B.E., Hausner, M., Loffler, F., Love, N.G., Muyzer, G.,
et al., 2006. A vista for microbial ecology and environmental
biotechnology. Environmental Science & Technology 40,
1096e1103.
Sadaie, T., Sadaie, A., Takada, M., Hamano, K., Ohnishi, J., et al.,
2007. Reducing sludge production and the domination of
comamonadaceae by reducing the oxygen supply in the
wastewater treatment procedure of a food-processing factory.
Bioscience, Biotechnology, and Biochemistry 71, 791e799.
Saikaly, P.E., Stroot, P.G., Oerther, D.B., 2005. Use of 16S rRNA gene
terminal restriction fragment analysis to assess the impact of
solids retention time on the bacterial diversity of activated
sludge. Applied and Environmental Microbiology 71,
5814e5822.
Schramm, A., Santegoeds, C.M., Nielsen, H.K., Ploug, H.,
Wagner, M., et al., 1999. On the occurrence of anoxic
microniches, denitrification, and sulfate reduction in aerated
activated sludge. Applied and Environmental Microbiology 65,
4189e4196.
Schütte, U., Abdo, Z., Bent, S., Shyu, C., Williams, C., et al., 2008.
Advances in the use of terminal restriction fragment length
polymorphism (T-RFLP) analysis of 16S rRNA genes to
characterize microbial communities. Applied Microbiology
and Biotechnology 80, 365e380.
Seviour, R.J., Nielsen, P.H., 2010. Microbial communities in
activated sludge plants. In: Seviour, R.J., Nielsen, P.H. (Eds.),
Microbial Ecology of Activated Sludge. IWA Publishing,
London, UK.
Slater, F.R., Johnson, C.R., Blackall, L.L., Beiko, R.G., Bond, P.L.,
2010. Monitoring associations between clade-level variation,
overall community structure and ecosystem function in
enhanced biological phosphorus removal (EBPR) systems
using terminal-restriction fragment length polymorphism
(T-RFLP). Water Research 44, 4908e4923.
Smith, N.R., Yu, Z., Mohn, W.W., 2003. Stability of the bacterial
community in a pulp mill effluent treatment system during
normal operation and a system shutdown. Water Research 37,
4873e4884.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2002.
Wastewater Engineering: Treatment and Reuse. McGraw-
Hill, New York.
van der Gast, C.J., 2008. Chapter 6-Islands Shaping Thought in
Microbial Ecology. In: Allen, I., Laskin, S.S., Geoffrey, M.G. (Eds.),
Advances in Applied Microbiology. Academic Press, pp.
167e182.
van der Gast, C.J., Ager, D., Lilley, A.K., 2008. Temporal scaling of
bacterial taxa is influenced by both stochastic and
5488 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 7 6 e5 4 8 8
deterministic ecological factors. Environmental Microbiology
10, 1411e1418.
Wagner, M., Loy, A., Nogueira, R., Purkhold, U., Lee, N., et al., 2002.
Microbial community composition and function in wastewater
treatment plants. Antonie Van Leeuwenhoek 81, 665e680.
Wells, G.F., 2011. Reexamining the Engineered Nitrogen Cycle:
Microbial Diversity, Community Dynamics, and Immigration
in Nitrifying Activated Sludge Bioreactors. PhD Thesis. Civil
and Environmental Engineering. Stanford University,
Stanford, CA.
Wells, G.F., Park, H.-D., Yeung, C.-H., Eggleston, B., Francis, C.A.,
et al., 2009. Ammonia-oxidizing communities in a highly
aerated full-scale activated sludge bioreactor:
betaproteobacterial dynamics and low relative abundance of
Crenarchaea. Environmental Microbiology 11, 2310e2328.
Werker, A.G., 2006. An evaluation of full-scale activated sludge
dynamics using microbial fatty acid analysis. Water Science &
Technology 54, 11e19.
White, E.P., Adler, P.B., Lauenroth, W.K., Gill, R.A., Greenberg, D.,
et al., 2006. A comparison of the speciesetime relationship
across ecosystems and taxonomic groups. Oikos 112, 185e195.
Wu, Q., Liu, W.-T., 2009. Determination of virus abundance,
diversity and distribution in a municipal wastewater
treatment plant. Water Research 43, 1101e1109.
Xia, S., Duan, L., Song, Y., Li, J., Piceno, Y.M., et al., 2010. Bacterial
community structure in geographically distributed biological
wastewater treatment reactors. Environmental Science &
Technology 44, 7391e7396.
Xia, Y., Kong, Y., Thomsen, T.R., Halkjaer Nielsen, P., 2008.
Identification and ecophysiological characterization of
epiphytic protein-hydrolyzing Saprospiraceae (“Candidatus
Epiflobacter” spp.) in activated sludge. Applied and
Environmental Microbiology 74, 2229e2238.
Yannarell, A.C., Triplett, E.W., 2005. Geographic and
environmental sources of variation in lake bacterial
community composition. Applied and Environmental
Microbiology 71, 227e239.
Yuan, Z., Blackall, L.L., 2002. Sludge population optimisation:
a new dimension for the control of biological wastewater
treatment systems. Water Research 36, 482e490.
Zhang, T., Liu, Y., Fang, H.H., 2005. Effect of pH change on the
performance and microbial community of enhanced
biological phosphate removal process. Biotechnology and
Bioengineering 92, 173e182.
Zhou, Y., Pijuan, M., Yuan, Z., 2007. Free nitrous acid inhibition on
anoxic phosphorus uptake and denitrification by poly-
phosphate accumulating organisms. Biotechnology and
Bioengineering 98, 903e912.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Morphological architecture of dual-layer hollow fiber for

membrane distillation with higher desalination performance

Peng Wang, May May Teoh, Tai-Shung Chung*

Department of Chemical & Biomolecular Engineering, National University of Singapore, 4 Engineering Drive 4, Singapore 117576, Singapore

article info abstract

Article history:
Received 26 January 2011
Received in revised form
21 June 2011
Accepted 7 August 2011
Available online 23 August 2011
Keywords:
Membrane distillation
Uniform finger-like macrovoids
Fully sponge-like structure
Liquid entry pressure
Heat and mass transfer model
A new strategy to enhance the desalination performance of polyvinylidene fluoride (PVDF)
hollow fiber membrane for membrane distillation (MD) via architecture of morphological
characteristics is explored in this study. It is proposed that a dual-layer hollow fiber con-
sisting of a fully finger-like macrovoid inner-layer and a sponge-like outer-layer may
effectively enhance the permeation flux while maintaining the wetting resistance. Dual-
layer fibers with the proposed morphology have been fabricated by the dry-jet wet spin-
ning process via careful choice of dopes composition and coagulation conditions. In
addition to high energy efficiency (EE ) of 94%, a superior flux of 98.6 L m
2 h
1 is obtained
during the direct contact membrane distillation (DCMD) desalination experiments. More-
over, the liquid entry pressure (LEP) and long-term DCMD performance test show high
wetting resistance and long-term stability. Mathematical modeling has been conducted to
investigate the membrane mass transfer properties in terms of temperature profile and
apparent diffusivity of the membranes. It is concluded that the enhancement in perme-
ation flux arises from the coupling effect of two mechanisms; namely, a higher driving
force and a lower mass transfer resistance, while the later is the major contribution. This
work provides an insight on MD fundamentals and strategy to tailor making ideal
membranes for DCMD application in desalination industry.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction (RO), MD is capable of integrating with various renewable energy

Water, a necessity to daily life, is currently in great deficiency
because of the rapid growth of global population and the
accelerated industrialization and urbanization (Escobar, 2010).
To mitigate the water crisis, the desalination and wastewater
reclamation processes have been studied extensively in past
years (Shannon et al., 2008). Among the developed technologies,
membrane distillation (MD) shows a promising perspective
credited to its large membrane contact area, high salt rejection,
small foot print and mild operation conditions (Song et al., 2008).
Furthermore, as compared with other desalination technologies
like multi-stage flash distillation (MSFD) and reverse osmosis
sources such as solar energy, geothermal energy and waste heat
source (Curcio and Drioli, 2005). However, at present, the
application of MD process is still challenging in the real desali-
nation industry application, mainly due to the low permeation
flux, high risk of membrane wetting and long term operation
instability (Khayet, 2011).
MD is a thermally driven process based on the principle of
vaporeliquid equilibrium and coupled heat and mass transfer.
During MD process, the volatile compounds in the hot feed
solution evaporate at the feed/membrane interface, diffuse
through the membrane pores and condense into liquid via
various approaches (Suárez et al., 2010). The non-volatile

* Corresponding author. Tel.: þ65 6516 6645; fax: þ65 6779 1936.
E-mail address: chencts@nus.edu.sg (T.-S. Chung).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.012
5490 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
compounds are left in the feed solution. In direct contact
membrane distillation (DCMD), the most widely MD configu-
ration in desalination application, the preheated brine feed
and cooled permeate are brought in direct contact with two
surfaces of micro-porous membrane, respectively. Owing to
the hydrophobic characteristic, MD membrane acts as barrier
between two phases. Even though in MD process, membrane
does not directly contribute the selectivity, its structure,
material and geometrical properties play an important role in
the production rate, mechanical property and wetting resis-
tance (Bonyadi et al., 2007). As a result, properties of MD
membranes have been investigated by many researchers in
recent years.
As a coupled mass and heat transfer process, the effective
driving force of DCMD process is the vapor pressure difference
between the feed and permeate interfaces. Unlike other
membrane separation processes, improvement of permeation
flux for a DCMD membrane can be ascribed to two possible
mechanisms: namely, improvement of mass transfer coeffi-
cient and enhancement of effective driving force. Via the first
mechanism, DCMD membrane structure should be tailored in
an attempt to reduce the mass transfer resistance. This can be
attained by fabricating membranes with large mean pore size
and porosity, open-cell pore structure, thin functional layer
and small tortuosity. Through the second mechanism, the
thermal conductivity of the membrane needs to be minimized
so as to reduce the temperature polarization and achieve
a higher driving force. It can be reduced by increasing the
membrane porosity, since the conductivity of filled air in
membrane pores is much lower than the polymer matrix
(Wang et al., 2009). Recently, by fabricating the single-layer
hollow fiber membranes with finger-like macrovoids, Wang
et al. reported that the fabrication of membrane with macro-
voids is an effective way for flux enhancement attributed to
high porosity and low tortuosity which allows fast diffusion of
water vapor (Wang et al., 2009). A remarkably high permeation
flux of 79 kg m
2 h
1 was achieved at 80  C feed. Apart from
improving permeation flux, the macrovoid structure also has
potential on energy recovery improvement ascribed to its
excellent thermal insulation because of air-filled macro pores.
However, the impact of the macrovoid structure on membrane
wetting and long-term stability are still concerned (Gryta and
Barancewicz, 2010; Wang et al., 2009).
During the continuous DCMD operation, there is a risk of
membrane pore wetting by vapor condensation and liquid
penetration. At a worst situation, entirely wetted pores would
result in the loss of membrane selectivity when the feed and
permeate solutions at both streams are in direct contact
(Tomaszewska, 1999). Therefore, wetting of membrane pores
should always be avoided. To improve the wetting resistance,
a membrane of high hydrophobicity, small pore size and high
tortuosity is highly desired (Phattaranawik et al., 2003).
Qtaishat et al. modified the polyetherimide (PEI) asymmetric
flat membrane by using fluorinated surface modifying macro-
molecules (SMM) to increase the membrane hydrophobicity
(Qtaishat et al., 2009). The contact angle of their modified
membrane increased by more than 10 and the wetting resis-
tance was also found to be improved greatly. On top of material
hydrophobicity, a membrane matrix with a fully sponge-like
structure can amplify the wetting resistance because of
narrow pore size distribution and higher tortuosity (Khayet,
2011). Teoh and Chung (2009) have produced composite poly-
vinylidene fluoride polytetrafluoroethylene (PVDF/PTFE) hollow
fiber membranes with a fully sponge-like structure. The resul-
tant fiber membranes showed a higher resistance to pore
wetting and stable performance in long-term tests (Gryta and
Barancewicz, 2010; Teoh and Chung, 2009).
On one hand, membranes with smaller tortuosity, higher
porosity, larger pore size, interconnected pore structure and
macrovoid structure may improve permeation flux. On the
other hand, larger tortuosity, smaller porosity and pore size as
well as sponge-like structure are desired for high wetting
resistance. The trade-off relationship makes it difficult for
traditional single-layer MD membranes to achieve both high
permeation flux and wetting resistance in a single configura-
tion. Hence, dual-layer hollow fiber membranes comprising
dual morphologies offer the opportunity to optimize these two
properties. The objective of this study is to elucidate a novel
approach to enhance the permeation flux and energy efficiency
(EE ) without compromising the wetting resistance via
morphological architecture. The inner-layer is designed to be
full of finger-like macrovoids whereas the outer-layer comprises
sponge-like structure. We also aim to measure liquid entry
pressure (LEP) using a homemade set-up as a quantitative
indication of membrane wetting resistance. In addition,
a mathematical model will be established based on the DCMD
permeation flux and operation properties to obtain and analyze
the essential transportation parameters including temperature
profiles and apparent diffusivity. This information could
provide the implicit understanding of vapor permeation
mechanism and relative contributions of aforementioned two
possible mechanisms for flux enhancement through these
newly developed MD membranes. Overall, this work may
provide profound implications for the strategies to enhance MD
performances and attempts to make the application of this
process more practical.
2. Materials and methods
2.1. Materials
The working polymer, Kureha PVDF#1300 resin, was kindly
provided by Kureha Corp. N-methyl-1-pyrrolidone (NMP,
>99.5%), ethylene glycol (EG, >99.5%) and isopropanol (IPA,
>99.5%) used in the hollow fiber fabrication were purchased
from Merck. The hydrophobic clay particles Closite 20A were
purchased from Southern Clay (Gonzales) while PTFE particles
(<1 mm) were supplied by SigmaeAldrich. The ultra-pure
water used in DCMD tests was produced by a Milli-Q unit
(MilliPore) with the resistivity of 18 MU cm.
2.2. Dope preparation
Prior to dope preparation, the PVDF resin and additive parti-
cles were dried in a vacuum oven (Model 282A, Thermo Fisher
Scientific Inc.) at 80  C to remove trapped moisture. A typical
procedure to prepare a dope is described as follows: The PVDF
polymer resin and the clay or PTFE particles with pre-
determined amounts were added into the NMP/EG solvent
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0 5491

mixture. After complete dissolution, the dope was maintained
at a slow stirring rate to prevent the sedimentation of the
particles before spinning. The compositions for each compo-
nent in all the dopes are listed in Table 1.
2.3. Hollow fiber fabrication
The dual-layer PVDF hollow fiber membranes were fabricated
via a dry-jet wet phase inversion spinning process by a tri-
orifice spinneret as documented elsewhere (Jiang et al.,
2004). During spinning, the inner-layer dope was held at
50  C to reduce the dope viscosity. Single-layer hollow fiber
with comparable overall thickness with the dual-layer fiber D3
was spun and denoted as S-out to investigate the effect of
sponge-like selective layer thickness on the DCMD perfor-
mance. Membrane S-out is produced by feeding the outer-
layer dope to the inner annulus of the spinneret and
employing a bore fluid contains water/NMP 60/40 wt%. Details
of the spinning conditions can be found in Table 1.
After spinning, the as-spun fibers were immersed in tap
water around 4 days to remove solvent residue. Subsequently,
the wet fibers were frozen in a refrigerator and dried overnight
in a freeze drier (S61-Modulyo-D, Thermo Electron Corp.). All
the spinning conditions have been repeated for three times to
ensure the reproducibility.
2.4. Membrane characterizations
2.4.1. Morphology study
Hollow fiber samples were observed by a field emission
scanning electron microscope (FESEM; JEOL JSM-6700). The
samples were prepared by being immersed and fractured in
liquid nitrogen.
Porosity of hollow fibers was obtained by Eq. (1).
 
mfiber =rfiber
3 ¼ 1
 100%
Vfiber
2.4.2. Contact angle measurements
The dynamic contact angle, q, was measured with a KSV
Sigma 701 tensiometer (0.01 , KSV Instruments Ltd.) through
a force tensiometry method at 25  C. During the tensiometry
measurement, a high sensitivity micro balance was utilized to
measure the interfacial force when the sample fiber was
brought into contact with water. Then, the dynamic contact
angle was calculated by a software module with the fiber
geometry.
2.4.3. Mechanical property measurements
The mechanical properties of fibers including Young’s modulus,
strain at break and maximum extension were measured by an
Instron tensiometer (Model 5542, Instron Corp.). A constant
elongation rate of 10 mm min
1 with a starting gauge length of
50 mm was applied. For each spinning condition, ten fiber
samples were tested so as to ensure the accuracy.
2.4.4. LEP measurements
The liquid entry pressure (LEP) was measured to examine the
membrane wetting resistance. The homemade set-up for LEP
test is shown in Fig. 1. A stainless steel test tube, which can
sustain high pressure, was used as the reservoir for testing
solution (3.5 wt% NaCl solutions). A digital pressure gauge
(Range: 0e2 bar, 0.001 bar, Wika) was installed on top of the
tube. To conduct the measurement, a hollow fiber module was
prepared by sealing one end of the fiber with epoxy whereas
leaving the other end open. During testing, compressed N2
was connected to the tube to generate the hydraulic pressure.
The outlet of the tube was connected to the lumen side of the
hollow fiber module. The whole module was put in a beaker of
deionized water. The hydraulic pressure was increased with
a step of 0.1 bar and each pressure was maintained for 10 min.
The conductivity of the water in the beaker was monitored by
a conductivity meter Lab 960 (0e500 ms cm
1, 0.1 ms cm
1,
SCHOOT instrument).
(1)

where Vfiber is the fiber volume, mfiber is the fiber weight and rfiber
is the density of the fiber material. Vfiber was estimated from
inner diameter, outer diameter and fiber length. mfiber was
measured by an accurate beam balance (A&D, GR-200). rfiber was
measured by a multi pycnometer (Quantachrome MVP-D160-E)
under helium purging at 20 psi. For each sample, 10 measure-
ments were carried out and an average was adopted.
2.5. DCMD desalination experiments
The DCMD desalination experiments were carried out to
evaluate the permeation flux at different conditions. Before
the experiments were conducted, the testing modules were
fabricated by assembling pre-determined number of fibers
into a plastic tube of 3/8 in. diameter, with both ends sealed by

Table 1 e Spinning conditions of single- and dual-layer PVDF hollow fibers.

Membrane ID D1 D2 D3 S-outa

Inner-layer dope composition (wt%) PVDF/NMP/EG/Closite 20A:10/77/10/3 Outer-layer dope

Inner-layer dope flow rate (ml min
1) 2.0 1.6 1.2 1.6
Outer-layer dope composition (wt%) PVDF/NMP/EG/PTFE:12/77/8/3 NA
Outer-layer dope flow rate (ml min
1) 0.3 0.3 0.3 NA
Bore fluid (wt%) water/NMP: 100/0 water/NMP: 100/0 water/NMP: 100/0 water/NMP: 60/40
Bore flow rate (ml min
1) 1.5 1.5 1.5 1.5
External coagulant (wt%) IPA/water: 60/40 IPA/water: 60/40 IPA/water: 60/40 IPA/water: 60/40
Air gap (cm) 2 2 2 2
Take up speed Free fall Free fall Free fall Free fall

a S-out is produced by feeding the outer dope to the inner annulus of the spinneret.
5492 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0

Fig. 1 e Experimental set-up for the LEP measurements.

epoxy. The number of fibers in each module was determined
by attaining a packing fraction of 30  5% and a filtration area
around 100 cm2.
A laboratory-scale DCMD set-up is shown in Fig. 2. A 3.5 wt
% NaCl solution was used as the feed whereby ultrapure water
was employed as the permeate. The inlet temperature of the
feed and permeate solutions were maintained at the target
value by a temperature circulator (F12, Julabo) and a water
cooler (RT7, Thermal Scientific), respectively. The feed solu-
tion was circulated by a centrifugal pump to the shell side of
the membrane module when a rotary pump circulates the
permeate solution to the lumen side of membrane module.
The linear velocities of feed and permeate solutions were kept
at 1.4 and 0.7 m s
1 respectively. To monitor the temperature
variation during the test, digital thermal couples with accu-
racy of 0.1  C were installed at the inlets and outlets of feed
and permeate streams, respectively. The ionic conductivity of
the permeate stream was measured before and after the
DCMD test to ensure there is no occasion of leakage. For each
test, the membrane module was tested for 30 min to ensure
the accuracy.
The permeation flux for each feed temperature was
calculated based on the outer surface of the fiber using Eq. (2):
DW
Nw ¼ (2)
Ao t
where Nw is the permeation flux, DW is the permeation weight
collected over a pre-determined time duration (t) and Ao is the
effective permeation area calculated based on the outer
diameter of hollow fiber.
The long-term DCMD desalination performance of the
fabricated dual-layer hollow fiber was conducted with fiber D3
and similar operation conditions with short-term DCMD
experiment. During the test, the inlet temperatures of feed
and permeate solutions were maintained at 60.0  C and
15.0  C, respectively. The produced water from the permeate

Fig. 2 e Laboratory-scale set-up for the DCMD experiment.

 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
tank was recycled to the feed tank to maintain the salinity of
feed solution. The separation factor (R) is calculated with Eq.
(3) (Wang et al., 2009):
 
cp
R¼ 1
 100%
cf
where cf and cp are the NaCl concentration of feed and
permeate solutions, respectively.
3. Results and discussion
3.1. Membrane characterizations
3.1.1. Hollow fiber membrane morphology
Fig. 3 displays the morphology of dual-layer hollow fiber
membrane D3. It can be seen that a novel morphology con-
sisting of an inner-layer full of finger-like macrovoids and an
outer-layer made of sponge-like structure has been attained.
From the enlarged images, large macrovoids of regular finger-
like shapes fully occupy the entire porous substrate layer,
whilst open cell sponge-like pores form at the outer selective
layer. As PVDF was used as the organic phase for the inner- and
outer-layer dopes, there is no delamination or interfacial
resistance between these two layers. These morphological
properties are in good agreement with the proposed
morphology. Based on the modeling result reported by Laganà
et al. (2000), an optimum selective layer thickness of the MD
membranes is about 30e60 mm by assuming the thermal
conductivity of the polymeric material to be 0.1e0.3 W m
1 K
1
and a typical pore tortuosity of 1.3. Applied into our proposed
morphology, the thickness of the sponge-like functional layer
should be within the recommended range. Hence, the spinning
condition was varied in order to optimize the membrane
morphology. Fibers D1, D2, D3 were spun using water as bore
Fig. 3 e Morphologies of the hollow fiber membrane D3 (IF [ 1.2 ml minL1). The dashed arrow shows the spinning direction.
5493
fluid and inner-layer dope flow rates of 2, 1.6 and 1.2 ml min
1,
respectively. The respective cross sections of the aforemen-
tioned fibers are shown in Fig. 4. Shown in Fig. 4 and Table 2,
a sponge-like layer as thin as w39 mm can be produced in fiber
(3) D3, which is within the range of the optimum membrane
thickness as proposed by Laganà et al. (2000). This thin sponge-
like layer is expected to enhance the vapor permeation effec-
tively without sacrificing in wetting resistance. As compared
with dual-layer fibers, single-layer fiber S-out exhibits smaller
macrovoids and sponge-like layer with thickness w104 mm. The
morphology is similar as the typical MD hollow fibers reported
previously (Teoh and Chung, 2009).
3.1.2. Porosity and mechanical properties measurements
Generally, membranes with high porosity have the benefits of
lower thermal conductivity and higher mass transfer coeffi-
cient, which are favorable for MD processes. In fact, due to the
high concentration of non-solvent (EG), the as-spun dual-layer
hollow fibers have elevated bulk porosities larger than 80%, as
shown in Table 2. Furthermore, the dual-layer hollow fibers
(D1eD3) exhibit superior properties than the single-layer one
(S-out), owing to the large amount of macrovoids formed in
the inner-layer.
The measured Young’s modulus, tensile stress at break and
strain at break are summarized in Table 2. Membranes with
macrovoid structure often suffer from the weak mechanical
strength (Widjojo and Chung, 2006). The Young’s modulus and
tensile stress of the self-fabricated dual-layer hollow fiber are
comparable with other microporous membranes (Tsai et al.,
2001), which make it acceptable for low pressure membrane
operation. Compared with the single-layer membrane (S-out),
the tensile stress at break of dual-layer hollow fibers is slightly
lower. This may be attributed to existence of long finger-like
macrovoids in the membrane inner-layer. On the other hand,
the observed mechanical properties for dual-layer membrane
decreases with a reduction in membrane overall thickness.
5494 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0

Fig. 4 e Morphology of the cross-sections for hollow fibers: D1, D2, D3 and S-out.

Since MD membranes are not operated at high pressure and calibrated correlation between conductivity and NaCl concen-
the sponge-like structure provides sufficient wetting resis- tration, the amount of testing solution permeates across the
tance (as discussed in Section 3.1.1), the newly developed wetted membrane can be calculated.
membranes can be further tested in LEP measurements and Fig. 5(a) illustrates the calculated permeation fluxes under
DCMD experiments. different pressures for hollow fiber D1. Under low hydraulic
pressures (0e0.6 bar), no obvious wetting can be observed
3.1.3. LEP measurements through the study. Nonetheless, when the applied pressure
Wetting resistance is one of essential performance indicators to increases to 0.7 bar, there is a trace amount of testing solution
assess the long-term prospective of any newly developed permeating through the membrane, which clearly indicates
membranes for MD applications. It is known that the LEP the wetting of bigger pores. Interestingly, a further increase in
measurement is a quantitative indication of wetting resistance. pressure not only results in partial wetting, but also promotes
During the LEP tests, a 3.5 wt% NaCl testing solution is directly the wetting of smaller pores. Since LEP is defined as the
in contact with the feed side of the respective membranes. minimum pressure and the sustainability of the membrane
Once the applied pressure reaches the LEP value, some of the toward liquid penetration, the LEP value for hollow fiber D1 is
membrane pores will be wetted and the testing solution will determined as 0.7 bar.
permeate through those wetted pores. Thus the conductivity of A similar procedure has been repeated for other fibers and
the deionized water will increase gradually. With the pre- the LEP values for the aforementioned membranes are tabu-
lated in Table 2. It can be shown that LEP values for all tested
membranes are higher than the typical operation pressure of
DCMD experiments, which is within the range of 0.1e0.3 bar.
It is worth noting that dual-layer membranes (D1, D2 and D3)
Table 2 e Characteristic properties of single- and dual- with similar sponge-like layer but different macrovoids layer
layer PVDF hollow fibers. thicknesses exhibit the same LEP values of 0.7 bar. This
Membrane ID D1 D2 D3 S-out implies that the sponge-like layer, rather than the macrovoid
layer, is the key factor in determining the capability of
OD (mm) 1192  21 1049  19 961  22 982  12
membrane wetting resistance. A comparison of LEP values for
Wall thickness 198  13 164  17 141  24 143  6
(mm) single- and dual-layer membranes further reveals the impact
Sponge-layer 78  6 70  7 39  1 104  5 of the sponge-like layer thickness on the membrane wetta-
thickness (mm)* bility. According to Garcia-Payo et al. (2000), more tortuous
Contact angle ( ) 106  5.0 107  4.6 110  2.9 113  5.5 pore structure and thicker sponge-like layer provide greater
Porosity (%) 89.4  1.9 86.7  2.0 84.0  1.1 73.6  1.6 resistance to pore wetting. Thus, membrane S-out with
LEP (bar) 0.7 0.7 0.7 1.0
a thicker sponge-like layer of around 106 mm can provide
Strain at 116  4 92  3 74  7 104  3
break (%)
a slightly higher LEP and better wetting resistance.
Tensile stress 0.78  0.03 0.66  0.05 0.62  0.09 1.08  0.02
at break (MPa)
Young’s modulus 30.1  1.4 26.9  1.4 25.4  1.7 33.5  2.3
3.2. DCMD performance
(MPa)
The permeation fluxes obtained for both single- and dual-layer
* Sponge-like layer thickness is measured as distribution of silicon
hollow fibers vary with feed inlet temperatures are shown in
element in Energy-dispersive spectroscopy (EDX).
Fig. 5 (b). For all the membrane samples, the conductivity of the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0 5495

permeate solutions is <5 ms cm
1 with separation factor single-layer fiber (S-out). This evidently discloses the note-
>99.98%, which indicates the membranes are not wetted. worthy function of finger-like macrovoids at the membrane
Among these dual-layer hollow fibers, fiber D3 which was spun inner-layer for the enhancement of DCMD flux. As proposed
at an inner-layer flow rate of 1.2 ml min
1 demonstrates earlier, the dual-layer fiber consisting of a macrovoids sup-
a superior permeation flux as high as 98.6 L m2 h
1 at an inlet porting layer has a lower mass transfer resistance and a higher
feed temperature of 80.2  C. Since an increase in inner-layer driving force resulting from the smaller tortuosity and higher
flow rate during spinning results in a thicker inner-layer and porosity. Further details about the relative contributions of
decays the flux dramatically, the values of the flux obtained for these two effects will be discussed in Section 3.3. Generally, the
fibers D1 and D2 are about 9.7% and 28.3% lower than fiber D3, permeation flux increases almost exponentially with increasing
respectively. This may be owing to the fact that larger overall feed temperature owing to the exponential dependence of
membrane and sponge-like layer thicknesses lead to decay in vapor pressure on temperature (Bonyadi and Chung, 2007).
membrane mass transfer coefficient. The normalized permeation with a cylindrical coordinate
With a similar overall wall thickness, the permeation flux and hollow fiber configuration has been calculated with Eq. (4)
obtained for dual-layer hollow fiber D3 is w96% higher than the (Hou et al., 2009):

a 6.0 b 120
D1
D2

L m-2 hr-1)
L m-2 hr-1)

5.0 100 D3
S-out
4.0
0.9bar 80

Permeation Flux, (L
Permeation Flux (L

3.0 60
0.8bar
2.0 40

1.0 0-0.6bar 20
0.7bar
P

0.0 0
0 3 6 9 12 45 55 65 75 85
c Time (min) Feed Inlet Temperature (°C)
d
Normalized Permeation Flux, (L m-1 s-1)

6.0E-06 100
D1 Tf
D2 Tf,m
Temperaature Profile (°C)

5.0E-06
80
D3 Tp
4.0E-06 S-out Tp,m
60
3.0E-06
40
2.0E-06

20
1.0E-06

0.0E+00 0
45 55 65 75 85 45 55 65 75 85
Feed Inlet Temperature (°C) Feed Inlet Temperature (°C)
e f
2.0E-05 1.00
parant Diffusivity, KD ,( m2 s-1)

D1
D2
1.6E-05 D3
Energy Efficiency, EE

0.80 S-out
E

1.2E-05
0.60
8.0E-06
D1
4.0E-06 1µmD2 0.40
100nm
D3
App

S-out
0.0E+00 0.20
45 55 65 75 85 45 55 65 75 85
Feed Inlet Temperature (°C) Feed Inlet Temperature (°C)

Fig. 5 e (a) Permeation fluxes obtained from LEP measurements for hollow fiber D1. (b) DCMD Permeation fluxes obtained for
PVDF hollow fibers. (c) Normalized DCMD Permeation fluxes obtained for PVDF hollow fibers. (d) Temperature profile for
PVDF hollow fiber D1: Tf and Tp are the average temperatures for feed and permeate sides; Tf,m and Tp,m are temperatures
calculated for membrane surfaces facing the feed and permeate sides, respectively. (e) Calculated apparent Diffusivities (Da)
for PVDF hollow fibers. (f) Calculated energy efficiencies for PVDF hollow fibers.
5496 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
 
ro
Nw;normal ¼ Nw ro ln
ri
Fig. 5(c) demonstrates the normalized flux for all the
membranes. The results show that all the dual-layer
membranes have the similar normalized fluxes, which may
be attributed to the similar structure (finger-like macrovoids
structure and sponge-like structure). The values of normal-
ized fluxes of dual-layer membranes are much higher than the
single-layer membrane.
3.3. Modeling of mass transfer in DCMD
Due to the complexity of DCMD processes, the vapor trans-
portation across a MD membrane is often determined by many
parameters including membrane micro-structure, membrane
module configuration and operation conditions. In order to
investigate the detailed mass transport of the self-fabricated
hollow fiber membranes, a mathematical model is proposed.
This model aims to acquire essential mass transfer parameters
as temperature profile and apparent diffusivity with the oper-
ation parameters and permeation flux data obtained from the
DCMD experiments. The key information and results of the
modeling is shown in this section and detailed modeling
equations can be found in the Supplementary material.
3.3.1. Temperature profile
Fig. 6 demonstrates the temperature changes and heat
transfer across the hollow fiber membrane. As a coupled heat
and mass transfer process, the vapor pressure difference
between the bulk feed and permeate solutions cannot be fully
utilized as the effective driving force. A considerable propor-
tion of heat loss (i.e., temperature polarization) is observed at
the boundary layers. Therefore, investigation of the temper-
ature profile across MD membrane is a critical approach to
understand the heat transfer during the DCMD process.
Fig. 6 e (a) Temperature profile and heat transfer across a DCMD hollow fiber and (b) total heat transfer along the hollow fiber
module.
An attempt to estimate the temperature profile across the
(4)
MD hollow fiber was performed as follows: (1) the steady state
heat flux (Qss) is equal to the enthalpy change between the
inlet and outlet of both feed and permeate streams, and (2) the
convective heat transfer coefficients (hf and hp) are obtained
from heat transfer correlations. Thus, the temperatures at
membrane surfaces can be calculated by Eqs. (5) and (6).
Qss
Tf ;m ¼ Tf
(5)
hf Ao
Qss
Tp;m ¼
Tp (6)
hp Ai
where Tf,m and Tp,m are the average temperatures at membrane
surfaces toward feed and permeate, respectively. Tf and Tp are
the average temperatures of bulk feed and permeate solutions,
respectively. hf and hp are the heat convection coefficients of
feed and permeate sides, respectively. Ao and Ai are the area of
outer and inner surface of hollow fiber, respectively, and Qss is
the average heat flux at steady state.
Fig. 5(d) illustrates the temperature profile of fiber D1. The
temperatures difference of bulk and membrane surface at the
feed side (i.e., 3e4  C) is relatively lower than the gap at the
permeate side (i.e., 10e20  C). This may be due to the fact that
the heat transfer coefficient and mass flow rate are much
higher at the feed as compared to the permeate side. At steady
state, where the heat fluxes at feed and permeate sides are
equal, the resultant heat transfer boundary layer at the
permeate side is thicker than the feed side. As a result,
a dramatic temperature drop is expected from the membrane
surface to the permeate bulk solution. In addition, the
temperature difference for the permeate side are more
significant at a higher feed temperature, but no noticeable
changes can be observed at the feed side. The increase of the
permeate side is owing to a higher total heat flux (Qss) arisen
from superior heat conduction flux and permeation flux. In
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
contrast, the heat transfer coefficient for the feed side
increases simultaneously as the feed temperature and total
heat flux increase, which results in a smaller change for the
temperature gap between Tp and Tp,m.
3.3.2. Apparent diffusivity
Usually, the permeation flux obtained in the DCMD is strongly
correlated by the fiber’s structure, such as wall thickness,
inner and outer diameters. In order to investigate the function
of the proposed morphology, the parameter of apparent
diffusivity is introduced which exclude the impacts of
membrane thickness, fiber diameters and MD operation
parameters. It is defined as Eq. (7):
3 Dw
a
Da ¼
s the percentage of latent heat flux over the total heat flux
where Dw
a is the diffusion coefficient of water vapor through
stagnant air, s is the tortuosity of vapor diffusion path in the
membrane.
For industrial processes, apparent diffusivity is one of the
most important modeling parameters to examine the diffu-
sion of gas or liquid through a porous structure (Merdas et al.,
2002). Hereby, it is introduced as a quantitative measurement
of the intrinsic diffusion property. The apparent diffusivity
can be obtained from the equation that describes the diffusion
of vapor through a porous hollow fiber membrane as following
(Fernandez-Pineda et al., 2002):
 
1 Mww
Nw ¼ Da Pf ;m
Pp;m
Ylm ro lnðro =ri Þ
where Nw is the permeation flux, Mww is the molecular weight
of water, Ylm is the logarithm mean fraction of air in
membrane pores, ro and ri are the outer radius and inner
radius of the hollow fiber respectively, Pf ;m and Pp;m are the
vapor pressure at the membrane surfaces facing feed and
permeate sides respectively. (The effect of concentration
polarization (CPC) on Pf ;m is taken into consideration with
calculated mass transfer coefficient. The detailed equations
are included in Supplementary material.)
The calculated Da values for all PVDF follow fibers are pre-
sented in Fig. 5(e). For dual-layer fibers, the Da values are fairly
comparable, which is possibly attributing to their similarity in
terms of membrane morphology. From Eq. (7), it can be
concluded that the value of Da is only determined by the
membrane characteristics such as mean pore size, porosity and
tortuosity. Owing to the similar coagulation conditions, dual-
layer membranes have a similar pore size and porosity.
Besides, Table 2 displays that the dual-layer fibers also show
similar percentages of sponge-like layer thickness to the overall
thickness (0.3e0.35). This result suggests a comparable average
tortuosity and consequently obtained similar Da values. A
comparison of Da values between dual-layer and single layer
provides solid evidence that the enhancement in mass transfer
rate is mainly due to a lower mass transfer resistance of finger-
like macrovoids. At the feed temperature of 80  C, the Da value
of the dual-layer membrane (D3) is w80% higher than that of
the single-layer fiber (S-out). As discussed in Section 3.2, the
finger-like macrovoid structure in the MD membrane enhances
the permeation flux through two possible mechanisms:
namely, increased mass transfer rate and enhanced driving
5497
force. In this regards, about 1-fold permeation flux improve-
ment is attained by membrane D3 as compared to the fiber S-
out, as shown in Fig. 5(b). Therefore, a conclusion can be drawn
that superior permeation fluxes obtained for the proposed MD
fibers is a coupling effect of both mechanisms, while the
enhancement of mass transfer rate is the major contribution.
3.4. Energy efficiency
As a coupled mass and heat transfer process, the total heat
transfer in the DCMD process consists of two mechanisms: (1)
the latent heat associated with the vaporization and
condensation of water (Ql), and (2) the heat conduction across
the membrane which usually indicated as undesirable heat
(7) loss (Qc). Hence, the energy efficiency, EE, can be denoted to
across the membrane (Smolders and Franken, 1989), and
calculated using Eq. (9) for hollow fibers:
Ql N w l m Ao
h¼ ¼
Q c þ Ql km

Tf
Tp Ao þ Nw lm Ao
ro lnðro =ri Þ
Nw lm Ao
¼
(9)
mp Cp;p Tp;out
Tp;in
where lm is the latent heat of water vaporization evaluated at
average membrane temperature, Tf and Tp are average
temperatures at feed and permeate sides, respectively. km is
the thermal conductivity of the membrane, mp is the mass
(8) flow rate of the permeate solution, Cp,p is the average specific
heat capacity of the permeate solution, Tp,in and Tp,out are inlet
and outlet temperatures of the bulk permeate solution,
respectively.
The calculated EE against feed inlet temperature for all
membranes are shown in Fig. 5(f). The EE increases with the
feed inlet temperature. This is attributed to the fact that the
driving force for vapor transport increases exponentially,
whereas the conduction heat increases linearly with the feed
temperatures. The EE of fiber D3 are almost twice of the fiber
S-out; this improvement is mainly caused by the co-action of
two factors: namely, enhanced mass transfer coefficient and
reduced thermal conductivity. As referring to Fig. 5(e), the
dual-layer fiber D3 exhibits a much higher apparent diffusivity
and a less mass transport resistance, and consequently larger
permeation fluxes (Nw) and latent heat flux (Ql). On the other
hand, because the thermal conductivity of the stagnant air in
pore structure is much lower than the membrane polymer
matrix (Curcio and Drioli, 2005). The unique micro-structure
with finger-like macrovoids with a high porosity in fiber D3
effectively reduces the heat conduction loss (Qc). Therefore,
compared with single-layer configuration, the unique dual-
layer morphology allows the combined advantages offered
by the macrovoid supporting layer (i.e., higher permeation
flux and energy efficiency) and a fully sponge-like outer
selective layer (i.e., withstand membrane wetting).
For dual-layer hollow fibers, the EE values increase gradually
with a decrease in the overall fiber thickness. This is mainly
attributed to the adverse effects of membrane thickness
on thermal conduction and evaporation, where a thinner
membrane may enhance both conduction heat flux (Qc) and
latent heat flux (Ql). In other words, a decrease in membrane
5498 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0

100 100.00%

Permeation Flux (L m-2 hr-1)

Separation factor (%)

80 99.99%

60 99.98%

40 99.97%

20 99.96%

0 99.95%
0 20 40 60 80 100 120 140 160 180 200

Time (hr)

Fig. 7 e Variation of permeation flux and separation factor during the long-term DCMD experiment.

thickness could increase both the de-numerator and numerator
in Eq. (9). Hence, the changes of EE values are depending on the
relative changing rate of de-numerator and numerator. Due to
the high apparent diffusivities of dual-layer membranes, the
rate of enhancement in latent heat flux (Ql) is much greater than
the conduction heat flux (Qc). As a result, the latent heat flux is
the dominant factor on the calculated EE for all the tested MD
membranes.
3.5. Long-term desalination performance
The long-term DCMD desalination performance of the fabri-
cated dual-layer hollow fiber was conducted for 200 h. The
permeation flux and separation factor are illustrated in Fig. 7. It
can be seen that the permeation flux and separation factor are
relatively stable during the whole test. The separation factor is
higher than 99.9%, which indicates a high wetting resistance of
the tested membranes. A slight decline of permeation flux and
separation factor can be observed at w around 80 h. This may
be attributed to the scaling of inorganic salts on the membrane
surfaces (Gryta and Barancewicz, 2010; Song et al., 2008).
3.6. Comparison with other MD membranes
Table 3 shows a comparison of DCMD performance between
this work and the literature data. For a better comparison, the
literature data consists of commercial and laboratory fabri-
cated flat sheet and hollow fiber membranes. A noticeable flux

Table 3 e DCMD performances of MD membranes for desalination.

Membrane Thickness Feed solution property Permeate solution property Permeation flux

(mm) Solution Tf,in vf Tp,in vp (L m2 h
1)

( C) (m s
1) ( C) (m s
1)

GVHP commercial flat sheet 118 Deionized water 90.7 500 rpm 19.7 500 rpm 48.6
(Khayet, 2011)
PVDF flat sheet e 1.0 wt% NaCl 60 e 20 e 10
(Tomaszewska, 1996)a
SMM/PEI flat sheet 160a 0.5 M NaCl 50 500 rpm 15 500 rpm 20.9
(Qtaishat et al., 2009)
Accurel Membrana GmbH PP 150 1.0 wt% NaCl 90 2.29 15e17 1.66 41.4
commercial hollow fiber
(Li and Sirkar, 2004)b
PVDF single-layer hollow fiber 110 3.5 wt% NaCl 79.5 1.9 17.5 0.9 40.4
(Teoh and Chung, 2009)
PVDF single-layer hollow fiber 130 3.5 wt% NaCl 81.8 0.5 20 0.15 40.5
(Hou et al., 2009)
PVDF single-layer hollow fiber 190 3.5 wt% NaCl 81.3 1.8 17 1.2 79.2
(Wang et al., 2009)
PVDF dual-layer hollow fiber 340 3.5 wt% NaCl 90.3 1.6 17 0.8 55.2
(Bonyadi and Chung, 2007)
PVDF dual-layer hollow fiber 271 3.5 wt% NaCl 78.2 1.8 16.6 0.72 66.9
(Su et al., 2010)
PVDF dual-layer hollow 141 3.5 wt% NaCl 79.8 1.4 17 0.7 98.6
fiber D3 (this work)

Boldface denotes result of the presented research. All other data are from the literature.
a The thickness was measured from the SEM image.
b The original data was reported based on the inner diameter, we had converted the permeation flux based on outer diameter, which is more
convenient for the comparison.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
enhancement has been achieved in this work. This is denoted
to the morphological architecture with the aid of the dopes
formulation and spinning conditions.
4. Conclusions
We have demonstrated that morphological architecture of
PVDF dual-layer hollow fiber membranes can be utilized to
enhance the DCMD performance. The dual-layer hollow fiber
with a fully finger-like inner-layer and a totally sponge-like
outer-layer is a preferred structure. The resultant dual-layer
membranes show enhanced MD performance with minimum
sacrifice of wetting resistance. The aforementioned
morphology can be manipulated and tailor made via careful
manipulations of inner- and outer-dope compositions as well
as coagulation conditions. As a result, a permeation flux of
98.6 L m2 h
1 has been obtained. Credited to the high thermal
resistance and excellent permeation flux, the dual-layer fiber
exhibits a remarkably high EE value of w94%. Furthermore,
dual-layer membranes with sponge-like layer thicknesses
between 30 and 40 mm reveal higher membrane wetting
resistance with the measured LEP values of 0.7 bar.
The mass transfer modeling shows that the apparent
diffusivity values for dual-layer fibers are remarkably higher
than the single-layer configuration. This implies that the
former has a much smaller resistance for vapor transfer than
the latter. The as-spun fibers membranes exhibit a better
permeation flux due to the coupling effects of lower mass
transfer resistance and higher effective driving force.
However, a lower mass transfer resistance is the dominant
factor for the higher performance.
Acknowledgments
The authors would like to acknowledge A*STAR and National
University of Singapore for funding the research through the
grant number R-279-000-291-305. The authors also appreciate
Kureha Corp., Japan for the provision of the Kureha PVDF
resin. Special thanks are due to Prof. E.L. Cussler (University of
Minnesota), Dr. K. Y. Wang, Dr. N. Widjojo, Dr. J. C. Su, Miss H.
Wang and Mr. Y. H. Sim for their valuable suggestions.
Appendix. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.012.
references
Bonyadi, S., Chung, T.S., 2007. Flux enhancement in membrane
distillation by fabrication of dual layer hydrophilic-
hydrophobic hollow fiber membranes. Journal of Membrane
Science 306 (1e2), 134e146.
5499
Bonyadi, S., Chung, T.S., Krantz, W.B., 2007. Investigation of
corrugation phenomenon in the inner contour of hollow fibers
during the non-solvent induced phase-separation process.
Journal of Membrane Science 299 (1e2), 200e210.
Curcio, E., Drioli, E., 2005. Membrane distillation and related
operations e a review. Separation and Purification Reviews 34
(1), 35e86.
Escobar, I.C., 2010. A summary of challenges still facing
desalination and water reuse. In: Escobar, I.C., Schäfer, A.I.
(Eds.), Sustainable Water for the Future e Water Recycling
Versus Desalination. Elsevier Science, The Netherlands, pp.
389e397.
Fernandez-Pineda, C., Izquierdo-Gil, M.A., Garcia-Payo, M.C.,
2002. Gas permeation and direct contact membrane
distillation experiments and their analysis using different
models. Journal of Membrane Science 198 (1), 33e49.
Garcia-Payo, M.C., Izquierdo-Gil, M.A., Fernandez-Pineda, C.,
2000. Wetting study of hydrophobic membranes via liquid
entry pressure measurements with aqueous alcohol solutions.
Journal of Colloid and Interface Science 230 (2), 420e431.
Gryta, M., Barancewicz, M., 2010. Influence of morphology of
PVDF capillary membranes on the performance of direct
contact membrane distillation. Journal of Membrane Science
358 (1e2), 158e167.
Hou, D., Wang, J., Qu, D., Luan, Z., Ren, X., 2009. Fabrication and
characterization of hydrophobic PVDF hollow fiber
membranes for desalination through direct contact
membrane distillation. Separation and Purification
Technology 69 (1), 78e86.
Jiang, L., Chung, T.S., Li, D.F., Cao, C., Kulprathipanja, S., 2004.
Fabrication of matrimid/polyethersulfone dual-layer hollow
fiber membranes for gas separation. Journal of Membrane
Science 240 (1e2), 91e103.
Khayet, M., 2011. Membranes and theoretical modeling of
membrane distillation: a review. Advances in Colloid and
Interface Science 164 (1e2), 56e88.
Laganà, F., Barbieri, G., Drioli, E., 2000. Direct contact membrane
distillation: modelling and concentration experiments.
Journal of Membrane Science 166 (1), 1e11.
Li, B.A., Sirkar, K.K., 2004. Novel membrane and device for direct
contact membrane distillation e based desalination process.
Industrial & Engineering Chemistry Research 43, 5300e5309.
Merdas, I., Thominette, F., Tcharkhtchi, A., Verdu, J., 2002. Factors
governing water absorption by composite matrices.
Composites Science and Technology 62 (4), 487e492.
Phattaranawik, J., Jiraratananon, R., Fane, A.G., 2003. Effect of
pore size distribution and air flux on mass transport in direct
contact membrane distillation. Journal of Membrane Science
215 (1e2), 75e85.
Qtaishat, M., Matsuura, T., Khayet, M., Khulbe, K.C., 2009.
Comparing the desalination performance of SMM blended
polyethersulfone to SMM blended polyetherimide membranes
by direct contact membrane distillation. Desalination and
Water Treatment 5 (1e3), 91e98.
Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G.,
Mar~ıas, B.J., Mayes, A.M., 2008. Science and technology for
water purification in the coming decades. Nature 452 (7185),
301e310.
Smolders, K., Franken, A.C.M., 1989. Terminology for membrane
distillation. Desalination 72 (3), 249e262.
Song, L., Ma, Z., Liao, X., Kosaraju, P.B., Irish, J.R., Sirkar, K.K.,
2008. Pilot plant studies of novel membranes and devices for
direct contact membrane distillation-based desalination.
Journal of Membrane Science 323 (2), 257e270.
Su, M., Teoh, M.M., Wang, K.Y., Su, J., Chung, T.S., 2010. Effect of
inner-layer thermal conductivity on flux enhancement of
dual-layer hollow fiber membranes in direct contact
membrane distillation. Journal of Membrane Science.
5500 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 4 8 9 e5 5 0 0
Suárez, F., Tyler, S.W., Childress, A.E., 2010. A theoretical study of
a direct contact membrane distillation system coupled to
a salt-gradient solar pond for terminal lakes reclamation.
Water Research 44 (15), 4601e4615.
Teoh, M.M., Chung, T.S., 2009. Membrane distillation with
hydrophobic macrovoid-free PVDF-PTFE hollow fiber
membranes. Separation and Purification Technology 66 (2),
229e236.
Tomaszewska, M., 1999. Membrane distillation. Environment
Protection Engineering 25 (1), 37e47.
Tomaszewska, M., 1996. Preparation and properties of flat-sheet
membranes from poly(vinylidene fluoride) for membrane
distillation. Desalination 104 (1e2), 1e11.
Tsai, H.A., Huang, D.H., Ruaan, R.C., Lai, J.Y., 2001. Mechanical
properties of asymmetric polysulfone membranes containing
surfactant as additives. Industrial and Engineering Chemistry
Research 40 (25), 5917e5922.
Wang, K.Y., Foo, S.W., Chung, T.S., 2009. Mixed matrix PVDF
hollow fiber membranes with Nanoscale pores for
desalination through Direct Contact Membrane Distillation.
Industrial & Engineering Chemistry Research 48 (9),
4474e4483.
Widjojo, N., Chung, T.S., 2006. Thickness and air gap dependence
of macrovoid evolution in phase-inversion asymmetric hollow
fiber membranes. Industrial & Engineering Chemistry
Research 45 (22), 7618e7626.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Effect of short-chain organic acids on the enhanced

desorption of phenanthrene by rhamnolipid biosurfactant
in soilewater environment

Chun-jiang An 1, Guo-he Huang*, Jia Wei 1, Hui Yu 1

Faculty of Engineering and Applied Science, University of Regina, Regina, Saskatchewan, Canada S4S 0A2

article info abstract

Article history: This study investigated the effect of short-chain organic acids on biosurfactant-enhanced
Received 14 December 2010 mobilization of phenanthrene in soilewater system. The desorption characteristics of
Received in revised form phenanthrene by soils were assessed in the presence of rhamnolipid and four SCOAs,
15 June 2011 including acetic acid, oxalic acid, tartaric acid and citric acid. The tests with rhamnolipid
Accepted 7 August 2011 and different organic acids could attain the higher desorption of phenanthrene compared
Available online 16 August 2011 to those with only rhamnolipid. Among the different combinations, the series with
rhamnolipid and citric acid exhibited more significant effect on the desorption perfor-
Keywords: mance. The removal of phenanthrene using rhamnolipid and SCOAs gradually increased
Phenanthrene as the SCOA concentration increased up to a concentration of 300 mmol/L. The effects of
Biosurfactant pH, soil dissolved organic matter and ionic strength were further evaluated in the presence
Short-chain organic acids of both biosurfactant and SCOAs. The results showed that the extent of phenanthrene
Desorption desorption was more significant at pH 6 and 9. Desorption of phenanthrene was relatively
Field soil lower in the DOM-removed soils with the addition of biosurfactant and SCOAs. The
presence of more salt ions made phenanthrene more persistent on the solid phase and
adversely affected its desorption from contaminated soil. The results from this study may
have important implications for soil washing technologies used to treat PAH-contaminated
soil and groundwater.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction been demonstrated to be a threat to human health and

Polycyclic aromatic hydrocarbons (PAHs) are frequently found
in the environment as a result of various natural processes
and anthropogenic activities (Nadal et al., 2004). During the
past decades, large amounts of PAHs have been generated in
incomplete combustion of organic materials and the refining
of crude oil (Ferrarese et al., 2008). Contamination from the
release of petroleum hydrocarbons into the environment by
deliberate discharges, accidental spills and leakages is
a particularly significant problem. These contaminants have
ecological safety (Djomo et al., 2004; Yu et al., 2011). Pollution
caused by these compounds is becoming a widespread
problem for the ecosystem across the world. The U.S. Envi-
ronmental Protection Agency (EPA) has already classified
PAHs as priority pollutants (Kanaly and Harayama, 2000). Due
to the low solubility and high hydrophobicity, a significant
proportion of these organic contaminants can bind to soil
sediments (Gomez et al., 2010; Wei et al., 2011). Mass transfer
limitation from the solid phase to the liquid phase seriously
impedes the bioavailability and reduces the bioremediation

* Corresponding author. Tel.: þ1 306 585 4095; fax: þ1 306 585 4855.
E-mail addresses: an209@uregina.ca (C.-j. An), huang@iseis.org (G.-h. Huang), jia.wei@iseis.org (J. Wei), yuhui200@uregina.ca (H. Yu).
1
Tel.: þ1 306 585 4095; fax: þ1 306 585 4855.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.011
5502 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
efficiency of these compounds. PAHs often persists in soil and
groundwater system, posing a long-term threat to the
environment.
Soil washing with extracting solutions has been proposed
as a promising technique for the remediation of contaminated
soils and groundwater in recent years (Deshpande et al., 1999;
Chu and Chan, 2003; Gan et al., 2009). These washing solutions
include different types of chemical agents, such as surfac-
tants, solvents, and cosolvents (Khodadoust et al., 2000; Silva
et al., 2005; Alcantara-Garduno et al., 2008). Among these
compounds, surfactant has been extensively used to increase
the desorption and mobilization of pollutants in soil due to its
unique characteristic. Surfactant is composed of both
a hydrophilic and hydrophobic moiety, which can form
micelle aggregates in aqueous phase when surfactant
concentration exceeds critical micelle concentration (CMC)
(Deshpande et al., 2000). These micelles significantly facilitate
the solubilization of hydrophobic organic compounds (HOCs)
and their transport from contaminated soil to aqueous phase.
However, concerns over some synthetic surfactants have
increased because they are toxic and thus present a potential
risk for environment and human health (Kanga et al., 1997). As
a more prominent and less toxic alternative, biosurfactants,
such as rhamnolipids, are starting to be widely used in soil
remediation. Rhamnolipids are anionic biosurfactants
produced by Pseudomonas sp. bacteria and they can show
greater environmental compatibility and surface activity
(Clifford et al., 2007). The application of rhamnolipid in field
remediation of PAH-contaminated soil has been successfully
demonstrated in many previous studies (Noordman et al.,
1998; Bordas et al., 2005; Mulligan, 2005). However, a good
understanding of the various factors involved in
biosurfactant-enhanced remediation is still a challenge in
many respects. Further study is necessary to define the role of
complex system composition and heterogeneity in contami-
nant behaviors.
The technology of soil washing with short-chain organic
acids (SCOAs) has also been employed to remediate polluted
soils. In natural environment, these organic acids can be
produced from incomplete carbohydrate oxidation and
chemical acidification within root exudates, microbial secre-
tions, and decomposition processes (Blank et al., 1994; Wang
et al., 2009). These simple carboxylic acids can act as
chelating agents and facilitate the translocation of metals in
soils (Jones, 1998; van Hees et al., 2003). Such properties have
been used to increase the treatment efficiency for heavy metal
contaminated soil in many applications (Yuan et al., 2007;
Zhang et al., 2008). Recently, it was found the presence of
SCOAs could have an influence on the behavior of HOCs.
White et al. (2003) reported that the availability of weathered
p,p0 -DDE increased in the presence of seven organic acids. It
was also observed in our previous studies that the addition of
SCOAs showed adverse effect on adsorption process of pyrene
(An et al., 2010). Although these findings are encouraging,
a number of important issues about the role of SCOAs remain
poorly understood under either laboratory or field conditions.
Moreover, with the prevalence and concern of biosurfactant,
few studies have focused on the interacted influence of bio-
surfactants and organic acids on the behaviors and mecha-
nisms of PAH translocations.
The present study therefore investigated the effect of
SCOAs on the mobilization of phenanthrene by rhamnolipid
biosurfactant in soilewater environment. Batch tests were
conducted to determine the desorption of phenanthrene in
two kinds of natural soil samples. The performances of
different biosurfactant-SCOA combinations, as well as the
effects of solution pH, dissolved organic matter and ionic
strength were evaluated. The results from this study may
have important implications for soil washing technologies
used to treat PAH-contaminated soil and groundwater.
2. Material and methods
2.1. Biosurfactant
The biosurfactant used in this study was a rhamnolipid
obtained from the Jeneil Biosurfactant Company (WI, USA).
Jeneil product RECO-10, with a mono- to di-rhamnolipid ratio
of 1:1, was used and the biosurfactant was supplied as a 10%
aqueous solution. It is an extra-cellular natural substance
produced during precisely controlled fermentation processes
utilizing certain bacterial strains (Wei et al., 2005). The two
major rhamnolipid components in this solution are a mono-
rhamnolipid (a-L-rhamnopyranosyl-b-hydroxydecanoyl-b-
hydroxydecanoate), and a dirhamnolipid (2-O-a-L-rhamno-
pyranosyl-a-L-rhamnopyranosyl-b-hydroxydecanoyl-b-hydr-
oxydecanoate). The molecular weight of monorhamnolipid
(C26H48O9) and dirhamnolipid (C32H58O13) are 504 g/mol and
650 g/mol, respectively. The critical micellar concentration of
this biosurfactant is 24 mg/L.
2.2. Chemicals
Phenanthrene was selected as representative PAH, and was
obtained from Aldrich Chemical Co. (WI, USA) with purity
greater than 99%. Four species of SCOAs, including acetic acid,
oxalic acid, tartaric acid and citric acid, obtained commer-
cially as pure chemical reagents (Sigma Aldrich, MO, USA),
were used in the study. Dichloromethane was obtained from
Fisher Scientific (PA, USA). All other chemicals used were of
reagent grade quality or higher.
2.3. Soil preparation
The clean natural soil was derived from Coleville site in Sas-
katchewan, Canada. Two soils, under-plant soil and ditch soil
which represent the predominant soil types of the site, were
chosen for evaluation in this study. The physical and textural
characteristics of these soils are given in Table 1. Soil samples
were dried and sieved through a 2-mm sieve to remove coarse
fragments. Contaminated soil was then prepared by dissolv-
ing an appropriate quantity of phenanthrene in dichloro-
methane and a known weight of soil was added with
continuous mixing. The resultant mixture was placed in
a ventilation hood to allow the complete evaporation of the
solvent. The contaminated soil was stored in a vessel at room
temperature for 7 days. Such soil had a final concentration of
100 mg/kg for phenanthrene and was used directly in the
batch experiments.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
Table 1 e Properties of the selected soils.
Sorbents Organic pH Sand Silt Clay Texture
matter (%) (H2O) (%) (%) (%)a
Under-plant soil 8.84 7.2 14.3 11.9 28.7 Sandy loam
Ditch soil 4.57 7.2 2.9 24.8 36.2 Clay loam
a Soil grain size classification is according to international crite-
rion: clay < 0.002 mm, silt < 0.02 mm, 0.02 mm < sand < 0.2 mm.
2.4. Desorption experiments
All the experiments were carried out in 25 mL glass centrifuge
tubes with Teflon-lined screw caps. The effects of bio-
surfactant and SCOAs on the behaviors of phenanthrene were
examined using the combinations of rhamnolipid and four
organic acids including acetic acid, oxalic acid, tartaric acid
and citric acid. For the desorption experiment, 0.5 g contam-
inated soil and 10 mL background solution. The background
solution was comprised of appropriate rhamnolipid and
organic acids, as well as 0.01 mol/L NaN3 as a biocide. System
pH was adjusted to 6 at a constant ion concentration. The
centrifuge tubes were vortexed for 20 s, and then placed on
a reciprocal shaker at 20  1  C and 125 rpm for 24 h to reach
the desorption equilibrium. Preliminary experiments showed
that 24 h were enough for the desorption of phenanthrene to
reach equilibrium, and the phenanthrene degradation or
adsorption by the tubes was negligible. The tubes were then
centrifuged at 5000 rpm for 25 min. Phenanthrene in the
aqueous phase were extracted with dichloromethane, and
their concentrations were analyzed by gas chromatography.
The amount of phenanthrene adsorbed to the soil was the
difference between the initial amount in the system and the
amount of the phenanthrene remaining in the solution. All
the tests were conducted in triplicate.
2.5. Effect of pH
The nature of biosurfactant and SCOAs can vary at different
pH, so combined effects of rhamnolipid and SCOAs under
different pH conditions should be studied. Prior to the test, pH
of background solution was adjusted to 3, 6 and 9 with
appropriate HCl and NaOH, respectively. The ion concentra-
tion in the system was kept constant by adding NaCl. Then
tests were performed in the same way as the batch desorption
experiments described above. The initial concentration for
each SCOA was 100 mmol/L, and the rhamnolipid concen-
trations were 50 mg/L.
2.6. Effect of DOM in the soil
To investigate the influence of soil dissolved organic matter
(DOM) on phenanthrene desorption in the presence of bio-
surfactant and SCOAs, the DOM in the soils was removed,
following the method as described by Cookson et al. (2005).
The treated soils were contaminated and mildly grounded to
pass through the sieve and homogenized. Then desorption
tests were done in the presence of rhamnolipid and different
SCOAs following the procedures above. The initial
5503
concentration for each SCOA was 100 mmol/L, and the
rhamnolipid concentrations were 50 mg/L.
2.7. Effect of ionic strength
To investigate the influence of ionic strength, the desorption
tests were done in the presence of rhamnolipid and different
SCOAs following the procedures detailed above. NaCl was
added at different concentrations (0.2, 0.3 and 0.4 mol/L). The
initial concentration for each SCOA was 100 mmol/L and
initial pH value was 6. When the pH level of the solution was
adjusted with NaOH or HCl, the increment of ionic strength
due to the pH-adjusting solution was corrected.
2.8. Analytical methods
The concentrations of phenanthrene were determined with
analysis by gas chromatography. The GC instrument, a Varian
GC 3800-FID system coupled with a Varian 8200 autosampler
(CA, USA), was equipped with a 25 m  0.32 mm ID (DB-5)
column with 0.25 mm film thicknesses (J&W Scientific Inc., CA).
Helium was used as the carrier gas (1.5 mL/min). The
temperature program was as follows: the oven temperature
was held at 40  C for 1.5 min then ramped to 175  C at a rate of
50  C/min. The temperature was held at 175  C for 1 min, then
ramped to 220  C at 7  C/min, and held at this final tempera-
ture for 1 min. The injector temperature was 250  C, and the
detector temperature was 230  C. Injection was made in the
split mode with a split ratio of 1:50 (since 1.75 min).
3. Results and discussion
3.1. Effect of biosurfactant on the phenanthrene
desorption
The effect of rhamnolipid on the desorption of phenanthrene
in soilewater system was quantified through measuring the
phenanthrene concentration in the bulk after desorption
equilibrium. Various biosurfactant concentrations were
applied to evaluate the performance of rhamnolipid in
removing phenanthrene from the contaminated soil. As Fig. 1
shows, in the absence of rhamnolipid, the amount of phen-
anthrene desorbed was 3.37 mg/kg for the under-plant soil
with an organic carbon content of 8.84%, and 7.98 mg/kg for
the ditch soil with an organic carbon content of 4.57%. The
amount of phenanthrene that could be desorbed in ditch soil
was greater than that in under-plant soil. In the presence of
rhamnolipid biosurfactant, the desorption of phenanthrene
by two soils was noticeably affected. The phenanthrene
removal amounts were of 8.76 and 12.24 mg/kg with 50 mg/L
rhamnolipid for under-plant and ditch soils, respectively. In
the presence of 100 and 200 mg/L rhamnolipid solution, the
results showed superior desorption efficiency and the des-
orbed amounts of phenanthrene were 12.05 and 13.12 mg/kg
for under-plant, respectively. The increase of rhamnolipid
concentration for ditch soil would result in a similar
enhancement of the phenanthrene desorption.
Desorption is the predominant process controlling move-
ment of HOCs through the soil matrix. The less desorbed
5504 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0

amount of phenanthrene observed in under-plant soil, which suitable for application for remediation and biostimulation of
has higher contents of organic matter and lower contents of PAH-polluted soil.
clay, was due to the stronger binding of phenanthrene to
under-plant soil. The removal efficiency often showed great
3.2. Effect of SCOAs on the phenanthrene desorption by
dependence on the soil composition and PAH properties (Zhou
biosurfactant
and Zhu, 2007). The different performance between these two
soils might be attributed mainly to their differences in soil
The desorption characteristics of phenanthrene by soils were
texture and organic contents.
assessed in the presence of rhamnolipid biosurfactant and
Biosurfactant concentration is a critical factor for the
four SCOAs. Fig. 2 illustrates the desorbed amount of phen-
removal of HOCs from soil. It was observed that increasing the
anthrene in the soilewater systems under different combi-
concentration of rhamnolipid could enhance phenanthrene
nations of agents. Performance with the combined treatments
desorption regardless of the soil used (Fig. 1). At concentra-
of rhamnolipid and organic acids exhibited an increase in
tions above the CMC, hydrophobic pollutants can readily
phenanthrene desorption compared to those that had
partition into the hydrophobic core at the center of the
received a treatment of rhamnolipid alone. The data in Fig. 2
micelle, thus increasing HOC aqueous concentration through
also indicate that desorption behaviors of phenanthrene in
micelle solubilization and promoting the desorption of HOCs
each soil in the presence of different biosurfactant-SCOA
from soils into aqueous phases (Pennell et al., 1997). In the
combinations are not similar. Among the combined treat-
present study, the rhamnolipid biosurfactant with concen-
ments of biosurfactant and SCOAs, the series with rhamno-
tration above 50 mg/L were applied to initiate the onset of
lipid and citric acid exhibited more significant effect on the
micellization and avoid the significant adsorption of bio-
desorption performance. The maximum removal of phenan-
surfactant (Mata-Sandoval et al., 2002; Cheng et al., 2004).
threne was 10.95 and 16.69 mg/kg for the combination of
Phenanthrene removed from soil in the presence of 100 and
rhamnolipid and citric acid with under-plant soil and ditch
200 mg/L rhamnolipid could achieve a greater desorption
soil, respectively.
efficiency when compared with that in the presence of 50 mg/
The rhamnolipid biosurfactant was still effective for the
L rhamnolipid. The results in this study are similar to previous
significant removal of phenanthrene from the soils in the
studies by Lai et al. (2009) who reported that increasing the
presence of SCOAs. The result indicates the presence of
concentration of rhamnolipid biosurfactant could enhance
SCOAs other than rhamnolipid is responsible for the different
the removal of total petroleum hydrocarbons from contami-
behaviors of phenanthrene desorption. The mechanism by
nated soil. However, the excessive biosurfactant addition
which rhamnolipid and SCOAs affected the desorption of
would make the remediation process less economically
phenanthrene was complex. Soil organic matter (SOM) func-
feasible. Furthermore, since the biodegradation of bio-
tions as an important partition medium in the retaining of
surfactant could compete with the biodegradation of PAHs,
HOCs (Rutherford et al., 1992). The organic matter can bond
too high concentration of biosurfactant would result in the
with the metals strongly in the soil through forming surface
retardation of contaminant bioremediation (Vipulanandan
complexes. The combined treatments with rhamnolipid and
and Ren, 2000). The potential sorption of biosurfactant on
citric acid showed better metal extraction performance than
soils, which can also show an adverse effect on remediation
individual application of rhamnolipid or citric acid by
efficiency, is mainly due to the influence of clays rather than
increasing ligand availability (Gunawardana et al., 2010). In
soil organic matters (Ochoa-Loza et al., 2007). Rhamnolipid
the presence of biosurfactant, therefore, the use of SCOAs
biosurfactant used in present study could facilitate the
might further enhance the release of these metals and disrupt
phenanthrene removal with high efficiency, thereby being
the linkage between organic matter and soil matrix.

16
Under-plant soil Ditch soil
14

12
Desorbed amount (mg/kg)

10

0
0 50 100 200
Rhamnolipid concentration (mg/L)

Fig. 1 e Effect of rhamnolipid biosurfactant on the Fig. 2 e Effect of SCOAs on the phenanthrene desorption by
phenanthrene desorption. biosurfactant.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0 5505

Phenanthrene may be released from the soil along with the were all higher than that where rhamnolipid had been applied
desorbed organic matter. This is in accordance with other alone. It can be seen that the desorbed phenanthrene
studies that the presence of organic acid was correlated with increased significantly with the increase of SCOA concentra-
the favorable desorption of organic pollutants (White et al., tion from 0 to 100 mmol/L, and then increased slightly when
2006; Zhang and Dong, 2008; An et al., 2010). Moreover, the further increasing the SCOA concentration from 100 to
rhamnolipid biosurfactant could lower the interfacial tension 300 mmol/L in most series. The desorbed amount with under-
against soil organic fractions in water, facilitating the trans- plant soil reached 9.25, 11.43, 10.89, 11.72 mg/kg in the pres-
location of organic matter and bounded contaminants into ence of 100 mg/L rhamnolipid and 300 mmol/L acetic acid,
aqueous phase. Then the surfactant molecule could bond with oxalic acid, tartaric acid and citric acid, respectively. A similar
the unbound phenanthrene compounds and dissolve them. trend was also found in other treatments with ditch soils.
Besides soil organic matter, mineral surfaces also play Additionally, the phenanthrene desorption with the combi-
a role in the distribution of PAHs, especially for the soil with nations of biosurfactant and different SCOAs showed
low organic matter content. SCOAs are known to interact with different increments when organic acid concentrations
inorganic soil particles via different mechanisms (Jones and increased. At the same concentration range of SCOA with
Brassington, 1998). The adsorption sites on the soil surface ditch soil, the desorbed phenanthrene with acetic acid varied
might be occupied by organic acids and thus the adsorption of from 12.29 to12.93 mg/kg, which was less remarkable than the
phenanthrene was hindered. In addition, the adsorption of series with rhamnolipid and other organic acids.
SCOAs could retard the adsorption of the biosurfactant on the It was observed that the removal of phenanthrene using
soil surface. More surfactant molecules, therefore, will be rhamnolipid and SCOAs gradually increased as the SCOA
diffused in the solution and possess a stronger capacity to concentration increased up to a concentration of 300 mmol/L.
desorb phenanthrene. Several effects discussed above could But the relative change in the amount of desorbed phenan-
all facilitate the desorption process of phenanthrene. threne in the presence of both biosurfactant and SCOAs
Enhanced phenanthrene availability in these combinations
could be attributed to the synergistic actions of rhamnolipid
and organic acids through potentially different modes of
action.
The observed removal efficiencies of phenanthrene from
soil were in the order of: rhamnolipid and citric
acid > rhamnolipid and oxalic acid > rhamnolipid and tartaric
acid > rhamnolipid and acetic acid. The differences are
mainly due to different chemical structures and properties of
organic acids. Among the SCOAs used, for example, citric acid
is a ternary organic acid and can provide more anions for
complexing than the other unary or binary acids. Such char-
acteristic could lead to the greater efficacy of citric acid in
desorbing metals and SOM from soils, carrying more adsorbed
phenanthrene molecules meanwhile. Furthermore, ions of
binary and ternary organic acids can form chelates with five-
or six-membered ring structure which are more stable (Qin
et al., 2004). Thus it can be seen the extent of increment was
less for the series with rhamnolipid and acetic acid than that
with rhamnolipid and other organic acids. In some other
studies, however, it was observed that oxalate acid and acetic
acid could enhance more desorption of organic and metal
contaminants, respectively (White et al., 2003; Maturi and
Reddy, 2008). Such difference is perhaps ascribed to the
presence of rhamnolipid biosurfactant, as well as the differ-
ences of contaminants and soil properties, which need to be
elucidated in further research.

3.3. Effect of SCOA concentrations on the phenanthrene

desorption in the presence of biosurfactant

The experiments were based on a series of batch tests with

same initial concentration of rhamnolipid but different
dosages of SCOAs. Considering the conditions in filed appli-
cations (Dermont et al., 2008), initial SCOA concentrations
ranged from 0 to 300 mmol/L. As shown in Fig. 3, when Fig. 3 e Effect of SCOA concentrations on the phenanthrene
rhamnolipid and SCOAs were applied in combination at desorption from (A) under-plant soil and (B) ditch soil in the
different ratios, the desorption percentage of phenanthrene presence of biosurfactant.
5506 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0

decreased with an increase in the concentration of SCOAs. As
the concentration of rhamnolipid was fixed, the enhancement
was most likely due to the variation of SCOA amount. At low
concentration of SCOAs, the effect of organic acids was not
significant and there was no obvious difference in the
desorption enhancement for various combinations. At this
stage, the enhanced desorption of phenanthrene could be
mainly ascribed to the performance of rhamnolipid bio-
surfactant. At higher concentration of SCOAs in the presence
of rhamnolipid, organics acids became more available to
destroy the barriers within soil matrix that normally func-
tioned to impede the translocation of organic contaminants.
More desorbed phenanthrene was observed at high SCOA
concentrations, meaning that the SCOA concentrations in the
presence of rhamnolipid at fixed concentration were posi-
tively correlated to desorbed amount of phenanthrene from
soil. In the absence of biosurfactant, it has been previously
noted that desorbed norfloxacin increased with increasing
concentrations of citric acid, malic acid and salicylic acid
(Zhang and Dong, 2008). In contrast, it was observed that
desorbed p,p0 -DDE decreased with organic acids above
50 mmol/L (White et al., 2003). Although the increased SCOA
concentrations were favorable for the enhanced desorption of
phenanthrene, the increments for each SCOA were different,
indicating the nature of organic acid is important for this
process. When the amount of organic acid was higher than
certain value, the low increasing rate of desorption was
possibly due to excess quantities of organic acids approaching
the maximum effectiveness to dissociate organic matter and
desorb contaminant from soil.
The results indicate that pH variation could influence the
transfer of phenanthrene from soil to aqueous phase in the
presence of surfactant and organic acids. The performance of
biosurfactant may be related quantitatively to the surface
tension and contaminant solubility, which are pH dependent
(Zhang and Miller, 1992; Shin et al., 2004). As for the SCOAs,
the possible existing forms of organic acids (dissociated and
undissociated) are also strongly related to pH conditions
(Jones and Brassington, 1998). Under the combined applica-
tion of rhamnolipid and SCOAs, the series at pH 3 could ach-
ieve a least desorption of phenanthrene when compared with
those at higher pH range. This was mainly due to limited
ability of rhamnolipid biosurfactant to enhance surface
activity and contaminant solubility at low pH. Also, less
organic acids could be dissociated from their protons and
existed as organic acid molecules. The maximum desorption
was achieved when the solution pH increased to 6. At this pH
range, both of the surface activity and contaminant solubility
with biosurfactant are near to the optimum conditions.
Organic acids could provide more ions, resulting in enhanced
complexing tendency (Qin et al., 2004).
Beyond pH 6, the increase of pH would result in a slight
decrease for the desorption efficiency of phenanthrene in
A 12
pH 3 pH 6 pH 9
10
Desorbed amount (mg/kg)

3.4. Effect of pH on the phenanthrene desorption in the 6

presence of biosurfactant and SCOAs

4
The effect of pH on the desorption of phenanthrene in the
presence of rhamnolipid and SCOAs was studied by con- 2
ducting batch studies at various pH between 3 and 9. Fig. 4
depicts the amount of initial PAHs desorbed from soil into
0
aqueous phase. The data indicates that the desorption Rhamnolipid Rhamnolipid Rhamnolipid Rhamnolipid Rhamnolipid
behavior of phenanthrene in two soils could be affected by the + Acetic acid + Oxalic Acid + Tartaric Acid + Citric Acid
pH conditions in the system. In the presence of biosurfactant B 20
pH 3 pH 6 pH 9
and SCOAs, the extent of phenanthrene desorption was rela-
tively lower at low pH range. At pH 3, the desorption amount
was from 4.64 to 5.63 mg/kg for different series with under- 15
D esorbed amount (mg/kg)

plant soil. Then the desorption increased sharply from pH 3

to pH 6, but decreased little with the change in pH from pH 6 to
pH 9. With the ditch soil, the phenanthrene desorption was
10
also less pronounced at pH 3. Above pH 3, similar trend was
observed except that the addition of oxalic, tartaric and citric
acids in the presence of rhamnolipid appeared to slightly
5
enhance the desorption of phenanthrene at pH 9 compared
with that at pH 6. The amounts of phenanthrene desorbed at
pH 9 were 16.34, 15.57, and 17.89 mg/kg in the presence of
rhamnolipid along with oxalic, tartaric and citric acids, 0
respectively. At the same pH value, desorption amounts of Rhamnolipid Rhamnolipid Rhamnolipid Rhamnolipid Rhamnolipid
+ Acetic acid + Oxalic Acid + Tartaric Acid + Citric Acid
phenanthrene differentiate for varied combinations. The use
of rhamnolipid and citric acid showed better desorption Fig. 4 e Effect of pH on the phenanthrene desorption from
performance than the combinations of rhamnolipid and other (A) under-plant soil and (B) ditch soil in the presence of
organic acids. biosurfactant and SCOAs.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
most tests. Although the individual performance of rhamno-
lipid could be limited at pH above 6, the removal of phenan-
threne still reached more than 7.61 and 12.19 mg/kg with
combined use of rhamnolipid and SCOAs in our experiment
with under-plant and ditch soil, respectively. The high
removal efficiency at high pH could be attributable to facili-
tation by organic acids. Moreover, the system at high pH might
lead to increased polarity of the organic material, showing
lower affinity for hydrophobic compounds (Ping et al., 2006).
At the same time, increases in pH could be accomplished with
increased soil-particle dispersion, as well as more dis-
associated soil minerals and organic matter (You et al., 1999).
These factors will contribute to the strong desorption even at
high pH range. Furthermore, since soil is negatively charged at
near-neutral or higher pH, rhamnolipid, present in anionic
forms, are expected to sorb minimally to the particle surface.
For both under-plant and ditch soils, elevated pH in
combination with rhamnolipid and citric acid had the largest
removal efficiency. The variations of phenanthrene desorp-
tion at different pH might relate to the different dissociation
constants of organic acids involved. The complexing ability of
organic acid will increase with enhanced pH, especially for the
polyprotic weak acid such as citric acid. The different phen-
anthrene desorption performances at varied pH may be
attributed to effects of biosurfactant and organic acids, as well
as the soil constitutes and other system conditions. The
relative importance of the different mechanisms depends on
the physical and chemical properties of the interactive sor-
bateesorbent system. Furthermore, based on our results, the
optimum pH values for desorption was relatively close to that
for microbial growth. It is thus important to also include such
environmental concern of pH for in situ application.
3.5. Effect of soil DOM on the phenanthrene desorption
in the presence of biosurfactant and SCOAs
The desorption amounts of phenanthrene in the presence of
biosurfactant and SCOAs with DOM-removed soils are illus-
trated in Fig. 5. When compared with Fig. 2, the results showed
that the phenanthrene desorption differentiated significantly
between the treated and untreated soils. In general, the des-
orbed amounts of phenanthrene for treated soils were less
than those for untreated soils. Without the amendments of
SCOAs, addition of rhamnolipid enhanced the mobilization of
phenanthrene with the desorption of 5.98 and 9.36 mg/kg for
DOM-removed under-plant and ditch soils, respectively. The
desorption capacities were greater in the presence of both
rhamnolipid of SCOAs, irrespective of whether the soil DOM
was removed. In the series with rhamnolipid and different
organic acids, the desorbed amounts of phenanthrene in the
soilewater systems were from 6.13 to 7.66 mg/kg for treated
under-plant soils, which were lower than the values of cor-
responding untreated soils. Similar trend was observed for
different combinations of biosurfactant and SCOAs with
treated ditch soils, except the phenanthrene desorption was
greater in the series with rhamnolipid and tartaric acid as
compared with the tests with rhamnolipid and oxalic acid.
The desorption of phenanthrene was adversely affected
with the use of DOM-removed soil. This difference indicates
that some soil properties other than biosurfactant and organic
5507
Fig. 5 e Effect of soil DOM on the phenanthrene desorption
in the presence of biosurfactant and SCOAs.
acids can be responsible for the changes of phenanthrene
desorption. This result is similar to previous study where
adsorption of phenanthrene on soil increased after removal of
DOM (Luo et al., 2008). The DOM in soil could exhibit a high
affinity for HOCs and enhance the dissolution of organic
compounds in soilewater system (Pignatello et al., 2006). With
a decreased quantity of DOM in soil, the concentration of DOM
in solution decreased so that less phenanthrene was trans-
ported into aqueous phase. Binding affinity of organic
contaminates to the sediment increased after the removal of
soil DOM. In contrast, if all the organic matter is removed,
organic contaminants would deposit mainly on mineral
surface and exhibits a stronger desorbing capacity (Gao et al.,
1998).
Although most DOM has been removed, larger amounts of
phenanthrene were desorbed from soil in the soilewater
systems containing both rhamnolipid and SCOAs than those
from soil in the systems containing only rhamnolipid. This is
in part due to the mobilization of phenanthrene by rhamno-
lipid biosurfactant, but also due to the influence of the organic
acids on remaining organo-mineral linkages, all of which
operate simultaneously. The variation of desorption
percentage with different combinations is closely related to
the speciation of the organic acids. The lower increases in
phenanthrene desorption with rhamnolipid and acetic acid
amendments are possibly due to weak complexing ability of
organic acid, restricting the diffusion of resident contami-
nants in soils. These observations suggested that the soil DOM
was a critical factor affecting the desorption processes by
biosurfactant in presence of organic acids.
3.6. Effect of ionic strength on the phenanthrene
desorption in the presence of biosurfactant and SCOAs
The phenanthrene desorption influenced by combined use of
rhamnolipid and SCOA at different ionic-strength conditions
is shown in Fig. 6. The desorption of phenanthrene in the
presence of rhamnolipid and different organic acids
decreased with increasing concentrations of the ions. At the
ionic strength of 0.3 mol/L, the phenanthrene desorption
5508 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
ranged from 7.92 to 10.16 mg/kg and 10.64e14.30 mg/kg for
different combinations with under-plant and ditch soils,
respectively. Compared with the tests without any amend-
ments, the desorption amount of phenanthrene was also
elevated with the addition of biosurfactant and SCOAs at
higher ionic-strength range. However, the increase was not
very significant as noticed at lower ionic strength. At the ionic
strength of 0.4 M NaCl, the phenanthrene desorption effi-
ciency with different combinations was from 6.52 to 9.10 and
8.45e13.77 mg/kg for under-plant and ditch soil, respectively.
At the same range of NaCl concentration, ditch soil followed
by under-plant soil, showed the greater desorption capacities
with different combinations of rhamnolipid and SCOAs.
Organic acids often coexist with metal ions in natural
environment. The above results indicates that the variation of
system ionic strength is likely to influence the overall phen-
anthrene desorption efficiency with the addition of rhamno-
lipid and SCOAs. A recent study reported that the surface
activity of rhamnolipid was not negatively affected to any
degree by high salinities (Abdel-Mawgoud et al., 2009). The
size of both mono- and dirhamnolipid biosurfactant vesicles
were reduced after the addition of NaCl, thereby improving
the water solubility of the biosurfactant (Pornsunthorntawee
et al., 2009). However, the findings of present study show
Fig. 6 e Effect of ionic strength on the phenanthrene
desorption from (A) under-plant soil and (B) ditch soil in the
presence of biosurfactant and SCOAs.
that the high ionic strength values were correlated with the
low mobility of phenanthrene. In the solution with high ionic
strength, solubility of HOCs might decrease due to the salt-out
effect process (Brunk et al., 1996). It was also postulated that
PAHs could be less partitioned to dissolved organic substances
in the presence of high-concentration metals (Saison et al.,
2004). The presence of more salt ions, therefore, can
decrease the diffusion of phenanthrene contaminants from
soil into aqueous phase. In addition, certain SCOAs could
enhance the removal efficiency for phenanthrene which
contains rhamnolipid biosurfactant and NaCl at high
concentrations.
The metal ions showed less inhibitory effect on phenan-
threne desorption in biosurfactant-enhanced process when
the organic acids were amended. This was of special interest
for application to in situ soil washing with high salinity level.
Among the different combinations at the same ionic strength,
the maximum desorption for phenanthrene was observed in
the series with rhamnolipid and citric acid. In the presence of
NaCl, the desorption of phenanthrene by two soils was still
different. Consequently, other factors, like soil fractions and
organic acid characteristics, might also be important for
deciding the combined role of biosurfactant and SCOAs in
contaminant behaviors within metal ion enrichment.
4. Conclusions
The study investigated the effect of SCOAs on the desorption
of phenanthrene in the presence of rhamnolipid bio-
surfactant. The results supported the combined use of bio-
surfactant and SCOAs could further enhance the desorption of
phenanthrene from soil into aqueous phase. The overall
desorption behavior of phenanthrene could be assumed to be
a sum of a large number of individual desorption events with
different mechanisms. Desorption amounts of phenanthrene
differentiated for varied combinations, in which the effect of
ternary and binary SCOAs were more remarkable. SCOAs at
high concentration caused a more significant increase in
desorption from the soils with the amendment of bio-
surfactant. The quantity and species of organic acids could
affect the tendency of phenanthrene distribution in the
presence of biosurfactant. Furthermore, the phenanthrene
desorption was higher in the ditch soil compared with the
under-plant soil. The variation of desorption with different
combinations was also closely related with the soil DOM.
Soil property might be an important factor influencing the
mobility behaviors of phenanthrene with combined use of
biosurfactant and organic acids. In addition, adjusting the pH
and ionic strength could control the effect of biosurfactant
and organic acids, allowing higher remediation efficiency at
specific values. The results presented here demonstrated that
the effect of SCOAs on the biosurfactant-enhanced desorption
might be directly influenced by the solution chemistry. The
results suggest that the combined application of rhamnolipid
and SCOAs can be regarded as a good alternative for the
removal of PAH compounds from contaminated soil. Further
study is also needed to help obtain more theoretical founda-
tion for the PAH mobility with additional biosurfactant and
organic acids.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
Acknowledgments
This research was supported by the Major Science and Tech-
nology Program for Water Pollution Control and Treatment,
the Canadian Water Network under the Networks of Centers
of Excellence (NCE), and the Natural Science and Engineering
Research Council of Canada. We are also grateful for anony-
mous reviewers for their helpful suggestions and advices.
references
Abdel-Mawgoud, A., Aboulwafa, M., Hassouna, N., 2009.
Characterization of rhamnolipid produced by Pseudomonas
aeruginosa isolate Bs20. Applied Biochemistry and
Biotechnology 157 (2), 329e345.
Alcantara-Garduno, M.E., Okuda, T., Tsai, T.Y., Nishijima, W.,
Okada, M., 2008. Experimental and mathematical evaluation
of trichloroethylene removal from saturated soil using acetic
acid with saturated ozone. Separation and Purification
Technology 60 (3), 299e307.
An, C.J., Huang, G.H., Yu, H., Wei, J., Chen, W., Li, G.C., 2010. Effect
of short-chain organic acids and pH on the behaviors of
pyrene in soilewater system. Chemosphere 81 (11),
1423e1429.
Blank, R.R., Allen, F., Young, J.A., 1994. Extractable anions in soils
following wildfire in a sagebrush-grass community. Soil
Science Society of America Journal 58 (2), 564e570.
Bordas, F., Lafrance, P., Villemur, R., 2005. Conditions for effective
removal of pyrene from an artificially contaminated soil using
Pseudomonas aeruginosa 57SJ rhamnolipids. Environmental
Pollution 138 (1), 69e76.
Brunk, B.K., Jirka, G.H., Lion, L.W., 1996. Effects of salinity changes
and the formation of dissolved organic matter coatings on the
sorption of phenanthrene: implications for pollutant trapping
in estuaries. Environmental Science & Technology 31 (1),
119e125.
Cheng, K.Y., Zhao, Z.Y., Wong, J.W.C., 2004. Solubilization and
desorption of PAHs in soil-aqueous system by biosurfactants
produced from Pseudomonas aeruginosa P-CG3 under
thermophilic condition. Environmental Technology 25 (10),
1159e1165.
Chu, W., Chan, K.H., 2003. The mechanism of the surfactant-
aided soil washing system for hydrophobic and partial
hydrophobic organics. Science of the Total Environment 307
(1e3), 83e92.
Clifford, J.S., Ioannidis, M.A., Legge, R.L., 2007. Enhanced aqueous
solubilization of tetrachloroethylene by a rhamnolipid
biosurfactant. Journal of Colloid and Interface Science 305 (2),
361e365.
Cookson, W.R., Abaye, D.A., Marschner, P., Murphy, D.V.,
Stockdale, E.A., Goulding, K.W.T., 2005. The contribution of
soil organic matter fractions to carbon and nitrogen
mineralization and microbial community size and structure.
Soil Biology and Biochemistry 37 (9), 1726e1737.
Dermont, G., Bergeron, M., Mercier, G., Richer-Laflèche, M., 2008.
Soil washing for metal removal: a review of physical/chemical
technologies and field applications. Journal of Hazardous
Materials 152 (1), 1e31.
Deshpande, S., Shiau, B.J., Wade, D., Sabatini, D.A., Harwell, J.H.,
1999. Surfactant selection for enhancing ex situ soil washing.
Water Research 33 (2), 351e360.
Deshpande, S., Wesson, L., Wade, D., Sabatini, D.A., Harwell, J.H.,
2000. DOWFAX surfactant components for enhancing
contaminant solubilization. Water Research 34 (3), 1030e1036.
5509
Djomo, J.E., Dauta, A., Ferrier, V., Narbonne, J.F., Monkiedje, A.,
Njine, T., Garrigues, P., 2004. Toxic effects of some major
polyaromatic hydrocarbons found in crude oil and aquatic
sediments on Scenedesmus subspicatus. Water Research 38 (7),
1817e1821.
Ferrarese, E., Andreottola, G., Oprea, I.A., 2008. Remediation of
PAH-contaminated sediments by chemical oxidation. Journal
of Hazardous Materials 152 (1), 128e139.
Gan, S., Lau, E.V., Ng, H.K., 2009. Remediation of soils
contaminated with polycyclic aromatic hydrocarbons (PAHs).
Journal of Hazardous Materials 172 (2e3), 532e549.
Gao, J.P., Maguhn, J., Spitzauer, P., Kettrup, A., 1998. Sorption of
pesticides in the sediment of the Teufelsweiher pond
(Southern Germany). I: equilibrium assessments, effect of
organic carbon content and PH. Water Research 32 (5),
1662e1672.
Gomez, J., Alcantara, M.T., Pazos, M., Sanroman, M.A., 2010.
Remediation of polluted soil by a two-stage treatment system:
desorption of phenanthrene in soil and electrochemical
treatment to recover the extraction agent. Journal of
Hazardous Materials 173 (1e3), 794e798.
Gunawardana, B., Singhal, N., Johnson, A., 2010. Amendments
and their combined application for enhanced copper,
cadmium, lead uptake by Lolium perenne. Plant and Soil 329
(1e2), 283e294.
Jones, D.L., 1998. Organic acids in the rhizosphere e a critical
review. Plant and Soil 205 (1), 25e44.
Jones, D.L., Brassington, D.S., 1998. Sorption of organic acids in
acid soils and its implications in the rhizosphere. European
Journal of Soil Science 49 (3), 447e455.
Kanaly, R.A., Harayama, S., 2000. Biodegradation of high-
molecular-weight polycyclic aromatic hydrocarbons by
bacteria. Journal of Bacteriology 182 (8), 2059e2067.
Kanga, S.A., Bonner, J.S., Page, C.A., Mllls, M.A., Autenrieth, R.L.,
1997. Solubilization of naphthalene and methyl-substituted
naphthalenes from crude oil using biosurfactants.
Environmental Science & Technology 31 (2), 556e561.
Khodadoust, A.P., Bagchi, R., Suidan, M.T., Brenner, R.C.,
Sellers, N.G., 2000. Removal of PAHs from highly
contaminated soils found at prior manufactured gas
operations. Journal of Hazardous Materials 80 (1e3), 159e174.
Lai, C.C., Huang, Y.C., Wei, Y.H., Chang, J.-S., 2009. Biosurfactant-
enhanced removal of total petroleum hydrocarbons from
contaminated soil. Journal of Hazardous Materials 167 (1e3),
609e614.
Luo, L., Zhang, S.Z., Ma, Y.B., 2008. Evaluation of impacts of soil
fractions on phenanthrene sorption. Chemosphere 72 (6),
891e896.
Mata-Sandoval, J.C., Karns, J., Torrents, A., 2002. Influence of
rhamnolipids and Triton X-100 on the desorption of pesticides
from soils. Environmental Science & Technology 36 (21),
4669e4675.
Maturi, K., Reddy, K.R., 2008. Extractants for the removal of mixed
contaminants from soils. Soil & Sediment Contamination 17
(6), 586e608.
Mulligan, C.N., 2005. Environmental applications for
biosurfactants. Environmental Pollution 133 (2), 183e198.
Nadal, M., Schuhmacher, M., Domingo, J.L., 2004. Levels of PAHs
in soil and vegetation samples from Tarragona County, Spain.
Environmental Pollution 132 (1), 1e11.
Noordman, W.H., Ji, W., Brusseau, M.L., Janssen, D.B., 1998.
Effects of rhamnolipid biosurfactants on removal of
phenanthrene from soil. Environmental Science & Technology
32 (12), 1806e1812.
Ochoa-Loza, F.J., Noordman, W.H., Jannsen, D.B., Brusseau, M.L.,
Maier, R.M., 2007. Effect of clays, metal oxides, and organic
matter on rhamnolipid biosurfactant sorption by soil.
Chemosphere 66 (9), 1634e1642.
5510 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 0 1 e5 5 1 0
Pennell, K.D., Adinolfi, A.M., Abriola, L.M., Diallo, M.S., 1997.
Solubilization of dodecane, tetrachloroethylene, and 1,2-
dichlorobenzene in micellar solutions of ethoxylated nonionic
surfactants. Environmental Science & Technology 31 (5),
1382e1389.
Pignatello, J.J., Kwon, S., Lu, Y., 2006. Effect of natural organic
substances on the surface and adsorptive properties of
environmental black carbon (char): attenuation of surface
activity by humic and fulvic acids. Environmental Science &
Technology 40 (24), 7757e7763.
Ping, L.F., Luo, Y.M., Wu, L.H., Qian, W., Song, J., Christie, P., 2006.
Phenanthrene adsorption by soils treated with humic
substances under different pH and temperature conditions.
Environmental Geochemistry and Health 28 (1e2), 189e195.
Pornsunthorntawee, O., Chavadej, S., Rujiravanit, R., 2009.
Solution properties and vesicle formation of rhamnolipid
biosurfactants produced by Pseudomonas aeruginosa SP4.
Colloids and Surfaces B e Biointerfaces 72 (1), 6e15.
Qin, F., Shan, X.Q., Wei, B., 2004. Effects of low-molecular-weight
organic acids and residence time on desorption of Cu, Cd, and
Pb from soils. Chemosphere 57 (4), 253e263.
Rutherford, D.W., Chiou, C.T., Kile, D.E., 1992. Influence of soil
organic matter composition on the partition of organic
compounds. Environmental Science & Technology 26 (2),
336e340.
Saison, C., Perrin-Ganier, C., Amellal, S., Morel, J.L., Schiavon, M.,
2004. Effect of metals on the adsorption and extractability of
14
C-phenanthrene in soils. Chemosphere 55 (3), 477e485.
Shin, K.H., Kim, K.W., Seagren, E.A., 2004. Combined effects of pH
and biosurfactant addition on solubilization and
biodegradation of phenanthrene. Applied Microbiology and
Biotechnology 65 (3), 336e343.
Silva, A., Delerue-Matos, C., Fiúza, A., 2005. Use of solvent
extraction to remediate soils contaminated with
hydrocarbons. Journal of Hazardous Materials 124 (1e3),
224e229.
van Hees, P.A.W., Vinogradoff, S.I., Edwards, A.C., Godbold, D.L.,
Jones, D.L., 2003. Low molecular weight organic acid
adsorption in forest soils: effects on soil solution
concentrations and biodegradation rates. Soil Biology and
Biochemistry 35 (8), 1015e1026.
Vipulanandan, C., Ren, X.P., 2000. Enhanced solubility and
biodegradation of naphthalene with biosurfactant. Journal of
Environmental Engineering-ASCE 126 (7), 629e634.
Wang, Y.J., Chen, J.H., Cui, Y.X., Wang, S.Q., Zhou, D.M., 2009.
Effects of low-molecular-weight organic acids on Cu(II)
adsorption onto hydroxyapatite nanoparticles. Journal of
Hazardous Materials 162 (2e3), 1135e1140.
Wei, J., Huang, G.H., Yu, H., An, C.J., 2011. Efficiency of single and
mixed gemini/conventional micelles on solubilization of
phenanthrene. Chemical Engineering Journal 168 (1), 201e207.
Wei, Q.F., Mather, R.R., Fotheringham, A.F., 2005. Oil removal
from used sorbents using a biosurfactant. Bioresource
Technology 96 (3), 331e334.
White, J.C., Mattina, M.I., Lee, W.Y., Eitzer, B.D., Iannucci-
Berger, W., 2003. Role of organic acids in enhancing the
desorption and uptake of weathered p, p0 -DDE by Cucurbita
pepo. Environmental Pollution 124 (1), 71e80.
White, J.C., Parrish, Z.D., Isleyen, M., Gent, M.P.N., Iannucci-
Berger, W., Eitzer, B.D., Kelsey, J.W., Mattina, M.I., 2006.
Influence of citric acid amendments on the availability of
weathered PCBs to plant and earthworm species.
International Journal of Phytoremediation 8 (1), 63e79.
You, S.J., Yin, Y., Allen, H.E., 1999. Partitioning of organic matter
in soils: effects of pH and water/soil ratio. The Science of the
Total Environment 227 (2e3), 155e160.
Yu, H., Huang, G.H., Wei, J., An, C.J., 2011. Solubilization of mixed
polycyclic aromatic hydrocarbons through a rhamnolipid
biosurfactant. Journal of Environmental Quality 40 (2), 477e483.
Yuan, S.H., Xi, Z.M., Jiang, Y., Wan, J.Z., Wu, C., Zheng, Z.H., Lu, X.
H., 2007. Desorption of copper and cadmium from soils
enhanced by organic acids. Chemosphere 68 (7), 1289e1297.
Zhang, J.Q., Dong, Y.H., 2008. Effect of low-molecular-weight
organic acids on the adsorption of norfloxacin in typical
variable charge soils of China. Journal of Hazardous Materials
151 (2e3), 833e839.
Zhang, W., Tsang, D.C.W., Lo, I.M.C., 2008. Removal of Pb by
EDTA-washing in the presence of hydrophobic organic
contaminants or anionic surfactant. Journal of Hazardous
Materials 155 (3), 433e439.
Zhang, Y., Miller, R.M., 1992. Enhanced octadecane dispersion and
biodegradation by a Pseudomonas rhamnolipid surfactant
(biosurfactant). Applied and Environmental Microbiology 58
(10), 3276e3282.
Zhou, W.J., Zhu, L.Z., 2007. Efficiency of surfactant-enhanced
desorption for contaminated soils depending on the
component characteristics of soil-surfactant- PAHs system.
Environmental Pollution 147 (1), 66e73.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Nitrification and potential control mechanisms in simulated

premises plumbing

Mohammad Shahedur Rahman a,1, Gem Encarnacion a,b, Anne K. Camper a,c,*
a
Center for Biofilm Engineering, Montana State University, 366 EPS Building, Bozeman, MT 59717, USA
b
Department of Microbiology, Montana State University, Bozeman, MT 59717, USA
c
Department of Civil Engineering, Montana State University, Bozeman, MT 59717, USA

article info abstract

Article history: Indigenous drinking water organisms were used to establish nitrification in glass reactors
Received 7 February 2011 containing copper or polyvinyl chloride (PVC) surfaces. The reactors were fed soil-derived
Received in revised form humics as the organic carbon source and ammonium sulfate as the nitrogen source in
19 July 2011 biologically treated tap water. Water in the reactors was stagnant for 8 h and then flowed
Accepted 7 August 2011 for 5 min to simulate conditions in household plumbing. Following the establishment of
Available online 16 August 2011 complete nitrification (conversion of ammonia to nitrate) in both reactor types, various
inhibitors of nitrification were tested followed by a period where recovery of nitrification
Keywords: was observed. In one PVC reactor, copper was gradually introduced up to 1.3 ppm. To
Nitrification ensure that most of the copper was in the ionic form, the pH of the influent was then
Biofilms gradually lowered to 6.6. No significant change in nitrification was observed in the pres-
Chlorite ence of copper. Chlorite was introduced into copper and PVC reactors at doses increasing
Chloramines from 0.2 ppm to 20 ppm. There was limited effect on the PVC system and inhibition in the
Copper copper reactor only at 20 ppm. Chloramine was tested at chlorine to ammonia ratios
Plumbing ranging from 0.5:1 to 5:1. Nitrification activity was impacted significantly at a 5:1 ratio and
ultimately stopped, with the fastest response being in the copper system. Whenever
a control mechanism was tested, there was increased release of copper from the reactors
with copper coupons. In all cases, nitrification recovered when inhibitors were removed
but the rates of recovery differed depending on the treatment method and coupon surface.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction disinfectant can reduce DBPs, there is the chance that

A topic of regulatory and public health concern in drinking
water is the creation of potentially carcinogenic disinfection
by-products (DBPs) when the water is chlorinated. As an
alternative to chlorination, many utilities have shifted to the
use of chloramines to reduce the levels of regulated DBPs to
meet the Stage 2 Disinfectants/Disinfection By-Product Rule
(USEPA, 2000). Although the use of chloramines as a secondary
increased levels of free ammonia in the distribution system
may serve as an energy source for indigenous nitrifying
organisms. Proliferation of these organisms can then result in
nitrification in the distribution system. Not surprisingly,
nitrification is one of the most frequent operational problems
encountered by drinking water utilities that use chloramine
for secondary disinfection (Skadsen, 1993; Wolfe and Lieu,
2001; Odell et al., 1996; Wilczak et al., 1996; Seidel et al., 2005).

* Corresponding author. Center for Biofilm Engineering, Montana State University, 366 EPS Building, Bozeman, MT 59717, USA.
Tel.: þ1 406 994 4906; fax: þ1 406 994 6098.
E-mail address: anne_c@biofilm.montana.edu (A.K. Camper).
1
Current address: Anderson Civil Consultants Nanaimo, BC V9R 2T5, Canada.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.009
5512 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
Nitrification is a microbial process by which reduced
nitrogen compounds (primarily ammonia) are sequentially
oxidized to nitrite and nitrate. This process can have a detri-
mental impact on water quality. Due to nitrification, chlora-
mine residual, pH, alkalinity and dissolved oxygen of water
decrease. As nitrification can cause a decrease in pH, some
utilities may be susceptible to elevated levels of soluble metal
contaminants such as lead (Zhang et al., 2009), leading to Lead
and Copper Rule (LCR) violations. Nitrification can also cause
biological instability through production of soluble microbial
products which may support the growth of heterotrophic
bacteria in low nutrient environments (Rittmann et al., 1994).
Studies on nitrification in drinking water have mostly been
done in distribution mains or at the treatment plant level and
studies on premises plumbing are lacking. Premises plumbing
not only has higher surface to volume ratios but also has
about 10 times more length than water mains (NRC, 2006). In
addition, it stands to reason that favorable conditions for
nitrification such as low or no disinfectant, long water age and
warmer temperatures exist in premises plumbing. This gap in
knowledge led to investigations on possible control for nitri-
fication in premises plumbing. The studies took place in
laboratory reactors designed to simulate the surface area to
volume ratios, flow conditions, and water quality that may be
encountered in these systems. Ammonia concentrations
chosen represented a worst case scenario where all the
chloramine added at the regulatory limit (4 ppm) had decayed
to release ammonia. Two commonly utilized premises
plumbing materials, copper and polyvinyl chloride (PVC), were
used. Copper is the most widely used metal for household
plumbing systems and more than 90% of domestic plumbing
material in the US is made of copper (Oskarsson and Norrgren,
1998) while PVC pipes are also a very common plumbing
material (NSF, 2008).
Three separate strategies based on realistic approaches for
drinking water systems were evaluated. First, the inhibitory
effect of copper was examined by (1) comparing nitrification in
PVC vs copper reactors and (2) by introducing known amounts
of copper (gradually increasing concentration) into a nitrifying
reactor with PVC coupons. Previous work has shown that
nitrification in pure cultures is either enhanced or inhibited by
copper, depending on its concentration. Loveless and Painter
(1968) found that 0.005e0.03 ppm of Cuþ2 stimulated the
growth of the ammonia oxidizer Nitrosomonas while Skinner
and Walker (1961) observed enhanced growth at higher
concentrations of 0.1e0.5 ppm copper. However, these higher
concentrations were found to be inhibitory by Loveless and
Painter (1968). Zhang and Edwards (2005) observed slight
inhibition of nitrification for pure cultures in the presence of
5 ppb copper while 25 ppb copper had a slightly stimulatory
effect. In the same study, a much higher concentration of
500 ppb copper significantly inhibited nitrification. Zhang and
Edwards (2010) also reported inhibition of nitrification at
copper levels greater than 100 mg/L. In cases where there is
free ammonia, coppereammonia complexes such as [Cu
þ2
(NH3)2þX ], (Sato et al., 1988) and cupric tetraamine [Cu(NH3)4 ]
(Lee et al., 1997) may also be responsible for inhibition of
nitrifiers. It is also important to know the form in which the
copper exists because Cuþ2 ions are believed to be responsible
for the inhibition of nitrifying bacteria (Braam and Klapwijk,
1981; Hu et al., 2003). Cupric ions in the vicinity of the cell
membrane may cause damage by depolarization and impair-
ment of receptors or transporter molecules (Alt et al., 1990),
and may bind proteins and block the function of proteins in
the exoploymeric substance (EPS) of the bacteria (Geesey and
Jang, 1989). Ion speciation can be controlled by the pH (Braam
and Klapwijk, 1981; Edwards et al., 1996), and this was
considered as part of the experimental design.
A second set of experiments was done to determine the
impact of chlorite on nitrification. The use of chlorite as
a control mechanism for nitrification has been proposed for
full scale distribution systems and storage tanks. Hynes and
Knowles (1983) showed that chlorite interfered with the first
step of nitrification, specifically in the oxidation of ammonia to
nitrite by Nitrosomonas europea. Several studies in both labo-
ratory and full scale drinking water distribution systems have
been conducted to investigate the effect of chlorite on nitrifi-
cation. McGuire et al. (1999) found that low levels (0.2 ppm) of
chlorite caused a significant reduction in the culturability of
ammonia oxidizing bacteria (AOB). In the same study, the
experience of the Gulf Coast (Texas) Water Authority (GCWA)
which uses chlorine dioxide as the primary disinfectant and
chloramine as the secondary disinfectant was reported.
Chlorite was detected in their distribution system in the range
of 0.25e0.35 ppm and although the conditions were conducive
to nitrification, it did not occur suggesting the inhibitory effect
of chlorite to nitrification. In another study conducted in plug
flow reactors in Tucson AZ, continuous feed of chlorite at
concentrations as low as 0.1 ppm was found to prevent nitri-
fication (McGuire et al., 1999). Chlorite was also used to prevent
nitrification in parts of the Glendale, CA distribution system
(McGuire et al., 2009). Interestingly, chlorite has been found to
be ineffective in controlling nitrification in other studies.
McGuire et al. (1999) reported a nitrification episode in the
Corpus Christi Texas distribution system that continued even
after dosing with chlorite. Similarly, Karim and LeChevallier
(2006) reported the recurrence of nitrification in a pilot
system where initially the use of 0.5 ppm chlorite controlled
nitrification. In another study the presence of chlorite in water
reservoirs prevented the onset of nitrification, but once nitri-
fication started, introducing chlorite was not effective and
a 0.2 ppm dose of chlorite in a nitrifying reservoir inhibited
nitrification for only two weeks (McGuire et al., 2006). To
investigate the impact of chlorite on nitrification in the simu-
lated premises plumbing system, doses were incrementally
increased from 0.2 to 20 ppm.
A third potential control mechanism was maintaining
a chloramine disinfectant residual within the plumbing
system. Chloramine, although considered as a weaker disin-
fectant than chlorine for suspended cells, is thought to be
more effective for disinfecting biofilms (LeChevallier et al.,
1988; Wahman et al., 2009), and since most nitrifying
bacteria are present as biofilms rather than planktonic cells in
both natural and engineered systems (Schramm et al., 1996;
Okabe et al., 2005; Kindaichi et al., 2006), the use of chlora-
mine was also examined. In an actual water distribution
system, it may be possible to either distribute water with
a stable residual or recreate chloramine through booster
chlorination. To investigate this experimentally, chlorine was
applied to attain different chlorine to ammonia ratios.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
All of these experiments took place in reactors that had
been undergoing complete nitrification (conversion of
ammonia to nitrate) for one year. Importantly, the reactors
had undefined mixed population biofilms originating from
Bozeman tap water and were not inoculated with specific
nitrifying organisms. The operating conditions were carefully
chosen to represent conditions in premises plumbing while
allowing for meaningful sampling strategies and control of
variables. During each test, effluent copper concentrations
were measured to assess the influence of that strategy on
copper release. As such, the results provide insights on how
indigenous nitrifying biofilm communities respond to poten-
tial control strategies and how these strategies influence
effluent water quality.
2. Materials and methods
2.1. Reactors
Domestic plumbing systems were simulated using a modified
CDC reactor (Goeres et al., 2005). Modifications included add-
ing parallel coupons to solid rods, adding a base plate, and
changing the stir blades to either polyvinyl chloride (PVC) or
copper (Fig. 1). The surface area of the coupons, base plate and
stir blades was calculated to create the same surface to
volume ratio as that of a six foot long 3/4 ” diameter domestic
plumbing pipe. The PVC or copper was washed with 0.1 N
NaOH three times prior to use. Volume of the reactors is
120 ml.
Fig. 1 e Modified CDC reactor showing the paired copper
coupons (1.5 3 1.7 cm inside, 1.5 3 1.3 cm outside on the
rods), base plate (bottom of the reactor) and the blade in the
center.
5513
2.2. Operational scheme
To simulate periods of stagnation in home plumbing the
reactors were flushed with peristaltic pumps with shear
created by the stirplate for 5 min followed by 8 h stagnation
periods. The feed pumps and stirplates were controlled by
timers that were offset by 1 min with the stirplate starting
before the pumps. At the end of 5 min the stirplates stopped,
followed by the pumps. The stirplates were set to create
a rotational speed of the blade of 300 rpm, which was
approximately equivalent to a velocity of 3 ft/s in the bulk
water. The cycle was repeated three times per day.
Two sets (four reactors each) of modified CDC reactors
equipped with two different types of coupons (PVC and
copper) were used in this investigation. All reactors had been
in operation for more than one year and showed signs of
stable, complete nitrification as measured by conversion of
ammonia to nitrate.
2.3. Stock/feed solution preparation
All reactors were fed with a combination of mineral amended
reverse osmosis (RO) water, biologically treated Bozeman tap
water, and a humic substances organic feed. The RO water
was amended to create an alkalinity of 35 mg/L as CaCO3 and
a stable pH of 8.15. Constituents of the RO water þ mineral
feed were MgSO4 (39.6 mg/L), NaHCO3 (59.6 mg/L), CaSO4$2H2O
(25 mg/L), Al2(SO4)$18H2O (0.62 mg/L), CaCl2$2H2O (20.80 mg/
L), and Na2SiO3$9H2O (26 mg/L). Ammonium sulfate was
added to provide a final concentration in the reactor of
0.71 mg/L as N. Biologically treated water provided as a sepa-
rate, parallel influent. It was created by flowing Bozeman tap
water (surface water source, no background ammonia, chlo-
rinated) through a granular activated carbon column followed
by flow through a biologically active carbon column to provide
a continuous inoculum of indigenous organisms (104 CFU/ml
of heterotrophic plate count (HPC)) that was the only source of
microorganisms to the reactors. This water also contained
sufficient background phosphate for microbial growth at the
organic carbon level used. Organic carbon was supplied to the
reactors in a third, separate influent in the form of soil-derived
humic substances. 50 grams of Elliot silt loam soil (Interna-
tional Humic Substances Society) was added to 500 ml of 0.1 N
NaOH and mixed for 48 h. This solution was centrifuged at
10,000  g for 20 min. The supernatant was collected in carbon
free glassware (baked at 390  C for 5 h) and stored at 4  C in the
dark. Total organic carbon content of the humics was
measured using a Dohrman DC-80 and subsequently diluted
to the appropriate concentration using the RO water feeding
the reactors to provide a concentration in the reactors of 4 mg/
L as dissolved organic carbon. The RO water/biologically
treated Bozeman tap water/organic carbon feed ratio was
50:5:1.
Other amendments and modifications of the influent feed
are described below for each set of specific experiments.
2.4. Sampling
Water was collected from the reactors at the end of the 8 h
stagnation period three times weekly. Samples were analyzed
5514 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
for ammonia, nitrate, and nitrite. Weekly samples were
analyzed for heterotrophic plate counts, ammonia oxidizing
bacteria, and nitrite oxidizing bacteria.
Free NH3eN was measured using a HACH 2000 spectro-
photometer using the salicylate method (HACH method
10023) at 655 nm immediately after the samples were
collected. Nitrite was analyzed using a HACH 2000 spectro-
photometer and the diazotization method. Reaction of nitrite
with sulfanilic acid and forms an intermediate diazonium salt
that couples with chromotropic acid to produce a pink colored
complex measured at 507 nm. Nitrate in filtered samples
(0.2 mm pore size polyethersulfone) was measured using
a Dionex ion chromatography system with a CD20 conduc-
tivity detector and GP40 gradient pump unit. An AS4A column
and DS3 detection stabilizer was also used in this method. The
Dionex ion chromatography system was calibrated using five
sodium nitrate standards (1, 0.5, 0.2, 0.1, 0 ppm of NO3eN). To
minimize experimental error, after every seven measure-
ments a standard solution of nitrate was measured to check
the accuracy of the measurement. If the obtained measure-
ment of the standard was outside 90e110% of the standard
value then the calibration was repeated and sample was
measured again (Standard Methods, 1995).
Heterotrophic plate counts were done according to
Standard Methods (1995) 9215A using R2A agar plates. Plates
were incubated at 20  C for 7 days, and then the number of
colonies in the plates was counted using a Quebec colony
counter. For the chloramine experiments, the disinfectant
was neutralized with sodium thiosulfate prior to dilution and
plating.
Ammonia oxidizing bacteria (AOB) and nitrite oxidizing
bacteria (NOB) populations were enumerated using the most
probable number (MPN) technique (Rowe et al., 1977) using
Costar Clear-Bottom 96 well microtiter plates. The mineral
medium used for AOB contained per liter: (NH4)2SO4, 330 mg;
KH2PO4, 100 mg; MgSO4$7H2O 40 mg; CaCl2, 15 mg and 1 ml of
a trace-element solution. The trace-element solution con-
tained per liter: Na2EDTA, 4292 mg; FeCl2$4H2O, 1988 mg;
MnCl2$H2O, 99 mg; NiCl2$6H2O, 24 mg; CoCl2$6H2O, 24 mg;
CuCl2$2H2O, 17 mg; ZnCl2, 68 mg; Na2MoO4$2H2O, 24 mg and
H3BO3, 62 mg. Bromothymol blue (5ml/L of 0.04% solution in
water) was added as a pH indicator. The pH was adjusted to 8
using 1 M NaOH before autoclaving at 110  C for 15 min. The
NOB medium had the same composition except that it did not
contain (NH4)SO4 and bromothymol blue, and was supple-
mented with 34.5 mg/L NaNO2. The pH was adjusted to 6.5
with 1 M NaOH before autoclaving at 110  C for 15 min. After
inoculation the microtiter plates were sealed with polyester
tape to prevent evaporation and incubated for nine weeks at
20  C in the dark. After the incubation period AOB and NOB
presence was determined by detecting nitrite and nitrate in
the medium by adding 40 ml of 0.2% diphenylamine in H2SO4 in
the well. The absorbance of the blue color was measured at
630 nm with a microplate reader (EL808 ultra microplate
reader BioTek instruments). The blue color indicated nitrite
or nitrate had formed and the well was scored as positive.
Griess Ilosvay reagent (Alexander and Clark, 1965) was used to
detect nitrite. A red color was produced within 5 min and
absorbance measured after 5 min at 540 nm with the micro-
plate reader. If nitrite in the well was detected it was scored as
negative for NOB. The difference between the two measure-
ments then allowed for differentiation between AOB and NOB.
The MPNs were calculated according to Rowe et al. (1977).
Biofilm samples were collected at the end of each experi-
ment’s test period prior to the recovery phase. One coupon
was removed from the reactor and placed in a glass tray
(baked at 390  C for 5 h) containing the autoclaved RO water.
The coupon was then scraped using an autoclaved rubber
policeman inside a laminar flow hood. The biomass with
water was poured in a sterilized 50 ml Falcon tube, which
was then homogenized with a homogenizer (Bio-
homogenizer Model M133/12810, ESGE) for 30 s. From the
homogenized biomass, samples were taken for MPN and HPC
analysis.
2.5. Statistical analysis
Paired t-test analysis was done using Microsoft Excel on the
data to see if there are significant differences between two
treatments. The level of significance for all tests was a ¼ 0.05.
2.6. Experimental approach
2.6.1. Effect of copper ion on nitrification
To test the effect of copper on nitrification, copper in the form
of CuSO4 was added to the influent of one nitrifying PVC
reactor. Initially15 ppb copper was continually dosed and
incrementally raised to 1.3 ppm, which is the action limit for
copper according to the Lead and Copper Rule. Each concen-
tration of copper was maintained for two weeks. Another PVC
reactor was used as the control where no copper was added.
Because nitrification is presumed to be more affected by free
copper (Cuþ2) (Braam and Klapwijk, 1981), and the amount of
Cuþ2 in solution is a function of pH, the pH of the influent with
1.3 ppm of copper was then reduced gradually from 8.15 to 6.6
by 0.3 units every two weeks. Another PVC reactor with the
influent at the same pH but with no copper was used a control.
When copper was added to the PVC reactor at lower
concentrations (less than 400 ppm) an ICPMS (inductively
coupled plasma mass spectrometer) was used to measure the
copper. At higher concentrations a HACH 2000 spectropho-
tometer (method 8506,560 nm wavelength) was used. Both
total and dissolved copper in the effluent water were
measured. Dissolved copper is operationally defined as the
portion of the copper which passes through a 0.45 mm pore
size syringe filter. In the presence of colloidal species that can
pass through the filter, the method represents an upper bound
to truly soluble copper. When the pH was adjusted, a copper
ion selective electrode (Cu-ISE), (Orion cupric electrode,
model, 94e29, Boston, MA) was used to measure the free
copper (Cuþ2) in the water. The electrode was calibrated using
standard cupric ion solutions according to manufacturer’s
direction before measuring the sample. Cupric ion at 1.3 ppm
total copper and pH 6.6 was measured and was found to be in
the 0.90  0.10 ppm range.
2.6.2. Effect of chlorite on nitrification
Laboratory grade sodium chlorite was added to the influent of
one PVC and one copper reactor. Initially 0.2 ppm of chlorite
was added and gradually increased to 2.0 ppm and then
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
a shock load of 20 ppm of chlorite was dosed. Each concen-
tration of chlorite was tested for two weeks. Chlorite was
measured by a modification (McGuire et al., 1999) of the US
Environmental Protection Agency (USEPA, 1993) method 300
with the Dionex ion chromatography system described
above equipped with an AS9 column and 100 ml sampling loop.
Chlorite was then discontinued and the recovery of nitrifica-
tion in both the PVC and copper system was observed. Copper
concentrations in the copper reactor effluent were also
monitored.
2.6.3. Effect of chloramine on nitrification
To represent different scenarios that can occur in a distribu-
tion system due to chloramine decay, different amounts of
chlorine were added in the ammonia-containing RO water
influent of two nitrifying copper and PVC reactors. The added
sodium hypochlorite formed chloramine with the ammonia in
the influent at different chlorine to ammonia ratios. Initially
a 0.5:1 ratio was applied and gradually raised to 5:1. Both free
and total chlorine were measured using a Lamotte DC1100
colorimeter with the DPD colorimetric method (Standard
Methods, 1995, method 4500-Cl). At the beginning of the
measurement phosphate buffer was added to the sample to
maintain a pH between 6.2 and 6.5. Dissolved and total copper
in the copper reactor effluents were also measured.
At the end of chloramine exposure, chlorine was dis-
continued and recovery of nitrification in PVC and copper
reactors was compared.
3. Results
All experiments were done in reactors where stable, complete
nitrification (conversion of ammonia to nitrate) had been
occurring for one year. Before initiation of the experiments,
there was no difference in ammonia conversion between PVC
and copper, and there was good reproducibility in effluent
water quality between pairs of reactors with the same coupon
materials (data not shown). All testing consisted of a phase
where the reactors were exposed to a stepwise change in
water quality (addition of copper/change of pH, addition of
chlorite, or increase in chlorine to ammonia ratio) and
compared to a control reactor with stable influent water
quality followed by a recovery period where complete nitrifi-
cation was again established. As such, the data demonstrated
the efficacy of the treatment method in addition to the
robustness of the nitrification process.
3.1. Effect of copper on nitrification
To eliminate the influence of the copper substratum on the
results, these experiments were conducted with copper ion
added to a PVC reactor. Copper doses were incrementally
increased from 15 to 1300 ppb. At the lower copper doses there
was little difference in NH3 utilization measured by percent
disappearance of influent to effluent concentrations in the
effluent. At higher copper doses (i.e., 600e1300 ppb) inter-
mittent small decreases in NH3 utilization were observed.
Overall, there was no statistically significant difference in the
effluent ammonia concentrations of the copper reactor
5515
compared to the control until 600 ppb was reached. After
600 ppb added copper, the difference was significant ( p < 0.05)
but actual measurements differed by a maximum of only
0.05 mg/L NH3eN (n ¼ 6 for each copper dose). A similar trend
was seen with nitrate measurements; there were very small
differences with a maximum variation of 0.04 mg/L NO3eN
between the control and copper dosed systems. After the
highest copper dose was reached, and to ensure all copper was
as Cuþ2, the pH of the reactor was gradually lowered to 6.6.
Another PVC reactor without copper where the pH was also
adjusted was used as a control. There were statistically
significant differences in effluent ammonia and nitrate levels
at all pH values (7.8e6.6 at 0.3 increments) but the actual
amounts were no more than 0.04 mg/L NH3eN or NO3eN. In all
cases, there was no more than 0.26 mg/L NH3eN in the control
or copper exposed reactor effluents (range 0.16e0.26) and
NO3eN ranged from 0.65 to 0.43 mg/L. Only trace amounts of
nitrite were measured (data not shown).
There were no significant differences in AOB or NOB MPN
counts in the reactor effluents from the copper vs control
reactors. Values ranged from 7 to 35 MPN/ml for AOB and from
4 to 41 MPN/ml for NOB. These results are reported because
utilities will typically only measure bulk water nitrifier
numbers since biofilm interrogation is not feasible. Similarly,
there were no differences in the HPC values, with the range
from 4  104 to 6  105 per ml. Increasing copper doses were
not statistically correlated with MPN or HPC values.
3.2. Effect of chlorite on nitrification
Because chlorite has been used in full scale distribution
systems to control nitrification with varying effect, this
chemical was added in increasing doses (0.2e20 ppm) to
nitrifying PVC and copper reactors. Reactors that did not
receive chlorite were maintained as controls. As seen in Fig. 2,
low range chlorite (i.e., 0.2e2 ppm) did not affect nitrification
in the PVC reactor. When the dose was increased to 20 ppm
the PVC reactor was slightly affected as it temporarily had
lower NH3eN utilization, but even in the presence of chlorite,
complete ammonia loss rebounded. The temporary decreases
in NH3eN utilization may have been the result of detaching
biomass, adaptation to the chlorite, or other unknown factors.
Fig. 2 e NH3eN utilization (%) in bulk water for chlorite
dosed and control PVC reactors.
5516 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
The copper reactor was relatively unaffected by lower chlorite
doses, but at 20 ppm chlorite, NH3eN utilization dropped to
zero (Fig. 3). After chlorite was discontinued it took almost two
months to re-establish complete nitrification.
The bulk water NO2eN and NO3eN as percentages of the
added NH3eN were calculated. For the both reactors, chlorite
at the low range (0.2e2.0 ppm) did not significantly affect the
NO2eN percentage. At 20 ppm chlorite, NO2eN percentage
increased noticeably to about 1e3% of the added ammonia in
both systems, but the magnitude of this change was only in
the ppb range. In the case of NO3eN, the percent of ammonia
converted decreased from near 100%e60% in the PVC reactor
the day that the chlorite was increased to 20 ppm and
rebounded to the original level of near 100% within 10 days
even when this level of chlorite was maintained. In the case of
the copper reactor, the percent conversion dropped to less
than 5% after a week of exposure to 20 ppm chlorite and did
not rebound for almost two months.
Effluent total and dissolved copper concentrations in
chlorite added and control copper were also measured. The
95% confidence interval for total and dissolved copper for the
chlorite dosed reactor (0.8  0.04 and 0.69  0.03 ppm) was
higher than that for control reactor (0.71  0.02 and
0.63  0.02 ppm). A paired t-test for both effluent total and
dissolved copper was averaged for chlorite concentrations
from 0.1 to 2 ppm and for 20 ppm. In all cases, there was
significantly more total and dissolved copper in reactors that
had received chlorite.
The MPN values for AOB and NOB were comparable to
those found in the previous copper experiment. The only
difference was that at 20 ppm chlorite, no NOB were found in
the effluent of both the PVC and copper reactors. HPC values
were unaffected and similar to those in the previous copper
experiment.
3.3. Effect of chloramine on nitrification
The effect of chloramine on nitrification was investigated by
gradually increasing the amount of free chlorine fed to the
influent of nitrifying PVC and copper reactors. Initially chlo-
rine was added at 0.5:1 chlorine to ammonia ratio and grad-
ually raised to a 5:1 ratio, with a total/combined chloramine
Fig. 3 e NH3eN utilization (%) in bulk water for chlorite
dosed and control copper reactors.
dose of 3.55 mg/L. NH3 utilization in the chloramine dosed
PVC, copper and control reactors are shown in Fig. 4. Bulk
NH3eN utilization decreased significantly only at the 5:1
chlorine to ammonia ratio, with occasional decreases at the
lower ratios. Note that the 5:1 ratio was maintained for two
months and there was a long period of acclimation where the
ammonia utilization gradually declined. Even at the 5:1 ratio
after two months of exposure, the copper reactor continued to
utilize around 20% of the ammonia. When chloramine was
discontinued, the copper reactor regained full nitrification
after three weeks. In contrast, the PVC reactor ceased
ammonia utilization at the higher ratio and it required
approximately six weeks to recover its nitrifying ability.
The percent conversion of ammonia in the influent to
NO2eN and NO3eN in the reactors after 8 h of stagnation was
calculated for all chlorine to ammonia ratios. Only in the case
of the 5:1 ratio in the PVC reactor did the nitrite level initially
increase from the detection limit and peaked at two weeks to
approximately 1.4% of the added nitrogen. This level then
again decreased to no detectable nitrite. There was no change
in nitrite in the copper reactors as a result of increased chlo-
rine to ammonia ratios; it remained at the limit of detection
suggesting complete conversion to nitrate. For the copper
reactor, the percent of ammonia converted to nitrate dropped
one week after attaining the 5:1 ratio to an average of 18%
conversion and it took two weeks to reach an average of 12%
in the PVC reactor. When chlorine addition was terminated,
there was a rebound of total conversion of ammonia to nitrate
approximately three weeks later for the PVC reactor. This time
was considerably longer (seven weeks) for the copper reactor,
and corresponds to the complete loss of ammonia (Fig. 4).
Total and dissolved copper concentrations in the effluent
of the monochloramine-dosed and control copper reactors
were measured at each chlorine to ammonia ratio. The 95%
confidence intervals for total and dissolved copper for the
monochloramine-dosed reactor (0.83  0.02 and 0.73  0.02)
are higher than that for control reactor (0.75  0.02 and
0.63  0.01). A paired t-test showed that the p values for total
and dissolved copper were significantly different. It appears
that the presence of monochloramine significantly increased
the copper concentration in the water.
Fig. 4 e NH3eN utilization (%) in bulk water for chloramine
dosed (PVC, copper) and control (PVC, copper) reactors.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2 5517

The average MPN value for AOB and NOB in the reactor
Table 2 e Cell numbers for autotrophic and heterotrophic
water are shown in Table 1. In general, there were fewer AOB populations in the biofilm at the end of each experiment.
and NOB in reactors that received chlorine. No NOB were AOB e ammonia oxidizing bacteria, NOB e nitrite
detected in the presence of chloramine at the 5:1 ratio in oxidizing bacteria, HPC e heterotrophic plate counts.
either reactor which is consistent with the increased level of ND [ none detected.
nitrite that was measured. HPC effluent counts were unaf- AOB NOB HPC
fected and comparable to those of the control reactors that did (MPN/cm2) (MPN/cm2) (CFU/cm2)
not receive chlorine (data not shown).
Copper added PVC 5.7  103 4.4  103 1.6  106
pH Control PVC 2.8  103 3.2  103 7.7  105
3.4. Effect of copper, chlorite and chloramine on biofilm Chlorite added PVC 3.5  102 7.7  101 1.5  106
cell numbers NH2Cl added PVC 9.2  101 7.9  102 1.2  105
PVC-control 2.4  103 2.8  103 1.2  106
Chlorite added copper 1.3  101 ND 1.8  105
AOB/NOB and HPC numbers in the biofilm were determined at
NH2Cl added copper 1.4  102 ND 9.8  101
the end of each experiment and before the recovery phase Copper-control 1.1  103 8.0  102 1.8  105
(Table 2). AOB/NOB abundance is typically about 3 logs lower
than the HPC with the exception of the highest chlorine to
ammonia (5:1) ratio in the copper reactor, where the AOB were
present at numbers greater than that of the heterotrophs. No been observed in nitrifying water distribution systems where
NOB were detected in the copper reactor biofilms at the there is incomplete nitrification, nitrite accumulation, and
termination of the chlorite and chloramine experiments. A high nitrite-N concentrations (Wolfe et al., 1988).
lack of detection of NOB in the biofilm and the bulk water is
consistent with the increased levels of nitrite measured in 4.1. Added copper
these reactors.
Copper introduced to the nitrifying PVC reactor did not have
a significant effect on ammonia utilization, nor on nitrate
concentration in the reactor. The numbers of HPC, AOB and
4. Discussion
NOB were also unaffected by copper. This observation
contradicts the results from other researchers (Skinner and
In this project three different strategies for controlling nitri-
Walker, 1961; Loveless and Painter, 1968; Martin and
fication in household plumbing were studied. The inhibitory
Richard, 1982; Zhang and Edwards, 2005, 2010), who repor-
effect of copper, chlorite, and a chloramine residual on nitri-
ted that low (5 ppbe0.56 ppm) copper concentrations inhibi-
fication were tested. The impact of each control measure on
ted nitrification. It is possible that inhibition may have
the release of total and ionic copper from copper coupons into
occurred if concentrations had been increased to higher levels
solution was also evaluated. Throughout the experiments in
as those used by Tomlinson et al. (1966) and Meiklejohn
the control reactors, there was total loss of ammonia, no
(1950). The discrepancy in copper sensitivity between the
detectable nitrite, and near complete conversion of ammonia
results of this work and that of others may be because the
to nitrate. This suggests that nitrite was likely formed but
previous work was done with pure cultures or activated
converted to nitrate within the 8 h stagnation period of the
sludge, and the copper tolerance for those biological systems
experimental system and that complete nitrification was
may be much different from the biofilm grown in the reactors
occurring. Complete nitrification is contrary to what has often
used in this project. Another possible explanation is that
organic matter contains functional groups (Sarathy and Allen,
2005) that bind metals to form less bioavailable complexes
Table 1 e Average (n [ 2) MPN per ml for ammonia (Loveless and Painter, 1968; Dodge and Theis, 1979; Crecelius
oxidizing bacteria (AOB) and nitrite oxidizing bacteria et al., 1982) that are therefore less inhibitory (Kim et al.,
(NOB) at different Cl2 to NH3eN ratios. ND [ none
2006). The humics used in this research may have acted in
detected.
this capacity although the copper in the reactor water was
Chlorine NH2Cl Control NH2Cl Control
detected in the ionic form.
to NH3eN added PVC PVC added copper
Added copper also had no impact on the HPC numbers in
ratio reactor reactor copper reactor
reactor the reactors. This is in contrast to the findings of others
(Thurman and Gerba, 1989; Artz and Killham, 2002; Kim et al.,
0.5:1 AOB 31 37.7 10.7 38.1
2002; Teitzel and Parsek, 2003; Lehtola et al., 2004) who
1.0:1.0 27.6 31.5 10.4 14.5
reported that copper has a toxic effect on heterotrophs in
1.5:1 9.7 52.2 6.9 27.6
2:01 7.2 9.5 8.7 17 drinking water. Again this discrepancy may be due to the
5:01 12.7 31.4 13.2 21.2 difference in experimental conditions and bacterial pop-
ulations. Another potential explanation is that the reactors
0.5:1 NOB 48.8 59.1 106.7 30.3 used in this study contained intact, natural biofilms that had
1.0:1.0 9.7 23.3 23.5 51.5 been present for over a year while other studies used sus-
1.5:1 21 351.9 28.5 170.9
pended organisms and/or fresh copper surfaces.
2:01 8.4 793.6 10.2 61.5
5:01 ND 310.1 ND 48.3
Copper toxicity depends on the concentration of cupric ion
(Cuþ2) (Braam and Klapwijk, 1981), and this ionic form
5518 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
becomes more abundant at a lower pH (Edwards et al., 1996).
Even when the pH of the reactor was lowered to 6.6 to ensure
that the added copper was in the cupric ion form, there was no
effect on nitrification or the numbers of AOB, NOB or HPC in
the reactor effluent or in the biofilm.
4.2. Chlorite
Chlorite was chosen as another potential control mechanism
because previous studies by McGuire et al. (1999) showed
that chlorite ion (0.2e1.0 ppm) in distribution systems can
inhibit nitrification. Results in this project contradict their
reports. The PVC reactor was initially impacted only at the
unrealistic dose of 20 ppm chlorite but then regained the
ability to nitrify in the presence of this concentration. At this
high concentration there was an impact on nitrification in
the copper system that persisted after the cessation of
chlorite addition. McGuire et al. (1999) mentions that chlorite
did not inhibit nitrification in one system, which according to
the author may be due to the presence of higher ammonia
(1.4 mg/L). Karim and LeChevallier (2006) also reported that
chlorite was unable to hinder nitrification. All these studies
were done with low doses of chlorite (0.2e1.0 ppm) and their
flow pattern, water quality and bacterial population/biofilm
characteristics may be significantly different from this
project. Another possible explanation is that chlorite inhibits
the activity of Nitrosomonas europaea and Nitrobacter winog-
radski (Hynes and Knowles, 1983), but may be inactive toward
other groups of AOB and NOB that may be predominant in
the reactors.
The planktonic and biofilm heterotrophic populations
were unaffected by chlorite exposure. The trend in hetero-
trophic numbers supports previous work by Gagnon et al.
(2005) where chlorite at 0.1e0.25 ppm was ineffective in
inactivating heterotrophic bacteria. Similar to HPC, AOB
values did not show any effect due to chlorite exposure. The
planktonic NOB population remained unchanged for all
concentrations of chlorite except at 20 ppm, when they could
not be detected. NOB were also not detected in the copper
reactor’s biofilm at the end of exposure period to the high
chlorite concentration. Hynes and Knowles (1983) reported
that pure cultures of AOB (N. europaea) and NOB (N. winog-
radski) were inhibited by chlorite, and that the AOB are 50
times more sensitive to chlorite inhibition than NOB. In our
experiments it appeared that the NOB were more sensitive
than the AOB. Again this may be due to the differences in
experimental setup, pure cultures vs environmental biofilm,
water quality and other factors. It is also possible that the
methods used in this work did not enumerate all of the
potential nitrifying organisms in the reactors.
A potential impact of chlorite on copper plumbing could be
an increased release of copper into the water. As chlorite
(ClO2
2 ) could be transformed to chlorine dioxide (ClO2) in an
acidic environment (Gates, 1989) that could created by
oxidation of ammonia during nitrification by biofilm on the
surfaces, copper corrosion could increase due to the oxidative
nature of chlorine dioxide. Therefore, there could be an
unintended result of elevated copper release when chlorite is
applied, and this was seen in these experiments. For this
reason, and because chlorite is a regulated compound, the
utilities should carefully evaluate chlorite before implement-
ing it as a nitrification control strategy.
4.3. Chloramine
According to a survey by Seidel et al. (2005), optimizing the
chlorine to ammonia ratio is the most common nitrification
control technique. The chlorine to ammonia-N weight ratio
used to form monochloramine typically varies from 3:1 to 5:1
(Wilczak, 2006). Several studies in full scale chloraminated
systems have determined that a minimum 2e3 mg Cl2/L
chloramine residual should be maintained to prevent nitrifi-
cation (Wolfe et al., 1990; Lieu et al., 1993; Kirmeyer et al., 1995;
Odell et al., 1996; Harrington et al., 2002). A combination of
chloramines does and optimizing the chlorine to ammonia
ratio has been shown to be the easiest and most cost-effective
way to control/prevent nitrification (Lieu et al., 1993).
The results of this work can be compared to several studies
where the chlorine to ammonia ratio has been measured in
pilot and full scale systems and related to nitrification.
McGuire et al. (2004) showed that nitrification occurred in
a pilot system where chloramine was applied at a 3:1 ratio.
Similarly, for two covered finished water reservoirs in
southern California which were initially chloraminated at
a 3:1 ratio, nitrification was significantly reduced after raising
the ratio to 5:1 (Wolfe et al., 1988). A Florida utility rarely
experienced nitrification when the combined chlorine
residual was above 1 mg/L at a chlorine to ammonia ratio of
5:1 (Liu et al., 2005). A study done by Karim and LeChevallier
(2006) showed that monochloramine applied at a 3:1 ratio
was not able to control nitrification, but it was effective at a 5:1
ratio.
Based on the above information, the impact of an incre-
mental increase in the chlorine to ammonia ratio on nitrifi-
cation was tested. The intent was to test situations in (1)
a system where chloramine residual was regained and/or (2)
households downstream of booster chlorination that was
used to recreate chloramines from free ammonia (Wolfe and
Lieu, 2001; Wilczak, 2006). At the starting ratio of 0.5:1, about
0.35 mg/L of total chlorine was present. This was chosen
because Holt et al. (1995) showed reduction of nitrification at
total chlorine concentrations of more than 0.3 mg/L. However,
at an 8 h stagnation time, this low ratio did not have an impact
on nitrification and organisms responsible for nitrification
could be found in the bulk water and biofilm. Very minor
transient impacts were seen with increasing ratios of 1:1,
1.5:1, and 2:1. There are several possible explanations for
these results. Nitrifying bacteria form protective layers (slime
layer or capsules) mainly composed of polysaccharides
(Prosser, 1986) as a defensive mechanism against unfavorable
environmental conditions such as low pH. These capsules
protect organisms so that they are more resistant to disin-
fectants (Stewart and Olson, 1996). Cunliffe (1991) detected
nitrifiers in 64% of the samples collected in South Australia
and of them 20.7% contained more than 5.0 mg Cl2/L of
monochloramine. The author hypothesized that as the nitri-
fiers grow in aggregates or in biofilm attached to the surface,
they remain unaffected by disinfectant and the nitrifiers
detected in samples containing high chloramine residual may
have been detached shortly before or during the sampling
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
period. Higher AOB have been detected in biofilm than water
in the Metropolitan Water District of Southern California
distribution system (Stewart and Lieu, 1997). Regan (2001)
postulated on the protective mechanism of biofilms. Mono-
chloramine is mass transport limited in biofilm, so bacteria
inside the biofilm are not exposed to the disinfectant. Another
protective mechanism may be the relative ratio of growth vs
disinfection. Harrington et al. (2003) stated that if AOB growth
rate driven by ammonia concentration exceeds the AOB
inactivation rate by monochloramine, then theoretically AOB
can grow in the presence of monochloramine. Fleming et al.
(2005) proposed nitrification potential curves based on the
relative concentration of chlorine (biocide) and free ammonia
(food). According to them the threshold chlorine value is 1.6,
above which nitrification would be prevented, without any
influence from free ammonia concentration. At chlorine
concentrations below 1.6, the nitrification potential depends
on the ratio of chlorine and free ammonia.
Only when the 5:1 Cl2 to NH3eN ratio was reached was
there a consistent reduction of nitrification during the 8 h of
stagnation, and the affect of this ratio agrees with results from
Karim and LeChevallier (2006) and Lieu et al. (1993). Likewise,
in work by Harrington et al. (2002), nitrification did not occur
when the total chlorine concentrations were more than
2.2 mg/L and the biocide to food ratio was 1.9 mg Cl2/mg of N
or more. The most pronounced effect was on the PVC reactor,
which is in contrast to the chlorite effect which was most
obvious on the copper reactor. Interestingly, the HPC numbers
in the biofilm on the copper surfaces were lower than those of
the AOB, suggesting that the disinfectant had greater activity
on the HPC than the nitrifiers. No planktonic NOB were
detected in monochloramine exposed reactors at the 5:1 ratio
and they were also not detected in the biofilm of the copper
reactor. A possible explanation is that the NOB were more
vulnerable to disinfection than the AOB (Wolfe and Lieu, 2001).
These results also support observations that once nitrifi-
cation starts in full scale systems, higher levels of chloramine
may not be an effective control method. Skadsen (1993)
reported that a chloramine dose of 8 mg Cl2/L was not effec-
tive in controlling nitrification in the Ann Arbor, Michigan
distribution system. This may be due to the fact that nitrite
can degrade chloramine residuals before it can inactivate the
nitrifying bacteria (Wolfe et al., 1988; Kirmeyer et al., 1995;
Odell et al., 1996). Also, maintaining monochloramine at
a high ratio close to 5:1 in full scale systems is not always easy,
and is sometimes associated with dicholoramine formation,
and taste/odor problems and higher DBP formation (Skadsen
and Cohen, 2006).
This project also provides some insight on effect of chlo-
ramine on copper corrosion. It has been found that chlora-
mine increased copper corrosion (Ingleson et al., 1949).
Enhanced copper solubility during periods of chloramination
with excess ammonia present was observed for Champaign IL
tap water (AwwaRF, 1990). Ammonia has a strong complexa-
tion constant for cupric ion (Schock et al., 1995). On the other
hand, according to MacQuarrie et al. (1997) and Rahman et al.
(2007), application of monochloramine results in a decrease in
copper pipe corrosion. Current results found higher total and
dissolved copper concentrations in the reactors where the
disinfectant was added.
5519
4.4. Copper vs PVC surfaces
Copper and PVC were chosen because of their use in plumbing
systems and also because of the potential antimicrobial effect
of copper vs PVC. One could also hypothesize that copper
materials may be more prone to nitrification in chloraminated
systems because monochloramine can rapidly decay through
reactions with a copper plumbing system (Edwards and
Nguyen, 2005), therefore providing more free ammonia for
nitrification. Therefore, pipe material could significantly
influence the nitrification process.
In the conditioned control reactors with no added chlorite
or chlorine, there was no difference in ammonia utilization
between the two surfaces, and the HPC, AOB and NOB numbers
were similar. The lack of efficacy of copper against these
organisms was supported by the experiments where copper
was added to the PVC reactor and no decrease in nitrification
was seen. In the presence of chorite or combined chlorine,
there were some differences between the two systems. In the
case of chlorite, nitrification in the copper system was affected
to a greater extent with a loss of nitrification at the highest
concentration (20 ppm) and an extended recovery time after
chlorite addition ceased. This could be because of the
production of chlorine dioxide (Gates, 1989) on the metal
surface. However, in the presence of combined chlorine, the
PVC system was more greatly impacted; nitrification ceased
and it required nearly six weeks for recovery. In the copper
reactor, there was never a complete loss of nitrification, and
complete nitrification was again attained only three weeks
after chlorine addition stopped. This suggests that the mech-
anism of chloramine decay on the copper surface, as reported
by Edwards and Nguyen (2005), could be at play.
4.5. Relationship between HPC and nitrification
Some research has shown that there is a correlation between
HPC values and nitrification in water systems with HPC
increasing when nitrification occurs. Wolfe et al. (1990)
reported that HPC and AOB population were highly corre-
lated in distribution system water, and that this may be
explained due to the dependence of HPC on AOB for carbon
fixation. However, other researchers reported that the rela-
tionship is very site specific (Donnelly and Giani, 2005), likely
because HPC can grow in response to other available organics
and their numbers are also affected by residual disinfectant.
Throughout these experiments there was no consistent rela-
tionship between HPC and AOB/NOB MPN values. This may be
due to the presence of humic substances in the reactors.
5. Conclusions
 Copper surfaces or copper added to reactors at concentra-
tions up to 1.3 ppm within a pH range of 6.6e8.15 did not
inhibit nitrification.
 Chlorite was effective at inhibiting nitrification only at an
unrealistic dose (20 ppm), on copper surfaces. There was
limited, transient control on reactors with PVC surfaces at
20 ppm chlorite.
5520 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
 Chloramine at a Cl2 to NH3eN ratio of 5:1 managed nitrifi-
cation in the copper reactor and was able to control it in the
PVC system.
 The addition of chlorite and chloramines may increase
copper corrosion.
 No correlation between HPC and AOB/NOB was found.
 Time to recovery after cessation of addition of chlorite and
chloramine varied but did occur, suggesting that nitrifica-
tion in premises plumbing is a robust process.
Acknowledgments
This material is based upon work supported by the
National Science Foundation under Grant No. 0329474. Any
opinions, findings, and conclusions or recommendations
expressed in this material are those of the authors and do
not necessarily reflect the views of the National Science
Foundation. The authors thank the MUSES team funded by
this grant and led by Dr. Andrea Dietrich at Virginia Tech
for their insights, with special appreciation to Dr. Marc
Edwards for his assistance. The authors also acknowledge
John Neuman and Daniel Swanson for their help in the
laboratory.
references
Alexander, M., Clark, F.E., 1965. Nitrifying bacteria. In: Black, C.A.
(Ed.), Methods of Soil Analysis. American Society of
Agronomy, Madison, WI, pp. 1477e1483. Part 2.
Alt, E.R., Sternlieb, I., Goldfischer, S., 1990. The cyto-pathology of
metal overload. International Review of Experimental
Pathology 31, 165e188.
Artz, R.R.E., Killham, K., 2002. Survival of Escherichia coli O157:H7
in private drinking water wells: influences of protozoan
grazing and elevated copper concentrations. FEMS
Microbiology Letters 216 (1), 117e122.
American Water Works Association Research Foundation, 1990.
Lead Control Strategies. AWWA Research Foundation and
AWWA, Denver, CO.
Braam, F., Klapwijk, A., 1981. Effect of copper on nitrification in
activated sludge. Water Research 15 (9), 1093e1098.
Clescerl, L.S., Greenberg, A.E., Eaton, A.E. (Eds.), 1995. Standard
Methods for Examination of Water & Wastewater, nineteenth
ed. American Public Health Association.
Crecelius, W.A., Hardy, J.T., Gibson, C.I., Schmid, T.R.L., Apts, C.W.,
Curtisen, J.M., Joyce, S.P., 1982. Copper bioavailability to marine
bivalves and shrimps: relationship to cupric ion activity.
Marine Environmental Research 6 (1), 13e26.
Cunliffe, D.A., 1991. Bacterial nitrification in chloraminated water
supplies. Applied and Environmental Microbiology 57 (11),
3399e3402.
Dodge, E.E., Theis, T.L., 1979. Effect of chemical speciation on the
uptake of copper by Chironomus tentans. Environmental
Science Technology 13 (10), 1287e1288.
Donnelly, M., Giani, R., 2005. Developing nitrification control
limits from distribution system data. In: Proceedings of
American Water Works Association Water Quality
Conference, Denver CO.
Edwards, M., Nguyen, C., 2005. Chemistry of rapid chloramine
decay in water contacting copper plumbing. In: Proceedings of
American Water Works Association Water Quality Technology
Conference, Quebec City, Canada.
Edwards, M., Schock, M.R., Meyer, T.E., 1996. Alkalinity, pH and
copper corrosion by-product release. Journal of American
Water Works Association 88 (3), 81e88.
Fleming, K.K., Harrington, G.W., Noguera, D.R., 2005. Nitrification
potential curves: a new strategy for nitrification prevention.
Journal of American Water Works Association 97 (8), 90e99.
Gagnon, G.A., Rand, J.L., O’Leary, K.C., Rygel, A.C., Chauret, C.,
Andrews, R.C., 2005. Disinfectant efficacy of chlorite and
chlorine dioxide in drinking water biofilm. Water Research 39
(9), 1809e1817.
Gates, D., 1989. Chlorine dioxide generation technology and
mythology. In: Proceedings of AWWA Water Quality
Conference, Denver CO.
Geesey, G.G., Jang, L., 1989. Interactions between metal ions and
capsular polymers. In: Beveridge, T.J., Doyle, R.J. (Eds.), Metal
Ions and Bacteria. Wiley & Sons, New York, pp. 325e357.
Goeres, M.D., Loetterle, L.R., Hamilton, M.A., Murga, R., Kirby, D.
W., Donlan, R.M., 2005. Statistical assessment of a laboratory
method for growing biofilms. Microbiology 151 (3), 757e762.
Harrington, G.W., Noguera, D.R., Kandou, A.I., Vanhoven, D.J.,
2002. Pilot-scale evolution of nitrification control strategies.
Journal of American Water Works Association 94 (11), 78e89.
Harrington, G.W., Noguera, D.R., Bone, C.C., Kandou, A.I.,
Oldenburg, P.S., Regan, J.M., Hoven, D.V., 2003. Ammonia from
Chloramine Decay: Effect on Distribution System Nitrification.
AWWA Research Foundation, Denver, CO.
Holt, D., Todd. R.D., Delanoue, A., Colborne, J.S., 1995. A study of
nitrite formation and control in chloraminated distribution
systems. In: Proceedings of American Water Works Association
Water Quality Technology Conference, Denver, CO.
Hu, Z., Chandran, K., Grasso, D., Smets, B.F., 2003. Impact of metal
sorption and internalization on nitrification inhibition.
Environmental Science Technology 37 (4), 728e734.
Hynes, R.K., Knowles, R., 1983. Inhibition of chemoautotrophic
nitrification by sodium chlorate and sodium chlorite:
a reexamination. Applied and Environmental Microbiology 45
(4), 1178e1182.
Ingleson, H., Sage, A.M., Wilkinson, R., 1949. Effect of chlorination
of drinking water on brass fittings. Journal of Industrial Water
Engineering 3 (1), 81e91.
Karim, M.R., LeChevallier, M.W., 2006. Alternative treatment for
nitrification control in chloraminating utilities. In: Proceedings
of the AWWA Annual Conference, San Antonio TX.
Kim, B.R., Andersson, J.E., Mueller, S.A., Gaines, W.A., Kendall, A.
M., 2002. Literature reviewdefficacy of various disinfectants
against Legionella in water systems. Water Research 36 (18),
4433e4444.
Kim, K.T., Kim, I.S., Hwang, S.H., Kim, S.D., 2006. Estimating the
combined effect of copper and phenol to nitrifying bacteria in
wastewater treatment plants. Water Research 40 (3), 561e568.
Kindaichi, T., Kawano, Y., Ito, T., Satoh, H., Okabe, S., 2006.
Population dynamics and in situ kinetics of nitrifying bacteria
in autotrophic nitrifying biofilms as determined by real-time
quantitative PCR. Biotechnology and Bioengineering 94 (6),
1111e1121.
Kirmeyer, G.J., Odell, L.H., Jacangelo, J.G., Wilczak, A., Wolfe, R.L.,
1995. Nitrification Occurrence and Control in Chloraminated
Water Systems. AWWARF and AWWA, Denver CO.
LeChevallier, M.W., Cawthon, C.D., Lee, R.G., 1988. Inactivation of
biofilm bacteria. Applied and Environvironmental
Microbiology 54 (10), 2492e2499.
Lee, Y.W., Ong, S.K., Sato, C., 1997. Effect of heavy metals on
nitrifying bacteria. Water Science and Technology 36 (12), 69e74.
Lehtola, M.J., Miettinen, I.T., Kienanen, M.M., Kekki, T.K., Laine, O.,
Hirvonen, A., Vartiainen, T., Martikainen, P.J., 2004.
Microbiology, chemistry and biofilm development in a pilot
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2
drinking water distribution system with copper and plastic
pipes. Water Research 38 (17), 3769e3779.
Lieu, N.I., Wolfe, R.L., Means, E.G., 1993. Optimizing chloramine
disinfection for control of nitrification. Journal of American
Water Works Association 85 (2), 84e90.
Liu, S., Taylor, J.S., Randall, A.A., Dietz, J.D., 2005. Nitrification
modeling in chloraminated distribution systems. Journal of
American Water Works Association 97, 98e108.
Loveless, J.E., Painter, H.A., 1968. The influence of metal ion
concentration and pH value on the growth of a Nitrosomonas
strain isolated from activated sludge. Journal of General
Microbiology 52 (1), 1e14.
MacQuarrie, D.M., Mavinic, D.S., Neden, D.G., 1997. Greater
vancouver water district drinking water corrosion
inhibitor testing. Canadian Journal of Civil Engineering 24
(1), 34e52.
Martin, G., Richard, Y., 1982. Elements of inhibition of nitrifying
bacteria. Water Science and Technology 14 (6e7), 781e794.
McGuire, M., Lieu, J.N., Pearthree, M., 1999. Using chlorite ion to
control nitrification. Journal of American Water Works
Association 91 (10), 52e62.
McGuire, M.J., Pearthree, M., Blute, N.K., Arnod, K.F., 2004. Tucson
nitrification control pilot study: a significant advance in the
state of the art of implementing chloramines in a desert
environment. Arizona Water and Pollution Control
Association Annual Conference.
McGuire, M.J., Pearthree, M.S., Blute, N.K., Arnold, K.F., Hoogerwerf, T.,
2006. Nitrification control by chlorite ion at pilot scale. Journal
of American Water Works Association 98 (1), 95e105.
McGuire, M.J., Wu, X., Blute, N.K., Askenaizer, D., Qin, G., 2009.
Prevention of nitrification using chlorite ion: results of
a demonstration project in Glendale, Calif. Journal of
American Water Works Association 101 (10), 47e59.
Meiklejohn, J., 1950. The isolation of Nitrosomonas europaea in pure
culture. Journal of General Microbiology 4 (2), 185e191.
National Research Council, 2006. Water Science and Technology
Board, Science and Technology Board. Drinking Water
Distribution Systems: Assessing and Reducing Risks. The
National Academies Press, Washington D.C.
National Sanitation Foundation, 2008. http://www.nsf.org/
consumer/plumbing/index.asp (accessed 18.06.08.).
Odell, L.H., Kirkrmeyer, G.J., Wilczak, A., Jacangelo, J.G.,
Marcinko, J.P., Wolfe, R.L., 1996. Controlling nitrification in
chloraminated systems. Journal of American Water Works
Association 88 (7), 86e98.
Okabe, S., Kinadaichi, T., Itoh, T., 2005. Fate of 14C-labeled
microbial products derived from nitrifying bacteria in
autotrophic nitrifying biofilms. Applied and Environmental
Microbiology 71 (7), 3987e3994.
Oskarsson, A., Norrgren, L., 1998. Copper pipes as a source of
copper exposure in man and environment. Environmental
Reviews 6, 139e150.
Prosser, J.I., 1986. Experimental and theoretical models of
nitrification. In: Prosser, J.I. (Ed.), Nitrification. IRL Press,
Oxford, England, pp. 74e75.
Rahman, S., McDonald, B.C., Gagnon, G.A., 2007. Impact of
secondary disinfectants on copper corrosion under stagnation
conditions. Journal of Environmental Engineering 133 (2),
180e185.
Regan, J. M., 2001. Microbial Ecology of Nitrification in
Chloraminated Drinking Water Distribution Systems, Ph.D
Dissertation, University of Wisconsin, Madison, Wisconsin.
Rittmann, B.E., Regan, J.M., Stahl, D.A., 1994. Nitrification as
a source of soluble organic substrate in biological treatment.
Water Science and Technology 30 (6), 1e8.
Rowe, R., Todd, R., Waide, J., 1977. Microtechnique for most
probable number analysis. Applied and Environmental
Microbiology 33 (3), 675e680.
5521
Sarathy, V., Allen, H.E., 2005. Copper complexation by dissolved
organic matter from surface water and wastewater effluent.
Ecotoxicology and Environmental Safety 62, 337e344.
Sato, C., Leung, S.W., Schnoor, J.L., 1988. Toxic response of
Nitrosomonas europaea to copper in inorganic media and
wastewater. Water Research 22 (9), 1117e1127.
Schock, M., Lytle, D., Clement, J., 1995. Effect of pH, DIC,
orthophosphate and sulfate on drinking water cuprosolvency.
US EPA Report/600/R-95/085 June 1995.
Schramm, A., Larsen, L.H., Revsbech, N.P., Ramsing, N.B.,
Amman, R., Schleifer, K.H., 1996. Structure and function of
a nitrifying biofilm as determined by in situ hybridization and
the use of microelectrodes. Applied and Environmental
Microbiology 62 (12), 4641e4647.
Seidel, C.J., McGuire, M.J., Summers, S., Via, S., 2005. Have utilities
switched to chloramines? Journal of American Water Works
Association 97 (10), 87e97.
Skadsen, J., 1993. Nitrification in a distribution system. Journal of
American Water Works Association 85 (7), 95e103.
Skadsen, J., Cohen, Y.K., 2006. Fundamentals and Control of
Nitrification in Chloraminated Drinking Water Distribution
System. AWWA Manual M56. American Water Works
Association, Denver, CO.
Skinner, F.A., Walker, N., 1961. Growth of Nitrosomonas europaea in
batch and continuous culture. Archives of Microbiology 38,
339e349.
Stewart, M.H., Olson, B.H., 1996. Bacterial resistance to potable
water disinfectants. In: Hurst, C.J. (Ed.), Modeling Disease
Transmission and its Prevention by Disinfection. Cambridge
University Press, Cambridge, England, pp. 140e192.
Stewart, M.H., Lieu, N.I., 1997. Nitrification in chloraminated
drinking water and its association with biofilms. In:
Proceedings of the AWWA Water Quality Technology
Conference, Denver, CO.
Teitzel, G.M., Parsek, M.K., 2003. Heavy metal resistance of biofilm
and planktonic Pseudomonas aeruginosa. Applied and
Environmental Microbiology 69 (4), 2313e2320.
Thurman, R.B., Gerba, C.P., 1989. The molecular mechanisms of
copper and silver ion disinfection of bacteria and viruses.
CRC Critical ReviewdEnvironmental Control 18 (4),
295e315.
Tomlinson, T.G., Boon, A.G., Trotman, C.N.A., 1966. Inhibition of
nitrification in the activated sludge process of sewage
disposal. Journal of Applied Bacteriology 29 (2), 266e291.
USEPA, 1993. Method for the Determination of Inorganic
Substances in Environmental Samples. EPA/600/R-93/100
Washington D.C.
USEPA, 2000. Microbial/Disinfection Byproducts (M-DBP) Federal
Advisory Committee: Stage 2 M-DBP Agreement in Principle.
EPA, Washington, D.C.
Wahman, D.G., Wulfeck-Kleier, K.A., Pressman, J.G., 2009.
Monochloramine disinfection kinetics of Nitrosomonas
europaea by propidium monoazide quantitative PCR and live/
dead BacLight methods. Applied and Environmental
Microbiology 75 (17), 5555e5562.
Wilczak, A., 2006. Overview of causes and control of
nitrification in water distribution systems. Chapter 4. In:
Kosyra, M.K. (Ed.), Fundamentals and Control of
Nitrification in Chloraminated Drinking Water Distribution
Systems, AWWA Manual M56. American Water Works
Association, Denver, CO.
Wilczak, A., Jacangelo, J.G., Marcinko, J.P., Odell, L.H., Kirmeyer, G.
J., Wolfe, R.L., 1996. Occurrence of nitrification in
chloraminated distribution systems. Journal of American
Water Works Association 88 (7), 74e85.
Wolfe, R.L., Lieu, N.I., 2001. Nitrifying bacteria in drinking water.
In: Bitton, G. (Ed.), Encyclopedia of Environmental
Microbiology. John Wiley and Sons, New York, NY.
5522 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 1 1 e5 5 2 2

Wolfe, R.L., Means III, E.G., Davis, M.K., Barrett, S.E., 1988.
Biological nitrification in covered reservoirs containing
chloraminated water. Journal of American Water Works
Association 80 (9), 109e114.
Wolfe, R.L., Lieu, N.I., Izaguirre, G., Means, E.G., 1990. Ammonia-
oxidizing bacteria in chloraminated distribution systems:
seasonal occurrence, distribution and disinfection resistance.
Applied and Environmental Microbiology 56 (2), 451e462.
Zhang, Y., Edwards, M., 2005. Water chemistry and growth of
nitrifying bacteria in drinking water distribution systems. In:
Proceedings of the American Water Works Association Water
Quality Technology Conference, Denver, CO.
Zhang, Y., Griffin, A., Rahman, M., Camper, A., Baribeau, H.,
Edwards, M., 2009. Lead contamination of potable water due to
nitrification. Environmental Science and Technology 43 (6),
1890e1895.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Disinfection of domestic effluents by gamma radiation: Effects

on the inactivation of Ascaris lumbricoides eggs

Gloria S.M.B. de Souza a, Ludmila A. Rodrigues a, Warllem J. de Oliveira a,

Carlos A.L. Chernicharo b,*, Marcos P. Guimarães a, Cristiano L. Massara c, Pablo A. Grossi d
a
Department of Parasitology, Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627, 31270-901 Belo Horizonte, MG, Brazil
b
Department of Sanitary and Environmental Engineering, Universidade Federal de Minas Gerais, Av. Antônio Carlos, 6627,
31270-901 Belo Horizonte, MG, Brazil
c
René Rachou Research Center, Av. Augusto de Lima, 1715, 30190-002 Belo Horionte, MG Brazil
d
Center for Development of Nuclear Technology, CDTN, Av. Antônio Carlos, 6627, 31270-901 Belo Horizonte, MG, Brazil

article info abstract

Article history: This work investigated the inactivation of Ascaris lumbricoides eggs in domestic effluents by
60
Received 2 February 2011 gamma radiation from a Co source. Domestic wastewater was treated in a compact
Received in revised form demo-scale system consisting of a UASB reactor and a trickling filter; treatment was
30 July 2011 carried out at the Center for Research and Training on Sanitation (CePTS), Federal
Accepted 7 August 2011 University of Minas Gerais, in Belo Horizonte-MG, Brazil. One-liter of treated wastewater
Available online 23 August 2011 samples was artificially contaminated with an average of 1000 non-embryonated Ascaris
lumbricoides eggs from human feces; samples were then irradiated in a multiple-purpose
Keywords: irradiator at different doses (0.5e5 kGy). Eggs were recovered from the wastewater and
Disinfection the viability of these irradiated eggs was evaluated; the description of the egg develop-
Gamma radiation mental phases with each dose of gamma radiation was recorded. Radiation doses of
Wastewater 3.5 kGy effectively disinfected effluents with lower concentrations of A. lumbricoides eggs;
Inactivation higher radiation doses of 5 kGy were necessary to disinfect effluents with higher eggs
Eggs Ascaris lumbricoides concentrations.
Viability ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
Although conventional wastewater treatment systems have
the capacity to improve effluent quality, they are not sufficient
to remove all contaminants (Von Sperling and Mascarenhas,
2005). Disinfectant agents act by inducing biochemical
changes in both pathogenic agents and in the effluent.
Different wastewater disinfection methods include chlorina-
tion, ozonation and ultraviolet radiation. However, none of
these methods are effective in removing or inactivating more
resistant pathogens such as nematode eggs (Al-Adawi et al.,
2006; Galal-Gorchev, 1996; Tahri et al., 2010).
To protect the public, prior to its use, treated and dis-
infected effluent should have its contamination potential
monitored by the examination of indicators such as the
presence of coliforms and nematode eggs (WHO, 1989, 2004,
2006; Ayres and Mara, 1996).
A parameter that has not yet been taken into consideration
by legislation, but which is epidemiologically relevant, is the
evaluation of nematode egg viability. The presence of nema-
tode eggs alone does not assess egg viability or infective
potential. Eggs from several nematode species, especially
from geohelminths, undergo a development step in the envi-
ronment that requires favorable temperature, humidity and

* Corresponding author. Tel.: þ55 3134091020; fax: þ55 3134091879.

E-mail address: calemos@desa.ufmg.br (C.A.L. Chernicharo).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.008
5524 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8
oxygenation conditions to become viable and infective. Ascaris
lumbricoides eggs are a good indicator of parasitological
quality; they are the most numerous species present in
wastewater and are more resistant to adverse environmental
situations compared to many other enteric organisms.
Therefore, nematode egg inactivation can be estimated by
viability differences between A. lumbricoides eggs before and
after the wastewater treatment process.
Although currently expensive, gamma radiation use is
justified by its capability of degrading organic matter and
pathogenic organisms while retaining nutrients such as
nitrogen and phosphorus, which are important in the use of
wastewater for agricultural purposes (Bao et al., 2002; Basfar
and Abdel Rehim, 2002; Borrely et al., 1998; Hill, 2003; Horak,
1994; Rawat et al., 1998; Wang and Wang, 2007).
The literature is controversial regarding the effectiveness of
gamma radiation on the inactivation of helminth eggs present
in wastewater and sewage sludge. Higher radiation doses were
necessary in the studies reported by Chmielewski et al. (1995)
and Enigk et al. (1975) cited by Capizzi and Schwartzbrod
(2001), who achieved 100% egg inactivation at 6 kGy and
4.8 kGy doses, respectively. Both irradiated the eggs in sludge,
but neither provided sample characteristics. Chmielewski
et al. (1995) used helminth eggs present in the sludge itself.
Also in the study by Melmed and Comninos (1979) high doses
were tested. They used Ascaris sp. eggs directly from sludge
with an 85% viability percentage and determined that the
irradiation of 1-l sample at a radiation dose of 5 kGy inactivated
99.5% of the eggs. Further, a radiation dose of 10 kGy inacti-
vated 98.6e99.6% of the eggs present in the sludge.
On the other hand, Horak (1994) showed that the devel-
opment of Ascaris suum eggs contained in aerobic sludge was
inhibited after irradiation at a dose of 1.1 kGy. However, in
their study the sludge was inoculated with a concentration of
100e1000 eggs recovered directly from the uterus of an adult
worm, and the initial egg viability ranged between 24.6 and
54.6%. Consistent with this report, Shamma and Al-Adawi
(2002) demonstrated that a gamma radiation dose of 1.5 kGy
inactivated 100% of the A. lumbricoides eggs present in sludge
samples. In this study, samples of sludge without A. lum-
bricoides eggs were inoculated with approximately 100 eggs
previously recovered from raw sewage sludge by the NaNO3
flotation method (WHO, 2004). The egg viability percentage in
the control was 59.6%.
Fig. 1 e Flow sheet, sampling points and main characteristics of the demo-scale UASB/TF system.
The conflicting studies differ on four important aspects of
experimental design: species of helminth eggs used in the
experiments; sources of the helminth eggs; percentages of
viable eggs in the control groups; and type of liquid media for
egg inoculation.
Taking these factors into consideration, this paper inves-
tigates the disinfection of domestic effluents by the action of
gamma radiation using the viability of A. lumbricoides eggs as
an indicator of efficacy.
2. Materials and methods
To clarify the existing differences found in the literature and
to render the experimental conditions more realistic, this
paper outlined its tests under the following conditions:
 Use of eggs of A. lumbricoides, which is the species more
prevalent in domestic wastewater and poses more risks to
human beings.
 Use of A. lumbricoides eggs with a high percentage of viable
eggs (above 80%).
 Inoculation of eggs in treated wastewater samples, to better
represent possible interactions between the eggs and the
medium during irradiation.
2.1. Area of study
The treated wastewater used in the disinfection experiments
was obtained from a demo-scale treatment system comprised
of an Upflow Anaerobic Sludge Blanket (UASB) reactor and
a Trickling Filter (TF) operating in series, as depicted in Fig. 1.
The system was installed at the Center for Research and
Training on Sanitation (CePTS), Belo Horizonte, Brazil (coor-
dinates 19 530 4200 S and 43 520 4200 W, altitude 800 m). The UASB
reactor was fed on raw sewage taken from the Arrudas
wastewater treatment plant, after being submitted to pre-
treatment for solids and grit removal, therefore representing
a typical urban wastewater. The main characteristics and
operational conditions of the treatment units are presented in
Fig. 1; the main physical-chemical characteristics of the raw
sewage and the final effluent from the treatment system are
depicted in Table 1.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8 5525

Table 1 e Physicalechemical characteristics of the raw sewage and the final effluent of the UASB/TF system.
Parameters (mg L1) Raw sewage Final effluent (treated wastewater)

Median Mean Standard deviation Median Mean Standard deviation

COD 405 392 100.1 75 76 21.2

BOD 198 211 44.6 24 23 4.5
TSS 180 172 46.4 38 42 18.1

COD, chemical oxygen demand; BOD, biochemical oxygen demand; and TSS, total suspended solids.

2.2. Preparation and collection of samples
The relatively low concentration of A. lumbricoides eggs in the
wastewater samples being studied required artificial
contamination with eggs in order to obtain results sufficient
for statistical analysis. Eggs were obtained from human fecal
sediments provided by a clinical analysis laboratory. To obtain
concentrations higher than 100 non-embryonated eggs L1
after the recovery procedure, portions of feces containing
approximately 1000 eggs were inoculated into the treated
wastewater samples.
Composite sampling was performed by an automatic
sampler programmed to collect 1-l of wastewater at hourly
intervals for 24 h. The samples were prepared from the
homogenization of the collected wastewater; duplicate 1-l
aliquots for every dose of irradiation were withdrawn.
Subsequently, the samples were artificially inoculated with
A. lumbricoides eggs. The samples were irradiated at a bench
scale. Irradiation was performed at the Gamma Irradiation
Laboratory of the Nuclear Technology Development Center.
This Gamma Irradiation Facility is a Multiple-Purpose Pano-
ramic Irradiator, classified as a Category II and manufactured
by MDS Nordion in Canada; model/series number IR-214 and
type GB-127, equipped with a dry-storage Cobalt-60 source
with maximum radioactive activity of 2200 TBq or 60,000 Ci.
The samples were irradiated at a distance of 15 cm from the
radioactive source where the radiation field was simulta-
neously measured by using Fricke dosimeters under elec-
tronic equilibrium conditions (ASTM-E-1026-95, 2002). Dose
homogenization techniques were taken into account and
decay time corrections were performed to assure a reliable
prediction of the radiation dose to be delivered to the samples.
A total of seven tests were performed in duplicate; thus, a total
of 14 samples were tested for each dose, except for doses of
0.5, 1.0 and 1.5 kGy, which corresponded to preliminary tests
(see Table 3), when only two tests and four samples were
tested for each dose. In parallel, non-irradiated eggs were
incubated as controls. The radiation doses were calculated
according to Eq. (1). Table 2 displays the gamma radiation
doses tested.
D¼RT
where D is the radiation dose (kGy); R, dose rate (Gy/h),
measured by Fricke dosimetry at 15 cm from the radiation
source; T, irradiation time (h).
Eggs were recovered from the wastewater by the adapted
incubation method (Zerbini and Chernicharo, 2001). Briefly, 1-l
of each sample was put at rest for a 24-h period to allow
sedimentation. Supernatants were removed with the aid of
a suction pump (FANEN-Model 089/CAL), leaving 200 mL
volume in the container. Successive additions of saline solu-
tion (0.85% NaCl) to the remaining material, followed by
centrifugation steps, were carried out until obtaining a clear
sediment. Next, eggs were floated with zinc sulfate solution
(ZnSO4$7H2O e Sinth) (1.18 g L1 density) and filtered out with
47 mm diameter, 0.45 mm porosity cellulose ester membranes
(Millipore). The retained material was stored in 0.1 N sulfuric
acid (H2SO4 e Sinth) solution in 50 mL Falcon tubes that were
incubated at 28  C for 28 days and aerated daily (Massara et al.,
1990). After the incubation period, the tubes were centrifuged,
and the supernatants were carefully removed. The remaining
2 mL were thoroughly mixed and counted in a Sedgewick
Rafter chamber with the aid of a microscope (Olympus, model
CX31). Eggs with a formed larva were considered viable. The
other developmental stages (morula and gastrula) were
quantified; these forms were considered non-viable.
The percentage of inactivation of A. lumbricoides eggs in the
samples was obtained using the following equation:
 
Nf
E¼ 1  100 (2)
No
where E is the efficiency of inactivation of A. lumbricoides eggs;
Nf, number of viable A. lumbricoides eggs after irradiation; No,
total number of fertile A. lumbricoides eggs.
The differences between the central tendency measure-
ments (medians) of the percentages of egg inactivation by
gamma ray-mediated disinfection of effluents were analyzed
with the aid of the non-parametric test for multiple compar-
isons among independent samples (Kruskal-Wallis); Statistica
6.0 software was used. For this purpose, the hypothesis of
equality between the central tendencies (medians) is denied
when the p value is lower than a, where a is the level of
significance ( p < 0.05).
Table 2 e Gamma radiation doses.
Gamma radiation dose (kGy) Time spent (min)
(1)
0.5 5.61e6.41
1.0 11.23e12.81
1.5 16.86e19.23
2.5 23.55e24.43
3.5 32.96e33.41
4.5 42.10e42.7
5.0 47.10e47.73
5526 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8
Fig. 2 e Distribution of the developmental stages of
A. lumbricoides eggs following gamma irradiation.
3. Results
3.1. Effects of radiation on egg development
Fig. 2 presents the percentage distributions of the different
developmental stages of the A. lumbricoides eggs in the irradi-
ated wastewater samples and controls. The different doses of
gamma radiation had little effect on egg development up to the
gastrula stage, as 60% of the eggs were able to achieve this stage.
Fig. 3A and B shows that gamma radiation affected A.
lumbricoides egg development and prevented the formation of
infective larva. Non-viable eggs are no longer considered
epidemiologically relevant.
3.2. Efficiency of Ascaris lumbricoides egg inactivation
Table 3 displays the descriptive statistics of the amounts of
total and viable eggs identified in the gamma-irradiated
Fig. 3 e (A, B) Non-viable A. lumbricoides eggs irradiated with 5.0 kGy of gamma radiation.
wastewater samples. Because the initial amount of eggs in
the different experiments was not constant, the samples were
compared from the calculation of the percentage of inactiva-
tion promoted by each dose of gamma radiation. In this way,
the results of the amount of total and viable A. lumbricoides
eggs allowed the percentage of inactivated eggs in each
sample to be calculated by the application of Eq. (2). The
calculation was individually made for each sample, and the
results are presented in Fig. 4. No infective larvae were
observed for the 5.0 kGy dose of gamma radiation.
Figs. 3 and 4 demonstrate that the use of gamma radiation
in the disinfection of domestic effluents is relatively effective,
considering its effect on A. lumbricoides egg development.
3.3. Doses of gamma radiation recommended for
disinfection
The radiation doses were compared according to the statis-
tical test Kruskal-Wallis. The p values calculated when the
medians among percentages of inactivation in the doses of
gamma radiation tested were compared. The hypothesis of
equality between the central tendencies (medians) is denied
when the p value is less than 0.05.
Significant differences between the median percentages of
inactivation of non-irradiated and irradiated samples with
doses above 3.5 kGy were confirmed by the statistical tests
( p < 0.05). No significant difference between the radiation
doses of 3.5 and 5.0 kGy was found ( p > 0.05).
4. Discussion
4.1. Effects of radiation on egg development
Gamma radiation has a short wavelength, on the order of
104 nm, making it highly penetrating. Thus, when gamma
radiation acts on a pathogenic organism, it can cause both
internal and external chemical changes (Hill, 2003; Horak,
1994). In the case of nematode eggs, the gamma irradiation
most likely affects the biomolecules inside the eggs. If the
radiation interacts with organelles such as the centrioles, it
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8 5527

Table 3 e Descriptive statistics of the amount of A. lumbricoides eggs.

Dose 0.0 0.5a 1.0a 1.5a 2.5 3.5 4.5 5.0
(kGy)
Total Viable Total Viable Total Viable Total Viable Total Viable Total Viable Total Viable Total Viable
eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs eggs

Mean 346 297 363 275 371 175 302 55 337 10 384 6 389 1 368 0
(g L1)
Max. 663 556 641 475 527 263 471 101 799 37 862 32 819 10 671 0
(g L1)
Min. 168 121 113 89 194 116 218 27 65 2 225 0 174 0 216 0
(g L1)
SD 197 175 285 213 148 70 115 33 174 9 157 9 180 3 137 0
N 6 6 4 4 4 4 4 4 14 14 14 14 14 14 14 14

SD, standard deviation; N, number of samples.

a Results of doses 0.5, 1.0 and 1.5 kGy refer to preliminary studies and used for the purpose of constructing the dose-response curve.

would damage the mitotic spindle that is essential for
embryonic development. Thus, the radiation would prevent
cell division and impact the initial developmental stages
(morula). However, if the radiation damaged other organelles
such as the ribosome or even the DNA itself, there may be
deficient production of the enzymes required to coordinate egg
development. Therefore, although the cells could divide, the
lack of enzymes would prevent the continuation of egg
development to the larva stage. In this scenario, egg develop-
ment would stop at one of the more advanced stages (gastrula).
As approximately 60% of the recovered eggs were found in the
gastrula stage, these results strongly suggest that the gamma
irradiation damaged the biomolecules and organelles required
for enzyme synthesis and regulation. These effects interrupt
the development of the egg, making it inactive and without
epidemiological importance. No data were found in the liter-
ature that takes into consideration the action of the gamma
radiation on the organelles within internal nematode egg cells.
4.2. Efficiency of Ascaris lumbricoides egg inactivation
a dose between 2.5 and 5 kGy for the inactivation of A. lum-
bricoides eggs in domestic wastewater.
A hypothesis to explain that the doses necessary for
helminth eggs inactivation were higher for wastewater
samples (present study) than for sludge samples (Horak, 1994;
Shamma and Al-Adawi, 2002), which present much higher
solids content and therefore should demand higher radiation
doses, is the origin of the eggs used as an efficiency indicator
in each study, as well as their viability percentage prior to
irradiation with gamma rays. Another hypothesis involves the
type of sample irradiated. It is known that some components,
such as oxygen, which is potentially present in aerobic sludge
samples, promote a synergetic effect with radiation and
increase the action of the gamma radiation (Melmed and
Comninos, 1979). This may explain, at least in some studies
(Horak, 1994), the lower radiation doses reported to achieve
100% inactivation of Ascaris sp. eggs.
4.3. Doses of gamma radiation recommended for
disinfection

Although 84.8% of the eggs in the control samples of this study
were viable, 100% of the A. lumbricoides eggs exposed to 5 kGy
of gamma radiation were inactivated. These results are in
agreement with the results of Bastos (2007), who suggested
Fig. 4 e Dose-response curve regarding the inactivation of
A. lumbricoides eggs following gamma irradiation.
According to the statistical tests, there was no significant
difference among the doses of gamma radiation of 3.5, 4.5 and
5.0 kGy; however, the 5 kGy dose resulted in 100% egg inacti-
vation in each of the seven experiments.
The quality of the effluent, in terms of concentration of
nematode eggs, is directly related to the disease, endemicity
and sanitation of the area as well as to the socioeconomic
situation of the population (WHO, 2004). In this context, this
research suggests that for effluents with lower concentrations
of nematode eggs, such as 10e80 nematode eggs per liter,
a 3.5 kGy dose of gamma radiation could be adequate to
ensure adequate effluent quality. However, in the case of
more concentrated effluents, such as those containing greater
than 500 nematode eggs per liter, the recommended dose
would be 5.0 kGy of gamma radiation for wastewater
disinfection.
Thus, the application of gamma radiation for the disin-
fection of domestic wastewater is promising because it fits
the effluent within the limits recommended by the World
Health Organization for use of treated wastewater (less than
1 egg per liter) (WHO, 1989, 2004, 2006; Ayres and Mara,
1996).
5528 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 3 e5 5 2 8
5. Conclusions
 This work suggests that gamma radiation acts on
the biomolecules and organelles of A. lumbricoides eggs,
affecting protein production. Although this allowed eggs to
develop to more advanced stages, the eggs were ultimately
non-viable and without epidemiological importance. The
radiation dose of 5 kGy hindered the development of infec-
tive larvae, whereas 85% of the non-irradiated eggs (control)
developed into the infective form.
 The radiation gamma doses of 3.5 kGy were effective for the
disinfection of effluents with lower concentrations of
A. lumbricoides eggs, whereas higher radiation doses of 5 kGy
were necessary for the disinfection of effluents with higher
concentrations of eggs.
 Although it is still expensive, the technology for the disin-
fection of sewage with gamma rays is potentially applicable
in the near future, especially in the case of worsening water
shortage.
Acknowledgments
The authors thank the following institutions for the support
provided: Centro de Tecnologia de Desenvolvimento Nuclear e
CDTN; Conselho Nacional de Desenvolvimento Cientı́fico
e Tecnológico e CNPq; Financiadora de Estudos e Projetos e
FINEP (PROSAB); Fundação de Amparo à Pesquisa de Minas
Gerais e FAPEMIG (project no 14.107); Instituto de Análises
Clı́nicas Hermes Pardini; Pro-Reitoria de Pesquisa da Uni-
versidade Federal de Minas Gerais e UFMG; Universidade
Federal de Minas Gerais (Department of Parasitology and
Department of Sanitary and Environmental Engineering).
references
Al-Adawi, M.A., Albarodi, H., Hammoudeh, A., Shamma, M.,
Sharabi, N., 2006. Accelerated larvae development of Ascaris
lumbricoides eggs with ultraviolet radiation. Radiation Physics
and Chemistry 75 (1), 115e119.
ASTM-E-1026-95, 2002. Practice for Using the Fricke Reference
Standard Dosimetry System. American Society for Testing and
Materials, West Conshohocken, PA.; USA.
Ayres, R.M., Mara, D.D., 1996. Analysis of wastewater for use in
agriculture. A Laboratory Manual of Parasitological and
Bacteriological Techniques. WHO e World Health
Organization, Geneva, pp. 1e35.
Bao, H., Liu, Y., Jia, H., 2002. A study of irradiation in the
treatment of wastewater. Radiation Physics and Chemistry 63
(3e6), 633e636.
Basfar, A.A., Abdel Rehim, F., 2002. Disinfection of wastewater
from a Riyadh Wastewater Treatment Plant with ionizing
radiation. Radiation Physics and Chemistry 65 (4, 5), 527e532.
Bastos, G.S.M., 2007. Application of gamma radiation for the
disinfection of sewage treated in UASB reactor. In: 24 Proc.
Brazilian Congress on Sanitary and Environmental
Engineering, ABES (in Portuguese).
Borrely, S.I., Cruz, A.C., Del Mastro, N.L., Sampa, M.H.O.,
Somessari, E.S., 1998. Radiation processing of sewage and
sludge. A review. Progress in Nuclear Energy 33 (1, 2), 3e21.
Capizzi, S., Schwartzbrod, J., 2001. Irradiation of Ascaris ova in
sludge using an electron beam accelerator. Water Research 35
(9), 2256e2260.
Chmielewski, A.G., Zimek, Z., Bryl-Sandelewska, T., Kosmal, W.,
Kalisz, L., Kaimierczuk, M., 1995. Disinfection of municipal
sewage sludges in installation equipped with electron
accelerator. Radiation Physics and Chemistry 46 (4e6),
1071e1074.
Enigk, K., Holl, P., Dey-Hazra, A., 1975. Eradication of parasitic
cysts and eggs in sewage sludge by irradiation with low energy
electrons. Zbl. Bakt. Hyg 161, 61e71.
Galal-Gorchev, H., 1996. Chlorine in water disinfection. Pure and
Applied Chemistry 68 (9), 1731e1735.
Hill, V.R., 2003. Prospects for pathogen reductions in livestock
wastewaters: a review. Critical Reviews in Environmental
Science and Technology 33 (2), 187e235.
Horak, P., 1994. Experimental destruction of ascarid ova in
sewage sludge by accelerated electron irradiation. Water
Research 28 (4), 939e941.
Massara, C.L., Costa, H.M.A., Carvalho, O.S., 1990. The study of
Ascaris lumbricoides in the laboratory. Revista da Sociedade
Brasileira de Medicina Tropical 23 (1), 43e47 (in Portuguese).
Melmed, L.N., Comninos, D.K., 1979. Disinfection of sewage
sludge with gamma radiation. Water SA 5 (4), 153e159.
Rawat, K.P., Sharma, A., Rao, S.M., 1998. Microbiological and
physicochemical analysis of radiation disinfected municipal
sewage. Water Research 32 (3), 737e740.
Shamma, M., Al-Adawi, M.A., 2002. The morphological changes of
Ascaris lumbricoides ova in sewage sludge water treated by gamma
irradiation. Radiation Physics and Chemistry 65 (3), 277e279.
Tahri, L., Elgarrouj, D., Zantar, S., Mouhib, M., Azmani, A.,
Sayah, F., 2010. Wastewater treatment using gamma
irradiation: Tétouan pilot station, Morocco. Radiation Physics
and Chemistry 79 (4), 424e428.
Von Sperling, M., Mascarenhas, L.C.A.M., 2005. Performance of
very shallow ponds treating effluents from an UASB reactor.
Water Science and Technology 51 (12), 83e90.
Wang, J., Wang, J., 2007. Application of radiation technology to
sewage sludge processing: a review. Journal of Hazardous
Materials 143 (1, 2), 2e7.
WHO e World Health Organization, 1989. Health Guidelines for
the Use of Wastewater in Agriculture and Aquaculture.
Geneva. Technical report 778, pp. 1e76.
WHO e World Health Organization, 2004. Integrated Guide to
Sanitary Parasitology. Regional Office for Eastern
Mediterranean. Regional Centre for Environmental Health
Activities, Amman Jordan, pp. 1-120.
WHO e World Health Organization, 2006. Guidelines for the safe
use of wastewater, excreta and greywater. Wastewater use in
agriculture 2, p. 222.
Zerbini, A.M., Chernicharo, C.A.L., 2001. Methodologies for
quantification, identification and analysis of viability of
helminth eggs in raw and treated sewage. In: Chernicharo, C.
A.L. (Ed.), Post Treatment of Effluents from Anaerobic
Reactors. Methodologic Aspects. PROSAB 2 Belo Horizonte, pp.
71e107 (in Portuguese).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Bromate formation in a hybrid ozonation-ceramic membrane

filtration system

Mohammadreza Moslemi a, Simon H. Davies b, Susan J. Masten a,b,*

a
Department of Civil Engineering, McMaster University, Hamilton, Ontario, Canada
b
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824, USA

article info abstract

Article history: The effect of pH, ozone mass injection rate, initial bromide concentration, and membrane
Received 30 March 2011 molecular weight cut off (MWCO) on bromate formation in a hybrid membrane
Received in revised form filtrationeozonation reactor was studied. Decreasing the pH, significantly reduced bromate
16 July 2011 formation. Bromate formation increased with increasing gaseous ozone mass injection
Accepted 8 August 2011 rate, due to increase in dissolved ozone concentrations. Greater initial bromide concen-
Available online 16 August 2011 trations resulted in higher bromate concentrations. An increase in the bromate concen-
tration was observed by reducing MWCO, which resulted in a concomitant increase in the
Keywords: retention time in the system. A model to estimate the rate of bromate formation was
Bromide developed. Good correlation between the model simulation and the experimental data was
Bromate achieved.
Ozonation ª 2011 Elsevier Ltd. All rights reserved.
Ceramic membrane
Water treatment
Modeling

1. Introduction drinking water and eliminates or greatly reduces the need

Ozonation has gained widespread use in drinking water
treatment over the past few decades (Gottschalk et al.,
2000). Recently, a number of researchers have shown that
when ozonation is used in combination with membrane
filtration, problems of fouling resulting from the deposition
of natural organic matter on the membrane surface or
within the membrane pores can be overcome (Van Geluwe
et al., 2011; Kim and Van der Bruggen, 2010; Mozia et al.,
2006; Karnik et al., 2005). In such hybrid systems, aqueous
ozone reacts with and degrades organic foulants accumu-
lated on the membrane surface, thereby decreasing fouling.
In this innovative method, ensuring a minimum dissolved
ozone concentration enables the continuous treatment of
for membrane cleaning procedures (Kim et al., 2009; You
et al., 2007; Mozia et al., 2006; Schlichter et al., 2003).
A major concern in the treatment of bromide-containing
waters using ozonation is the formation of brominated by-
products (Siddiqui et al., 1995). Bromate (BrO
3 ), an inorganic
by-product of ozonation, is a human carcinogen, and the
United States Environmental Protection Agency (USEPA), the
Ontario (Canada) Safe Drinking Water Act, and the European
Union have set the maximum contaminant level (MCL) for
bromate at 10 mg/L (Ontario drinking-water quality standards,
2002; European Union, 1998; USEPA, 1998).
Bromate is formed by the reaction of molecular ozone with
the bromide ion (Br
), which is naturally present in many
waters. The reactions involved are complex and involve both

* Corresponding author. Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48824, USA.
Tel.: þ1 517 355 2254; fax: þ1 517 355 0250.
E-mail address: masten@egr.msu.edu (S.J. Masten).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.015
5530 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4
O3
Br• BrO•
• •
HO HO
BrO2-
O3 O3
Br- HOBr/OBr-
Fig. 1 e Bromate formation e molecular ozone and
hydroxyl radical pathways. Adapted from Pinkernell and
von Gunten (2001).
molecular ozone and OH radicals, as shown in Fig. 1
(Crittenden and Harza, 2005; Legube et al., 2004; Pinkernell
and von Gunten, 2001; Haag and Hoigné, 1983). Bromide ion
reacts with molecular ozone to form hypobromous acid
(HOBr), which is at equilibrium with hypobromite ion (OBr
).
Subsequently, hypobromite ion reacts with ozone to form
bromite ion, which further reacts with molecular ozone to
form bromate. Bromide can also react with the hydroxyl

radical ( OH) to form the bromine radical, which then reacts
with molecular ozone in a complex set of reactions to form
bromate. As such, both direct and indirect pathways play
a role in bromate formation; however the rate constants for
the reaction of hydroxyl radical with bromide are appreciably
greater than that of those involving molecular ozone and it
has been also shown that the contribution of OH to bromate
formation is more significant compared to that of molecular
ozone (Mizuno et al., 2004; von Gunten and Oliveras, 1998). It is
reported that in Milli-Q water, almost 70% of bromate

formation occurs through OH mediated oxidation reactions
and the remaining 30% depends on the molecular ozone
reactions (Ozekin et al., 1998).
While the formation of various bromine containing by-
products in conventional ozonation systems has been
studied (Haag and Hoigné, 1983; Krasner et al., 1993; von
Gunten and Pinkernell, 2000), the formation of such
O3 generator
N2 O2
P Retentate
Feed tank
Main tank
Fig. 2 e Schematic of the experimental setup.
disinfection by-products in a hybrid ozonation-membrane
filtration has not been previously investigated. Hence, it is
O3 crucial to study bromate formation under various operating
BrO3-
conditions in such hybrid systems. In this study, the effect
of pH, inlet ozone mass rate, initial bromide concentration,
and membrane molecular weight cut off on bromate
formation in a membrane filtrationeozonation system was
investigated.
2. Materials and methods
2.1. Experimental setup
Fig. 2 illustrates the schematic of the reactor that was used in
this study. The system was designed to withstand pressures
of up to 550 kPa (80 psi) and pressurized using nitrogen gas.
The experimental apparatus was equipped with two high-
pressure stainless steel tanks with capacities of 5 and 8 L.
The smaller tank was located in the recycle line to supply the
system with the feed water. The second tank operated as
reserve raw water supply to replace the water lost from the
system through permeation and bleeding and was connected
to the main tank through a solenoid valve (Type 6013, Bürkert
Corp., USA), which was programmed to maintain a constant
water level in the main tank. The total water volume and the

pressure inside the reactor were maintained at 1.5 L and
138 kPa (20 psi), respectively, throughout the experiments.
All equipment, tubing and connections were made of ozone
resistant materials and were either Teflon or 316-stainless
steel. Water was circulated in the loop at a flow rate of
200 mL/min using a gear pump (Model 000-380, Micropump
Inc., USA). The water inside the tank was mixed as a result of
the turbulence produced by the jet of water as it was returned
to the tank.
Gaseous ozone was generated from pressurized oxygen gas
(having a purity of 99.999%) using a corona discharge ozone
O3 monitor O3 destruction Vent
Valve
Pump
P Pressure gauge
F P
F Flow meter
Bleed
P
F P F P
Permeate Bleed
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4 5531

generator (Absolute Ozone AE15MC80P, Absolute System Inc.,
Edmonton, AB, Canada). The gaseous ozone concentration
was adjusted by varying the ozone generator’s voltage. Ozone
gas was injected into the water stream flowing in the recir-
culation loop through a Swagelok Tee fitting (Model SS-400-3,
Swagelok, USA). Ozone was transferred from the gas phase to
the aqueous phase, as a result of the hydraulic mixing inside
the reactor. The inlet gas flow rate was adjusted to the desired
values using a digital flow-controller (Model MC-500SCCM-D,
Alicat Scientific Inc., USA). Any gas exhausted from the
system was destroyed by purging the gas through a potassium
iodide (2% KI) solution. A pressure regulator (Model
KCB1F0A2A5P20000, Swagelok, USA) was employed to
monitor and regulate the pressure in the recirculation loop.
The water temperature was kept at 20  0.1  C by recycling
water through a water coil which was installed inside the
main tank. The temperature of the chilled water was
controlled using a refrigerated circulator bath (NESLAB RTE-
10, Thermo Fisher Scientific, USA).
In these experiments, tubular ultrafiltration (UF)
membranes (TAMI North America, QC, Canada) with nominal
molecular weight cutoffs of 5, 8 and 15 kilodaltons (kD) were
used. UF membranes, rather than nanofiltration (NF)
membranes, were chosen because the desired flux can be
achieved at a lower operating pressure with a UF membrane,
than with an NF membrane. Ceramic membranes were used
as they are ozone resistant. The active length and the external
diameter of the employed membranes were 25 cm and 10 mm,
respectively. A stainless steel housing (TAMI North America,
QC, Canada) was used to hold the membrane.
A bleed line was installed in the recycle loop to allow the
retentate to be sampled. The bleed flow rate was set to 5.0 mL/
min using a digital flow-controller (Model LC-50CCM-D, Alicat
Scientific Inc., USA). A digital flow meter (Model L-5LPM-D,
Alicat Scientific Inc., USA) and a digital pressure gauge (Model
2074, Ashcroft Inc., USA) were mounted in the loop to monitor
the water flow rate and pressure within the system. The
permeate flux was continuously measured using a balance
(Adventurer Pro, Ohaus Corporation, USA). Temperature,
pressure, and cross flow rate were continuously recorded
using a data acquisition system (LabView, National Instru-
ments, USA).
of a high performance liquid chromatograph (Model 230,
Varian ProStar, USA) with an anion-exchange column (IonPac
AS23, Dionex Corp., USA) a guard column (IonPac AG23, Dio-
nex Corp., USA), an anion micromembrane suppressor (AMMS
III e 4 mm, Dionex Corp., USA) and a conductivity detector
(CD25A, Dionex Corp., USA). The analysis was carried out in
accordance with the method proposed by Dionex (Application
Note 184, Dionex Corp., USA). A solution of 4.5 mM sodium
carbonate (NaCO3) (99.5%, EMD, USA) and 0.8 mM sodium
bicarbonate (NaHCO3) (99.7%, SigmaeAldrich, Germany) was
used as the mobile phase. The column flow rate was 1 mL/min.
pH was measured using a pH meter (pHTestr 30, Eutech
Instruments, Illinois, USA).
2.3. Reagents
Ultrapure water (Milli-Q water with a resistivity greater than
18 MU) was used to prepare all reagents and solutions. Milli-Q
water was spiked with sodium bromide (NaBr) to obtain the
desired concentrations. Subsequently, the pH of water was
adjusted to the desired value by adding phosphoric acid
(H3PO4) (99.99%, SigmaeAldrich, USA) and/or sodium
hydroxide (NaOH) (98.0þ%, Fluka, Germany). The water was
not buffered, since our experimental results showed that the
pH did not vary over the course of the experiment. Tertiary
butyl alcohol (Fisher Scientific, Germany) was used as
hydroxyl radical scavenger. Bromide and bromate standards
were prepared using sodium bromide (99.995% (metals basis),
Fluka, Germany) and sodium bromate (NaBrO3) (99.7þ%,
Fisher Scientific, USA).
3. Results and discussion
The effects of pH, ozone mass injection rate, initial bromide
concentration, and membrane molecular weight cut off
(MWCO) on bromate formation were investigated. Steady
state bromae concentrations observed under various oper-
ating conditions are presented in Table 1. All bromate
concentrations reported are averages of triplicate analyses
within experiments. In all cases, the relative standard devia-
tion for the measurements was less than 3.6%.

2.2. Analytical methods

The concentrations of ozone in the permeate and the reten- Table 1 e Steady state bromate concentration in the
tate were continuously monitored by measuring the absor- permeate under various operating conditions.
bance at a wavelength of 258 nm using flow-through cells
pH Inlet O3 Initial MWCO Temperature Steady
mounted in a UV/Vis spectrophotomer (Model 4054, Phar- mass [Br
] (kD) ( C) state [BrO

3]
macia LKB, Biochrom Ltd., UK). The aqueous ozone concen- (mg/min) (mg/L) (mM)
tration was then determined using an extinction coefficient
3 1.5 1.00 5 20 0.58
of 2900 M
1 cm
1 (Hoigné and Bader, 1976). The inlet ozone
6 1.5 1.00 5 20 4.13
gas concentration was also continuously measured using 8 1.5 1.00 5 20 8.03
a UV ozone monitor (Model 454H, Teledyne Instruments, 6 0.5 1.00 5 20 1.57
USA). To quench further ozone reactions, the residual dis- 6 1.0 1.00 5 20 2.86
solved ozone in the samples for bromide and bromate anal- 6 0.75 1.00 5 20 2.55
ysis were purged by bubbling nitrogen gas through the 6 2.25 1.00 5 20 4.67
6 1.5 0.05 5 20 0.055
sample for 2 min.
6 1.5 0.25 5 20 0.72
Bromide and bromate concentrations in the samples were 6 1.5 0.50 5 20 1.75
measured in triplicate using ion chromatography, consisting
5532 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4

10 5
Inlet O3 mass=2.25 mg/min
9 Inlet O3 mass=1.5 mg/min
Inlet O3 mass=1.0 mg/min
4 Inlet O3 mass=0.75 mg/min
8
[BrO3-] (micromole/L)

Inlet O3 mass= 0.5 mg/min

[BrO3-] (micromole/L)
7

6 3

4 2
pH=8.0
3
pH=6.0
1
2 pH=3.0

1
0
0
0 30 60 90 120 150 180
0 30 60 90 120 150 180
Time (min)
Time (min)
Fig. 5 e Bromate concentration in the permeate vs. time at
Fig. 3 e Bromate concentration in the permeate vs. time at
various inlet O3 mass injection rates. Membrane: 7 channel
various pH values. Membrane: 7 channel e 5 kD, inlet O3
e 5 kD, initial [BrL]: 1 mg/L, temperature: 20  C, pH: 6.0,
mass injection rate: 1.5 mg/min, initial [BrL]: 1 mg/L,
ionic strength: 13.0 mM.
temperature: 20  C, ionic strength: 512.5, 13.0 and 12.5 mM
corresponding to pH 3, 6 and 8, respectively.

is an effective way to control bromate formation, doing so is

3.1. Effect of pH neither cost effective nor practical in terms of full-scale
drinking water treatment (Crittenden and Harza, 2005).
As shown in Fig. 3, the steady state concentration of bromate
ion measured in the permeate line at a reactor pH of 3.0 is
3.2. Effect of ozone mass injection rate
significantly less then that observed at pH 6.0 or 8.0. This is not

surprising since the hydroxyl radical ( OH) is more reactive
The effects of the ozone mass injection rate (0.50, 0.75, 1.00, 1.50,
with bromide than is the ozone molecule (von Gunten and
2.25 mg/min) on the dissolved ozone and bromate concentration
Oliveras, 1998) and at low pH, the ozone molecule is more
are shown in Figs. 4 and 5, respectively. The aqueous ozone
stable and the concentration of hydroxyl radicals in the
concentrations increased with time until steady state conditions
reactor would be expected to be less than that found at higher
were attained. The steady state aqueous ozone concentration
pH (Pinkernell and von Gunten, 2001; von Gunten, 2003).
([O3]ss) was found to be proportional to the ozone mass injection
Furthermore, HOBr is much less reactive with ozone than is
rate (see inset in Fig. 4), which is also consistent with the exper-
OBr
(Mizuno et al., 2004). The pKa for the dissociation of
imental results presented by Moslemi et al. (2010). Higher ozone
hypobromous acid to hypobromite (HOBr 4 OBr
þ Hþ) is 8.8
mass injection rates resulted in greater bromate concentrations
(Haag and Hoigné 1983; Siddiqui and Amy, 1993), so the
in the permeate. As shown in Fig. 6, there is a strong linear
concentration of OBr
is low at low pH, hence the rate of
relationship between the steady state bromate concentration in
formation of bromate is slower at low pH. While pH reduction
the permeate and the steady state dissolved ozone concentration

14
Inlet O3 mass= 2.25 mg/min
Inlet O3 mass=1.5 mg/min 5
12 Inlet O3 mass=1.0 mg/min
[O ]

R = 0.933 Inlet O3 mass=0.75 mg/min

Inlet O3 mass=0.5 mg/min y = 0.480x + 0.114
[BrO3 ]ss (micromole/L)

10 4 2
Aqueous [O ] (mg/L)

R = 0.996
Inlet O mass
8
3

6
2
-

1
2

0 0
0 30 60 90 120 150 180 0 1 2 3 4 5 6 7 8 9 10
Time (min) Aqueous [O3]ss (mg/L)

Fig. 4 e Dissolved ozone concentration in the permeate vs. Fig. 6 e Steady state bromate concentration in the
time at various inlet O3 mass injection rates. Membrane: permeate vs. steady state aqueous O3 concentration.
7 channel e 5 kD, initial [BrL]: 1 mg/L, temperature: 20  C, Membrane: 7 channel e 5 kD, initial [BrL]: 1 mg/L,
pH: 6.0, ionic strength: 13.0 mM. temperature: 20  C, pH: 6.0, ionic strength: 13.0 mM.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4 5533

5
Table 3 e Retention time in the membrane for various
MWCOs.
[BrO3 ]ss (micromole/L)

4
MWCO Vmp Wm (g) Vmd VP J (mL/s) q (s)
y = 4.186x - 0.176 (kD) (mL) (mL) (mL)
2
3 R = 0.992
5 14.1 38.7 10.2 3.96 0.116 34
8 14.1 38.8 10.2 3.93 0.169 23
2 15 14.1 40.2 10.6 3.57 0.198 18
-

0 there is a strong linear correlation (r2 ¼ 0.999) between the

0 0.25
Initial [Br ] (mg/L)
Fig. 7 e Steady state bromate concentration in the
permeate vs. initial bromide concentration. Membrane:
7 channel e 5 kD, inlet O3 mass injection rate: 1.5 mg/min,
temperature: 20  C, pH: 6.0.
(r2 ¼ 0.996). This result shows the importance of developing
processes that effectively minimize DBP formation by operating
at the lowest possible dissolved ozone concentrations.
3.3. Initial bromide concentration
In another set of experiments, the initial bromide concentra-
tion was varied (0.05, 0.25, 0.5 or 1.0 mg/L) (see Fig. S1). As it is
shown in Fig. 7, there is a strong correlation between the
steady state bromate concentration in the permeate and the
initial bromide concentration in the feed solution (r2 ¼ 0.992)
indicating that as more bromide ion was available to react
with molecular ozone and/or hydroxyl radicals, more bromate
was produced.
3.4. Effect of MWCO
As presented in Table 2, lower bromate concentrations were
observed with membranes having higher MWCOs (also see
Fig. S2). The permeability of the membranes tested increases
with MWCO, so at a constant operating pressure, as the
MWCO of the membrane increases, the permeate flux also
increases, and the residence time inside the reactor decreases.
Therefore, for membranes with greater MWCOs, the ozone
exposure (the dissolved ozone concentration multiplied by
0.5 0.75 1 steady state bromate concentration and the retention times in
-
the membrane (r2 ¼ 0.999) and the entire system (r2 ¼ 0.993).
This implies that in hybrid systems, bromate formation is
significantly affected by the contact times in the membrane
and the entire system, although the latter likely predominates
simply because of its greater magnitude. These results suggest
that bromate formation can be minimized by employing
membranes with the largest MWCO feasible for the particular
operation.
4. Bromate formation model
A kinetic-based model was developed to predict the rate of
bromate formation in the hybrid ozone-membrane system.
Experimental results reported herein were used to calculate
the reaction rate constant. Bromate formation in the
employed reactor can be expressed using the following mass
balance equation:
dCR
V ¼ Qin Cin
QP CP
QB CR þ kCR V (1)
dt
where V, total volume of water in the system (1500 mL); t, time
(min); CR, concentration of bromate in the system (retentate)
(mM); Cin, concentration of bromate in the inflow into the
system (mM); CP, concentration of bromate in the permeate
(mM); Qin, inlet flow rate (mL/min); QP, permeate flow rate (mL/
min); QB, bleed flow rate (mL/min); k, rate constant of bromate
formation (1/min).
5
y = 0.090x + 1.060
[BrO3 ]ss (micromole/L)

4 2
retention time) is reduced. The retention times for each R = 0.999

system and the three different membranes were calculated 3

(see Supporting Documentation for details). The retention
times for the membranes with MWCOs of 5, 8, and 15 kD were
2
-

34, 23, and 18 s, respectively (Table 3). As can be seen in Fig. 8

Table 2 e Steady state [BrOL

3 ] at various MWCOs.
0
0 10 20 30 40
MWCO pH Inlet O3 Initial Steady state Retention time (s)
(kD) mass (mg/min) [Br
] (mg/L) [BrO

3 ] (mM)
Fig. 8 e Steady state bromate concentration in the
5 6.0 1.5 1.00 4.13
permeate vs. retention time in the membrane. Membrane:
8 6.0 1.5 1.00 3.18
7 channel, inlet O3 mass: 1.5 mg/min, initial [BrL]: 1 mg/L,
15 6.0 1.5 1.00 2.67
temperature: 20  C, pH: 6.0, ionic strength: 13.0 mM.
5534 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 2 9 e5 5 3 4
Table 4 e Rate constants for the formation of bromate in
the hybrid ozonation-membrane system.
Inlet O3 mass Rate constant Relative standard
(mg/min) (min
1) deviation (%)
0.5 0.0061 4.8
0.75 0.0063 6.7
1.0 0.0064 2.4
1.5 0.0067 4.3
2.25 0.0068 5.6
At steady state, the bromate concentration in the system
will not vary with time (dC/dt ¼ 0). Moreover, the feed water
was free of bromate ion (Cin ¼ 0). Hence, after substitution into
Eq. (1), the rate constant of bromate formation can be obtained
using the following equation.
QP CP þ QB CR
k¼
CR V
Using Eq. (2), the rate constants for the experiments with
varying ozone mass injection rate were calculated (see Table
4). At a pH of 6.0, the rate constants do not vary significantly
with ozone mass injection rate. Hence, one can assume that
the mechanism for bromate formation is independent of
ozone mass injection rate at a constant pH. At an ozone mass
injection rate of 1.5 mg/min, the rate constants for bromate
formation at pH 3, 6, and 8 were determined to be 0.0061,
0.0067, and 0.0085 min
1, respectively. Under these condi-
tions, the aqueous ozone concentrations were 9.6, 8.3, and
4.9 mg/L, for solutions at pH values of 3, 6, and 8, respectively.
As discussed earlier, the increased overall rate of reaction is
likely due to an increase in the OH radical concentration at
higher pH and the dissociation of hypobromous acid to OBr
,
which is more reactive with ozone than is HOBr.
Acknowledgements
The authors acknowledge financial support of this work from
the National Science Foundation Research (Grant No. CBET-
0506828) and from the Natural Sciences and Engineering
Research Council of Canada Discovery Grant Program.
Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.015.
references
Crittenden, J., Harza, M.W., 2005. In: Water Treatment: Principles
and Design, second ed. MWH, John Wiley, New Jersey.
European Union, 1998. Official Journal of the European
Communities, L 330, Council Directive 98/83/EC.
Gottschalk, C., Libra, J.A., Saupe, A., 2000. Ozonation of Water and
Waste Water: A Practical Guide to Understanding Ozone and
Its Application. Wiley-VCH, New York.
von Gunten, U., Oliveras, Y., 1998. Advanced oxidation of
bromide-containing waters: bromate formation mechanisms.
Environ. Sci. Technol. 32, 63e70.
von Gunten, U., Pinkernell, U., 2000. Ozonation of bromide-containing
drinking waters: a delicate balance between disinfection and
bromate formation. Water Sci. Technol. 41, 53e59.
von Gunten, U., 2003. Ozonation of drinking water: part II.
Disinfection and by-product formation in presence of
bromide, iodide or chlorine. Water Res. 37, 1469e1487.
Haag, W.R., Hoigné, J., 1983. Ozonation of bromide-containing
waters: kinetics of formation of hypobromous acid and
bromate. Environ. Sci. Technol. 17, 261e267.
Hoigné, J., Bader, H., 1976. The role of hydroxyl radical reactions
in ozonation processes in aqueous solutions. Water Res. 10,
377e386.
Karnik, B.S., Davies, S.H.R., Chen, K.C., Jaglowski, D.R., Baumann, M.
J., Masten, S.J., 2005. Effects of ozonation on the permeate flux of
nanocrystalline ceramic membranes. Water Res. 39, 728e734.
Kim, J., Van der Bruggen, B., 2010. The use of nanoparticles in
(2) polymeric and ceramic membrane structures: review of
manufacturing procedures and performance improvement for
water treatment. Environ. Pollut. 158, 2335e2349.
Kim, J., Shan, W., Davies, S.H.R., Baumann, M.J., Masten, S.J.,
Tarabara, V., 2009. Interactions of aqueous NOM with
nanoscale TiO2: implications for ceramic membrane
filtrationeozonation hybrid process. Environ. Sci. Technol. 43,
5488e5494.
Krasner, S.W., Glaze, W.H., Weinberg, H.S., Daniel, P.A., Najm, I.N.
, 1993. Formation and control of bromate during ozonation of
waters containing bromide. J. AWWA 85, 73e81.
Legube, B., Parinet, B., Gelinet, K., Berne, F., Croue, J.P., 2004.
Modeling of bromate formation by ozonation of surface
waters in drinking water treatment. Water Res. 38, 2185e2195.
Mizuno, T., Yamada, H., Tsuno, H., 2004. Formation of bromate
ion through a radical pathway in a continuous flow reactor.
Ozone Sci. Eng. 26, 573e584.
Moslemi, M., Davies, S.H., Masten, S.J., 2010. Ozone mass transfer
in a recirculating loop semibatch reactor operated at high
pressure. Adv. Oxidation Technol. 13, 79e88.
Mozia, S., Tomaszewska, M., Morawski, A.W., 2006. Application of
an ozonation-adsorption-ultrafiltration system for surface
water treatment. Desalination 190, 308e314.
Ontario drinking-water quality standards, 2002. Safe Drinking
Water Act, Ontario regulation 169/03.
Ozekin, K., Westerhoff, P., Amy, G.L., Siqqiqui, M., 1998. Molecular
ozone and radical pathways of bromate formation during
ozonation. J. Environ. Eng. 124, 456e462.
Pinkernell, U., von Gunten, U., 2001. Bromate minimization
during ozonation: mechanistic considerations. Environ. Sci.
Technol. 35, 2525e2531.
Schlichter, B., Mavrov, V., Chmiel, H., 2003. Study of a hybrid
process combining ozonation and membrane filtration -
filtration of model solutions. Desalination 156, 257e265.
Siddiqui, M.S., Amy, G.L., 1993. Factors affecting DBP formation
during ozone-bromide reactions. J. AWWA 85, 63e72.
Siddiqui, M.S., Amy, G.L., Rice, R.G., 1995. Bromate ion formation:
a critical review. J. AWWA 87, 58e70.
USEPA, 1998. National primary drinking water regulations:
disinfection and disinfection byproducts. Federal Reg. 63,
69390e69476.
Van Geluwe, S., Braeken, L., Van der Bruggen, B., 2011. Ozone
oxidation for the alleviation of membrane fouling by natural
organic matter: a review. Water Res. 45, 3551e3570.
You, S.H., Tseng, D.H., Hsu, W.C., 2007. Effect and mechanism of
ultrafiltration membrane fouling removal by ozonation.
Desalination 202, 224e230.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Prediction of contamination potential of groundwater arsenic

in Cambodia, Laos, and Thailand using artificial neural
network

Kyung Hwa Cho a,1, Suthipong Sthiannopkao b,1, Yakov A. Pachepsky c,

Kyoung-Woong Kim d, Joon Ha Kim d,e,*
a
Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, 130 Natural Resources Road,
Amherst, MA 01003, USA
b
Department of Environmental and Occupational Health, National Cheng Kung University, Tainan City, Taiwan, ROC
c
USDA-ARS, Environmental Microbial & Food Safety Laboratory, 10300 Baltimore Ave., Beltsville, MD 20705, USA
d
School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), 261 Cheomdan-gwagiro,
Buk-gu, Gwangju 500-712, Republic of Korea
e
Sustainable Water Resource Technology Center (SWRTC) at GIST, 261 Cheomdan-gwagiro, Buk-gu, Gwangju 500-712, Republic of Korea

article info abstract

Article history:
Received 10 February 2011
Received in revised form
29 May 2011
Accepted 8 August 2011
Available online 27 August 2011
Keywords:
Multiple linear regression
Principal component regression
Artificial neural network
Principal component-artificial
neural network
The arsenic (As) contamination of groundwater has increasingly been recognized as
a major global issue of concern. As groundwater resources are one of most important
freshwater sources for water supplies in Southeast Asian countries, it is important to
investigate the spatial distribution of As contamination and evaluate the health risk of As
for these countries. The detection of As contamination in groundwater resources, however,
can create a substantial labor and cost burden for Southeast Asian countries. Therefore,
modeling approaches for As concentration using conventional on-site measurement data
can be an alternative to quantify the As contamination. The objective of this study is to
evaluate the predictive performance of four different models; specifically, multiple linear
regression (MLR), principal component regression (PCR), artificial neural network (ANN),
and the combination of principal components and an artificial neural network (PC-ANN) in
the prediction of As concentration, and to provide assessment tools for Southeast Asian
countries including Cambodia, Laos, and Thailand. The modeling results show that the
prediction accuracy of PC-ANN (NasheSutcliffe model efficiency coefficients: 0.98 (traning
step) and 0.71 (validation step)) is superior among the four different models. This finding
can be explained by the fact that the PC-ANN not only solves the problem of collinearity of
input variables, but also reflects the presence of high variability in observed As concen-
trations. We expect that the model developed in this work can be used to predict As
concentrations using conventional water quality data obtained from on-site measure-
ments, and can further provide reliable and predictive information for public health
management policies.
ª 2011 Elsevier Ltd. All rights reserved.

* Corresponding author. School of Environmental Science and Engineering, Gwangju Institute of Science and Technology (GIST), 261
Cheomdan-gwagiro, Buk-gu, Gwangju 500-712, Republic of Korea. Tel.: þ82 62 970 3277; fax: þ82 62 970 243.
E-mail address: joonkim@gist.ac.kr (J.H. Kim).
1
Cho and Sthiannopkao contributed equally to this work and are listed in alphabetical order.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.010
5536 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4
1. Introduction
Groundwater resources are the most important components
of drinking water supplies for Southeast Asian countries (Berg
et al., 2007). Arsenic (As) contamination of groundwater has
become a major problem on a worldwide scale because As is
a carcinogenic element, which is mostly present as an inor-
ganic species in natural water systems (Bagla and Kaise, 1996;
AWWA, 2001; Berg et al., 2006; Kocar and Fendorf, 2009;
Sthiannopkao et al., 2010). Naturally occurring As-enriched
groundwater has been observed in tube well/hand pump
drinking water supplies in South and Southeast Asia,
including Thailand, Vietnam, Loa PDR, Cambodia, Myanmar,
Bangladesh, India, Nepal, and Pakistan (Berg et al., 2001a,b;
Smedley and Kinniburgh, 2002; Sun et al., 2002; Polya et al.,
2003, 2005; Stanger et al., 2005; Kohnhorst, 2005; Tetsuro
et al., 2006; Chiew et al., 2009). It is estimated that about 200
million people are at risk of harmful toxic effects of As in these
countries (Sun et al., 2002).
In Southeast Asian countries, monitoring strategies of As
contamination need to be improved in order to quantify As
concentrations in groundwater, which can then be used to
provide further information to better assess and manage
public health. The detection of As contamination of ground-
water resources, however, is hampered by a substantial labor
and cost burden for Southeast Asian countries; it requires
sophisticated equipment, highly skilled technicians, and
incurs a high maintenance cost. As such, dealing with local
As-associated problems in Cambodia and Laos remains
problematic, where facilities for determining arsenic
concentrations as well as human resources and funding for
analyzing arsenic are still very much lacking.
Therefore, modeling approaches for As concentrations
using on-site measurement data can be an alternative to
characterizing the As contamination potential, to provide
predictive information for better public health management.
Indeed, obtaining on-site data such as pH, redox potential
(Eh), and salinity are less expensive and much easier to
perform than to detect and measure As contamination via
graphite atomic absorption spectrophotometry (AAS) or
inductively coupled plasma mass spectroscopy (ICP-MS). It is
thought that these simple measurements can be used to fill
the gap incurred by the missing facilities, human resources,
and funding for arsenic determination using sophisticated
equipment.
The most popular modeling approaches currently in use
are multiple linear regression (MLR) and principal component
regression (PCR), both of which attempt to find the most
appropriate predictive model by fitting a linear equation to
multiple observed data (explanatory variables). However, MLR
and PCR may not be successful due to their statistical
assumptions, including absence of outliers, normality, and
randomness, which can be easily violated (Gros, 1997). In
particular, the existence of correlations among explanatory
variables (referred to here as “collinearity”) diminishes the
statistical stability (or robustness) and may cause significantly
high prediction errors (Mac Nally, 2002). Moreover, incorpo-
ration of too many redundant (or insignificant), or too few,
explanatory variables into models is not practical; thus,
a stepwise regression technique is often required to deter-
mine the optimal number of variables to use as explanatory
variables (over- or underspecification) (Luan et al., 2008).
Furthermore, even though the predictive power of MLR and
PCR can be acceptable, the models often result in the poor
accuracies when making predictions with new datasets in the
validation step.
Therefore, more complex nonlinear models have been
developed that achieve better predictions, and have subse-
quently been applied to water resources problems such as
hydrological processes, water quality problems, and dam
operations (Maier and Dandy, 1996; Wen and Lee, 1998; Lee
et al., 2003; Riad et al., 2004; Sarangi and Bhattacharya, 2005;
Tayfur et al., 2005; Holmberg et al., 2006; Kuo et al., 2008).
These nonlinear models such as artificial neural networks
(ANNs), however, are not only difficult to construct, but also
often cause over-fit problems in predictions; furthermore, the
straightforward interpretation of relationships between
explanatory and dependent variables cannot be achieved by
nonlinear models. A few researchers have applied ANN to the
prediction of As in groundwater (Purkait et al., 2008; Chang
et al., 2010). For example, Purkait et al. (2008) used conven-
tional water quality parameters to predict As contamination
in Eastern India in an attempt to select the best model among
linear and nonlinear models. Chang et al. (2010) applied an
ANN model to recover missing values in As concentration
datasets in an area of Taiwan.
This study utilizes a comprehensive dataset from three
different countries, having a wide range of As concentration
levels and conventional water quality parameters. The
objective of this study is to analyze and evaluate the predictive
performance of MLR, PCR, ANN, and the sequential combi-
nation of principal component and ANN (PC-ANN) in the
prediction of As concentrations in groundwater and thereby
provide improved assessment tools for Southeast Asian
countries. In addition, a sensitivity analysis is also applied to
investigate the cause and effect relationship between input
parameters and As concentrations.
2. Materials and methods
2.1. Field and sampling sites
Groundwater samples were taken in an attempt to investigate
the As concentration and five in-situ parameters for three
countries: Cambodia, Loas, and Thailand (see Fig. 1). Table 1
shows the information on the timing and study sites. In
Cambodia, thirty groundwater samples were collected from
six villages in Kandal Province in 2008 ((Prek Thom village,
Kbal Kaoh commune, n ¼ 5), (Phoum Thom village, Phoum
Thom commune, n ¼ 5), (Chounlork village, Korki commune,
n ¼ 5), (Tuol Tnort village, Koki commune, n ¼ 5), (Doun Sor
village, Koki commune, n ¼ 5), (Poul Pear Ker village, Khom
Day Eth commune, n ¼ 5)). In 2010, 49 samples were taken
from south of Phnom Penh. In Laos, a total of 62 tube well
samples were collected in 2008 from households located in
provinces of Champasack (n ¼ 27), Attapeu (n ¼ 10), Saravane
(n ¼ 11), Savannakhet (n ¼ 4), Borikhamxay (n ¼ 7), and
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4 5537

Vientiane (n ¼ 3). Vientiane is located in the upstream of the

Table 1 e Timing and sites of Arsenic sampling in
Mekong River, followed by Borikhamxay, Savannakhet, Sar- Southeast Asia.
avane, Champasack and Attapeu, respectively. The areas
Study Sampling Number of Number of
along the Mekong River located in middle and southern parts
area time samples villages
of Laos starting from Borikhamxay downward are floodplain or provinces
areas. In Thailand, the concentrations of As in groundwater
Cambodia 2008 30 6
were examined in Tambon Ongphra (n ¼ 10) in Suphan buri
2010 49 1
province in 2008. In the three countries, samples for total As
Laos 2008 62 6
were collected from tube wells by following this sequence: 1) Thailand 2008 10 1
pumping the tube well for several minutes, 2) washing a clean
polyethylene bottle with the well water, and 3) taking water
without filtering. relative standard deviation of the measurement exceeded 5%.
Dilution was made with 2% HNO3 when the concentration of
2.2. Sample analysis and on-site measurement the sample was over the upper limit of the standard range
(100 mg L1). During the sample collection in three countries,
As concentrations in groundwater samples were measured by a series of in-situ measurements were conducted: pH, Eh,
graphite atomic absorption spectrophotometry (GF-AAS; Per- water temperature (Wt) (HORIBA d-54 meter), electrical
kin Elmer 5100 PC, with a detection limit of 0.5 mg L1). GF-AAS conductivity (EC), and total dissolved solids (TDS) (ORION 3
was calibrated by an external standard technique in the range STAR, Thermo Electron Corporation).
of 0e100 mg L1. For a quality control, standard reference
material (SRM) for natural water (National Institute of Stan- 2.3. Modeling approaches: MLR, PCR, ANN, and PC-
dards & Technology NIST 1640) was used to assure the preci- ANN
sion of the measurement. After every tenth sample during
analysis, the SRM sample and calibration standards were All predictive models were developed using on-site data for
analyzed to check the analysis accuracy. All samples were pH, EC, TDS, Wt, and Eh. These variables were logarithmically
measured at least twice in order to assess the measurement transformed in order to normalize them for four different
reliability; samples were reanalyzed if the error either from models (Rawlings et al., 1998). A pattern search algorithm and
the SRM or from the calibration standards exceeded 10% or the the Latin hypercube-one-factor-at-a-Time (LH-OAT) method
were used to optimize parameters and investigate parameter
sensitivity analysis of ANN, respectively (see Supplementary
Information). Strategies of training, validation, and testing
for ANN are also addressed in Supplementary Information.

Table 2 e Results of MLR and PCR for As concentration of

groundwater.
Regression coefficients Collinearity statistics

Independent bia Std. VIF Mean VIF

variable (i) Error SEbi

(Constant)b 2.83 2.73 2.02

pH 2.60 0.89 1.14
Electrical 0.00 0.05 2.67
conductivity
Total dissolved 0.23 0.04 2.83
solids
Temperature 1.62 1.74 1.16
Redox potential 0.21 0.19 2.29

(Constant)b 1.80 0.04 0.99

PC1 0.18 0.04 0.98
PC2 0.13 0.04 0.99
PC3 0.01 0.04 0.99
PC4 0.16 0.06 1.00
PC5 0.34 0.04 0.98

a The subscript i indicates the water quality parameter, and bi the

computed coefficients of the water qualities in the MLR and PCR
models,
log(As) ¼ b0 þ b1  pH þ b2  Conductivity þ b3  TDS þ b4  Water
temperature þ b5  Redox, log(As) ¼ b0 þ b1  PC1 þ
b2  PC2 þ b3  PC3 þ b4  PC4 þ b5  PC5.
Fig. 1 e Map of study the site, showing sampling stations
b (Constant) in the table represents b0 in the MLR and PCR models.
and Mekong River in Cambodia, Laos, and Thailand.
5538 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4
2.3.1. Multiple linear regression (MLR)
The MLR model was developed as follows:
X
n
logðyÞ ¼ b0 þ bi logðxi Þ
i¼1
where xi is the explanatory variable i, y is the dependent vari-
able, bi is the regression coefficient of explanatory variables i,
and b0 is the value of the intercept in the log-linear fitting.
2.3.2. Principal component regression (PCR)
PCR combines a principal component analysis (PCA) decom-
position with MLR (Jolliffe, 2002; Cho et al., 2009). As a result of
PCA, a new set of variables (the principal components (PCs))
and PC scores are generated from the orthogonal linear
transform of the original data. The PC scores are then used in
the regression as explanatory variables.
2.3.3. Artificial neural network (ANN)
ANN is a useful method for determining pattern classifica-
tions of multi-variable datasets as well as the prediction of
complex processes. The multilayer perceptron ANN consists
of two or more layers of nodes, including an input layer,
a hidden layer, and an output layer, which are connected by
links with varying weights. The nodal data are multiplied by
the weights to compute the signal strength, and then are
transferred to the next node in the network; the input layer
nodes accept the input vectors and forward the signals to the
next layer according to the connection. This process is
continued until the signals reach the output layer.
2.3.4. Principal component-artificial neural network (PC-
ANN)
PC-ANN merges PCA decomposition with ANN (Sousa et al.,
2007). The main difference between this approach and ANN
is that PC scores generated from the orthogonal linear trans-
formation of the original data are used as the input variables
Fig. 2 e Mean values of As concentration, pH, redox potential, total dissolved solids, electrical conductivity, and temperature
along with their standard deviations for Cambodia, Laos, and Thailand.
of ANN; other procedures for the optimization, training, and
validation are the same as those for the ANN model.
(1)
3. Results and discussion
3.1. Relationship between As concentrations and on-site
measurements
Fig. 2 shows the average values and standard deviations of As
concentrations and conventional water quality parameters in
Cambodia, Laos, and Thailand. Arsenic concentrations of
groundwater were observed to be the highest in Cambodia
and the lowest in Laos. Similarly, pH was the highest in
Cambodia, and the lowest in Laos. Conversely, Eh was the
highest in Laos, and the lowest in Cambodia. Significant
positive (Pearson correlation: 0.25; p-value: 0.00) and negative
correlations (Pearson correlation: 0.10, p-value: 0.12) were
found between the As concentrations and pH and between the
As concentrations and Eh, respectively. Even though the p-
value for Eh is greater than 0.05, the presence of this correla-
tion is consistent with previous studies (Berg et al., 2007;
Buschmann et al., 2007), indicating that high arsenic concen-
trations might be triggered by reducing conditions.
3.2. Linear models for predicting As concentrations
3.2.1. MLR for predicting As concentrations
Fig. 3(A) and (B) respectively shows the observed and the
predicted As concentrations of groundwater in both the
generation and validation steps of two different regression
models (i.e., MLR and PCR). Overall, the MLR model did not
reproduce the variations of observed As concentrations in
either the generation or validation steps; the developed MLR
model tends to underestimate As concentrations. Table 2
shows the regression coefficients bi, the corresponding
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4 5539

Fig. 3 e Comparison results between the observed and predicted As concentrations of groundwater, (A) and (B): MLR and
PCR, (C) and (D): ANN and PC-ANN, closed circles: MLR, open circles: PCR, closed squares: ANN, open squares: PC-ANN.

standard error SEbi ; and the collinearity statistics of the MLR 3.2.2. PCR for predicting As concentrations
model. Note that if the absolute value of bi in the MLR model is In the PCR model, five PCs were generated from five water
greater than twice its standard error (i.e., SEbi ), the ith variable quality parameters and then used as input variables for
can be regarded as a significant variable (Rawlings et al., 1998).
Here, TDS and pH have regression coefficients greater than
twice their standard errors.
Table 3 e MAEs and NasheSutcliffe model efficiency
Furthermore, collinearity was found among the explanatory
coefficients of the model predictions for As.
variables because the largest variance inflation factor (VIF) was
Generation/Training steps Validation step
greater than 10, with an average VIF value being substantially
greater than 1 (Bowerman and O’Connell, 1990; Myers, 1990). MAEa NSEb MAEa NSEb
Consequently, the computed values of bi and SEbi for a certain
MLR 141.59 0.29 345.26 0.75
explanatory variable strongly rely on the degree of its correla- PCR 93.68 0.14 102.54 0.19
tion with the other variables in the MLR model. The modeling ANN 79.68 0.71 88.93 0.47
accuracy was also compared using the NasheSutcliffe model PC-ANN 72.08 0.61 101.17 0.66
efficiency coefficient (NSE) computed from the predicted and Group A ANN 24.93 0.83 24.78 0.34
observed As concentrations (Nash and Sutcliffe, 1970). In PC-ANN 22.63 0.84 44.09 0.52
Group B ANN 49.68 0.96 111.21 0.74
essence, the closer the NSE is to 1, the more accurate the model
PC-ANN 35.10 0.98 75.15 0.71
is. In particular, an acceptable NSE value needs to be greater
than 0.5, while a good agreement value should be greater than a Mean absolute error (MAE) is a statistical approach used to assess
0.7 (Moriasi et al., 2007). As shown in Table 3, the NSE values of the model performance, and its unit is mg L1
1
Pn
the MLR model (0.29 and 0.75) indicate that the MLR model MAE ¼ ½n i¼1 jxobs  xpre j where n indicates the number of
observations of As. Here, xobs and xpre indicate the observed and
developed in this study cannot be considered a suitable model
predicted As concentrations, respectively.
for predicting the As concentration of groundwater using on- PT 2
t¼1 ðxobs  xpre Þ
t t
site measurement data. This result is similar to a previous b NSE ¼ 1  PT t 2
where xobs is the observed As
ðx
t¼1 obs
t  x pre Þ
study in Eastern India; Purkait et al. (2008) found that MLR
concentrations, xpre is the predicted As concentrations, and xtobs is
showed good accuracy for low levels, but that it did not repro-
the averaged value of the observed As.
duce the high levels of As concentration.
5540 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4
predicting As concentrations. Here, the cumulative
percentage of the variations explained by the five PCs was
100%, implying that these PCs were able to explain all the
variations in the original dataset. The five extracted PCs are
significantly related to each water quality parameter, PC1 (Eh,
0.95), PC2 (conductivity, 0.93), PC3 (water temperature, 0.99),
PC4 (pH, 1.00), and PC5 (TDS, 0.81).
Fig. 3(A) and (B) respectively compares the observed and
the predicted As concentrations of groundwater in both the
generation and validation steps of PCR. Overall, however, the
PCR model did not reproduce variations of the observed As
concentrations, usually by underestimating the relatively
high As concentrations similar to the MLR. As shown in Table
2, PC1, PC2, PC4, and PC5 had regression coefficients with
absolute values being greater than twice their standard errors.
In addition, all VIF values for the explanatory variables were
equal to or less than 1, thereby implying that the collinearity
problem in MLR was completely overcome through the PC
application. As shown in Table 2, however, the calculated
NSEs imply that the PCR model developed in this study still
cannot be used as a reasonable model for predicting ground-
water As concentrations using on-site measurement data.
Consequently, it demonstrates that the linearity of MLR and
PCR cannot reproduce the dynamic variations of observed As
concentrations in groundwater.
3.3. Nonlinear models for predicting As concentrations
3.3.1. Optimization processes for ANN and PC-ANN models
The optimal momentum rate, the number of hidden nodes,
and the learning rate were obtained by the pattern search
algorithm. The pattern search was used to determine the
optimal parameter set from the ranges of three parameters;
hidden nodes were ranged from 5 to 20, and learning and
momentum rates were ranged from 0.01 to 0.6. Finally, the
pattern search provided the optimal parameter set of the
learning and momentum rates and the number of hidden
layers, resulting in a minimum objective function. Moreover,
while too few hidden nodes in ANN results in a poor predictive
power, too many hidden nodes may cause a large computa-
tional time and over-fitting. The optimized number of hidden
nodes for the ANN and PC-ANN models were respectively set
to 13 and 15, which is consistent with past works on ANN
because they range between 2 and 3 times the number of input
nodes (Brion and Lingireddy, 1999). Note that the “tan-
sigmoid” transfer function was used in the neurons of the
hidden and output layers.
After the determining the learning rate, the momentum
rate, and the number of hidden nodes in the ANN and PC-ANN
models, the errors for the prediction of As concentrations
were computed in terms of MAE and NSE, and then compared
to those of MLR and PCR.
3.3.2. ANN for predicting As concentrations
Fig. 3(C) and (D) compares the observed and the predicted As
concentrations of groundwater in both the training and vali-
dation steps of ANN, where the horizontal and vertical axes
respectively indicate the observed and the predicted As
concentrations in groundwater. Overall, it can be seen that the
ANN model well reproduced the variations of the observed As
concentrations of groundwater in Cambodia, Laos, and
Thailand. As shown in Table 2, NSE values of ANN in training
and validation step are respectively greater than 0.5 and
approximately 0.5. In this case, it is clear that the MAE values
for ANN are much less than those for MLR and PCR. Conse-
quently, it can be seen that the optimized ANN model can be
a useful tool in predicting groundwater As concentrations
using on-site measurement data. This result is in good
agreement with a previous study by Purkait et al. (2008), in
which ANN showed better results than either a linear model.
3.3.3. PC-ANN for predicting As concentrations
Fig. 3(C) and (D) illustrates the observed and predicted
groundwater As concentrations in both the training and
validation steps of PC-ANN, where the open circles and
squares indicate values from PC-ANN. The figure shows that
the PC-ANN model also well reproduced the variations of
observed groundwater As concentrations in Cambodia, Laos,
and Thailand. As shown in Table 2, whereas the NSE value for
PC-ANN in the training step was 0.61, the NSE in the validation
step was 0.66; i.e., the NSE of PC-ANN was greater than that of
ANN in the validation step. Consequently, it demonstrates
that the PC-ANN model can be useful in predicting ground-
water As concentrations using on-site measurement data.
3.4. Comparison between MLR, PCR, ANN, and PC-ANN
As shown in Table 2, the prediction accuracies of ANN and PC-
ANN are better than those of MLR and PCR models in both the
generation/training and validation steps. The table also
demonstrates that the developed ANN and PC-ANN models
show acceptable accuracies for predicting As concentrations
in the groundwater in Cambodia, Laos, and Thailand, which
can be explained by the fact that the nonlinearity of ANN can
reproduce the vigorous variations of As concentrations. As
mentioned above, the MLR and PCR models developed in this
study are not deemed to be suitable models because linearity
is not sufficient for explaining the dynamic variations of
observed As concentrations. This study demonstrates,
however, that the ANN model is a more acceptable approach
than the MLR model in terms of modeling As concentrations
using data from on-site measurements. In particular, it can be
posited that PC-ANN is the best model because of its highest
NSE in the validation step.
3.5. Redox potential vs As concentration
Fig. 4 shows the As concentration versus Eh. Based on the Eh,
the dataset was divided into two groups, Group A (<0.00) and
Group B (0.00). The 89.23% of data for Group A are samples
collected from Kandal Province in Cambodia and the 84.52% of
data for Group B were measured from Lao PDR and Thailand.
Kandal Province is located in the Mekong Delta, which
receives a substantial volume of sediment (160 millon t yr1)
from the Mekong River (Meybeck and Carbonnel, 1975; Ta
et al., 2001); the delta is mainly composed of young alluvial
soil (Nguyen et al., 2000). In addition, a reduced state of As (III)
was found to be the dominant species in Kandal Province
(Polya et al., 2003; Rowland et al., 2008; Polizzotto et al., 2008).
Consequently, Group A in the figure can be characterized as
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4 5541

predominantly As in highly reducing aquifer regions having

a low Eh level. In the figure, the As concentrations are mostly
stable in the elevated level (Group A), but show a steep
negative trend with increasing Eh in Group B (i.e., Lao PDR and
Thailand). Hydrological conditions in Lao PDR, Cambodia, and
Vietnam have many characteristics in common and As (III)
was also found to be a dominant species in Lao PDR. Here,
Group B can also be interpreted as As in a reducing region,
though it showed a negative correlation with Eh as opposed to
Group A. This correlation may result from a combination of
low iron (ferrous) concentrations and high Eh in Lao PDR
(Chanpiwat et al., 2011). This different type of dependence of
As on Eh found may result in difficulty in using the entire
dataset for a single modeling approach. Therefore, Groups A
and B were utilized to train ANN and PC-ANN models, which
showed superior accuracy than linear models, in an attempt
to improve the prediction performance.
Fig. 5 compares the observed and predicted values of As
Fig. 4 e As concentration versus redox potential, showing
concentrations for the two datasets (Groups A and B). The two
Groups A and B.
models show a better prediction for Group A, but relatively
poor accuracies for Group B. Table 2 also demonstrates that
the NSE of the models for Group A were lower than the models
for Group B in the validation steps, and even worse than the

Fig. 5 e The prediction results of ANN and PC-ANN for Groups A and B, (A) ANN model for Group A, (B) ANN model for Group
B, (C) PC-ANN model for Group A, and (D) PC-ANN model for Group B.
5542 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4

Table 4 e Sensitivity rank of the conventional on-site measurement data in ANN and PC-ANN models from the LH-OAT
sensitivity analysis.
Group A Group B

ANN PC-ANN ANN PC-ANN

 1
Rank 1 Temperature [ C] pH [e] Total dissolved solids [g L ] Redox potential [mV]
Rank 2 Total dissolved solids [g L1] Total dissolved solids [g L1] Redox potential [mV] Electrical conductivity [ms cm1]
Rank 3 Electrical conductivity [ms cm1] Temperature [ C] Temperature [ C] pH [e]
Rank 4 Redox potential [mV] Redox potential [mV] pH [e] Temperature [ C]
Rank 5 pH [e] Electrical conductivity [ms cm1] Electrical conductivity [ms cm1] Total dissolved solids [g L1]

models trained with a whole dataset. This result may be
attributed to the insignificant relationship between As
concentration and the other water quality parameters, and
the limited variability of the observed As concentration in
Group A. In particular, the prediction for the lower As
concentration in Group A is poor, resulting in a lower NSE. In
contrast, the model accuracies for Group B in the training and
validation steps are satisfactory in terms of NSE (Moriasi et al.,
2007).
3.6. Sensitivity analysis
The sensitivities of the model outputs (i.e., As concentrations)
to water quality parameters were investigated in an attempt
to optimize ANN and PC-ANN models using the LH-OAT
method. Table 4 presents the sensitivity ranks for the five
parameters of ANN and PC-ANN for Groups A and B. For Group
A, the most significant parameters of ANN and PC-ANN are
temperature and pH, respectively. It seems to be counter-
intuitive because pH, TDS, and Eh were identified as signifi-
cant parameters in previous studies (Buschmann et al., 2007;
Amini et al., 2008; Chanpiwat et al., 2011). Conversely, pH is
the most sensitive parameter for PC-ANN, followed by TDS
and temperature [ C]; this is an acceptable result because
neutral to high pH conditions favor As release by promoting
desorption processes compared to the predominantly acidic
(Buschmann et al., 2007). In addition, pH  7 might possibly
enhance the mobilization of As, as explained by Buschmann
et al. (2007). The sensitivity results for Group B indicate that
Eh is the most sensitive parameter for PC-ANN. This coincides
with a previous study on As in Laos (Chanpiwat et al., 2011).
Also, it is clearly seen that As is negatively correlated with Eh,
as shown in Fig. 4. Consequently, even though the predictive
power of ANN is sufficient for following variations of the
observed As concentration, the role of temperature is theo-
retically different from that in literaturedindeed, it may be
caused by a collinearity problem among the input variables.
Conversely, the roles of pH and Eh in PC-ANN coincide with
preliminary studies, and thereby imply that PC-ANN is the
superior model for predicting the As concentrations of
groundwater in Cambodia, Laos, and Thailand.
Table 5 compares the averaged observed concentrations
and the As concentrations predicted by PC-ANN for each
province. In general, the model performances for Cambodia
showing a higher As concentration are more accurate than
those for the others countries. This may be caused by the
training process, which tends to follow a higher observed As,
though may not be as useful for lower As concentrations. In
other words, the model developed in this study would be more
informative for a high-risk area, such as Kandal Province in
Cambodia.
4. Conclusions
Groundwater resources are one of the most important
components of drinking water supplies, especially in rural
areas of Southeast Asian countries including Cambodia, Laos,
and Thailand. Over the years, several researchers have
attempted to explore the levels of As contamination from
various viewpoints. However, few studies have focused on
a statistical modeling of As that involved other conventional
water quality parameters. As such, the main conclusions
drawn from this study are as follows:

Table 5 e Comparison results between the observed and 1) The poor accuracies of MLR and PCR indicated that linear
predicted mean As concentrations (PC-ANN) for each models for conventional water quality parameters cannot
province. reproduce dynamic variations of observed groundwater As
Countries Province Observed As Predicted As concentrations.
[mg L1] [mg L1] 2) The prediction accuracies of ANN and PC-ANN were better
than those of MLR and PCR models in both the generation/
Thailand Suphan Buri 70.0 70.4
Cambodia Kandal 440.0 325.6 training and validation steps, showing acceptable accura-
Phnompen 136.4 127.6 cies for predicting As concentrations.
Lao PDR Vientiane 14.4 24.4 3) The results of the sensitivity analysis demonstrated that
Borikhamxay 14.0 30.0 the predictive power of ANN was satisfactory to follow
Savannakhet 24.0 6.4 variations of the observed As, but the roles of the input
Saravance 18.8 13.6
parameters are theoretically different from those in litera-
Champasack 40.0 25.6
ture, which might be caused by a collinearity problem
Attapeu 11.2 31.6
among the input variables.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4
4) Conversely, the roles of pH and Eh in PC-ANN coincided
with preliminary studies, and thereby imply that PC-ANN is
the superior model for predicting As concentrations of
groundwater in Cambodia, Laos, and Thailand.
We expect the PC-ANN model developed in this study to be
valuable to not only environmental scientists when designing
an efficient As monitoring and removal plan, but also for policy
makers and NGOs for establishing effective public health
management policies. In particular, because the As testing in
a laboratory is a complicated and costly process, the model can
be better applied to establish an effective As monitoring and
public health management in developing countries. The
prediction of As, however, is still a challenging work, especially
for developing a reliable model. In this study, the PC-ANN
model developed for Cambodia, Laos, and Thailand will be
more robust and reliable once new datasets are obtained from
other countries. As clearly shown in this study, As contami-
nation could not be explained by a linear combination of
conventional water quality parameters. Therefore, this model
still needs to be validated by using new datasets in order to
more precisely investigate the relationship between As
contamination and conventional water quality parameters.
Acknowledgments
This research was supported by the Basic Science Research
Program through the National Research Foundation of Korea
(NRF) funded by the Ministry of Education, Science and
Technology, Korea (No. 2010-0011822). We would also like to
acknowledge the assistance of the Ministry of Land, Transport
and Maritime Affairs (MLTMA), and the Korea Meteorological
Administration (KMA).
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.010.
references
American Water Works Association (AWWA), 2001. Arsenic Rule,
vol. 45. Mainstream.
Amini, M., Mueller, K., Abbaspour, K.C., Rosenberg, T., Afyuni, M.,
Møller, K.N., Sarr, M., Johnson, C.A., 2008. Statistical modeling
of global geogenic fluoride contamination in groundwaters.
Environmental Science and Technology 42 (10), 3662e3668.
Bagla, P., Kaise, J., 1996. India’s spreading health crisis draws
global arsenic experts. Science 274, 174e175.
Berg, M., Tran, H.C., Nguyen, T.C., Pham, H.V., Schertenleib, R.,
Giger, W., 2001a. Arsenic contamination of groundwater and
drinking water in Vietnam: a human health threat.
Environmental Science and Technology 35 (13), 2621e2626.
Berg, M., Tran, H.C., Nguyen, T.C., Pham, H.V., Schertenleib, R.,
Giger, W., 2001b. Magnitude of arsenic pollution in the Mekong
and Red River DeltasdCambodia and Vietnam. Science of the
Total Environment 372 (2e3), 413e425.
5543
Berg, M., Luzi, S., Kim, P.T., Viet, P.H., Giger, W., Stuben, D., 2006.
Arsenic removal from groundwater by household sand filters:
comparative field study, model calculations, and health
benefits. Environmental Science and Technology 40 (17),
5567e5573.
Berg, M., Stengel, C., Trang, P.T.K., Hung Viet, P., Sampson, M.L.,
Leng, M., Samreth, S., Fredericks, D., 2007. Magnitude of
arsenic pollution in the Mekong and Red River
DeltasdCambodia and Vietnam. Science of the Total
Environment 372 (2e3), 413e425.
Bowerman, B.L., O’Connell, R.T., 1990. Linear Statistical
Models: An Applied Approach, second ed. Duxbury Press,
Belmont, CA.
Brion, G.M., Lingireddy, S., 1999. A neural network approach to
identifying non-point sources of microbial contamination.
Water Research 33 (14), 3099e3106.
Buschmann, J., Berg, M., Stengel, C., Sampson, M.L., 2007. Arsenic
and manganese contamination of drinking water resources in
Cambodia: coincidence of risk areas with low relief
topography. Environmental Science and Technology 41 (7),
2146e2152.
Chang, F.J., Kao, L.S., Kuo, Y.M., Liu, C.W., 2010. Artificial neural
networks for estimating regional arsenic concentrations in
a blackfoot disease area in Taiwan. Journal of Hydrology 388
(1e2), 65e76.
Chanpiwat, P., Sthiannopkao, S., Cho, K.H., Kim, K.W., San, V.,
Suvanthong, B., Vongthavady, C., 2011. Contamination by
arsenic and other trace elements of tube-well water along the
Mekong River in Lao RDR. Environmental Pollution 159 (2),
567e576.
Chiew, H., Sampson, M.L., Huch, S., Ken, S., Bostick, B.C., 2009.
Effect of groundwater iron and phosphate on the efficacy of
arsenic removal by iron-amended biosand filters.
Environmental Science and Technology 43 (16), 6295e6300.
Cho, K.H., Kang, J.-H., Ki, S.J., Park, Y., Cha, S.M., Kim, J.H., 2009.
Determination of the optimal parameters in regression
models for the prediction of chlorophyll-a: a case study of the
Yeongsan Reservoir, Korea. Science of the Total Environment
407, 2536e2545.
Gros, P., 1997. Practical linear regression modelling. Basic
principles. Validation methods. Oceanis 23 (3), 359e515.
Holmberg, M., Forsius, M., Starr, M., Huttunen, M., 2006. An
application of artificial neural networks to carbon, nitrogen
and phosphorus concentrations in three boreal streams and
impacts of climate change. Ecological Modelling 195 (1e2),
51e60.
Jolliffe, I.T., 2002. Principal Component Analysis, second ed.
Springer, New York.
Kocar, B.D., Fendorf, C., 2009. Thermodynamic constraints on
reductive reactions influencing the biogeochemical of Arsenic
in soils and sediments. Environmental Science and
Technology 43 (13), 4871e4877.
Kohnhorst, A., 2005. Arsenic in groundwater in selected countries
in South and Southeast Asia: a review. The Journal of Tropical
Medicine and Parasitology 28 (2), 73e82.
Kuo, J.T., Hsieh, P.H., Jou, W.S., 2008. Lake eutrophication
management modeling using dynamic programming. Journal
of Environmental Management 88 (4), 677e687.
Lee, J.H., Yu, M.J., Bang, K.W., Choe, J.S., 2003. Evaluation of the
methods for first flush analysis in urban watersheds. Water
Science and Technology 48 (10), 167e176.
Luan, F., Liu, H.T., Wen, Y., Zhang, X., 2008. Prediction of
quantitative calibration factors of some organic compounds in
gas chromatography. Analyst 133 (7), 881e887.
Mac Nally, R., 2002. Multiple regression and inference in ecology
and conservation biology: further comments on identifying
important predictor variables. Biodiversity and Conservation
11 (8), 1397e1401.
5544 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 3 5 e5 5 4 4
Maier, H.R., Dandy, G.C., 1996. The use of artificial neural
networks for the prediction of water quality parameters.
Water Resources Research 32 (4), 1013e1022.
Meybeck, M., Carbonnel, J.P., 1975. Chemical transport by Mekong
river. Nature 255, 134e136.
Moriasi, D.N., Arnold, J.G., Van Liew, M.W., Bingner, R.L.,
Harmel, R.D., Veith, T.L., 2007. Model evaluation guidelines for
systematic quantification of accuracy in watershed
simulations. Transactions of the ASABE 50 (3), 885e900.
Myers, R., 1990. Classical and Modern Regression with
Applications, second ed. Duxbury Press, Boston, MA.
Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through
conceptual models part Ida discussion of principles. Journal
of Hydrology 10 (3), 282e290.
Nguyen, V.L., Ta, T.K.O., Tateishi, M., 2000. Late Holocene
depositional environments and coastal evolution of the
Mekong River delta, southern Vietnam. Journal of Asian Earth
Sciences 18 (4), 427e439.
Polizzotto, M.L., Kocar, B.D., Benner, S.G., Sampson, M.,
Fendorf, S., 2008. Near-surface wetland sediments as a source
of arsenic release to ground water in Asia. Nature 454 (7203),
505e508.
Polya, D.A., Gault, A.G., Bourne, N.J., Lythgoe, P.R., Cooke, D.A.,
2003. Coupled HPLC-ICP-MS analysis indicates highly
hazardous concentrations of dissolved arsenic species in
Cambodian groundwaters. In: Holland, J.G., Tanner, S.D. (Eds.),
Plasma Source Mass Spectrometry: Applications and Emerging
Technologies. Royal Society of Chemistry, Cambridge, pp.
27e140.
Polya, D.A., Gault, A.G., Diebe, N., Feldman, P., Rosenboom, J.W.,
Gilligan, E., Fredericks, D., Milton, A.H., Sampson, M.,
Rowland, H.A.L., Lythgoe, P.R., Middleton, C., Cooke, D.A.,
2005. Arsenic hazard in shallow Cambodian groundwaters.
Mineralogical Magazine 69 (5), 807e823.
Purkait, B., Kadam, S.S., Das, S.K., 2008. Application of artificial
neural network model to study arsenic contamination in
groundwater of Malda District, Eastern India. Journal of
Environmental Informatics 12 (2), 140e149.
Rawlings, J.O., Pantula, S.G., Dickey, D.A., 1998. Applied
Regression Analysis: A Research Tool. Springer, New York.
Riad, S., Mania, J., Bouchaou, L., Najjar, Y., 2004. Rainfall-runoff
model using an artificial neural network approach.
Mathematical and Computer Modelling 40 (7e8), 839e846.
Rowland, H.A.L., Gault, A.G., Lythgoe, P., Polya, D.A., 2008.
Geochemistry of aquifer sediments and arsenic-rich
groundwaters from Kandal Province, Cambodia. Applied
Geochemistry 23 (11), 3029e3046.
Sarangi, A., Bhattacharya, A.K., 2005. Comparison of artificial
neural network and regression models for sediment loss
prediction from Banha watershed in India. Agricultural Water
Management 78 (3), 195e208.
Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source,
behaviour and distribution of arsenic in natural waters.
Applied Geochemistry 17 (5), 517e568.
Sousa, S.I.V., Martins, F.G., Alvim-Ferraz, M.C.M., Pereira, M.C.,
2007. Multiple linear regression and artificial neural networks
based on principal components to predict ozone
concentrations. Environmental Modelling and Software 22 (1),
97e103.
Stanger, G., Truong, T.V., My Ngoc, K.S.L.T., Luyen, T.V.,
Tuyen, T.T., 2005. Arsenic in groundwaters of the lower
Mekong. Environmental Geochemistry and Health 27 (4),
341e357.
Sthiannopkao, S., Kim, K.W., Cho, K.H., Wantala, K., Sotham, S.,
Sokuntheara, C., Kim, J.H., 2010. Arsenic levels in human hair,
Kandal Province, Cambodia: the influences of groundwater
arsenic, consumption period, age and gender. Applied
Geochemistry 25 (1), 81e90.
Sun, G., Liu, J., Luong, T.V., Sun, D., Wang, L., 2002. Endemic
Arsenicosis: A Clinical Diagnostic with Photo Illustrations.
UNICEF East Asia and Pacific Regional Office, Bangkok,
Thailand, ISBN 974-685-040-7.
Ta, T.K.O., Nguyen, V., Tateishi, M., Kobayashi, I., Saito, Y.,
2001. Sedimentary facies, diatom and foraminifer
assemblages in a late PleistoceneeHolocene incised-valley
sequence from the Mekong River Delta, Bentre Province,
southern Vietnam: the BT2 core. Journal of Asian Earth
Sciences 20 (1), 83e94.
Tayfur, G., Swiatek, D., Wita, A., Singh, V.P., 2005. Case study:
finite element method and artificial neural network models
for flow through Jeziorsko Earthfill dam in Poland. Journal of
Hydraulic Engineering 131 (6), 431e440.
Tetsuro, A., Takashi, K., Junko, F., Reiji, K., Tu, B.M., Pham, T.K.T.,
Hisato, I., Annamalai, S., Pham, H.V., Shinsuke, T., 2006.
Contamination by arsenic and other trace elements in tube-
well water and its risk assessment to humans in Hanoi,
Vietnam. Environmental Pollution 139 (1), 95e106.
Wen, C.G., Lee, C.S., 1998. A neural network approach to
multiobjective optimization for water quality management in
a river basin. Water Resources Research 34 (3), 427e436.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Phytoaccumulation of antimicrobials from biosolids: Impacts

on environmental fate and relevance to human exposure

Niroj Aryal, Dawn M. Reinhold*

Department of Biosystems and Agricultural Engineering, Michigan State University, 216 A.W. Farrall Agricultural Engineering Hall, East
Lansing, MI 48824, USA

article info abstract

Article history: Triclocarban and triclosan, two antimicrobials widely used in consumer products, can
Received 5 January 2011 adversely affect ecosystems and potentially impact human health. The application of
Received in revised form biosolids to agricultural fields introduces triclocarban and triclosan to soil and water
7 August 2011 resources. This research examined the phytoaccumulation of antimicrobials, effects of
Accepted 9 August 2011 plant growth on migration of antimicrobials to water resources, and relevance of phy-
Available online 23 August 2011 toaccumulation in human exposure to antimicrobials. Pumpkin, zucchini, and switch grass
were grown in soil columns to which biosolids were applied. Leachate from soil columns
Keywords: was assessed every other week for triclocarban and triclosan. At the end of the trial,
Triclosan concentrations of triclocarban and triclosan were determined for soil, roots, stems, and
Triclocarban leaves. Results indicated that plants can reduce leaching of antimicrobials to water
Phytoremediation resources. Pumpkin and zucchini growth significantly reduced soil concentrations of tri-
Plant uptake closan to less than 0.001 mg/kg, while zucchini significantly reduced soil concentrations of
triclocarban to 0.04 mg/kg. Pumpkin, zucchini, and switch grass accumulated triclocarban
and triclosan in mg per kg (dry) concentrations. Potential human exposure to triclocarban
from consumption of pumpkin or zucchini was substantially less than exposure from
product use, but was greater than exposure from drinking water consumption. Conse-
quently, research indicated that pumpkin and zucchini may beneficially impact the fate of
antimicrobials in agricultural fields, while presenting minimal acute risk to human health.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction 51  15 mg triclocarban and 30  11 mg triclosan per g dry sludge

Antimicrobials, specifically triclocarban (3,4,40 -trichloroca-
rbanilide) and triclosan (5-chloro-2-(2,4-dichlorophenoxy)-
phenol), are widely used in personal care products (Ying et al.,
2007). Triclocarban and triclosan primarily enter the envi-
ronment through domestic sewage discharge to wastewater
treatment plants, where removal is predominantly due to
sorption (78  11% for triclocarban and 80  22% for triclosan)
to wastewater particulate matter (Chu and Metcalfe, 2007;
Heidler and Halden, 2007; Heidler et al., 2006; Sapkota et al.,
2007). Consequently, digested municipal sludge accumulates
(Heidler and Halden, 2007; Heidler et al., 2006). Greater than
97% of triclocarban and triclosan in sewage is discharged to
water resources and biosolids, leading to a 0.6 to 1 million kg/
year combined input into U.S. environment (EPA, 2003; Halden
and Paull, 2005). The greatest discharge of triclocarban and
triclosan into the environment is through municipal applica-
tion of biosolids to fields, as more than 50% of biosolids are
land applied (Heidler et al., 2006).
Triclocarban is generally detected in U.S. surface waters at
concentrations from 10 to 1550 ng/L (Halden and Paull, 2005;
Sapkota et al., 2007), whereas triclosan is detected in U.S.

* Corresponding author. Tel.: þ1 517 432 7732; fax: þ1 517 432 2892.
E-mail address: reinhold@egr.msu.edu (D.M. Reinhold).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.027
5546 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
rivers and streams from 3.0 to 75 ng/L (Bester, 2003; Ying and
Kookana, 2007). Additionally, agricultural soils previously
amended with biosolids can accumulate 1.2e65 ng/kg triclo-
carban and 0.16e1.0 ng/kg triclosan (Cha and Cupples, 2009).
Once introduced into the environment, triclocarban and tri-
closan sorb to soils or sediments and are not predicted to
readily degrade (Halden and Paull, 2005; Ying et al., 2007).
Experimental half-lives of triclocarban and triclosan range
from 87 to 231 and 18e58 days respectively in aerobic soil, with
longer half-lives in anaerobic soils (Wu et al., 2009; Ying et al.,
2007). Triclocarban and triclosan are lowly soluble in water
with solubilities of 45 mg/L for triclocarban (Snyder et al., 2010)
and 462 mg/L for triclosan (EPI Suite 4.0). Triclocarban and tri-
closan have high octanolewater partitioning coefficients (KOW)
with log KOW of 4.90 for triclocarban and log KOW of 4.76 for
triclosan (EPI Suite 4.0). Henry Law constants of
4.52  1011 atm-m3/mole for triclocarban and 2.13  108 atm-
m3/mole for triclosan (EPI Suite 4.0) indicate that they are
nonvolatile.
Release of antimicrobials into the environment may result
in bioaccumulation of antimicrobials in aquatic organisms and
humans. For example, up to 58 mg/kg triclosan and 299 mg/kg
triclocarban was measured in snails near wastewater treat-
ment plant (WWTP) effluent (Coogan and La Point, 2008).
Furthermore, concentrations of 2.4e3790 mg/L triclosan were
measured in 74.6% of urine samples collected from the U.S.
general population (Calafat et al., 2008). Triclosan, at the
concentration of 0.01e19 mg/kg, was also detected in plasma of
Swedish women who did not use personal care products con-
taining triclosan, indicating unintentional systemic exposure
to antimicrobials through sources other than personal care
products (Allmyr et al., 2006).
Antimicrobials have the potential to adversely impact
human and ecosystem health. In humans, disruption of
endocrine activity by triclosan and triclocarban is expected at
concentrations of 29e3150 mg/L (Ahn et al., 2008). At much
lower concentrations, triclocarban and triclosan disrupt crit-
ical ecological processes. In rivers, antimicrobials adversely
affect biofilm structure and function (Lawrence et al., 2009).
Freshwater microbial communities are sensitive to 2.9 mg/L
triclosan (Johnson et al., 2009) and concentrations as low as
150 ng/L triclosan can have physiological effects on thyroid
hormone, body weight, and hind limb development in frogs
(Fraker and Smith, 2004; Veldhoen et al., 2007). Consequently,
current environmental concentrations of antimicrobials have
the potential to disrupt aquatic ecosystems. In terrestrial
ecosystems, triclocarban and triclosan can inhibit soil respi-
ration, nutrient recycling, and plant growth (Liu et al., 2009).
Development of microbial and drug resistance has also been
reported (Heath et al., 1998; Walsh et al., 2003).
Studies have demonstrated that plants such as pumpkin
and zucchini have potential to accumulate hydrophobic
chlorinated organic pollutants. In field experiments of poly-
chlorinated biphenyls (Aroclor 1254/1260), Cucurbita pepo ssp
pepo cv. Howden (pumpkin) plants took up, translocated, and
accumulated 7.6 mg/kg PCBs in plant shoots (Aslund et al.,
2007). Studies have also demonstrated that C. pepo species
(pumpkin and zucchini) can extract and translocate dichlor-
odiphenyltrichloroethane (DDT) and its metabolites (White
et al., 2003) and polychlorinated dibenzo-p-dioxins and
dibenzofurans (Hülster et al., 1994). Pumpkin and zucchini
showed more potential for phytoaccumulation than other
crops. For example, pumpkin and zucchini translocated
1.8e36 times more 2,4,8-trichlorodibenzo-p-dioxin and 2e4.3
times more 1,3,6,8-tetrachlorodibenzo-p-dioxin than 10 other
food crops (Zhang et al., 2009). Additionally, the capability of
switch grass (Panicum variegatum L.) to reduce PCB concen-
tration in soils through unidentified mechanism has been
demonstrated (Dzantor et al., 2000). Recent studies have also
detected triclocarban and triclosan in shoot tissues and beans
of soybean plants that were planted in biosolids-amended
soils (Wu et al., 2010).
The study presented herein evaluates the hypothesis that
triclocarban and triclosan, being chlorinated aromatic organic
pollutants similar to DDT and PCBs, can be taken up by plants
to reduce antimicrobial concentrations in soil and the
migration of antimicrobials to water resources. Under-
standing the fate of triclocarban and triclosan in vegetated
fields to which biosolids have been applied is crucial to
understanding the entrance of antimicrobials into water
resources and quantifying the human health effects of
consuming food grown on land to which biosolids are applied.
The aims of this study were to: (i) assess the effects of plant
growth on leaching of antimicrobials from land applied
biosolids, (ii) evaluate accumulation of antimicrobials in
pumpkin, zucchini and switch grass, and (iii) preliminarily
evaluate relevance of triclocarban and triclosan phytoaccu-
mulation to human health.
2. Materials and methods
2.1. Soil, seeds and biosolids
The plant varieties used for this study were pumpkin (C. pepo
cultivar Howden), zucchini (C. pepo cultivar Gold Rush), and
switch grass (P. variegatum). The specific varieties of C. pepo
were chosen based on their reported accumulation potential
for hydrophobic organic pollutants (Aslund et al., 2008;
Lunney et al., 2004; Wang et al., 2004; White et al., 2003).
Switch grass was selected as a non-vegetable plant that has
potential to stimulate rhizosphere microbial degradation of
hydrophobic aromatic pollutants like PCBs (Dzantor et al.,
2000). Pumpkin and zucchini seeds were obtained from
Johnny Seeds, Maine, whereas switch grass seedlings were
obtained from the Department of Crop and Soil Sciences, MSU.
Soil for the study, a screened sandy clay loam, was obtained
from East Lansing, MI. Biosolids were collected from a nearby
wastewater treatment plant and were analyzed for triclo-
carban and triclosan prior to use by previously developed
methods (Cha and Cupples, 2009) using pressurized solvent
extraction and tandem mass spectrometry. Triclocarban and
triclosan were present at concentrations of 8.18  0.56 mg/kg
and 0.18  0.01 mg/kg dry mass of biosolids, respectively.
2.2. Chemicals
Triclocarban [CAS 101-20-2] (>98%) was obtained from Tokyo
Chemical Industry and triclosan [CAS 3380-34-5] from
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
Calbiochem. Mobile phase and extraction solvents were
obtained from VWR, Inc. LCeMS solvents were of MS grade.
2.3. Experimental columns
Plants were grown in experimental columns with diameters of
14.7 cm and lengths of 30 cm. The bottom 7.6 cm of the soil
column was filled with soil without mixing biosolids. The next
15.2 cm of soil was thoroughly mixed with biosolids at the
application rate of 0.73 dry Mg per 1000 m2 (Cha and Cupples,
2009). Solid content of biosolids was 4.8% (dry basis) and 200 g
of wet biosolids were applied before seed sowing, with an
additional 60 g of wet biosolids applied after 8 weeks to
simulate a second field application.
Experimental design included quadruple columns for
pumpkin and zucchini and triplicate columns for switch grass
and no plant controls. Seeds were sown at the rate of a seed
per column except for switch grass, for which plants were
transplanted directly. Plants were maintained at a constant
temperature (23  2  C) using a light regime of 16 h light: 8 h
dark.
2.4. Sampling
For leachate sampling, columns were flooded with equal
volumes of water so that a minimum of 100 mL water leached
from each column. Leachate samples were collected in amber
bottles under each soil column once every two weeks.
Samples were stored immediately at 4  C until prepared for
analysis. Plants were harvested at the end of 22 weeks. Plant
tissues were separated into roots, leaves, and stems for
pumpkin and zucchini and into roots and shoots for switch
grass. Plant tissues were rinsed carefully to remove soil and
dust particles, air-dried at room temperature (23  2  C),
weighed and stored in amber bottles in the refrigerator until
sample extraction. Soil samples from each column, at depths
of 5 cm, 10 cm, 15 cm, and 20 cm, were collected in amber
bottles, screened to remove plant roots, homogenized by
mixing, and stored at 8  C until sample preparation.
2.5. Sample preparation
Aqueous samples were prepared as published previously
(Halden and Paull, 2004, 2005). Samples were passed through
a solid phase extraction (SPE) cartridge (Oasis HLB 3 cc, Waters
Corporation) and eluted with 4 ml of 50% methanol and 50%
acetone containing 10 mM acetic acid. Elutes were dried under
nitrogen, reconstituted in 1 ml of 50% methanol and 50%
acetone, filtered through a 0.2 mm PTFE membrane, and
analyzed by liquid chromatography mass spectrometry in
negative electrospray ionization mode (LCeMS/ESI(e)).
Soil samples were prepared and extracted as previously
published (Cha and Cupples, 2009) by pressurized liquid
extraction (PLE), using a Dionex ASE 200 accelerated solvent
extractor. Triplicate subsamples of soil (5 dry g each) from
each column sample were extracted. A fourth subsample was
weighed and dried for at least 24 h at 105  C and again weighed
for moisture determination. Extraction on the ASE utilized
acetone with oven temperature of 100  C, extraction pressure
of 1500 psi, static time of 5 min, and flush volume of 100%. The
5547
extracts were evaporated to dryness under nitrogen, recon-
stituted in 50% methanol and 50% acetone, spiked with 6 ppm
triclosan and triclocarban, and filtered through a 0.2 mm filter
before analysis by LCeMS. For quality assurance, all analytical
runs included a blank, a control, and a sample triplicate. The
recovery from the methods was 94.03  9.76% for triclocarban
and 83.39  19.48% for triclosan. For plant samples, frozen
plant tissues were oven dried, grounded in mortar and pestle,
and extracted using the same method as for soil described
above. All analytical runs included a blank, a control, and
a sample triplicate except for roots, where sufficient biomass
for replicate analysis was not available.
2.6. LC/MS analysis
A Shimadzu LCeMS 2010 EV was used to analyze samples for
triclocarban and triclosan. Samples were separated using
Allure biphenyl column (5 mm, 150  2.1 mm) from Restek Cor.
using a binary gradient of 75% methanol and 25% 5-mM
ammonium acetate to 100% methanol. MS parameters were:
curved desolvation line (CDL) of 1.5 V, block and CDL
temperature of 30  C and nitrogen desolvation gas flow rate of
1.5 L/min. MS negative electrospray ionization mode with
scan mode was used for method development and identifi-
cation, whereas selected ion monitoring (SIM) mode was used
for quantification. Retention time (tR  0.1 min), detection of
characteristic molecular ions (m/z 313 for triclocarban and m/z
287 for triclosan), and detection of reference ions (m/z 315 and
317 for triclocarban and m/z 289 and 291 for triclosan) were
used to identify the target molecules (Halden and Paull, 2005).
Quantification was performed using external, linear calibra-
tion and a minimum of six calibration levels. The limits of
detection were 10e100 ng/L for water and 0.1e1.0 ng/kg for
soils and plants.
2.7. Statistical analysis
All statistical analysis was performed in Sigma Plot (v 11.0).
A one tailed t-test was used for all pair-wise comparisons and
one-way ANOVA was used for all other comparisons. The
reported values are in mean  standard error.
2.8. Relevance of plant accumulation of antimicrobials to
human health
Accumulation of antimicrobials by pumpkin and zucchini
represents the potential for direct exposure of humans to
antimicrobial through ingestion of vegetables. To quantify the
potential impacts of accumulation of antimicrobials by
pumpkin and zucchini on human health, doses of exposure
for different routes were compared. The resulting dose of
triclocarban and triclosan from consumption of pumpkin and
zucchini exposed to biosolids was predicted by assuming fruit
concentrations equal to the range of observed stems and
leaves concentrations. This assumption is conservative in
that it likely over predicts actual risk because: (i) leaves had
lower concentrations of antimicrobials than that of stems in
this study and another similar study with soybeans (Wu et al.,
2010) and (ii) decreases in concentrations of PCB and DDE were
observed in the stem of pumpkin and zucchini as the distance
5548 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
from root increased (Aslund et al., 2007; White et al., 2003).
Predicted doses were compared to doses from other routes of
exposure: product use (EPA, 2002), drinking water exposure
(EPA, 2002) and exposure from eating soybean produced from
fields that receive biosolids (Wu et al., 2010). The dose calcu-
lations were based on 5 g/day average soybean (Reinwald
et al., 2010) and 11.5 g/day average pumpkin and zucchini
consumption per capita in USA (USDA, 2011).
3. Results and discussions
3.1. Leaching of antimicrobials
Both triclocarban and triclosan were detected in leachate from
the experimental soil columns. As shown in Fig. 1, the
concentrations increased initially for two weeks and then
decreased. The lag in peak concentration was likely due to
sorption of antimicrobials to bottom 7.6 cm of soil where
biosolids were not applied. In saturated soil systems, sorption
and biodegradation were primary removal mechanisms for
triclocarban (Drewes et al., 2003; Essandoh et al., 2010). The
maximum triclosan concentration in leached water from
pumpkin, zucchini, switch grass and control were
530  180 mg/L, 1670  540 mg/L, 460  280 mg/L, 710  450 mg/mL
respectively, all observed in second week. Triclocarban
concentrations also followed a similar trend with maximum
concentrations of 210  160 mg/L, 190  110 mg/L, 120  40 mg/L,
and 370  360 mg/L for pumpkin, zucchini and switch grass and
Fig. 1 e Triclocarban and triclosan concentrations in
leached water (mg/ml) with time (weeks). Points represent
mean and error bars represent standard error. Volume of
water leached each week was statistically similar between
columns.
control, respectively. An initial peak with maximum concen-
tration of 110  81 mg/L triclosan and 3.4  2.2 mg/L triclo-
carban, followed by a rapid decrease in concentrations, has
previously been observed in runoff from agricultural field after
application of biosolids (Sabourin et al., 2009). Higher
concentrations in this experiment were likely due to the
relatively short columns. The second addition of biosolids
increased antimicrobial concentrations immediately. For tri-
clocarban, the observed second peak was not statistically
different from the first peak (P ¼ 0.6). In contrast, triclosan was
observed at lower concentrations after the second biosolids
application (P ¼ 0.03).
Triclosan, despite being present at a lower initial concen-
tration in the applied biosolids, was collected at higher
concentrations than triclocarban in leachate during the first
four weeks. The higher concentrations of triclosan in the
leachate were attributed to its higher solubility and decreased
affinity for sorption when compared to triclocarban. The
aqueous solubility of triclosan is approximately 10 times that
of triclocarban. Sorption of triclocarban to soils is also much
stronger than sorption of triclosan, with sorption coefficients
(Kd) of 1029 L/kg and 231 L/kg (respectively) for sandy loam
soils (Wu et al., 2009). Despite triclosan and triclocarban being
weak acids with acid dissociation constants (Ka) of 107.9 and
1012.7 (respectively), the role of soil pH, as affected by addi-
tion of biosolids, was not expected to greatly impact sorption
of triclosan or triclocarban. While sorption of the anionic form
of triclosan is less than the sorption of the neutral form of
triclosan, sorption of the anionic form of triclosan is still
considerable, resulting in only a slight decrease in net triclo-
san sorption in soils when the pH increased from 4 to 8 (Wu
et al., 2009). Triclocarban sorption was almost unaltered
when pH of soil was increased from 4 to 8 (Wu et al., 2009). As
measurements indicated the soil pH after addition of biosolids
to the experimental columns were between 7 and 8, the
increase in pH from the addition of biosolids to the soil
columns was expected to only minimally decrease the sorp-
tion of triclosan to the soil and was not expected to affect
triclocarban sorption to the soil.
In contrast to the first four weeks, more triclocarban
leached from the columns than triclosan over the remaining
18 weeks. Reduction in the relative leaching of triclosan was
most likely due to increased microbial degradation of triclo-
san. In aerobic soils, degradation of triclosan was character-
ized by a half-life of 18e58 days, while degradation of
triclocarban exhibited a half-life of 87e231 days (Wu et al.,
2009; Ying et al., 2007). Additionally, the initial concentra-
tions of triclocarban in the biosolids were much higher than
were the initial concentrations of triclosan.
The total masses of antimicrobials leached from the
vegetated soil columns over 22 weeks were not significantly
different from the total masses leached from the control
columns. However, the highest concentrations of antimicro-
bials were leached in week 2, prior to full establishment of the
plants. When only considering the total antimicrobials
leached after the second addition of biosolids, there were
significant differences between the masses of antimicrobials
leached from the control and vegetated columns, with P
values of <0.01 for pumpkin, 0.01 for zucchini and 0.01 for
switch grass. These results suggest that established plants can
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2 5549

reduce leaching of antimicrobials, prompting additional

analysis of the fate of antimicrobials in vegetated soils.

3.2. Soil concentration of antimicrobials

Soil concentrations of antimicrobials at the end of 22 weeks

are summarized in Table 1. Soil concentrations of triclocarban
in zucchini columns and of triclosan in zucchini and pumpkin
columns were substantially lower than were soil concentra-
tions in control columns (0.02 < P < 0.05). However, soil
concentrations of triclocarban in pumpkin columns were
similar to those in control columns (P ¼ 0.09). Soil concen-
trations of triclosan and triclocarban in switch grass columns
were similar to or greater than soil concentrations in control
columns. The observed difference in final soil concentration
of antimicrobials between pumpkin, zucchini, and switch
grass columns may have resulted from difference in plant
growth in the column. Shoot mass of switch grass plants was
0.49  0.24 g as opposed to 3.74  0.93 g for pumpkin and
5.39  0.28 g for zucchini. Root masses were 0.91  0.40 g for Fig. 2 e Plant concentrations of triclocarban. Points
switch grass, 0.08  0.03 g for pumpkin and 0.10  0.03 g for represent mean and error bars represent standard error.
zucchini. Development of an extensive root mass in the
absence of robust above ground growth did not appear to
influence soil concentration of triclosan in switch grass accumulation of PCBs (Aslund et al., 2007) and soybean
columns. However, results indicated pumpkin and zucchini accumulation of triclocarban and triclosan (Wu et al., 2010).
may decrease soil concentrations of triclocarban and/or tri- A similar decrease from stem to leaf was observed for the
closan after land application of biosolids. accumulation of antimicrobials by pumpkin; however, the
decrease was only significant for triclosan. In contrast,
3.3. Concentrations of antimicrobials in plant tissues concentration of antimicrobials in zucchini increased from
stem to leaf, with a significant increase observed for triclo-
Triclocarban and triclosan were detected in roots, stems, and carban. Further studies with higher number of replicates
leaves of pumpkin, zucchini, and switch grass. Antimicrobials could clarify the significance of the observed trends in plant
concentrations in plant tissues (Figs. 2 and 3) ranged from concentrations of antimicrobials.
1.10 mg/kg in leaves to 39.5 mg/kg in roots. Root concentra- The hydrophobicities of triclocarban (log KOW ¼ 4.9) and
tions of antimicrobials were generally higher than concen- triclosan (log KOW ¼ 4.76) indicate the potential for bio-
trations in stems and leaves, with the exception of accumulation. Soil concentrations of antimicrobials were
triclocarban concentrations in pumpkin tissues. However, significantly less than the concentrations of antimicrobials in
plant tissue concentrations were highly variable and higher stems, leaves, and water for all plant species (P values from
root concentrations were usually not statistically significant
compared to shoot concentrations. No consistent trends in
change in concentration of antimicrobials in stem and leaf
tissues in pumpkin and zucchini were observed. A general
decrease in concentration from root to stem to leaves to fruits
has been previously observed for pumpkin and zucchini

Table 1 e Soil antimicrobial concentrations and P values

for comparison with controls. ND refers to values which
were detected but less than quantitation limit
(<0.001 mg/kg).
Triclocarban, Triclosan, Sum,
mg/kg mg/kg mg/kg

Pumpkin 0.055  0.003 ND (P ¼ 0.021) 0.055  0.003

(P ¼ 0.091) (P ¼ 0.073)
Zucchini 0.038  0.004 ND (P ¼ 0.021) 0.038  0.004
(P ¼ 0.040) (P ¼ 0.032)
Switch 0.241  0.026 0.001  0.000 0.166  0.014
grass (P ¼ 0.045) (P ¼ 0.457) (P ¼ 0.155)
Fig. 3 e Plant concentrations of triclosan. Points represent
Control 0.097  0.012 0.007  0.001 0.105  0.013
mean and error bars represent standard error.
5550 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
<0.05 except for between soil and zucchini roots). Triclo-
carban root bioaccumulation factors, the ratio of root
concentration to soil concentrations (g/g), were 11.01  5.06
for pumpkin, 40.27  46.34 for zucchini, and 30.92  9.41 for
switch grass. For triclosan, the root bioaccumulation factors
(g/g) were 972  398 for pumpkin, 1822  260 for zucchini, and
874  706 for switch grass respectively. These values are
comparably higher than 1.0e3.3 for DDT bioaccumulation for
the same plant varieties (Lunney et al., 2004) and higher than
root bioaccumulation factors of 2.2e5.8 for triclosan and
1.7e2.0 for triclocarban observed for soybean (Wu et al., 2010).
Triclocarban translocation factors (the ratios of shoot
concentration to root concentration) were 0.78  0.55 g/g for
pumpkin, 0.27  0.19 g/g for zucchini, and 0.52 g/g for switch
grass. For triclosan, translocation factors were 0.30  0.21 g/g
for pumpkin, 0.16  0.05 g/g for zucchini and 0.81 g/g for
switch grass. Low translocation factor imply limited transport
from root to shoot. These translocation factors are compa-
rable to those reported in Aslund et al. (2007) and Lunney et al.
(2004) for accumulation of DDT, DDD, and DDE and PCBs by
the same species. Wu et al. (2010) reported translocation
factors of 0.01e0.28 and 0.16e1.77 for triclocarban and triclo-
san, respectively, in soybean plants, indicating the trans-
location of antimicrobials is comparable for pumpkin,
zucchini, and soybean. However, as root bioaccumulation
factors were greater for pumpkin and zucchini than for
soybeans, results indicated that pumpkin and zucchini have
greater potential to accumulate antimicrobials than soybeans.
A mass balance for each column was completed (Fig. 4).
A significant mass fraction of triclocarban was not accounted
for in most of the columns, indicating other mechanisms of
antimicrobial loss, such as microbial degradation, phytosti-
mulation, or phytodegradation. After eight weeks, the largest
portion of triclocarban remained in the soil, indicating that
sorption plays an important role in fate of triclocarban in
Fig. 4 e Mass balance analyses of triclosan and triclocarban
in vegetated and control columns.
vegetated soils. For zucchini, total mass accumulation in
leaves was highest followed by stems and roots accumulation.
In contrast, stems accumulated the greatest mass, followed by
leaves and then by roots, in pumpkin. Accumulation of anti-
microbials was greater in roots than in shoots for switch grass
owing to its limited shoot production. With the exception of
triclosan in zucchini columns (where all mass was accounted
for in leachate and plant samples), the unaccounted mass of
antimicrobials was greater in planted columns than in
unplanted controls, supporting the observation that presence
of plants may promote removal of antimicrobials by microbial
degradation or other unidentified mechanisms. Analysis of
coefficients of variation showed that there was minimal
variation between the samples in the same column, but rather
large variation between the columns. Additionally, large
variability was observed in plant masses between columns.
Consequently, the observed variability in the experiment was
most likely due to inherent biological variability within plant-
based systems, rather than variability due to sampling or
procedure.
3.4. Relevance of plant accumulation of antimicrobials to
human health
The doses calculated for multiple routes of exposure to anti-
microbials, shown in Fig. 5, are substantially less than the no-
observable adverse effect level (NOAEL) of 25 mg/kg bw/d for
triclocarban (EPA, 2002). Therefore, none of the examined
exposure routes present concerns, on an acute basis, to
human health. Exposure of triclocarban from eating pumpkin
and zucchini grown in fields receiving biosolids is two orders
less than exposure from using products containing triclo-
carban, about 35 times greater than exposure from drinking
water, and about 250 times greater than exposure from eating
soybeans grown in fields receiving biosolids. Assuming a per
capita consumption of 534 g/day fresh vegetables (USDA,
2001e2002) and that the triclocarban concentrations in
pumpkin shoots are representative of all vegetables, the
Fig. 5 e Comparison of triclocarban dose associated with
multiple routes of exposure for an adult. Error bars
represent maximum and minimum doses. The NOAEL (no-
observable adverse effect level) for triclocarban is 25 mg/
kg-bw/day.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
resulting antimicrobial exposure from vegetable consumption
is (at worst case) similar to the exposure from using personal
care products. Therefore, the acute health risk from antimi-
crobial accumulation by vegetables in fields receiving
biosolids is likely minimal. However, additional studies,
utilizing a more diverse range of vegetable and fruit crops, are
needed to confirm this assessment.
However, there is also the issue of antibiotic resistance due
to antimicrobial exposure. While there is dearth of informa-
tion about the development of microbial resistance from using
products containing triclocarban, many reports indicate
development of antibiotic resistance due to the triclosan. For
example, the susceptibility of Escherichia coli, Proteus mirabilis,
and Staphylococcus aureus to antibiotics was reduced by 40e400
times from exposure to triclosan (Saleh et al., 2011; Stickler
and Jones, 2008; Suller and Russell, 2000). Though reduction
of soil antimicrobial concentrations by plants may help to
reduce antimicrobial resistance in soils, the ingestion of
antimicrobials through food might increase drug and anti-
microbial resistance in humans. More research is needed to
explore this potential ramification of the accumulation of
antimicrobials by food crops.
4. Conclusions
Research indicated that established plants reduce the migra-
tion of antimicrobials from biosolids to water resources
through phytoaccumulation and additional unidentified
mechanisms. The soil concentrations of triclocarban and/or
triclosan in pumpkin and zucchini columns were less than
soil concentrations in unplanted columns. Additionally, the
masses of antimicrobials for which the study was unable to
account were generally greater in columns with plants than in
columns without plants. Pumpkin, zucchini, and switch grass
plants accumulated triclosan and triclocarban in mg per kg
(dry weight) concentrations. In general, roots had higher
concentrations of antimicrobials than shoots, but less total
mass accumulation of antimicrobials due to high production
of shoots. Consequently, results indicate that plants (i) reduce
leaching of antimicrobials from fields to which biosolids have
been applied, (ii) directly impact the fate of antimicrobials
through phytoaccumulation, and (iii) decrease the persistence
of antimicrobials in soil systems through additional, uniden-
tified mechanisms. There was no acute human risk to humans
due to eating pumpkins and zucchini produced from field to
which biosolids have been applied. However, additional field-
scale studies with fruit production and more food crops are
necessary to more clearly understand human exposure to
antimicrobials through food crops. More research is also
needed to identify and quantify mechanisms that dictate the
fate of antimicrobials in vegetated fields receiving biosolids.
Acknowledgments
The authors are grateful to Reinhold Research group
members, Department of Biosystems and Agricultural Engi-
neering, Michigan State University. Also, special thanks to the
5551
funding agencies: Center for Water Science, MSU, Michigan
Agriculture Experiment Station and College of Engineering,
MSU. We would like to thank Dr. Alison Cupples and Dr. Jong
M. Cha for assistance with extractions and biosolids analysis.
references
Ahn, K.C., Zhao, B., Chen, J., Cherednichenko, G., Sanmarti, E.,
Denison, M.S., Lasley, B., Pessah, I.N., Kultz, D., Chang, D.P.Y.,
Gee, S.J., Hammock, B.D., 2008. In vitro biologic activities of the
antimicrobials triclocarban, its analogs, and triclosan in
bioassay screens: receptor-based bioassay screens.
Environmental Health Perspectives 116, 1203e1210.
Allmyr, M., Adolfsson-Erici, M., McLachlan, M.S., Sandborgh-
Englund, G., 2006. Triclosan in plasma and milk from Swedish
nursing mothers and their exposure via personal care
products. Science of the Total Environment 372, 87e93.
Aslund, M.L.W., Rutter, A., Reimer, K.J., Zeeb, B.A., 2008. The
effects of repeated planting, planting density, and specific
transfer pathways on PCB uptake by Cucurbita pepo grown in
field conditions. Science of the Total Environment 405, 14e25.
Aslund, M.L.W., Zeeb, B.A., Rutter, A., Reimer, K.J., 2007. In situ
phytoextraction of polychlorinated biphenyl (PCB) -
contaminated soil. Science of the Total Environment 374, 1e12.
Bester, K., 2003. Triclosan in a sewage treatment process e
balances and monitoring data. Water Research 37, 3891e3896.
Calafat, A.M., Ye, X., Wong, L.Y., Reidy, J.A., Needham, L.L., 2008.
Urinary concentrations of triclosan in the US population:
2003e2004. Environmental Health Perspectives 116, 303e307.
Cha, J.M., Cupples, A.M., 2009. Detection of the antimicrobials
triclocarban and triclosan in agricultural soils following land
application of municipal biosolids. Water Research 43,
2522e2530.
Chu, S.G., Metcalfe, C.D., 2007. Simultaneous determination of
triclocarban and triclosan in municipal biosolids by liquid
chromatography tandem mass spectrometry. Journal of
Chromatography A 1164, 212e218.
Coogan, M.A., La Point, T.W., 2008. Snail bioaccumulation of
triclocarban, triclosan, and methyltriclosan in a North Texas,
USA, stream affected by wastewater treatment plant runoff.
Environmental Toxicology and Chemistry 27, 1788e1793.
Drewes, J.E., Heberer, T., Rauch, T., Reddersen, K., 2003. Fate of
pharmaceuticals during ground water recharge. Ground
Water Monitoring and Remediation 23, 64e72.
Dzantor, E.K., Chekol, T., Vough, L.R., 2000. Feasibility of using
forage grasses and legumes for phytoremediation of organic
pollutants. Journal of Environmental Science and Health Part
A-Toxic/Hazardous Substances & Environmental Engineering
35, 1645e1661.
EPA, 2002. High Production Volume (HPV) Chemical Challenge
Program Data Availability and Screening Level Assessment for
Triclocarban. The TCC Consortium.
EPA, 2003. TSCA Chemical Substances Inventory Washington D.C.
Essandoh, H.M.K., Tizaoui, C., Mohamed, M.H.A., Amy, G.,
Brdjanovic, D., 2010. Fate of triclocarban during soil aquifer
treatment: soil column studies. Water Science and
Technology 61, 1779e1785.
Fraker, S.L., Smith, G.R., 2004. Direct and interactive effects of
ecologically relevant concentrations of organic wastewater
contaminants on Rana pipiens tadpoles. Environmental
Toxicology 19, 250e256.
Halden, R.U., Paull, D.H., 2004. Analysis of triclocarban in aquatic
samples by liquid chromatography electrospray ionization
mass spectrometry. Environmental Science and Technology
38, 4849e4855.
5552 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 4 5 e5 5 5 2
Halden, R.U., Paull, D.H., 2005. Co-occurrence of triclocarban and
triclosan in US water resources. Environmental Science and
Technology 39, 1420e1426.
Heath, R.J., Yu, Y.T., Shapiro, M.A., Olson, E., Rock, C.O., 1998.
Broad spectrum antimicrobial biocides target the FabI
component of fatty acid synthesis. Journal of Biological
Chemistry 273, 30316e30320.
Heidler, J., Halden, R.U., 2007. Mass balance assessment of
triclosan removal during conventional sewage treatment.
Chemosphere 66, 362e369.
Heidler, J., Sapkota, A., Halden, R.U., 2006. Partitioning,
persistence, and accumulation in digested sludge of the
topical antiseptic triclocarban during wastewater treatment.
Environmental Science and Technology 40, 3634e3639.
Hülster, A., Müller, J.F., Marschner, H., 1994. Soil-plant transfer of
polychlorinated dibenzo-p-dioxins and dibenzofurans to
vegetables of the cucumber family (Cucurbitaceae).
Environmental Science and Technology 28, 1110e1115.
Johnson, D.R., Czechowska, K., Chevre, N., van der Meer, J.R.,
2009. Toxicity of triclosan, penconazole and metalaxyl on
Caulobacter crescentus and a freshwater microbial community
as assessed by flow cytometry. Environmental Microbiology
11, 1682e1691.
Lawrence, J.R., Zhu, B., Swerhone, G.D.W., Roy, J., Wassenaar, L.I.,
Topp, E., Korber, D.R., 2009. Comparative microscale analysis
of the effects of triclosan and triclocarban on the structure
and function of river biofilm communities. Science of the
Total Environment 407, 3307e3316.
Liu, F., Ying, G.G., Yang, L.H., Zhou, Q.X., 2009. Terrestrial
ecotoxicological effects of the antimicrobial agent triclosan.
Ecotoxicology and Environmental Safety 72, 86e92.
Lunney, A.I., Zeeb, B.A., Reimer, K.J., 2004. Uptake of weathered
DDT in vascular plants: potential for phytoremediation.
Environmental Science and Technology 38, 6147e6154.
Reinwald, S., Akabas, S.R., Weaver, C.M., 2010. Whole versus the
piecemeal approach to evaluating soy. Journal of Nutrition
140, 2335Se2343S.
Sabourin, L., Beck, A., Duenk, P.W., Kleywegt, S., Lapen, D.R., Li, H.
X., Metcalfe, C.D., Payne, M., Topp, E., 2009. Runoff of
pharmaceuticals and personal care products following
application of dewatered municipal biosolids to an
agricultural field. Science of the Total Environment 407,
4596e4604.
Saleh, S., Haddadin, R.N.S., Baillie, S., Collier, P.J., 2011. Triclosan
e an update. Letters in Applied Microbiology 52, 87e95.
Sapkota, A., Heldler, J., Halden, R.U., 2007. Detection of
triclocarban and two co-contaminating chlorocarbanilides in
US aquatic environments using isotope dilution liquid
chromatography tandem mass spectrometry. Environmental
Research 103, 21e29.
Snyder, E.H., O’Connor, G.A., McAvoy, D.C., 2010. Measured
physicochemical characteristics and biosolids-borne
concentrations of the antimicrobial triclocarban (TCC).
Science of the Total Environment 408, 2667e2673.
Stickler, D.J., Jones, G.L., 2008. Reduced susceptibility of Proteus
mirabilis to triclosan. Antimicrobial Agents and Chemotherapy
52, 991e994.
Suller, M.T.E., Russell, A.D., 2000. Triclosan and antibiotic
resistance in Staphylococcus aureus. Journal of Antimicrobial
Chemotherapy 46, 11e18.
USDA (United States Department of Agriculture), 2011. Economic
Research Service. Vegetables and Melon Data. www.ers.usda.
gov/Data/Vegetables/ (accessed 15.03.11.).
USDA (United States Department of Agriculture), Office of
Communications, 2001e2002. Agriculture Fact Book:
2001e2002. www.usda.gov/factbook, Washington, DC, p. 161,
ISBN: 001-000-04709-4.
Veldhoen, N., Skirrow, R.C., Osachoff, H., Wigmore, H.,
Clapson, D.J., Gunderson, M.P., Van Aggelen, G., Helbing, C.C.,
2007. The bactericidal agent triclosan modulates thyroid
hormone-associated gene expression and disrupts
postembryonic anuran development (vol. 80, p. 217, 2006).
Aquatic Toxicology 83, 84.
Walsh, S.E., Maillard, J.Y., Russell, A.D., Catrenich, C.E.,
Charbonneau, D.L., Bartolo, R.G., 2003. Development of
bacterial resistance to several biocides and effects on
antibiotic susceptibility. Journal of Hospital Infection 55,
98e107.
Wang, X.P., White, J.C., Gent, M.P.N., Iannucci-Berger, W.,
Eitzer, B.D., Mattina, M.J.I., 2004. Phytoextraction of weathered
p,p0 -DDE by zucchini (Cucurbita pepo) and cucumber (Cucumis
sativus) under different cultivation conditions. International
Journal of Phytoremediation 6, 363e385.
White, J.C., Wang, X.P., Gent, M.P.N., Iannucci-Berger, W.,
Eitzer, B.D., Schultes, N.P., Arienzo, M., Mattina, M.I., 2003.
Subspecies-level variation in the phytoextraction of
weathered p,p’-DDE by Cucurbita pepo. Environmental Science
and Technology 37, 4368e4373.
Wu, C., Spongberg, A.L., Witter, J.D., Fang, M., Czajkowski, K.P.,
2010. Uptake of pharmaceutical and personal care products by
soybean plants from soils applied with biosolids and irrigated
with contaminated water. Environmental Science and
Technology 44, 6157e6161.
Wu, C.X., Spongberg, A.L., Witter, J.D., 2009. Adsorption and
degradation of triclosan and triclocarban in soils and
biosolids-amended soils. Journal of Agricultural and Food
Chemistry 57, 4900e4905.
Ying, G.G., Kookana, R.S., 2007. Triclosan in wastewaters and
biosolids from Australian wastewater treatment plants.
Environment International 33, 199e205.
Ying, G.G., Yu, X.Y., Kookana, R.S., 2007. Biological degradation of
triclocarban and triclosan in a soil under aerobic and
anaerobic conditions and comparison with environmental
fate modelling. Environmental Pollution 150, 300e305.
Zhang, H., Chen, J., Ni, Y., Zhang, Q., Zhao, L., 2009. Uptake by
roots and translocation to shoots of polychlorinated dibenzo-
p-dioxins and dibenzofurans in typical crop plants.
Chemosphere 76, 740e746.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Speciation of trace inorganic contaminants in corrosion scales

and deposits formed in drinking water distribution systems

Ching-Yu Peng*, Gregory V. Korshin

Department of Civil and Environmental Engineering, University of Washington, Box 352700, Seattle, WA 98105-2700, United States

article info abstract

Article history: Sequential extractions utilizing the modified Tessier scheme (Krishnamurti et al., 1995) and
Received 27 December 2010 measurements of soluble and particulate metal released from suspended solids were used
Received in revised form in this study to determine the speciation and mobility of inorganic contaminants (As, Cr, V,
21 July 2011 U, Cd, Ni, and Mn) found in corrosion scales and particles mobilized during hydraulic
Accepted 9 August 2011 flushing events. Arsenic, chromium and vanadium are primarily associated with the
Available online 30 August 2011 mobilization-resistant fraction that is resistant to all eluents used in this study and also
bound in highly stable crystalline iron oxides. Very low concentrations of these elements
Keywords: were released in resuspension experiments. X-ray absorbance measurements demon-
Arsenic strated that arsenic in the sample with the highest As concentration was dominated by
Nickel As(V) bound by iron oxides. Significant fractions of uranium and cadmium were associated
Manganese with carbonate solids. Nickel and manganese were determined to be more mobile and
Uranium significantly associated with organic fractions. This may indicate that biofilms and natural
Corrosion organic matter in the drinking water distributions systems play an important role in the
Surface scales accumulation and release of these inorganic contaminants.
Inorganic contaminant ª 2011 Elsevier Ltd. All rights reserved.
Mobilization
Fractionation

1. Introduction These studies have demonstrated that while heavy

Accumulation of inorganic contaminants in corrosion solids
and sediments commonly found in drinking water distribu-
tion systems (DWDSs) has been addressed in significant detail
in prior research (Lytle et al., 2004; Schock, 2005; Schock et al.,
2008; Friedman et al., 2010). These studies have demonstrated
that several heavy elements can be found in scales formed on
metals corroding in drinking water. For instance, Lytle et al.
(2004) demonstrated that iron-based corrosion products and
sediments adsorb and concentrate arsenic in DWDSs. Lead
pipe scales have also been reported to have significant levels
of As, Cd, Cr, Hg, and V (Schock et al., 2008; Gerke et al., 2009;
Kim and Herrera, 2010).
metals and allied elements are typically present only at trace
levels in treated potable water entering DWDS, they tend to
accumulate in DWDS corrosion solids where their concen-
trations can exceed those in the influent treated water by
several orders of magnitude (Valentine and Stearns, 1994;
Reiber and Dostal, 2000; Lytle et al., 2004; Schock et al.,
2008; Gerke et al., 2009). For instance, Reiber and Dostal
(2000) and Lytle et al. (2004) determined that corrosion
solids formed in treated water having arsenic levels <10 mg/L
had arsenic concentrations of up to several hundred mg/kg.
This corresponds to more than a hundred-fold enrichment of
corrosion solids with As compared to its concentration in the
ambient water.

* Corresponding author. Tel.: þ1 206 660 2233.

E-mail address: cypeng@u.washington.edu (C.-Y. Peng).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.017
5554 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
Our prior research (Friedman et al., 2010; Peng et al.,
submitted for publication) demonstrated similar trends for
both radionuclides (radium-226, radium-228) and inorganic
contaminants (antimony, arsenic, barium, chromium, lead,
nickel, selenium, thallium, uranium, vanadium) present in
stable surface scales and particles released during hydrant
flushing that mobilizes relatively loosely bound DWDS solids
(Vreeburg et al., 2009; Husband and Boxall, 2011). While the
concentrations of antimony, selenium, cadmium and thal-
lium in these samples were always low (<2 mg/g), levels of
other toxic elements such as lead and arsenic were variable
ranging from <0.1 mg/g to >1000 mg/g (Hill et al., 2010; Peng
et al., submitted for publication).
The accumulation of trace inorganic contaminants in
DWDS solids can have consequences for the health of exposed
populations if these contaminants are released from the
scales thus resulting in their high levels at consumers’ tap
(Fisher et al., 2000; Lytle et al., 2010). Remobilization of
contaminants bound by corrosion solids can occur through
several mechanisms, notably physical mobilization of partic-
ulates from the solid matrix (e.g., during hydrant flush events
or due to the destabilization caused by changes of water
chemistry) and/or chemical release via desorption and/or
dissolution.
Ultimately, the mobility and availability of contaminants
accumulated in DWDS solids depend on their physico-
chemical speciation that encompasses states ranging from
mobile species that can be released via desorption to those
tightly bound by stable crystalline oxides of iron and, in lesser
extent, Ca and Mg carbonates, silica and manganese oxides
predominating typical corrosion scales (Linge, 2008). Thus, the
determination of mobility and other aspects of the speciation
of trace level inorganic contaminants in DWDS solids require
that the modes of binding of these elements in the solid
matrix be ascertained.
That goal can be achieved via selective extractions that
allow estimating contributions of the target elements bound
by a priori defined types of solid phases. One sequential
extraction procedure that has been extensively employed to
speciate inorganic contaminants in soils and sediments was
proposed by Tessier et al. (1979). It fractionates heavy metal
species found in soils or sediments into five operationally
defined fractions that correspond to exchangeable, carbonate-
bound, iron and manganese oxide-bound, organically bound
and residual metal. The Tessier’s scheme was later modified
to separate the fraction comprising species bound by iron and
manganese oxides into narrower-defined fractions such as
metaleorganic complex-bound, easily-reducible metal oxide-
bound, amorphous mineral colloid-bound, and crystalline
iron oxide-bound species (Shuman, 1985; Krishnamurti et al.,
1995). For this matter, iron and manganese minerals
compounds, notably ferrihydrite, goethite, lepidocrocite,
magnetite, hydrous manganese oxides, are commonly found
in the corrosion scales and deposits of DWDSs (Benjamin
et al., 1996; Sarin et al., 2001; Peng et al., 2010). Although the
composition and properties of DWDS corrosion solids and
deposits may be different from those of soils and sediments
for which the Tessier’s and related schemes have been
developed, it is reasonable to assume that the sequential
extraction can be a powerful fractionation method to
investigate the speciation of heavy metals accumulated in
DWDS solids.
To our knowledge, sequential extraction methods have not
been consistently applied to determine speciation of heavy
metals found in DWDS corrosion solids and deposits. In this
study, we applied this approach to examine the speciation and
potential mobility of several inorganic contaminants (e.g., As,
Cr, V, U, Cd, Ni, Mn) occurring at various levels in represen-
tative samples of DWDS corrosion scales. Of the elements
targeted in this study, arsenic, chromium, uranium and
cadmium are regulated by the National Primary Drinking
Water Standards (NPDWSs). Manganese is regulated by the
National Secondary Drinking Water Standards (NSDWSs),
which are non-enforced guidelines concerned with contami-
nants that may cause aesthetic or cosmetic effects in drinking
water (U.S. EPA, 2009a). Vanadium has been listed in the
Drinking Water Contaminant Candidate List 3 (CCL3) (U.S.
EPA, 2009b). Though currently unregulated by the U.S. EPA,
nickel was regulated in the past due to its adverse health
effects. The current WHO guideline for nickel is 70 mg/L
(Guidelines for Drinking-Water Quality, 2008).
In addition to sequential extractions, separate experiments
were carried to determine concentrations of the target
elements in soluble and particulate fractions (passing
nominal pore sizes 5, 2, 1, and 0.4 mm) released to the ambient
water. Finally, the structure-sensitive method of X-ray
adsorption spectroscopy (XAS) was used to examine the
speciation of As bound by one representative DWDS corrosion
solids sample.
2. Materials and methods
2.1. Solid samples characteristics
Three pipe specimens (CC-A, CC-B, and CC-D) and two
hydrant flush samples (J-E and J-J) were selected for the
experiments based on higher concentrations of trace inor-
ganic contaminants in them. These samples were obtained
from the earlier study funded by Water Research Foundation
(Friedman et al., 2010). Relevant characteristics and elemental
compositions of the examined samples are shown in Table 1.
Prior to all experimental procedures, the solids were dried at
103  C for 1 day, crushed using a mortar and pestle, passed
through a number 50 sieve (300-mm mesh) and homogenized.
Four samples of the samples that had a sufficient mass were
analyzed by X-ray Diffraction (XRD) measurement to identify
their mineralogical phases. XRD analyses were described in
more detail in Peng et al. (2010) and Friedman et al. (2010).
Specific surface area was carried out using a Quantachrome
NOVA 4200e instrument and calculated using multi-point BET
method.
2.2. Sequential extraction procedures and operationally
defined metal fractions
Speciation of the selected inorganic contaminants (As, Cr, V, U,
Cd, Ni, Mn) in the solid samples was determined based on the
sequential extraction scheme described by Krishnamurti et al.
(1995). Important details of the scheme are summarized in
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3 5555

Table 1 e Characteristics and elemental compositions of solid samples examined in this study.
Sample ID Sample Pipe Fe Mn As Cr V U Cd Ni BETb Mineralogyc
type materiala (mg/g) (mg/g) (mg/g) (mg/g) (mg/g) (mg/g) (mg/g) (mg/g) (m2/g)

CC-A Pipe Cast iron 235,000 46,692 40.2 7.13 196 5.08 11.2 121 97.42 Magnetite, quartz
specimen
CC-B Pipe Cast iron 117,000 28,957 234 29.1 71.3 1.8 3.4 92.5 NMd NM
specimen
CC-D Pipe Cast iron 252,000 21,654 140 78.5 37.4 15.6 2.51 296 20.08 Siderite, quartz,
specimen hydroxyapatite
J-E Hydrant Cast iron 146,000 760 3.88 197 22.2 1.12 0.18 136 23.4 Calcite, dolomite,
flush quartz
J-J Hydrant Cast iron 335,000 1,091 30.9 65.2 13.9 1.25 0.48 107 50.82 Goethite, magnetite,
flush ferrihydrite, calcite,
quartz

a For hydrant flush samples, pipe material refers to the type of pipe used to distribute water in the flushing zone.
b Crushed samples were used to analyze the BET surface area.
c Mineralogy was based on X-ray diffraction (XRD) measurement.
d NM: inadequate mass for testing.

Table 2. The following fractions that correspond to different
operationally defined modes of binding of inorganic contami-
nants in the solid phases were established:
(1) Exchangeable metal: heavy elements associated with
exchangeable binding sites. This mode of binding is
expected to result in relatively easy displacement of the
bound metals by competing ions.
(2) Carbonate-bound metal: heavy elements associated with
sorption on or occlusion into carbonates, typically calcite
(CaCO3) and dolomite (CaMg(CO3)2). Elements coprecipi-
tated with calcite can be dissolved by acidic sodium
acetate.
(3) Metaleorganic complexes: heavy elements associated with
humic species present in the solid matrix.
(4) Easily reducible metal oxide-bound metal: heavy elements
associated with easily reducible metal oxides such as
manganese (III, IV) oxides. Hydroxylamine hydrochloride
(NH2OH$HCl) is specific to Mn oxides, leaving crystalline
iron oxides unaffected and dissolving minimal amounts of
amorphous iron oxides.
(5) Organic-bound metal: heavy elements associated with
organic matter other than humic substances.
(6) Amorphous mineral colloid-bound metal: heavy elements
associated with amorphous iron oxides and poorly crys-
talline aluminosilicate mineral colloids. Acidic ammonium
oxalate ((NH4)2C2O4) performed in dark is specific for dis-
solving amorphous minerals.
(7) Crystalline iron oxide-bound metal: heavy elements associ-
ated with crystalline iron oxides.
(8) Mobilization-resistant metal: heavy elements associated
with mineral lattices resistant to the above successive
sequential extractions (Krishnamurti and Naidu, 2002;
Linge, 2008).
Sequential extractions were carried out in duplicate using
1 g of each solid sample (except CC-B, in which case 0.4 g was
used). Requisite amounts of the solids were placed in 50 mL
metal free polypropylene centrifuge tubes (VWR). After each
extraction, the tubes containing solids were centrifuged
washing for 10 min at 4000 rpm to separate the supernatant.
The residual solids were rinsed with 10 mL of de-ionized
distilled water, centrifuged again for 10 min at 4000 rpm and
the resulting supernatant collected for further analyses. All
chemicals used in the sequential extraction procedure were all
ACS grade. Their solutions were checked to determine the

Table 2 e Sequential extraction scheme for speciation used in this study (Krishnamurti et al., 1995).
Step Fractions Reagent Reaction time and
temperature

1 Exchangeable 10 mL of 1 M Mg(NO3)2 at pH 7 4 h at 25  C
2 Carbonate-bound 25 mL of 1 M CH3CO2Na at pH 5 6 h at 25  C
3 Metaleorganic complex-bound 30 mL of 0.1 M Na4P2O7$10H2O at pH 10 20 h at 25  C
4 Easily reducible metal oxide-bound 20 mL of 0.1 M NH2OH$HCl in 0.01 M HNO3 30 min at 25  C
5 Organic-bound 5 mL of 30% H2O2 at pH 2, 3 mL of 0.02 M HNO3 2 h at 85  C
3 mL of 30% H2O2 at pH 2 2 h at 85  C
Cool, 10 mL Mg(NO3)2 in 20% HNO3 30 min at 25  C
6 Amorphous mineral colloid-bound 10 mL of 0.2 M (NH4)2C2O4 (adjusted to pH 3 with 0.2 M H2C2O4) 4 h at 25  C (dark)
7 Crystalline iron oxide-bound 25 mL of 0.2 M (NH4)2C2O4 (adjusted to pH 3 with 0.2 M H2C2O4) 30 min at 95  C
in 0.1 M ascorbic acid
8 Mobilization-resistant Digestion with HNO3 and 30% H2O2
5556 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
concentration of the target trace metals in them
(Supplementary data Table S1).
2.3. Separation of soluble and particulate fractions of
metal released from suspended corrosion solids
To determine concentration of the target elements in solids
assigned to soluble and particulate fractions released from
suspended corrosion solids, 20 mg of each sample was
dispersed into 100 mL water with alkalinity 100 mg/L as CaCO3
and pH 7.6 placed in 125 mL Nalgene metal free bottles.
Samples were placed on a shaker and aliquots were taken for
analyses after 30, 60 and 120 min of exposure. The samples
were filtered through filters with nominal pore size of 5, 2, 1,
and 0.4 mm (Whatman Nuclepore polycarbonate track-etched
membranes). The filter units and syringes were prewashed
with 1% HNO3 and then rinsed with de-ionized distilled water.
Sample aliquots of 15 mL were filtered directly into 15 mL
metal free tubes. The first 5 mL of filtrate was discarded.
Filtered samples were acidified to 1% HNO3 with ultrapure
HNO3 and stored at 4  C prior to analysis.
2.4. X-ray absorption spectroscopy
X-ray absorbance near edge structure (XANES) and extended
X-ray absorption fine structure spectra of three arsenic model
compounds and CC-D sample were carried out at beamline
X19-A of the National Synchrotron Light Source (NSLS) at
Brookhaven National Laboratory (BNL), Upton, NY. The X-ray
energy was varied from 200 eV below to 1000 eV above the
absorption K-edges of As (Ek ¼ 11,868 eV) using a double
crystal Si(111) monochromator. Samples were ground and
mounted on a piece of 3M scotch tape via brushing; the tape
was then folded several times until the intensity of the signal
reached a suitable range for numerical analysis. Two model
compounds (sodium arsenite NaAsO2 and sodium arsenate
(Na2HAsO4$7H2O)) were measured in the transmission mode;
while the adsorbed As(V)eFe (synthetic model compound) and
CC-D were analyzed in the fluorescence mode. Fluorescence
mode was carried out by using the Lytle detector with
a Germanium filter and Soller-type slits to minimize the
fluorescence background. To improve the signal-to noise
ratio, at least 4 measurements were obtained and averaged for
the diluted samples. Gold metal foil was always measured in
the transmission mode simultaneously with all other samples
and used as the reference for the alignment of energies.
Energy calibration was performed with Athena software
and achieved by assigning the first inflection point of the
simultaneously measured Au foil to 11,919 eV. The edge jump
of pre- and post-edge regions was then normalized. The
function was then transformed from energy unit (eV) to
photoelectron wave vector (k) unit (A1) to produce the EXAFS
3 Average (%)
function (c(k)). The k -weighted c(k) function was then Fourier
transformed (FT) using a Hanning window to create the
radical structure function (RSF) in R-space ( A). The experi-
mental EXAFS data were then fitted with coordination number
(N ), interatomic distance (R), and the DebyeeWaller param-
eter (s2) using Artemis program. Linear least-squared fitting
(LSF) was done on the XANES spectra after background
subtraction and normalization with Athena program.
2.5. Elemental analysis
Concentrations of several elements in samples generated
using sequential extractions and filtration experiments that
utilized filters with varying sizes were determined by induc-
tively coupled plasma-mass spectroscopy (ICP-MS) (Perki-
nElmer ELAN DRC-e ICP-MS). The elements quantified in these
measurements included arsenic (As), chromium (Cr), vana-
dium (V), uranium (U), cadmium (Cd), nickel (Ni) and manga-
nese (Mn). Quality control samples, including laboratory-
fortified blanks and laboratory-fortified samples, were per-
formed for every ten samples analyzed. Average elemental
recoveries ranged from 85.2 to 92.8% for the laboratory-
fortified samples.
A certified reference material (CRM) (Loamy Sand, CRM
024-050) was purchased from RTC (Laramie, Wyoming) to
quality check the analytical method. Results obtained from
ICP-MS analysis of certified reference material are summa-
rized in Supplementary data Table S2.
3. Results and discussion
Average extraction efficiencies for several elements examined
in this study are presented in Table 3. They ranged from 95 to
114% with standard deviations shown in Table 3. The percent
distribution of inorganic contaminants species extracted from
individual solids can be found in Fig. 1. Three pipe specimens
(CC-A, CC-B, and CC-D) originated from the same drinking
water distribution system but were taken from three different
locations. Very similar distributions of inorganic contaminants
were found in samples CC-A and CC-D. However, the data for
sample CC-B differed from them. Since these specimens orig-
inated from the utility that relies on groundwater from
multiple wells, ambient water chemistry was likely to vary
between the wells thus potentially resulting in variations of
DWDS solids’ properties in the distribution system. Two
hydrant flush samples (J-E and J-J) were provided by the utility
that relies on groundwater from multiple wells and operates
a larger water system than utility CC. Accordingly, water
quality conditions of distribution system J are likely to vary
spatially and temporally because of the different source waters
and their blending in the system. Indeed, the compositions of
sample J-E and J-J were very different. The hydraulically
mobilized material in sample J-E had comparatively low levels
Table 3 e Average extraction efficiencies and
corresponding standard deviations (SD) of five corrosion
solids and deposits examined in this study.
SD (%)
Arsenic (As) 107.2 13.6
Chromium (Cr) 113.5 14.6
Vanadium (V) 99.6 17.9
Uranium (U) 106.1 15.2
Cadmium (Cd) 94.8 28.3
Nickel (Ni) 107.1 31.2
Manganese (Mn) 103.2 24.5
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3 5557

a 100%
b100%
90% 90%
Mobilization-resistant Mobilization-resistant
80% 80%
Crystalline Fe oxide-bound Crystalline Fe oxide-bound
70% 70%
Amorphous mineral colloid-bound Amorphous mineral colloid-bound
60% 60%
H2O2 extractable organic-bound
H2O2 extractable organic-bound 50%
50% Easily reducible metal oxide-bound
Easily reducible metal oxide-bound 40%
40% Metal-organic complex-bound
Metal-organic complex-bound
30% Carbonate-bound
30% Carbonate-bound
20% Exchangeable
20% Exchangeable
10%
10%
0%
0% As Cr V U Cd Ni Mn
As Cr V U Cd Ni Mn
c d100%
100% 90%
90% Mobilization-resistant
Mobilization-resistant 80%
80% Crystalline Fe oxide-bound
Crystalline Fe oxide-bound 70%
70% Amorphous mineral colloid-bound
Amorphous mineral colloid-bound 60%
60% H2O2 extractable organic-bound
H2O2 extractable organic-bound
50% 50%
Easily reducible metal oxide-bound Easily reducible metal oxide-bound
40% Metal-organic complex-bound 40% Metal-organic complex-bound
30% Carbonate-bound 30% Carbonate-bound
20% Exchangeable
20% Exchangeable
10%
0% 10%
As Cr V U Cd Ni Mn 0%
As Cr V U Cd Ni Mn

e
100%
90%
Mobilization-resistant
80%
Crystalline Fe oxide-bound
70%
Amorphous mineral colloid-bound
60%
H2O2 extractable organic-bound
50%
Easily reducible metal oxide-bound
40% Metal-organic complex-bound
30% Carbonate-bound
20% Exchangeable
10%
0%
As Cr V U Cd Ni Mn

Fig. 1 e Percent distribution of inorganic contaminants in pipe specimens (a) CC-A, (b) CC-B, (c) CC-D, and hydrant flush
solids (d) J-E, (e) J-J examined following the sequential extraction scheme.

of iron (14.6 wt%) and manganese (760 mg/g), but a relatively
high level of total carbon (6.1 wt%). By comparison, levels of Fe,
total carbon and Mn in sample J-J were 33.5 wt%, 3.4 wt% and
1091 mg/g, respectively. The distributions of inorganic
contaminants found in samples J-E and J-J were also different,
especially for elements associated with the organic-bound
fraction. Although pipe specimens and hydrant flush samples
examined in this study were not from the same distribution
system, intrinsic differences in the generation of these two
types of samples warrant their comparison that can reveal
potentially important differences in retention and release of
retained contaminant. The fraction of arsenic associated with
crystalline iron oxides was next in importance (19.8%), with
contributions of amorphous mineral colloid-bound, organic-
bound, metaleorganic complex-bound fractions becoming
progressively smaller (9.3%, 6.5% and 4%, respectively).
Arsenic fractions assumed to have highest potential mobility
had lowest, almost negligible contributions to the total
arsenic, with the arsenic bound to easily reducible metal
Percentage of inorganic contaminants in various fractions

100%
inorganic contaminants from these substrates. Average 90%
percent distributions of inorganic contaminants species 80%
Mobilization-resistant

shown in Fig. 1(a)e(e) are present in Fig. 2. The following 70%

Crystalline Fe oxide-bound
Amorphous mineral colloid-bound
discussion of chemical fractionation data will focus on the data 60%
H2O2 extractable organic-bound
summarized in Fig. 2. A summary of average values and stan- 50%
Easily reducible metal oxide-bound
dard deviations of contributions of different fractions of heavy 40%
Metal-organic complex-bound
metals established using the sequential extraction scheme 30% Carbonate-bound
(Fig. 2) is provided in the Supplementary data Table S3. 20% Exchangeable
10%

3.1. Arsenic 0%
As Cr V U Cd Ni Mn

Chemical fractionation data (Fig. 2) showed that on the Fig. 2 e Average percent distributions of inorganic
average, the distribution of As in five corrosion solids was contaminants species in the corrosion solids and deposits
predominated by the mobilization-resistant fraction (59.3%) examined in this study following the sequential extraction
that is deemed to correspond to a very low mobility of the scheme.
5558 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3

Table 4 e Average concentrations (mg/L) of inorganic contaminants released from suspension of corrosion solids and
deposits.
Dissolved (<0.45 mm) Particulate (0.45e1 mm) Particulate (1e2 mm) Particulate (2e5 mm)

PS HF PS HF PS HF PS HF

As 0.170 0.132 0.039 0 0.067 0.001 0.163 0.002

U 0.019 0.006 0.001 0.001 0.003 0.001 0.019 0.004
Cd 0.051 0.024 0.002 0.004 0.005 0.004 0.037 0.022
Ni 3.487 0.837 0.114 0.116 0.122 0.162 0.194 0.868
Mn 377.132 7.294 14.038 0.111 15.594 0.515 39.542 1.947

PS: pipe specimen. Three pipe specimens include CC-A, CC-B and CC-D.
HF: hydrant flush solid. Two hydrant flush solids include J-E and J-J.

oxides, carbonate-bound and exchangeable sites having
concentrations of 0.6%, 0.3% and 0.2%, respectively.
These results demonstrate that arsenic in the corrosion
solids was dominated by the mobilization-resistant and
crystalline iron oxide fractions contributing together to ca.
80% of the total. This may also indicate that As in these solids
is mostly occluded in the stable mineral structures and, as
a result, it is not likely to be easily mobilized. This is not
surprising as As(V) is well known to have strong retention on
mineral surfaces, especially iron minerals (Wilkie and Hering,
1996; Dixit and Hering, 2003). This also shows that the level of
soluble arsenic release associated with possible chemical or
physical perturbation affecting corrosion solids is expected by
largely negligible.
This conclusion was confirmed by the data obtained when
these samples were suspended in water having the pH 7.6 and
alkalinity 100 mg/L as CaCO3, respectively. The average
concentrations of inorganic contaminants released into
ambient water from the five examined solids and belonging to
different size fractions are shown in Table 4. Average concen-
trations (in log scale) of released inorganic contaminants of
individual solids in different size fractions are provided in Fig. 3.
For arsenic, 2.8% and 8.7%, respectively, of the total As
were released from suspended pipe specimens and hydrant

a U 0.4 µm
b
U 0.4 µm
Cd 1 µm 1 µm
Cd
As 2 µm 2 µm
As
5 µm
Ni Ni 5 µm
Mn Mn

-3 -2 -1 0 1 2 3 4 -3 -2 -1 0 1 2 3 4
c d
U 0.4 µm U 0.4 µm

Cd 1 µm Cd 1 µm

2 µm As 2 µm
As
5 µm
5 µm Ni
Ni
Mn
Mn
-3 -2 -1 0 1 2 3 4
-3 -2 -1 0 1 2 3 4

e
U 0.4 µm
Cd 1 µm

As 2 µm
5 µm
Ni
Mn

-3 -2 -1 0 1 2 3 4

Log (released inorganic concentrations, in µg/L)

Fig. 3 e Concentrations of inorganic contaminants released from suspended corrosion solids (200 mg/L) at pH 7.6, alkalinity
100 mg/L passing filters with varying nominal pore sizes. (a) CC-A, (b) CC-B, (c) CC-D, (d) J-E, and (e) J-J solid samples.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3 5559

Percentage of inorganic contaminants in various fractions

a
100%

80%

Normalized µ(E)
2~5 µm
60%
1~2 µm
0.4~1 µm
40% <0.4 µm

20%
LCF
0%
PS HF PS HF PS HF PS HF PS HF As(V)-Fe (91.7%)
NaAsO2 (8.3%)
As U Cd Ni Mn

11840 11860 11880 11900 11920 11940

Fig. 4 e Average contributions of inorganic contaminants
released from suspended corrosion solids and deposits
and associated with different size fractions. (PS: pipe b
specimen, including CC-A, CC-B and CC-D. HF: hydrant

Fourier Transform of X(k)k3

flush solid, including J-E and J-J.)

flush solids and passing through a 5 mm filter. The average As

levels released from suspended pipe specimens and hydrant
flush solids into water were 0.44 mg/L and 0.14 mg/L,
respectively.
Most of the arsenic released in the resuspension experi-
ments from three pipe specimens was in the operationally
defined soluble dissolved metal fraction (<0.4 mm) followed by
the fraction assigned to particles with sizes between 2 and 0 1 2 3 4 5
5 mm. These two major size fractions contributed to ca. 80% of
the total As passing a 5 mm filter while the 1e2 mm and Fig. 5 e (a) Linear combination fitting of XANES and (b)
0.4e1 mm fractions constituted the rest, as shown in Fig. 4. Fourier transforms of the EXAFS for As in corrosion solid
The speciation of arsenic released from suspended samples CC-D. The solid bold line in (a) represents the XANES data
exhibited a different distribution pattern for pipe specimens for corrosion solid CC-D. The dotted line in (a) corresponds
and hydrant flush solids. In all cases, actual concentrations of to the linear combination fitting (LCF) of two model
arsenic were very low. In the case of solids retained on the compounds (As(V)eFe and NaAsO2). The solid and dashed
surface of corroding pipe, the average released As concentra- lines in (a) correspond to the XANES data for As(V)eFe and
tions in dissolved fraction (<0.4 mm) and those corresponding NaAsO2 model compounds. The peak positions in (b) are
to three ranges of varying particles sizes (2e5 mm, 1e2 mm and uncorrected for phase shits. The dashed line in (b) is the
0.4e1 mm) were 39%, 37%, 15% and 9%, respectively. In model fits.
contrast, for hydrant flush solids, almost all (98%) released
arsenic concentrations existed in the dissolved fraction.
X-ray adsorption spectroscopy (XAS) was employed to
further explore the chemical nature of arsenic in sample CC-D The local coordination environment of arsenic in corrosion
that had the highest As concentration among all studied solid CC-D was also examined by EXAFS spectroscopy. Fourier
solids. The white-line position of As in its X-ray absorbance transformed spectra of As in CC-D are shown in Fig. 5(b). Peak
spectra is known to be sensitive to the oxidation state of this positions shown in that figure are uncorrected for phase shifts
element, with the white-line locations at 11,871.5 and so that they are slightly shifted from the true interatomic
11,875 eV for As(III) and As(V), respectively (e.g., Manning distances. The experimental spectrum was fitted with a theo-
et al., 2002; Cancès et al., 2008). Accordingly, we examined retical model to yield interatomic distance (R), coordination
the normalized As K-edge XANES spectrum of sample CC-D number (CN), and the DebyeeWaller parameter (s2) based on
and processed the data using linear combinations of XANES the optimized fit. These calculations (their results are
spectra of model compounds with known oxidation states summarized in Table 5) showed that in the first As coordina-
fitted by means of linear least-squared fitting (LSF). Results of tion shell, four oxygen atoms located at a 1.67 
A distance from
this fitting are shown in Fig. 5(a). The LSF procedure indicated the central As atom were present. This AseO distance was
that the arsenic in sample CC-D is dominated by As(V) statistically identical to the average AseO distances in the
components (91%), especially by As(V) having Fe atoms in its model compounds, such as scorodite (FeAsO4$4H2O) and
chemical environment. Less than 9% of the total As present in sodium arsenate (Na2HAsO4$7H2O) (Waychunas et al., 1993),
sample CC-D appears to exist as As(III). and arsenate adsorbed on or coprecipitated with goethite,
5560 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
Table 5 e EXAFS parameters defining local coordination
environment of As in corrosion solid (CC-D).
Shell Distance (
A) Coordination DebyeeWaller
number (CN) factor (s2)
AseO 1.67 4 0.001
AseFe 3.30 2 0.002
lepidocrocite, hematite or ferrihydrite (Fendorf et al., 1997;
Sherman and Randall, 2003). The second peak in the spectra
shown in Fig. 5(b) appears to indicate the presence of two iron
atoms located at distance of 3.3 A from the central As atom.
This AseFe distance appears to correspond to As associated
with the surfaces of iron oxides by double-corner sharing
(Fendorf et al., 1997; Sherman and Randall, 2003; Cancès et al.,
2005). These data confirm that As present in the examined
sample is in primarily þ5 oxidation state and it is at least
partly associated with Fe-containing matrixes. This is in
accord with data of prior research showing that As(V) tends to
sorb very strongly on and be bound by mineral surfaces,
especially iron minerals; while As(III) is considerably more
mobile (Manning et al., 2002; Meng et al., 2002).
3.2. Chromium and vanadium
Chromium and vanadium exhibited somewhat similar frac-
tionation patterns (Fig. 2). The majority of Cr in the examined
solids was associated with the mobilization-resistant (48.4%),
crystalline iron oxide-bound (18.6%) and organic-bound (15.6%)
fractions. The proportions of V occurring in the mobilization-
resistant, crystalline iron oxide-bound and organic-bound
fractions were 27.8%, 20.3% and 24.3%, respectively. A very
low percentage of Cr (0%) and V (0.04%) were found in the
exchangeable fraction. These findings suggest that Cr and V in
these solids are located primarily in the stable matrixes and
appear to be relatively immobile. This result is in agreement
with the results of prior research demonstrating that iron-
based oxides are highly effective sorbents that remove chro-
mium and vanadium in a wide range of water conditions
(Smith and Ghiassi, 2006; Naeem et al., 2007).
This conclusion was supported by the data of determina-
tions of soluble and particulate Cr and V concentrations
released from suspended solids. In these experiments, the
concentrations of Cr and V released were consistently below
the method detection limits (Cr: 0.054 mg/L; V: 0.078 mg/L) and
hence they were not included in Table 4, Figs. 3 and 4.
3.3. Uranium
Sequential extractions showed that in the case of uranium the
mobilization-resistant (31.2%), organic-bound (22.6%), and
carbonate-bound (16.2%) fractions had the highest contribu-
tions. The sum of the contributions of the mobilization-
resistant and crystalline iron oxide-bound fractions accoun-
ted for 41.4% of the total U, which is less compared to all other
elements examined in this study (except Mn).
The data also show that combined contributions of the
organic-bound and metaleorganic complex-bound fractions
were 34.6% of the total uranium, while the contribution of the
carbonate fraction was 16.2% (Fig. 2). The associations of
uranium with these fractions are in agreement with the
strong complexation of the uranyl ion (UO2þ 2 ) prevalent in the
aerobic environment with carbonate and natural organic
matter (e.g., Zhou and Gu, 2005; Bednar et al., 2007; Stewart
et al., 2010).
In experiments with suspended solids, 2.3% (0.042 mg/L)
and 5.2% (0.012 mg/L) of U were released on the average from
three pipe specimens and two hydrant flush solids, respec-
tively. For the three pipe specimens, the average contributions
of uranium associated with the 2e5 mm, 1e2 mm, 0.4e1 mm and
<0.4 mm (soluble) fractions were 46.1%, 6.9%, 1.6% and 45.3%,
respectively. The size fractions of uranium for the hydrant
flush solids were similar, with 50% of released U found in the
dissolved fraction (<0.4 mm), followed by 31.9%, 11.1% and
6.9% associated with the particular fraction of 2e5 mm, 1e2 mm
and 0.4e1 mm, respectively (Fig. 4).
3.4. Cadmium
Chemical fractionation data (Fig. 2) showed that 25.3% of Cd in
five corrosion solids was in the mobilization-resistant frac-
tion. The next dominant fraction was cadmium associated
with crystalline iron oxides (18.5%), followed by organic-
bound (16.1%) and carbonate-bound (12.6%) metal. The
remaining fractions had smaller contributions that decreased
in the order: easily reducible metal oxides (9%) > amorphous
mineral colloid (8.2%) > metaleorganic complex (7.9%) >
exchangeable cadmium (2.4%). Only 43.8% of Cd was retained
in the stable mineral structures; while 31.9% of Cd, including
exchangeable, carbonate, metaleorganic complex and easily
reducible metal oxides, was associated with potential mobile
fractions.
The percentage of Cd released from the suspended solids
was higher than that determined for As, V, Cr and U. Specifi-
cally, 8.8% (0.096 mg/L) and 59.7% (0.054 mg/L) of the total Cd
were released from three pipe specimens and two hydrant
flush solids, respectively. The dominant fraction of Cd
released from three pipe specimens (53.8%) and two hydrant
flush samples (43.8%) was in dissolved (<0.4 mm) fraction. For
three pipe specimens, the next predominant fraction was in
particular fraction of 2e5 mm (38.3%), followed by particular
fraction of 1e2 mm (5.4%) and particular fraction of 0.4e1 mm
(2.6%). For two hydrant flush samples, the particular fraction
of 2e5 mm (41.3%) was also the second important fraction,
followed by particular fraction of 0.4e1 mm (7.8%) and partic-
ular fraction of 1e2 mm (7.1%) (Fig. 4).
3.5. Nickel and manganese
Sequential extractions and filtration results demonstrated the
existence of similar fractionation profiles for nickel and
manganese, especially for pipe specimens (Figs. 2 and 4). This
is possibly associated with a high affinity of hydrated
manganese oxides to nickel and resulting co-accumulation of
these two metals (Green-Pedersen et al., 1997; Trived and Axe,
2001; Trived et al., 2001).
In the case of nickel, the mobilization-resistant fraction
contained the largest percentage (39.1%) of the metal. Organ-
ically bound nickel accounted for 21.3% of the total followed
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
by other potentially mobile fractions including nickel associ-
ated with easily reducible metal oxides (9.1%), followed by
the exchangeable and carbonate-bound Ni (8.8 and 7.6%,
respectively).
The organically bound manganese was the predominant
fraction (28.1%), followed by the mobilization-resistant frac-
tion (27.6%). Nominally mobile fractions of manganese
accounted for 36.4% of the total. These fractions included easily
reducible metal oxides (18.6%), carbonate-bound (9.1%), and
exchangeable (8.7%). The least important fractions are still
amorphous mineral colloid-bound (3.9%), crystalline iron
oxide-bound (2.6%), and metaleorganic complex-bound (1.4%).
Filtration data obtained using suspended solids showed
that on the average 8.8% (3.92 mg/L) and 7.7% (1.98 mg/L) of the
total Ni present in these samples were released into water from
three pipe specimens and two hydrant flush samples, respec-
tively. 89% of Ni released from the pipe specimens was in the
dissolved (<0.4 mm) fraction while in case of hydrant flush
samples only 42.2% of the released Ni passed through 0.4 mm
filter. At the same time, a large part of released Ni (43.8%) from
two hydrant flush samples is in particular fraction of 2e5 mm.
On the other hand, 6.1% (446.3 mg/L) and 5% (9.87 mg/L) of
total Mn were released into water from three pipe specimens
and two hydrant flush samples, respectively. 84.5% and 73.9%
of released Mn from three pipe specimens and two hydrant
flush samples are in the dissolved (<0.4 mm) fraction. It needs
to be recognized however, that manganese concentrations
tend to be two or more magnitudes larger than those of nickel
in the pipe specimens; although the release percentages of
these two elements are very similar, manganese will have
overwhelming concentrations compared to that of nickel.
3.6. Practical implications
The experimental data presented above demonstrate the
existence of pronounced differences in the mobility of inor-
ganic contaminants retained by DWDS corrosion scales and,
when released, potentially affecting human health. The data
show that the physico-chemical properties of As, Cr and V
retained by DWDS corrosion scales accumulated are similar.
These three elements were found to be tightly bound by the
solid matrixes and are expected to exhibit little mobility under
most conditions typical for drinking water systems. However,
because arsenic and, very likely, vanadium exist in anionic
forms in these conditions, their mobility can hypothetically be
increased via competition with and displacement by phos-
phate ions utilized for corrosion control (Jain and Loeppert,
2000; Copeland et al., 2007). Actual occurrence of such
competition remains to be ascertained.
On the other hand, arsenic release may be associated with
particulate matter, as was observed by Lytle et al. (2010) for
a small drinking water system in which high arsenic levels
were associated with iron oxide particles carried in the system
by the source water. Our speciation data similarly indicate
that because As present in corrosion solids is dominated by
the mobilization-resistant and crystalline iron oxide-bound
fractions, its release can occur primarily via particulates
physically dislodged by hydraulic events or via colloidal
mobilization of iron-based corrosion solids caused by changes
of water chemistry. Conditions possibly associated with such
5561
phenomena in drinking water distribution systems need to be
investigated in the future.
Similarly to our observations, high concentrations of vana-
dium (35 to 899 mg/g) in scales formed on surfaces of galvanized
iron were reported by Gerke et al. (2010). In that study, the
vanadium was determined to be present as discrete grains of
the mineral vanadinite Pb5(V5þO4)3Cl that were formed “up
stream” from lead pipe present in the examined systems; Gerke
et al. (2010) concluded that the vanadinite solid phase was
mostly embedded in near-surface regions of iron corrosion by-
products and, in the absence of physical dislodging or colloidal
mobilization, it was not likely to increase V concentrations in
the ambient water to dangerous levels. Our vanadium specia-
tion data are in agreement with the above observations as they
confirm that, the vanadium in the tested samples was primarily
associated with the mobilization-resistant and crystalline iron
oxide-bound fractions and the concentration of vanadium
released from iron-based corrosion solids were very low.
Similar observation was made in this study for chromium but,
in contrast with vanadium, the accumulation and release of Cr
have not been sufficiently addressed in prior research.
In contrast with As, Cr and V, significant fractions of
uranium and cadmium (ca. 16% and 13%, respectively) were
associated with carbonate-type solids. Prior research (e.g.,
Rihs et al., 2004; Khaokaew et al., 2011) shows that calcite and
other solid carbonate minerals can strongly sorb uranium and
cadmium. Such minerals are common in corrosion scales and
especially in hydrant flush samples where they can be domi-
nant solid phases (Peng et al., 2010). The retention of cadmium
by these phases appears to be at least partially reversible as
evidenced by the results that ca. 60% of the total cadmium was
released from the resuspended hydrant flush samples. While
actual levels of U and Cd in DWDS solids are very low and per
se not likely to be of concern, this finding suggests that
changes of drinking water alkalinity can be accompanied by
changes of levels of these contaminants.
Nickel and manganese were determined to be the most
mobile of all contaminants examined in our study. About 9% of
the retained Ni and Mn were found to exist as the exchange-
able fraction that is expected to be highly susceptible to vari-
ations of the ambient water quality. While estimation of
health effects potentially associated with the release of nickel
and manganese accumulated in DWDS solids goes beyond the
scope of this paper, this finding highlights the importance of
manganese removal from drinking water and control of its
status within drinking water distribution networks.
Notable contributions (more than 20%) of organic-bound
fractions of vanadium, uranium, nickel and manganese may
also indicate that biofilms and natural organic matter in the
DWDSs play an important role in the accumulation of these
inorganic contaminants. This is in agreement with prior
findings showing that extracellular polymer substances (EPS)
and biofilms per se can sorb inorganic contaminants (e.g.,
Flemming, 1995; Lalonde et al., 2007; Hitchcock et al., 2009). On
the other hand, this finding indicates that changes of water
chemistry causing destabilization of biofilms can be accom-
panied by release the sorbed contaminants. Further charac-
terization of natural organic matter and biofilms found in
DWDS solids is needed to provide more insight into the nature
of binding of the examined and other inorganic contaminants.
5562 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
4. Conclusions
Results of sequential extractions and measurements of soluble
and particulate metal fractions released from suspended
corrosion solids indicate that all examined trace-level heavy
metals (As, Cr, V, U, Cd, Ni, and Mn) exhibit specific features of
their fractionation and mobility. Arsenic, chromium and
vanadium are primarily associated with the mobilization-
resistant fraction that was unaffected by all eluents used in
this study. At the same time, low percentages of As, Cr and V
were associated with mobile fractions (exchangeable,
carbonate-bound, metaleorganic complex, and easily reduc-
ible metal oxide-bound). X-ray absorbance measurements
demonstrated that the arsenic in the sample with the highest
As concentration was dominated by As(V) bound by iron oxides.
Measurements of soluble and particulate metal concen-
trations demonstrated that only in the case of Ni and Mn
released from solids suspended at pH 7.6 and alkalinity
100 mg/L the majority of the released metal was in the dis-
solved (<0.4 mm) fraction while the other elements were
mostly associated particles with sizes between 0.4 and 5 mm.
However, in the case of arsenic, the concentrations of release
As were very low and almost all (98%) As released from the
hydrant flush solids was in the dissolved fraction. The data of
the resuspension and sequential extraction experiments were
in close agreement and showed that Ni and Mn are much
more mobile than all the other inorganic contaminants
examined in this study.
Acknowledgments
This study was supported by National Science Foundation
(Grant # 0931676) and partially by Water Research Foundation
(Project #3118) and the USEPA. The content and conclusions
are the views of the authors and do not necessarily reflect the
views of the funding agencies. The authors would also like to
thank Prof. Anatoly I. Frenkel and the Synchrotron Catalysis
Consortium for the support and facilitation of the EXAFS
experiments at Brookhaven National Laboratory.
Appendix. Supplementary data
Supplementary data associated with this article can be found
in the online version, at doi:10.1016/j.watres.2011.08.017.
references
Bednar, A.J., Medina, V.F., Ulmer-Scholle, D.S., Frey, B.A.,
Johnson, B.L., Brostoff, W.N., Larson, S.L., 2007. Effects of
organic matter on the distribution of uranium in soil and plant
matrices. Chemosphere 70 (2), 237e247.
Benjamin, M.M., Sontheimer, H., Leroy, P., 1996. Corrosion of iron
and steel. In: Internal Corrosion of Water Distribution
Systems. Cooperative Research Report. AWWA Research
Foundation, Denver, CO, pp. 29e70.
Cancès, B., Juillot, F., Morin, G., Laperche, V., Polya, D.,
Vaughan, D.J., Hazemann, J.-L., Proux, O., Brown Jr., G.E.,
Calas, G., 2008. Changes in arsenic speciation through
a contaminated soil profile: a XAS based study. Science of the
Total Environment 397 (1e3), 178e189.
Cancès, B., Juillot, F., Morin, G., Laperche, V., Alvarez, L., Proux, O.,
Hazemann, J.-L., Brown Jr., G.E., Calas, G., 2005. XAS evidence
of As(V) association with iron oxyhydroxides in
a contaminated soil at a former arsenical pesticide processing
plant. Environmental Science & Technology 39 (24),
9398e9405.
Copeland, R.C., Lytle, D.A., Dionysiou, D.D., 2007. Desorption of
arsenic from drinking water distribution system solids.
Environmental Monitoring and Assessment 127 (3), 523e535.
Dixit, S., Hering, J.G., 2003. Comparison of arsenic(V) and
arsenic(III) sorption onto iron oxide minerals: implications for
arsenic mobility. Environmental Science & Technology 37 (18),
4182e4189.
Fendorf, S., Eick, M.J., Grossl, P., Sparks, D.L., 1997. Arsenate and
chromate retention on goethite: I. Surface structure.
Environmental Science & Technology 31 (2), 315e320.
Fisher, E.L., Fuortes, L.J., Valentine, R.L., Mehrhoff, M., Field, R.W.,
2000. Dissolution of 226Radium from pipe-scale deposits in
a public water supply. Environment International 26 (1),
69e73.
Flemming, H.C., 1995. Sorption sites in biofilms. Water Science &
Technology 32 (8), 27e33.
Friedman, M.J., Hill, A.S., Reiber, S.H., Valentine, R.L., Larsen, G.,
Young, A., Korshin, G.V., Peng, C.-Y., 2010. Assessment of
Inorganics Accumulation in Drinking Water Systems Scales
and Sediments. Water Research Foundation, Denver, CO.
Gerke, T.L., Scheckel, K.G., Schock, M.R., 2009. Identification and
distribution of vanadinite (Pb5(V5þO4)3Cl) in lead pipe
corrosion by-products. Environmental Science & Technology
43 (12), 4412e4418.
Gerke, T.L., Scheckel, K.G., Maynard, J.B., 2010. Speciation and
distribution of vanadium in drinking water iron pipe corrosion
by-products. Science of the Total Environment 408 (23),
5845e5853.
Guidelines for Drinking-Water Quality, third ed., 2008 World
Health Organization, Geneva.
Green-Pedersen, H., Jensesn, B.T., Pind, N., 1997. Nickel
adsorption on MnO2, Fe(OH)3, montmorillonite, humic acid
and calcite: a comparative study. Environmental Technology
18 (8), 807e815.
Hill, A.S., Friedman, M.J., Reiber, S.H., Korshin, G.V., Valentine, R.
L., 2010. Behavior of trace inorganic contaminants in drinking
water distribution systems. Journal of the American Water
Works Association 102 (7), 107e118.
Hitchcock, A.P., Dynes, J.J., Lawrence, J.R., Obst, M., Swerhone, G.
D.W., Korber, D.R., Leppard, G.G., 2009. Soft X-ray
spectromicroscopy of nickel sorption in a natural river biofilm.
Geobiology 7 (4), 432e453.
Husband, P.S., Boxall, J.B., 2011. Asset deterioration and
discolouration in water distribution systems. Water Research
45 (1), 113e124.
Jain, A., Loeppert, R.H., 2000. Effect of competing anions on the
adsorption of arsenate and arsenite by ferrihydrite. Journal of
Environmental Quality 29 (5), 1422e1430.
Khaokaew, S., Chaney, R.L., Landrot, G., Ginder-Vogel, M.,
Sparks, D.L., 2011. Speciation and release kinetics of cadmium
in an alkaline paddy soil under various flooding periods and
draining conditions. Environmental Science & Technology 45
(10), 4249e4255.
Kim, E.J., Herrera, J.E., 2010. Characteristics of lead corrosion scales
formed during drinking water distribution and their potential
influence on the release of lead and other contaminants.
Environmental Science & Technology 44 (16), 6054e6061.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 5 3 e5 5 6 3
Krishnamurti, G.S.R., Huang, P.M., Van Rees, K.C.J., Kozak, L.M.,
Rostad, H.P.W., 1995. Speciation of particulate-bound
cadmium of soils and its bioavailability. Analyst 120 (3),
659e665.
Krishnamurti, G.S.R., Naidu, R., 2002. Solidesolution speciation
and phytoavailability of copper and zinc in soils.
Environmental Science & Technology 36 (12), 2645e2651.
Lalonde, S.V., Amskold, L.A., Warren, L.A., Konhauser, K.O., 2007.
Surface chemical reactivity and metal adsorptive properties of
natural cyanobacterial mats from an alkaline hydrothermal
spring, Yellowstone National Park. Chemical Geology 243 (1),
36e52.
Linge, K.L., 2008. Methods for investigating trace element binding
in sediments. Critical Reviews in Environmental Science and
Technology 38 (3), 165e196.
Lytle, D.A., Sorg, T.J., Frietch, C., 2004. Accumulation of arsenic in
drinking water distribution systems. Environmental Science &
Technology 38 (20), 5365e5372.
Lytle, D.A., Sorg, T.J., Muhlen, C., Wang, L., 2010. Particulate
arsenic release in a drinking water distribution system.
Journal of the American Water Works Association 102 (3),
87e98.
Manning, B.A., Hunt, M.L., Amrhein, C., Yarmoff, J.A., 2002.
Arsenic(III) and arsenic(V) reactions with zerovalent iron
corrosion products. Environmental Science & Technology 36
(24), 5455e5461.
Meng, X., Korfiatis, G.P., Bang, S., Bang, K.W., 2002. Combined
effects of anions on arsenic removal by iron hydroxides.
Toxicology Letters 133 (1), 103e111.
Naeem, A., Westerhoff, P., Mustafa, S., 2007. Vanadium removal
by metal (hydr)oxide adsorbents. Water Research 41 (7),
1596e1602.
Peng, C.-Y., Hill, A.S., Friedman, M.J., Valentine, R.L., Larson, G.S.,
Young, A.M., Reiber, S.H., Korshin, G.V. Occurrence of trace
inorganic contaminants in drinking water distribution
systems. Journal of the American Water Works Association,
submitted for publication.
Peng, C.-Y., Korshin, G.V., Valentine, R.L., Hill, A.S., Friedman, M.J.,
Reiber, S.H., 2010. Characterization of elemental and structural
composition of corrosion scales and deposits formed in
drinking water distribution systems. Water Research 44 (15),
4570e4580.
Reiber, S., Dostal, G., 2000. Well water disinfection sparks
surprises. Opflow 26 (3), 1e6.
Rihs, S., Sturchio, N.C., Orlandini, K., Cheng, L., Teng, H., Fenter, P.,
Bedzyk, M.J., 2004. Interaction of uranyl with calcite in the
presence of EDTA. Environmental Science & Technology 38 (19),
5078e5086.
Sarin, P., Snoeyink, V.L., Bebee, J., Kriven, W.M., Clement, J.A.,
2001. Physico-chemical characteristics of corrosion scales in
old iron pipes. Water Research 35 (12), 2961e2969.
Schock, M.R., 2005. Distribution systems as reservoirs and
reactors for inorganic contaminants. In: Distribution System
5563
Water Quality Challenges in the 21st Century. AWWA Reserch
Foundation, Denver, CO, pp. 105e140.
Schock, M.R., Hyland, R., Welch, M., 2008. Occurrence of
contaminant accumulation in lead pipe scales from domestic
drinking water distribution systems. Environmental Science &
Technology 42 (12), 4285e4291.
Sherman, D.M., Randall, S.R., 2003. Surface complexation of
arsenic(V) to iron(III) (hydr)oxides: structural mechanism from
ab initio molecular geometries and EXAFS spectroscopy.
Geochimica et Cosmochimica Acta 67 (22), 4223e4230.
Shuman, L.M., 1985. Fractionation method for soil
microelements. Soil Science 140 (1), 11e22.
Smith, E., Ghiassi, K., 2006. Chromate removal by an iron sorbent:
mechanism and modeling. Water Environment Research 78
(1), 84e93.
Stewart, B.D., Mayes, M.A., Fendorf, S., 2010. Impact of uranyl-
calcium-carbonato complexes on uranium(VI) adsorption to
synthetic and nature sediments. Environmental Science &
Technology 44 (3), 928e934.
Tessier, A., Campbell, P.G.C., Bisson, M., 1979. Sequential
extraction procedure for the speciation of particulate trace
metals. Analytical Chemistry 51 (7), 844e851.
Trived, P., Axe, L., 2001. Predicting divalent metal sorption to
hydrous Al, Fe, and Mn oxides. Environmental Science &
Technology 35 (9), 1779e1784.
Trived, P., Axe, L., Tyson, T.A., 2001. XAS studies of Ni and Zn
sorbed to hydrous manganese oxide. Environmental Science &
Technology 35 (22), 4515e4521.
U.S. Environmental Protection Agency, 2009a. National Primary
Drinking Water Standards and National Secondary Drinking
Water Standards. EPA 816-F-09-004. U.S. EPA, Office of Water,
Washington DC.
U.S. Environmental Protection Agency, 2009b. Final Third
Contaminant Candidate List 3. EPA 815-F-09-001. U.S. EPA,
Office of Water, Washington DC.
Valentine, R.L., Stearns, S.W., 1994. Radon release from water
distribution system deposits. Environmental Science &
Technology 28 (3), 534e537.
Vreeburg, J.H.G., Blokker, E.J.M., Horst, P., Van Dijk, J.C., 2009.
Velocity-based self-cleaning residential drinking water
distribution systems. Water Science and Technology: Water
Supply 9 (6), 635e641.
Waychunas, G.A., Rea, B.A., Fuller, C.C., Davis, J.A., 1993. Surface
chemistry of ferrihydrite: part 1. EXAFS studies of the
geometry of coprecipitated and adsorbed arsenate.
Geochimica et Cosmochimica Acta 57 (10), 2251e2269.
Wilkie, J., Hering, J.G., 1996. Adsorption of arsenic onto hydrous
ferric oxide: effects of adsorbate/adsorbent ratios and co-
occurring solutes. Colloids and Surfaces A 107, 97e110.
Zhou, P., Gu, B., 2005. Extraction of oxidized and reduced forms of
uranium from contaminated soils: effects of carbonate
concentration and pH. Environmental Science & Technology
39 (12), 4435e4440.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

QMRAspot: A tool for Quantitative Microbial Risk Assessment

from surface water to potable water

Jack F. Schijven a,*, Peter F.M. Teunis b, Saskia A. Rutjes c, Martijn Bouwknegt c,
Ana Maria de Roda Husman c
a
National Institute for Public Health and the Environment, Expert Centre for Methodology and Information Services, PO Box 1, 3720 BA,
Bilthoven, The Netherlands
b
National Institute for Public Health and the Environment, Epidemiology and Surveillance, The Netherlands
c
National Institute for Public Health and the Environment, Laboratory for Zoonoses and Environmental Microbiology, The Netherlands

article info abstract

Article history: In the Netherlands, a health based target for microbially safe drinking water is set at less
Received 9 March 2011 than one infection per 10,000 persons per year. For the assessment of the microbial safety
Received in revised form of drinking water, Dutch drinking water suppliers must conduct a Quantitative Microbial
1 August 2011 Risk Assessment (QMRA) at least every three years for the so-called index pathogens
Accepted 12 August 2011 enterovirus, Campylobacter, Cryptosporidium and Giardia. In order to collect raw data in the
Available online 23 August 2011 proper format and to automate the process of QMRA, an interactive user-friendly
computational tool, QMRAspot, was developed to analyze and conduct QMRA for
Keywords: drinking water produced from surface water. This paper gives a description of the raw data
Quantitative Microbial Risk requirements for QMRA as well as a functional description of the tool. No extensive prior
Assessment knowledge about QMRA modeling is required by the user, because QMRAspot provides
Tool guidance to the user on the quantity, type and format of raw data and performs a complete
Drinking water analysis of the raw data to yield a risk outcome for drinking water consumption that can be
Index pathogen compared with other production locations, a legislative standard or an acceptable health
based target. The uniform approach promotes proper collection and usage of raw data and,
warrants quality of the risk assessment as well as enhances efficiency, i.e., less time is
required. QMRAspot may facilitate QMRA for drinking water suppliers worldwide. The tool
aids policy makers and other involved parties in formulating mitigation strategies, and
prioritization and evaluation of effective preventive measures as integral part of water
safety plans.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction contamination depending on water, treatment and pathogen

1.1. General
Drinking water may be derived from surface water that is
often contaminated with human pathogens and that may
cause disease upon ingestion (Leclerc et al., 2002). Applied
treatment processes may efficiently remove such microbial
characteristics. Quantitative data on pathogen concentrations
in the source water and on the efficiency of the treatment
processes are required to assess the risks from drinking water
consumption (ILSI, 1996; WHO, 2011). The World Health
Organization (WHO) Guidelines for Drinking water Quality
(WHO, 2011) outline a preventive management framework for
safe drinking water entailing health based targets, system

* Corresponding author. Tel.: þ31 30 274 2994; fax: þ31 30 274 4434.
E-mail address: Jack.schijven@rivm.nl (J.F. Schijven).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.024
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
assessment from source through treatment to the point of
consumption, operational monitoring of the control measures
in the drinking water production, management plans doc-
umenting the system assessment and monitoring plans and
a system of independent surveillance that verifies that the
above are operating properly.
1.2. Dutch Drinking Water Act
In line with the WHO Guidelines, the Dutch Drinking Water
Act of 2001 (Anonymous, 2001) requires risk assessment for
waterborne pathogens to demonstrate microbially safe
drinking water. The infection risk is set at complying with
a provisional health based target of less than one infection per
10,000 individuals per year. If the assessed infection risk
exceeds this target value, the drinking water company must
consult with the national environmental inspector about
necessary measures.
1.3. Dutch Inspectorate Guideline 5318
In the Dutch Drinking Water Act, no specific directives were
given on how to perform risk assessment. Therefore, the
Inspectorate Guideline 5318 for the assessment of the microbial
safety of drinking water (Anonymous, 2005) was drafted in close
consultation between the government (Environmental Inspec-
torate), the National Institute of Public Health and the Environ-
ment (RIVM), Bilthoven, the Netherlands and the drinking water
industry. Inspectorate Guideline 5318 specifies the demand for
safe drinking water by ways of Quantitative Microbial Risk
Assessment (QMRA) primarily using production location-
specific data and (inter)national knowledge mainly on infec-
tivity of pathogenic microorganisms and treatment efficiency.
Risk assessment for exposure to pathogenic microorgan-
isms in drinking water was described by Teunis et al. (1997),
Haas et al. (1999), Haas and Eisenberg (2001), the ILSI frame-
work (Benford, 2001) and Medema et al. (2003). These publi-
cations and development of risk based approaches by the
WHO (2011) served as the basis for the specification of QMRA
in Inspectorate Guideline 5318. It is emphasized that risk
assessment is an iterative process which is directed by prac-
tical and theoretical progress.
1.4. Index pathogens
The Dutch Drinking Water Act states that microorganisms
should not be present in drinking water at such concentra-
tions that they form a threat to public health (Anonymous,
2001). This demand essentially concerns all waterborne
pathogens. Because it is not feasible to monitor each water-
borne pathogen, four so-called index pathogens were selected
to represent waterborne pathogenic viruses, bacteria and
parasitic protozoa (Anonymous, 2005). The selection criteria
were prevalence, disease outcome, and possibilities for
prevention and/or treatment. Obviously, waterborne trans-
mission should be a significant pathway for these pathogens.
In addition, a detection method, possible pathogen sources
and efficiency of water treatment processes have to be
established. This way, enterovirus, Campylobacter, Cryptospo-
ridium and Giardia were selected as the index pathogens. It is
5565
believed that if a drinking treatment is effective in removing
these index pathogens, adequate safety is warranted against
other waterborne pathogens.
1.5. Surface water monitoring
Monitoring of the surface water should be aimed at achieving
a representative quantification of the numbers of pathogenic
microorganisms in the source water, considering seasonal
variability as well as short term fluctuations of pathogen
concentrations (Westrell et al., 2006b).
According to Inspectorate Guideline 5318, a QMRA needs to
be repeated at least every three years and three years of
monitoring for pathogens may be condensed into one year to
allow a higher monitoring frequency and hence provide more
information about variability in pathogen concentrations.
Inspectorate Guideline 5318 defined the monitoring frequency
for source waters as dependent on the drinking water
production volume (Table 1). In addition to regular moni-
toring, a number of incidental samples must be collected and
analyzed at so-called peak moments, when peak concentra-
tions in pathogen counts are assumed to occur, for example
due to heavy rainfall (Kistemann et al., 2002; Kay et al., 2007).
1.6. Treatment efficiency
A treatment or production site should be designed, organized
and operated such that under any circumstance safe drinking
water is produced. In that regard, the so-called multiple
barrier principle applies: at times or instances when a treat-
ment stage performs less well the other treatment stages
should still provide adequate removal in order to produce safe
drinking water (WHO, 2011).
To quantify the efficiency of the treatment, data should be
collected for each drinking water production location,
because, commonly treatment efficiency is highly location-
specific. Any changes in the treatment process require new
collection of data.
Monitoring programs for determining treatment efficiency
should also account for any temporal variability (Kistemann
et al., 2002; Westrell et al., 2006b).
1.7. Indicator organisms
Commonly, concentrations of index pathogens will decrease
below detection limits by drinking water treatment. However,
so-called indicator organisms that are assumed to have
similar properties as the index pathogens, so that they are
Table 1 e Number of samples per period of three years for
each index pathogen according to the drinking water
production of a drinking water treatment location
(Anonymous, 2005).
Production Regular Incidental Total
(m3/day)
<10,000 6 (one every eight weeks) 3 9
10,000e100,000 13 (one every four weeks) 6 19
>100,000 26 (one every two weeks) 9 35
5566 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
removed equally or less well by drinking water treatment, are
used to characterize treatment. Appropriate indicator organ-
isms not only resemble the corresponding index pathogens in
their fate and behavior in water, but also occur in higher
numbers and are easier to enumerate with higher recovery.
Nevertheless, after each treatment stage, indicator numbers
decrease and in many cases drop below detection limits as
found for index pathogens.
Inspectorate Guideline 5318 (Anonymous, 2005) prescribes
by default F-specific or somatic bacteriophages as the indi-
cator organisms for determining removal efficiency by
drinking water treatment of enterovirus. Similarly, Escherichia
coli is set as the indicator organism for Campylobacter and
spores of sulphite reducing clostridia (SSRC) for both Crypto-
sporidium and Giardia.
1.8. Data analysis and new QMRAs
The iterative process of risk assessment in the Netherlands
started with the enforcement of Inspectorate Guideline 5318 in
2006 at the five Dutch drinking water companies that treat
surface water for drinking water production at fourteen loca-
tions. For preparation of this first QMRA, they not only collected
all available microbiological data on their source waters and on
removal efficiency of their treatment, but also provided
detailed descriptions on the dimensions and characteristics of
all treatment processes. All this information was documented
in reports that were sent to the Environmental Inspectorate,
who forwarded these reports to the National Institute of Public
Health and the Environment, Bilthoven, the Netherlands
(RIVM). Using the reported data, the actual QMRA was con-
ducted at RIVM. RIVM reported the outcomes of the QMRA to
the Environmental Inspectorate encompassing conclusions on
compliance with the allowed maximum infection risk as well
as on the completeness of the data. Outcomes were also dis-
cussed in meetings with each drinking water company, the
environmental inspector and RIVM. The outcomes of this study
are summarized by Schijven and de Roda Husman (2009).
Extracting all historic data from the first QMRA reports
assembled by the Dutch drinking water companies as well as
evaluating those reports appeared to be very laborious
(Schijven and de Roda Husman, 2009), which led to the desire
to process and analyze new data in a more efficient and
standardized manner with a risk outcome in relation to the
legislative health based target. To that aim, a spreadsheet was
designed to be used by the drinking water companies for
entering the data from the next round of monitoring and
a computational tool was designed to automatically analyze
the data and conduct the QMRA.
1.9. Objectives
This paper presents and describes an interactive user-friendly
computational tool, named QMRAspot, that was developed to
analyze and conduct QMRA for a drinking water production
chain from surface water to potable water. The objectives of
developing this tool are the following:
- Collection and automated reading of raw microbial data,
entailing index pathogen data from source water and
indicator organism data for determining drinking water
treatment efficiency, all provided in a standard format;
- Automatic analysis of those data, entailing fitting of
appropriate probability distributions to quantify concen-
trations and treatment efficiencies;
- A uniform and transparent way of conducting QMRA;
- A risk outcome in relation to the legislative health based
target.
By describing the tool, this paper is intended to also serve
as a user and reference manual of how to conduct QMRA for
drinking water.
2. Data
2.1. Raw data
For a QMRA, it is essential to collect quantitative microbial
data as raw unprocessed data. Raw data on enumerated
microorganisms in water are the counted numbers of the
microorganisms as well as the corresponding investigated
volume of the sample. Commonly, counts are numbers of
plaque-forming units (pfu) for viruses and colony-forming
units (cfu) for bacteria (Schets et al., 2008; Teunis et al.,
2005a). Oocysts of Cryptosporidium and cysts of Giardia may be
counted manually or automatically under a microscope using
fluorescent dye (Schets et al., 2008). Raw presence/absence
data are presence or absence of micoorganisms in replicate
dilutions of a water sample (De Roda Husman et al., 2009).
Obviously, the concentration of, for example, 1 pathogen
particle in 1 mL of water is the same as a 100 particles in 100 mL
of water, but if 100 particles were observed that count is more
accurate, hence it is essential to use counts and sample
volumes as they were observed and not concentrations (ratios
of count/volume). All raw data must include a sample date.
2.2. Source water data
Source water data are raw data of index pathogens in the
source water (e.g. Rutjes et al., 2009; Lodder et al., 2010). At many
production locations, river water first passes a storage reservoir
before further treatment. For enterovirus, river water may
appropriately be designated as source water. However, for
Campylobacter, Cryptosporidium and Giardia, the storage reservoir
should be considered the starting point of the QMRA, because of
contamination of the storage reservoir water with Campylobacter
from birds, wildlife, or runoff from agricultural land.
2.3. Recovery data
In order to determine the recovery efficiencies of the detection
method for index pathogens, ideally, each sample of source
water, or a fraction used for analysis, is spiked with a suffi-
ciently high number of, for example, a specific type of indi-
cator organism. The spiked and recovered numbers can then
be used to estimate the recovery efficiency. Therefore, raw
recovery data consist of counts and samples sizes of the
spiked and recovered microorganisms that are paired
according to sampling date (Teunis et al., 1999; Rutjes and de
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
Roda Husman, 2004; Schets et al., 2004). If the source water
sample contains sufficiently high numbers of indicator
organisms, like bacteriophages, to allow direct counting after
plating, these indicator organisms can be used to estimate the
recovery efficiency of a concentrating step needed for detec-
tion of the associated pathogen.
2.4. Treatment data
Raw data of indicator organisms for a treatment process consist
of counts and sample sizes of samples taken before and after
treatment. If available, raw data of index pathogens can be used.
2.5. Consumption data
QMRAspot offers four choices for consumption of unboiled
drinking water per person per day. These are all lognormal
distributions, with parameters defined by various studies.
Parameters m ¼ 1085779 and s ¼ 1.07487 are for the
Netherlands, corresponding to a mean of 0.27 L per person per
day (Teunis et al., 1997), a lognormal distribution with
parameters m ¼ 0.03598 and s ¼ 0.77218 for the USA, corre-
sponding to a mean of 1.3 L per person per day (USEPA, 2006),
a fixed volume of 2 L per person per day (WHO, 2011) and,
finally, the possibility of putting in the parameter values for m
and s for any other log normally distributed consumption
data, if available for another country or for a specific
subpopulation. Consumption data may differ between coun-
tries and also between subpopulations; climate may also play
a role. For more data and a discussion about such variability,
the reader is referred to USEPA (2006), Westrell et al. (2006a)
and WHO (2011).
2.6. Dose response data
The dose response relation for rotavirus has been published
(Teunis et al., 1996), as well as for Giardia and Cryptosporidium
(Teunis et al., 1996, 2002a, b). The dose response relation of
rotavirus, an enteric virus, is applied as a worst case for virus
infectivity (Regli et al., 1991). The hierarchical dose response
relation for three isolates of Cryptosporidium parvum has been
updated to include two additional isolates for which challenge
studies have been published (Okhuysen et al., 2002). For
Campylobacter, the human challenge dose response study
updated with outbreak data has been used (Teunis et al.,
2005b). For any index pathogen the same dose response data
are used for each QMRA simulation.
3. Standard data file: QMRAdata.xls
The tool reads the raw data from a standard Microsoft Excel
spreadsheet file, here, for convenience, named QMRAdata.xls.
0 0  11

 
Y
n B B 
1 
L r; ¼ 2log B B
@f ðrÞNegBin@ni ; Vi ExpðlnrÞ;
1 þ lVi
i¼1  1þ
5567
It contains three sheets: SCHEME, RAW DATA and HELP. The
SCHEME sheet provides a description of the drinking water
production location and defines a table with column headers
that is used by the tool to make the appropriate data selec-
tions. Through the SCHEME sheet, the user has control over
what data should be used for QMRA. Obviously, the RAW
DATA sheet contains all raw data and the HELP sheet provides
background information on how to fill the RAW DATA sheet
with raw data in the required format.
Table 2 shows a summarized setup of the SCHEME sheet. In
the SCHEME sheet, the effluent data of a treatment step may be
the influent data for the next treatment step. This need not be so
if the data for the next treatment step concern other types of
microorganisms, or were from a different location in the drinking
water utility, or from pilot plant experiments. The SCHEME sheet
can be modified easily. For example, two treatment steps may be
combined using the influent data of the first of the two and the
effluent data of the second of the two treatment steps.
In the RAW DATA sheet, every row is a full record of raw
data (Table 3). This simple design allows for automated filling
from a Laboratory Information Management System (LIMS),
for example, as records in the form of Comma Separated
Values (CSV).
The source of the data for the treatment steps can be
selected from the following list: Plant scale, pilot plant scale,
laboratory scale. According to Inspectorate Guideline 5318
(Anonymous, 2005), location-specific plant scale data are
generally preferred, followed by pilot plant scale data, and if
these are not available, data from laboratory experiments. In
other words, location-specific data are recommended. If the
use of data from other locations is desired, applicability
should be verified by comparison of treatment conditions.
References to data from literature should be listed in the
accompanying QMRA reports.
4. Fitting of distributions to the data
4.1. General
All raw data sets should include three or more samples:
smaller data sets are ignored and parameters are not esti-
mated. Counts in QMRAdata.xls may only be integers.
4.2. Source water concentration
Based on the assumption that counts n within each sample of size
V are Poisson distributed, while the concentration (Poisson
parameter) is gamma-distributed among samples, the counts
n1 .nN of N samples with samples sizes V1 .VN have a Negative
Binomial distribution (NegBin) with parameters r and 1=1 þ lVi
(Teunis et al., 2009). Parameters r and l are estimated by mini-
mizing the following deviance function:
CC
1
CC
AA (1)
V ExpðlnlÞ
i
s
5568 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6

Table 2 e Summary of the SCHEME sheet in QMRAdata.xls.

Row\column A B C D E F

1 Drinking water Drinking water

utility utility name
2 Production Production
location location name
3, 13, 23, 33 Index Enterovirus,
pathogen Campylobacter,
Cryptosporidium, Giardia
4, 14, 24, 34 Source water Source water name Spike Recovery
for each index pathogen
5, 15, 25, 35 Recovery Recovery indicator name: Spike name Recovery
indicator e F-specific RNA name
bacteriophage,
somatic coliphage
or enterovirus
e E. coli or Campylobacter
e SSRC or Cryptosporidium
e SSRC or Giardia
6, 16, 26, 36 Treatment Treatment indicator Influent Effluent Data source
7e12 z1 .z6 Treatment Treatment indicator name: Influent Effluent For each
17e22 for each step names e F-specific RNA name name treatment step:
27e32 of the bacteriophage, of each of each e Plant scale
37e42 four index somatic coliphage treatment treatment e Pilot plant scale
pathogens or enterovirus step step e Laboratory scale
e E. coli or Campylobacter
e SSRC or Cryptosporidium
e SSRC or Giardia

where f ðrÞ ¼ 1  Fððx  20Þ=20Þ is a prior function for the shape Campylobacter. These observations are used to calculate
factor r, with Fððx  20Þ=20Þ the cumulative normal distribution a concentration for each sample by minimizing the following
with a mean and variance of 20. This prior prevents extremely deviance function:
small values of r and facilitates robust parameter estimation
n

Y 
without strongly affecting the estimates. s is a scaling factor. If ni ¼ 0 0 Poisð0jcVi Þ
Lðc; VÞ ¼ 2log (2)
ni > 0 0 1  Poisð0jcVi Þ
the mean sample volume is less than 1 L, then s ¼ 0.001. This i¼1

scaling avoids computational underflows.

In QMRAdata.xls presence/absence data may be given for where Pois denotes Poisson distribution, c is concentration
any microorganism, although this is usually only the case for and Vi is the sample size of the i-th sample.

Table 3 e The RAW DATA sheet in QMRAdata.xls.

Column Name Description

A Name Name of source water, "Spike", "Recovery", treatment influent, treatment effluent.
B Sampling code Specific code included in RAW DATA for reference. Different codes may be included
for different sampling points of the same influent or effluent. In the QMRA the
data designated by different codes are combined.
C Microorganism Index pathogens and indicator microorganisms.
D Date Date format: DD-MM-YY.
E Count Only whole positive numbers (integers) allowed. Maximum counts per plate according
to the standard method.
F Sample volume The actual sample volume (liter) used for counting of the microorganisms.
Example 1: a 10 L sample was collected, concentrated to 100 mL, 5 ml was plated
for counting, then the sample size is 10/20 ¼ 0.5 L.
Example 2: 98 colonies were counted on a plate with 1 ml sample and 11 colonies
were counted in the ten-fold dilution. Count is 109 and sample size is 0.0011 L
G/J/M/P/S V1/V2/V3/V4/V5 V1.5 are the sample volumes (liter) of five dilution steps in the MPN scheme.
H/K/N/Q/T R1/R2/R3/R4/R5 R1.5 are the replicate numbers for each dilution step in the MPN scheme. Only
whole numbers allowed.
I/L/O/R/U MPN1/MPN2/MPN3/ MPN1.5 are the Most Probable Numbers in the MPN scheme. Only whole numbers
MPN4/MPN5 from 0 - the replicate number allowed. The tool does not use available MPN data
if counts (column E) and sample sizes (column F) are available.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
Subsequently, a Gamma distribution with parameters r
and l is fitted to the concentration data by minimizing the
following deviance function:
0 
ci ¼ N 0 Gammacdf Lðc;
 VÞ ¼ c95%2 ðdf ¼ 1Þ lnr;
2
Y
n
Lðr; lÞ ¼ 2log @ ci ¼ 0 0 1  Gammacdf Lðc; VÞ ¼ c95% ðdf ¼ 1Þlnr; ln l A
i¼1 0 < ci < N 0 Gammaðlnr; ln lÞ
where ci is the concentration of the i-th sample. Gammacdf
denotes a cumulative Gamma distribution. Lðc; VÞ ¼ c295% ðdf ¼ 1Þ
is the root of the likelihood function equal to the 95-percentile of
a X -squared distribution with one degree of freedom (df).
If raw data of the index pathogen in the source water
consist of nondetects only, a Gamma distribution with
P
parameters r ¼ 0.01 and l ¼ 1=ð Vi þ 1=100Þ is assumed.
4.3. Recovery, R
In order to estimate the recovery efficiency of the detection
method of the index pathogen, samples are spiked with
a known number of the specified indicator organisms. After
processing of the samples, a fraction of the spiked organisms
will be recovered. The data on the initial spike and on the
recovery are paired. The recovered fraction is assumed Beta-
distributed with parameters a and b. Estimation of these
parameters by means of the paired Beta model is explained in
detail by Teunis et al. (1999, 2009). If recovery data are lacking,
then it is assumed that R ¼ 1.
4.4. Treatment, z
Here, we assume that treatment is in effect, implying that
microorganisms are removed and thus 0  z  1. It is assumed
that microorganisms passing treatment do so independently
with a probability or fraction z. This may be modeled as
a binomial process, either with paired or unpaired samples
(Teunis et al., 1999, 2009).
Collection of paired data from a treatment step requires
exact timing of the sampling. The pairing may be lost if
mixing occurs during treatment. Residence times in treat-
ment may vary from a few hours to several days. In many
cases, even with short residence times and samples of
influent and effluent collected on the same day, pairing is not
evident.
Estimation of the parameters a and b by means of the
unpaired Beta model is explained in detail by Teunis et al. (1999,
2009). In case the effluent of a treatment stage produces only
nondetects, it is still possible to evaluate treatment.
5. Monte Carlo simulation
From all distributions, 10,000 Monte Carlo (MC) samples are
generated. The source water concentration of the index
pathogens, Csource, is Gamma-distributed with parameters r
and l/s. For recovery R, and treatment steps z1 .z6 , Beta-
distributed MC samples are generated.
5569
For the unboiled drinking water consumption, W liter, MC
samples of lognormal distributions are generated, but in case
of the WHO-data set, a fixed value of 2 L is used.
  1

 ln l 
(3)
Monte Carlo samples of the dose response parameters
a and b are provided as pregenerated data and included in the
tool in a packed form to save memory space.
6. QMRAspot tool code
QMRAspot has been developed in Mathematica 8.0.0
(Wolfram, Inc., Champaing IL, USA). It can be run in as
a Dynamic Module in Mathematica Player Pro versions 7 and 8.
When the calculations are complete, the user can page
through the results. A QMRA report can be generated for each
of the index pathogens and saved as a Mathematica notebook
and/or pdf file. Fig. 1 shows the splash screen of the tool.
7. Exposure and infection risk
Exposure to the index pathogens is given as the dose D, the
number of ingested index pathogens per person per day and is
calculated by multiplying the MC samples of source concen-
tration Csource, recovery R, treatment zi and consumption data W:
1Y 6
D ¼ Csource zi W (4)
R i¼1
Infection risk per person per day is calculated by applying
the dose response relation (Teunis and Havelaar, 2000):
Pinf;person;day ¼ 1  1 F1 ða; a þ b; DÞ (5)
where a and b are infectivity parameters that are pathogen
specific and 1 F1 is the confluent hypergeometric function.
Parameters a and b are MC sample pairs (joint distribution),
reflecting variability of infectivity.
Infection risk per person per year is calculated from MC
samples of daily infection risk by applying Eq. (5) 10,000 times
for each day in a year to obtain 365 MC sample distributions,
which are then multiplied with each other (Teunis et al., 1997):
Y
365
 
Pinf;person;year ¼ 1  1  Pinf;person;day;i (6)
i¼1
8. Results
As examples, a number of distributions will be shown of
source water concentrations, treatment and infection risks
based on data from a number of Dutch drinking water
suppliers (Figs. 2e4). These are the raw data for the second
round of QMRA, where data were delivered in the standard
5570 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6

Fig. 1 e Splash screen of QMRA spot. Only a QMRAdata.xls file with raw data needs to be opened. After pushing the “Run
QMRA”-button, a full analysis of the data will be done and a QMRA will be conducted.

format in QMRAdata.xls. The names of the drinking water
locations are kept anonymous, because it is not the intention
of this paper to present the QMRA results for specific drinking
water locations. Using fictive data could provide unrealistic
examples. The presented data are a combination of data from
different suppliers to form a fictive drinking water production
location.
Figs. 2e4 show the full QMRA for Cryptosporidium as
example. Fig. 5 gives the box-whisker plots for the infection
risk per person per year for all four index pathogens. Table 4
summarizes all QMRA steps for all four index pathogens by
giving mean and 95-percentile values of all distributions on
log10 scale, as well as the distribution parameter values.
8.1. Index pathogen concentrations in source water
QMRAspot provides time plots for inspecting variations of
concentrations over time (Fig. 2). The histograms illustrate
MC samples of the Gamma-distributed concentrations,
which appear to be left-skewed on a logarithmic scale. This
was also observed for enterovirus by Teunis et al. (2009). For
reference, mean concentration and 95-percentiles of these
distributions are included in the time plot. Peak concen-
trations may be defined as those concentrations above the
95-percentile. In winter, lower temperatures and increased
precipitation lead to shorter residence times between
wastewater discharge locations and intake points for
drinking water production, resulting in less inactivation
than during summer (Schijven et al., 1996; Schijven and de
Roda Husman, 2005). Note that for the index pathogens,
except Campylobacter, in the majority of the samples, no
index pathogens were detected, resulting in low mean
concentrations (Table 4).
8.2. Recovery
Fig. 2 includes recovery data for Cryptosporidium. Given these
observations, the need for including recovery data into QMRA
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6 5571

Fig. 2 e QMRA of Cryptosporidium. Time plot of measured source water concentrations and histograms of MC samplings of
fitted distributions to source water concentrations and recovery data, and histogram of calculated for recovery corrected
source water concentration.

may be clear. In the Netherlands some data on recovery of 8.3. Treatment

index pathogens are available (e.g. Rutjes and de Roda
Husman, 2004; Schets et al., 2004), but location-specific data
on recovery of index pathogens are usually lacking. Moreover,
recoveries may vary between samples, necessitating estima-
tion of recovery efficiency for every single sample. Such an
approach has been used for detection of Cryptosporidum
oocysts and Giardia cysts (Quintero-Betancourt et al., 2003;
Ferguson et al., 2004).
Fig. 3 shows the estimated removal of SSRC as indicator for
removal of Cryptosporidium (and Giardia). Time plots of the
influent and effluent concentrations are shown. The Beta
distributions of the z-values on logarithmic scale are generally
left-skewed. In this example, the effluent data of Z1 are the
influent data for Z2. For all Z1 and Z2 influent and effluent
samples, SSRC concentrations were above detection limit. For

Fig. 3 e QMRA of Cryptosporidium. Time plots of measured source influent and effluent concentrations of SSRC for treatment
steps z1, z2 and z3 and histograms of MC samplings of fitted distributions the fractions of SSRC that pass treatment.
5572 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6

Fig. 4 e QMRA of Cryptosporidium. Histograms of total treatment and drinking water concentration as calculated from the MC
samples of treatment steps z1, z2 and z3 and the source water concentration. Histogram of MC samples from the Dutch
drinking water consumption data. Histogram of exposure to Cryptosporidium as calculated from the drinking water
concentration and consumption MC samples.

Z3 a data set from another drinking water production location
was used, showing a large number of samples, of which in the
effluent, only a few samples had concentrations above the
detection limit.
8.4. Exposure
Fig. 4 shows the MC-histograms of SSRC removal from
multiplying the MC-data of all treatment steps, the MC-
histograms of the drinking water concentrations as calcu-
lated from the estimated source water concentrations,
recovery and treatment, the MC-histogram of the lognormal
distributed drinking water consumption, and, finally, the MC-
histogram of the exposure or dose as calculated from the
estimated drinking water concentrations and drinking water
consumption data.
8.5. Infection risk
Fig. 5 shows box-whisker plots of the annual infection risk for
each of the index pathogens, with arithmetic mean (horizontal
line), quartiles (box) and 95% interval (whiskers). If the mean
infection risk is above the target value of 104 per person per
year, as in this example for Campylobacter, the box-whisker plot
is red. If only the 95-percentile exceeds the target value, the box-
whisker plot is yellow, as is the case for Cryptosporidium. In all
other cases the box-whisker plot is green. This is to emphasize
that it is highly recommended that at least 95% of the time the
infection risk is below the health based target value.
9. Discussion
9.1. Raw data
QMRAspot imposes no restrictions on microbial count ranges.
Detection methods are restricted to a maximum count,
usually 100 microorganisms per plate. Higher counts suffer
large systematic errors because of nutrient exhaustion, over-
lapping colonies or plaques, or a higher probability that colo-
nies or plaques did not originate from a single bacterial cell or
virus particle (Teunis et al., 2005a). Counts per sample could

Fig. 5 e Box-whisker plots of the annual infection risk for each index pathogen, with mean (line), quartiles (box) and 95%
interval (whiskers).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6 5573

Table 4 e Summary of QMRA steps for all four index pathogens for a fictitious drinking water production location with three
subsequent treatments Z1, Z2 and Z3.
QMRA step Dimensions Log10 mean Log10 95% Distribution

Enterovirus
Source water concentration N/litre 3.6 3.0 Gamma(0.63; 0.00044)
Z1 Fraction 0.86 0.48 Beta(1.7; 10)
Z2 Fraction 2.3 1.6 Beta(0.086; 18)
Total treatment (somatic coliphage) Fraction 3.2 2.5
Drinking water concentration N/litre 6.8 6.2
Consumption Litre/person/day 0.55 0.051 Lognormal(1.9; 1.1)
Exposure N/person/day 7.3 6.9
Infection risk /person/day 7.7 7.2
Infection risk /person/year 5.1 4.9
Campylobacter
Source water concentration N/litre 2.1 2.8 Gamma(0.34; 380)
Z1 Fraction 0.64 0.26 Beta(1.2; 4.1)
Z2 Fraction 2.3 1.5 Beta(0.099; 21)
Z3 Fraction 4.8 1.1 Beta(0.068; 5.5)
Total treatment (E. coli) Fraction 1.9 5.0
Drinking water concentration N/litre 2.9 3.4
Consumption Litre/person/day 0.55 0.051 Lognormal(1.9; 1.1)
Exposure N/person/day 3.5 4.1
Infection risk /person/day 3.9 4.8
Infection risk /person/year 1.3 0.81
Cryptosporidium
Source water concentration N/litre 2.7 2.1 Gamma(0.033; 0.069)
Recovery Fraction 0.77 0.47 Beta(2.7; 13)
Corrected source water concentration N/litre 1.7 1.2
Z1 Fraction 0.46 0.16 Beta(1.9; 3.6)
Z2 Fraction 1.8 1.2 Beta(0.42; 28)
Z3 Fraction 1.5 0.69 Beta(0.076; 2.5)
Total treatment (SSRC) Fraction 3.8 3.2
Drinking water concentration N/litre 5.7 7
Consumption Litre/person/day 0.55 0.051
Exposure N/person/day 6.3 7.9
Infection risk /person/day 7 8.6
Infection risk /person/year 4.4 3.8
Giardia
Source water concentration N/litre 2.9 2.2 Gamma(0.21; 0.0058)
Total treatment (SSRC) Fraction 3.8 3.2
Drinking water concentration N/litre 6.7 6.7
Consumption Litre/person/day 0.55 0.051 Lognormal(1.9; 1.1)
Exposure N/person/day 7.4 7.4
Infection risk /person/day 9.1 9.1
Infection risk /person/year 6.5 6.1

nevertheless be high numbers if replicate plates were counted
and totaled. High counts with trailing zeroes, for example
2200, are doubtful and may be processed data, for example,
extrapolations of 22 in a 100 times diluted sample.
Often, low counts per plate are assumed to be unreliable
and discarded. Although accuracy of estimated concentra-
tions may be low, the method in QMRAspot for fitting
a distribution accounts for such uncertainty and discarding
information should be discouraged.
Ideally, volumes of source water should be large enough to
enable detection of index pathogens. In some cases, analysis
of a 10 L sample of source water is sufficient whereas other
locations may require a 100- to 2000 L sample.
pathogen concentrations in source water for QMRA should be
a continuous activity in order to evaluate possible trends, for
example, due to climate changes or local changes such as in
discharges of wastewater. Also, regular monitoring programs
are inherently limited and may often not capture peak events;
however, by continuating monitoring programs for index
pathogen concentrations in source water, gradually more
insight and knowledge on peak events will be collected
(Kistemann et al., 2002; Kay et al., 2007). This may lead to
sufficient data that enable adaptive dynamic filtering for early
warning of peak concentrations (Westrell et al., 2006b) and
timely and appropriate management actions may be under-
taken to prevent waterborne disease.

9.2. Index pathogen concentrations in source water 9.3. Recovery, infectivity and typing

In discussions with representatives of the drinking water Location-specific data on the recovery of detecting index
companies, it became apparent that collection of index pathogens in the source water are still sparse. Recoveries may
5574 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
vary widely, especially for the parasitic protozoa (Schets et al.,
2004), and, therefore, can have a strong impact on the risk
assessment outcome. It should also be noted that recovery
may not only be dependent on the detection method, the
pathogen and the type of water, but also on microbial counts
(Schmidt and Emelko, 2011). Schmidt and Emelko (2011)
demonstrated this with a model that describes variability in
microorganism counts as a function of sample volume and
the analytical recovery of the enumeration method and that
was expanded to include temporal concentration variability
and sample-specific recovery information.
Detection methods vary widely from counting of entero-
virus plaques in a cell culture assay, determining most prob-
able numbers of Campylobacter, to microscopic viability
counting of Cryptosporidium oocysts and Giardia cysts. The
performance characteristics of detection methods lead to
specific meanings of the risk estimates regarding the infec-
tivity of the counted microorganisms. This should be
accounted for when comparing risks between different
waterborne pathogens. More thought should go into the
meaning of risk estimates in the light of newly developed,
rapid molecular techniques such as PCR for significant
waterborne pathogens, such as, for example, noroviruses that
cannot be determined with the use of the abovementioned
detection methods (Laverick et al., 2004). QMRAspot should be
adapted to not only account for recovery but also for the
infectious fraction of counted index pathogens.
To correctly assess an infection risk for humans from
drinking water consumption, one should limit quantification
to human pathogens. Some pathogens may be zoonotic with
an animal origin but nevertheless infectious to humans. If
a substantial fraction of the pathogen is of animal origin, one
may want to determine the contribution of the animal sources
relative to human sources to the infection risk estimate. In
this regard, quantitative microbial source tracking comes into
play (Schijven and de Roda Husman, 2011).
9.4. Treatment
If a treatment step is well-characterized and the process
conditions remain the same, it is not necessary to collect new
data before and after the treatment step for each new QMRA.
Nevertheless, it is common practice for Dutch drinking water
companies to constantly monitor treatment performance;
hence it is obvious to include new data into a new QMRA.
Moreover, information on treatment failure may accumulate
over time. Obviously, if treatment process conditions have
changed, which may include changes in source water quality
or any of the other determinants of drinking water quality,
a QMRA should be conducted to evaluate any effects of such
changes. In that regard, specific treatment models that predict
treatment efficiency for a range of process conditions are very
useful and extensive monitoring to evaluate treatment effi-
ciency under the altered conditions may not be necessary.
Such a predictive model is in development for slow sand
filtration (Schijven et al., 2008), which is not only applicable to
altered process conditions, but also for other production
locations.
The default indicator organisms may not always be the
best choice for representing the removal of the index
pathogens. For example, parasitic protozoa are removed
much better by slow sand filtration than SSRC (Hijnen et al.,
2007). However, currently, for most treatment processes, the
abovementioned index organisms are used by default,
awaiting research that provides data for more appropriate
index organisms in particular cases.
Although QMRA based on location-specific raw data for
treatment at full scale is by far to be preferred, the tool
provides the option of including distribution parameters
values of fraction z of the microorganisms that were able to
pass treatment instead of raw data. This option is not included
to move away from collecting raw data, but often location-
specific data at plant scale are not available, because indi-
cator organism levels were (expected to be) below detection
limits. This often occurs with very efficient treatment steps
and/or at the end of the production chain. In those cases, one
has to rely on data from pilot plant experiments that mimic
full scale conditions, or on data from laboratory scale experi-
ments, or use treatment data that were collected at other
plants operating under similar conditions. In all those cases,
the applicability of the data needs to be verified.
The option of including a treatment step by means of its
distribution parameters can also be used to determine the
required additional treatment if a drinking water location
exceeds the health based targets. This option allows for
scenario studies, and therefore, greatly increases the versa-
tility of QMRAspot.
9.5. Consumption data
QMRAspot provides the option of defining distribution
parameter values for any set of consumption data. Also here,
over time, a database of consumption data for different
countries may emerge. It should be noted that in the current
version of QMRAspot, the choice for drinking water
consumption data encompasses the range of mean drinking
water consumption per person per day from about a quarter of
a liter to 2 L, which may cover all countries (Westrell et al.,
2006a).
9.6. Dose response relation
Applied dose response relations were generally derived from
studies in which a specific strain of the index pathogen was
given to human volunteers (Teunis et al., 1996, 2002a, 2002b).
However, one pathogen strain does not represent the suite of
strains that may occur in source waters for drinking water
production. A hierarchical dose response relation, as was
performed for multiple isolates of C. parvum, produced esti-
mates that differed very much between isolates (Teunis et al.,
2002a). Predictions based on multilevel dose response rela-
tions may aid probabilistic risk assessments such as pre-
sented here to properly reflect the variation in pathogen
strains. Moreover, dose response data from outbreaks may
inform the dose response relation as was shown for
Campylobacter (Teunis et al., 2005b). Such additions, both
hierarchical analysis and the use of outbreak data, could aid
the estimation of the enterovirus dose response relation for
which now the rotavirus dose response relation is used. For
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
such analysis data are currently not available, and additional
research is required.
9.7. Experience of Dutch drinking water companies
After the first round of QMRA using historic data, the Dutch
drinking water companies reported their experiences with
conducting QMRA at a workshop. There was a general
consensus about the benefits of QMRA (Schijven and de Roda
Husman, 2009): The QMRA framework provided integral
insight into the robustness of the drinking water treatment. It
enabled identification of weak links in the treatment and what
additional or improved treatment would be needed. It also
facilitated communication between management and opera-
tors, and, therefore, provided a basis for the appropriate
implementation of Water Safety Plans (Summerill et al., 2010).
9.8. Future developments
Because treatment efficiencies are described generically by
a Beta distribution, the launch of the QMRAspot tool intends to
initiate a database with treatment data in the form of the
parameters of the Beta distribution for any specific indicator
organism under any specific treatment process conditions.
Such an effort was started for enterovirus removal by Teunis
et al. (2009).
Currently, the QMRA ends with finished water, i.e. effects of
the distribution system are not included. Pathogen intrusion and
interaction of pathogens with biofilms in distribution systems,
for example, as described for viruses by Skraber et al. (2005), need
to be evaluated for their significance to infection risks.
QMRAspot is based on the Dutch Inspectorate Guideline
5318 (Anonymous, 2005) including the four index pathogens
enterovirus, Campylobacter, Cryptosporidium and Giardia. Given
the need to prioritize for emerging pathogens, this may
change in the future and require extending the tool with
QMRA of other pathogens.
10. Conclusions
An interactive user-friendly computational tool, QMRAspot,
has been developed for use without extensive prior knowledge
about QMRA modeling to estimate a risk outcome for
consumption of drinking water produced from surface water.
This risk outcome for drinking water consumption can be
compared to those of other production locations, to a legisla-
tive standard or to a generally acceptable health based target.
Because of the ease of use and the standardization of data
formats, the tool not only facilitates QMRA for Dutch drinking
water companies, but, in fact, for any drinking water company
or other interested party such as a policy maker, inspector or
consumer. User guidance is provided on how to structure the
required raw microbial data. Then, by a simple push on the
button, data are analyzed and a QMRA report is produced for
four index pathogens in less than a few minutes. This may be
described as a quick-and-clean (not dirty) risk assessment.
Although a QMRA based on actual raw data is highly
preferred, it is also possible to conduct a QMRA using distri-
bution parameters for pathogen concentrations in the source
5575
water, for each treatment step, and for drinking water
consumption. Such parameters values may be taken from
literature (e.g. Teunis et al., 2009). Moreover, QMRAspot can be
used to estimate the distribution parameters for a specific
(additional) treatment step to achieve the required pathogen
removal. Generic information can be applied to define
scenarios to answer a variety of what-if questions.
QMRAspot has a strong educative character by providing
guidance on how to structure data collection, analysis and risk
assessment, how to report the results, and by aiding policy
makers and decision makers in formulating mitigation strat-
egies, preventive measures and prioritization of measures. The
tool provides insight into the efficiency of the treatment steps.
In all these regards, the tool facilitates risk communication.
Acknowledgements
This work was performed by order and on the account of The
Environmental Inspectorate, who is highly acknowledged for
its contribution in defining of and observing compliance with
Inspectorate Guideline 5318. The Dutch drinking water
companies are highly acknowledged for their contribution
with data and fruitful discussions as participants in the
Infection Risk Working Group are the other participants.
references
Anonymous, 2001. Besluit van 9 januari 2001 tot wijziging van het
waterleidingbesluit in verband met de richtlijn betreffende de
kwaliteit van voor menselijke consumptie bestemd water.
Staatsblad van het Koninkrijk der Nederlanden 31, 1e53.
Anonymous, 2005. VROM-Inspectierichtlijn Analyse
microbiologische veiligheid drinkwater. VROM-Inspectie.
Artikelcode 5318.
Benford, D., 2001. Principals of Risk Assessment of Food and
Drinking Water Related to Human Health. ILSI Europe concise
monograph SERIES. International Life Science Institute,
Belgium, pp. 1e43.
De Roda Husman, A.M., Lodder, W.J., Rutjes, S.A., Schijven, J.F.,
Teunis, P.F.M., 2009. Long-term inactivation study of three
enteroviruses in artificial surface and groundwaters, using
PCR and cell culture. Applied Environmental Microbiology 75
(4), 1050e1057.
Ferguson, C., Kaucner, C., Krogh, M., Deere, D., Warnecke, M.,
2004. Comparison of methods for the concentration of
Cryptosporidium oocysts and Giardia cysts from raw waters.
Canadian Journal of Microbiology 50, 657e682.
Haas, C.N., Eisenberg, J.N.S., 2001. Risk assessment. In: Fewtrell, L.,
Bartram, J. (Eds.), Water Quality: Guidelines, Standards and
Health. IWA Publishing, London, pp. 161e183.
Haas, C.N., Rose, J.B., Gerba, C.P., 1999. Quantitative Microbial
Risk Assessment. Wiley, New York, USA.
Hijnen, W.A.M., Dullemont, Y.J., Schijven, J.F., Hanzens-
Brouwer, A.J., Rosielle, M., Medema, G.J., 2007. Removal and
fate of Cryptosporidium parvum, Clostridium perfringens and
small-sized centric diatoms (Stephanodiscus hantzschii) in slow
sand filters. Water Research 41 (10), 2151e2162.
ILSI Risk Science Institute Pathogen Risk Assessment Working
Group, 1996. A conceptual framework to assess the risks of
human disease following exposure to pathogens. Risk
Analysis 16, 841e848.
5576 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 6 4 e5 5 7 6
Kay, D., Watkins, J., Francis, C.A., Wyn-Jones, A.P., Stapleton, C.M.,
Fewtrell, L., Wyer, M.D., Drury, D., 2007. The microbiological
quality of seven large commercial private water supplies in the
United Kingdom. Journal of Water and Health 5 (4), 523e538.
Kistemann, T., Classen, T., Koch, C., Dangendorf, F., Fischeder, R.,
Gebel, J., Vacata, V., Exner, M., 2002. Microbial load of drinking
water reservoir tributaries during extreme rainfall and runoff.
Applied Environmental Microbiology 68 (5), 2188e2197.
Laverick, M.A., Wyn-Jones, A.P., Carter, M.J., 2004. Quantitative
RT-PCR for the enumeration of noroviruses (Norwalk-like
viruses) in water and sewage. Letters in Applied Microbiology
39 (2), 127e136.
Leclerc, H., Schwartzbrod, L., Dei-Cas, E., 2002. Microbial agents
associated with waterborne diseases. Critical Reviews in
Microbiology 28 (4), 371e409.
Lodder, W.J., van den Berg, H.H., Rutjes, S.A., de Roda Husman,
A.M., 2010. Presence of enteric viruses in source waters for
drinking water production in The Netherlands. Applied
Environmental Microbiology 76 (17), 5965e5971.
Medema, G.J., Hoogenboezem, W., van der Veer, A.J., Ketelaars,
H.A.M., Hijnen, W.A.M., Nobel, P.J., 2003. Quantitative risk
assessment of cryptosporidium in surface water treatment.
Water Science & Technology 47, 241e247.
Okhuysen, P.C., Rich, S.M., Chappell, C.L., Grimes, K.A.,
Widmer, G., Feng, X.C., Tzipori, S., 2002. Infectivity of
a Cryptosporidium parvum isolate of cervine origin for
healthy adults and interferon-gamma knockout mice. Journal
of Infectious Diseases 185 (9), 1320e1325.
Quintero-Betancourt, W., Gennaccaro, A.L., Scott, T.M., Rose, J.B.,
2003. Assessment of methods for detection of infectious
Cryptosporidium oocysts and Giardia cysts in reclaimed effluents.
Applied Environmental Microbiology 69 (9), 5380e5388.
Regli, S., Rose, J.B., Haas, C.N., Gerba, C.P., 1991. Modeling the risk
from Giardia and viruses in drinking water. Journal AWWA 83
(11), 76e84.
Rutjes, S.A. and de Roda Husman, A.M, 2004. Procedure for virus
detection in water for use in the Dutch drinking-water
guidelines. RIVM Report 330000007, 38 p. (in Dutch).
Rutjes, S.A., Lodder, W.J., van Leeuwen, A.D., de Roda Husman, A.
M., 2009. Detection of infectious rotavirus in naturally
contaminated source waters for drinking water production.
Journal of Applied Microbiology 107 (1), 97e105.
Schets, F.M., Medema, G.J. and Schijven, J.F, 2004. The recovery of
the detection method for Cryptosporidium and Giardia in
water. RIVM Report 330000008, 71 p. in Dutch.
Schets, F.M., van Wijnen, J.H., Schijven, J.F., Schoon, H., de Roda
Husman, A.M., 2008. Monitoring of waterborne pathogens in
surface waters in amsterdam, the Netherlands, and the
potential health risk associated with exposure to
cryptosporidium and giardia in these waters. Applied
Environmental Microbiology 74 (7), 2069e2078.
Schijven, J.F., de Roda Husman, A.M., 2011. QMST and QMRA
(Chapter 25).
Schijven J.F., Medema G.J., Nijs A.C.M. de and Elzenga J.G, 1996.
Emissie en verspreiding van Cryptosporidium, Giardia en
enterovirussen via huishoudelijk afvalwater. RIVM report no.
289202014.
Schijven, J.F., de Roda Husman, A.M., 2005. Effect of climate
changes on waterborne disease in the Netherlands. Water
Science & Technology 51 (5), 79e87.
Schijven, J.F. and de Roda Husman, A.M., 2009. Analysis of the
microbiological safety of drinking water. Experiences with
handling records 2006e7. RIVM report 703719038, (in
Dutch).
Schijven, J.F., Colin, M., Dullemont, Y., Hijnen, W.A.M., Magic-
Knezev, A., Oorthuizen, W., Wubbels, G., Bruins, J., Rutjes, S.A.
and de Roda Husman, A.M., 2008. Effect filtratiesnelheid,
temperatuur en korrelgrootte op de verwijdering van micro-
organismen door langzame zandfiltratie, 2008 RIVM report
330204002, (in Dutch).
Schmidt, P.J., Emelko, M.B., 2011. QMRA and decision-making:
are we handling measurement errors associated with
pathogen concentration data correctly? Water Research 45
(2), 427e438.
Skraber, S., Schijven, J.F., Gantzer, C., de Roda Husman, A.M.,
2005. Pathogenic viruses in drinking-water biofilms: a public
health risk? Biofilms 2, 105e117.
Summerill, C., Pollard, S.J., Smith, J.A., 2010. The role of
organizational culture and leadership in water safety plan
implementation for improved risk management. Science of
the Total Environment 408 (20), 4319e4327.
Teunis, P.F.M., Chappell, C.L., Okhuysen, P.C., 2002a.
Cryptosporidium dose response studies: variation between
isolates. Risk Analysis 22 (1), 175e183.
Teunis, P.F.M., Rutjes, S.A., Westrell, T., de Roda Husman, A.M.,
2009. Characterization of drinking water treatment for virus
risk assessment. Water Resarch 43 (2), 395e404.
Teunis, P.F.M., Havelaar, A.H., 2000. The Beta Poisson model is not
a single hit model. Risk Analysis 20, 511e518.
Teunis, P.F.M., Chappell, C.L., Okhuysen, P.C., 2002b.
Cryptosporidium dose-response studies: variation between
hosts. Risk Analysis 22 (3), 475e485.
Teunis, P.F.M., Evers, E.G., Slob, W., 1999. Analysis of variable
fractions resulting from microbial counts. Quantitative
Microbiology 1, 63e88.
Teunis, P.F.M., Lodder, W.J., Heisterkamp, S.H., de Roda
Husman, A.M., 2005a. Mixed plaques: statistical evidence how
plaque assays may underestimate virus concentrations.
Water Research 39 (17), 4240e4250.
Teunis, P.F.M., Medema, G.J., Kruidenier, L., Havelaar, A.H., 1997.
Assessment of the risk of infection by Cryptosporidium or
Giardia in drinking water from a surface water source. Water
Research 31 (6), 1333e1346.
Teunis, P.F.M., Van den Brandhof, W., Nauta, M., Wagenaar, J.,
Van den Kerkhof, H., Van Pelt, W., 2005b. A reconsideration of
the Campylobacter dose-response relation. Epidemiology and
Infections 133 (4), 583e592.
Teunis, P.F.M., Van der Heijden, O.G., Van de Giessen, J.W.B.,
Havelaar, A.H., 1996. The dose-response relation in human
volunteers for gastro-intestinal pathogens. RIVM Report
284550002.
USEPA, 2006. Economic Analysis for the Final Ground Water Rule.
United States Environmental Protection Agency. 2006. EPA
815-R-06e014. www.epa.gov/safewater.
Westrell, T., Andersson, Y., Stenström, T.A., 2006a. Drinking
water consumption patterns in Sweden. Journal of Water and
Health 04.4 551-522.
Westrell, T., Teunis, P., Van den Berg, H., Lodder, W., Ketelaars, H.,
Stenström, T.A., de Roda Husman, A.M., 2006b. Short- and long-
term variations of norovirus concentrations in the Meuse river
during a 2-year study period. Water Research 40, 2613e2620.
WHO, 2011. Guidelines for Drinking-Water Quality, fourth ed.
World Health Organisation, Geneve, Zwitserland, ISBN 978 92
4 154815 1.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Embodied energy comparison of surface water

and groundwater supply options

Weiwei Mo a, Qiong Zhang a,*, James R. Mihelcic a, David R. Hokanson b

a
Civil and Environmental Engineering Department, University of South Florida, Tampa, FL 33620, USA
b
Trussell Technologies, Inc., 232 North Lake Ave, Suite 300, Pasadena, CA, USA

article info abstract

Article history: The embodied energy associated with water provision comprises an important part of water
Received 16 February 2011 management, and is important when considering sustainability. In this study, an
Received in revised form inputeoutput based hybrid analysis integrated with structural path analysis was used to
4 August 2011 develop an embodied energy model. The model was applied to a groundwater supply system
Accepted 9 August 2011 (Kalamazoo, Michigan) and a surface water supply system (Tampa, Florida). The two systems
Available online 16 August 2011 evaluated have comparable total energy embodiments based on unit water production.
However, the onsite energy use of the groundwater supply system is approximately 27%
Keywords: greater than the surface water supply system. This was primarily due to more extensive
Embodied energy pumping requirements. On the other hand, the groundwater system uses approximately 31%
Hybrid analysis less indirect energy than the surface water system, mainly because of fewer chemicals used
Life cycle assessment for treatment. The results from this and other studies were also compiled to provide a rela-
Energy path tive comparison of embodied energy for major water supply options.
Groundwater ª 2011 Elsevier Ltd. All rights reserved.
Surface water
Reclaimed water
Water supply
Water treatment

1. Introduction withdrawal, treatment, and distribution. The energy used

Global water withdrawals have increased rapidly over the past
several decades, and are expected to continue to grow in the
near future (Shah et al., 2003; Konikow and Kendy, 2005; USGS,
2010). Extensive groundwater and surface water withdrawals
have led to environmental problems, such as groundwater
depletion, land subsidence, seawater intrusion, and surface
water quality deterioration, which have consequently impacted
water availability in many regions (Taylor and Alley, 2001;
Barlow, 2003; USGS, 2003; Konikow and Kendy, 2005).
The environmental impacts associated with water supply
are further compounded by energy requirements during
onsite for constructing, operating, and maintaining water
supply systems is referred to here as “direct energy.” It
comprises around 33% of a typical city’s government energy
budget for public utilities in California (CEC, 1992; AwwaRF,
2004) and around 2e3% of global energy demand (ASE, 2002).
The energy associated with material use and administrative
services is referred to here as “indirect energy.” Previous
studies suggest that indirect energy of water supply is
comparable to, or even greater than, direct energy (Mo et al.,
2009). The embodied energy (direct and indirect energy)
associated with water provision also increases with growing
water demand. For instance, direct energy increases with

* Corresponding author. Tel.: þ1 813 974 6448; fax: þ1 813 974 2957.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.016
5578 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
a declining water table and well yield, while indirect energy
increases when more sophisticated technologies and addi-
tional chemicals are used to treat water sources of poorer
quality.
Reduction of energy use and associated carbon emissions
from water supply is also gaining increased attention. For
example, in the US, states like California (under Assembly Bill
32) are requiring a reduction in carbon emissions from water
supply and treatment. In light of global water management
issues, consideration of the energy embodied in water
systems should become more important in the future.
Accordingly, this study focused on the energy embodiment in
water supply systems. Other impact categories associated
with material use were not considered as they are beyond the
scope of the study.
In the last decade, efforts have been made to evaluate the
embodied energy of water importation, reclamation, and
desalination, driven by the specific regional needs (Peters,
2005; Raluy et al., 2005; Tangsubkul et al., 2005; Stokes and
Horvath, 2006; Lyons et al., 2009). The energy embodied in
surface water systems has also been studied in countries such
as Canada (Racoviceanu et al., 2007) and South Africa
(Friedrich, 2002). Embodied energy values associated with
specific water supply options are summarized in Table 1.
Although environmental impacts such as greenhouse effects,
acidification, and nutrient enrichment of groundwater and
surface water supply have been compared (Godskesen et al.,
2011), no direct comparison has been made in terms of
energy embodiment between surface water and groundwater
systems as shown in Table 1.
Direct energy use associated with groundwater and surface
water supply systems, on the other hand, has previously been
examined on large scales (e.g., Wilkinson (2000) performed
a study for the state of California; EPRI (2002) performed
a study for the US). Specifically, the study published by the
Electric Power Research Institution (EPRI, 2002) concluded that
a groundwater supply system requires about 30% more elec-
tricity on a unit basis than a surface water supply system.
Table 1 e Life cycle energy associated with water supply systems identified in previous studies.
Water Sources Embodied Energy Methodology
(MJ/m3 of water)a
Imported water 18 Process based hybrid LCA
5 Process LCA
Desalinated water 42 Process based hybrid LCA
41 Process based hybrid LCA
27 Process based hybrid LCA
24 Process LCA
Recycled water 17 Process based hybrid LCA
3 Process LCA
Surface water 3 Process based hybrid LCA
2 Process LCA
a Energy was reported in the primary energy form, which includes the direct use of energy found in nature and the use of secondary energy
such as electricity in forms of fossil fuels, nuclear energy and renewable energy.
Neither of the studies, however, addresses indirect energy
consumption.
Three methods are primarily used by previous researchers
for estimating embodied energy: (1) traditional life cycle
assessment, (2) process based hybrid approach, and (3)
inputeoutput based hybrid approach. The traditional life
cycle assessment tends to underestimate the energy
embodiments because of limited data sources and truncated
system boundaries (Crawford, 2008). The process based hybrid
approach sums the direct energy and the inputeoutput results
of the energy embodied in each type of materials. It is more
complete than the traditional life cycle assessment; however,
it usually suffers from limited data sources for material use,
and thus cannot be readily applied to other systems. Accord-
ingly, an inputeoutput based hybrid approach was utilized in
this study. This approach involves substituting available
process data into an inputeoutput model in order to minimize
the errors associated with the traditional life cycle assessment
and the process based hybrid analysis (Crawford, 2008).
Previous studies (Crawford, 2008; Mattila et al., 2010) have
shown that the inputeoutput based hybrid approach is more
comprehensive and less labor intensive than the traditional
life cycle assessment. Additionally, the inputeoutput based
hybrid approach enables flexibility by first providing a rough
estimation, and then allowing detailed modifications based on
site and system-specific data using structural path analysis.
One weakness of this approach is that neither differences in
water consumption patterns nor temporal differences asso-
ciated with water supply systems can be reflected in the
model results.
The objective of this study was therefore to estimate the
“cradle to gate” (source to customer) energy embodiment
(direct and indirect energy) of one groundwater and one
surface water supply system and to provide a relative
comparison of embodied energy for major water supply
options through the compilation of results from this and
previous studies. The novelty of this study lies in the use of an
inputeoutput based hybrid approach with structural path
Comments Source
Conveyance pipe length: 575 km Stokes and Horvath, 2009
Conveyance pipe length: 261 km Lyons et al., 2009
Reverse osmosis with Stokes and Horvath, 2009
conventional pretreatment
Reverse osmosis with Stokes and Horvath, 2009
membrane pretreatment
Brackish groundwater Stokes and Horvath, 2009
Reverse osmosis Lyons et al., 2009
Stokes and Horvath, 2009
Lyons et al., 2009
Only considers operation phase Racoviceanu et al., 2007
of the treatment plant
Friedrich, 2002
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6 5579

analysis to provide more comprehensive results with insights mining, (3) power generation and supply, (4) natural gas
into the energy flow. distribution, and (5) petroleum refineries.
The total embodied energy intensities (i.e., the total energy
in primary energy form per dollar output of the water systems)
2. Methodology for system construction, operation and maintenance phases
were estimated using Eqs. (1)e(2) based on data from the 2002
An inputeoutput based hybrid approach was used in this inputeoutput tables. Additional details on the calculation of
study for estimating embodied energy. Basic steps involved in total embodied energy intensities can be found in Mo et al.
the approach are presented in Fig. 1. This same approach can (2010).
be used for estimating energy embodied in other water, 0 1
X
N X
N
wastewater, and industrial systems as long as the user has 3
P ¼ @ dik1 ;jk 3 ik1 A (1)
t
identified appropriate economic target sectors and has access k¼1 i;j¼1
to system-specific data.
The system boundary in this study includes the construc- X
5

tion and operation stages of water intake infrastructures
(wells/exposed tower), treatment plants (administrative
buildings included), water storage tanks, pipeline systems,
and pumping stations. The end-of-life stage was not consid-
ered because the embodied energy associated with it has been
shown to be insignificant in previous studies (Friedrich, 2002;
Raluy et al., 2005). Among the 424 commodity sectors provided
3 i0 ¼ dn;i0  tarriff n  an (2)
n¼1
In Eqs. (1) and (2), 3 P t ¼ total embodied energy intensity of
the target sector “t”, TJ/$ output of sector “t”; k ¼ stage index;
N ¼ number of sectors in stage k. dik1 ;jk ¼ direct coefficient
from sector “i” at stage k  1 to sector “j” at stage k;
3 ik1 ¼ energy intensity of sector “i” at k  1 stage, TJ/$ output of

in the 2002 inputeoutput tables provided by the Bureau of
Economic Analysis (BEA, 2007), the water-related sectors were
identified as the “water, sewage and other systems” sector
(WSOS), representing system operation and maintenance
(O&M) phase, and the “other nonresidential structures” sector
(NS), representing the system construction phase. Addition-
ally, five energy supply sectors were identified within the 424
commodity sectors. They are: (1) oil and gas extraction, (2) coal
sector “i”; 3 i0 ¼ direct energy intensity of sector “i” at stage 0,
TJ/$ output of sector “i”; n ¼ energy supply sector index;
dn;i0 ¼ direct coefficient from energy supply sector n into sector
“i”; tarriffn ¼ energy tariff of the energy supply sector n, TJ/$
energy; and, an ¼ nationally averaged primary energy factor of
energy supply sector n.
After calculating the energy intensities of the water-related
sectors, they were multiplied with their corresponding

Fig. 1 e Flow chart for the development of the embodied energy model.
5580 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
economic activities (expenses) to obtain the initial embodied
energy. The initial embodied energy was then adjusted for
individual systems through structural path analysis (Treloar
et al., 2001; Lenzen and Crawford, 2009). This reduces errors
caused by sector aggregation in the inputeoutput tables and
provides an insight of the energy flow in water supply. The
final embodied energy was obtained after the adjustments.
The manuscript follows with an introduction to the methods
in expense estimation, structural path analysis, and modifi-
cation of the total embodied energy.
2.1. Expense estimation
The WSOS sector represents the O&M activities in water
supply systems. The monetary output of the WSOS sector is
the annual expenses for operating and maintaining water
supply systems, which were obtained from the selected water
supply systems.
The NS sector represents the activities in constructing
water supply systems. The monetary output of the NS sector
is the capital costs of water systems. Because it is very diffi-
cult to obtain the total capital costs directly from the water
systems due to expansions and renovations over time, in this
study, cost equations and curves were carefully selected to
best estimate the capital costs of the existing systems
including the capital costs of the treatment processes,
equipments, and administrative buildings (Gumerman et al.,
1979; Mickley, 2001; Traviglia and Characklis, 2008; McGivney
and Kawamura, 2008). Costs estimated by the equations
and curves from years other than 2002 were adjusted to 2002
$USD.
2.2. Structural path analysis
In order to modify the initial embodied energy, energy paths
(supply chains starting from the energy involved in one
material or service supply sector, and ending at the water-
related sector) representing high percentages of the total
Fig. 2 e Description of Energy path, stages and relationship between different stages with a sample of a 3-stage energy path.
embodied energy intensities were extracted through struc-
tural path analysis. The terms of energy paths and stages are
illustrated in Fig. 2. To reduce the amount of calculations, up
to 5-stage energy paths were checked. Threshold values were
selected to determine the amount of energy paths to be
extracted. The large amount of commodity sectors in the US
inputeoutput tables leads to an extremely large number of
energy paths to be extracted. Thus, in order to represent
greater than 90% of the initial total embodied energy intensi-
ties, threshold values were selected to extract paths repre-
senting 90% of the initial total embodied energy intensities for
the two water-related sectors in this study.
2.3. Modification of the total embodied energy
To modify the initial direct energy, system-specific data from
the water supply systems were substituted in to replace the
initial model estimations. Due to data limitations, however,
this adjustment was only performed for the WSOS sector.
To modify the initial indirect energy, the method presented
by Treloar (1997) and Lenzen and Crawford (2009) was used.
For a certain 1-stage energy paths (from sector s1 to target
sector), the energy involved can be calculated as:
Es1 ;0 ¼ 3 s1 Cs1 ¼ 3 s1 ds1 ;t Ct (3)
where: Es1 ;0 ¼ the initial energy for the energy path from sector
“s1” (the sector in stage 1) to the target sector “t”, TJ; Es1 ¼ direct
energy intensity of sector “s1”, TJ/$ output of sector “s1”;
Cs1 ¼ direct purchase from sector “s1” by the target sector “t”, $;
ds1 ;t ¼ direct coefficient from sector “s1” to the target sector “t”,
$/$ output of the target sector “t”; and, Ct ¼ total monetary
output of the target sector “t”, $.
According to Eq. (3), the calculation of a 1-stage energy path
contains two parts, the direct energy intensity of “s1” sector
(3 s1 ) and the amount of “s1” commodity directly used by the
target sector (Cs1 ). Both parts were adjusted based on available
data. As shown in Eq. (3), Cs1 was calculated by multiplying the
direct coefficient with the total monetary output of the water-
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6 5581

related sector “t” in the inputeoutput analysis. It can be Environmental Protection Agency according to the population
adjusted using detailed expenses associated with different they serve (both systems serve > 100,000 people) (EPA, 2010);
items obtained from the selected water supply systems. To (2) they represent typical groundwater and surface water
adjust 3 s1 energy use for manufacturing sectors in 2002 was treatment processes; and (3) data for these two systems are
obtained from the Energy Information Administration (EIA, readily available to the authors. Geographic differences of the
2010). The modified energy can be calculated using Eq. (4) two systems were not considered in this study. A detailed
! ! comparison of the two systems is provided in Table 2.
adj adj
3
s1
Cs1
Es1 ;D ¼ Es1 ;0 rs1 ¼ Es1 ;0 (4)
3 s1 Cs1 3.1. Kalamazoo Public Water Supply System
In Eq. (4), rs1 ¼ ratio of the modified energy to the initial
The Kalamazoo Public Water Supply System (referred to as
energy for the energy path from sector “s1” to the target sector
Kalamazoo system) is the largest groundwater based water
“t”; Es1 ;D ¼ modified energy for the energy path from sector “s1”
adj supply system in the Kalamazoo River watershed, serving
to the target sector “t”, TJ; 3 s1 ¼ adjusted direct energy
over 121,000 customers. The Kalamazoo system pumps an
intensity of sector “s ”, TJ/$ output of sector “s1”; and,
1

adj average of 76.8 thousand m3 of water per day and deploys

Cs1 ¼ adjusted direct purchase from sector “s1” by the target
1276 km of water mains. Raw water is withdrawn from 101
sector “t”, $.
local wells with an average well depth of around 58 m. Limited
For energy paths with i stages, the initial energy involved
treatment (disinfection) is provided in two of the total 18
can be determined using Eq. (5):
pumping stations, after which the water is supplied to the end
Y
i
  users (Kalamazoo, 2008).
Esi ;0 ¼ 3 si dsk ;sk1 Ct (5) The annual O&M expense in the Kalamazoo system is
k¼1
approximately $11.1 million. Of the $11.1 million annual
where, Esi ;0 ¼ initial energy for the energy path from the sector expense, $1.16 million are used for purchasing electricity, and
in stage i “si” to the target sector “t” (target sector is the sector $0.08 million are used for purchasing natural gas (CKWD,
in stage 0, s0), TJ; 3 si ¼ direct energy intensity of the sector in 2010). The commodity output of the NS sector (construction)
stage i “si”, TJ/$ output of sector “si”; and, dsk ;sk1 ¼ direct was estimated based on the capital cost of the Kalamazoo
coefficient from sector “sk” to sector “sk1”, (s0 represents the system. Since the Kalamazoo system only has limited
sector in stage 0 which is the target sector “t”). treatment within the pumping stations, the water treatment
Similarly, the modified energy for energy path from the infrastructure was not considered separately. The well data
sector “si” to the target sector “t” can be calculated using Eq. (6) of the Kalamazoo system were obtained from the “Water
with rsi . Well Viewer” (MDEQ, 2009a) and “Wellogic” (MDEQ, 2009b)
adj
! adj
! managed by the Michigan Department of Environmental
3
si
Cs1 Quality.
Esi ;D ¼ Esi ;0 rsi ¼ Es1 ;0 (6)
3 si Cs1

For the commodity use at stage “i”, an assumption has been
made that the change of direct commodity use will cause the
upstream supply of this commodity to change proportionally.
Also, the indirect energy was modified by substituting the
original energy embodied in each energy path with the
modified energy.
3. Description of water systems used in
study
One groundwater supply system (Kalamazoo Public Water
Supply System, Michigan) and one surface water supply
system (City of Tampa Waterworks, Florida) were studied.
These two systems were chosen because: (1) both of them are
classified as “very large” water supply systems by the US
3.2. City of Tampa Waterworks
The City of Tampa Waterworks (referred to as Tampa system)
is one of the largest water supply systems in Florida, serving
a population of 657,000. The average daily flow in the system is
approximately 287 thousand m3, about 3.7 times higher than
the average flow in the Kalamazoo system. However, the
impact of such differences on direct energy use per unit water
produced is negligible at the production scale between 38
thousand m3 per day (10 MGD) and 380 thousand m3 per day
(100 MGD) (EPRI, 2002). As a result, it is allowable for us to
compare the total embodied energy of the two systems.
The Tampa system has more than 3541 km of water mains.
Raw water is withdrawn from the Hillsborough River, and
treated with pre-ozonation and GAC filters in addition to
a conventional process that consists of flash mix, flocculation
and sedimentation. The raw water has a turbidity of 15e220

Table 2 e Key information of the Kalamazoo system and the Tampa system.
Water supply systems Water source Daily flow Serving Percentage of chemical Length of the
(thousand m3/day) population cost with total O&M cost pipelines (km)

The Kalamazoo system Groundwater 76.8 121,000 2% 1276

The Tampa system Surface water 287 657,000 13% 3541
5582 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6

NTU with an average of 117 NTU. The detected dissolved

Table 3 e Threshold value that represents 90% of the
oxygen has a range of 1.9e14.3 mg/L with an average of initial total embodied energy intensity for WSOS (the
4.1 mg/L. The bromide detected ranges from 31 to 180 mg/L “water, sewage and other systems” sector) and NS (the
with an average of 85 mg/L. This is greater than the Maximum “other nonresidential structures” sector).
Daily Level of 0.5 mg/L. Total organic carbon ranges from 3.3 to Sector WSOS NS
24.2 mg/L with an average of 15.1 mg/L.
The annual O&M expense is $68.3 million. Of the $68.3 Threshold Value (GJ/$) 1.48E-07 2.40E-09
million annual expense, $3.95 million is used for purchasing 1-stage energy paths a
179 190
electricity. The commodity output of the NS (construction) is 2-stage energy paths 1031 8889
estimated based on the capital cost of the Tampa system. Key 3-stage energy paths 240 19,356
4-stage energy paths 8 4819
information used for estimating capital cost was collected
5-stage energy paths 0 499
directly from the Tampa system.
a Energy path: Supply chains start from the energy involved in one
material or service supply sector, and end at the water-related
sector.
4. Results and discussion

4.1. Expense estimation
The estimated total capital expense in the Kalamazoo system
is $118.4 million, and the total capital expense in the Tampa
system is $416.0 million. The breakdowns of the capital costs
in both systems are provided in Fig. 3. Assuming life spans for
both systems of 100 years (Peters, 2005; Stokes and Horvath,
2006), the unit capital expense for the Kalamazoo system is
around $42 per thousand m3 of water produced, and the unit
O&M expense is around $394 per thousand m3 of water
produced. The total cost (construction and O&M) for
producing one thousand m3 of water in the Kalamazoo system
is $436. Similarly, the unit capital expense for the Tampa
system is around $40 per thousand m3 of water produced, and
the unit O&M expense is around $653 per thousand m3 of
water. The total cost for producing one thousand m3 of water
in the Tampa system is $692.
The unit O&M expense of the Tampa system is much larger
than the Kalamazoo system. This may be because of the much
greater use of water treatment chemicals in the Tampa
system. On the other hand, the unit capital expenses of both
systems are similar, even though the Tampa system has an
additional water treatment plant. The percentage of the
Kalamazoo system pipeline capital expense within the total
capital expense is much larger than that of the Tampa system.
This may result from the more distributed water intake
infrastructure in the Kalamazoo system compared with the
Tampa system and the lower population density in the City of
Kalamazoo compared with the City of Tampa (USCB, 2010). For
both systems, pipeline construction is the largest capital cost
contributor.
Overall, the results show that surface water supply
systems may be more expensive to operate than the ground-
water supply systems depending on the raw water quality, but
may be less expensive to construct than the groundwater
supply systems depending on the length of pipelines.
4.2. Top energy paths for water-related sectors
For the structural path analysis, the threshold values selected
for both sectors and the numbers of energy paths in each of
the five stages checked are provided in Table 3. Stage 2 has the
most energy paths extracted for the WSOS sector; while stage
3 has the most energy paths for the NS sector. Overall, the NS
sector has significantly more energy paths than the WSOS

28 0.9 0.8 10 0.1

Tampa Pipeline systems

System
Wells (Kalamazoo)

Pumping stations

37 0 .1 2 .0 33.2
2 Water storage tanks

Water treatment plant

Kalamazoo
System Exposed tower
(Tampa))

0 5 10 15 20 25 30 35 40 45
Capital Cost (Dollar per thousand m3 in $ 2002 under 100-year lifetime)

Fig. 3 e Breakdown of capital costs per thousand cubic meter of water produced under 100 year life-time associated with the
groundwater sourced Kalamazoo system and the surface water sourced Tampa system in $ 2002.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
Table 4 e Total embodied energy for groundwater sourced Kalamazoo system and the surface water sourced Tampa
system.
Water Supply Systems Direct Energy (MJ/m3)
O&M Construction Total
The Kalamazoo system 6.1 0.2 6.3
The Tampa system 4.8 0.2 5.0
Differencesa 28% 6% 27%
a Differences ¼ [(Data from the Kalamazoo System -Data from the Tampa System)/Data from the Tampa System].
sector in the top 90% of the initial total embodied energy
intensity, and the paths in the NS sector are more evenly
distributed in different stages than in the WSOS sector. The
top energy paths of the WSOS sector are mainly related to
maintenance and engineering services, production of the
treatment and maintenance materials, and transportation.
The top energy path of the NS sector are mainly involved
with the production of the building materials, such as
asphalt, steel, cement, stone etc., engineering services and
transportation.
Engineering services have a large impact on the indirect
energy use of water systems because large amounts of energy
and materials are required to provide such services. Chemical
and construction material production is another major
contributor to the indirect energy because a large amount of
energy is consumed during each stage of the production
processes. Lastly, transportation plays a significant role in
energy consumption for both constructing and operating
water systems.
4.3. Modification and calculation of the total
embodied energy
The system-specific O&M direct energy use was estimated
through the annual energy expenditures and local average
energy prices. The average electricity retail price in Michigan
is 9.18 cents/kWh, and the average price of natural gas is 6.1
dollars/GJ. Thus, the direct energy for operating and main-
taining the Kalamazoo system was estimated to be 170 TJ. The
direct energy for both O&M and construction amounts to
6.3 MJ per m3 of water produced at the Kalamazoo system. On
the other hand, the average electricity retail price in Florida is
10.13 cents/kWh. Thus, the direct energy for the O&M of the
Tampa system was estimated to be 497 TJ. The direct energy
for both O&M and construction amounts to 5.0 MJ/m3 of water
produced at the Tampa system.
For indirect energy, the available system-specific data from
the Kalamazoo system and the Tampa system were
substituted to adjust the original embodied energy of the two
systems. The direct energy intensities of 25 manufacturing
sectors were also modified (EIA, 2010). Under a 100-year life
Table 5 e Breakdown of the major contributors to the total O&M embodied energy.
Energy use categories Direct energy use Chemicals
Kalamazoo System 61.9% 5.7%
Tampa System 46.1% 9.6%
5583
Indirect Energy (MJ/m3) Total Embodied Energy (MJ/m3)
O&M Construction Total O&M Construction Total
3.7 0.3 4.0 9.8 0.5 10.3
5.5 0.3 5.8 10.3 0.5 10.8
32% 9% 31% 5% 8% 4%
span, the indirect energy used for the Kalamazoo system to
supply 1 m3 of water is 4.0 MJ, and the indirect energy used for
the Tampa system to supply 1 m3 of water is 5.8 MJ after
modification.
After the modification of both direct and indirect energy,
the total embodied energies for the two water supply systems
are provided in Table 4. The total embodied energy in the
Kalamazoo system for supplying 1 m3 of water is 10.3 MJ, and
the total embodied energy in the Tampa system for supplying
1 m3 of water is 10.8 MJ. The unit total embodied energy in the
Tampa system is slightly larger than that of the Kalamazoo
system. Compared with initial total embodied energy, the
modified total embodied energy of the Kalamazoo system
increased by 68%, and the modified total embodied energy of
the Tampa system increased by 10%. The differences show the
necessity of the modification step using the system-specific
data in the analysis.
The unit direct energy consumption of the Kalamazoo
system is 27% higher than the Tampa system, which is
consistent with EPRI’s estimation. This result can be explained
by the large pumping requirement for water delivery in the
Kalamazoo system. It is also consistent with a previous result
that the pipeline system in the Kalamazoo system accounts
for a more important portion of energy consumption than the
Tampa system. Groundwater supply systems usually have
deep and widely distributed wells for water intake, which may
increase their pumping energy requirements.
Unlike the direct energy consumption, the unit indirect
energy consumption at the Kalamazoo system is around 31%
less than the Tampa system. This is primarily because of the
greater use of chemicals and engineering services at the
Tampa system. Groundwater supply systems typically have
better raw water quality than surface water supply systems.
Systems such as the Kalamazoo system require only limited
treatment, which significantly reduces the amount of required
chemicals. In contrast, the Tampa system uses a large quan-
tity of chemicals to treat the lower quality raw water. In
addition to disinfectants, other chemicals such as ferrous
sulfate (for coagulation) and ozone (for pre-ozonation) are
used. Manufacturing these chemicals is very energy intensive
based on the data from the inputeoutput tables. Moreover, the
Maintenance Engineering service Customer service
12.6% 0.7% 0.4%
13.9% 3.2% 0.4%
5584 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
Table 6 e Sensitivity analysis of the Kalamazoo system and the Tampa system.
Selected water supply systems Direct energy
þ50%
Total embodied energy
The Kalamazoo system þ30%
The Tampa system þ22%
surface water supply systems are usually more complicated
than the groundwater supply systems, thus more engineering
services are involved, which also contributes to the large
indirect energy demand of the Tampa system. Breakdown of
the major contributors to the total O&M embodied energy of
the two systems is provided in Table 5.
4.4. Comparison with other studies
The results from this study are higher than the embodied
energy provided by Racoviceanu et al. (2007) and Friedrich
(2002) partly due to the different system boundaries
selected. Unlike this study, Racoviceanu et al. (2007) only
considered the operation phase of the treatment plant, while
Friedrich considered all operation, construction, and decom-
mission phases of the treatment plant.
Furthermore, the estimated embodied energy varies a lot
based on different estimation methods used, different raw
water qualities and treatment technologies, and different
geographical locations. For instance, as shown in Table 1, even
the energy embodiments of the similar three water supply
options studied by Stokes and Horvath (2009) and Lyons et al.
(2009) differ by 2e4 fold. Although there is some variance in
previous results, desalination consistently appears as the
most energy intensive water supply option. Furthermore, the
embodied energy of surface and groundwater supplies is
comparable with options of water reclamation and importa-
tion. Additional studies are needed to compare groundwater,
surface water, and reclaimed water supply options in a similar
geographical area, with more details on raw water quality and
treatment process characteristics, in order to better under-
stand the energy and material use of these options.
4.5. Uncertainty and sensitivity analysis
Uncertainties in this study are primarily from the
inputeoutput tables, varied life span of different components,
different geographical location of the selected systems,
and capital expense estimation. Bullard and Sebald (1988)
found a standard error of 1% for row sums in the US 1967
inputeoutput tables, while Lenzen (2000) assumed an error
bound of 3% for the Australian inputeoutput tables. Because
there is a lack of studies on the truncation errors and sensi-
tivity of the recent US inputeoutput tables, uncertainty of our
results was not quantified.
A sensitivity analysis was carried out to determine how
direct energy and different inputs used for the estimation
would affect the results (Table 6). The analysis showed that
the results are very sensitive to the direct energy consumption
because it accounts for the largest portion of the total
Infrastructure life span Total chemical use
50% þ50%
Total embodied energy Indirect energy Total embodied energy
þ5% þ8% þ3%
þ4% þ9% þ5%
embodied energy. Additionally, the Kalamazoo system is
more sensitive to direct energy than the Tampa system, which
is consistent with the previous discussion that the Kalamazoo
system has higher unit direct energy use. The results are
however not very sensitive to the change of the system life
span. This is because the construction life stage only
comprises a small portion of the total embodied energy. In
regards to chemical use, the Tampa system is more sensitive
to it than the Kalamazoo system. This observation is also
consistent with the previous discussion that the Tampa
system has a larger indirect energy requirement, primarily
because of the greater use of chemicals.
5. Conclusions
The results from this study show that Kalamazoo ground-
water supply system that only employs disinfection with no
additional treatment is more energy intensive than Tampa
surface water supply system in terms of direct energy. This is
caused by higher pumping requirements; however, the
surface water supply system is more energy intensive in terms
of indirect energy because of greater requirements for mate-
rial use.
The results from this study are also higher than previous
life cycle studies performed on surface water systems due to
different system boundaries selected and different estimation
methods used. This study shows the flexibility of using the
inputeoutput based hybrid analysis based on data availability.
It can be easily used by researchers and utilities to evaluate
embodied energy of water supply systems. This method,
however, still has various uncertainties including errors
propagated from inputeoutput tables and uncertainties in the
capital cost estimation for the selected water supply systems.
Additionally, this study did not consider the geographical
differences between the two systems, which may also affect
the total embodied energy. The sensitivity analysis indicated
that the results are very sensitive to the direct energy use.
However, the results are not very sensitive to the system life
span. In addition, the embodied energy of the Tampa system
is more sensitive to the chemical use than that of the Kala-
mazoo system.
Although there is no significant difference on the total
embodied energy consumption for the specific groundwater
and surface water supply systems evaluated, the results
suggest there is a trade-off between direct and indirect energy
for different systems. It is thus important for water managers
to differentiate direct and indirect energy in future life cycle or
energy studies for water supply systems.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
Acknowledgments
This material is based in part upon work supported by the
National Science Foundation under Grant Numbers CBET
0725636. Any opinions, findings, and conclusions or recom-
mendations expressed in this material are those of the
authors and do not necessarily reflect the views of the
National Science Foundation. We would also like to thank the
Kalamazoo Water Department and Mr. Skip Pierpont from the
City of Tampa Waterworks for their assistance.
references
ASE, Alliance to Save Energy/US Agency for International
Development, 2002. Watergy: taking advantage of untapped
energy and water efficiency opportunities in municipal water
systems.
AwwaRF, Awwa Research Foundation, 2004. Water and
wastewater industry energy efficiency: a research roadmap,
Project #2923.
Barlow, P.M., 2003. Ground water in fresh water salt water
environments of the Atlantic coast. U.S. Geological Survey
Circular 1262.
BEA, Bureau of Economic Analysis, 2007. http://www.bea.gov/
industry/index.htm#benchmark_io (accessed 15.12.08.).
Bullard, C.W., Sebald, A.V., 1988. Monte Carlo sensitivity analysis
of inputeoutput models. The Review of Economics and
Statistics 70 (4), 708e712.
CEC, California Energy Commission, 1992. Energy efficiency
programs for cities, counties, and schools. P400-91-030.
CKWD, City of Kalamazoo Water Department, 2010. Personal
communication.
Crawford, R.H., 2008. Validation of a hybrid life-cycle inventory
analysis method. Journal of Environmental Management 88
(3), 496e506.
EIA, Energy Information Administration. http://www.eia.doe.gov/
emeu/mecs/mecs2002/data02/shelltables.html (accessed 15.
07.10.).
EPA, US Environmental Protection Agency. http://water.epa.gov/
infrastructure/drinkingwater/pws/factoids.cfm (accessed 01.
04.10.).
EPRI, Electric Power Research Institute, 2002. Water &
Sustainability. In: U.S. Electricity Consumption for Water
Supply & Treatment e The Next Half Century, vol. 4 No.
1006787, Palo Alto, California.
Friedrich, E., 2002. Life-cycle assessment as an environmental
management tool in the production of potable water. Water
Science and Technology 46 (9), 29e36.
Godskesen, B., Zambrano, K.C., Trautner, A., Johansen, N.B.,
Thiesson, L., Andersen, L., Clauson-Kaas, J., Neidel, T.L.,
Rygaard, M., Kloverpris, N.H., Albrechtsen, H.J., 2011. Life cycle
assessment of three water systems in Copenhagen e
a management tool of the future. Water Science and
Technology 63 (3), 565e572.
Gumerman, R.C., Culp, R.L., Hansen, S.P., 1979. Estimating water
treatment costs: volume 2-cost curves applicable to 1 to 200
MGD treatment plants. EPA-600/2-79-162b. National Technical
Information Service, Springfield, VA.
Kalamazoo, The City of Kalamazoo, 2008. 2007 Water quality
report. Environmental Service Division, Public Service
Department.
5585
Konikow, L.F., Kendy, E., 2005. Groundwater depletion: a global
problem. Hydrogeology Journal 13 (1), 317e320.
Lenzen, M., 2000. Errors in conventional and input-output-based
life-cycle inventories. Journal of Industrial Ecology 4 (4),
127e148.
Lenzen, M., Crawford, R., 2009. The path exchange method for
hybrid LCA. Environmental Science and Technology 43 (21),
8251e8256.
Lyons, E., Zhang, P., Benn, T., Sharif, F., Li, K., Crittenden, J.,
Costanza, M., 2009. Life cycle assessment of three water
supply systems: importation, reclamation and desalination.
Water Science & Technology 9 (4), 439e448.
Mattila, T.J., Pakarinen, S., Sokka, L., 2010. Quantifying the total
environmental impacts of an industrial symbiosis e
a comparison of process-, hybrid and input-output life cycle
assessment. Environmental Science and Technology 44 (11),
4309e4314.
McGivney, W., Kawamura, S., 2008. Cost estimating manual
for water treatment facilities. John Wiley & Sons,
Hoboken, N.J.
MDEQ, Michigan Department of Environmental Quality, 2009a.
Water Well Viewer. http://wellviewer.rsgis.msu.edu/
(accessed 31.08.09.).
MDEQ, Michigan Department of Environmental Quality, 2009b.
Welllogic. http://www.deq.state.mi.us/wellogic/main.html
(accessed 31.08.09.).
Mickley, M.C., 2001. Membrane Concentrate Disposal: Practices
and Regulation. Department of the Interior: Bureau of
Reclamation, Denver, CO, U.S, pp. 1e266.
Mo, W., Zhang, Q., Mihelcic, J.R., Hokanson, D.R., 2009.
Development and application of an embodied energy model
for individual water supply systems in Great Lakes Region. In:
Proceedings of the 80th Annual Water Environment
Federation Conference & Exposition Orlando, Florida, October
10e14, 2009.
Mo, W., Nasiri, F., Eckelman, M.J., Zhang, Q., Zimmermman, J.B.,
2010. Measuring the embodied energy in drinking water
supply systems: a case study in Great Lakes Region.
Environmental Science and Technology 44 (24), 9516e9521.
Peters, G.A., RK., 2005. Environmental sustainability in water
supply planning e an LCA approach for the Eyre Peninsula,
South Australia. In: 4th Australian Life Cycle Assessment
Conference, Sydney.
Racoviceanu, A.I., Karney, B.W., Kennedy, C.A., Colombo, A.F.,
2007. Life-cycle energy use and greenhouse gas emissions
inventory for water treatment systems. Journal of
Infrastructure Systems 13 (4), 261e270.
Raluy, R.G., Serra, L., Uche, J., 2005. Life cycle assessment of
desalination technologies integrated with renewable energies.
Desalination 183, 81e93.
Shah, T., Roy, A.D., Qureshi, A.S., Wang, J., 2003. Sustaining Asia’s
groundwater boom: an overview of issues and evidence.
Natural Resource Forum 27 (2), 130e141.
Stokes, J.R., Horvath, A., 2006. Life cycle energy assessment of
alternative water supply systems. International Journal of Life
Cycle Assessment 11 (5), 335e343.
Stokes, J.R., Horvath, A., 2009. Energy and air emission effects of
water supply. Environmental Science & Technology 43 (8),
2680e2687.
Tangsubkul, N., Beavis, P., Moore, S.J., Lundie, S., Waite, T.D.,
2005. Life cycle assessment of water recycling technology.
Water Resource Management 19, 521e537.
Taylor, C.J., Alley, W.M., 2001. Ground-water level monitoring and
the importance of long-term water-level data. U.S. Geological
Survey Circular 1217, 68.
Traviglia, A.M., Characklis, G.W., 2008. An expert system for
decision making in the use of desalination for augmenting
5586 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 7 7 e5 5 8 6
water supplies. Department of Environmental Sciences and
Engineering, University of North Carolina at Chapel Hill.
DWPR Report No. 107.
Treloar, G.J., 1997. Extracting embodied energy paths from input-
output tables: towards an input-output based hybrid energy
analysis method. Economic System Research 9 (4), 375e391.
Treloar, G., Love, P., Faniran, O.O., 2001. Improving the reliability
of embodied energy methods for project life-cycle decision
making. Logistics Information Management 14 (5/6), 303e318.
USCB, U.S. Census Bureau, 2010. State and County Quick Facts. http://
quickfacts.census.gov/qfd/index.html (accessed 15.12.10.).
USGS, U.S. Geographical Survey, 2003. Ground-water depletion
across the nation. U.S. Geological Survey Fact Sheet 103e03.
USGS, U.S. Geographical Survey, 2010. Groundwater Use in the
United States. http://ga.water.usgs.gov/edu/wugw.html
(accessed 15.07.10.).
Wilkinson, R., 2000. Methodology for analysis of the energy
intensity of California’s water systems and an assessment of
multiple potential benefits through integrated water energy
efficiency measures. Environmental Studies Program at the
University of California, Santa Barbara. Agreement No.
4910110.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Nitrate reduction in a simulated free-water surface

wetland system

Teresa M. Misiti, Malek G. Hajaya, Spyros G. Pavlostathis*

School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA 30332-0512, USA

article info abstract

Article history: The feasibility of using a constructed wetland for treatment of nitrate-contaminated
Received 13 June 2011 groundwater resulting from the land application of biosolids was investigated for a site
Received in revised form in the southeastern United States. Biosolids degradation led to the release of ammonia,
7 August 2011 which upon oxidation resulted in nitrate concentrations in the upper aquifer in the range
Accepted 11 August 2011 of 65e400 mg N/L. A laboratory-scale system was constructed in support of a pilot-scale
Available online 22 August 2011 project to investigate the effect of temperature, hydraulic retention time (HRT) and
nitrate and carbon loading on denitrification using soil and groundwater from the biosolids
Keywords: application site. The maximum specific reduction rates (MSRR), measured in batch assays
Denitrification kinetics conducted with an open to the atmosphere reactor at four initial nitrate concentrations
Groundwater from 70 to 400 mg N/L, showed that the nitrate reduction rate was not affected by the initial
Constructed wetland nitrate concentration. The MSRR values at 22
C for nitrate and nitrite were 1.2  0.2 and
Carbon loading 0.7  0.1 mg N/mg VSSCOD-day, respectively. MSRR values were also measured at 5, 10, 15
Nitrate loading and 22
C and the temperature coefficient for nitrate reduction was estimated at 1.13.
Temperature effect Based on the performance of laboratory-scale continuous-flow reactors and model simu-
lations, wetland performance can be maintained at high nitrogen removal efficiency
(>90%) with an HRT of 3 days or higher and at temperature values as low as 5
C, as long as
there is sufficient biodegradable carbon available to achieve complete denitrification. The
results of this study show that based on the climate in the southeastern United States,
a constructed wetland can be used for the treatment of nitrate-contaminated groundwater
to low, acceptable nitrate levels.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction applied to serve as a nutrient source for plant growth as well

World-wide, nitrate is among the most common groundwater
contaminants, mainly introduced into the environment from
agricultural activities related to the excessive use of nitrate-
containing fertilizers and manure (Burkart and Stoner, 2002;
Murgulet and Tick, 2009; Rivett et al., 2008). In addition, as
an alternative to landfilling, biosolids generated by the
anaerobic digestion of municipal primary and waste activated
sludge or other stabilization processes are in some cases land
as to enrich the soil in organic matter. In the case of land
application of biosolids, ammonia, which is either already
present or produced as a result of biosolids degradation on
site, is utilized as a nitrogen source for plant growth. When
biosolids are land applied at recommended agronomic
frequencies and rates, ammonia is not anticipated to be of
environmental concern. However, when biosolids are applied
in excess of recommended rates or during non-growing
seasons, excess ammonia is oxidized to nitrate, which can

* Corresponding author. Tel.: þ1 404 894 9367; fax: þ1 404 894 8266.
E-mail address: spyros.pavlostathis@ce.gatech.edu (S.G. Pavlostathis).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.019
5588 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8
then leach into the groundwater (Surampalli et al., 2008; US
EPA, 2000a).
A wastewater treatment plant located in the southeastern
United States, which services an area with a total population
of approximately 230,000, generates biosolids at approxi-
mately 5.4  103 dry metric tons/year. A portion of the
biosolids had been land applied to an agricultural field
adjacent to the wastewater treatment plant and served as
a source of nutrients for crop production. As a result of long-
term biosolids application, groundwater nitrate concentra-
tions in the upper aquifer of this site reached levels between
65 and 400 mg NO 3 eN/L, which are above the regulated limit
of 10 mg NO 3 eN/L (Maltais-Landry et al., 2009c; US EPA,
2009). Nitrate and nitrite concentrations higher than the
regulated limit present health concerns as they are toxic
to humans and livestock (US EPA, 2009). As a remediation
approach, the wastewater treatment facility proposed
groundwater pumping and development of an overland-flow
wetland system for the treatment of the nitrate-
contaminated groundwater before being released into the
nearby river.
>Wetland-based treatment systems are commonly used
for the biological removal of nitrogen, phosphorus, sulfur,
heavy metals and other pollutants, acting as a buffer between
the pollution source and the natural aquatic ecosystem
(Bachand and Horne, 2000; Kadlec and Wallace, 2009; Kjellin
et al., 2007; Maltais-Landry et al., 2009c). Among all the
nitrogen transformation processes taking place in wetlands,
the one related to nitrate removal is denitrification, i.e., the
reduction of nitrate to dinitrogen (N2). The effectiveness of
wetlands is largely affected by the biological activity and
temperature of the wetland location. Treatment wetlands are
often constructed in regions with moderate to cold climates
that experience large seasonal temperature variations (Kadlec
and Reddy, 2001).
Wetlands are complex biological systems and their
performance depends on a number of chemical, physical and
biological processes (US EPA, 2000b). For the biologically-
mediated nutrient removal in wetlands, both empirical and
mechanistic models have been used. Kadlec and Reddy (2001)
assessed the effect of temperature on treatment wetlands by
using a simple model where all reaction mechanisms are
grouped into a pseudo-first order removal rate. On the other
hand, Kjellin et al. (2007) used Menten and first-order kinetics
to model nitrate removal in wetland sediment. Denitrification
kinetics have also been modeled using both the single- and
dual-substrate Monod model (Hajaya et al., 2011; Heinen,
2006; Kornaros et al., 1996).
A three-cell pilot-scale, free-water surface wetland system
was constructed at the above-mentioned biosolids application
site to test the effectiveness of treating the nitrate-
contaminated groundwater using various carbon sources
during the system’s start-up period while vegetation was
being established. One cell served as a control (i.e., no external
carbon addition) and the other two cells used either MicroC
G or hay as carbon source, respectively. The pilot-scale
system was designed to demonstrate the efficiency of micro-
bial nitrate reduction under conditions open to the atmo-
sphere and as affected by various environmental and
operational conditions. In particular, the high groundwater
nitrate concentrations and the presence of oxygen in the free-
surface wetland system presented conditions which may
affect nitrogen removal efficiency.
In support of the pilot-scale constructed wetland
demonstration project, a continuous-flow, laboratory-scale
system was built to assess the treatment of nitrate-
contaminated groundwater as a function of carbon (i.e.,
electron donor) and nitrate loading, hydraulic retention time,
and temperature. The objective of the laboratory study was
to determine the feasibility of using a constructed wetland
for treatment of the nitrate-contaminated groundwater
through a series of batch and continuous-flow nitrate
reduction tests using MicroC G as the electron donor, as
well as soil and nitrate-contaminated groundwater from the
biosolids application site. Nitrate and nitrite reduction rate
estimates, resulting from the laboratory batch tests, were
used in a mathematical model to simulate the nitrogen
removal efficiency of continuous-flow, free-water surface
systems as a function of both operational and environmental
conditions.
2. Materials and methods
2.1. Sample collection and characterization
Groundwater and surface soil were collected at the biosolids
application site located in the southeastern United States. The
groundwater samples were stored in plastic containers under
refrigeration (4
C). The soil sample was passed through a US
No. 10 sieve, spread thin to air dry for 24 h at room tempera-
ture, and then stored in covered plastic containers at room
temperature (22e24
C). All samples were characterized by
measuring pH, soluble and total chemical oxygen demand
(sCOD and tCOD), dissolved organic carbon (DOC), moisture
content, NHþ  
4 eN, NO3 eN, NO2 eN, and other ions. To measure
soil pH, DOC, soluble COD, ammonia and ions, a soil filtrate
solution was prepared by adding 5 g of dry soil to 300 mL of
deionized (DI) water and mixing for 1 day at room tempera-
ture. The soil solution was then centrifuged at 10,000 rpm for
30 min. The results of soil and groundwater characterization
are shown in Table 1. Both the soil and groundwater samples
were slightly acidic, with pH values of 4.4 and 5.7, respectively.
The soil sample was mostly inorganic matter (w95%) and did
not contribute significant soluble COD or ions to the solution
(Table 1).
MicroC G, a plant-derived complex carbohydrate
mixture, was obtained from Environmental Operating Solu-
tions Inc. (Bourne, MA) and used as the electron donor and
carbon source in all experiments. A MicroC G solution was
prepared by a 1000-fold dilution in DI water and analyzed for
pH, DOC, soluble COD and ions. The concentrated MicroC G
stock and the diluted solution were stored in the dark at 22
and 4
C, respectively. The MicroC G solution was acidic
with a pH of 3.9 and the measured COD of the undiluted
solution was approximately 640 g/L, which agrees closely
with the technical specifications provided by the manufac-
turer (Table 1). MicroC G did not contain any anions or
ammonia, is soluble in water and has a freezing point of 8
C
(EOS, 2008).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8 5589

was conducted at room temperature (22e24
C) and lasted for

Table 1 e Characteristics of soil, groundwater and MicroC
G samples used in this study. a total of 121 days.
The second continuous-flow test, Run 2, was conducted to
Parameter Soil Groundwater MicroC
investigate the effect of the COD:N ratio on nitrate removal in
G
the system. Similarly to Run 1, site groundwater was fed with
pH 4.4 5.7 3.9 a nitrate concentration of 69 mg NO 3 eN/L. The MicroC G was
Water content (%) 3.9  0.1a
fed every 2 h at a flow rate dependent on the HRT and target
Dry weight (%) 96.1  0.1
COD:N ratios of 6, 5, 4, 3, 2, 1 and 0.5. Run 2 was conducted at
Organic matter (% of dry) 4.9  0.1
DOC (filtrate; mg C/L) 9.8  0.4 9.1  1.5 411  32c room temperature (22e24
C) and lasted for a total of 250 days.
Soluble COD (filtrate; 14.5  7 58.1  4.8 642  41c The last continuous-flow test, Run 3, was designed to
mg/L) investigate the effect of temperature on the nitrate reduction.
Total COD (mg/g dry 68.3  5.2 The reactor was housed in a controlled temperature room and
weight) its temperature was stepwise decreased from 22 to 5
C. The
Ions (filtrate)
rate of temperature change between the four target temper-
Chloride (mg Cl/L) NDb 14.4  0.8 ND
ature values was 2
C/day. The system performance was
Nitrite (mg N/L) ND ND ND
Nitrate (mg N/L) 0.3 69.3  1.3 ND assessed at four temperature values: 5, 10, 15 and 22
C. Site
Sulfate (mg S/L) 0.2 28.4  0.1 ND groundwater and MicroC G were fed at flow rates dependent
Phosphate (mg P/L) 1.3 ND ND on the HRT and to maintain a COD:N ratio of 6. At 5
C, the
Ammonia (filtrate; ND ND ND MicroC G had low solubility and a mixer was installed to
mg N/L) incorporate the feed into the groundwater (see Section 3.1.3
a Mean  standard deviation (n ¼ 3). for more details). Run 3 lasted for a total of 220 days.
b ND, not detected. In all continuous-flow runs, overflow reactor effluent was
c 1000-fold diluted solution. periodically collected and analyzed for nitrate, nitrite,
ammonia, pH, DOC, and soluble COD.

2.2. Continuous-flow laboratory-scale system 2.3. Batch assays

Three continuous-flow, laboratory-scale reactors were con-
structed in order to quantify nitrate removal under different
operational conditions including, effect of initial nitrate
concentration (Run 1), COD:N ratio (Run 2) and temperature
(Run 3). Three 15-L cubic Plexiglas reactors were filled with
10.5 kg of soil and approximately 9 L of nitrate-bearing
groundwater and were kept static for 1 day in order to expel
all air from the soil and uniformly wet the soil. The water
column depth was approximately 12 cm. Run 1 was conducted
in a single-compartment continuous-flow reactor; however, in
order to more closely simulate the flow regime of a full-scale
wetland system, two baffles were inserted into the reactors
for Run 2 and 3, thus dividing the liquid volume into three,
equal-volume compartments, resulting in a flow regime that
simulated 1.5e2 continuous-flow stir tank reactors (CSTRs) in
series (see Supplementary Material, Text S1 and Fig. S1).
Plastic reservoirs filled with groundwater were attached to
peristaltic pumps (Masterflex; Cole-Parmer) and the nitrate-
bearing groundwater was fed to the reactors continuously at
a specific flow rate to achieve the target hydraulic retention
times (HRTs). MicroC G was used as the electron donor and
carbon source and a 200 g COD/L diluted solution was fed
using a positive displacement pump (Fluid-Metering, Inc.) at
predetermined flow rates to achieve the target COD:N ratios.
The first continuous-flow test, Run 1, was designed to
investigate the effect of HRT and initial nitrate concentration
on system nitrate removal. Site groundwater, with a nitrate
concentration of 69 mg NO 3 eN/L, was fed to the system and
when the influent nitrate concentration was increased to
150 NO 3 eN/L, the site groundwater was amended with
a stock solution of NaNO3. The MicroC G was fed every 2 h
with the help of an electronic timer (ChronTrol) at a flow rate
dependent on the HRT to maintain a COD:N ratio of 6. Run 1
A 15-L cubic Plexiglas reactor was used in all batch assays
open to the atmosphere, filled with 10.5 kg of soil and
approximately 9 L of nitrate-bearing groundwater. Batch
assays were performed to investigate the effect of initial
nitrate concentration and temperature on denitrification
kinetics. To investigate the effect of initial nitrate concen-
tration, the site groundwater, containing approximately
69 mg NO 3 eN/L, was used, and in selected batch assays was
amended with a volume of a NaNO3 stock solution to achieve
initial concentrations of 150, 300, and 400 mg NO 3 eN/L. This
batch assay was conducted at room temperature (22e24
C).
To investigate the effect of temperature on nitrate reduction,
the site groundwater was amended with a volume of NaNO3
stock solution to achieve an initial concentration of
150 mg NO 3 eN/L at all temperature values tested: 5, 10, 15 and
22
C. The reactor was housed in a controlled temperature room
and its temperature decreased stepwise from 22 to 5
C. The rate
of temperature change between the four target temperature
values was 2
C/day. This batch assay was conducted simulta-
neously with the above-described continuous-flow Run 3.
MicroC G was used as the electron donor and carbon source
in all batch assays at a COD:N ratio of 6. In order to maintain
similar initial biomass concentrations in the soil layer, after
each batch assay was complete, the reactor was drained, back-
flushed with deionized water three times and nitrate-bearing
groundwater once before being refilled with nitrate-bearing
groundwater for the next batch assay. After the first batch
assay, which assessed the effect of initial nitrate concentration
on nitrate reduction, was completed at room temperature,
approximately 1 inch of the top soil layer was replaced with
fresh site soil and the reactor was transferred to the controlled
temperature room to assess the effect of temperature on
denitrification kinetics. Similarly to the first batch assay, the
5590 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

reactor was drained and back-flushed in between each batch
assay at the various target temperatures. In all open batch
assays, liquid samples were taken daily and nitrate, nitrite, pH
and periodic COD concentrations were measured.
In order to assess the effect of oxygen on the denitrification
kinetic rates measured in the open to the atmosphere reactor,
a closed batch assay was conducted using duplicate 160-mL
serum bottles sealed with rubber stoppers and aluminum
crimps and flushed with helium. In each serum bottle, 5 g dry
soil and 100 mL nitrate-containing groundwater were added.
A volume of different NaNO3 stock solutions were added to
each serum bottle resulting in initial nitrate concentrations
ranging from 70 to 400 mg NO 3 eN/L. Different volumes of
a MicroC G stock solution were added to each bottle resulting
in a COD:N ratio of 6. Incubation was carried out at room
temperature (22e24
C). During the incubation period, the
following parameters were measured: nitrate, nitrite, gas
production and gas composition (CO2, NO, N2O and N2).
2.4. Batch denitrification kinetics and parameter
estimation
experimental data, an initial biomass concentration was
chosen to fit the nitrate experimental data. Typical KS values
for denitrification have been reported in the range of
0.8e153 mg N/L (Kjellin et al., 2007; Tugtas and Pavlostathis,
2007; Zumft, 1997). The half-saturation constants for nitrate
and nitrite (KSNO3 and KSNO2 ) were estimated for each batch
assay based on reported value ranges to best fit the experi-
mental data. Parameter estimation was conducted following
a previously reported procedure (Hajaya et al., 2011).
Parameters sensitivity and identifiability analysis was per-
formed using the Berkeley Madonna Software Version 8.3
(Macey and Oster, 2006) and Matlab ode15 solver (MATLAB
7.0.1; The Mathworks, Natick, MA) following the procedure
described by Gujer (2008). The Fit ODE toolbox in Igor
Professional v.5.057 (WaveMetrics, Inc., Lake Oswego, OR)
was used to calculate the standard deviation values for the
evaluated parameters.
The effect of temperature on the MSRR was quantified by
fitting the resulting MSRR values at each temperature to the
modified Arrhenius model using nonlinear regression (Sig-
maPlot, Version 10.0 software; Systat Software Inc., San Jose,
CA, USA):

For this work, a two-step denitrification model (nitrate to kT2 ¼ kT1 qðT2 T1 Þ (4)
nitrite to dinitrogen) was used. Monod kinetic equations were T2 T1
where k and k are MSRR values (mg N/mg VSSCOD-day) at
used to describe microbial growth utilizing nitrate and nitrite
two different temperatures (
C) and q is the dimensionless
in all batch assays. Assuming that nitrate and nitrite are the
temperature coefficient.
limiting substrates (electron donor in excess with a COD/N
ratio of 6), the following differential equations were used:
2.5. Continuous-flow model
!
dSNO3 kNO3 SNO3
¼ XNOx (1) The denitrification kinetic rates estimated in the batch assays
dt KSNO3 þ SNO3
can be used to simulate and predict the performance of the
! ! continuous-flow wetland system under various operational
dSNO2 kNO3 SNO3 kNO2 SNO2 conditions. Based on a modified version of the dual-substrate
¼ XNOx  XNOx (2)
dt KSNO3 þ SNO3 KSNO2 þ SNO2 Monod model presented by Kornaros et al. (1996) and a system
mass balance, the continuous-flow system was modeled
! !
dXNOx YNO3 kNO3 SNO3 YNO2 kNO2 SNO2 using the series of differential equations (5) through (8) for
¼ XNOx þ XNOx  bXNOx
dt KSNO3 þ SNO3 KSNO2 þ SNO2 nitrate and nitrite reduction, cell growth and electron donor
utilization as follows:
(3)
  ! 
where SNO3 , SNO2 , and XNOx are nitrate, nitrite and denitrifiers dSNO3 SNO3 ;o  SNO3 kNO3 SNO3 C
¼  XNOx (5)
concentrations (mg NO 
3 eN/L, mg NO2 eN/L and mg VSSCOD/L, dt s kSNO3 þ SNO3 KC þ C
respectively); t is time (days); kNO3 and kNO2 are the nitrate and
nitrite maximum specific reduction rates (MSRR; mg N/   ! 
dSNO2 SNO2 ;o  SNO2 kNO3 SNO3 C
gVSSCOD day); KSNO3 and KSNO2 are the nitrate and nitrite half- ¼  XNOx
dt s kSNO3 þ SNO3 KC þ C
saturation constants (mg N/L); YNO3 and YNO2 are the theoret- ! 
ical yield coefficients (g VSSCOD/mg N); and b is the microbial kNO2 SNO2 C
 XNOx (6)
decay coefficient (day1). KSNO2 þ SNO2 KC þ C
Based on bioenergetic calculations, the yield coefficients
! 
for nitrate (YNO3 ) and nitrite (YNO2 ) used for all simulations dC ðCo  CÞ YNO3 kNO3 SNO3 C
were calculated to be 1.14 and 1.72 g VSSCOD/g N, respectively ¼  XNOx
dt s kSNO3 þ SNO3 KC þ C
(Rittmann and McCarty, 2001). The decay rate values for ! 
denitrifiers are generally in the range of 0.05e0.15 day1 YNO2 kNO2 SNO2 C
 XNOx (7)
KSNO2 þ SNO2 KC þ C
(Rittmann and McCarty, 2001; Tchobanoglous et al., 2003). A
microorganism decay rate of 0.1 day1 was chosen for all
! 
simulations; however, preliminary simulations using values dXNOx ðXNOx ;o  XNOx Þ YNO3 kNO3 SNO3 C
¼ þ XNOx
of 0.05 and 0.15 day1 resulted in small variations in nitrate dt s kSNO3 þ SNO3 KC þ C
concentration patterns. ! 
YNO2 kNO2 SNO2 C
Given the fact that the initial, active denitrifiers concen- þ XNOx  bXNOx (8)
KSNO2 þ SNO2 KC þ C
tration in the soil (XNOx ) was not measurable, for each set of
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8 5591

where SNO3 ;o , SNO2 ;o , and XNOx ;o are influent nitrate, nitrite and
biomass concentrations (mg N/L, mg N/L and mg VSSCOD/L,
respectively); Co and C are the electron donor (MicroC G)
concentrations in the influent and effluent (mg COD/L); KC
is the half-saturation constant for the electron donor
(mg COD/L); and s is the HRT (¼V/Q) (days). The KC value used
in all simulations was 20 mg COD/L, a value reported for
MicroC G as the electron donor for denitrification (Cherchi
et al., 2009).
The effluent nitrate and nitrite concentrations were
simulated in Matlab at various HRT, COD:N and temperature
values using the previously estimated biokinetic constants
(kNO3 , kNO2 , KSNO3 , KSNO2 , YNO3 , YNO2 , and b) and initial biomass
concentrations based on the denitrification kinetics described
in Section 2.4, above.
2.6. Analytical methods
DOC, COD, pH, dissolved oxygen (DO), water content, NHþ 4 eN,
NO 
3 eN, NO2 eN, and other ions were measured following
procedures outlined in Standard Methods (APHA, 2005). Soluble
COD was measured by the HACH colorimetric method and the
total COD was measured by the Open Reflux Method. DOC was
measured using a Shimadzu Total Organic Carbon (TOC)
Analyzer equipped with an infrared detector for CO2
measurement (Shimadzu Scientific Instruments, Inc.,
Columbia, MD). DO was measured using the polarographic
method with a YSI Model 58 oxygen meter in conjunction with
a YSI 5750 oxygen probe (Yellow Springs Instruments, Yellow
Springs, OH). Anions were measured using a Dionex DX-100
ion chromatography unit (Dionex Corporation, Sunnyvale,
CA) equipped with a conductivity detector, a Dionex IonPac
AG14A (4  50 mm) precolumn, and a Dionex IonPac AS14A
(4  250 mm) analytical column. The unit was operated in
autosuppression mode with 1 mM NaHCO3/8 mM Na2CO3
The effluent pH ranged from 6.5 to 7.5 and the DO remained
below 1 mg/L after addition of MicroC G. Ammonia was not
detected in any effluent samples. In all reactors, the effluent
COD and DOC remained constant and low (DOC below 40 mg/L
and COD below 75 mg/L). Sulfate was also periodically
measured and was consistently in the range of 20e30 mg S/L
in both influent and effluent samples, indicating that signifi-
cant sulfate reduction was not occurring in the open to the
atmosphere reactors.
3.1.1. Effect of HRT and influent nitrate concentration
Run 1 was performed in a single-compartment, continuous-
flow reactor which was operated at HRT values of 2.8, 3.5 and 5
days, while being continuously fed with groundwater at
67 mg NO 3 eN/L and MicroC G at a COD:N of 6 after the first
week, during which external carbon was not added. The
effluent nitrate concentration over the entire run period is
shown in Fig. 1, along with other operational parameters.
Upon addition of MicroC G directly to the reactor on day 8,
the effluent nitrate concentration decreased and reached
non-detectable levels within 3 days. For the remainder of
this run, the effluent nitrate concentration did not exceed
20 mg NO 3 eN/L at any of the three hydraulic retention times
tested. To illustrate that the nitrate removal follows
Monod kinetics, according to which the effluent nitrate
concentration is not a function of influent nitrate concentra-
tion, the influent groundwater concentration was increased
on day 88e130 mg NO 3 eN/L and the concentration of the
MicroC G solution changed accordingly to maintain a COD:N
ratio of 6. The effluent nitrate concentration increased slightly
to approximately 7 mg NO 3 eN/L until steady-state was
NITRATE (mg N/L)

150
eluent at a flow rate of 1 mL/min. The minimum detection 125
INFLUENT

limit for nitrate and nitrite was 0.05 and 0.1 mg N/L, 100
75
respectively. 50
25
Gas composition was determined by a gas chromatography 0
(GC) unit (Agilent Technologies, Model 6890N; Agilent Tech- 6
HRT (Days)

nologies, Inc., Palo Alto, CA) equipped with two columns and
4
two thermal conductivity detectors. Dinitrogen (N2) was
separated with a 15 m HP-Molesieve fused silica, 0.53 mm i.d. 2
column (Agilent Technologies, Inc.). Carbon dioxide (CO2), 0
nitric oxide (NO) and nitrous oxide (N2O) were separated with 100
EFFLUENT NITRATE (mg N/L)

a 25 m Chrompac PoraPLOT Q fused silica, 0.53 mm i.d. 90

80
column (Varian, Inc., Palo Alto, CA). Helium was used as the
70
carrier gas at a constant flow rate of 6 mL/min. The 10:1 split 60
injector was maintained at 150
C, the oven was set at 40
C 50
and the detector temperature was set at 150
C. The minimum 40
30
detection limit for CO2, NO, N2O and N2 was 800, 500, 7 and 20
50 ppmv, respectively. 10
0
0 10 20 30 40 50 60 70 80 90 100 110 120 130
TIME (Days)
3. Results and discussion
Fig. 1 e Effluent nitrate concentration in a continuous-flow
3.1. Continuous-flow reactor system performance reactor operated at room temperature (22e24
C) at various
HRT values, influent nitrate concentrations and with
The effluent streams from the three continuous-flow reactors MicroC G at a COD:N ratio of 6:1 at/after day 8 (MicroC G
were periodically analyzed for pH, ammonia, COD and DOC. was not added from day 92 to 104).
5592
achieved, after which the effluent concentration quickly
returned to non-detectable levels (Fig. 1).
In order to qualitatively evaluate the effect of carbon
source on the nitrate removal in this reactor, at approximately
92 days, the MicroC G pump was turned off and only the
groundwater at approximately 130 mg NO 3 eN/L was fed to
the reactor at an HRT of 5 days. As shown in Fig. 1, the effluent
nitrate concentration increased and reached about
88 mg NO 3 eN/L within 10 days, further demonstrating the
necessity of a continuous addition of a degradable carbon
source. On day 104, the groundwater pump was turned off to
simulate a batch system while the MicroC G pump continued
supplying carbon. A gradual decrease of the nitrate concen-
tration to about 20 mg NO 3 eN/L in 20 days was observed.
Thus, even at an elevated influent nitrate concentration,
continuous addition of biodegradable organic carbon in
excess of stoichiometric levels achieved high nitrate removal
efficiency.
3.1.2. Effect of COD:N ratio
Run 2 was conducted in a three-compartment, continuous-
flow reactor with an influent groundwater nitrate concentra-
tion kept constant at 70 mg NO 3 eN/L, an HRT of 2 and then 5
days, while the COD:N ratio was stepwise decreased to lower
values. Fig. 2 shows the reactor effluent nitrate concentration
along with other operational parameters. For the first 18 days,
the reactor was operated at an HRT of 2 days and a COD:N ratio
of 6, during which the effluent nitrate concentration
decreased sharply to less than 10 mg NO 3 eN/L. The HRT was
then increased to 5 days to achieve more stable operation and
consistent effluent nitrate concentration of less than
10 mg NO 3 eN/L. The COD:N ratio was stepwise decreased
from an initial value of 6:1 to the lowest value of 0.5:1. On day
6 biweekly from the soil/water interface by using an inverted
HRT (Days)
4
2 collected in the vial by water displacement. Then, the vial was
0 sealed with a stopper while under water, positioned upright
6 and its headspace analyzed by gas chromatography after
COD:N
4
2
0 nitrogen gas (N2) as the main nitrate reduction process in the
EFFLUENT NITRATE (mg N/L)
70
60
50
40
30
20
10
0
0 20 40 60 80 100 120 140 160 180 200 220 240 260
TIME (Days)
Fig. 2 e Effluent nitrate concentration in a continuous-flow
reactor operated at room temperature (22e24
C), mean
influent groundwater nitrate concentration of 70 mg N/L
and with MicroC G at several COD:N ratios.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8
25, the COD:N ratio was decreased to 5:1, during which the
effluent nitrate concentration remained below 4 mg NO 3 eN/L.
When on day 75, the COD:N ratio was further decreased to 4:1,
the effluent nitrate concentration increased sharply to about
27 mg NO 3 eN/L. A decrease of the COD:N ratio to 3:1 and
then to 2:1 resulted in an effluent nitrate concentration
ranging between 32 and 40 mg NO 3 eN/L. A further decrease
of the COD:N ratio to 1:1 and then to 0.5:1 resulted in a grad-
ual increase of the effluent nitrate concentration to
42 mg NO 3 eN/L. On day 232, the COD:N ratio was increased to
5:1, which resulted in a rapid decrease of the effluent nitrate
concentration to below 4 mg NO 3 eN/L. Based on these results,
for a system open to the atmosphere at ambient temperature
between 22 and 24
C, influent nitrate concentration of
67 mg NO 3 eN/L, and an HRT value of 5 days, the minimum
COD:N ratio is approximately 5:1 in order to achieve an
effluent nitrate concentration of less than 10 mg N/L.
The theoretical requirement for complete denitrification,
ignoring microbial growth, is 2.85 mg COD/mg nitrateeN
reduced to N2. At relatively low COD:N values, incomplete
denitrification is possible, which could lead to the formation
of nitric oxide (NO) and nitrous oxide (N2O), both potent
greenhouse gases (Maltais-Landry et al., 2009a, 2009b). It has
been reported that N2O emissions in wetlands are highly
dependent on the COD:N ratio, as well as the pH, dissolved
oxygen, and temperature among other parameters (Inamori
et al., 2008; Wu et al., 2009). Wu et al. (2009) found that
significant amounts of N2O were released from constructed
wetlands at very high and very low COD:N ratios
(2 > COD:N > 10), with minimum emissions at a ratio of 5:1. In
the present study, in order to investigate if NO and N2O were
released in the laboratory reactor due to incomplete denitri-
fication, on days 104, 143, 192 and 215 when the continuous-
flow reactor was operated with a COD:N ratio of 3:1, 2:1, 1:1
and 0.5:1, respectively, gas bubbles and water were collected
glass vial fully submerged in the water and partially imbedded
into the soil. Gas bubbles released from the surface soil were
30 min equilibration at room temperature. NO and N2O were
not detected at any of the COD:N ratios tested, confirming that
complete denitrification occurred, leading to the production of
laboratory reactor.
3.1.3. Effect of temperature
The three-compartment, continuous-flow reactor was housed
and operated in a temperature-controlled room to simulate
the effect of temperature on nitrate reduction. For the first 15
days, the continuous-flow reactor was operated at 22
C with
an HRT of 2 days, during which the effluent nitrate concen-
tration decreased sharply to less than 13 mg NO 3 eN/L (Fig. 3).
The HRT was then increased to 5 days to achieve more stable
operation and an effluent nitrate concentration of less than
10 mg NO 3 eN/L. On day 30 the room temperature was
decreased and by day 32 reached 15
C. While at 15
C, the
reactor performance did not change and the effluent nitrate
concentration was kept at non-detectable levels. On day 51,

25
TEMP. ( C)
20
o
15
10
5
0 temperature values above 5
C. Based on the experimental
12
HRT (Days)
10
8 tration of 10 mg N/L or less can be achieved even at a water
6 temperature as low as 5
C, as long as sufficient degradable
4
2
0 groundwater.
70
EFFLUENT NITRATE (mg N/L)
60
50
40
30
20
10
0 temperature on the wetland performance of the pilot-scale
0 20 40 60 80 100
TIME (Days)
Fig. 3 e Effluent nitrate concentration in a continuous-flow
reactor operated at a range of temperature (22e5
C), mean
influent groundwater nitrate concentration of 70 mg N/L
and with MicroC G at a COD:N ratio of 6:1 (arrow indicates
start of complete MicroC G and groundwater mixing; see
text).
the room temperature was decreased again and reached 10
C
by day 55. There was a slight increase in the effluent nitrate
concentration at this time, but within 24 h it returned to non-
detectable levels.
After the room temperature was reduced to 5
C by day 82,
the reactor effluent concentration increased rapidly to
a maximum 47 mg NO 3 eN/L, and then started to decrease.
However, for over 35 days at 5
C, the reactor’s performance
was not stable and the effluent nitrate concentration fluctu-
ated between 10 and 35 mg NO 3 eN/L, albeit with a down-
wards trend (Fig. 3). In an attempt to achieve a stable effluent
concentration, on day 120 the HRT was increased to 10 days.
Although the effluent nitrate concentration decreased signif-
icantly at an HRT of 10 days, it continued to fluctuate between
5 and 20 mg NO 3 eN/L. Upon further observation, it was real-
ized that MicroC G was not well mixed with the groundwater
in the reactor as it was delivered intermittently by a micro
pump every 2 h at the point where the groundwater was
constantly pumped into the reactor (head of reactor). It
appears that MicroC G has a low solubility at 5
C. Therefore,
the unstable and poor performance of the reactor was attrib-
uted to lack of uniform electron donor distribution and thus
availability. On day 172, a mixer was installed in the influent
portion of the reactor and turned on by an electronic timer
every 2 h while the MicroC G was fed, and for an additional
10 min after feeding was stopped. With intermittent mixing,
even at 5
C, MicroC G was well incorporated into the reactor
groundwater, which resulted in stable reactor performance
with non-detectable effluent nitrate concentrations. On day
190, the HRT was returned to 5 days and the reactor operated
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8 5593
for approximately another 25 days, during which period the
effluent nitrate concentration remained below 2 mg N/L
(Fig. 3). During this period at 5
C, biofilm formation on the
reactor walls was more noticeable than during operation at
results of the temperature study, an effluent nitrate concen-
carbon is provided and well incorporated into the
It is noteworthy that the conditions used in this test
differed from what was typically observed at the biosolids
application site where the groundwater temperature did not
change significantly throughout the year (it ranged between
18 and 20
C). In addition, during winter with ambient air
temperature between 4 and 15
C, the temperature at the
sediment/water interface in the pilot-scale wetland ranged
from 5 to 20
C. Therefore, even during winter, the impact of
120 140 160 180 200 220 wetland at the biosolids application site, where the lowest
average monthly air temperature, usually in January, is about
8e9
C, is expected to be less drastic. Therefore, the results of
the laboratory study at a water temperature as low as 5
C are
conservative, but show the resilience and efficiency of the
denitrification process at low temperature values.
3.2. Denitrification kinetics
3.2.1. Effect of initial nitrate concentration under open and
closed conditions
Denitrification at initial nitrate concentrations of 70, 140, 300
and 400 mg N/L was tested under conditions open to the
atmosphere. The nitrate and nitrite concentrations in each
assay over the incubation period are shown in Fig. 4. A lag
period of approximately 1 day was observed in the first batch
assay performed with site soil and groundwater, which was
attributed to the very low active denitrifying population size of
the surface soil used in these assays. Nitrate reduction pro-
ceeded immediately in all subsequent assays, indicating that
some active biomass was retained in the soil despite the
rinsing procedure performed in between each batch assay.
Transient nitrite concentrations were observed in all assays,
but after the complete removal of nitrate, nitrite was
completely removed in less than 4 days, except in the first
assay, in which nitrite reduction was slower and nitrite was
removed in approximately 5 days.
Similarly to the open batch assays, closed batch assays
were conducted in serum bottles in the absence of oxygen
(data not shown). A relatively low nitrate removal rate was
also observed in the first 20 h of incubation, which is attrib-
uted to the low population size of active denitrifying bacteria
in the soil. Significant nitrite levels were observed in series
with an initial nitrate concentration of 140 mg NO 3 eN/L and
above. Nevertheless, the nitrite reduction rate was fast and all
series achieved complete denitrification in less than 5 days.
Using the nitrate and nitrite reduction data and assuming
a two-step model (nitrate to nitrite to dinitrogen), the MSRR
values were estimated. Based on the data from the open batch
assays and applying Monod kinetics with the biokinetic
parameter values described in Section 2.4 above, the MSRR
5594 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8

values were determined to be very similar in all assays con-

80 ducted with different initial nitrate concentrations (Table 2).
A The best model fit to the experimental data is shown in Fig. 4.
Nitrate The root mean squared deviation (RMSD) at initial nitrate
60 concentration of 70, 140, 300 and 400 mg N/L was 7.0, 11.1, 36.4
Nitrite
and 44.8, respectively.
Sensitivity and identifiability analysis (Section 2.4, above)
40 showed that the highest degree of sensitivity was for the MSRR,
while the lowest degree of sensitivity was for the half-
saturation constant (see Supplementary Material, Fig. S2).
20 Although the sensitivity analysis indicates that the half-
saturation constant cannot be uniquely identified from the
provided data sets, the KSNO3 and KSNO2 values were estimated at
0
20.7  6.3 and 18.8  7.5 mg N/L for nitrate and nitrite, respec-
150
B tively. These values are apparent half-saturation constants, as
125 opposed to intrinsic, which take into consideration any mass
transfer limitations (Tugtas and Pavlostathis, 2007). Such
100 limitations are expected for a system where most of the
microbial activity occurred at the lower water layers, soil/water
75 interface and within the soil matrix. Visual inspection lead to
the observation that most N2 gas production was occurring in
50 the top soil layer, indicated by the entrapment of gas bubbles
and thick biofilm layer on the soil/water interface compared to
NITROGEN (mg N/L)

25 a relatively clear reactor water column. When estimating

denitrification kinetics for each initial nitrate concentration of
0
70, 140, 300 and 400 mg N/L, the initial biomass was assumed to
300 be 5, 10, 10 and 15 mg VSS/L, respectively, due to increased
C retention of active biomass in the sediment despite flushing of
250
the reactor in between each batch assay. Similarly to the open
200 assay, the MSRR values for the closed batch assays for nitrate
and nitrite were estimated to be 1.5  0.2 and 0.4  0.1 mg N/
150 mg VSSCOD-day, respectively. The KSNO3 and KSNO2 values were
16.1  2.8 and 20.5  6.2 mg N/L, respectively.
100 Oxygen affects (increases) the electron donor requirement
for denitrification; however, comparing the nitrate reduction
50 rate achieved in closed systems (1.5  0.2 mg NO 3 eN/
mg VSSCOD-day) to that achieved in open to the atmosphere
0 systems (1.2  0.2 mg NO 3 eN/mg VSSCOD-day) at an initial
400 COD:N ratio of 6, the kinetics of nitrate reduction were not
D severely affected in the open to the atmosphere reactor.
Therefore, as long as a bioavailable carbon source is supplied
300 in excess of that required for the complete nitrate reduction,
the nitrate reduction kinetics are not impacted by other
alternative electron acceptors (e.g., oxygen for open systems)
200
for similar systems (e.g., low mixing intensity and reaeration).

100
Table 2 e Estimated maximum specific reduction rate
(MSRR; k) and half-saturation constant (KS) values for
0 batch assays conducted with an open to the atmosphere
0 1 2 3 4 5 6 reactor and four initial nitrate concentrations (70, 140, 300
and 400 mg N/L).
TIME (Days)
Parameter Value
Fig. 4 e Measured (data points) and simulated (lines)
kNO3 (mg N/mg VSSCOD-day) 1.2  0.2a (1.0e1.5)b
nitrogen species in batch assays conducted with an open
KSNO3 (mg N/L) 20.7  6.3 (16.5e30.0)
to the atmosphere reactor at an initial nitrate kNO2 (mg N/mg VSSCOD-day) 0.7  0.1 (0.6e0.9)
concentration of (A) 70, (B) 140, (C) 300 and (D) 400 mg N/L KSNO2 (mg N/L) 18.8  7.5 (15.0e30.0)
using MicroC G as the carbon source at an initial COD:N
a Estimate  standard deviation.
ratio of 6:1.
b Range.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8 5595

The laboratory reactors were static and mixing was minimal.

A In both the continuous-flow and batch assays, most of the
microbial activity was in the soil/water interface and top soil
layer where dissolved oxygen concentrations were typically
less than 1 mg/L and maintained only by diffusion from the
free-water surface. Alternatively, under field conditions,
a higher rate of aeration is expected (e.g., wind action), which
may negatively impact the nitrate reduction rate as a result of
a higher competition for the carbon/electron source between
oxygen and nitrate-reducing processes.

3.2.2. Effect of temperature

In order to assess the effect of temperature on denitrification
kinetics, batch assays were performed at target temperatures
B of 5, 10, 15 and 22
C in a reactor open to the atmosphere
(Fig. 5). Similarly to the previous batch assays, a lag of
approximately 1 day was observed for nitrate removal. The
rates of both nitrate and nitrite reduction were very similar at
22 and 15
C, with a maximum transient nitrite concentration
slightly higher at 15
C. At 10
C, a significantly lower rate of
nitrate and nitrite reduction was observed; however, at 5
C,
the time required for the complete removal of nitrate and
nitrite was more than double of that under 10
C. The nitrate
and nitrite reduction rates at 5
C were much lower than at the
other three temperature values, and the maximum transient
nitrite concentration was the lowest. Therefore, the nitrate
reduction rates were not severely affected until the tempera-
C ture dropped below 10
C, which agrees with the findings of
other studies on denitrification at low temperature values
(Burgoon, 2001; Darbi and Viraraghavan, 2004; Hajaya et al.,
2011; Lee et al., 2009). Hajaya et al. (2011) reported a decrease
of more than 50% in the nitrate and nitrite reduction rates for
a temperature decrease from 15 to 10
C. Studies investigating
the seasonal temperature effects on denitrification in wetland
systems have reported significant decrease in denitrification
rates at temperature values below 15
C (Kadlec and Reddy,
2001; Poe et al., 2003).
Using the nitrate reduction data, estimated biokinetic
parameters and assuming a two-step denitrification model
as previously discussed in Section 2.4 above, the MSRR
values were estimated and are summarized in Table 3. The
half-saturation constants were similar at all temperature
D values; KSNO3 and KSNO2 were estimated as 57.0  4.8 and
16.5  2.4 mg N/L, respectively. As expected, the MSRR
increased with increasing temperature; however, the MSRR
values were very similar at 15 and 22
C, which agrees with
previous reports stating that the optimum temperature for
nitrate reduction is closer to 15 than 22
C, and that in this
temperature range, variations in temperature only slightly
affect denitrification rates (Hajaya et al., 2011; Kristiansen,
1983; Lee et al., 2009). The model accurately simulated the
nitrate reduction at all four temperature values; however, as
temperature decreased, the discrepancy between the
model and experimental nitrite concentrations increased.
At lower temperature values the model overestimated the

10, and (D) 5
C with groundwater at an initial nitrate

Fig. 5 e Measured (data points) and simulated (lines) concentration of 150 mg N/L using MicroC G as the carbon
nitrogen species in batch assays conducted with an open source at an initial COD:N ratio of 6:1 (note the different x-
to the atmosphere reactor maintained at (A) 22, (B) 15, (C) axis scale of panel D).
5596
Table 3 e Estimated maximum specific reduction rate
(MSRR; k) values for batch assays conducted with an open
to the atmosphere reactor maintained at different
temperatures.
Temperature (
C) kNO3
mg N/mg VSSCOD-day
5 0.41  0.01b
10 0.87  0.01
15 1.61  0.02
22 1.64  0.03
a Root mean square deviation
qXffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ ðmeasured value  estimated valueÞ2 :
b Estimate  standard deviation.
experimentally measured nitrite concentrations, with the
deviation increasing with decreasing temperature, indi-
cated by the increasing RMSD values (Table 3). The best
model fit for the four temperature batch assays is shown
in Fig. 5.
Based on the MSRR values at the four temperature values
tested and the modified Arrhenius model (equation (4)), the
dimensionless temperature coefficient (q) was estimated. By
taking 15
C as the basis, the temperature coefficient was 1.13
for nitrate reduction, a value which is within the temperature
coefficient range of 1.04 and 1.16 reported by Kadlec and
Reddy (2001) for removal of nitrate in wetlands. Cherchi
et al. (2009) reported a temperature coefficient of 1.11 for
nitrate reduction in a chemostat using MicroC G as the
carbon source. Kadlec and Wallace (2009) also reported
a mean temperature coefficient of 1.11 for nitrate reduction in
wetland systems with temperatures ranging from 6 to 24
C.
Temperature coefficient values of 1.07e1.14 were estimated
for bacteria in biofilms and artificially encapsulated at
temperatures as low as 3
C, indicating that the temperature
dependence of microbial activity is not affected in immobi-
lized bacteria (Vackova et al., 2011; Welander and Mattiasson,
2003). Thus, the q value found in the present study agrees well
with those previously reported.
3.3. Continuous-flow system simulation
Based on the estimated denitrification kinetic rates, the
continuous-flow wetland system was simulated using the
series of differential equations and parameters described in
Section 2.5, above. Model conditions were chosen based on the
results of both the batch and continuous-flow laboratory-
scale experiments. The obtained MSSR values indicated that
nitrate reduction rates were not affected by initial nitrate
concentration and trial simulations showed that the effluent
nitrate concentration at high COD:N ratios was not a function
of influent nitrate concentrations, agreeing with Monod
kinetics. As a result, for all simulations the value for SNO3 ;o was
kept constant at 150 mg N/L. Nitrite was not detected in the
site groundwater and the groundwater biomass was assumed
to be negligible (i.e., SNO2 ;o and XNOx ;o were assumed to be 0).
The influent electron donor concentration, Co, was adjusted to
simulate different COD:N ratios.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8
In addition to HRT, COD:N ratio and temperature, microbial
biomass retention is another variable parameter for a contin-
uous-flow system. Full-scale wetland systems are generally
rich with various types of biomass, such as trees, cattails,
kNO2 RMSDa algae and bacteria, most of which is retained within the
system, providing nutrients and serving as a carbon source to
drive biological processes and cell growth (Kadlec and
0.20  0.02 23.7 Wallace, 2009). Retention of the denitrifying biomass ulti-
0.31  0.03 19.7 mately controls the solids retention time (SRT) of the system,
0.50  0.04 26.1 which is usually very high and not easily controlled in full-
0.53  0.50 15.5
scale wetland systems. To simulate a wetland system with
a high degree of microbial biomass retention, the biomass
retention factor (b) was introduced into the system biomass
mass balance equation as follows:
! 
dXNOx ðXNOx ;o  bXNOx Þ YNO3 kNO3 SNO3 C
¼  XNOx
dt s kSNO3 þ SNO3 KC þ C
! 
YNO2 kNO2 SNO2 C
þ XNOx  bXNOx (9)
KSNO2 þ SNO2 KC þ C
The factor b takes values from 0 (i.e., 100% biomass retention)
to 1 (i.e., 0% biomass retention).
The pilot-scale wetland system at the biosolids application
site was open to the atmosphere and plant and vegetation
growth was significant in all cells, likely contributing signifi-
cant biodegradable carbon to the system and increasing
retention of the denitrifying biomass. Alternatively, the
laboratory-scale system was constructed with only soil and
groundwater, with MicroC G serving as the sole carbon
source in all laboratory assays. Although there was no vege-
tation, most of the biomass was retained in the laboratory-
scale continuous-flow reactors, with minimum amounts
released in the effluent streams. Because the fraction of
biomass retained could not be quantified accurately, a value of
75% biomass retention (i.e., b ¼ 0.25) was assumed for all
continuous-flow simulations. Using this biomass retention
value and an HRT of 5 days at 22
C, the effluent steady-state
nitrogen concentration (nitrate and nitrite) was predicted to
be approximately 5 mg N/L using model simulation. Alterna-
tively, assuming no biomass retention (i.e., b ¼ 1) and oper-
ating under the same conditions (HRT of 5 days at 22
C), the
effluent steady-state nitrogen concentration was predicted to
be approximately 13.5 mg N/L. The steady-state effluent
nitrate concentration in the laboratory reactor, Run 1, under
the same conditions ranged from 0 to 3 mg N/L, which is
slightly lower than the simulation results. Therefore, these
results verify that the assumption for biomass retention is
reasonable and that the model closely predicts the perfor-
mance of the laboratory continuous-flow system.
Using the continuous-flow design equations (equations (5)
through (7) and (9)), the system performance was evaluated as
a function of HRT, temperature and electron donor availability
(i.e., COD:N ratio). System performance is expressed as
nitrogen removal efficiency (%), which includes both nitrate
and nitrite in the effluent. Other possible denitrification
intermediates, NO and N2O, which are not included in the
model and were not detected in any of the continuous-flow
runs, are assumed to react rapidly and therefore cannot be
detected (Kornaros et al., 1996). Thus, the nitrogen removal
efficiency was evaluated using the following relationship:
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8 5597

NITROGEN REMOVAL (%)

0 20 40 60 80 100

A B C

REMOVAL (%)
100 100 100
80 80 80
60 60 60
40 40 40
20 20 20
06 06 06
5 15 5 15 5
CO 4 3 2 9 1113 CO 4 3 2
D:N 1 9 1113 CO 4 3 2 23
1720 )
D:N 1 5 7 s ) 5 7 ) D:N 1 1114
13 T (D
ay 13 (Day
s 5 8 MP. (°C
HR HRT TE

Fig. 6 e Model simulation of the effect of COD:N ratio and HRT on nitrogen removal in a continuous-flow system at (A) 5 and
(B) 15
C; (C) effect of COD:N ratio (1e6) and temperature (5e22
C) on nitrogen removal at an HRT of 5 days (75% denitrifying
biomass retention is assumed in all simulations).

 
SNO3 ;o  SNO3 þ SNO2 Based on results obtained with the laboratory-scale
Nitrogen Removal Efficiency ð%Þ ¼  100
SNO3 ;o continuous-flow system and model simulations, high
(10) nitrogen removal efficiencies (>90%) can be consistently
achieved and maintained by free-water surface wetland
Fig. 6A and B shows the effect of COD:N and HRT on nitrogen
systems while treating nitrate-contaminated groundwater
removal efficiency of the system at 5 and 15
C, respectively.
with high nitrate levels with an HRT of 3 days or higher and at
Assuming that 75% of biomass is retained (i.e., b ¼ 0.25),
temperature values as low as 5
C, as long as there is suffi-
changes in the HRT only slightly affect nitrogen removal
cient biodegradable carbon available to achieve complete
efficiency. Due to the high SRT, only at low HRT values, below
denitrification. Influent COD:N ratios should be maintained at
3 days, is the system performance impacted.
3:1 and higher. The results of this study show that con-
Similarly, Fig. 6C illustrates the system nitrogen removal
structed wetland technology is a technically feasible and
efficiency (%) as a function of COD:N ratio and temper-
attractive alternative for the treatment of nitrate-bearing
ature for values ranging from 1 to 6 and 5 to 15
C, respec-
groundwater at the biosolids application site.
tively. System performance is only slightly impacted at
a decreased temperature and is more severely impacted at
COD:N ratio values below 3. The experimental results of the
laboratory system and simulations indicate that denitrifi- Acknowledgments
cation can be successful in free-water surface wetland
systems at low temperature and HRT values as long as there This research was supported by the Columbus Water
is enough biodegradable carbon to achieve complete Works (CWW), Columbus, GA through Jordan, Jones and
denitrification. Goulding, Inc. (JJG), Norcross, GA. Special thanks to Camp,
Dresser and McKee, Inc. (CDM) for a graduate fellowship to
T. Misiti.
4. Conclusions

A laboratory-scale system was designed and developed to Appendix. Supplementary data

simulate a pilot-scale, free-water surface constructed wetland
system proposed for the treatment of nitrate-contaminated
groundwater at a biosolids application site. Although oxygen
can increase the electron donor requirement for denitrification,
this study showed that fast nitrate reduction rates can be
achieved even in systems open to the atmosphere, as long as
the electron/carbon source is not limiting. The kinetics of
nitrate reduction in open to the atmosphere reactors were not
severely affected by oxygen competition at initial nitrate
concentrations as high as 400 mg N/L. The rate of nitrate
reduction was not affected by nitrate concentrations as high as
400 mg N/L, but decreased with decreasing temperature;
however, even at temperature values as low as 5
C, complete
denitrification occurred in both batch and continuous-flow
systems as long as sufficient biodegradable carbon was
bioavailable (dissolved).
Supplementary data related to this article can be found online
at doi:10.1016/j.watres.2011.08.019.
references
American Public Health Association, 2005. Standard Methods for
the Examination of Water and Wastewater, 21st ed. APHA-
AWWA-WEF, Washington, D.C.
Bachand, P.A.M., Horne, A.J., 2000. Denitrification in constructed
free-water surface wetlands: I. Very high nitrate removal rates
in a macrocosm study. Ecological Engineering 14 (1e2), 9e15.
Burgoon, P.S., 2001. Denitrification in free water surface wetlands
receiving carbon supplements. Water Science and Technology
44 (11e12), 163e169.
5598 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 8 7 e5 5 9 8
Burkart, M.R., Stoner, J.D., 2002. Nitrate in aquifers beneath
agricultural systems. Water Science and Technology 45 (9), 19e28.
Cherchi, C., Onnis-Hayden, A., El-Shawabkeh, I., Gu, A.Z., 2009.
Implication of using different carbon sources for
denitrification in wastewater treatments. Water Environment
Research 81 (8), 788e799.
Darbi, A., Viraraghavan, T., 2004. Effect of low temperature on
denitrification. Fresenius Environmental Bulletin 13 (3B), 279e282.
Environmental Operating Solutions, 2008. Material Safety Data
Sheet e EOS MicroC GTM. http://www.eosenvironmental.com/
product/microc_g.htm (accessed 01.12.10). Bourne,
Massachusetts.
Gujer, W., 2008. Systems Analysis for Water Technology.
Springer, Berlin, Germany.
Hajaya, M.G., Tezel, U., Pavlostathis, S.G., 2011. Effect of
temperature and benzalkonium chloride on nitrate reduction.
Bioresource Technology 102 (8), 5039e5047.
Heinen, M., 2006. Simplified denitrification models: overview and
properties. Geoderma 133 (3e4), 444e463.
Inamori, R., Wang, Y., Yamamoto, T., Zhang, J., Kong, H., Xu, K.,
Inamori, Y., 2008. Seasonal effect on N2O formation in nitrification
in constructed wetlands. Chemosphere 73 (7), 1071e1077.
Kadlec, R.H., Reddy, K.R., 2001. Temperature effects in treatment
wetlands. Water Environment Research 73 (5), 543e557.
Kadlec, R.H., Wallace, S.D., 2009. Treatment Wetlands, second ed.
CRC-Press, Boca Raton, FL.
Kjellin, J., Hallin, S., Worman, A., 2007. Spatial variations in
denitrification activity in wetland sediments explained by
hydrology and denitrifying community structure. Water
Research 41 (20), 4710e4720.
Kornaros, M., Zafiri, C., Lyberatos, G., 1996. Kinetics of
denitrification by Pseudomonas denitrificans under growth
conditions limited by carbon and/or nitrate or nitrite. Water
Environment Research 68 (5), 934e945.
Kristiansen, S., 1983. The temperature optimum of the nitrate
reductase assay for marine-phytoplankton. Limnology and
Oceanography 28 (4), 776e780.
Lee, C.G., Fletcher, T.D., Sun, G.Z., 2009. Nitrogen removal in
constructed wetland systems. Engineering in Life Sciences 9
(1), 11e22.
Macey, R.I., Oster, G.F., 2006. Berkeley Madonna, 8.3. Berkeley
Madonna Inc.
Maltais-Landry, G., Maranger, R., Brisson, J., 2009a. Effect of
artificial aeration and macrophyte species on nitrogen cycling
and gas flux in constructed wetlands. Ecological Engineering
35 (2), 221e229.
Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009b.
Greenhouse gas production and efficiency of planted and
artificially aerated constructed wetlands. Environmental
Pollution 157 (3), 748e754.
Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009c.
Nitrogen transformations and retention in planted and
artificially aerated constructed wetlands. Water Research 43
(2), 535e545.
Murgulet, D., Tick, G.R., 2009. Assessing the extent and sources of
nitrate contamination in the aquifer system of southern
Baldwin County, Alabama. Environmental Geology 58 (5),
1051e1065.
Poe, A.C., Piehler, M.F., Thompson, S.P., Paerl, H.W., 2003.
Denitrification in a constructed wetland receiving agricultural
runoff. Wetlands 23 (4), 817e826.
Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology:
Principles and Applications. McGraw-Hill, New York, NY.
Rivett, M.O., Buss, S.R., Morgan, P., Smith, J.W.N., Bemment, C.D.,
2008. Nitrate attenuation in groundwater: a review of
biogeochemical controlling processes. Water Research 42 (16),
4215e4232.
Surampalli, R.Y., Lai, K.C.K., Banerji, S.K., Smith, J., Tyagi, R.D.,
Lohani, B.N., 2008. Long-term land application of biosolids e
a case study. Water Science and Technology 57 (3), 345e352.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater
Engineering: Treatment and Reuse, fourth ed. McGraw-Hill,
Boston.
Tugtas, A.E., Pavlostathis, S.G., 2007. Electron donor effect on
nitrate reduction pathway and kinetics in a mixed
methanogenic culture. Biotechnology and Bioengineering 98
(4), 756e763.
United States Environmental Protection Agency (US EPA), 2000a.
Biosolids Technology Fact Sheet e Land Application of
Biosolids. Office of Water, Washington, D.C.
United States Environmental Protection Agency (US EPA), 2000b.
Constructed Wetlands Treatment of Municipal Wastewaters.
Office of Research and Development, Cincinnati, Ohio.
United States Environmental Protection Agency (US EPA), 2009.
Drinking Water Contaminants. http://water.epa.gov/drink/
contaminants/index.cfm (accessed 31.05.11).
Vackova, L., Srb, M., Stloukal, R., Wanner, J., 2011. Comparison of
denitrification at low temperature using encapsulated
Paracoccus denitrificans, Pseudomonas fluorescens and mixed
culture. Bioresource Technology 102 (7), 4661e4666.
Welander, U., Mattiasson, B., 2003. Denitrification at low
temperatures using a suspended carrier biofilm process.
Water Research 37 (10), 2394e2398.
Wu, J., Zhang, J., Jia, W.L., Xie, H.J., Gu, R.R., Li, C., Gao, B.Y., 2009.
Impact of COD/N ratio on nitrous oxide emission from
microcosm wetlands and their performance in removing
nitrogen from wastewater. Bioresource Technology 100 (12),
2910e2917.
Zumft, W., 1997. Cell biology and molecular basis of denitrification.
Microbiology and Molecular Biology Reviews 61 (4), 533e616.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Diversity and antibiotic resistance of Aeromonas spp.

in drinking and waste water treatment plants

Vânia Figueira, Ivone Vaz-Moreira, Márcia Silva, Célia M. Manaia*

CBQF/Escola Superior de Biotecnologia, Universidade Católica Portuguesa, R. Dr. António Bernardino de Almeida, 4200-072 Porto, Portugal

article info abstract

Article history: The taxonomic diversity and antibiotic resistance phenotypes of aeromonads were
Received 14 June 2011 examined in samples from drinking and waste water treatment plants (surface, ground
Received in revised form and disinfected water in a drinking water treatment plant, and raw and treated waste
13 August 2011 water) and tap water. Bacteria identification and intra-species variation were determined
Accepted 13 August 2011 based on the analysis of the 16S rRNA, gyrB and cpn60 gene sequences. Resistance
Available online 22 August 2011 phenotypes were determined using the disc diffusion method.
Aeromonas veronii prevailed in raw surface water, Aeromonas hydrophyla in ozonated
Keywords: water, and Aeromonas media and Aeromonas puntacta in waste water. No aeromonads were
Aeromonas detected in ground water, after the chlorination tank or in tap water. Resistance to cefta-
Quinolone resistance zidime or meropenem was detected in isolates from the drinking water treatment plant
Surface water and waste water isolates were intrinsically resistant to nalidixic acid. Most of the times,
Waste water quinolone resistance was associated with the gyrA mutation in serine 83. The gene qnrS,
but not the genes qnrA, B, C, D or qepA, was detected in both surface and waste water
isolates. The gene aac(6’)-ib-cr was detected in different waste water strains isolated in the
presence of ciprofloxacin. Both quinolone resistance genes were detected only in the
species A. media. This is the first study tracking antimicrobial resistance in aeromonads in
drinking, tap and waste water and the importance of these bacteria as vectors of resistance
in aquatic environments is discussed.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction causative agents of fish disease (Janda and Abbott, 2010).

The genus Aeromonas comprises ubiquitous bacteria, consid-
ered indigenous to aquatic environments (Janda and Abbott,
2010). Members of this genus are able to inhabit surface
water (rivers, lakes), sewage, drinking water (tap and bottled
mineral), thermal waters and sea water (Biscardi et al., 2002;
Maalej et al., 2003; Pablos et al., 2009). Some species, mainly
the psychrophilic Aeromonas salmonicida and the mesophilic
Aeromonas hydrophila and Aeromonas veronii are recognized
Aeromonas spp. are also important human opportunistic
pathogens with ability to cause various types of diseases,
which include intestinal, blood, skin and soft tissue and
trauma-related infections (Aminov, 2009; Lamy et al., 2009;
Janda and Abbott, 2010). Among the leading pathogenic
species are A. hydrophila, Aeromonas caviae (later synonym of
Aeromonas punctata) and A. veronii (Lamy et al., 2009). The
environmental ubiquity associated with the potential patho-
genicity of these bacteria has been illustrated also in recent

* Corresponding author. Tel.: þ351 22 5580059; fax: þ351 22 5090351.

E-mail address: cmmanaia@esb.ucp.pt (C.M. Manaia).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.021
5600 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
natural disasters (Dixon, 2008). Evidences for the water-
human transmission of Aeromonas spp. are also available
(Khajanchi et al., 2010).
Over the last years, a greater public awareness and scien-
tific understanding of antimicrobial resistance contributed to
consider environmental reservoirs and paths of dissemina-
tion as critical points for antimicrobial resistance control.
Among such reservoirs and paths of dissemination, water
environments in which enter antibiotic resistant organisms
from human and animal sources play a pivotal role (Baquero
et al., 2008; Kümmerer, 2009; Taylor et al., 2011). The impor-
tance of municipal waste water treatment plants as sources of
antimicrobial resistance and the risks of contamination of
surface waters has been demonstrated in numerous publica-
tions (e.g. Gõni-Urriza et al., 2000; Ferreira da Silva et al., 2007;
Servais and Passerat, 2009; Novo and Manaia, 2010). As
a consequence of surface and ground water contamination,
emerges the hypothesis that antimicrobial resistance can
reach drinking water, serving as a vehicle of resistance
transfer to the water consumers. Indeed, antimicrobial resis-
tance has been detected in drinking water (Schwartz et al.,
2003; Faria et al., 2009; Xi et al., 2009; Vaz-Moreira et al.,
2011). Considering this urban water cycle, ubiquitous bacteria,
which can colonize different types of water, are of special
interest to assess potential forms of antimicrobial resistance
dissemination. Given their ubiquity and patterns of acquired
antimicrobial resistance, members of the genus Aeromonas are
good example of such bacteria.
In a recent comprehensive review on the genus Aeromonas,
Janda and Abbott (2010) alerted for the little attention given to
the general low susceptibility of aeromonads to various
classes and combinations of antimicrobial agents. Neverthe-
less, the potential of aeromonads to develop and disseminate
antibiotic resistance either in clinical settings or in the envi-
ronment has been demonstrated in numerous publications
(Walsh et al., 1997; Gõni-Urriza et al., 2000; Huddleston et al.,
2006; Blasco et al., 2008; Cattoir et al., 2008; Gordon et al.,
2008; Lamy et al., 2009; Arias et al., 2010a,b). Moreover,
recent and emerging antibiotic resistance seems to be
common in different species of Aeromonas. For instance,
different variants of the plasmid-mediated quinolone resis-
tance qnr gene were detected in environmental isolates of the
species A. punctata, Aeromonas media or Aeromonas allo-
saccharophila (Cattoir et al., 2008; Picão et al., 2008; Xia et al.
2010). In spite the ubiquity of aeromonads in aquatic envi-
ronments and the likelihood to develop antimicrobial resis-
tance, the ecology and patterns of resistance of members of
this genus present in drinking and waste water treatment
plants has not been addressed in scientific literature. The
current work aimed at filling this gap and was based on the
hypothesis that Aeromonas spp. could serve as a vehicle for
antibiotic resistance dissemination within the urban water
cycle. According to this hypothesis, this work was designed to
track aeromonads and their antibiotic resistance profiles in
different parts of the urban water cycle. Our main goal was the
identification of major sources of antibiotic resistant aero-
monads and of critical points for their elimination. Specifi-
cally, it was intended to i) identify the different aquatic
environments within the urban water cycle where aero-
monads are more prevalent, and possible factors contributing
for their reduction; ii) determine the predominant species in
each of those environments; iii) infer about the possible
relationship Aeromonas species-antibiotic resistance pattern.
2. Materials and methods
2.1. Water sampling
Water samples were collected from different aquatic envi-
ronments within an urban water cycle in the region of
Northern Portugal (Fig. 1). These sites included: i) a drinking
water treatment plant (WTP) and the respective distribution
system, which supplies water to a population of about 1.5
million of inhabitants; ii) household tap water, served by the
WTP and iii) a municipal waste water treatment plant (WWTP),
serving 85 000 inhabitant equivalents, in the same geograph-
ical area (Ferreira da Silva et al., 2006; Vaz-Moreira et al., 2011).
In the WTP, samples were collected from raw surface water
(A), ground water (alluvial wells) (B), after sand filtration and
ozonation (C) and after chlorination (D), the drinking water
final treatment step. Four samples were collected downstream
the WTP, in the drinking water distribution system respec-
tively, after the first or the second re-chlorination stations (E-
H). Tap water samples were collected in eleven houses (I-S),
from taps used 1e4 times a month, served by the WTP referred
to above and situated within an area of 25 km (Vaz-Moreira
et al., 2011). Waste water samples corresponded to raw (T)
and treated waste water (U). Sites A to H (drinking water
treatment plant and distribution system) were sampled in
November 2007 and in September 2009, sites I to S (taps) were
sampled in April, July and October 2009 and sites T and U
(waste water treatment plant) were sampled nine times
between November 2004 and November 2009.
2.2. Isolation, enumeration and preliminary
identification
This work was integrated in a wider study designed to assess
the diversity and antibiotic resistance of culturable bacteria,
belonging to different groups, present in selected niches
within the urban water cycle. For the microbiological char-
acterization of the water samples it was used the membrane
filtration method, as described before for waste and surface
water (Ferreira da Silva et al., 2006; Vaz-Moreira et al., 2011).
Volumes of 10e500 ml (WTP, drinking water distribution
system, taps) or of 1e10 ml (WWTP) of water samples or
decimal dilutions thereof were filtered through cellulose
nitrate membranes (0.45 mm pore size, 47 mm diameter,
Albet), which were placed onto different culture media and
incubated up to 7 days. No selective culture medium for aer-
omonads was used. Aeromonas spp. analysed in this study
were isolated among the culturable heterotrophs recovered on
different culture media - Plate Count Agar (PCA, Pronadisa), on
m-endo-agar-LES (Difco), on mFC agar (Difco), on Tergitol-7
agar (TTC, Oxoid), on Pseudomonas Isolation Agar (PIA,
Difco), on R2A agar (Difco) or on Bile Esculin Agar (BEA, Pro-
nadisa). The culture media PCA, PIA, BEA and R2A were
incubated at 30  C and mFC agar, m-endo-agar-LES and TTC
were incubated at 37  C. Given this work was designed to
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1 5601

B (101 CFU mL-1)

2
C (101-103 CFU mL-1)

3 O3 4 O3 5 6
O3
1 Cl 7 Cl
- -
A (103 CFU mL-1)

8 D (10-2-101 CFU mL-1)

WWTP

10 9 Water Distribution
E-H (10-2-102 CFU mL-1)
T (105-106 CFU mL-1)I-S System
1 4 -1
(10 -10 CFU mL )
11 U (104-105 CFU mL-1)

1 – Raw Surface Water 5 – Coagulation/Flocculation 9 – Drinking Water

2 – Ground Water 6 - Flotation and Filtration 10 – Raw Waste Water
3 – Filtration 7 – Chlorination 11 – Treated Waste Water
4 – Ozonation 8 – Treated Water Reservoir

Fig. 1 e Schematic representation of the drinking and waste water treatment plants, indicating the treatment stages and
sampled sites (AeU). Ground water disinfection involves only stages 6e7. CFU mLL1 corresponds to culturable heterotrophic
counts in the plates from which the aeromonads were isolated.

recover culturable bacteria from different bacterial groups,
representatives of all colony types were selected for further
culture isolation and purification, according to the following
criterion: all colonies of morphotypes represented by less than
five colonies, half of the colonies of morphotypes represented
by five to 10 colonies, and about one third of colonies with
morphotypes represented by more than 10 colonies. Cyto-
chrome c oxidase positive isolates with the morphology of
Gram-negative rods and forming yellow colonies on GSP agar
(Merck) at 30  C were presumptively identified as aeromonads
(Corry et al., 2003; Abulhamd, 2009; Kivanc et al., 2011). This
set of aeromonads included a total of 121 isolates, from which
72 and 8 were isolated, respectively, from the first and second
sampling campaigns in the WTP; 1, 3, 2, 6, 1, 1, 5 and 13 were
isolated, respectively, from each sampling date in the WWTP.
A group of 9 WWTP isolates recovered on PCA or mFC agar
supplemented with 4 mg L1 ciprofloxacin (AppliChem) (Novo
and Manaia, 2010) were also examined in this study.
2.3. Identification at the species level and determination
of intra-species variation
Identification at the species level was based on the analysis of
the 16S rRNA gene sequence and intra-species variation was
assessed on basis of the comparison of two additional
housekeeping genes, gyrB and cpn60 (Yáñez et al., 2003;
Miñana-Galbis et al., 2009). PCR amplifications of fragments
of the genes 16S rRNA, gyrB and cpn60 were conducted using
the primers and the conditions described before (Table 1). PCR
products were purified with GFX PCR DNA purification kit (GE
Healthcare) and the nucleotide sequences were determined.
Partial nucleotide sequences of the genes 16S rRNA, gyrB
and cpn60 were aligned using Clustal W from MEGA 4.0 soft-
ware (Tamura et al., 2007) and compared with the homologous
sequences of the type strains of the different Aeromonas
species, available in the GenBank database. For the gene cpn60,
the nucleotide sequences of the type strains of the species
Aeromonas sanarellii and Aeromonas taiwanensis were not avail-
able in the GenBank database, and were determined in this
study using the type strains provided by BCCM/LMG culture
collection with the numbers LMG 24682T and LMG 24683T,
respectively. These nucleotide sequences were deposited in
the GenBank database with the accession numbers JF920655
and JF920656 for A. sanarellii and A. taiwanensis, respectively.
Nucleotide sequence relatedness was estimated based on
the model of Jukes and Cantor (1969) and dendrograms were
created using the neighbour-joining method. The maximum-
likelihood method was also applied to assess tree stability. In
the analysis were used 1283, 779 and 555 nucleotide positions
of the 16S rRNA, gyrB and cpn60 gene sequences, respectively.
Non-repetitive nucleotide sequences were deposited in the
GenBank database with the accession numbers JF920473-
JF920563, JF938599-JF938689, and JF920564-JF920654 for 16S
rRNA, gyrB and cpn60 sequences, respectively.
In an attempt to discriminate strains of the same species,
each pair of isolates was compared based on the nucleotide
sequence of each of the three genes. Strains differing at least
in a nucleotide position were classified as representing
distinct sequence types (ST). This comparison was repre-
sented in a dendrogram constructed based on 2617 nucleotide
positions of the concatenated sequences of 16S rRNA, cpn60
and gyrB genes (Fig. 2).
2.4. Determination of antibiotic resistance phenotypes
The susceptibility to 12 antibiotics was determined using the
agar diffusion method (CLSI, 2007). The antibiotics tested were
nalidixic acid (NA, 30 mg); ciprofloxacin (CIP, 5 mg); amoxicillin
5602 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1

Table 1 e Primers and PCR conditions used.

Gene Primers Sequence Fragment length Annealing temp. Reference
(bp) ( C)

16S rRNA 27F GAGTTTGATCCTGGCTCAG 1465 55  C Lane, 1991

1492R TAC CTT GTT ACG ACT T
gyrB gyrB-3F TCCGGCGGTCTGCACGGCGT 1130 58  C Yáñez et al., 2003
gyrB-14R TTGTCCGGGTTGTACTCGTC
cpn60 C175 GAAATYGAACTGGAAGACAA 763 52  C Miñana-Galbis et al., 2009
C938 GTYGCTTTTTCCAGCTCCA
gyrA AsalgyrAF TCCTATCTTGATTACGCCATG 481 50  C Goñi-Urriza et al., 2002
AsalgyrAR CATGCCATACCTACCGCGAT
parC AcparCF GTTCAGCGCCGCATCATCTAC 245 54  C Goñi-Urriza et al., 2002
EcparCR TTCGGTGTAACGCATTGCCGC
qnrA qnr Am F AGAGGATTTCTCACGCCAGG 580 54  C Cattoir et al., 2007
qnrAm R TGCCAGGCACAGATCTTGAC
qnrB qnrBm F GGMATHGAAATTCGCCACTG 264 54  C Cattoir et al., 2007
qnrBm R TTTGCYGYYCGCCAGTCGAA
qnrS qnrSm F GCAAGTTCATTGAACAGGGT 428 54  C Cattoir et al., 2007
qnrSm R TCTAAACCGTCGAGTTCGGCG
qnrC qnrC-F GGGTTGTACATTTATTGAATC 447 50  C Wang et al., 2009
qnrC-R TCCACTTTACGAGGTTCT
qnrD qnrD-F CGAGATCAATTTACGGGGAATA 582 55  C Cavaco et al., 2009
qnrD-R AACAAGCTGAAGCGCCTG
aac(6’)-Ib aac(6)-F TTGCGATGCTCTATGAGTGGCTA 482 55  C Park et al., 2006
aac(6)-R CTCGAATGCCTGGCGTGTTT
qepA qepA-F TGGTCTACGCCATGGACCTCA 1137 56  C Périchon et al., 2007
qepA-R TGAATTCGGACACCGTCTCCG
cphA cphA-F TCTATTTCGGGGCCAAGGG 230 55  C Balsalobre et al., 2009
cphA-R TCTCGGCCCAGTCGCTCTTCA
blaTEM blaTEM fw ATAAAATTCTTGAAGACGAAA 939 55  C DiPersio et al., 2005
blaTEM rv GACAGTTACCAATGCTTAATCA

(AML, 25 mg); ticarcillin (TIC, 75 mg); cephalothin (CP, 30 mg);
ceftazidime (CEF, 30 mg); streptomycin (STR, 10 mg); sulpha-
methoxazole/trimethoprim (SXT, 25 mg); tetracycline (TET,
30 mg); gentamicin (GEN, 10 mg); colistin sulphate (CT, 50 mg)
and meropenem (MER, 10 mg). For the antibiotics AML and CT,
which are not included in the CLSI list, were used the
following criteria: S  21/R < 14 and S  10/R < 10, respectively.
Whenever diameters larger than R but smaller than S were
observed, were referred to as intermediary, and were excluded
from the resistance percentage calculations. The strains
Escherichia coli ATCC 25922 and Pseudomonas aeruginosa DSM
1117 (¼ATCC 27853) were included as quality controls.
(qnrA1 þ), Klebsiella pneumoniae B1 (qnrB1 þ) and Enterobacter
cloacae S1 (qnrB4þ and qnrS1þ) were used as positive controls
for the presence of the determinants qnrA, qnrB and qnrS.
The strains E. coli DH10B transformant pHS11 and E. coli DH10B
transformant p2007057 were used as positive controls for
qnrC and qnrD, respectively. Salmonella enterica serovar typhi-
murium GSS-HN-2007-003 was used as positive control (Xia
et al., 2009) for the presence of gene aac(6’)-Ib. Strain E. coli
TOP10 þ pAT851 was used as positive control for gene qepA.
PCR products were purified and the nucleotide sequences
were determined and compared. A representative of each
distinct nucleotide sequence was deposited in the GenBank
(JF938596-JF938598).

2.5. Screening of resistance genetic determinants

Mutations in the chromosomal genes gyrA and parC and the
presence of resistance genes qnrA, qnrB, qnrS, qnrC, qnrD,
aac(6’)-Ib and qepA were screened in quinolone resistant
isolates. The presence of genes cphA and blaTEM associated
with beta-lactam resistance were screened in all isolates. The
primers and PCR conditions used were described before
(Table 1).
Point mutations in the genes gyrA and parC were identified
after comparison with homologous nucleotide sequences of
quinolone susceptible strains available in the GenBank - A.
punctata CIP 7616T (AY027899 and AF435418), A. hydrophila
subsp. hydrophila CIP 7614T (AY027901 and AF435419) and
Aeromonas sobria CIP 7433T (AY027900 and AF435420) as
described before (Goñi-Urriza et al., 2002). Strains E. coli L0
2.6. Statistical analysis
The chi-squared test was used to compare the prevalence
values of antibiotic resistance phenotypes or genotypes and
sequence types in different water sampled sites (SPSS 19.0 for
Windows, SPSS Inc., Chicago, IL).
3. Results
3.1. Species diversity and intra-species variation
A collection of 741 Gram-negative cytochrome c oxidase-
positive isolates recovered on the culture media and condi-
tions described above was screened for the presence of
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1 5603

Aeromonas. On basis of 16S rRNA gene sequence analysis,

among these isolates, 121 (80 from the WTP, 32 from WWTP
and nine from WWTP isolated in the presence of cipro-
floxacin) were identified as belonging to 11 species of the
genus Aeromonas. The other strains were affiliated to genera
such as Pseudomonas, Ralstonia, Comamonas, Acidovorax, Bre-
vundimonas, Cupriavidus, Chryseobacterium, Achromobacter and
to the family Sphingomonadacae.
The group of Aeromonas spp. isolates comprised 80 from
the drinking water treatment plant (51 from raw surface
water - SC and 29 recovered after water ozonation - PO) and
32 from the waste water treatment plant (17 from raw - RWW
and 15 from treated waste water - TWW). Additionally, nine
aeromonads were isolated in the presence of ciprofloxacin
(five from RWW and four from TWW). Aeromonads were not
detected in ground water samples, neither downstream the
chlorination tank of the WTP, including in tap water (Table 2,
Fig. 1).
Aeromonas spp. identification to the species level (Table 2)
was supported by the 16S rRNA gene sequence analysis. Most
of the times (114/121), these identifications were the same as
those determined based on the analysis of the genes gyrB or
cpn60. Only for seven isolates (five from TWW and two from
SC) of the species Aeromonas aquariorum, A. punctata and
A. veronii, the genes gyrB or cpn60 would lead to a different
identification. Among the eleven species, eight were detected
in raw surface water (Fig. 1, site A) and only four after the

Fig. 2 e Neighbour-joining dendrogram based on 16S rRNA,

cpn60 and gyrB concatenated nucleotide sequences.
Bootstrap values (‡50%) generated from 1000 replicates are
indicated at branch points. Bold circles indicate branches
recovered by the maximum-likelihood method. Sequence
types (ST) and16S rRNA GenBank accession numbers are
indicated in parenthesis. GenBank accession numbers for
cpn60 and gyrB sequences are, respectively, for A.
allosaccharophila EU306795 and AY101777, A. aquariorum
FJ936120 and EU268444, A. enteropelogenes EU306837 and
EF465526, A. eucrenophila EU306803 and AY101776, A.
hydrophila subsp. hydrophila EU306804 and AY101778, A.
jandaei EU306807 and AY101780, A. media EU306808 and
AY101782, A. punctata EU306800 and AY101783, A.
sanarellii (JF920655) and FJ807277, A. taiwanensis (JF920656)
and FJ807272, A. veronii EU306839 and AY101795.* Strains
for which 16S rRNA based identification differed from that
given by the genes gyrB and cpn60. Strains designation:
Isolates from the drinking water treatment plant were
generically designated as SxMn, with S standing for site of
isolation (A, raw surface water; C, after ozonation), x for the
sampling date (2, second sampling date), M, for the culture
medium of isolation (F, mFC agar; T, Tergitol-7 agar; P,
Pseudomonas isolation agar; R, R2A agar; E, Bile Esculin
agar), and n for the number of the isolate. Isolates from the
raw or treated wastewater were generically designated as
AxMn or ExMn, respectively (A relative to RWW and E to
TWW); x for the sampling date; M for the culture medium
of isolation (P, PCA; PC, PCA with ciprofloxacin; EL, m-
endo-agar-LES; FC, mFC agar with ciprofloxacin); and n for
isolate number of the isolate.
5604 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1

Table 2 e Diversity and percentage (number of isolates, number of sequence types) of Aeromonas species in the different
types of water.
Species % (n) Raw surface After Ozonation Raw Waste water (22) Treated Waste
water (51) (29) water (19)

A. allosaccharophila (3) 2.0% (1, 1) e 9.1% (2, 2) e

A. aquariorum (5) e 6.9% (2, 1) e 15.8% (3, 3)
A. enteropelogenes (1) 2.0% (1, 1) e e e
A. eucrenophila (1) 2.0% (1, 1) e e e
A. hydrophila subsp. hydrophila (25) 13.7% (7, 7) 58.6% (17, 4) 4.5% (1, 1) e
A. jandaei (5) 3.9% (2, 2) 10.3% (3, 1) e e
A. media (25) 19.6% (10, 10) e 36.4% (8, 8) 36.8% (7, 7)
A. punctata (18) 7.8% (4, 4) e 36.4% (8, 6) 31.6% (6, 6)
A. sanarellii (4) e e 4.5% (1, 1) 15.8% (3, 2)
A. taiwanensis (1) e e 4.5% (1, 1) e
A. veronii (33) 49.0% (25, 22) 24.1% (7, 3) 4.5% (1, 1) e

No aeromonads were isolated from ground water or from any sampling point after water chlorination, including in 11 household taps.

ozonation process (Fig. 1, site C). Whereas in raw surface
water the species A. veronii and A. media predominated, after
ozonation, the species A. hydrophila subsp. hydrophila repre-
sented more than half of the isolates. The species A. media and
A. punctata were not detected after the ozonation process.
These same two species, A. media and A. punctata, prevailed in
raw and in treated waste water (Table 2).
In order to infer about intra-species variability and to track
bacteria in the different water samples, bacterial isolates were
compared on basis of the nucleotide sequences of the genes
16S rRNA, cpn60 and gyrB. This procedure allowed the identi-
fication of 91 sequence types (Table 2). For sake of simplicity,
the relationship of the isolates was represented in a dendro-
gram based on the comparative analyses of the 16S rRNA,
cpn60 and gyrB concatenated sequences (Fig. 2). Not surpris-
ingly, the same sequence type was not detected in non-
directly-communicating water compartments, i.e, in surface
and waste water. In contrast, the same sequence types were
observed occasionally in communicating zones, separated by
water treatment, i.e. ozonation or waste water treatment. In
the drinking water treatment plant, the same sequence type
of A. veronii (V1) was detected in raw surface water and after
ozonation. In spite of this, water ozonation seemed to impose
a serious bottleneck on strain diversity with the reduction of
48 sequence types in raw surface water to only nine in ozo-
nated water. Moreover, the sequence types of A. hydrophila
subsp. hydrophila detected in ozonated water (isolates C of
sequence type HH1, Fig. 2) were, most of the times, distinct
from those detected in raw surface water, suggesting that
minor population representatives may have gained advantage
under the oxidative stress imposed by ozone. In the waste
water treatment plant, the same sequence types of the species
A. media (M1), A. punctata (P2) and A. sanarellii (S1) were
detected in raw and in treated waste water (Fig. 2) and P2 and
S1 were also present in waste water samples collected in
different dates.
3.2. Antibiotic resistance phenotypes
Amoxicillin resistance was observed in all isolates except in
an Aeromonas enteropelogenes strain, suggesting that aero-
monads are intrinsically resistant to this beta-lactam. For the
other antibiotics, in general, the highest resistance prevalence
values were observed for the penicillin ticarcillin and for the
cephalosporin cephalothin, reaching percentages superior to
40%, irrespective of the type of water. Resistance phenotypes
to the cephalosporin ceftazidime and the carbapenem mer-
openem were observed at low rates, exclusively in surface
water (Table 3). Ceftazidime resistance was observed in
a single isolate of A. media of raw surface water. Meropenem
resistance was observed in three isolates of A. veronii, before
and after water ozonation. Colistin resistance was another
rare phenotype, detected only after the ozonation process in
two isolates of Aeromonas jandaei with the same sequence
type, presumably representing the same strain.
Nalidixic acid resistance was about five times more prev-
alent among waste water isolates (90.6%) than in surface
water (17.6%, p < 0.001) and was not detected after ozonation
(Table 3). In waste water, quinolone resistance was mainly
related to the species A. media and A. punctata, which pre-
dominated in that type of water. Curiously, these two species
were not observed after water ozonation, a fact that may
explain the apparent efficiency of that disinfection process on
quinolone resistance elimination. The potential of A. media as
reservoir of quinolone resistance in waste water was
confirmed by the fact that nine waste water strains (five from
RWW and four from TWW) isolated in the presence of 4 mg L1
of ciprofloxacin were all members of this species. These nine
isolates, represented by eight distinct sequence types, were,
not surprisingly, resistant to nalidixic acid and had at least an
intermediary resistance phenotype to ciprofloxacin (four
resistant and five intermediary). Six out of these nine strains
were resistant to at least three different classes of antibiotics
(R3) (gentamycin, tetracycline and sulfamethoxazole/
trimethoprim), exhibiting resistance phenotypes rare among
the aeromonads isolated in the absence of ciprofloxacin.
Multi-resistance was also frequent among A. punctata (six out
of 18 isolates) (Table 3). In contrast, none of the 33 A. veronii
isolates presented resistance to three different classes of
antibiotics, probably due to the fact that most of these isolates
were recovered from raw surface or ozonated water. Multi-
resistance (R3) did not differ significantly between raw
surface and ozonated water. In contrast, R3 was significantly
( p < 0.05) higher in waste water than in raw surface water.
Table 3 e Antibiotic resistance prevalence (%) in the different sampled sites and Aeromonas species.
NA CIP TIC CP CEF MER STR SXT TET GEN CT R3

Type of water Species distribution

SC (n ¼ 51) 17.6 2.0 68.6 51.0 2.0 3.9 54.9 0 2.0 0 0 5.9

PO (n ¼ 29) 0 0 44.8 44.8 0 3.4 44.8 0 3.4 0 6.9 6.9

RWW (n ¼ 22)
- without CIP (n ¼ 17) 94.1 0 52.9 70.6 0 0 23.5 5.9 5.9 0 0 23.5
- with CIP (n ¼ 5) 100.0 40.0 100.0 80.0 0 0 100.0 20.0 20.0 40.0 0 100.0

TWW (n ¼ 19)
- without CIP (n ¼ 15) 86.7 0 40.0 66.7 0 0 26.7 20.0 13.3 0 0 20.0
- with CIP (n ¼ 4) 100.0 50.0 100.0 100.0 0 0 25.0 25.0 25.0 25.0 0 25.0

Species Water types distribution

A. A. hydrophila subsp. 12.0 0 32.0 52.0 0 0 24.0 0 0 0 0 8.0
hydrophila (n ¼ 25)

B. A. media (n ¼ 25)
ewithout CIP (n ¼ 16) 43.8 6.3 87.5 81.3 6.3 0 12.5 6.3 0 0 0 6.3
ewith CIP (n ¼ 9) 100.0 44.4 100.0 88.9 0 0 66.7 22.2 22.2 33.3 0 66.7

C. A. punctata (n ¼ 18) 88.9 0 22.2 72.2 0 0 44.4 16.7 11.1 0 0 33.3

D. A. veronii (n ¼ 33) 3.0 0 81.8 21.2 0 9.1 78.8 0 0 0 0 0

E. Other (n ¼ 20) 55.0 0 50.0 75.0 0 0 35.0 0 15.0 0 10.0 15.0

NA, nalidixic acid; CIP, ciprofloxacin; TIC, ticarcillin; CP, cephalothin; CEF, ceftazidime; STR, streptomycin; SXT, sulphamethoxazole/trimeth-
oprim; TET, tetracycline; GEN, gentamicin; CT, colistin sulphate; MER, meropenem. R3 represents isolates resistant to 3 or more distinct
antibiotic classes (except AML). SC, surface water captation; PO, post ozonation; RWW, raw waste water; TWW, treated waste water.
5606 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
3.3. Genetic determinants of quinolone resistance and
cphA gene distribution
Given the significantly higher prevalence of quinolone resis-
tance in waste- than in surface water and in the species
Aeromonas puntacta and A. media than in the others ( p < 0.001)
it was decided to investigate if similar mechanisms of resis-
tance were present in both types of water and in the different
species. Irrespective of the type of water or Aeromonas species,
nalidixic acid resistance was associated with mutations in the
gene gyrA (n ¼ 45) and sometimes also on the gene parC (n ¼ 15
and a silent mutation) (Table 4). The most common gyrA
mutations were transversions in the position 83 (AGC / ATC,
in 31 isolates or AGT / ATT, in six isolates) corresponding to
the substitution of a serine for an isoleucine residue. In two
isolates, one of A. allosaccharophila and one of A. jandaei, it was
not possible to achieve a successful amplification of the gene
gyrA, even using alternative primer sets and protocols (n.d. in
Table 4). Among the plasmid-mediated quinolone resistance,
only the genes qnrS and aac(6’)-ib-cr were detected. Although
being found exclusively in the species A. media (Table 4), these
genes were observed in different strains (different sequence
types). The gene qnrS was detected in strains of both surface
and waste water, isolated either in the presence or in the
absence of ciprofloxacin. The qnrS positive strains isolated in
the presence of 4 mg L1 ciprofloxacin harboured also the gene
aac(6’)-ib-cr which, in contrast to qnrS, was associated with
a resistance or intermediary phenotype for ciprofloxacin. The
gene aac(6’)-ib-cr was detected exclusively in strains isolated
on ciprofloxacin-supplemented medium, suggesting that
these strains represent a minor fraction of the bacterial pop-
ulation, which can gain advantage in the presence of selective
pressure. One isolate from surface water harboured the gene
aac(6’)-ib, but not the cr variant that confers resistance to
ciprofloxacin. The cr variant of the gene aac(6’)-ib presented
mutations in the position 102, with an arginine residue (AGG
or, in one RWW isolate, CGG) instead of tryptophan (TGG), on
position 117, with a leucine residue (TTA) instead of serine
(TCA) and on position 179, with a tyrosine residue (TAT)
instead of an aspartate (GAT).
The most common metallo-beta-lactamase expressed by
Aeromonas spp. is encoded by the chromosomal gene cphA,
reported mainly in the species A. hydrophila, A. veronii and A.
jandaei (Janda and Abbott, 2010). The presence and diversity
of this gene was screened in an attempt to identify a differ-
ential pattern between isolates from the drinking and waste
water treatment plants or between different species. The
gene cphA was detected in the species Aeromonas allosachar-
ophila (one isolate from SC and one isolate from RWW), A.
aquariorum (2 isolates from PO), A. hydrophila subsp. hydro-
phila (in all except in one isolate from SC and one from PO), A.
jandaei (in all isolates) and A. veronii (in all except in one
isolate from SC, seven from PO and one from RWW). Unex-
pectedly, it was also detected in a raw waste water isolate of
the species A. media, recovered from ciprofloxacin-supple-
mented medium. The nucleotide sequences of the gene cphA
were different among these isolates. However, those differ-
ences corresponded to silent mutations, as the amino acid
sequences were identical among the waste- and drinking
water treatment plant isolates (data not shown). Neverthe-
less, a noticeable contrast was found in terms of prevalence.
The gene cphA was significantly ( p  0.001) more prevalent
among surface water (65%) than in ozonated water (97%)
isolates. It was also significantly ( p < 0.001) less prevalent
(18%) in the waste water treatment plant than in the drinking
water treatment plant (76%). None of the treated waste water
isolates harboured the gene cphA (Fig. 3). The plasmid related
beta-lactamase gene blaTEM was detected in a single raw
waste water strain of A. media isolated on culture medium
supplemented with 4 mg L1 of ciprofloxacin, and which
harboured also the gene aac(6’)-ib-cr.
4. Discussion
This study was based on the hypothesis that Aeromonas spp.
can serve as vehicle for antibiotic resistance dissemination
within the urban water cycle. The experimental planning
comprised the detection, diversity typing and determination
of antimicrobial resistance patterns of aeromonads within
different parts of the urban water cycle. In respect to detec-
tion, Aeromonas spp. were isolated from raw and treated waste
water, as well as, from surface water, including after ozone
disinfection. In contrast, culturable aeromonads were not
detected in locations with pristine or disinfected water,
ground and tap water, respectively. Apparently the drinking
water treatment, mainly water chlorination, could remove
aeromonads to, at least, less than one CFU in 100 mL of water.
Although the failure to detect aeromonads in tap water could
be attributed to low bacterial densities, in some taps, hetero-
trophs reached 101e104 CFU mL1 (Fig. 1). The presence of
Aeromonas spp. in drinking water is undesirable, as may have
implications for user health, mainly via contact transmission
(WHO, 2008). Nevertheless, aeromonads have been detected in
different types of drinking water, namely tap, mineral bottled
and wells (Kühn et al., 1997; Biscardi et al., 2002; Pablos et al.,
2009). Some authors referred to the seasonality of aero-
monads, which increase may coincide with the raise in the
environmental temperature (Janda and Abbott, 2010). In this
study, a priori, the failure to detect Aeromonas spp. in ground
and tap water cannot be attributed to such seasonality, given
the fact that these samples were collected in Summer and
Winter (ground water) or in Spring, Summer and Autumn
(taps). It is noteworthy that the absence of culturable Aero-
monas spp. in tap water contrasts to what was observed in the
same samples for other bacterial groups. For example, it was
observed that sphingomonads, pseudomonads, and Acineto-
bacter spp., in spite the sharp decrease of total heterotrophs
observed after water chlorination, were present in tap water
(Vaz-Moreira et al., 2011; our data unpublished). The fact that
the examined taps had a low usage rate (one to four times
a month) may be part of the possible explanation for the
absence of Aeromonas in tap water, as stagnancy of water in
pipes is described as promoting bacterial community rear-
rangements (Lautenschlager et al. 2010). However, a deeper
study would be needed to confirm such hypothesis. Thus, the
apparent reduced risk of Aeromonas spp. to contribute for
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1 5607

Table 4 e Diversity, antibiotic resistance phenotypes, origin and genetic determinants of quinolone resistance of nalidixic
acid resistant isolates.

Resistance phenotype Mutations Quinolone

Species (n) ST (n) Origin (n)
CIP TIC CP STR Other gyrA parC resistance genes
A. allosaccharophila A2 (1) GAA (Glu)87 - -
RWW (2)
(2) A3 (1) n.d. AGA (Arg)80 -
Aq2 (1) ATC (Ill)83 - -
A. aquariorum (3) Aq3 (1) TWW (3) ATC (Ill)83 ATC (Ill)80 -
Aq4 (1) TET ATC (Ill)83 - -
HH7 (1) ATT (Ill)83 ATT (Ill)80 -
A. hydrophila subsp. SC (2)
HH8 (1) ATT (Ill)83 - -
hydrophila (3)
HH12 (1) RWW (1) ATC (Ill)83 ATC (Ill)80 -
A. jandaei (1) J2 (1) SC (1) n.d. - -
aac(6’)-ib-cr
RWW (1) GEN TET CGC (Arg)83 AGT (Ser)80
M1 (2)* blaTEM
TWW (1) ATC (Ill)83 - aac(6’)-ib-cr
83 aac(6’)-ib
M6 (1) CAZ ATC (Ill) -
qnrS
M7 (1) SC (3) AGA (Arg)83 - -
M10 (1) ATC (Ill)83 - -
M12 (1) ATC (Ill)83 - -
M13 (1)* GEN ATC (Ill)83 ATC (Ill)80 aac(6’)-ib-cr
aac(6’)-ib-cr
A. media (16) M14 (1)* SXT TET ATC (Ill)83 -
qnrS
RWW (6)
aac(6’)-ib-cr
M15 (1)* GTC (Val)83 -
qnrS
83
M16 (1)* ATC (Ill) - aac(6’)-ib-cr
M18 (1) ATC (Ill)83 - -
M19 (1)* ATC (Ill)83 - aac(6’)-ib-cr
M20 (1)* ATC (Ill)83 ATC (Ill)80 aac(6’)-ib-cr
M21 (1)* TWW (5) SXT TET GEN ATC (Ill)83 ATC (Ill)80 aac(6’)-ib-cr
M23 (1) SXT TET ATC (Ill)83 CGC (Arg)80 -
M24 (1) AGA (Arg)83 - qnrS
ATC (Ill)83 - -
P1 (3) RWW (3) TET ATC (Ill)83 - -
SXT TET ATC (Ill)83 - -
RWW (1) ATC (Ill)83 - -
P2 (2)
TWW (1) ATC (Ill)83 - -
P3 (1) ATC (Ill)83 ATC (Ill)80 -
P4 (1) SC (3) AGA (Arg)83 AAA (Lys)84 -
P6 (1) ATC (Ill)83 AAA (Lys)84 -
A. punctata (16)
P7 (1) AGG (Arg)83 - -
P8 (1) SXT ATC (Ill)83 - -
RWW (5)
P9 (1) AGA (Arg)83 - -
P10 (1) ATC (Ill)83 - -
P12 (1) ATC (Ill)83 - -
P13 (1) SXT ATC (Ill)83 ATC (Ill)80 -
TWW (4)
P14 (1) SXT ATC (Ill)83 ATC (Ill)80 -
P15 (1) ATC (Ill)83 AAA (Lys)84 -
RWW (1) ATT (Ill)83 - -
S1 (3) ATT (Ill)83 - -
A. sanarellii (4) TWW (2)
ATT (Ill)83 - -
S2 (1) TWW (1) TET ATT (Ill)83 - -
A. taiwanensis (1) T1 (1) RWW (1) ATC (Ill)83 ATC (Ill)80 -
A. veronii (1) V25 (1) RWW (1) ATC (Ill)83 - -
n.d., not determined; (shadowing; black, resistant; grey, intermediary; white, susceptible) .
* isolated in medium supplemented with 4 mg L1 of ciprofloxacin.
5608 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
Fig. 3 e Percentage of isolates of each type of water that
harbored the cphA gene or the gyrA mutation.
antimicrobial resistance dissemination via tap water,
observed in this study, must be interpreted with precaution.
In respect to diversity, it was observed that the pattern of
Aeromonas species was distinct in the different types of water.
Although A. media was abundant in both raw surface and waste
water, these types of water differed on the predominance of A.
veronii and A. punctata (Table 2). Nevertheless, the percentage of
distinct sequence types was not significantly different in
surface and waste water. Apparently, and in spite the low
number of isolates, waste water treatment did not lead to
a significant reduction of the number of sequence types. In
contrast, water ozonation seemed to impose a bottleneck both
in the number of species and of sequence types (Fig. 2). Indeed,
the species Aeromonas hydrophyla subsp. hydrophyla became
over represented after ozonation with a significant ( p < 0.001)
reduction in the percentage of distinct sequence types. More-
over, in general, the sequence types detected after ozonation
were different from those found in raw surface water, sug-
gesting some kind of rearrangement in the aeromonads pop-
ulation due to water disinfection (Fig. 2).
A. media and A. punctata were the species in which quinolone
resistance presented the highest prevalence ( p < 0.001) and the
predominance of these species in waste water contributed to
explain the elevated rates of nalidixic acid resistance in waste
water, significantly ( p < 0.001) higher than in surface water.
Similarly, sulfamethoxazole/trimethoprim resistance found
exclusively in those two species, was observed only in waste
water. In contrast, ceftazidime and meropenem resistance were
detected only in surface water, although it is acknowledged that
these resistance phenotypes could have been detected also in
waste water if a larger number of isolates had been examined.
The search for genetic determinants related to quinolone
resistance showed that the gyrA mutations were the primary,
even not the unique, mechanism (Table 4). The higher preva-
lence of these mutations in waste water isolates in compar-
ison with the prevalence values observed in surface water
(Fig. 3), can be supported by previous studies which demon-
strate that quinolone resistance may arise from the contact
with mutagenic substances, widely found in the environment
(Miyahara et al., 2011). In fact, in waste water the occurrence
of such potential mutagens is much more probable than in
uncontaminated waters. Additionally, some effect of selective
pressure may take place in waste waters, in which the detec-
tion of quinolones is common, with concentrations of cipro-
floxacin up to 0.7 mg/L detected in Portuguese municipal waste
water treatment plants (Seifrtová et al., 2008; our data for the
same plant, unpublished). It is not possible to know the rele-
vance of de novo mutation or of selection (vertical trans-
mission) for the observed chromosomal mutations associated
with quinolone resistance. But, although probably both forms
can contribute for resistance spreading, de novo events may be
frequent as gyrA or parC gene mutations were observed in
different strains (sequence types). In any case, the species
A. punctata and A. media seem to play an important role on this
form of dissemination. These results are in agreement with
the work of Goñi-Urriza et al. (2000), who assessed the impact
of an urban effluent on antibiotic resistance of Aeromonas
spp. in a riverine area. As in the current study, nalidixic acid
resistance was observed in the majority of the aeromonads
(72%), most of them of the species A. punctata (A. caviae), and
was exclusively chromosomally encoded. The conclusion
reached by Goñi-Urriza et al. (2000), applies also to the present
study e urban effluents are responsible for the increase of
quinolone resistance in the receptor water courses.
Among the quinolone resistance determinants associated
with mobile genetic elements, only the genes qnrS and aac(6’)-
ib-cr were detected and only in the species A. media. In both
cases, these genetic determinants were found in different
strains (sequence types), as expected if horizontal gene trans-
fer is equated. Both determinants, and mainly aac(6’)-ib-cr
which was detected only in isolates recovered in the presence
of ciprofloxacin, were rare in the analysed samples. Never-
theless, the gene qnrS is apparently widespread in waters
(waste water, rivers, aquaculture), detected not only from
total DNA and but also from cultures of aeromonads and
Enterobacteriaceae (Cattoir et al., 2008; Picão et al., 2008;
Szczepanowski et al., 2009; Ishida et al., 2010; Cummings
et al., 2011). Similarly, the gene aac(6’)-ib-cr is found in
different types of water (lake water, river sediments, aqua-
culture), either in total DNA or in bacterial isolates (aero-
monads and Enterobacteriaceae) (Picão et al. 2008; Ishida et al.,
2010; Cummings et al., 2011). The presence of these genes,
although conferring low-levels of resistance, can favour and
complement the selection of other resistance mechanisms
(Rodrı́guez-Martı́nez et al., 2010). The fact that the determi-
nants qnrS and aac(6’)-ib-cr were detected only in A. media
suggests that this species may represent an important vector
of quinolone resistance. Ceftazidime resistance was also
detected only in this same species in surface water. Recent
evidences that water A. media can colonize humans (Khajanchi
et al., 2010) may give additional relevance to this species on the
dispersal of resistance. Although the number of isolates
examined was too low to strongly support this conclusion, the
data suggested that water ozonation may promote the reduc-
tion of A. media. For instance, no quinolone or ceftazidime
resistance were observed downstream of this point. In
contrast, meropenem resistance, in this study associated to
the species A. veronii, was observed also in ozonated water.
Nevertheless, the data gathered in this study suggests that
water chlorination may contribute to control resistance prop-
agation by aeromonads via drinking water.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
In general, the results obtained suggest that different aer-
omonads populations and antibiotic resistance determinants
prevail in different parts of the urban water cycle. The clearest
example of this was the distribution of the gene cphA and of
the gyrA mutation, observed in the majority of the surface and
waste water isolates, respectively (Fig. 3). As expected, the
gene cphA was predominant among the species A. hydrophila
subsp. hydrophila, A. veronii and A. jandaei, also the most
prevalent in the drinking water treatment plant. In the same
way, the gyrA mutations prevailed in A. media and A. punctata,
the predominant species in waste water. The contrast
observed in the distribution of both genetic determinants is
mainly due to the patterns of species occurring in both types
of water and which, probably, are due to the environmental
conditions and selective pressures imposed in both types of
habitat. This also demonstrates that, in each type of water,
aeromonads may represent a source of distinct types of anti-
biotic resistance.
The importance of a given Aeromonas species for the anti-
microbial resistance patterns in each type of water, observed
in the current work, confirm previous studies conducted with
other bacterial groups. Figueira et al. (2011) studied different
populations of waste water E. coli and concluded that varia-
tions on the prevalence of quinolone resistance were corre-
lated with the dynamics of some population sub-sets. Vaz-
Moreira et al. (2011) characterizing the patterns of antimicro-
bial resistance in sphingomonads from tap water and cup
fillers of dental chairs also concluded that antibiotic resis-
tance patterns were often species- rather than site-related.
Nevertheless, in the current work, and in contrast to what
was suggested by other authors studying A. salmonicida from
fish farms and environmental samples (Giraud et al., 2004;
Kim et al., 2011), no clonal spreading of antibiotic resistance
was observed. In contrast, rarely were observed the same
sequence types in different water samples. This suggests that
the acquisition of a specific resistance type, either by hori-
zontal gene transfer or by adaptive mutation, may take place
preferentially in a given habitat, in which a species is preva-
lent or has a higher fitness than the others. In other words, the
success of resistance acquisition may depend on the fitness of
the target bacterium (receptor of horizontal gene transfer or
mutant) in a specific environment.
5. Conclusions
The patterns of Aeromonas species and antimicrobial resis-
tance varied over different parts of the urban water cycle;
In each type of water, the antimicrobial resistance patterns
were primarily function of the prevailing species;
In raw surface and waste water no strong evidences for
clonal dissemination of antimicrobial resistance were
detected;
Water ozonation imposed a bottleneck on species diversity,
with evidences of clonal selection, and promoted a significant
reduction of quinolone resistance and the increase of cphA
metallo-beta-lactamase;
Waste water aeromonads, particularly A. media and A.
punctata, were confirmed as relevant environmental harbours
5609
of quinolone resistance, either chromosomally ( gyrA muta-
tion) or plasmid encoded (qnrS and aac(60 )-Ib-cr).
Water aeromonads were confirmed as relevant agents for
antimicrobial resistance spreading in the environment, which
presence in tap water could be significantly reduced by water
chlorination.
Acknowledgements
Authors gratefully acknowledge Prof. P. Nordmann, Dr. L.M.
Cavaco, and Prof. M. Wang for the positive controls for the
detection of the genes qnrA, qnrB and qnrS; qnrD and aac(6’)Ib-
cr; and qnrC, respectively. This study was financed by Funda-
ção para a Ciência e a Tecnologia (projects PTDC/AMB/70825/
2006; PTDC/AMB/71236/2006, IVM grant SFRH/BD/27978/2006;
MS grant Integration into Research).
references
Abulhamd, A.T., 2009. Characterization of Aeromonas hydrophila
isolated from aquatic environments using phenotypic and
genotyping methods. Research Journal of Agriculture and
Biological Sciences 5, 923e931.
Aminov, R.I., 2009. The role of antibiotics and antibiotic
resistance in nature. Environmental Microbiology 11 (12),
2970e2988.
Arias, A., Seral, C., Navarro, F., Miro, E., Coll, P., Castillo, F.J.,
2010a. Plasmid-mediated QnrS2 determinant in an Aeromonas
caviae isolate recovered from a patient with diarrhoea. Clinical
Microbiology and Infectious Diseases 16, 1005e1007.
Arias, A., Seral, C., Gude, M.J., Castillo, F.J., 2010b. Molecular
mechanisms of quinolone resistance in clinical isolates of
Aeromonas caviae and Aeromonas veronii bv. sobria.
International Microbiology 13, 135e141.
Balsalobre, L.C., Dropa, M., Lincopan, N., Mamizuka, E.M.,
Matté, G.R., Matté, M.H., 2009. Detection of metallo-b-
lactamases-encoding genes in environmental isolates of
Aeromonas hydrophila and Aeromonas jandaei. Letters in Applied
Microbiology 49, 142e145.
Baquero, F., Martı́nez, J.-L., Cantón, R., 2008. Antibiotics and
antibiotic resistance in water environments. Current Opinion
in Biotechnology 19, 260e265.
Biscardi, D., Castaldo, A., Gualillo, O., Fuscos, R., 2002. The
occurrence of cytotoxic Aeromonas hydrophila strains in Italian
mineral and thermal waters. Science of the Total Environment
292, 255e263.
Blasco, M.D., Esteve, C., Alcaide, E., 2008. Multiresistant
waterborne pathogens isolated from water reservoirs and
cooling systems. Journal of Applied Microbiology 105 (2),
469e475.
Cattoir, V., Poirel, L., Rotimi, V., Soussy, C.J., Nordmann, P., 2007.
Multiplex PCR for detection of plasmid-mediated quinolone
resistance qnr genes in ESBL-producing enterobacterial
isolates. Journal of Antimicrobial Chemotherapy 60, 394e397.
Cattoir, V., Poirel, L., Aubert, C., Soussy, C.J., Nordmann, P., 2008.
Unexpected occurrence of plasmid-mediated quinolone
resistance determinants in environmental Aeromonas spp.
Emerging Infectious Diseases 14, 231e237.
Cavaco, L.M., Hasman, H., Xia, S., Aarestrup, F.M., 2009. qnrD,
a novel gene conferring transferable quinolone resistance in
Salmonella enterica serovar Kentucky and Bovismorbificans
5610 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
strains of human origin. Antimicrobial Agents and
Chemotherapy 53, 603e608.
Clinical and Laboratory Standards Institute, 2007. M100eS17.
Performance Standards for Antimicrobial Susceptibility
Testing, vol. 27, Wayne, PA.
Corry, J.E.L., Curtis, G.D.W., Baird, R.M., 2003. Handbook of Culture
Media for Food Microbiology, vol. 34. Elsevier Science BV,
Amsterdam, The Netherlands, pp. 331.
Cummings, D., Archer, K.F., Arriola, D.J., Pieter, A.B., Faucett, K.G.,
Laroya, J.B., Pfeil, K.L., Ryan, C.R., Ryan, K.R.U., Zuill, D.E., 2011.
Broad dissemination of plasmid-mediated quinolone
resistance genes in sediments of two urban coastal wetlands.
Environmental Science & Technology 45, 447e454.
DiPersio, J.R., Deshpande, L.M., Biedenbach, D.J., Toleman, M.A.,
Walsh, T.R., Jones, R.N., 2005. Evolution and dissemination of
extended-spectrum â-lactamase-producing Klebsiella
pneumoniae: epidemiology and molecular report from the
SENTRY Antimicrobial Surveillance Program (1997e2003).
Diagnostic Microbiology & Infectious Disease 51, 1e7.
Dixon, B., 2008. Natural disaster microbiology. Microbe 3,
312e313.
Faria, C., Vaz-Moreira, I., Serapicos, E., Nunes, O.C., Manaia, C.M.,
2009. Antibiotic resistance in coagulase negative staphylococci
isolated from wastewater and drinking water. Science of the
Total Environment 407, 3876e3882.
Ferreira da Silva, M., Tiago, I., Verı́ssimo, A., Boaventura, A.R.,
Nunes, O.C., Manaia, C.M., 2006. Antibiotic resistance of
enterococci and related bacteria in an urban waste water
treatment plant. FEMS Microbiololy Ecology 55, 322e329.
Ferreira da Silva, M., Vaz-Moreira, I., Gonzalez-Pajuelo, M.,
Nunes, O.C., Manaia, C.M., 2007. Antimicrobial resistance
patterns in Enterobacteriaceae isolated from an urban waste
water treatment plant. FEMS Microbiology Ecology 60,
166e176.
Figueira, V., Serra, E., Manaia, C.M., 2011. Differential patterns of
antimicrobial resistance in population subsets of Escherichia
coli isolated from waste- and surface waters. Science of the
Total Environment 409, 1017e1023.
Giraud, E., Blanc, G., Bouju-Albert, A., Weill, F.-X., Donnay-
Moreno, C., 2004. Mechanisms of quinolone resistance and
clonal relationship among Aeromonas salmonicida strains
isolated from reared fish with furunculosis. Journal of Medical
Microbiology 53, 895e901.
Gordon, L., Cloeckaert, A., Doublet, B., Schwarz, S., Bouju-
Albert, A., Ganière, J.-P., Le Bris, H., Le Flèche-Matéos, A.,
Giraud, E., 2008. Complete sequence of the floR-carrying
multiresistance plasmid pAB5S9 from freshwater Aeromonas
bestiarum. Journal of Antimicrobial Chemotherapy 62 (1),
65e71.
Goñi-Urriza, M., Capdepuy, M., Arpin, C., Raymond, N.,
Caumette, P., Quentin, C., 2000. Impact of an urban effluent on
antibiotic resistance of riverine Enterobacteriaceae and
Aeromonas spp. Applied and Environmental Microbiology 66
(1), 125e132.
Goñi-Urriza, M., Arpin, C., Capdepuy, M., Dubois, V., Caumette, P.,
Quentin, C., 2002. Type II topoisomerase quinolone resistance-
determining regions and mutations asociated with quinolone
resistance. Antimicrobial Agents and Chemotherapy 46,
350e359.
Huddleston, J.R., Zak, J.C., Jeter, R.M., 2006. Antimicrobial
susceptibilities of Aeromonas spp. isolated from environmental
sources. Applied and Environmental Microbiology 72 (11),
7036e7042.
Ishida, Y., Ahmed, A.M., Mahfouz, N.B., Kimura, T., El-Khodery, S.
A., Moawad, A.A., Tadashi Shimamoto, T., 2010. Molecular
analysis of antimicrobial resistance in Gram-negative bacteria
isolated from fish farms in Egypt. Journal of Veterinary
Medical Science 72 (6), 727e734.
Janda, J.M., Abbott, S.L., 2010. The genus Aeromonas: taxonomy,
pathogenicity, and infection. Clinical Microbiology Reviews 23,
35e73.
Jukes, T.H., Cantor, C.R., 1969. Evolution of protein molecules. In:
Mammalian Protein Metabolism, vol. 3. Academic Press,
Munro, pp. 21e132. Academic Press, NY.
Khajanchi, B.K., Fadl, A.A., Borchardt, M.A., Berg, R.L.,
Horneman, A.J., Stemper, M.E., Joseph, S.W., Moyer, N.P.,
Sha, J., Chopra, A.K., 2010. Distribution of virulence factors
and molecular fingerprinting of Aeromonas species isolates
from water and clinical samples: suggestive evidence of
water-to-human transmission. Applied and Environmental
Microbiology 76, 2313e2325.
Kim, J.H., Hwang, S.Y., Son, J.S., Han, J.E., Jun, J.W., Shin, S.P.,
Choresca Jr., C., Choi, Y.J., Park, T.H., Park, S.C., 2011.
Molecular characterization of tetracycline- and quinolone-
resistant Aeromonas salmonicida isolated in Korea. Journal of
Veterinary Science 12 (1), 41e48.
Kivanc, M., Yilmaz, M., Demir, F., 2011. The occurrence of
Aeromonas in drinking water, tap water and the porsuk river.
Brazilian Journal of Microbiology 42, 126e131.
Kühn, I., Huys, G., Coopman, R., Kersters, K., Janssen, P., 1997. A
4-year study of the diversity and persistence of coliforms and
Aeromonas in the water of a Swedish drinking water well.
Canadian Journal of Microbiology 43, 9e16.
Kümmerer, K., 2009. Antibiotics in the aquatic environment e
a review e part II. Chemosphere 75, 435e441.
Lamy, B., Kodjo, A., Laurent, F.colBVH Study Group, 2009.
Prospective nationwide study of Aeromonas infections in
France. Journal of Clinical Microbiology 47, 1234e1237.
Lane, D.J., 1991. 16S/23S rRNA sequencing. In: Stackebrandt, E.,
Goodfellow, M. (Eds.), Nucleic Acid Techniques in Bacterial
Systematics. Wiley, Chichester, pp. 115e175.
Lautenschlager, K., Boon, N., Wang, Y., Egli, T., Hammes, F., 2010.
Overnight stagnation of drinking water in household taps
induces microbial growth and changes in community
composition. Water Research 44 (17), 4868e4877.
Maalej, S., Mahjoubi, A., Elazri, C., Dukan, S., 2003. Simultaneous
effects of environmental factors on motile Aeromonas
dynamics in an urban effluent and in the natural seawater.
Water Research 37, 2865e2867.
Miñana-Galbis, D., Urbizu-Serrano, A., Farfán, M., Fusté, M.C.,
Lorén, J.G., 2009. Phylogenetic analysis and identification of
Aeromonas species based on sequencing of the cpn60 universal
target. International Journal of Systematic and Evolutionary
Microbiology 59, 1976e1983.
Miyahara, E., Nishie, M., Takumi, S., Miyanohara, H., Nishi, J.,
Yoshiie, K., Oda, H., Takeuchi, M., Komatsu, M., Aoyama, K.,
Horiuchi, M., Takeuchi, T., 2011. Environmental mutagens
may be implicated in the emergence of drug-resistant
microorganisms. FEMS Microbiology Letters 317, 109e116.
Novo, A., Manaia, C.M., 2010. Factors influencing antibiotic
resistance burden in municipal waste water treatment plants.
Applied Microbiology and Biotechnology 87, 1157e1166.
Pablos, M., Rodrı́guez-Calleja, J.M., Santos, J.A., Otero, A., Garcı́a-
López, M.L., 2009. Occurrence of motile Aeromonas in
municipal drinking water and distribution of genes encoding
virulence factors. International Journal of Food Microbiology
135, 158e164.
Park, C.H., Robicsek, A., Jacoby, G.A., Sahm, D., Hooper, D.C., 2006.
Prevalence in the United States of aac(6’)-Ib-cr encoding
a ciprofloxacin-modifying enzyme. Antimicrobial Agents and
Chemotherapy 50, 3953e3955.
Périchon, B., Courvalin, P., Galimand, M., 2007. Transferable
resistance to aminoglycosides by methylation of G1405 in 16S
rRNA and to hydrophilic fluoroquinolones by QepA-mediated
efflux in Escherichia coli. Antimicrobial Agents and
Chemotherapy 51, 2464e2469.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 5 9 9 e5 6 1 1
Picão, R.C., Poirel, L., Demarta, A., Silva, C.S., Corvaglia, A.R.,
Petrini, O., Nordmann, P., 2008. Plasmid-mediated quinolone
resistance in Aeromonas allosaccharophila recovered from a Swiss
lake. Journal of Antimicrobial Chemotherapy 62, 948e950.
Rodrı́guez-Martı́nez, J.-M., Nordmann, P., Ronco, E., Poirel, L.,
2010. Extended-spectrum cephalosporinase in Acinetobacter
baumannii. Antimicrobial Agents and Chemotherapy 54 (8),
3484e3488.
Schwartz, T., Kohnen, W., Jansen, B., Obst, U., 2003. Detection of
antibiotic-resistant bacteria and their resistance genes in
wastewater, surface water, and drinking water biofilms. FEMS
Microbiology Ecology 43, 325e335.
Seifrtová, M., Pena, A., Lino, C.M., Solich, P., 2008. Determination
of fluoroquinolone antibiotics in hospital and municipal
wastewaters in Coimbra by liquid chromatography with
a monolithic column and fluorescence detection. Analytical
Bioanalytical Chemistry 391, 799e805.
Servais, P., Passerat, J., 2009. Antimicrobial resistance of fecal
bacteria in waters of the Seine river watershed (France).
Science of the Total Environment 408, 365e372.
Szczepanowski, R., Linke, B., Krahn, I., Gartemann, K.-H.,
Gützkow, T., Eichler, W., Pühler, A., Schlüter, A., 2009.
Detection of 140 clinically relevant antibiotic resistance genes
in the plasmid metagenome of wastewater treatment plant
bacteria showing reduced susceptibility to selected
antibiotics. Microbiology 155, 2306e2319.
Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular
572 Evolutionary genetics analysis (MEGA) software version 4.0.
Molecular Biology and Evolution 24, 1596e1599.
Taylor, N.G.H., Verner-Jeffreys, D.W., Baker-Austin, C., 2011.
Aquatic systems: maintaining, mixing and mobilising
antimicrobial resistance? Trends in Ecology and Evolution 26
(6), 278e284.
5611
Vaz-Moreira, I., Nunes, O.C., Manaia, C.M., 2011. Diversity and
antibiotic resistance patterns of Sphingomonadaceae isolates
from drinking water. Applied and Environmental Microbiology
77 (16), 5697e5706.
Walsh, T.R., Stunt, R.A., Nabi, J.A., MacGowan, A.P., Bennett, P.M.,
1997. Distribution and expression of beta-lactamase genes
among Aeromonas spp. Journal of Antimicrobial
Chemotherapy 40, 171e178.
Wang, M., Guo, Q., Xu, X., Wang, X., Ye, X., Wu, S., Hooper, D.C.,
Wang, M., 2009. New plasmid-mediated quinolone resistance
gene, qnrC, found in a clinical isolate of Proteus mirabilis.
Antimicrobial Agents and Chemotherapy 53, 1892e1897.
World Health Organization, 2008. Guidelines for drinking water,
third ed. WHO, Geneva.
Xi, C., Zhang, Y., Marrs, C.F., Ye, W., Simon, C., Foxman, B.,
Nriagu, J., 2009. Prevalence of antibiotic resistance in drinking
water treatment and distribution systems. Applied and
Environmental Microbiology 75 (17), 5714e5718.
Xia, S., Hendriksen, R.S., Xie, Z., Huang, L., Zhang, J., Guo, W.,
Xu, B., Ran, L., Aarestrup, F.M., 2009. Molecular
characterization and antimicrobial susceptibility of Salmonella
isolates from infections in humans in Henan Province, China.
Journal of Clinical Microbiology 47, 401e409.
Xia, R., Guo, X., Zhang, Y., Xu, H., 2010. qnrVC-like gene located in
a novel complex class 1 integron harboring the ISCR1 element
in an Aeromonas punctata strain from an aquatic environment
in Shandong Province, China. Antimicrobial Agents and
Chemotherapy 54, 3471e3474.
Yáñez, M.A., Catalán, V., Apráiz, D., Figueras, M.J., Martı́nez-
Murcia, A.J., 2003. Phylogenetic analysis of members of the
genus Aeromonas based on gyrB gene sequences. International
Journal of Systematic and Evolutionary Microbiology 53,
875e883.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Long-term effect of ZnO nanoparticles on waste activated

sludge anaerobic digestion

Hui Mu, Yinguang Chen*

State Key Laboratory of Pollution Control and Resources Reuse, School of Environmental Science and Engineering, Tongji University,
1239 Siping Road, Shanghai 200092, China

article info abstract

Article history: The increasing use of zinc oxide nanoparticles (ZnO NPs) raises concerns about their
Received 9 May 2011 environmental impacts, but the potential effect of ZnO NPs on sludge anaerobic digestion
Received in revised form remains unknown. In this paper, long-term exposure experiments were carried out to
10 August 2011 investigate the influence of ZnO NPs on methane production during waste activated sludge
Accepted 14 August 2011 (WAS) anaerobic digestion. The presence of 1 mg/g-TSS of ZnO NPs did not affect methane
Available online 23 August 2011 production, but 30 and 150 mg/g-TSS of ZnO NPs induced 18.3% and 75.1% of inhibition
respectively, which showed that the impact of ZnO NPs on methane production was
Keywords: dosage dependant. Then, the mechanisms of ZnO NPs affecting sludge anaerobic digestion
Zinc oxide nanoparticles were investigated. It was found that the toxic effect of ZnO NPs on methane production
Waste activated sludge was mainly due to the release of Zn2þ from ZnO NPs, which may cause the inhibitory
Anaerobic digestion effects on the hydrolysis and methanation steps of sludge anaerobic digestion. Further
Mechanisms investigations with enzyme and fluorescence in situ hybridization (FISH) assays indicated
that higher concentration of ZnO NPs decreased the activities of protease and coenzyme
F420, and the abundance of methanogenesis Archaea.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
With the rapid development of nanotechnology, nanoparticles
(NPs) are now widely used in some industrial products, such as
antibactericide coatings, catalysts, biomedicine, skin creams
and toothpastes because of their unique physicochemical
properties of enhanced magnetic, electrical, optical, and etc
(Maynard et al., 2006; Roco, 2005). It is inevitable for the release
of NPs from discover source to environment receptor, and some
NPs have been found in wastewater treatment plants (WWTPs)
and waste sludge (Brar et al., 2010). It is therefore necessary to
evaluate their impacts on the environment.
Zinc oxide (ZnO) NPs, one of metal oxide NPs, have received
increasing interest due to their widespread industrial, medical
and military applications (Ellsworth et al., 2000; Miziolek,
2002; Serda et al., 2009). There are some publications discus-
sing the toxicity of ZnO NPs on microbes. For example, Adams
et al. (2006) reported that 500 mg/L of ZnO NPs significantly
inhibited the growth of Bacillus subtilis up to 90%, but only
induced 38% of the growth inhibition of Escherichia coli,
meaning the different toxicity of ZnO NPs on different species
of bacteria. Previous study also showed that ZnO NPs reduced
the microbial biomass, and altered the diversity and compo-
sition of soil bacterial community (Ge et al., 2011).
The release of ZnO NPs to WWTPs, which are usually
operated with an activated sludge process, has been reported
recently (Gottschalk et al., 2009). The released ZnO NPs were
observed to be removed by activated sludge via adsorption,
aggregation and settling in WWTPs (Kiser et al., 2010, 2009;
Kiser et al., 2010; Limbach et al., 2008). Large amounts of WAS

* Corresponding author. Tel.: þ86 21 65981263; fax: þ86 21 65986313.

E-mail address: yinguangchen@yahoo.com (Y. Chen).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.022
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0
are produced in municipal WWTPs, which need to be treated
before being discharged to the environment. It can be antici-
pated that most of ZnO NPs will enter into sludge treatment
system. Among several sludge treatment methods anaerobic
digestion for methane is a preferred one because of WAS being
reused and the energy being recovered. Nevertheless, the
toxicity of ZnO NPs to sludge anaerobic digestion has seldom
been investigated.
Some studies addressed that the toxicity of ZnO NPs came
from the released zinc ions (Zn2þ) (Franklin et al., 2007; Wong
et al., 2010; Xia et al., 2008), but others found that the toxicity
of ZnO NPs to some microorganisms (such as E. coli and
Pseudomonas fluorescens) was not caused by the released Zn2þ
but ZnO NPs themselves (Jiang et al., 2009). Thus, the role of
released Zn2þ from ZnO NPs on sludge anaerobic digestion
should be taken into account. Moreover, oxidative stress
induced by ZnO NPs was reported to cause the loss of cell
viability, and the increase of intracellular reactive oxygen
species (ROS) was found to be toxic to cytoplasmic lipids,
proteins and other intermediates in cells (Sharma et al., 2009;
Xia et al., 2008). This study was to evaluate the impact of ZnO
NPs on methane production during sludge anaerobic diges-
tion and to explore the mechanisms. Furthermore, fluores-
cence in situ hybridization (FISH) technique with 16S rRNA-
targeted oligonucleotide probes was employed to monitor
the quantity change of bacteria and Archaea community after
WAS anaerobic digestion system long-term exposed to ZnO
NPs.
2. Materials and methods
2.1. Nanoparticles and waste activated sludge
ZnO NPs were purchased from Sigma Aldrich (St. Louis, MO).
The X-ray diffraction (XRD) pattern of ZnO NPs was measured
using a Rigaku D/Max-RB (Rigaku, Japan) diffractometer
equipped with a rotating anode and a Cu Ka radiation source
and shown in Fig. S1 (Supplementary Information). In this
study, stock dispersion of ZnO NPs was produced by adding 2 g
ZnO NPs to 1.0 L distilled water (pH 7.0) containing 0.1 mM
sodium dodecylbenzene sulfonate (SDBS) (Sigma Aldrich, St.
Louis, MO) to enhance the stability of nano-suspension
because the particles almost immediately aggregated in
surrounding medium (Adams et al., 2006; Franklin et al., 2007;
Ganesh et al., 2010; Keller et al., 2010; Simon-Deckers et al.,
2009; Xia et al., 2008). The stock dispersion was sonicated
(25  C, 250 W, 40 kHz) for 1 h to break aggregates before being
diluted to the exposure concentrations. Analysis of the
suspension by dynamic light scattering (DLS) (Franklin et al.,
2007; Simon-Deckers et al., 2009) using a Malvern Autosizer
4700 (Malvern Instruments, UK) indicated that the average
particle size of ZnO NPs was approximately 140  20 nm on the
basis of the number distribution with more than five separate
measurements per sample.
The WAS used in this study was withdrawn from the
secondary sedimentation tank of a municipal WWTP in
Shanghai, China. The sludge was concentrated by settling at 4  C
for 24 h, and its main characteristics (average data plus standard
deviations of triplicate tests) are as follows: pH 6.7  0.2, total
5613
suspended solids (TSS) 10070  780 mg/L, volatile suspended
solids (VSS) 7690  452 mg/L, soluble chemical oxygen demand
(SCOD) 90  5 mg/L, total chemical oxygen demand (TCOD)
10710  220 mg/L, total carbohydrate 899  530 mg-COD/L, total
protein 5685  149 mg-COD/L, and total zinc 0.8  0.2 mg/g-TSS.
2.2. Determination of ZnO nanoparticles dissolution
In order to measure the concentration of released Zn2þ from
ZnO NPs, three concentrations of ZnO NPs in 0.1 mM SDBS
solutions were prepared with the stock dispersion, and the
mixtures were maintained in an air-batch shaker (150 rpm) at
35  1  C for 48 h. At different time, the samples were with-
drawn and centrifuged at 12000 rpm for 30 min, and the
supernatant was collected, and filtered through 0.22 mm mixed
cellulose ester membrane (Jiang et al., 2009; Li et al., 2011). The
released zinc was determined by inductively coupled plasma
optical emission spectrometry (ICP-OES, PerkinElmer Optima
2100 DV, USA) after acidified with 4% ultrahigh purity HNO3.
2.3. Experiments of effects of ZnO nanoparticles and
their released Zn2þ on WAS anaerobic digestion for methane
production
Three dosages (1, 30 and 150 mg/g-TSS) of ZnO NPs were used to
investigate the impact of ZnO NPs on WAS digestion in this
paper. The dosage of 1 mg/g-TSS of ZnO NPs was chosen to be
the environmentally relevant concentration according to the
literature (EPA, 2009; Gottschalk et al., 2009). Also some scien-
tists suggested that although lower nanomaterial content
(50 mg C60/g-TSS in their study) showed almost no influence on
anaerobic community, a much higher nanomaterial dosage
should be investigated before the final conclusion regarding the
toxicity of nanomaterial was reached (Nyberg et al., 2008).
Moreover, since the environmental release of NPs might be
increased due to their large-scale production, the potential
effects of higher concentrations (30 and 150 mg/g-TSS) of ZnO
NPs were also investigated in this study according to the refer-
ence (Adams et al., 2006). The influence of ZnO NPs long-term
exposure on methane production was conducted in series of
serum bottles (500 mL each), with a sludge volume of 300 mL
each. As SDBS was used as the dispersing reagent in this study,
two controls, one with only sludge, and another one with sludge
plus 4 mg/g-TSS of SDBS, were conducted to investigate
whether methane production was affected by SDBS addition.
After flushed with nitrogen gas for 5 min to remove oxygen, all
bottles were capped with rubber stoppers, sealed and placed in
an air-bath shaker (150 rpm) at 35  1  C. Every day, 15 mL
fermentation mixture was manually withdrawn from each
serum bottle and the same amounts of raw sludge, SDBS and
ZnO NPs were supplemented, which resulted in a hydrolytic
retention time (HRT) or sludge retention time (SRT) of 20 d. The
sampling was operated in a glove box. After the reactors were
operated for 105 d, the daily methane production did not change
significantly with time, and then the analyses of methane
production, enzyme activity, biomass viability and microbial
community were conducted. The total gas volume was
measured by releasing the pressure in the bottles using a glass
syringe (100 mL) to equilibrate with the room pressure accord-
ing to our previous publication (Zhao et al., 2010).
5614 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0

Stock solution (50 mg/L) of ZnCl2 (Sigma-Aldrich) was SCFA compound) in each serum bottle. All other operations
prepared in 0.1 mM SDBS solution (pH 7.0). The long-term were the same as described above. By the analysis of methane
experiments of released Zn2þ from ZnO NPs affecting WAS production, the long-term effect of ZnO NPs on methanation
digestion were conducted with the same method described was obtained.
above except that the ZnCl2 was used to replace ZnO NPs, and
the total amount of Zn2þ added to the serum bottles was 1.2, 2.5. Analytical methods
11.6, and 17.6 mg/L, respectively.
Gas component was measured via a gas chromatograph
2.4. Effects of ZnO nanoparticles on each step involved (Agilent 6890N, USA) equipped with a thermal conductivity
in methane production detector using nitrogen as the carrier gas. The zinc concen-
tration was analyzed by ICP-OES (PerkinElmer Optima 2100
It is well known that sludge anaerobic digestion usually DV, USA). To measure the zinc content in sludge, the sample
undergoes solubilization of sludge particulate organic-carbon, was digested according to EPA Method 200.2 prior to ICP
hydrolysis, acidification and methanation. The experiments analyses. The pH value was measured by a pH meter. The
of long-term effects of ZnO NPs on these four stages were determinations of SCFA, protein, carbohydrate, TSS and VSS
conducted with the inoculum seeds from four long-term were the same as those described in the previous publication
operated reactors with ZnO NPs dosage of 0, 1, 30 and (Yuan, et al., 2006). The total SCFA was calculated as the sum
150 mg/g-TSS, respectively. The experiments of long-term of measured acetic, propionic, n-butyric, iso-butyric, n-valeric
effects of ZnO NPs on sludge particulate organic maters and iso-valeric acids. The COD (chemical oxygen demand)
solubilization were conducted as follows. WAS of 300 mL and conversion factors of protein, carbohydrate and SCFA were
30 mL inocula were added to each serum bottle. After flushed performed according to Grady et al. (1999). The detailed
with nitrogen gas for 5 min to remove oxygen, all bottles were analytical procedures of scanning electron microscopy (SEM),
capped with rubber stoppers, sealed and placed in an air-bath intracellular ROS, Cell counting kit-8 (CCK-8), FISH, protease,
shaker (150 rpm) at 35  1  C. The concentrations of soluble acetate kinase (AK) and coenzyme F420 activities are presented
protein and carbohydrate were measured after fermentation in Supplementary Information.
for 2 d.
As soluble protein and polysaccharide were the main 2.6. Statistical analysis
sludge solubilized products, in order to investigate the long-
term effects of ZnO NPs on the hydrolysis of sludge solubi- All assays were conducted in triplicate and the results were
lized products, the following batch tests with synthetic expressed as mean  standard deviation. An analysis of
wastewater containing bovine serum albumin (BSA, average variance (ANOVA) was used to test the significance of results
molecular weight Mw 67000, model protein compound used in and p < 0.05 was considered to be statistically significant.
this study) and dextran (Mww23800, model polysaccharide
compound) were conducted. The synthetic wastewater con-
Particulate organic
sisted of (mg/L of distilled water) 1000 KH2PO4, 400 CaCl2, 600
matters of
MgCl2$6H2O, 100 FeCl3, 0.5 ZnSO4$7H2O, 0.5 CuSO4$5H2O, 0.5 Solubilization
Solubilization waste active sludge
CoCl2$6H2O, 0.5 MnCl2$4H2O, 1 NiCl2$6H2O and 34.8 SDBS.
After 4.8 g BSA and 1.2 g dextran (the mass ratio of protein to Soluble polysaccharide
carbohydrate was almost the same as that in WAS) were
Soluble protein
dissolved in 1200 mL synthetic wastewater, the mixture liquid Hydrolysis protease Hydrolysis
was divided equally into 4 bottles, and then 30 mL inocula,
Amino acids Monosaccharide
which was heat-pretreated at 102  C for 30 min to kill
methanogens (Oh et al., 2003), was added before the pH in
Pyruvic acid
each bottle was adjusted to 7.0 by adding 4 M NaOH or 4 M HCl.
After flushed with nitrogen gas to remove oxygen, all bottles Acidification
were capped with rubber stoppers, sealed and placed in an
air-bath shaker (150 rpm) at 35  1  C. By analyzing Acetyl-CoA
the degradation efficiencies of protein and dextran, the long- AK
term effects of ZnO NPs on sludge hydrolysis were obtained.
The same operations were conducted when the long-term Propionic Acetic Butyric
effects of ZnO NPs on the acidification of hydrolyzed prod- acid acid acid
ucts were investigated except that the synthetic wastewater
containing 4.8 g L-glutamate (model amino acid compound) Methanation Carbon
and 1.2 g glucose (model monosaccharide compound). dioxide Hydrogen
F420
As acetic acid was the main short-chain fatty acid (SCFA) of
F420
sludge acidification product (Yuan et al., 2006) and the
Methane
preferred substrate for methane production (Fig. 1), the long-
term effect of ZnO NPs on methanation of the acidification Fig. 1 e Proposed metabolic pathway for methane
product was conducted with 300 mL synthetic wastewater (see production from WAS anaerobic digestion. Only the key
the above description) containing 0.72 g sodium acetate (model enzymes assayed in this study are labeled.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0 5615

TSS. Apparently, higher concentrations (30 and 150 mg/g-TSS)

Relativ e m eth an e p ro d uction ( % o f co ntro l)

100 * of ZnO NPs were capable of inhibiting the methane production.

* Some nanomaterials, such as fullerene, Au, Ag and Fe3O4,
were reported to give marginal influence on anaerobic
80 community (Barrena et al., 2009; Nyberg et al., 2008). Never-
theless, the Gram-positive B. subtilis was observed to be more
sensitive to ZnO NPs than Gram-negative E. coli. (Adams et al.,
60
2006), and ZnO NPs were found to negatively affect the soil
* bacterial community (Ge et al., 2011). It seems that it is difficult
40 to figure out the toxicity of ZnO NPs on microorganism
* involved in WAS digestion according to the current ZnO NPs
toxicology information as various species of bacteria are in
20 sludge anaerobic digestion system, and WAS anaerobic
digestion for methane production usually includes sludge
solubilization, hydrolysis, acidification and methanation
0
1 30 150 1.2 11.6 17.6 (Fig. 1). According to our knowledge, the effects of ZnO NPs on
the microbial community and each step involved in anaerobic
ZnO NPs (mg/g-TSS) Zn2+ (mg/L) digestion have never been documented, which will be inves-
Fig. 2 e Effects of different dosages of ZnO NPs (1, 30 and tigated in detail in the following text to understand the
150 mg/g-TSS) and the corresponding released Zn2D (1.2, mechanisms of ZnO NPs affecting methane production during
11.6 and 17.6 mg/L) on methane production during WAS WAS anaerobic digestion.
digestion. Asterisks indicate statistical differences
( p < 0.05) from the control. Error bars represent standard 3.2. Effects of ZnO nanoparticles on sludge surface and
deviations of triplicate tests. Zn2þ release as well as ROS change

The SEM analysis has been applied in literature to investigate

3. Results and discussions the adsorption of NPs to sludge (Kiser et al., 2009). As seen in
Fig. 3, there were large numbers of ZnO NPs on the surface of
3.1. Effect of ZnO nanoparticles on methane production sludge after long-term exposed to ZnO NPs. The same obser-
vations were reported by other researchers when the behavior
In this study, the addition of dispersing reagent (SDBS) at of NPs in wastewater treatment system was studied (Kiser
a dosage of 4 mg/g-TSS in sludge digestion experiments or et al., 2010; Limbach et al., 2008).
0.1 mM in synthetic wastewater tests was not observed to At ZnO NPs dosages of 1, 30 and 150 mg/g-TSS, respectively,
affect the methane production. This observation is consistent the corresponding released Zn2þ concentrations were 1.2, 11.6
with Garcia et al. (2006). In the coming text, the control repre- and 17.6 mg/L (Fig. S2, Supplementary Information). The long-
sents the reactor without ZnO NPs addition but with an SDBS term impact of released Zn2þ on methane production during
dosage of 4 mg/g-TSS in sludge digestion experiments or WAS anaerobic digestion is shown in Fig. 2. The presence of
0.1 mM in synthetic wastewater tests. As shown in Fig. 2, when 1.2 mg/L of Zn2þ did not give any significant impact on the
ZnO NPs were added to sludge fermentation system, their methane production ( p > 0.05). It might be that some chemical
influence on methane production was relevant to the dosage. compounds, such as sulfate (11.5 mg/L) in sludge, was bio-
At a lower ZnO NPs dosage (1 mg/g-TSS), no inhibitory converted to sulfide by sulfate reducing bacterial under anaer-
effect was observed (Fobserved ¼ 0.05, Fsignificance ¼ 7.71, obic conditions, and then the sulfide reacted with Zn2þ and thus
P (0.05) ¼ 0.83 > 0.05). When the dosage of ZnO NPs was 30 mg/g- reduced the toxicity of Zn2þ. However, the methane production
TSS, however, the average methane production decreased to was 90.6% of the control at a Zn2þ concentration of 11.6 mg/L.
81.7% of the control, which was further decreased to 24.9% of When the Zn2þ was 17.6 mg/L, a much lower methane
the control as the dosage of ZnO NPs increased to 150 mg/g- production (36.2% of the control) was observed. It can be seen

Fig. 3 e Scanning electron micrographs imaging of sludge long-term exposed to 0 mg/g-TSS (A), 1 mg/g-TSS (B), 30 mg/g-TSS
(C), and 150 mg/g-TSS (D) of ZnO NPs during WAS anaerobic digestion.
5616 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0

200 200
ROS production * 1 mg/g-TSS 30 mg/g-TSS 150 mg/g-TSS

Relativ e activ ity o f en zy m e ( % o f co nt r o l)

Relative biomass viability (% of control)
Biomass viability
100 *
Relative ROS production (% of control)

160 160
* *
80
*
120 120

* 60
80 80
*
40

40 40
20

0 0
1 30 150 0
ZnO NPs (mg/g-TSS) Protease AK F 420

Fig. 4 e Effects of different dosages of ZnO NPs (1, 30 and Fig. 6 e Comparisons of the activities of protease, AK and
150 mg/g-TSS) on the intracellular ROS production and coenzyme F420 in the long-term operated reactors exposed
biomass viability. Asterisks indicate statistical differences to different dosages of ZnO NPs (1, 30 and 150 mg/g-TSS).
( p < 0.05) from the control. Error bars represent standard Asterisks indicate statistical differences ( p < 0.05) from
deviations of triplicate tests. the control reactor. Error bars represent standard
deviations of triplicate tests.

A Protein Polysaccharide C Acetic Propionic

1500 iso-Butyric n-Butyric
100
Individual SCFA concentration
Concentrations (mg-COD/L)

80 1400
(mg-COD / L)

60
400

40

200
20

0 0

80
B BSA Dextran D
200 *
Cumulative methane production
Degradation efficiency (%)

* *
60
150
(mL / g-COD )

40 100

*
20 50

0 0
0 61 30 150 0 16 30 150
Innoculum sludge long-term exposed to different Innoculum sludge long-term exposed to different
dosages of ZnO NPs (mg/g-TSS) dosages of ZnO NPs (mg/g-TSS)

Fig. 5 e Effects of ZnO NPs on each step of sludge anaerobic digestion. A: the concentrations of soluble protein and
carbohydrate during the initial 2 d; B: the degradation of solubilized products (BSA and dextran) with time of 4 d; C: the
concentrations of acidification products (individual SCFA) with time of 4 d; D: the methanation products (methane) at time of
14 d. Asterisks indicate statistical differences ( p < 0.05) from the control. Error bars represent standard deviations of
triplicate tests.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0 5617

from Fig. 2 that the impact of ZnO NPs on methane production
mainly resulted from the dissolved Zn2þ. In a recent publication
Liu et al. (2011) also reported that the released Zn2þ from ZnO
NPs played an important role on the adverse effect of ZnO NPs
on the performance of biological wastewater treatment
process. In the literature the toxicity of ZnO NPs to some
microbes was also observed to come from the released Zn2þ, but
those studies focused on cell growth instead of microbial
function (Franklin et al., 2007; Wong et al., 2010; Xia et al., 2008).
The ROS induced by ZnO NPs was reported to be one reason
for their toxicity, which caused the loss of cell viability (Xia
et al., 2008). ZnO NPs were regarded as an exogenous source
of ROS for cells or organisms in some previous reports (Joshi et
al., 2009; Xia et al., 2008). As seen in Fig. 4, an increase of the
intracellular ROS production was observed with the increase
of ZnO NPs. Usually, ROS, including superoxide (O2-),
hydrogen peroxide (H2O2), and the hydroxyl radical (OH), are
produced in the presence of oxygen (Murphy, 2009). However,

Fig. 7 e Fluorescence in situ hybridization of sectios of biomass long-term (more than 105 d) cultured respectively in the
absence of ZnO NPs (A1-A3) and in the presence of 1 mg/gTSS (B1-B3), 30 mg/g-TSS (C1-C3) and 150 mg/g-TSS of ZnO NPs
(D1-D3) viewed by CLSM and photographed at higher (362) magnification. The sections were simultaneously hybridized
with Cy-3-labeled bacterial-domain probe (EUB338) (red) and FITC-labeled archaeal-domain probe (ARC915) (green). Overlay
of ARC915 (green) and EUB338 (red) are shown in A3, B3, C3 and D3. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this article.)
5618 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0
it has been reported that H2O2 can also be produced under
anaerobic conditions (Degli-Esposti and McLennan, 1998). The
increase of ROS in the sludge exposed to higher dosages of
ZnO NPs was a likely reason for their adverse effect on sludge
anaerobic digestion. It can be seen from Fig. 4 that the result of
biomass viability assay was consistent with the ROS produc-
tion, which decreased from 97.3% of the control ( p > 0.05) to
88.7% of the control when ZnO NPs increased from 1 to 30 mg/
g-TSS. At ZnO NPs dosage of 150 mg/g-TSS, the relative
biomass viability further decreased to 62.4% of the control.
3.3. Effects of ZnO nanoparticles on each step involved
in methane production
Protein and carbohydrate, the main constituents of WAS
(accounting for 61.5% of sludge TCOD), are usually in partic-
ulate state. The batch experiments were conducted to inves-
tigate the long-term effects of ZnO NPs on sludge particulate
protein and carbohydrate solubilization. The effects of ZnO
NPs on solubilization of sludge particulate organic matters
were expressed by the changes of soluble protein and carbo-
hydrate production in this study. As seen from Fig. 5A, there
were no significant differences in the concentrations of
soluble protein and carbohydrate after the initial 2 d
fermentation ( p > 0.05). It might be that the solubilization of
sludge particulate organic matters was not a microbial
process, which resulted in the influences of ZnO NPs on the
concentrations of soluble protein and carbohydrate not being
observed.
The long-term effects of three dosages of ZnO NPs on the
hydrolysis of sludge solubilized products (soluble protein and
carbohydrate) with time of 4 d are shown in Fig. 5B. The
degradation of dextran (model carbohydrate mater) in the
control reactor was almost the same as those in other three
ZnO NPs reactors. Nevertheless, the influence of ZnO NPs on
the degradation of BSA (model protein) was dosage depen-
dent. At dosages of 1 and 30 mg/g-TSS, the influences of ZnO
NPs were insignificant ( p > 0.05), but the degradation of BSA
at 150 mg/g-TSS of ZnO NPs was lower than that in the
control (58.5% versus 65.1%). It might be one reason for the
decreased methane production exposed to higher concen-
trations of ZnO NPs.
Fig. 5C illustrates the long-term effects of different
concentrations of ZnO NPs on the acidification of main
hydrolyzed products (amino acid and monosaccharide) to
SCFA during the initial 4 d. The influences of ZnO NPs on the
composition of SCFA were insignificant (see Table S1 for
statistical analysis, Supplementary Information). The total
SCFA concentrations, which calculated from Fig. 5C, were
2078  80, 2057  80, 2045  69 and 2050  50 mg-COD/L in the
reactors of control, and 1, 30 and 150 mg/g-TSS of ZnO NPs,
respectively. Obviously, the acidification step involved in
sludge digestion was not affected by ZnO NPs.
As to the influence of ZnO NPs on the methanation step, the
data in Fig. 5D indicated that there was no significant differ-
ence in the cumulative methane production between the
control and the 1 mg/g-TSS of ZnO NPs reactors at time of 14 d
( p > 0.05). However, the methane productions were 83.0% and
28.1% of the control at 30 and 150 mg/g-TSS of ZnO NPs,
respectively, suggesting that ZnO NPs significantly inhibited
the bio-conversion step of acetic acid to methane. At other time
the same observations could be made (Fig. S3, Supplementary
Information). By comparing the data in Fig. 5C and D, it
seems that methanogens are more sensitive to the toxicity of
ZnO NPs than acidogens. In the literature, some researchers
reported that acidogens were more resistant to metal toxicity
than methanogens (Zayed and Winter, 2000). In addition, the
data in Fig. 5AeD suggested that the negative influence of ZnO
NPs on the methanation step was the most serious one among
the four steps.
3.4. Determination of key enzyme activity
Further investigation showed that ZnO NPs influenced the
activities of enzymes relevant to sludge anaerobic digestion.
Although large numbers of enzymes took part in methane
production during sludge anaerobic digestion, in this study
only three enzymes responsible respectively for sludge
hydrolysis (i.e., protease), acidification (AK) and methanation
(coenzyme F420) (Fig. 1) were assayed as examples. The relative
activities of these enzymes in the long-term operated reactors
are demonstrated in Fig. 6. The AK activity did not change
significantly with ZnO NPs dosages ( p > 0.05). To the protease,
the dosage of 150 mg/g-TSS of ZnO NPs remarkably reduced its
activity. The coenzyme F420 activity, however, was ZnO NPs
dosage dependent, which was respectively 99.3%, 89.8% and
66.2% of the control at ZnO NPs of 1, 30 and 150 mg/g-TSS.
Apparently, not only the hydrolysis of soluble protein but the
transformation activity of electron donors of the redox-driven
proton translocation in methanogenic Archaea (expressed by
coenzyme F420 (Deppenmeier, 2002)) was significantly
restrained by higher concentrations of ZnO NPs. All these
consisted well with the above observed synthetic wastewater
experimental results.
3.5. FISH analysis results
For the purpose of investigating the influence of ZnO NPs on
the abundance of bacteria and Archaea, the FISH assay was
further conducted (Fig. 7), and the results were analyzed with
image analysis system (Image-Pro Plus, V6.0). It was found
that there were 39.5% of Archaea and 52.6% of bacteria in the
control reactor. In ZnO NPs reactors, the Archaea were 38.6% (1
mg/g-TSS), 27.1% (30 mg/g-TSS), and 3.5% (150 mg/g-TSS), and
the corresponding bacteria were 51.3%, 60.8% and 87.4%,
respectively. The ratios of Archaea to bacteria were 0.8:1, 0.9:1,
0.4:1 and 0.04:1, respectively, in the reactors of control, and 1,
30 and 150 mg/g-TSS of ZnO NPs, respectively. Obviously, the
more ZnO NPs appeared in sludge anaerobic digestion system,
the less Archaea remained, which was consistent with the
observed methane production when WAS was long-term
exposed to different dosages of ZnO NPs.
4. Conclusions
The above studies indicated that the methane production
during sludge anaerobic digestion was not affected by ZnO
NPs of 1 mg/g-TSS. Nevertheless, due to large numbers of Zn2þ
release the higher dosages (30 and 150 mg/g-TSS) of ZnO NPs
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0
inhibited the production of methane. By investigating the four
stages involved in sludge anaerobic digestion, i.e., solubiliza-
tion, hydrolysis, acidification and methanation, it was found
that the activities of protease and coenzyme F420 were nega-
tively influenced by higher dosages of ZnO NPs, which sug-
gested that only the steps of hydrolysis and methanation were
inhibited. The molecular biology studies indicated that
a lower abundance of methanogenesis Archaea was observed
at higher ZnO NPs dosages exposure.
Acknowledgements
This work was financially supported by the Foundation of
State Key Laboratory of Pollution Control and Resource Reuse
(PCRRK09002).
Appendix. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.022.
references
Adams, L.K., Lyon, D.Y., Alvarez, P.J.J., 2006. Comparative eco-
toxicity of nanoscale TiO2, SiO2, and ZnO water suspensions.
Water Research 40 (19), 3527e3532.
Barrena, R., Casals, E., Colon, J., Font, X., Sanchez, A., Puntes, V.,
2009. Evaluation of the ecotoxicity of model nanoparticles.
Chemosphere 75 (7), 850e857.
Brar, S.K., Verma, M., Tyagi, R.D., Surampalli, R.Y., 2010. Engineered
nanoparticles in wastewater and wastewater sludge-Evidence
and impacts. Waste Management 30 (3), 504e520.
Degli-Esposti, M, McLennan, H, 1998. Mitochondria and cells
produce reactive oxygen species in virtual anaerobiosis:
relevance to ceramide-induced apoptosis. FEBS Letters 430 (3),
338e342.
Deppenmeier, U., 2002. Redox-driven proton translocation in
methanogenic archaea. Cellular and Molecular Life Sciences
59 (9), 1513e1533.
Ellsworth, D.K., Verhulst, D., Spitler, T.M., Sabacky, B.J., 2000.
Titanium nanoparticles move to the marketplace. Chemical
Innovation 30 (12), 30e35.
EPA, 2009. Targeted National Sewage Sludge Survey Sampling and
Analysis Technical Report. Environmental Protection Agency,
Washington, DC: United States. http://www.epa.gov/
waterscience/biosolids/tnsss-tech.pdf.
Franklin, N.M., Rogers, N.J., Apte, S.C., Batley, G.E., Gadd, G.E.,
Casey, P.S., 2007. Comparative toxicity of nanoparticulate
ZnO, bulk ZnO, and ZnCl2 to a freshwater microalga
(Pseudokirchneriella subcapitata): the importance of particle
solubility. Environmental Science and Technology 41 (24),
8484e8490.
Ganesh, R., Smeraldi, J., Hosseini, T., Khatib, L., Olson, B.H.,
Rosso, D., 2010. Evaluation of nanocopper removal and
toxicity in municipal wastewaters. Environmental Science
and Technology 44 (20), 7808e7813.
Garcia, M.T., Campos, E., Sanchez-Leal, J., Ribosa, I., 2006. Effect of
linear alkylbenzene sulphonates (LAS) on the anaerobic
digestion of sewage sludge. Water Research 40 (15),
2958e2964.
5619
Ge, Y., Schimel, J.P., Holden, P.A., 2011. Evidence for negative
effects of TiO2 and ZnO nanoparticles on soil bacterial
communities. Environmental Science and Technology 45 (4),
1659e1664.
Gottschalk, F., Sonderer, T., Scholz, R.W., Nowack, B., 2009. Modeled
environmental concentrations of engineered nanomaterials
(TiO2, ZnO, Ag, CNT, Fullerenes) for different regions.
Environmental Science and Technology 43 (24), 9216e9222.
Grady, C.P.L., Daigger, G.T., Lim, H.C., 1999. Biological Wastewater
Treatment. Marcel Dekker Inc., New York.
Jiang, W., Mashayekhi, H., Xing, B., 2009. Bacterial toxicity
comparison between nano- and micro-scaled oxide particles.
Environmental Pollution 157 (5), 1619e1625.
Joshi, P., Chakraborti, S., Chakrabarti, P., Haranath, D.,
Shanker, V., Ansari, Z.A., Singh, S.P., Gupta, V., 2009. Role of
surface adsorbed anionic species in antibacterial activity of
ZnO quantum dots against Escherichia coli. Journal of
Nanoscience and Nanotechnology 9 (11), 6427e6433.
Keller, A.A., Wang, H.T., Zhou, D.X., Lenihan, H.S., Cherr, G.,
Cardinale, B.J., Miller, R., Ji, Z.X., 2010. Stability and aggregation
of metal oxide nanoparticles in natural aqueous matrices.
Environmental Science and Technology 44 (6), 1962e1967.
Kiser, M.A., Ryu, H., Jang, H.Y., Hristovski, K., Westerhoff, P., 2010.
Biosorption of nanoparticles to heterotrophic wastewater
biomass. Water Research 44 (14), 4105e4114.
Kiser, M.A., Westerhoff, P., Benn, T., Wang, Y., Perez-Rivera, J.,
Hristovski, K., 2009. Titanium nanomaterial removal and
release from wastewater treatment plants. Environmental
Science and Technology 43 (17), 6757e6763.
Li, M., Zhu, L., Lin, D., 2011. Toxicity of ZnO nanoparticles to
Escherichia coli: mechanism and the influence of medium
components. Environmental Science and Technology 45 (5),
1977e1983.
Limbach, L.K., Bereiter, R., Mueller, E., Krebs, R., Gaelli, R.,
Stark, W.J., 2008. Removal of oxide nanoparticles in a model
wastewater treatment plant: influence of agglomeration and
surfactants on clearing efficiency. Environmental Science and
Technology 42 (15), 5828e5833.
Liu, G.Q., Wang, D.M., Wang, J.M., Mendoza, C., 2011. Effect of ZnO
particles on activated sludge: Role of particle dissolution.
Science of the Total Environment 409 (14), 2852e2857.
Maynard, A.D., Aitken, R.J., Butz, T., Colvin, V., Donaldson, K.,
Oberdorster, G., Philbert, M.A., Ryan, J., Seaton, A., Stone, V.,
Tinkle, S.S., Tran, L., Walker, N.J., Warheit, D.B., 2006. Safe
handling of nanotechnology. Nature 444 (7117), 267e269.
Miziolek, A., 2002. Nanoenergetics: an emerging technology area
of national importance. AMPTIAC Quarterly 6 (1), 43e48.
Murphy, M.P., 2009. How mitochondria produce reactive oxygen
species. Biochemical Journal 417, 1e13.
Nyberg, L., Turco, R.F., Nies, L., 2008. Assessing the impact of
nanomaterials on anaerobic microbial communities.
Environmental Science and Technology 42 (6), 1938e1943.
Oh, S.E., Van Ginkel, S., Logan, B.E., 2003. The relative
effectiveness of pH control and heat treatment for enhancing
biohydrogen gas production. Environmental Science and
Technology 37 (22), 5186e5190.
Roco, M.C., 2005. The emergence and policy implications of
converging new technologies integrated from the nanoscale.
Journal of Nanoparticle Research 7 (2e3), 129e143.
Serda, R.E., Ferrati, S., Godin, B., Tasciotti, E., Liu, X.W., Ferrari, M.,
2009. Mitotic trafficking of silicon microparticles. Nanoscale 1
(2), 250e259.
Sharma, V., Shukla, R.K., Saxena, N., Parmar, D., Das, M.,
Dhawan, A., 2009. DNA damaging potential of zinc oxide
nanoparticles in human epidermal cells. Toxicology Letters
185 (3), 211e218.
Simon-Deckers, A., Loo, S., Mayne-L’Hermite, M., Herlin-
Boime, N., Menguy, N., Reynaud, C., Gouget, B., Carriere, M.,
5620 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 1 2 e5 6 2 0
2009. Size-, composition- and shape-dependent toxicological
impact of metal oxide nanoparticles and carbon nanotubes
toward bacteria. Environmental Science and Technology 43
(21), 8423e8429.
Wong, S.W.Y., Leung, P.T.Y., Djurisic, A.B., Leung, K.M.Y., 2010.
Toxicities of nano zinc oxide to five marine organisms:
influences of aggregate size and ion solubility. Analytical and
Bioanalytical Chemistry 396 (2), 609e618.
Xia, T., Kovochich, M., Liong, M., Madler, L., Gilbert, B., Shi, H., Yeh, J.,
Zink, J.I., Nel, A.E., 2008. Comparison of the mechanism of toxicity
of zinc oxide and cerium oxide nanoparticles based on dissolution
and oxidative stress properties. ACS Nano 2 (10), 2121e2134.
Yuan, H., Chen, Y., Zhang, H., Jiang, S., Zhou, Q., Gu, G., 2006.
Improved bioproduction of short-chain fatty acids (SCFAs)
from excess sludge under alkaline conditions. Environmental
Science and Technology 40 (6), 2025e2029.
Zayed, G., Winter, J., 2000. Inhibition of methane production from
whey by heavy metals - protective effect of sulfide. Applied
Microbiology and Biotechnology 53 (6), 726e731.
Zhao, Y., Chen, Y., Zhang, D., Zhu, X., 2010. Waste activated
sludge fermentation for hydrogen production enhanced by
anaerobic process improvement and acetobacteria inhibition:
the role of fermentation pH. Environmental Science and
Technology 44 (9), 3317e3323.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Genetic potential for N2O emissions from the sediment

of a free water surface constructed wetland

Arantzazu Garcı́a-Lledó a, Ariadna Vilar-Sanz a, Rosalia Trias a, Sara Hallin b,

Lluı́s Bañeras a,*
a
Molecular Microbial Ecology Group, Institute of Aquatic Ecology, Universitat de Girona, C/ Maria Aurèlia Capmany, 69, 17071 Girona, Spain
b
Swedish University of Agricultural Sciences, Department of Microbiology, Box 7025, Uppsala 750 07, Sweden

article info abstract

Article history: Removal of nitrogen is a key aspect in the functioning of constructed wetlands. However,
Received 13 April 2011 incomplete denitrification may result in the net emission of the greenhouse gas nitrous
Received in revised form oxide (N2O) resulting in an undesired effect of a system supposed to provide an ecosystem
22 July 2011 service. In this work we evaluated the genetic potential for N2O emissions in relation to the
Accepted 14 August 2011 presence or absence of Phragmites and Typha in a free water surface constructed wetland
Available online 30 August 2011 (FWS-CW), since vegetation, through the increase in organic matter due to litter degra-
dation, may significantly affect the denitrification capacity in planted areas. Quantitative
Keywords: real-time PCR analyses of genes in the denitrification pathway indicating capacity to
Constructed wetlands produce or reduce N2O were conducted at periods of different water discharge. Genetic
Denitrification potential for N2O emissions was estimated from the relative abundances of all denitrifi-
Genetic potential cation genes and nitrous oxide reductase encoding genes (nosZ ). nosZ abundance was
N2O emission invariably lower than the other denitrifying genes (down to 100 fold), and differences
qPCR increased significantly during periods of high nitrate loads in the CW suggesting a higher
Vegetation genetic potential for N2O emissions. This situation coincided with lower nitrogen removal
efficiencies in the treatment cell. The presence and the type of vegetation, mainly due to
changes in the sediment carbon and nitrogen content, correlated negatively to the ratio
between nitrate and nitrite reducers and positively to the ratio between nitrite and nitrous
oxide reducers. These results suggest that the potential for nitrous oxide emissions is
higher in vegetated sediments.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
Constructed wetlands (CWs) are recognized as feasible alter-
natives for removal of nitrogen in wastewater from agricul-
tural, industrial or municipal activities and have been
exploited as secondary or tertiary treatment alternatives to
promote water reuse (Reed et al., 1995; DeBusk and DeBusk,
2000). There are two types of CWs, the subsurface flow
systems (SSF) and the free-water surface systems (FWS) that
mainly differ in the presence of a free water flow over the
sediment surface. Water treatment in CWs involves processes
driven by the sediment, the vegetation and its associated
microbial communities (EPA, 2000). It is generally assumed
that areas planted with emergent macrophytes positively
affect the water restoration capacity of the wetland (Zhu and
Sikora, 1995; Lin et al., 2002; Ibekwe et al., 2007). This is due to
stimulation of microbial growth at either the epiphyton or the
root surface, e.g. by influencing the oxygen conditions and

* Corresponding author. Tel.: þ34 972 418 177; fax: þ34 972 418 150.
E-mail addresses: arantzazu.garcia@udg.edu (A. Garcı́a-Lledó), ariadna.vilar@udg.edu (A. Vilar-Sanz), rosalia.trias@udg.edu (R. Trias),
Sara.Hallin@slu.se (S. Hallin), lluis.banyeras@udg.edu (L. Bañeras).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.025
5622 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2
creating microenvironments with oxiceanoxic zones
(Gersberg et al., 1986; Zhu and Sikora, 1995; Boudraa et al.,
1999) or by the excretion of organic compounds from the
roots (Prade and Trolldenier, 1990; Brix, 1997; Nguyen, 2003),
as well as causing changes in hydraulic retention time
(Whitney et al., 2003; Toet et al., 2005; Kjellin et al., 2007). All
these processes influence nitrogen removal in CWs since they
affect the nitrifying and denitrifying microorganisms, which
are ultimately responsible for nitrogen transformations
resulting in nitrogen removal (Johnston, 1991; Zhu and Sikora,
1995; Boudraa et al., 1999; Körner, 1999; Vymazal, 2001; Lin
et al., 2002; Francis et al., 2007).
The microbial process denitrification is especially inter-
esting in CWs as it represents the net nitrogen loss from the
system. Denitrifying bacteria use nitrate as an alternative
electron acceptor, which is sequentially reduced to nitrogen
gas. The first step is catalyzed by nitrate reductases, either
a membrane-bound enzyme (Nar type) or a soluble peri-
plasmic enzyme (Nap) (Zumft, 1997; Moreno-Vivian et al.,
1999; Philippot, 2002). These two reductases are not exclusive,
but can be found simultaneously in the same organism (Carter
et al., 1995; Gregory et al., 2003; Roussel-Delif et al., 2005).
Dissimilatory nitrite reductases (Nir) catalyze the second step
of denitrification and two functionally equivalent enzymes
have been described, a cytochrome cd1 type and a copper
containing type, encoded by the nirS and nirK genes, respec-
tively. They are mutually exclusive and have not been found
in the same strain so far (Zumft, 1997), although different
strains of the same species may contain different Nir genes
(Coyne et al., 1989; Philippot, 2002). Nitric oxide reductases
(Nor) catalyze the reduction of nitric oxide to nitrous oxide
and the last step of denitrification is catalyzed by nitrous oxide
reductases (Nos) which lead to the production of nitrogen gas
(Zumft, 1997). All denitrifying genes described so far have
been used as molecular markers for qualitative and quanti-
tative studies of denitrifying bacteria in the environment. The
expression of denitrifying genes is dependent on the presence
of the respective enzymes substrates, i.e. sequential oxidized
forms of inorganic nitrogen, and low oxygen concentration
due to the facultative nature of this process (Tiedje, 1988).
Most denitrifiers use organic compounds as electron donors
and easily available carbon is another prerequisite.
Emissions of the potent greenhouse gas N2O may be due to
the activity of some denitrifying bacteria that are unable to
perform the final step of denitrification, due to the lack of the
nosZ gene (Wood et al., 2001; Kandeler et al., 2006; Vial et al.,
2006; Jones et al., 2008; Abell et al., 2010; Philippot et al.,
2011). Cheneby et al., 2004 compared denitrifying bacteria
between non-planted and maize planted soil and found that
nosZ lacking bacteria were dominant in the rhizosphere sug-
gesting that plants can affect N2O emission by selecting for
denitrifiers that do not have the capacity to reduce N2O. Plants
increase easily available carbon compounds by root exudates
and through the increase in organic matter due to litter
degradation. In addition, plants cause radial oxygen gradients
around the roots. These factors affect denitrifying bacteria, as
well as other heterotrophic bacteria (Boudraa et al., 1999;
Henry et al., 2008). Studies specifically directed to the anal-
ysis of denitrification in the rhizosphere have shown that
changes in the activity, composition and the abundance of
denitrifying bacteria may be plant-specific (Cheneby et al.,
2004; Henry et al., 2008; Ruiz-Rueda et al., 2009). In the sedi-
ment of planted areas the presence of decomposing litter may
be an additional factor supporting and shaping the denitrify-
ing community (Ingersoll and Baker, 1998). If emergent
macrophytes in CWs also select for denitrifiers genetically
incapable of reducing N2O is not known. If so, although
promoting nitrogen removal in CWs, macrophytes potentially
contribute to increase the N2O/N2 end-product ratio.
We evaluated the genetic potential for N2O emissions of
the sediment in relation to the presence or absence of the
main plant species used in treatment wetlands, Phragmites
australis and Typha latifolia. For that purpose, we determined
the abundance of denitrifiers by quantitative real-time PCR
(qPCR) of key functional genes in the denitrification pathway
(narG, napA, nirS, nirK and nosZ ) in comparison to the total
bacterial community size targeted by the 16S rRNA gene in the
sediments of the Empuriabrava FWS-CWs at periods of
different water discharge over the year.
2. Material and methods
2.1. Study site
The Empuriabrava FWS-CWs (Girona, NE Spain) were
designed in 1998 as a tertiary treatment to increase the water
quality of the effluent of a nearby located wastewater treat-
ment plant (WWTP). The CWs are included in the natural
preserved area of Els Aiguamolls de l’Empordà and are designed
to provide additional water to avoid excessive desiccation of
the flooded area in summer (Ruiz-Rueda et al., 2007). Nitrogen
is removed at the WWTP by a combined nitrifica-
tionedenitrification process using carrousel-type bioreactors
and enhanced aeration, resulting in nitrate loads to the CWs
changing between 251 and 1016 kg N per month. The
Empuriabrava CWs consists of three parallel cells followed by
a shallower lagoon and all measurements were conducted on
treatment cell 3. The cell is planted with independent
communities of reed (P. australis) and cattail (T. latifolia)
showing an almost equivalent distribution of both plant
species (49.3% of reed and 50.7% of cattail; Fig. 1). The average
water depth in the cell was about 0.6 m and the sediment was
water saturated during the sampling period. The sediment
depths varied between 5 and 10 cm in non-vegetated areas
and increased to 15e20 cm in planted surfaces. The location of
the vegetation spots and the plant species were designed in
the original project and planted shortly after construction of
the CW (1998). Harvest of the aerial biomass is performed
every one or two years during winter.
2.2. Sampling and DNA extraction
Temperature, conductivity, oxygen and pH at the sampling
locations were measured with a portable multiparametric
probe (Yellow Spring Instruments 650MDS). Water samples
(20 ml) from the water column in the wetland were collected
and analyzed for nitrate, nitrite and ammonia concentration
as described previously (Garcı́a-Lledó et al., 2011). Water
samples (100 ml) were regularly also collected at the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2
Fig. 1 e Scheme of vegetation stands and sampling locations in the third cell of the Empuriabrava FWS-CWs. Sampling plots
are indicated as black dots with the following two-letter codes: SE e bulk sediment, TY e areas planted with Typha latifolia,
PH e areas planted with Phragmites australis.
secondary effluent of the WWTP and at the end point of the
treatment cells for analyses of pH, conductivity, biological
oxygen demand (BOD), chemical oxygen demand (COD),
oxygen saturation, and concentrations of nitrate, nitrite and
ammonium, using conventional standard methods for
wastewater analyses (APHA, 1998). Total water flow into the
wetland system was automatically monitored daily.
Sediment samples were collected May 2008, August 2008,
March 2009 and July 2009 near the inlet and outlet areas. Three
sampling plots of one square meter each were defined at the
two areas. Sampling plot 1 was not covered with any emergent
plant and is referred as bulk sediment (SE). Plot 2 consisted of
an area covered with T. latifolia (TY) and Plot 3 was covered
with P. australis (PH). Three sediment replicates were taken in
each plot (Fig. 1).
Sediment cores were obtained from planted and unplanted
areas using a 2-cm-diameter methacrylate tube mounted in
a manual sampler. The upper 3 cm of the sediment core were
aseptically transferred to a container and chilled on ice for
transportation. Once in the laboratory, sediment was
completely homogenized and triplicates of 2 g aliquots were
stored at 80  C until processed.
Total nucleic acids were extracted using a modified CTAB
protocol previously described for the simultaneous recovery
of DNA and RNA from soils (Hurt et al., 2001). Purification of
DNA was done with AllPrep DNA/RNA Mini Kit (Qiagen)
according to the manufacturer’s instructions. DNA extraction
and quality was checked with a 0.8% agarose gel. DNA
concentration was quantified with a NanoVue Plus Spec-
trophotometer (GE Healthcare).
2.3. Sediment chemical analyses
The content of TC and TN in the sediment were analyzed by
combustion of dried samples at 975  C in a Perkin Elmer AE
SeriesII equipped with a TCD detector. The results were
evaluated using the K factor method with cystine
(C6H12N2O4S2) as a standard. Duplicates were always per-
formed for all chemical determinations. Due to the low
nitrogen content of the sediment, in most cases near the
detection limit, total Kjeldahl nitrogen (TKN) was also
5623
measured using standard methods for sediments (APHA,
1998). Sediment pH was measured from 1/5 (dry weight/
volume) sediment samples homogenized with double distilled
water.
2.4. Quantitative PCR of functional genes
Bacteria involved in the denitrification processes were esti-
mated by the quantification of key functional genes using
quantitative real-time PCR. The 16S rRNA gene was also
quantified to estimate the total amount of bacteria. Primers
and thermal cycling conditions used for each reaction have
previously been described for narG and napA (Bru et al., 2007)
and for nirS, nirK, nosZ and 16S rRNA genes (Hallin et al., 2009).
All qPCR reactions were performed on a Bio-Rad IQ5 thermal
cycler (Bio-Rad Laboratories, Inc.) in a total volume of 20 ml,
containing 1X Phire Hot Start II DNA Polymerase (Finnzymes
Oy. Espoo, Finland), 1 mM of each primer, 103 ng ml1 of Bovine
Serum Albumin (BSA) and 1 ng of DNA template. Results were
analyzed using Bio-Rad IQ5 software. Standard curves were
obtained using serial dilutions from 102 to 108 copies of line-
arized plasmids containing the respective functional genes.
Controls without templates gave null or negligible values. To
ensure that sediment samples did not have inhibitory effects
on PCR performance an inhibitory test were run with all
samples at the working concentration together with a known
amount of circular plasmid. The measured threshold cycle (Ct)
values were compared with those of a control of the plasmid
mixed with water. Despite the use of highly diluted DNA
extractions (down to 1 ng), inhibition effects could not be
removed from one sample of a Phragmites covered sediment
collected in May 2008 and three samples from sediments
covered with Typha collected in August 2008 and March 2009.
These samples have been removed from the statistical
analyses.
2.5. Statistical analyses
All statistical analyses were performed using SPSS for
Windows 15.0 (SPSS, Inc). Measured gene abundances were
log transformed in order to ensure a normal distribution of
5624 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2
data, which was checked by KolmogoroveSmirnov and Sha-
piroeWilk tests. Differences in gene abundances or gene
ratios were tested for the effects of planted vs. non-planted
sediments, sampling time, location of sampling plots and the
corresponding interactions by one-way ANOVA and Welch
tests. Differences between means of either gene were
analyzed using GameseHowell test (for unequal variance
data) post hoc analyses for multiple comparisons. To test for
effects of sampling period, type of sediment, position inside
the treatment cell and interactions on denitrifying commu-
nity, univariate general linear model (GLM) analyses were
performed. Correlation analysis of microbial abundances with
physical and chemical variables was performed using non-
parametric Spearman’s correlation test. Paired samples t-test
was chosen to compare differences between genes coding for
enzymes catalyzing the same reaction (i.e. narG vs. napA). The
significance level for all tests was 0.05.
3. Results
3.1. Characterization of the Empuriabrava FWS-CWs
During the studied period, the water inflow to the wetlands
presented stable pH conditions, with values around 7.6. BOD
was low, usually under 3 mg O2/l, except for occasional dates
when higher concentrations were recorded (Fig. 2). COD varied
from 30 to 121 mg O2/l, but no seasonal trends were observed
in the variation. Ammonia concentrations were low, with
slightly higher values between July and September (up to
4.5 mg NeNHþ 4 /l). In contrast, nitrate concentrations showed
a higher variation, between 0.1 and 15.8 mg NeNO 3 /l, with
concentration peaks occurring intermittently throughout the
year due to changes in nitrogen removal efficiencies in the
WWTP. The water flowing through the wetlands had strong
seasonal variations, from 1500 m3/day (February) to almost
6000 m3/day (August). This large variation is due to the intense
tourism in a nearby located touristic area that increases water
consumption during summer. Physicochemical parameters of
the water and sediment measured at the sampling locations
are presented in Table 1.
Water conductivity values varied from 2.7 (August 2008) to
7.8 mV/cm (March 2009). There was a significant negative
correlation between conductivity and water temperature
(Spearman’s correlation coefficient r ¼ 0.641, p < 0.001,
results not shown). The higher conductivity values during
winter are due to the infiltration of sea water in some of the
wastewater network collectors, which is diluted during
summer due to a higher wastewater volume to be treated at
the WWTP. The water flow to the wetland varied significantly
during the four sampling periods, receiving 2437 and 2202 m3/
day in May 2008 and March 2009, and 6007 and 4379 m3/day in
August 2008 and July 2009, respectively. Hydraulic retention
times (HRT) for treatment cell 3 were calculated during the
sampling periods by independent measurements of the water
flow, with values ranging from 4.2 to 15.5 days. Unfortunately,
the water flow was unexpectedly cut for a short period in July
2009 and these conditions are not indicative of the average
situation in the cell. To circumvent this problem, the monthly
average water flow to the wetland system was used in the
statistical analyses.
Differences in both water and sediment pH were not
extremely large and values ranged from 7.2 to 8.7 and from 6.9
to 7.9, respectively. Water nitrate and ammonium concen-
trations at the sampling locations ranged between 0.01 and
1.24 mg NeNO þ
3 /l and 0.06 and 0.49 mg NeNH4 /l, respectively.
Nitrite concentrations were not detectable. Positive correla-
tions were obtained between conductivity and nitrate
(r ¼ 0.350, p < 0.01) or ammonium (r ¼ 0.275, p < 0.05). Higher
TC and TKN values were usually obtained in sediments
collected from areas planted with P. australis ( p < 0.001).
Nitrogen removal efficiencies (%) were calculated on the
basis on the total nitrogen load and changes in the inlet and
outlet concentrations in treatment cell 3 (Garcı́a-Lledó et al.,
2011). Efficiencies were estimated to be 48 and 61% in
periods of low water discharge (May 2008 and March 2009,
respectively) and increased up to 71% when higher inflows
entered the CWS (July 2009). Nitrogen removal efficiencies
correlated positively with ammonia concentration (r ¼ 0.685,
p < 0.01) and negatively with nitrate concentrations
(r ¼ 0.640, p < 0.01) at the influent. Moreover, oxygen and
temperature showed a negative (r ¼ 0.484, p < 0.01) and
positive (r ¼ 0.703, p < 0.01) correlation to nitrogen removal
efficiencies, respectively.
3.2. Abundance of 16S rRNA and denitrification genes
The abundance of 16S rRNA genes ranged from 1.3  1011 to
5.8  1012 copies/g dw sediment and appeared to be fairly
stable among samples and over time (Fig. 3; Supplementary
Table 1). These numbers were always higher than those
obtained for any of the functional genes. Mean values for all
determined functional gene abundances were between 1 and
4 logs below the 16S rRNA gene.
The narG and napA genes ranged from 8.8  108 to 7.1  1010
and from 1.2  109 to 6.4  1010 copies/g dw sediment,
respectively. Mean napA gene numbers were significantly
higher than narG according to a paired sample t-test
(t ¼ 2.578, p < 0.05), but these differences were exclusively
due to dominance of napA genes in bulk sediments in March
2009 and no differences were detected when the March
samples were excluded (t ¼ 1.116, p > 0.05).
Sediment samples showed between 4.9  107 and
6.9  109 copies/g dw sediment for nirS and from 1.6  109 to
2.1  1011 copies/g dw sediment for nirK, suggesting a signifi-
cantly greater abundance of nirK-type denitrifiers (t ¼ 28.828,
p < 0.001). Finally, nosZ ranged from 3.5  107 to
1.3  109 copies/g dw sediment, and were significantly lower
than both nirS and nirK (t ¼ 5.184, p < 0.001 and t ¼ 51.770,
p < 0.001, respectively).
3.3. Factors determining the denitrifying community
size
Univariate general linear model tests (GLM) were used to
analyze the effect of time (sampling period), type of sediment,
sampling position in the treatment cell and their interactions,
as defined factors to explain differences in gene abundances.
Highly significant effects ( p < 0.001) were found when
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2 5625

Fig. 2 e Time variation for the 2008e2009 period of the water flow, concentrations of BOD and COD and ammonia and nitrate
at the inlet of the constructed wetland. Sediment sampling dates are indicated with arrows.

sampling period and types of sediment were considered
(Table 3). The abundance of the nirS gene was the only
exception and was not influenced by the presence of
vegetation. Additionally, the sampling position only had an
effect on 16S rRNA and narG ( p < 0.05). All possible interac-
tions between factors had no significant effect on gene
abundances, except for nitrate reductase genes in some cases.
Nevertheless, partial Eta squared values (%) for these inter-
actions were rather low, indicating a low significance in the
overall variance of the considered genes. Accordingly, in
subsequent analyses only the sampling time and the type of
sediment were considered.
In all cases, the lowest abundance of all analyzed genes
was found in samples obtained during August 2008, except for
the nitrate reductase gene narG (Fig. 3AeC). In addition, the
presence of vegetation, especially of Phragmites, yielded
a significant increase for all genes but napA and nirS according
to a non-parametric Welch test (F > 3.671, p < 0.05) (Fig. 3DeF).
Differences in the abundance of napA could only be assessed
by a post hoc analysis and resulted in higher values in sedi-
ments also covered with Phragmites. The presence of Typha did
not cause any clear influence on the gene abundances and,
depending on the genetic marker considered, was similar to
either the bulk sediment or the one with Phragmites.
In order to analyze differences in the gene abundances
according to physicochemical characteristics of the sediment
and water, pair-wise correlation tests were performed. nosZ
abundance, being indicative of capacity for N2O reduction,
5626 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2

Table 1 e Chemical properties of the sediment and overlying water of the different sampling locations and periods. Data
from sediments correspond to mean values and standard deviations of three replicate samples.
Type of Date Water characteristics Sediment properties
sediment 
Temp ( C) Cond (mS/cm) O2 (mg/l) pH NeNO
3 (mg/l) NeNHþ
4 (mg/l) pH TC (%) TKN (%)

SE inlet 15-May-08 20.7 3.8 6.1 7.6 0.45 0.37 7.4  0.1 4.0  0.2 0.21  0.02
TY inlet 21-May-08 20.4 3.6 7.6 8.0 1.07 0.21 7.7  0.1 5.6  1.1 0.30  0.06
PH inlet 20-May-08 20.0 3.5 8.8 8.0 1.24 0.21 7.4  0.1 13.5  1.9 1.23  0.08

SE inlet 20-Aug-08 24.8 2.7 4.3 7.7 0.08 0.21 7.7  0.0 3.8  0.4 0.16  0.02
SE outlet 11-Aug-08 25.9 2.8 7.8 8.0 0.01 0.08 7.7  0.1 3.3  0.4 0.09  0.02
TY inlet 21-Aug-08 25.3 2.8 5.4 7.9 0.12 0.27 7.9  0.2 4.2  0.7 0.18  0.10
PH inlet 18-Aug-08 24.9 3.1 5.2 7.9 0.21 0.12 7.9  0.0 4.3  0.3 0.23  0.04
PH outlet 12-Aug-08 26.6 2.9 3.2 8.1 0.01 0.06 7.8  0.1 5.3  0.2 0.32  0.07

SE inlet 9-Mar-09 12.6 7.1 6.1 8.4 0.26 0.19 6.9  0.1 4.2  0.2 0.07  0.04
SE outlet 3-Mar-09 12.2 7.8 7.8 8.3 0.40 0.25 7.2  0.1 3.9  0.2 0.13  0.05
TY inlet 11-Mar-09 13.3 7.2 5.5 8.1 0.11 0.17 7.0  0.1 4.9  0.4 0.16  0.08
TY outlet 4-Mar-09 11.1 7.8 6.8 8.3 0.33 0.27 7.0  0.1 4.3  1.4 0.16  0.16
PH inlet 10-Mar-09 14.0 7.2 5.6 8.3 0.19 0.11 7.4  0.1 9.8  0.8 0.90  0.11
PH outlet 5-Mar-09 10.0 7.6 5.7 8.3 0.57 0.22 7.3  0.4 8.7  0.2 0.54  0.22

SE inlet 8-Jul-09 22.6 4.5 3.6 7.3 0.18 0.32 7.7  0.2 5.7  1.0 0.47  0.19
SE outlet 7-Jul-09 25.5 4.7 3.9 7.7 0.08 0.49 7.2  0.1 3.8  0.1 0.18  0.04
TY inlet 13-Jul-09 26.4 4.6 14.9 8.7 0.03 0.28 7.6  0.3 5.0  1.3 0.29  0.22
TY outlet 14-Jul-09 26.1 4.7 10.2 8.3 0.03 0.15 7.2  0.1 4.8  0.9 0.29  0.18
PH inlet 10-Jul-09 22.4 4.4 3.0 7.7 0.03 0.25 7.1  0.2 12.0  2.9 1.04  0.67
PH outlet 16-Jul-09 27.6 4.5 3.8 8.2 0.03 0.11 7.5  0.2 9.0  0.7 0.83  0.21

SE, bulk sediment; TY, areas planted with Typha latifolia; PH, areas planted with Phragmites australis.

correlated positively with sediment TC (r ¼ 0.661, p < 0.01),
TKN (r ¼ 0.612, p < 0.01) and negatively with the water flow
and COD (r ¼ 0.330, p < 0.05; r ¼ 0.277, p < 0.05, respectively;
Table 2). For the other genes, both sediment carbon and
nitrogen content also correlated positively and significantly
with abundances and negative correlations were observed
when the water flow was considered, except for narG. The
nitrate content in the water was also an important factor that
specifically correlated with nitrite reductase genes, nirS and
nirK, and the nitrate reductase napA. Surprisingly, no signifi-
cant correlations were found for any of the genes and BOD in
the influent water or sediment pH, with the exception of water
pH and napA.
3.4. Relative abundance of denitrification genes
The variation in the relative proportion of genes coding for
enzymes catalyzing different steps in the denitrification can
be used to assess removal efficiencies in terms of potential
accumulation of intermediates in the denitrification pathway.
Variation of calculated ratios according to time (sampling
period), type of sediment, sampling position in the treatment
cell and the corresponding interactions was estimated by GLM
(Table 4). The relative contribution of the functional genes
(narG, napA, nirS, nirK and nosZ ) to the total bacteria pop-
ulation, represented by 16S rRNA did not vary according to the
presence of vegetation ( p > 0.05). When sampling period was
considered, significant variations could be found for the
relative amount of the functional genes narG, napA and nosZ
( p < 0.05) (results not shown). Interesting results can be
deduced when ratios between two functional genes are
calculated. The ratio between nitrate and nitrite reductases
(qnarG þ qnapA)/(qnirS þ qnirK ) varied from 0.2 to 5.4 and was
significantly higher in samples obtained in August 2008
independently of the presence of vegetation or not ( p < 0.05).
TC and TKN in the sediment were negatively correlated with
(qnarG þ qnapA)/(qnirS þ qnirK ) (r ¼ 0.366, p < 0.01 and
r ¼ 0.261, p < 0.05, respectively). The ratio (qnirS þ qnirK )/
qnosZ was significantly influenced by the sampling period
( p < 0.05) and the interaction between sampling period and
type of vegetation. The highest difference between the abun-
dance of nitrite reductases (qnirS þ qnirK ) and nitrous oxide
reductase (qnosZ ) was found in March and May, when higher
nitrate concentrations were measured, and accounted for
almost two orders of magnitude. When qnirK/qnosZ and qnirS/
qnosZ ratios were analyzed separately, a significant effect of
sampling time was found for both of them ( p < 0.05), but the
interaction between sampling time and vegetation was only
found significant for the former ( p < 0.05). Relevant variables
affecting the ratio between nitrite reducers and nitrous oxide
reducers were temperature (r ¼ 0.332, p < 0.05), nitrate
content in water (r ¼ 0.314, p < 0.05) and total carbon in
sediment (r ¼ 0.309, p < 0.05). Finally, the (qnarG þ qnapA)/
qnosZ ratio was fairly stable between the four sampling
periods ( p > 0.05) and no effect of the a priori defined factors
was detected.
4. Discussion
The location of samples was a minor factor related to the
abundance of denitrifying genes in the sediment of the
treatment wetland. Instead, sampling period and type of
sediment, mainly differing by nutrient availability, seems to
be more important factors controlling the total abundance of
denitrification genes in this system. Nevertheless, data from
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2 5627

Fig. 3 e Mean values of the number of copies of the bacterial 16S rRNA and the functional genes, narG, napA, nirS, nirK and nosZ
according to sampling period (AeC) and the type of sediment (DeF). Standard errors of the mean are indicated. Different letters
above the bars indicate significant differences ( p < 0.05) between sampling period (left) or type of sediment (right).

qPCR must be examined carefully since bacteria may have
more than one copy of the same gene per genome. Bacterial
genomes can harbor up to 13 copies of the 16S rRNA gene
(Fogel et al., 1999). The denitrification genes most often only
exist in one copy, although two or three copies have been
shown for all denitrification genes (Philippot, 2002; Jones et al.,
2008, 2011). Despite these limitations, qPCR studies provide
a realistic quantification of the size of the denitrification gene
pool and ratios between gene pools can be used to infer
community dynamics among samples of similar characteris-
tics (Henry et al., 2006; Kandeler et al., 2006; Geets et al., 2007;
Enwall et al., 2010).
The relative abundance of functional gene densities in
relation to the bacterial 16S rRNA gene abundance indicated
that the denitrifying and nitrate respiring bacteria in the
sediment constituted a fraction of 1.6  0.4 and 1.2  0.7% of
the bacterial communities, respectively. The values we
obtained in the sediment of Empuriabrava FWS-CWs
compared well with those found in the only wetland sedi-
ment characterized in terms of denitrifier abundances so far
(Chon et al., 2011) and studies performed in soils (e.g. Henry
et al., 2006; Kandeler et al., 2006; Hallin et al., 2009). The
strong correlation between narG and napA genes and the fact
that both have similar ratios to the 16S rRNA gene, suggests
a high proportion of Proteobacteria in the sediment, since both
narG and napA gene can be found in members of this phylum
(Philippot and Hojberg, 1999; Roussel-Delif et al., 2005; Bru
et al., 2007). Accordingly, quantitative analyses of bacteria at
5628 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2

Table 2 e Spearman correlation coefficients (r) between environmental variables in the wetland sediment and water,
influent water and abundances of the 16S rRNA and functional genes.
q16S rRNA qnarG qnapA qnirS qnirK qnosZ

Wetland characteristics Water

Temp ( C) 0.186 0.076 0.282 0.230 0.248 0.139
ns ns * ns ns ns
Cond (mS/cm) 0.356 0.174 0.407 0.241 0.387 0.341
** ns ** ns ** **
O2 (mg/l) 0.167 0.070 0.134 0.102 0.086 0.113
ns ns ns ns ns ns
RedOx (mV) 0.114 0.043 0.002 0.210 0.124 0.098
ns ns ns ns ns ns
pH 0.141 0.071 0.281 0.073 0.182 0.195
ns ns * ns ns ns
NeNO
3 (mg N/l) 0.308 0.205 0.325 0.366 0.351 0.224
* ns * ** ** ns
NeNHþ
4 (mg N/l) 0.179 0.177 0.240 0.200 0.168 0.155
ns ns ns ns ns ns
Sediment
TC (%) 0.712 0.544 0.428 0.613 0.658 0.661
** ** ** ** ** **
TKN (%) 0.653 0.573 0.395 0.539 0.600 0.612
** ** ** ** ** **
C:N 0.582 0.546 0.34 0.468 0.526 0.561
** ** ** ** ** **
pH 0.251 0.062 0.219 0.143 0.234 0.225
ns ns ns ns ns ns
Influent characteristics Water flow (m3/day) 0.421 0.160 0.361 0.384 0.443 0.330
** ns ** ** ** *
COD (mg O2/l) 0.376 0.134 0.254 0.346 0.356 0.227
** ns ns ** ** *
BOD (mg O2/l) 0.047 0.078 0.112 0.099 0.107 0.11
ns ns ns ns ns ns
NO
3 (mg N/l) 0.243 0.074 0.285 0.282 0.310 0.182
ns ns * * * ns
NHþ
4 (mg N/l) 0.168 0.120 0.235 0.123 0.230 0.116
ns ns ns ns ns ns

ns, not significant; *p < 0.05; **p < 0.01.

the inlet and outlet of WWTPs in other studies show a clear
dominance of Proteobacteria (Juretschko et al., 2002; Chouari
et al., 2010; McLellan et al., 2010).
The denitrifier community, assessed by the quantification
of nirS and nirK genes, was dominated by NirK-type denitri-
fiers, which were up to 2 logs more abundant than the NirS-
type. This contrasts with previous studies in the same
wetland using a PCR-TRFLP approach in which positive nirK
PCR amplifications could not be obtained for any of the sedi-
ment samples (Ruiz-Rueda et al., 2007). This discrepancy is
likely due to significant changes in the quality of the water
entering the Empuriabrava FWS-CWs due to reconstructions
in June 2007 (Garcı́a-Lledó et al., 2011). Even though the two
nitrite reductases are functionally equivalent, denitrifiers
harboring either nitrite reductase seem to show a preference
for certain environments and are likely not under the same
community assembly rules (Jones and Hallin, 2010). Recent
studies in agricultural soils also suggested that the existence
of the two types of nitrite reductase is due to differential niche
preferences (Philippot et al., 2009; Enwall et al., 2010).
However, the stability of the present physicochemical condi-
tions of the influent to the Empuriabrava FWS-CWs and in the
sediments prevents us from finding the environmental drivers
affecting the qnirS/qnirK ratio in this environment, other than
that the sediment and water provided conditions that favor
the NirK-type denitrifiers.
In agreement with our results, the nosZ abundance in soils
is often lower than that of other denitrifying genes (Henry
et al., 2006; Hallin et al., 2009; Philippot et al., 2009). The nosZ
gene encodes the final step of denitrification, which makes the
net nitrogen removal from CWs all the way to N2 possible, but
denitrifier communities with low nosZ to nir gene ratios can
result in increased N2O/(N2 þ N2O) end-product ratios
(Philippot et al., 2011). The lack of nitrous oxide reductase in
some denitrifying bacteria was reported a decade ago when
the Agrobacterium tumefaciens genome was sequenced (Wood
et al., 2001) and recently it has been shown that approxi-
mately 1/3 of genome sequenced denitrifying bacterial
isolates have this truncated pathway (Jones et al., 2008). The
relatively low abundance of nosZ genes in the treatment
wetland was especially pronounced in August 2008, coin-
ciding with low nitrate levels and an increase in the nitrogen
load in the form of ammonia to the FWS-CWs. In contrast to
the other periods, the ammonium was also decreasing
between the inlet and outlet of the different vegetation plots
(Table 1). At the same time, the relative abundance of the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2 5629

Table 3 e Results of general linear model (GLM) analyses for the effect of the three factors established (sampling period, type
of sediment and sampling location) and their interactions with the abundances of 16S rRNA and functional genes.
Time Sed Loc

F p Eta F p Eta F p Eta

q16S rRNA 23.21 *** 0.65 16.49 *** 0.46 6.01 * 0.14
qnarG 4.27 * 0.25 11.93 *** 0.39 6.77 * 0.15
qnapA 6.76 ** 0.35 10.04 *** 0.35 3.41 ns 0.08
qnirS 11.13 *** 0.47 2.87 ns 0.13 2.25 ns 0.06
qnirK 14.42 *** 0.53 10.96 *** 0.37 3.49 ns 0.08
qnosZ 10.48 *** 0.45 14.52 *** 0.43 3.18 ns 0.08

Time  Sed Time  Loc Sed  Loc Time  Sed  Loc

F p Eta F p Eta F p Eta F p Eta

q16S rRNA 1.67 ns 0.21 1.69 ns 0.08 1.65 ns 0.08 1.11 ns 0.08
qnarG 2.71 * 0.30 2.15 ns 0.10 1.08 ns 0.05 2.69 ns 0.18
qnapA 1.38 ns 0.18 3.92 * 0.17 0.88 ns 0.04 3.00 * 0.19
qnirS 1.03 ns 0.14 0.31 ns 0.02 0.63 ns 0.03 0.84 ns 0.06
qnirK 2.23 ns 0.26 1.14 ns 0.06 1.26 ns 0.06 0.27 ns 0.02
qnosZ 1.06 ns 0.14 1.14 ns 0.06 0.56 ns 0.03 2.54 ns 0.17

ns, not significant; *p < 0.05; **p < 0.01; ***p < 0.001 Time, sampling period; Sed, type of sediment; Loc, sample location in the treatment cell. Eta,
partial Eta Squared.

nitrate reductases, assessed by narG and napA genes, was
higher compared to the other sampling occasions. These
results suggest that the conditions favored bacteria capable of
dissimilatory nitrate reduction to ammonium in addition to
denitrifiers that do not have the complete denitrification
pathway. Higher potential rates of nitrate reduction to
ammonia were measured during August 2008 in bulk sedi-
ments and areas covered with P. australis thus confirming the
previous results (Garcı́a-Lledó et al., 2011).
Conditions that invariably will affect denitrification
activity, such as temperature and lower nitrate to carbon
ratios varied during the studied period. Relatively high nitrate
loads to the Empuriabrava FWS-CW occurred intermittently
but frequently throughout the year. Although only two of
these periods, May 2008 and March 2009, were sampled in this
study, the increase in the measured gene abundances let us
hypothesize that the probability of N2O production in the
Empuriabrava FWS-CWs would increase in relation to the
nitrate concentration. However, additional measurements in
summer coinciding with the increase of nitrate at the inlet
would be needed to confirm this hypothesis. Other studies
conducted on CWs mesocosms have reported significantly
higher N2O emissions during summer (Sovik and Klove, 2007).
In the Empuriabrava FWS-CWs significant differences of
the genetic potential for N2O emissions according to plant
species were only detected when interactions with sampling

Table 4 e Results of general linear model (GLM) analyses for the effect of the three factors established (sampling period, type
of sediment and sampling location) and their interactions with the ratios between the functional genes involved in
denitrification process.
Time Sed Loc

F p Eta F p Eta F p Eta

(qnarG þ qnapA)/(qnirS þ qnirK ) 3.64 * 0.23 1.06 ns 0.05 0.15 ns 0.00

(qnirS þ qnirK )/qnosZ 3.81 * 0.24 0.62 ns 0.03 0.27 ns 0.01
qnirS/qnirK 4.16 * 0.25 4.79 * 0.21 0.00 ns 0.00
qnirS/qnosZ 4.14 * 0.25 1.57 ns 0.08 0.18 ns 0.01
qnirK/qnosZ 3.76 * 0.23 0.73 ns 0.04 0.29 ns 0.01
(qnarG þ qnapA)/qnosZ 1.17 ns 0.09 0.31 ns 0.17 1.08 ns 0.03

Time  Sed Time  Loc Sed  Loc Time  Sed  Loc

F p Eta F p Eta F p Eta F p Eta

(qnarG þ qnapA)/(qnirS þ qnirK ) 1.91 ns 0.24 1.12 ns 0.06 1.15 ns 0.06 1.74 ns 0.12
(qnirS þ qnirK)/qnosZ 2.88 * 0.32 0.11 ns 0.01 2.09 ns 0.10 1.68 ns 0.12
qnirS/qnirK 0.88 ns 0.13 1.85 ns 0.09 0.29 ns 0.02 2.22 ns 0.15
qnirS/qnosZ 1.55 ns 0.20 0.34 ns 0.02 0.68 ns 0.04 0.50 ns 0.04
qnirK/qnosZ 2.91 * 0.32 0.17 ns 0.01 2.17 ns 0.11 1.83 ns 0.13
(qnarG þ qnapA)/qnosZ 1.29 ns 0.17 2.82 ns 0.13 0.27 ns 0.01 0.29 ns 0.02

ns, not significant; *, p < 0.05; Time, sampling period; Sed, type of sediment; Loc, sample location in the treatment cell. Eta, partial Eta Squared.
5630 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2
time were considered. However, according to the calculated
(qnirS þ qnirK )/qnosZ ratio, the highest genetic potential
would be found mainly in vegetated sediments where higher
nitrogen availability occurred due to leaf litter decomposition.
Others have measured higher N2O fluxes in mesocosms
planted with P. australis when compared to other macrophytes
and unvegetated sediments (Maltais-Landry et al., 2009).
Nevertheless, the effect of planted areas for the production of
N2O in wetlands has been poorly studied, and further research
in this direction is still needed to fully understand the
dynamics of constructed wetlands in relation to greenhouse
gas emissions.
5. Conclusions
The Empuriabrava FWS-CWs showed variations in physico-
chemical parameters analyzed both in water and sediment
between the different periods. These conditions favored the
maintenance of a denitrifying community that significantly
changed according to nutrient concentration and the types of
vegetation. The low abundance of nosZ genes compared with
the other denitrification genes is an indicative of the genetic
capacity of the system to potentially accumulate the N2O
intermediary and the relative proportion of nosZ genes
decreased during periods of high nitrate content to the
wetlands. Overall, the quantitative data on denitrification
gene abundances provide evidence of a high potential for
nitrous oxide emissions along the entire sediment surface and
in particular during periods of high nitrate loading. The
vegetation effect was mainly detected in combination with
sampling time and resulted in an increase of the potential for
nitrous oxide emissions in vegetated areas. The increase in
the nitrite to nitrous oxide reductase genes ratio has been
related to the higher total carbon content in these sediments.
Acknowledgments
A.G-L and A.V-S. are recipients of pre-doctoral grants from the
Ministerio de Ciencia y Educación and the Universitat de
Girona, respectively. The authors thank the contributions of
Anna Huguet, Jordi Sala and Lluı́s Sala for field analyses. This
research has been funded by the Spanish Ministerio de Cien-
cia y Educación (grant CGL2009-08338).
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.025.
references
Abell, G.C., Revill, A.T., Smith, C., Bissett, A.P., Volkman, J.K.,
Robert, S.S., 2010. Archaeal ammonia oxidizers and nirS-type
denitrifiers dominate sediment nitrifying and denitrifying
populations in a subtropical macrotidal estuary. The ISME
Journal 4 (2), 286e300.
APHA, A.a.W., 1998. Standard Methods for the Examination of
Water and Wastewater. American Public Health Association,
Washington.
Boudraa, A.O., Champier, J., Djebali, M., Behloul, F., Beghdadi, A.,
Bachand, P.A.M., Horne, A.J., 1999. Denitrification in
constructed free-water surface wetlands: II. Effects of
vegetation and temperature. Ecological Engineering 14, 17e32.
Brix, H., 1997. Do macrophytes play a role in constructed treatment
wetlands? Water Science and Technology 35, 11e17.
Bru, D., Sarr, A., Philippot, L., 2007. Relative abundances of
proteobacterial membrane-bound and periplasmic nitrate
reductases in selected environments. Applied and
Environmental Microbiology 73 (18), 5971e5974.
Carter, J.P., Hsaio, Y.H., Spiro, S., Richardson, D.J., 1995. Soil and
sediment bacteria capable of aerobic nitrate respiration.
Applied and Environmental Microbiology 61 (8), 2852e2858.
Cheneby, D., Perrez, S., Devroe, C., Hallet, S., Couton, Y.,
Bizouard, F., Iuretig, G., Germon, J.C., Philippot, L., 2004.
Denitrifying bacteria in bulk and maize-rhizospheric soil:
diversity and N2O-reducing abilities. Canadian Journal of
Microbiology 50 (7), 469e474.
Chon, K., Chang, J.-S., Lee, E., Lee, J., Ryu, J., Cho, J., 2011.
Abundance of denitrifying genes coding for nitrate (narG),
nitrite (nirS), and nitrous oxide (nosZ) reductases in estuarine
versus wastewater effluent-fed constructed wetlands.
Ecological Engineering 37 (1), 64e69.
Chouari, R., Le Paslier, D., Daegelen, P., Dauga, C., Weissenbach, J.,
Sghir, A., 2010. Molecular analyses of the microbial
community composition of an anoxic basin of a municipal
wastewater treatment plant reveal a novel lineage of
proteobacteria. Microbial Ecology 60 (2), 272e281.
Coyne, M.S., Arunakumari, A., Averill, B.A., Tiedje, J.M., 1989.
Immunological identification and distribution of dissimilatory
heme cd1 and nonheme copper nitrite reductases in
denitrifying bacteria. Applied and Environmental
Microbiology 55 (11), 2924e2931.
DeBusk, T., DeBusk, W., 2000. In: Kent, D.M. (Ed.), Applied
Wetlands Science and Technology, second ed. CRC Press, Boca
Raton, Florida (USA), pp. 241e279.
Enwall, K., Throback, I.N., Stenberg, M., Soderstrom, M., Hallin, S.,
2010. Soil resources influence spatial patterns of denitrifying
communities at scales compatible with land management.
Applied and Environmental Microbiology 76 (7), 2243e2250.
EPA, 2000. Guiding Principles for Constructed Treatment Wetlands
Providing for Water Quality and Wildlife Habitat. United States
Environmental Protection Agency, Office of Water.
Fogel, G.B., Collins, C.R., Li, J., Brunk, C.F., 1999. Prokaryotic
genome size and SSU rDNA copy number: estimation of
microbial relative abundance from a mixed population.
Microbial Ecology 38 (2), 93e113.
Francis, C.A., Beman, J.M., Kuypers, M.M., 2007. New processes
and players in the nitrogen cycle: the microbial ecology of
anaerobic and archaeal ammonia oxidation. The ISME Journal
1 (1), 19e27.
Garcı́a-Lledó, A., Ruiz-Rueda, O., Vilar-Sanz, A., Sala, L.,
Bañeras, L., 2011. Nitrogen removal efficiencies in a free water
surface constructed wetland in relation to plant coverage.
Ecological Engineering 37, 678e684.
Geets, J., de Cooman, M., Wittebolle, L., Heylen, K., Vanparys, B.,
De Vos, P., Verstraete, W., Boon, N., 2007. Real-time PCR assay
for the simultaneous quantification of nitrifying and
denitrifying bacteria in activated sludge. Applied Microbiology
and Biotechnology 75 (1), 211e221.
Gersberg, R.M., Elkins, B.V., Lyon, S.R., Goldman, C.R., 1986. Role
of aquatic plants in wastewater treatment by artificial
wetlands. Water Research 20 (3), 363e368.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2
Gregory, L.G., Bond, P.L., Richardson, D.J., Spiro, S., 2003.
Characterization of a nitrate-respiring bacterial community
using the nitrate reductase gene (narG) as a functional marker.
Microbiology 149 (1), 229e237.
Hallin, S., Jones, C.M., Schloter, M., Philippot, L., 2009.
Relationship between N-cycling communities and ecosystem
functioning in a 50-year-old fertilization experiment. The
ISME Journal 3 (5), 597e605.
Henry, S., Bru, D., Stres, B., Hallet, S., Philippot, L., 2006.
Quantitative detection of the nosZ gene, encoding nitrous
oxide reductase, and comparison of the abundances of 16S
rRNA, narG, nirK, and nosZ genes in soils. Applied and
Environmental Microbiology 72 (8), 5181e5189.
Henry, S., Texier, S., Hallet, S., Bru, D., Dambreville, C.,
Cheneby, D., Bizouard, F., Germon, J.C., Philippot, L., 2008.
Disentangling the rhizosphere effect on nitrate reducers and
denitrifiers: insight into the role of root exudates.
Environmental Microbiology 10 (11), 3082e3092.
Hurt, R.A., Qiu, X., Wu, L., Roh, Y., Palumbo, A.V., Tiedje, J.M.,
Zhou, J., 2001. Simultaneous recovery of RNA and DNA from
soils and sediments. Applied and Environmental Microbiology
67 (10), 4495e4503.
Ibekwe, A.M., Lyon, S.R., Leddy, M., Jacobson-Meyers, M., 2007.
Impact of plant density and microbial composition on water
quality from a free water surface constructed wetland. Journal
of Applied Microbiology 102 (4), 921e936.
Ingersoll, T.L., Baker, L.A., 1998. Nitrate removal in wetland
microcosms. Water Research 32 (3), 677e684.
Johnston, C.A., 1991. Sediment and nutrient retention by
freshwater wetlands: effects on surface water quality. Critical
Reviews in Environmental Control 21 (5), 491e565.
Jones, C.M., Hallin, S., 2010. Ecological and evolutionary factors
underlying global and local assembly of denitrifier
communities. The ISME Journal 4 (5), 633e641.
Jones, C.M., Stres, B., Rosenquist, M., Hallin, S., 2008. Phylogenetic
analysis of nitrite, nitric oxide, and nitrous oxide respiratory
enzymes reveal a complex evolutionary history for
denitrification. Molecular Biology and Evolution 25 (9), 1955e1966.
Jones, C.M., Welsh, A., Throback, I.N., Dorsch, P., Bakken, L.R.,
Hallin, S., 2011. Phenotypic and genotypic heterogeneity
among closely related soil-borne N2- and N2O-producing
Bacillus isolates harboring the nosZ gene. FEMS Microbiology
Ecology 76 (3), 541e552.
Juretschko, S., Loy, A., Lehner, A., Wagner, M., 2002. The microbial
community composition of a nitrifyingedenitrifying activated
sludge from an industrial sewage treatment plant analyzed by
the full-cycle rRNA approach. Systematic and Applied
Microbiology 25 (1), 84e99.
Kandeler, E., Deiglmayr, K., Tscherko, D., Bru, D., Philippot, L., 2006.
Abundance of narG, nirS, nirK, and nosZ genes of denitrifying
bacteria during primary successions of a glacier foreland.
Applied and Environmental Microbiology 72 (9), 5957e5962.
Kjellin, J., Hallin, S., Worman, A., 2007. Spatial variations in
denitrification activity in wetland sediments explained by
hydrology and denitrifying community structure. Water
Research 41 (20), 4710e4720.
Körner, S., 1999. Nitrifying and denitrifying bacteria in epiphytic
communities of submerged macrophytes in a treated
sewage channel. Acta hydrochimica et hydrobiologica 27 (1),
27e31.
Lin, Y.F., Jing, S.R., Wang, T.W., Lee, D.Y., 2002. Effects of
macrophytes and external carbon sources on nitrate removal
from groundwater in constructed wetlands. Environmental
Pollution 119 (3), 413e420.
Maltais-Landry, G., Maranger, R., Brisson, J., Chazarenc, F., 2009.
Greenhouse gas production and efficiency of planted and
artificially aerated constructed wetlands. Environmental
Pollution 157 (3), 748e754.
5631
McLellan, S.L., Huse, S.M., Mueller-Spitz, S.R.,
Andreishcheva, E.N., Sogin, M.L., 2010. Diversity and
population structure of sewage-derived microorganisms in
wastewater treatment plant influent. Environmental
Microbiology 12 (2), 378e392.
Moreno-Vivian, C., Cabello, P., Martinez-Luque, M., Blasco, R.,
Castillo, F., 1999. Prokaryotic nitrate reduction: molecular
properties and functional distinction among bacterial nitrate
reductases. Journal of Bacteriology 181 (21), 6573e6584.
Nguyen, C., 2003. Rhizodeposition of organic C by plants:
mechanisms and controls. Agronomie 23 (5e6), 375e396.
Philippot, L., 2002. Denitrifying genes in bacterial and Archaeal
genomes. Biochimica et Biophysica Acta 1577 (3), 355e376.
Philippot, L., Andert, J., Jones, C.M., Bru, D., Hallin, S., 2011.
Importance of denitrifiers lacking the genes encoding the
nitrous oxide reductase for N2O emissions from soil. Global
Change Microbiology 17, 1497e1504.
Philippot, L., Cuhel, J., Saby, N.P., Cheneby, D., Chronakova, A., Bru, D.,
Arrouays, D., Martin-Laurent, F., Simek, M., 2009. Mapping field-
scale spatial patterns of size and activity of the denitrifier
community. Environmental Microbiology 11 (6), 1518e1526.
Philippot, L., Hojberg, O., 1999. Dissimilatory nitrate reductases in
bacteria. Biochimica et Biophysica Acta 1446 (1e2), 1e23.
Prade, K., Trolldenier, G., 1990. Denitrification in the rhizosphere
of rice and wheat seedlings as influenced by plant K status,
air-filled porosity and substrate organic matter. Soil Biology
and Biochemistry 22 (6), 769e773.
Reed, S.C., Crites, R.W., Middlebrooks, E.J., 1995. Natural Systems for
Waste Management and Treatment. McGraw-Hill, New York.
Roussel-Delif, L., Tarnawski, S., Hamelin, J., Philippot, L.,
Aragno, M., Fromin, N., 2005. Frequency and diversity of
nitrate reductase genes among nitrate-dissimilating
Pseudomonas in the rhizosphere of perennial grasses grown
in field conditions. Microbial Ecology 49 (1), 63e72.
Ruiz-Rueda, O., Hallin, S., Baneras, L., 2009. Structure and
function of denitrifying and nitrifying bacterial communities
in relation to the plant species in a constructed wetland. FEMS
Microbiology Ecology 67 (2), 308e319.
Ruiz-Rueda, O., Trias, R., Garcia-Gil, L.J., Baneras, L., 2007.
Diversity of the nitrite reductase gene nirS in the sediment of
a free-water surface constructed wetland. International
Microbiology 10 (4), 253e260.
Sovik, A.K., Klove, B., 2007. Emission of N2O and CH4 from
a constructed wetland in southeastern Norway. Science of the
Total Environment 380 (1e3), 28e37.
Tiedje, J., 1988. Ecology of denitrification and dissimilatory nitrate
reduction to ammonium. In: Zehnder, A.J.B. (Ed.), Biology of
Anaerobic Microorganisms. Wiley, New York, pp. 179e244.
Toet, S., Bouwman, M., Cevaal, A., Verhoeven, J.T., 2005. Nutrient
removal through autumn harvest of Phragmites australis and
Thypha latifolia shoots in relation to nutrient loading in
a wetland system used for polishing sewage treatment plant
effluent. Journal of Environmental Science and Health, Part A
40 (6), 1133e1156.
Vial, L., Lavire, C., Mavingui, P., Blaha, D., Haurat, J., Moenne-
Loccoz, Y., Bally, R., Wisniewski-Dye, F., 2006. Phase variation
and genomic architecture changes in Azospirillum. Journal of
Bacteriology 188 (15), 5364e5373.
Vymazal, J., 2001. Types of constructed wetlands for wastewater
treatment: their potential for nutrient removal. In: Vymazal, J.
(Ed.), Transformations of Nutrients in Natural and
Constructed Wetlands. Backhuys Publishers, Leiden, The
Netherlands, pp. 1e93.
Whitney, D., Rossman, A., Hayden, N., 2003. Evaluating an
existing subsurface flow constructed wetland in Akumal,
Mexico. Ecological Engineering 20 (1), 105e111.
Wood, D.W., Setubal, J.C., Kaul, R., Monks, D.E., Kitajima, J.P.,
Okura, V.K., Zhou, Y., Chen, L., Wood, G.E., Almeida Jr., N.F.,
5632 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 2 1 e5 6 3 2

Woo, L., Chen, Y., Paulsen, I.T., Eisen, J.A., Karp, P.D., Bovee Sr., D.,
Chapman, P., Clendenning, J., Deatherage, G., Gillet, W., Grant, C.,
Kutyavin, T., Levy, R., Li, M.J., McClelland, E., Palmieri, A.,
Raymond, C., Rouse, G., Saenphimmachak, C., Wu, Z., Romero, P.,
Gordon, D., Zhang, S., Yoo, H., Tao, Y., Biddle, P., Jung, M.,
Krespan, W., Perry, M., Gordon-Kamm, B., Liao, L., Kim, S.,
Hendrick, C., Zhao, Z.Y., Dolan, M., Chumley, F., Tingey, S.V.,
Tomb, J.F., Gordon, M.P., Olson, M.V., Nester, E.W., 2001. The
genome of the natural genetic engineer Agrobacterium tumefaciens
C58. Science 294 (5550), 2317e2323.
Zhu, T., Sikora, F.J., 1995. Ammonium and nitrate removal in
vegetated and unvegetated gravel bed microcosm wetlands.
Water Science and Technology 32 (3), 219e228.
Zumft, W.G., 1997. Cell biology and molecular basis of
denitrification. Microbiology and Molecular Biology Reviews 61
(4), 533e616.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Optimization of the coagulation-flocculation process for pulp

mill wastewater treatment using a combination of uniform
design and response surface methodology

Jian-Ping Wang a,b, Yong-Zhen Chen a, Yi Wang b, Shi-Jie Yuan a, Han-Qing Yu a,*
a
Department of Chemistry, University of Science and Technology of China, Hefei, 230026, China
b
Department of Agricultural and Biological Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA

article info abstract

Article history: Pulp mill wastewater was treated using the coagulation-flocculation process with
Received 7 June 2011 aluminum chloride as the coagulant and a modified natural polymer, starch-g-PAM-g-
Received in revised form PDMC [polyacrylamide and poly (2-methacryloyloxyethyl) trimethyl ammonium chloride],
6 August 2011 as the flocculant. A novel approach with a combination of response surface methodology
Accepted 14 August 2011 (RSM) and uniform design (UD) was employed to evaluate the effects and interactions of
Available online 1 September 2011 three main influential factors, coagulant dosage, flocculant dosage and pH, on the treat-
ment efficiency in terms of the supernatant turbidity and lignin removals as well as the
Keywords: water recovery. The optimal conditions obtained from the compromise of the three
Coagulation-flocculation desirable responses, supernatant turbidity removal, lignin removal and water recovery
Optimization efficiency, were as follows: coagulant dosage of 871 mg/L, flocculant dosage of 22.3 mg/L
Pulp mill wastewater and pH 8.35. Confirmation experiments demonstrated that such a combination of the UD
Response surface methodology and RSM is a powerful and useful approach for optimizing the coagulation-flocculation
Uniform design process for the pulp mill wastewater treatment.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
The gross output of paper and paperboard is 79.8 million tons
in 2008 in China, and about 22% of them are using straw pulp
as staffs. The wastewater generated by the pulp mills is not
easy to treat due to the presence of a large amount of chem-
icals, such as many sodium salts of organic acids. In addition,
a large amount of lignin present in the wastewater, which
causes colority, turbidity and high COD (chemical oxygen
demand), is usually wasted and overburdens the treatment
process. Thus, both the lignin reclamation and wastewater
treatment are crucial. An efficient and cost-effective process
for the treatment of pulp mill wastewater should be pursued.
Coagulation-flocculation is a simple and efficient method for
wastewater treatment, and has been widely used for the
treatment of palm oil mill effluent (Ahmad et al., 2005), textile
wastewater (Meric et al., 2005) and abattoir wastewater
(Amuda and Alade, 2006), etc. Recently, coagulation-
flocculation or flocculation processes have also been exten-
sively used for the treatment of pulp mill wastewater. In such
studies, polyaluminium chloride (PAC), chitosan, polymeric
phosphate-aluminum chloride, cationic and anionic poly-
acrylamides (PAMs) and polydiallyldimethylammonium
chloride (polyDADMAC) have all been tested as a flocculant in
the flocculation process, and various levels of removal effi-
ciency for turbidity and lignin have been achieved (Razali
et al., 2011; Renault et al., 2009; Wong et al., 2006; Zheng
et al., 2011). In the coagulation-flocculation process, the effi-
ciency is governed by various factors, such as the type and
dosage of coagulant/flocculant (Desjardins et al., 2002; Hu

* Corresponding author. Fax: þ86 551 3601592.

E-mail address: hqyu@ustc.edu.cn (H.-Q. Yu).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.023
5634 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0
et al., 2005; Nandy et al., 2003; Spicer and Pratsinis, 1996; Wang
et al., 2002), pH (Elmaleh et al., 1996; Miller et al., 2008;
Rohrsetzer et al., 1998; Syu et al., 2003), mixing speed and
time (Gurses et al., 2003; Rossini et al., 1999), temperature and
retention time (Howe et al., 2006; Zhu et al., 2004). A proper
optimization of these factors could significantly increase its
treatment efficiency.
Response surface methodology (RSM) is an efficient way to
achieve such an optimization by analyzing and modeling the
effects of multiple variables and their responses and finally
optimizing the process. This method has been widely used for
the optimization of various processes in food chemistry,
material science, chemical engineering and biotechnology
(Granato et al., 2010; Hong et al., 2011; Liu et al., 2010; Singh
et al., 2010). In the traditional experimental design
approaches used in RSM, such as central composite design,
with an increase in experimental factors, the number of coef-
ficients of the quadratic model equation increases exponen-
tially and so does the number of experimental trials (Cheng
et al., 2002). To overcome this shortcoming of RSM, uniform
design (UD) can be used to investigate more factors with
substantially fewer experimental trials, since it determines the
number of the experimental trials only by the level of factors,
rather than by the number of factors. The UD method was first
proposed by Fang (1978). Compared to the traditional experi-
mental design methods, UD is capable of selecting experi-
mental points uniformly in the experimental region and highly
representative in the experimental domain; it imposes no
strong assumption on the model and may be used when the
underlying model between the responses and factors is
unknown or partially unknown; it can accommodates the
largest possible number of levels for each factor among all
experimental designs (Leung et al., 2000; Li et al., 2003; Wen
et al., 2005; Zhang et al., 1998). A combination of UD and RSM
would be able to achieve the optimization of a complex
multivariate process with the fewest multilevel experiments.
Therefore, in this study, UD and RSM were integrated to opti-
mize the coagulation-flocculation process for pulp mill
wastewater treatment.
The selection of high efficient coagulants and flocculants is
essential for a successful coagulation-flocculation process. In
this work, on the basis of our previous studies (Wang et al.,
2007, 2009), the conventional coagulant Al(OH)3 was chosen
as the coagulant, while a novel modified natural polymer,
starch-g-PAM-g-PDMC [polyacrylamide and poly (2-methacr-
yloyloxyethyl) trimethyl ammonium chloride] with both
strong charge neutralization and bridging abilities, was used
as the flocculant.
The main objective of this work was to treat the pulp mill
wastewater using the coagulation-flocculation process, which
was optimized using an integrative UD-RSM approach.
Removal efficiencies of both supernatant turbidity and lignin,
and recovery efficiency of clean water were chosen as the
dependent output variables. The compromise optimal condi-
tions for these three responses were obtained using the
desirability function approach. The novel optimization
strategy used for the pulp mill wastewater treatment process
in this study is expected to provide valuable information for
other complicated systems in environmental engineering and
other fields.
2. Materials and methods
2.1. Chemicals and operation
Aluminum chloride, hydrochloric acid and sodium hydroxide,
purchased from Shanghai Chemical Reagent Co., China, were
of analytical reagent grade and used without further purifi-
cation. The flocculant, starch-g-PAM-g-PDMC, was prepared
as follows: starch, (2-methacryloyloxyethyl) trimethyl
ammonium chloride and acrylamide were dissolved in water
in Pyrex glass vessels, and were heated in a preset water bath
using potassium persulphate as the initiator after deoxyge-
nating. Thereafter, the sample solutions were precipitated in
acetone and separated by filtration. Homopolymers formed in
the reactions were removed using the soxhlet extraction
method in ethanol. All the grafted samples were dried in
a vacuum oven at 50  C until a constant weight. The grafting
percentage and the cationic degree of the graft copolymer
used in this experiment was 269% and 1.96  103 mol/g,
respectively. The point of zero charge of the graft copolymer
was measured as 7.80. Its molecular structure is as shown in
Fig. 1 and its image is illustrated in Fig. S1 (Appendix).
Pulp mill wastewater was blending black liquor from the
primary sedimentation tank of Guoyang Paper and Pulp Mill
Co., China. The stuffs of the paper were wheat straw. The
initial pH, chemical oxygen demand (COD) and turbidity were
6.99, 1358 mg/L and 1209 NTU, respectively. The average sizes
of the colloid particles in the wastewater were 544 nm.
The coagulation-flocculation experiments were carried out
using the jar test method in 1-L beakers. After the coagulant
(stock solution of 50.0 g/L) was added with a dosage varying
from nil to 2100 mg/L, the solution pH was adjusted to 2.5e11.5
by adding 0.1 mol/L HCl or NaOH solutions. Then, the floccu-
lant at a concentration of 1.0 g/L was added with a dosage
varying from nil to 48 mg/L. The sample was immediately
stirred at a constant speed of 200 rpm for 2 min, followed by
a slow stirring at 40 rpm for 10 min; thereafter, a settlement for
5 min was performed. After that, samples were taken from the
water level around 2 cm underneath the surface for measuring
the turbidity and lignin concentrations of the supernatant.
Meanwhile, the volume of produced sludge was calculated
directly from the reading on the beakers, and then the volume
of the recovered clean water was calculated accordingly.
2.2. Experimental design and data analysis
UD tables can be described as Un(qm), where U, n, q and m stand
for the UD, the number of experimental trials, the number of
levels and the maximum number of factors, respectively. For
a given measure of uniformity M, a uniform design has the
smallest M-value overall fractional factorial design with n
runs and m q-level factors.
CH3 CONH2
S O CH2 C m CH2 C n
H
COOC2H4N(CH3)3Cl
Fig. 1 e Molecular structure of the graft copolymer starch-g-
PAM-g-PDMC (S: Starch).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0 5635

In this work, coagulant dosage (X1), flocculant dosage (X2)

Table 2 e Guide for selecting columns of generating
and pH (X3) were chosen as three independent variables in the vectors in U14(143).
coagulation-flocculation process. Seven levels for each factor
No. of factors Columns Discrepancy in
were selected to investigate the influence and interaction of
studied to be used uniformity
the factors. In order to improve the accuracy, the experiment
were carried out using U14(143), the range and levels of each 2 1, 4 0.0957
3 1, 2, 3 0.1455
factor, the guide for using (U14(143)) are listed in Tables 1 and 2,
4 1, 2, 3, 5 0.2091
respectively.
Efficiencies of turbidity removal, lignin removal and clean
water recovery were selected as the dependent variables in the regression model using the backward regression method
order to represent the overall wastewater treatment effi- with the experimental results:
ciency. The response variable was fitted by a sufficient model,
which is able to describe the relationship between the Y ¼ 29:9 þ 22:1X1 þ 11:6X2  2:5X21  0:6X22  0:7X23
dependent output variable and the independent variables 1:5X1 X2 þ 1:2X1 X3
using the regression method. R2 ¼ 0:917; F ¼ 8:81 (2)

X
j X
k i<j X
X Statistical testing of the model was performed with the
Y ¼ b0 þ bi Xi þ bii X2i þ bij Xi Xj (1) Fisher’s statistical test for analysis of variance (ANOVA). The
i¼1 i¼1 i j
quadratic regression shows that the model was significant
where Y is the response variable to be modeled; Xi and Xj are because the value of Fstatistic (the ratio of mean square due to
the independent variables which influence Ym; b0, bi, bii and bij regression to mean square to real error) of 8.81 was greater
are the offset terms, the ith linear coefficient, the quadratic than F0.001,7,6 (4.27). The value of the correlation coefficient
coefficient and the ijth interaction coefficient, respectively. (R2 ¼ 0.917) indicates that only 8.3% of the total variation could
The actual design of this work is given in Table 3 (Fang, 1994). not be explained by the empirical model (Leung et al., 2000;
The parameters of the response equations and corre- Meric et al., 2005).
sponding analysis on variations were evaluated using The p-value ( p ¼ 0.05) of Eq. (2) also implies that the
Uniform Design Software 2.1 (http://www.math.hkbu.edu.hk/ second-order polynomial model fitted the experimental
UniformDesign/software) and MATLAB 6.5, respectively. The results well. This is confirmed by Fig. 2a, in which the plots of
interactive effects of the independent variables on the predicted turbidity removal efficiencies versus measured ones
dependent ones were illustrated by three- and two- are shown. Most points distributed near to the straight line
dimensional contour plots. Finally, two additional experi- where the measured and predicted removal efficiencies are
ments were conducted to verify the validity of the statistical the same, indicating that the regression model is able to
experimental strategies. predict these removal efficiencies.
From Eq. (2), the optimal conditions for the supernatant
turbidity removal efficiency were obtained as follows: coagu-
lant dosage of 917 mg/L, flocculant dosage of 33.0 mg/L and pH
3. Results

The experimental results are listed in Table 3. The variance

trend was discrepant for the three responses. Therefore, the Table 3 e UD and response results for the study of three
operational conditions have been optimized respectively for experimental variables in coded units.
different responses. Run X1 X2 X3 Factors Response

X1 X2 X3 Turbidity Lignin Water

3.1. Optimization for supernatant turbidity removal removal removal recovery
(%) (%) efficiency
The supernatant turbidity removal efficiency listed in Table 3 (%)
is an important denotation for the treatment efficiency of the 1 1 4 7 1 2 4 63.9 36.0 66.0
coagulation-flocculation process. The following equation is 2 2 8 14 1 4 7 56.9 34.5 50.0
3 3 12 6 2 6 3 95.4 71.6 60.0
4 4 1 13 2 1 7 54.3 24.5 62.0
5 5 5 5 3 3 3 92.7 80.9 68.0
Table 1 e Levels of the variable tested in the U7(73) 6 6 9 12 3 5 6 92.9 70.9 60.0
uniform designs. 7 7 13 4 4 7 2 95.8 74.3 48.0
Variables Range and levels 8 8 2 11 4 1 6 87.8 47.6 68.0
9 9 6 3 5 3 2 92.6 76.2 50.0
1 2 3 4 5 6 7 10 10 10 10 5 5 5 98.7 82.5 66.0
11 11 14 2 6 7 1 71.2 40.1 50.0
X1, coagulant 0 350 700 1050 1400 1750 2100
12 12 3 9 6 2 5 95.1 59.5 64.0
dosage (mg/L)
13 13 7 1 7 4 1 64.8 45.3 42.0
X2, flocculant 0 8 16 24 32 40 48
14 14 11 8 7 6 4 67.2 32.5 64.0
dosage (mg/L)
X3, pH 2.5 4.0 5.5 7.0 8.5 10.0 11.5 Source: Fang, 1994.
5636 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0

of the lignin removal efficiency in the coagulation-flocculation

a 100
experiments:

90 Y ¼ 45:3 þ 35:6X1 þ 24:6X2 þ 8:2X3  4:0X21  2:4X22

1:3X23  1:1X1 X2
Predicted (%)

80
R2 ¼ 0:889; F ¼ 4:64 (3)
2
70 The results of F ¼ 4.64 > F (0.05,7,6) ¼ 4.207 and R ¼ 0.889 for
the lignin removal efficiency show that the second-order
polynomial model was significant and fitted the experi-
60
mental results well. Fig. 2b shows that the measured versus
predicted plot values were distributed evenly near to the
50
straight line. From Eq. (3), the optimal conditions for maximal
50 60 70 80 90 10 0
efficiency of lignin removal were estimated to be: coagulant
Measured (%)
dosage of 1004 mg/L, flocculant dosage of 25.9 mg/L and pH
90 5.7. Under these conditions, the maximal lignin removal effi-
b ciency was estimated to be 84.1%.
80
70 3.3. Optimization for water recovery efficiency
Predicted (%)

60
The regression model to describe the water recovery efficiency
50 of the coagulation-flocculation experiments was obtained
40 using the total regression method (Eq. (4)).

30 Y ¼ 15:3  4:0X1 þ 24:4X2 þ 22:4X3  0:4X21  2:4X22  1:8X23

20 þ1:0X1 X2 þ 0:8X1 X3  2:4X2 X3
R2 ¼ 0:935; F ¼ 6:35 (4)
20 30 40 50 60 70 80 90
Measured (%) The regression results, i.e., F ¼ 6.35 > F (0.05,9,4) ¼ 5.999 and
R2 ¼ 0.935, show that the second-order polynomial model was
significant and fitted the experimental results for the water
c 70 recovery efficiency well. Again, the experimental values were
distributed near to the straight line (Fig. 2c). From Eq. (4), the
60 optimal conditions for maximal water recovery efficiency
Predicted (%)

were estimated to be: coagulant dosage of 1040 mg/L, floccu-

50 lant dosage of 20.6 mg/L and pH 8.17, under which the
maximal water recovery efficiency was estimated to be 73.4%.

40
3.4. Confirmation experimental results
30
To confirm the validity of the statistical experimental strate-
30 40 50 60 70 gies, additional confirmation experiments were conducted in
Measured (%) duplicates. The chosen conditions for the coagulant dosage,
flocculant dosage and pH are all listed in Table 4, along with
Fig. 2 e Relationship between the predicted and measured the predicted and measured results. As shown in Table 4, the
(a) turbidity removal efficiency; (b) lignin removal measured efficiencies of the supernatant turbidity removal,
efficiency; and (c) water recovery efficiency. lignin removal and clean water recovery were close to the
predicted values using their respective regression models.
This demonstrates that the UD-RSM approach was appro-
priate for optimizing the operational conditions of the
of 5.67. Under the optimal conditions, the maximal turbidity coagulation-flocculation process.
removal efficiency was estimated to be 99.7%.

3.5. Multiple-response optimization

3.2. Optimization for lignin removal
Lignin is the main pollutant in the pulp mill wastewater. The
efficient removal of lignin from wastewater by the
coagulation-flocculation method is crucial to reclaim water
and lignin. The following equation (Eq. (3)) is the regression
model using the backward regression method with the results
Removal efficiency of supernatant turbidity, removal effi-
ciency of lignin and water recovery efficiency are three indi-
vidual responses, and their optimizations were achieved
under different optimal conditions. Thus, a compromise
among the conditions for the three responses is desirable. The
desirability function approach was used to achieve such a goal
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0 5637

agreement with the model predictions. After flocculation, the

Table 4 e Measured and calculated values for the
confirmation experiments. aluminum ion concentration in the supernatant was
measured as 6.8 mg/L, and the final pH was 6.2.
Run Conditions Parameter Measured Calculated
An optimal pH of 8.35 was predicted for multi-response
15 Coagulant dosage: Turbidity 99.6  0.1 99.7 optimization purpose, at which the optimal compromised
917 mg/L removal removal efficiencies were obtained. However, it is worthwhile
Flocculant dosage: (%)
noting that, decent efficiencies can be obtained at neutral pH
33.0 mg/L
(the initial pH of the wastewater) based on our experiment
pH: 5.67
16 Coagulant dosage: Lignin 88.4  0.3 84.1 results and the prediction by Eq. (5). Thus, in the practical
1004 mg/L removal application, this wastewater could be treated without pH
Flocculant dosage: (%) adjustment when the treatment efficiency is fulfilled and the
25.9 mg/L cost for pH adjustment is a big concern. This is a tradeoff
pH: 5.7 between the efficiency of the process and the cost for pH
17 Coagulant dosage: Water 74.0  2.0 73.4
adjustment.
1040 mg/L recovery
Flocculant dosage: (%)
20.6 mg/L
pH: 8.17 4. Discussion

4.1. Supernatant turbidity removal

(Zhang et al., 1998). The regression equation of the compro-
mise was obtained as follows:
Y ¼ 14:401 þ 0:0918X1 þ 0:923X2 þ 11:084X3  0:0000395X21
 0:0463X22  0:818X23 þ 0:000401X1 X2  0:00759X1 X3
þ 0:156X2 X3
The optimal conditions calculated from the regression
equation were as follows: coagulant dosage of 871 mg/L,
flocculant dosage of 22.3 mg/L and pH 8.35, respectively. The
corresponding removal efficiency of turbidity, removal effi-
ciency of lignin and water recovery efficiency were 95.7%,
83.4% and 72.7%, respectively. The overlay plot for the optimal
region is presented in Fig. 3. The shaded portion gave the
permissible values of the two variables by defining the desired
limits of removal efficiency of supernatant turbidity, removal
efficiency of lignin and water recovery efficiency.
A confirmation experiment under the compromised
conditions was carried out in triplicates, and the average
removal efficiencies of turbidity and lignin, and water
recovery efficiency were obtained as 95.0%, 83.5% and 72.0%,
respectively (Fig. S4 in Appendix). These results were in good
Fig. 3 e Overlay plot for the optimal region.
With the turbidity removal efficiency as the response, the
response surfaces of the quadratic model with one variable
kept at the optimal level and the other two varying within the
experimental ranges are shown in Fig. 4. The obvious peak in
(5) the response surfaces indicates that the optimal conditions
were exactly located inside the design boundary. In other
words, there were significant interactive effects on turbidity
between coagulant dosage and the flocculant dosage, coagu-
lant dosage and pH, as well as flocculant dosage and pH.
The turbidity removal efficiency was high when the coag-
ulant dosage and the flocculant dosage were within the range
of 700e1400 mg/L and 16e48 mg/L, respectively, at the optimal
pH 5.3 (Fig. 4b). Charge neutralization and sweep-floc were the
two main mechanisms leading to the aggregation of particles
in the coagulation process. In general, the appropriate pH for
the charge neutralization in the coagulation process is in
a range of 4.0e5.5 (Chang et al., 1993). When the aluminum ions
is used as a coagulant, pH 6.0e8.0 is suitable for the formation
of amorphous Al(OH)3, which removes organic matters by
adsorption on the precipitation of Al(OH)3(s) through the
sweep-floc mechanism (Chang et al., 1993). Thus, pH 5.3 is
favorable only for the charge neutralization. However, the
acidic condition (pH 5.3) was in favor of the improvement of
cationic charge density as well as the extension of the grafting
chain in the solution. In this case, both the charge neutraliza-
tion ability and the sweep-floc ability were improved. Taking
into account the two factors, pH 5.3 was appropriate for the
turbidity removal of the pulp mill wastewater.
Likewise, at the optimal flocculant dosage, the coagulant
dosage was within the range of 350e1050 mg/L and pH was
2.0e8.5 (Fig. 4b). Under these conditions, the coagulant AlCl3
exhibited good charge neutralization and sweep-floc abilities,
and thus the flocculant, in addition to the two abilities above,
had an improved adsorption bridging ability.
In addition, at the optimal coagulant dosage, the acidic
condition and flocculant dosage within the range of
24e48 mg/L were favorable for the supernatant turbidity
removal. As mentioned above, for a given coagulant dosage,
the acidic condition was appropriate for the removal of
turbidity with the graft copolymer used as the flocculant.
5638 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0
Fig. 4 e 3D surface graphs and contour plots of turbidity
removal efficiency showing the effect of variables: (a)
X1eX2; (b) X1eX3; and (c) X2eX3.
4.2. Lignin removal
When the lignin removal efficiency was selected as the
response, the response surfaces of the quadratic model with
one variable kept at the optimal level and the other two
varying within the experimental ranges are illustrated in
Fig. S2 (Appendix). The peak in the response surfaces indicates
that the optimal conditions were exactly located inside the
design boundary. Furthermore, there were significant inter-
active effects on turbidity between coagulant dosage and
flocculant dosage, coagulant dosage and pH, as well as floc-
culant dosage and pH.
Under the experimental conditions, the lignin was indis-
cerptible and suspended in the wastewater as colloids (about
500 nm). Thus, the coagulation-flocculation method was
appropriate for lignin removal. Acidic condition was in favor
of the precipitation of lignin and thus its removal. However,
the lignin removal efficiency was less than the turbidity
removal efficiency under their respective optimal conditions.
This is attributed to the fact that there is a great amount of
soluble phenols in the wastewater and phenols could not be
efficiently removed by the coagulation-flocculation method.
4.3. Water recovery efficiency
The response surfaces of the quadratic model with one vari-
able kept at the optimal level and the other two varying within
the experimental ranges, with the water recovery efficiency as
the response, are shown in Fig. S3 (Appendix). The elliptical
contour plots indicate that there were significant interactive
effects between coagulant dosage and flocculant dosage,
coagulant dosage and pH, as well as flocculant dosage and pH,
even the water recovery efficiency percentage increased at the
center of the three regions. This was evidenced by the obvious
peak in the response surfaces, in which the optimal condi-
tions were exactly located inside the design boundary.
In the coagulation-flocculation process, the coagulant was
dispersed in the wastewater to destabilize the colloidal
particles and the flocculant was used to agglomerate the
destabilized colloidal particles into large particles and then
precipitates. The flocculant used in our experiment was
a modified natural polymer, which was synthesized by graft-
ing two monomers onto starch backbone in order to improve
its charge neutralization and bridging ability (Wang et al.,
2009). The flexible grafting chain grafted onto the rigid
starch backbone increased the chances for the flocculant to
approach to the contaminant particles in the wastewater.
Therefore, the graft copolymer used as the flocculant would
conduce to the formation of larger and denser flocs, which are
readily separated from the wastewater. As a result, better
quality and higher water quantity could be obtained.
4.4. Significance of the integrated UD-RSM approach
A combination of UD and RSM applied in this work has
a rational statistical basis, and is demonstrated to be
a powerful approach for the optimization of the coagulation-
flocculation process for pulp mill wastewater treatment in
this study. With the UD method, the selected experimental
points were distributed uniformly in the factor space for all
the three key factors influencing the efficiency of this process,
i.e., coagulant dosage, flocculant dosage and pH. This facili-
tated the acquisition of most response information through
the fewest numbers of experiments. In addition, the applica-
tion of number theory in the experimental design facilitates
the computer statistical modeling and the subsequent
regression analysis in the UD software, such as linear
regression, non-linear regression and quadratic regression,
etc. In addition, the number of the experimental trials in UD is
determined only by the level of factors, not by the number of
factors. The UD method can also be used when the levels of
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0
factors are different. These are the advantages of UD over
other experiment design approaches.
A combination of UD and RSM provided a straightforward
way to evaluate the individual effects and interactions of the
experimental factors for desirable responses. Following the
RSM optimization, a desirability function approach could be
employed to obtain the compromise optimal conditions. This
study demonstrates that this integrated approach could
optimize the coagulation-flocculation process effectively
using few data sets, which becomes very attractive for the
processes where data are obtained costly. Thus, this inte-
grated optimization approach can be useful for other complex
wastewater treatment processes and multivariate systems in
other fields.
5. Conclusions
A coagulation-flocculation process with aluminum chloride as
the coagulant and a modified natural polymer, starch-g-PAM-
g-PDMC [polyacrylamide and poly (2-methacryloyloxyethyl)
trimethyl ammonium chloride], as the flocculant was
employed for pulp mill wastewater treatment. A novel
approach combined response surface methodology (RSM) and
uniform design (UD) was used to optimize the process and
evaluate the effects and interactions of three main influential
factors, coagulant dosage, flocculant dosage and pH, on the
treatment efficiency in terms of the supernatant turbidity and
lignin removals as well as the water recovery. An optimal
condition of coagulant dosage 871 mg/L, flocculant dosage
22.3 mg/L and pH 8.35 was obtained from the compromise of
the three desirable responses, i.e. supernatant turbidity
removal, lignin removal and water recovery efficiency.
Further confirmation experiments demonstrated that such
a combination of the UD and RSM is an effective and powerful
approach for the optimization of the coagulation-flocculation
process for pulp mill wastewater treatment.
Acknowledgments
We wish to thank the Key Special Program on the S&T for the
Pollution Control (2008ZX07103-001 and 2008ZX07010-003) for
the partial support of this study.
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.023.
references
Ahmad, A.L., Ismail, S., Bhatia, S., 2005. Optimization of
coagulation-flocculation process for palm oil mill effluent
using response surface methodology. Environmental Science
& Technology 39 (8), 2828e2834.
5639
Amuda, O.S., Alade, A., 2006. Coagulation/flocculation process in
the treatment of abattoir wastewater. Desalination 196 (1e3),
22e31.
Chang, Q., Fu, J.Y., Li, Z.L., 1993. Principles of Flocculation.
Lanzhou University Press, Lanzhou, China.
Cheng, B.J., Zhu, N., Fan, R.L., Zhou, C.X., Zhang, G.N., Li, W.K.,
Ji, K.J., 2002. Computer aided optimum design of rubber recipe
using uniform design. Polymer Testing 21 (1), 83e88.
Desjardins, C., Koudjonou, B., Desjardins, R., 2002. Laboratory
study of ballasted flocculation. Water Research 36 (3),
744e754.
Elmaleh, S., Yahi, H., Coma, J., 1996. Suspended solids abatement
by pH increase-upgrading of an oxidation pond effluent.
Water Research 30 (10), 2357e2362.
Fang, K.T., 1978. Probability and Statistics Bulletin. Institute of
Mathematics, Academia Sinica, Beijing, China.
Fang, K.T., 1994. Uniform Design and Uniform Tables. Science
Publication Co, Beijing, China.
Granato, D., Ribeiro, J.C.B., Castro, I.A., Masson, M.L., 2010.
Sensory evaluation and physicochemical optimisation of soy-
based desserts using response surface methodology. Food
Chemistry 121 (3), 899e906.
Gurses, A., Yalcin, M., Dogar, C., 2003. Removal of Remazol Red RB
by using Al (III) as coagulant-flocculant: effect of some
variables on settling velocity. Water Air and Soil Pollution 146
(1e4), 297e318.
Hong, F.L., Peng, J.C., Lui, W.B., 2011. Optimization of the process
variables for the synthesis of starch-based bioadable resin
using response surface methodology. Journal of Applied
Polymer Science 119 (3), 1797e1804.
Howe, K.J., Marwah, A., Chiu, K.P., Adham, S.S., 2006. Effect of
coagulation on the size of MF and UF membrane foulants.
Environmental Science & Technology 40 (24), 7908e7913.
Hu, C., Liu, H., Qu, J., Wang, D., Ru, J., 2005. Coagulation behavior
of aluminum salts in eutrophic water: significance of Al13
species and pH control. Environmental Science & Technology
40 (1), 325e331.
Leung, S.Y.L., Chan, W.H., Luk, C.H., 2000. Optimization of
fabrication conditions of high-efficiency ultrafiltration
membranes using methods of uniform design and regression
analysis. Chemometrics and Intelligent Laboratory Systems 53
(1e2), 21e35.
Li, J.F., Liao, H., Normand, B., Cordier, C., Maurin, G., Foct, J.,
Coddet, C., 2003. Uniform design method for optimization of
process parameters of plasma sprayed TiN coatings. Surface &
Coatings Technology 176 (1), 1e13.
Liu, B.B., Yang, M.H., Qi, B.K., Chen, X.R., Su, Z.G., Wan, Y.H., 2010.
Optimizing L-(þ)-lactic acid production by thermophile
Lactobacillus plantarum As. 1.3 using alternative nitrogen
sources with response surface method. Biochemical
Engineering Journal 52 (2e3), 212e219.
Meric, S., Selcuk, H., Belgiorno, V., 2005. Acute toxicity removal in
textile finishing wastewater by Fenton’s oxidation, ozone and
coagulation-flocculation processes. Water Research 39 (6),
1147e1153.
Miller, S.M., Fugate, E.J., Craver, V.O., Smith, J.A., Zimmerman, J.B.,
2008. Toward understanding the efficacy and mechanism of
Opuntia spp. as a natural coagulant for potential application in
water treatment. Environmental Science & Technology 42 (12),
4274e4279.
Nandy, T., Shastry, S., Pathe, P.P., Kaul, S.N., 2003. Pre-treatment
of currency printing ink wastewater through coagulation-
flocculation process. Water Air and Soil Pollution 148 (1e4),
15e30.
Razali, M.A.A., Ahmad, Z., Ahmad, M.S.B., Ariffin, A., 2011.
Treatment of pulp and paper mill wastewater with various
molecular weight of polyDADMAC induced flocculation.
Chemical Engineering Journal 166 (2), 529e535.
5640 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 3 3 e5 6 4 0
Renault, F., Sancey, B., Charles, J., Morin-Crini, N., Badot, P.M.,
Winterton, P., Crini, G., 2009. Chitosan flocculation of
cardboard-mill secondary biological wastewater. Chemical
Engineering Journal 155 (3), 775e783.
Rohrsetzer, S., Paszli, I., Csempesz, F., 1998. Colloid stability of
electrostatically stabilized sols. Part III: the role of pH in
hydration coagulation and peptization of SiO2- and Al2O3-sols.
Colloid and Polymer Science 276 (3), 260e266.
Rossini, M., Garrido, J.G., Galluzzo, M., 1999. Optimization of the
coagulation-flocculation treatment: influence of rapid mix
parameters. Water Research 33 (8), 1817e1826.
Singh, K.P., Gupta, S., Singh, A.K., Sinha, S., 2010. Experimental
design and response surface modeling for optimization of
Rhodamine B removal from water by magnetic
nanocomposite. Chemical Engineering Journal 165 (1),
151e160.
Spicer, P.T., Pratsinis, S.E., 1996. Shear-induced flocculation: the
evolution of floc structure and the shape of the size
distribution at steady state. Water Research 30 (5), 1049e1056.
Syu, M., Chen, B., Chou, S., 2003. A study on the sedimentation
model and neural network online adaptive control of
a benzoic acid imitated wastewater oxidation process.
Industrial & Engineering Chemistry Research 42 (26),
6862e6871.
Wang, D.S., Tang, H.X., Gregory, J., 2002. Relative importance of
charge neutralization and precipitation on coagulation of
kaolin with PACl: effect of sulfate ion. Environmental Science
& Technology 36 (8), 1815e1820.
Wang, J.P., Chen, Y.Z., Ge, X.W., Yu, H.Q., 2007. Optimization of
coagulation-flocculation process for a paper-recycling
wastewater treatment using response surface methodology.
Colloids and Surfaces A-Physicochemical and Engineering
Aspects 302 (1e3), 204e210.
Wang, J.P., Chen, Y.Z., Yuan, S.J., Sheng, G.P., Yu, H.Q., 2009.
Synthesis and characterization of a novel cationic chitosan-
based flocculant with a high water-solubility for pulp mill
wastewater treatment. Water Research 43 (20), 5267e5275.
Wen, S.H., Zhang, T., Tan, T.W., 2005. Optimization of the amino
acid composition in glutathione fermentation. Process
Biochemistry 40 (11), 3474e3479.
Wong, S.S., Teng, T.T., Ahmad, A.L., Zuhairi, A., Najafpour, G.,
2006. Treatment of pulp and paper mill wastewater by
polyacrylamide (PAM) in polymer induced flocculation.
Journal of Hazardous Materials 135 (1e3), 378e388.
Zhang, L., Liang, Y.Z., Jiang, J.H., Yu, R.Q., Fang, K.T., 1998.
Uniform design applied to nonlinear multivariate calibration
by ANN. Analytica Chimica Acta 370 (1), 65e77.
Zheng, H.L., Zhu, G.C., Jiang, S.J., Tshukudu, T., Xiang, X.Y.,
Zhang, P., He, Q.A., 2011. Investigations of coagulation-
flocculation process by performance optimization, model
prediction and fractal structure of flocs. Desalination 269
(1e3), 148e156.
Zhu, K., El-Din, M.G., Moawad, A.K., Bromley, D., 2004. Physical
and chemical processes for removing suspended solids and
phosphorus from liquid swine manure. Environmental
Technology 25 (10), 1177e1187.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

In situ disinfection of sewage contaminated shallow

groundwater: A feasibility study

Morgan M. Bailey a, William J. Cooper b, Stanley B. Grant a,b,*

a
Department of Chemical Engineering and Material Sciences, University of California, Henry Samueli School of Engineering,
University of California, Irvine, CA 92697, USA
b
Department of Civil and Environmental Engineering, and Urban Water Research Center, Henry Samueli School of Engineering,
University of California, Irvine, CA 92697, USA

article info abstract

Article history: Sewage-contaminated shallow groundwater is a potential cause of beach closures and
Received 18 June 2011 water quality impairment in marine coastal communities. In this study we set out to
Received in revised form evaluate the feasibility of several strategies for disinfecting sewage-contaminated shallow
9 August 2011 groundwater before it reaches the coastline. The disinfection rates of Escherichia coli (EC)
Accepted 14 August 2011 and enterococci bacteria (ENT) were measured in mixtures of raw sewage and brackish
Available online 22 August 2011 shallow groundwater collected from a coastal community in southern California. Different
disinfection strategies were explored, ranging from benign (aeration alone, and aeration
Keywords: with addition of brine) to aggressive (chemical disinfectants peracetic acid (PAA) or per-
Disinfection oxymonosulfate (Oxone)). Aeration alone and aeration with brine did not significantly
Shallow groundwater reduce the concentration of EC and ENT after 6 h of exposure, while 4e5 mg L
1 of PAA or
Sewage Oxone achieved >3 log reduction after 15 min of exposure. Oxone disinfection was more
Remediation rapid at higher salinities, most likely due to the formation of secondary oxidants (e.g.,
Water quality bromine and chlorine) that make this disinfectant inappropriate for marine applications.
Peroxymonosulfate Using a Lagrangian modeling framework, we identify several factors that could influence
Peracetic acid the performance of in-situ disinfection with PAA, including the potential for bacterial
Oxone regrowth, and the non-linear dependence of disinfection rate upon the residence time of
Brine water in the shallow groundwater. The data and analysis presented in this paper provide
a framework for evaluating the feasibility of in-situ disinfection of shallow groundwater,
and elucidate several topics that warrant further investigation.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction identified as a source of fecal pollution at some coastal bea-

Beach closures and advisories are generally caused by the
coastal discharge of land-side sources of fecal pollution such
as urban runoff or sewage effluent (Boehm et al., 2002; Grant
and Sanders, 2010). However, contamination of shallow
groundwater by aging sewage collection systems and failing
onsite sewage treatment and disposal systems has also been
ches (Lipp et al., 2001). The latter typically involves a two-step
process: (1) shallow groundwater is first contaminated by the
surface or subsurface release of sewage, and (2) the shallow
groundwater is then discharged to coastal waters under the
influence of land hydraulic gradients, tidal pumping, and/or
current induced pressure gradients (Burnett et al., 2003b).
The coastal discharge of sewage-contaminated shallow

* Corresponding author. Tel.: þ1 949 824 8277; fax: þ1 949 824 2541.
E-mail addresses: mbailey@uci.edu (M.M. Bailey), sbgrant@uci.edu (S.B. Grant).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.020
5642 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
groundwater can impact near-shore water quality by
vectoring fecal-oral pathogens (or their indicators) directly
from sources of human sewage to receiving waters (Boehm
et al., 2003). The discharge may also deliver nutrients that
promote the survival, and perhaps growth, of fecal bacteria in
near-shore waters (Boehm et al., 2004).
To date, most studies on shallow groundwater discharge
focus on its quantification using isotopic (McIlvin and Altabet,
2005), geophysical (Manheim et al., 2004; Santos et al., 2008;
Swarzenski et al., 2006), or hydrological methods (Burnett
et al., 2003a; Michael et al., 2005; Taniguchi et al., 2002), but
to our knowledge there are no published reports on the very
practical and important question of how to remediate shallow
groundwater once it becomes contaminated with sewage.
Given the advancing age and poor condition of sewage
collection systems and onsite sewage treatment and disposal
systems in many coastal areas of the U.S. (Food and Water
Watch, 2008), a cost-effective strategy for remediating
sewage-derived fecal pollution in brackish shallow ground-
water is urgently needed for the sustainable use of coastal
beaches and bays.
While the best remediation strategy should always be to
identify and eliminate the source(s) of sewage responsible for
groundwater contamination, logistical and economic
constraints may make such action infeasible in the near term,
in which case interim solutions are needed. Further, even if
a source of sewage contamination is identified and repaired, it
may take some time before pathogens and indicator bacteria
decay to acceptable concentrations, particularly when the
sediments in question are subsurface and thus not exposed to
sunlight (Mika et al., 2009). Subsurface injection of disinfec-
tant (or “in-situ disinfection”) is one obvious interim strategy
for remediating sewage-contaminated shallow groundwater,
but several issues need to be resolved before such an approach
can be implemented in practice. (1) Disinfectants can react
with organic matter and/or trace anions (e.g., bromide and
chloride ions) in brackish coastal groundwater to produce
toxic disinfection by-products, and their discharge to coastal
zones could pose a health risk to bathers along the shoreline
and negatively impact sensitive marine ecosystems (Monarca
et al., 2000). (2) The shallow groundwater adjacent to marine
coastal areas can be thought of as “subterranean estuaries”, in
which fresh (meteoric) waters mix with seawater before dis-
charging to the ocean (Moore, 1999). Consequently, the flow in
shallow groundwater can vary substantially in space and
time, potentially affecting the disinfection contact times
water parcels experience before discharging to coastal waters.
(3) Sewage constituents, such as pathogens and their indica-
tors, may not mix fully over the vertical dimension of the
shallow groundwater, and therefore disinfection strategies
may need to be tailored to target sewage constituents where
they are located, for example near the top of the water table.
(4) Planning and implementation of in-situ disinfection would
greatly benefit from quantitative “design criteria” that account
for the rate at which the target organisms decay with time
upon exposure to disinfectant, the decay in disinfectant with
time, the time scale over which water parcels reside in the
shallow groundwater before discharge, and ecological
processes that promote the subsurface regrowth of fecal
bacteria.
In this paper we present laboratory data and modeling
efforts intended to, at least preliminarily, address the issues
raised above, using as a test case Avalon Bay, Catalina Island,
California, where chronic fecal contamination of near-shore
waters has been linked to sewage contamination of shallow
groundwater by leaking sewage collection systems beneath
the City (Boehm et al., 2003, 2009b). The goals of this study are
to: (1) test three strategies for inactivating sewage-associated
fecal bacteria from shallow groundwater, including aeration
alone, aeration with addition of brine, and aeration with
addition of either peracetic acid (PAA) or peroxymonosulfate
(commercially known as Oxone); (2) characterize the salinity
dependence and disinfection by-product formation potential
of the latter two disinfectants; (3) evaluate several models for
PAA disinfection; and (4) develop a Lagrangian modeling
framework for evaluating, and potentially designing, in-situ
disinfection of sewage-contaminated coastal shallow ground
waters.
2. Materials and methods
2.1. Choice of disinfectant
Bench-scale experiments were carried out to measure the
decay of sewage-associated fecal indicator bacteria in shallow
groundwater upon exposure to: (1) aeration alone, (2) aeration
with addition of brine, (3) aeration with addition of PAA (35%
solution, CAS-79-21-0, Pfaltz & Bauer, CT), and (4) aeration
with addition of Oxone (43% dry weight, CAS-37222-66-5, Alfa
Aesar, Ward Hill, MA). The impact of aeration on bacterial die-
off was evaluated on the premise that, if efficacious, injection
of ambient air into the subsurface would be a straightforward,
economical, and environmentally benign approach for reme-
diating sewage-contaminated shallow groundwater. The
choice of brine as a candidate disinfectant was motivated by
the fact that its injection into shallow groundwater is unlikely
to cause environmental harm, and because the field site under
consideration, Avalon Bay, has a desalination plant from
which brine is produced as a waste product. PAA is a prom-
ising disinfectant for this application, because it produces
little or no toxic disinfection by-products when mixed with
seawater (it degrades into acetic acid, vinegar) (Kitis, 2004), is
a strong biocide in marine waters as evidenced by its use as an
antifouling agent in cooling water systems for coastal power
plants, and is regarded as relatively safe for discharge to
sensitive marine waters (Sanchez-Ruiz et al., 1995) including
the highly regulated Italian Lagoon of Venice (Cristiani, 2005).
To our knowledge, Oxone has not been tested as a biocide in
coastal marine settings, but its application to this particular
problem was motivated by the fact that, upon addition to
water, it generates hydroxyl radicals and sulfate ions
(Anipsitakis and Dionysiou, 2003); the former should accel-
erate the die-off of bacteria and viruses while the latter is
already present at high concentrations in marine waters.
Many commonly used disinfectants (e.g., ozone, chlorine gas,
sodium hypochlorite) were excluded from consideration for
logistical reasons and, although highly effective biocides, they
would likely react with trace anions and organics in sewage-
contaminated brackish groundwater to form toxic
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
disinfection by-products (Abarnou and Miossec, 1992; Allonier
et al., 1999).
2.2. Experimental characterization of disinfection kinetics
Raw (untreated) sewage was diluted 1:100 using various
mixtures (with various final salinities) of three source waters: (1)
relatively fresh (salinity ca. 1) water from an inland groundwater
well, (2) Avalon Bay water (salinity ca. 32), and (3) desalination
plant brine (salinity ca. 45). All source waters were aseptically
collected from the City of Avalon using sterile polypropylene
bottles and mixed with raw sewage collected from influent to
the City of Avalon wastewater treatment plant. Source waters
were not sterilized prior to mixing with sewage, and therefore
bacteria populations present in the final mixtures of sewage and
source waters probably included (minor) contributions from the
latter. Bench-scale disinfection experiments commenced
within 6 h of sample collection and were carried out in 4 L Nal-
gene polypropylene reaction vessels maintained in the dark and
partially submerged in a temperature controlled and recircu-
lating water bath at 15  1  C. Contents of the reactor vessel were
continuously aerated using a frit through which ambient air was
driven by an aquarium pump. To capture a range of salinities
and disinfectant concentrations, a matrix design was adopted in
which over 45 separate disinfection experiments (including 10
no-disinfectant-added control experiments) were conducted at
five different disinfectant concentrations (ranging from 0 to
7.3 mg L
1 for Oxone, and 0e8 mg L
1 for PAA) and four different
salinities (0, 15, 32, 45). The concentration range adopted for PAA
was based on previously published studies with this disinfec-
tant (Kitis, 2004; Santoro et al., 2007). To our knowledge, this is
the first study to evaluate Oxone disinfection in saline mixtures,
and thus the concentration range adopted for this disinfectant
was based on pilot studies (not reported) with this disinfectant.
Each batch reactor was sampled seven times over a period of
1e6 h (depending on experiment). This sampling schedule was
chosen to resolve disinfection over a single ebb tide, which
represents a theoretical minimum time a fluid parcel would be
in contact with disinfectant before discharging to Avalon Bay,
assuming that the discharge of shallow groundwater was under
tidal control. Samples were extracted from the reactor using
either a sterile syringe or pipet, analyzed for pH and conduc-
tivity, quenched by addition of approximately 0.4 mL of 0.1 N
sodium thiosulfate (CAS-7772-98-7, Mallinckrodt Chemicals),
immediately diluted either 1:10 or 1:100 in sterile deionized
water (Hardy Scientific, California), and enumerated for Escher-
ichia coli (EC) and enterococci bacteria (ENT) using Colilert-18 and
Enterolert defined substrate tests implemented in a 97-Well
Quanti-Tray format (IDEXX Laboratories, Maine). The adoption
of IDEXX Colilert-18 and Enterolert was motivated based on the
fact that both tests are approved by the U.S. Environmental
Protection Agency for enumerating EC and ENT bacteria in
ambient waters (USEPA, 2003) and, more to the point, are used
by the Los Angeles Department of Health Services in their
routine monitoring of recreational beach water quality in
Avalon Bay, and as a basis for management decisions regarding,
for example, the posting of Avalon beaches as unfit for swim-
ming. For a subset of the disinfection experiments, dissolved
organic carbon (DOC) and total organic carbon (TOC) concen-
trations were measured: (1) on the source water (groundwater or
5643
bay water) prior to addition of sewage and disinfectant; (2) on the
source water plus sewage prior to the addition of disinfectant;
and (3) at the end of the disinfection experiment. Samples
analyzed for DOC and TOC were collected in 500 mL amber glass
bottles and acidified with 2 mL of hydrochloric acid, and then
analyzed by TestAmerica using Standard Method 5310B.
2.3. Disinfection rate constants
Two different disinfection rate constants are reported in this
study: (1) an effective first-order disinfection rate constant k0d ,
and (2) an intrinsic disinfection rate constant kd. Values of the
effective first-order rate constant k0d were estimated by
regressing log-transformed bacteria concentration against
time based on Chick’s Law (Chick, 1908):
NðtÞ
ln ¼
k0d t (1)
N0
In Eq. (1), N(t) represents the concentration of bacteria in the
reaction vessel at any time t and N0 represents the bacterial
concentration present in the mixture at the start of the
disinfection experiment. The regression excluded any bacte-
rial measurements in the initial lag period and bacterial
measurements that fell below the lower-limit of detection for
the Colilert and Enterolert assays (in most cases, the lower
limit of detection was 10 most probable number (MPN) per
100 mL of sample).
In the case of PAA disinfection, an intrinsic disinfection
rate constant was also calculated, which represents the
susceptibility of bacteria to disinfectant independent of
disinfectant concentration:
kd ¼ k0d =C0 (2)
The variable C0 represents the initial (molar) concentration of
peroxycompounds which, in the case of commercial prepa-
rations of PAA, includes both PAA and hydrogen peroxide
(Wagner et al., 2002).
2.4. Activation energy for PAA disinfection
Using the reactor set-up described above, four separate
disinfection experiments were carried out at five tempera-
tures (T ¼ 5, 10, 15, 20 and 30  C) and a fixed concentration of
PAA (6 mg L
1) to determine the temperature sensitivity of EC
and ENT disinfection by PAA. Intrinsic disinfection rate
constants kd calculated from these experiments (see Section
2.3) were used to determine activation energies for the PAA
disinfection of EC and ENT, based on an Arrhenius plot of
ln(kd) against 1/T.
3. Disinfection results and discussion
3.1. Measurements of pH, TOC, and DOC
Over aeration of the reactor vessel caused the pH to increase
slightly from 7.9  0.2 to 8.3  0.2. No other systematic pH
trends were observed across the different experiments and
different source waters tested. The average DOC concentra-
tion in groundwater ranged from 1.6  0.14 mg L
1 (N ¼ 7,
5644 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
foreshore well, salinity 20) to 2.9  0.2 mg L
1 (N ¼ 4, inland
well, salinity 1). The TOC/DOC ratio was near unity in water
collected from the inland well (1.1 0.1, N ¼ 4) implying that
most of the organic carbon was dissolved. Raw sewage
collected from the Avalon Sewage Treatment Plant had much
higher levels of DOC (25 mg L
1, N ¼ 1). However, sewage
contributed only background levels of DOC (<0.25 mg L
1) to
the final 1:100 mixtures of sewage and source waters used in
the disinfection experiments. For disinfection experiments
carried out with PAA, the largest source of DOC was frequently
the PAA solution itself, which contained upwards of
280 mg L
1 of carbon as acetic acid. The acetic acid was
present as an equilibrium component of the commercial PAA
preparation, and generated as a breakdown product of PAA.
3.2. Experimental characterization of disinfection
kinetics
First-order effective disinfection coefficients ðk0d Þ estimated
from the 45 separate disinfection experiments are contoured
using Delaunay triangulation (Igor Pro v 6.10, Lake Oswego,
Oregon) against disinfectant concentration and salinity in
Fig. 1. These results are described in the sections below. A more
detailed analysis of k0d values, and an evaluation of different
disinfection models, is presented later in the paper (Section 4).
3.2.1. Aeration alone and aeration with addition of brine
We started by evaluating two environmental benign
approaches for removing fecal indicator bacteria from sewage-
contaminated shallow groundwater; namely, injection of
ambient air and/or injection of brine produced from a local
Fig. 1 e Contour plots of the effective first-order rate constants k0d [minL1] for PAA disinfection of EC (panel A) or ENT (panel
B), and Oxone disinfection of EC (panel C) or ENT (panel D).
desalination plant. Aeration alone and aeration with brine did
not significantly reduce EC and ENT concentrations after 6 h of
exposure; i.e., within the resolution of these experiments
k0d ¼ 0. Thus these two strategies are unlikely to be effective
against indicator bacteria in shallow groundwater over the time
scale of a single ebb tide.
3.2.2. Aeration with addition of PAA and oxone
EC and ENT concentrations were reduced by more than 1000
fold (3 log units) after 15 min of exposure to 4e5 mg L
1 of PAA.
The effective disinfection rate constants calculated for these
experiments (Fig. 1A and B): (1) increased monotonically with

1
PAA dose up to the maximum rate ðk0d ¼ 0:2 min Þ resolvable
with our experimental set-up; (2) do not depend, at least not
dramatically, on the salinity of the disinfection mixture; and
(3) are larger for EC than ENT at a fixed PAA dose. The last
observation is consistent with the results reported in Stampi
et al. (2002). Averaged across all experiments (and excluding
any experiments where the effective disinfection rate
constant exceeded 0.2 min
1), the average intrinsic rate
constants for PAA disinfection are kENT d ¼ 2:3  0:57 and

1
1
d ¼ 3:7  0:2 mM min : The estimate for kd is similar to
kEC EC
intrinsic rate constants estimated from previously published
data for PAA disinfection of EC in secondary settled effluent
(Dell’Erba et al., 2004) and PAA disinfection of fecal coliform
(FC) in secondary-treated sewage effluent (Wagner et al., 2002)
(Table 1). The fact that the intrinsic rate constants for PAA
disinfection of EC and FC are similar across these three studies
is notable, given that EC is a subset of FC, and the very
different initial bacterial concentrations (and presumably
sewage content) associated with the different source waters
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5645

Table 1 e Intrinsic disinfection rates and parameters from multiple studies.

Source Value kd (mM
1 min
1)a %H2 O2=%PAAb Temp ( C) N0 c Solutiond

FC EC ENT

This study NA 3.7  0.2 2.3  0.57 0.183 15 104 GW, BW,
Br, TPI
Dell’Erba et al. NA 3.65  2.41 NA 1.53 NR 102 TPE
Wagner et al. 1.04  1.26 NA NA 1.54 NR 105 TPE

NR-not reported. NA-not applicable.

a Intrinsic disinfection rate.
b Ratio of percent mass of PAA and hydrogen peroxide in PAA equilibrium solution.
c Magnitude of initial bacteria concentration.
d Sources waters for disinfection studies (GW-groundwater, BW-way water, Br-brine, TPI-treatment plant influent, TPE-treatment plant
effluent).

used in these studies, ranging from 102 to 105 bacteria per
100 mL (Table 1).
EC and ENT concentrations were reduced by more than 3
log units after 15 min of exposure to 4 mg L
1 Oxone. However,
unlike PAA, the disinfection rate depends on both Oxone dose
and solution salinity (Fig. 1C and D). At low Oxone dose
(<2 mg L
1), the effective disinfection rate increases mono-
tonically with Oxone dose, and exhibits no obvious salinity
dependence (i.e., the contour lines are near vertical in this
region of the plot). At higher Oxone dose, the effective disin-
fection rate increased monotonically with salinity, and
exhibits no obvious dose dependence (i.e., the contour lines
were near horizontal in this region of the plot). One possible
explanation for this pattern is that Oxone may oxidize anions
in the shallow groundwater and brine to yield, for example,
the secondary oxidants chlorine and bromine.
To explore this idea, a set of control experiments were
carried out in which either PAA or Oxone was added to an
aqueous solution consisting of 35 g L
1 NaCl to mimic the
background salinity of seawater and varying concentrations
(0e70 mg L
1) of KBr to mimic the presence of trace anions.
The formation of oxidized forms of chloride and bromide ions
(e.g., chlorine and bromine) was monitored by measuring
absorbance of the indicator dye N,N-diethyl-p-phenylenedi-
amine (DPD) (Standard Methods, 4500-CL G). Fig. 2A shows
DPD spectra measured after addition of 10 mg L
1 of either
Oxone (solid lines) or PAA (dashed lines) to distilled water
(Solution 1), an aqueous solution consisting of 35 g L
1 NaCl
(Solution 2), or an aqueous solution consisting of 35 g L
1 NaCl
and 70 mg L
1 KBr (Solution 3). In this set of experiments, the
disinfectant was allowed to react in the solution for 5 min,
whereupon DPD was added, and 1 min later the DPD absor-
bance spectrum was measured. Referring to Fig. 2A, the
absorbance of DPD increased in the order Solution
1 < Solution 2 < Solution 3, consistent with the idea that
Oxone oxidizes both chloride and bromide ions to form
secondary oxidants. DPD absorbance did not increase when
PAA was allowed to react for 5 min in Solution 2 and increased
only slightly when PAA was allowed to react for 5 min in
Solution 3. These latter results are consistent with the findings

Fig. 2 e Panel A: DPD absorbance spectra measured after 10 mg LL1 of either Oxone (solid lines) or PAA (dashed lines) is
allowed to react for 5 min in: Solution 1 (DI water), Solution 2 (DI water and 35 g LL1 NaCl), or Solution 3 (DI water, 35 g LL1
NaCl, and 70 mg LL1 KBr). Panel B: DPD absorbance at 515 nm after 10 mg LL1 of either Oxone (solid lines) or PAA (dashed
lines) are allowed to react for 5 min in Solution 2 with KBr concentrations shown. Panel C: the kinetics of oxidant formation
by 10 mg LL1 of either Oxone (solid line) or PAA (dashed line) in Solution 3.
5646 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
of Booth and Lester (1995), who report that PAA could not
oxidize chloride ion to hypochlorous acid, but could oxidize
bromide ion to hypobromous acid. In the presence of Oxone,
DPD absorbance increased with increasing KBr concentration
in Solution 2 (Fig. 1B), and the oxidation kinetics in Solution 3
are relatively rapidly (<5 min of exposure time, Fig. 1C). In the
presence of PAA, on the other hand, DPD absorbance
increased only slightly with increasing KBr concentration in
Solution 2 (Fig. 2B), and oxidization kinetics in Solution 3 are
relatively slow (>150 min exposure time) (Fig. 2C).
Collectively, the results presented above are consistent with
the idea that Oxone quickly oxidized trace anions in brackish
shallow groundwater to form secondary oxidants that act
synergistically with Oxone to enhance disinfection rates.
However, the reaction of Oxone with trace anions could also
increase the toxicity of the shallow groundwater, by producing
compounds that are carcinogenic (e.g., bromateion) and/or by
producing secondary oxidants (e.g., bromine and chlorine) that
subsequently react with organic compounds to form toxic
disinfection by-products (e.g., trihalomethanes) (Guo and Lin,
2009). For these reasons, and despite its obvious utility as
a disinfectant, Oxone should not be used for in-situ disinfection
of sewage contaminated shallow groundwater in settings, like
Avalon Bay, where the groundwater is under marine influence.
PAA, on the other hand, does not appear to react strongly with
chloride ion, or quickly with bromide ion, and thus is less likely
to produce toxic disinfection by-products; a conclusion sup-
ported by several published studies (Liberti et al., 1999; Dell’Erba
et al., 2007; Kitis, 2004). Although PAA does not appear to
generate the quantity and spectrum of disinfection by-products
associated with many disinfectants, some researchers have
raised concern about the potential environmental impact of
releasing disinfected effluents that contain a PAA residual
(Antonelli et al., 2009; de Lafontaine et al., 2008), which can be
toxic to crustaceans and microorganisms (Antonelli et al., 2009;
de Lafontaine et al., 2008). However, Lafontaine et al. (2008) note
that, because PAA breaks down rapidly in seawater, any
potential impacts would be localized around the region where
disinfection residual is discharged to the environment. Thus, if
PAA is used for in-situ disinfection of sewage contaminated
shallow groundwater, care should be taken to minimize the
PAA residual discharged to coastal waters.
3.3. Activation energy for PAA disinfection
Based on measurement of intrinsic disinfection rates over
a range of temperatures, from 5 to 30  C, the following acti-
vation energies were estimated for PAA disinfection of EC and
ENT: 37.5  7.8 kJ mol
1 and 38.0  9.3 kJ mol
1, respectively.
These activation energies are used later in the paper to esti-
mate intrinsic disinfection rates for PAA over a range of
temperatures relevant to the shallow groundwater in Avalon
Bay (see Section 5.1).
4. Modeling PAA disinfection kinetics
Of the strategies evaluated above, PAA disinfection appears
the most viable, given that it is both a potent biocide and less
likely to form toxic disinfection by-products. In this section we
evaluate several published models for PAA disinfection, all of
which are special cases of Hom’s Law (Hom, 1972):
dN
¼
kd Ntm Cn (3)
dt
where, N, C, kd, and t represent bacterial concentration,
disinfectant concentration, intrinsic disinfection rate
constant, and time, respectively. Chick’s Law (Eqs. (1) And (2)
of this paper) corresponds to a choice of exponent values
m ¼ 0, n ¼ 1, and a constant disinfectant concentration, C ¼ C0.
Wagner et al. (2002) suggested that, during the disinfection
process, the concentration of peroxycompounds in commer-
cial PAA mixtures decays with time in accordance with the
following second-order rate law:
dC
¼
k C2 (4)
dt
The rate constant k* depends on the initial peroxycompound
concentration C0 (Wagner et al., 2002):
k ¼ aC
b
0 (5)
Where the pre-factor and power-law exponents are given by
a ¼ 0.0093 mM(b
1) min
1 and b ¼ 1.420. In their analysis of PAA
disinfection, Santoro et al. (2007) suggest that peroxycompounds
exhibit zero-order decay; however, zero-order decay models
predict negative concentration in finite time, and therefore will
not be considered further here. Combining Eqs. (4) and (5), and
solving the differential equation, yields the following prediction
for disinfectant concentration as a function of time:
1
CðtÞ ¼ (6)
1=C0 þ k t
Given this time-dependence for the disinfectant concentra-
tion, Hom’s Law becomes:
dN
n
¼
kd Ntm ½1=C0 þ k t (7)
dt
Wagner et al. (2002) solved Eq. (7) for two different choices of
the exponents n and m: (1) n ¼ 1 and m ¼ 0 (no-tailing model),
and (2) n ¼ 1 and m ¼
1 (tailing model). The tailing model
provided a better empirical fit to the PAA disinfection data,
although their intrinsic disinfection rate constant kd varied
with the initial disinfectant concentration, and a new fitting
parameter (the time t at which N(t) ¼ N0) was introduced. Here
we opt for the more parsimonious no-tailing model, for which
an exact solution can be derived:

kd =k
N ¼ N0 ðk C0 t þ 1Þ (8)
PAA disinfection data collected in this study were tested
against Chick’s Model (Eq. (1)) and the no-tailing model (Eq. (8))
in Fig. 3. The data are plotted so that the intrinsic disinfection
rate constant kd can be estimated directly from the slope b of
the best-fit line: kd ¼ 2.303b (Chick’s Law) or kd ¼ b (no-tailing
model). Also shown in Fig. 3 are two estimates of model
performance, including the Pearson’s r2 correlation between
logN/N0 and the x-axis (either C0t or log½k C0 t þ 1=k ), and the
root mean square error (RMSE) between modeled and
measured values of bacterial log reduction. In general, for
a given choice of fecal bacteria group (either ENT or EC) the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5647

Fig. 3 e PAA disinfection data for EC (panel A and B) and ENT (panel C and D) fit to either Chick’s model (Eq. (1), panel A and C)
or Wagner et al.’s no-tailing model (Eq. (8), panel B and D). Values for kd are mML1 minL1 and units for RMSE are logN/N0.

two disinfection models have similar r2 and RMSE values
(compare panels A and C with panels B and D in Fig. 3). Both
models are a better description of ENT disinfection (r2 ¼ 0.68 to
0.72, RMSE ¼ 0.32e0.37) than EC disinfection (r2 ¼ 0.52 to 0.55,
RMSE ¼ 0.6e0.62). The fact that ENT exhibited less dispersion
around the no-tailing model may reflect less variability
(compared to EC) in the disinfection resistance of enterococci
bacteria populations present in the sewage and source waters.
Intrinsic rate constants estimated from the slope b are also
similar for the two models. Chick’s Law yields intrinsic rate
constants for EC and ENT of 3.0 and 1.9 mM
1 min
1, respec-
tively (panels A and C). The no-tailing model yields intrinsic
rate constants for EC and ENT of 3.3 and 3.5 mM
1 min
1,
respectively (panels B and D). The intrinsic disinfection rate
for EC is also similar to values estimated by averaging rate
constants obtained from our individual experiments (see
Section 3), and from data reported in other studies of PAA
disinfection (Table 1). In summary, Chick’s Law and the no-
tailing model were both reasonably good predictors of bacte-
rial decay caused by PAA disinfection, and both yield values of
the intrinsic disinfection rate constant kd that were consistent
across models, and across studies. Relative to the modeling
effort described in the next section, the primary benefit of the
no-tailing model is that it accounts for the decay in disinfec-
tant concentration that will inevitably occur following injec-
tion of PAA into the subsurface.
5. Lagrangian model of in situ disinfection
In this section we develop a quantitative model that accounts
for the physical, chemical, and biological factors that might
influence the in-situ disinfection of sewage contaminated
shallow groundwater with PAA. The model is developed in
two stages. First, a Lagrangian framework is used to predict
the concentration of both sewage constituents (fecal indicator
bacteria) and PAA residual in a parcel of shallow groundwater
as it travels from the point of injection to the point where it is
discharged to the coastal ocean. Second, an analytical model
is derived for fluid parcel residence time in shallow
5648 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

groundwater, for the situation (relevant to Avalon Bay) where
shallow groundwater discharge is dominated by tidal pump-
ing and meteoric flow.
5.1. Lagrangian framework for modeling disinfection
kinetics
Fig. 4 illustrates the physical transport processes that might
affect fecal bacteria concentration in a parcel of groundwater
as it moves from the point where it first encounters subsur-
face injected disinfectant (point A) to the point where it exits
the shallow groundwater and is discharged to the coastal
ocean (point B). As the parcel moves from A to B, the
concentration of bacteria in the fluid parcel changes with time
due to disinfection, non-disinfection related die-off, regrowth,
and attachment to the porous matrix (filtration). Mass balance
over a fluid parcel as it travels from A to B yields the following
rate expression for the concentration of bacteria (N, bacteria
L
3), where s [min
1] represents the time a sewage contami-
nated water parcel has been in contact with disinfectant
(referred to here as “residence time”), C is the disinfectant
concentration [mM], and the rate constants for disinfection,
inactivation, filtration, and growth are kd [mM
1 min
1], ki
[min
1], kf [min
1], and mg [min
1]:
dN
1

¼
kd N½1=C0 þ k s
ki þ kf N þ mg N
ds
In formulating Eq. (9), we adopted the no-tailing version of
Hom’s Law described earlier (Eq. (7) with m ¼ 0 and n ¼ 1)
and assume that inactivation, filtration, and regrowth of
bacteria all follow first-order kinetics. While we did not
(e.g., Lazarova et al., 1998; Lefevre et al., 1992). Here we use
the Monod equation (Levenspiel, 1980) to model the
dependence of growth rate mg on dissolved organic carbon
(DOCT):
mg;max DOCT ðsÞ
mg ðsÞ ¼ (10)
Ks þ DOCT ðsÞ
where, mg,max and Ks represent the maximum growth rate and
saturation constant, respectively. Based on the measurements
of DOC presented in Section 3.1, potential sources of DOC
include ambient groundwater (DOCGW), sewage (DOCSewage),
and acetic acid associated with the PAA mixture, including
acetic acid present as an equilibrium component of commer-
cial PAA preparations (DOCAA,0), and acetic acid formed by the
decomposition of PAA during disinfection (DOCA(s)), where the
latter can be estimated from the loss of peroxycompound
concentration with time:
DOCAA ðsÞ ¼ 0:82MðC0
CðsÞÞ (11)
Eq. (11) assumes stoichiometric conversion of PAA to acetic
acid, taking into account the molar fraction (0.82) of the per-
oxycompound concentration that is PAA and the weight of
carbon associated with every mole of acetic acid, M ¼ 25 g of
carbon per mole. Combining Eqs. (10) and (11), we have the
following prediction for the total DOC available for growth of
fecal indicator bacteria:
(9) DOCT ðsÞ ¼ DOCGW þ DOCSewage þ DOCAA;0 þ 0:82MðC0
CðsÞÞ: (12)
Combining Eqs. (9)e(12) and solving the resulting differential
equation, yields the following formula for the concentration of
bacteria in a parcel of shallow groundwater as a function of
residence time (s):

 

kf ðC0 M þ Ks Þ þ C0 Mki
C0 Mmg;max þ DOCT0 kf þ ki
mg;max þ ki Ks 
a
N0 ðDOCT0 þ Ks Þ exp
s
C0 M þ DOCT0 þ Ks
NðsÞ ¼ kd =k a
(13a)
ðC0 k s þ 1Þ ½C0 k sðMC0 þ DOCT0 þ Ks Þ þ DOCT0 þ Ks 

observe regrowth in the disinfection experiments presented

earlier, it is a well-known that the acetic acid in commercial Ks mg M
a¼ (13b)
PAA mixtures can serve as a carbon source for the growth k ðC0 M þ DOCT0 þ Ks Þ2
of heterotrophic bacteria, including fecal indicator bacteria

Fig. 4 e A conceptual model for the in-situ disinfection of sewage-contaminated shallow groundwater in coastal marine
environments.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5649

Fig. 5A presents the bacterial log reduction predicted by Eq.
(13) (solid curves) as a function of initial PAA concentration
(vertical axis) and residence time (horizontal axis). This graph
was generated using the parameter values listed in Table 2,
which are based on the experimental measurements pre-
sented earlier and field conditions relevant to the Avalon Bay
field site, including a shallow groundwater temperature of
15  C. For the range of peroxycompound concentrations
considered in Fig. 5A, the dependence of bacteria concentra-
tion on residence time exhibits two patterns. At small resi-
dence times, very little of the PAA has been converted to acetic
acid, disinfection dominates, and bacteria concentration
declines rapidly with increasing residence time. At longer
residence times, PAA has been mostly converted to acetic
acid, regrowth of bacteria dominates, and bacterial concen-
tration increases with increasing residence time. This decay/
regrowth pattern is illustrated for a single initial perox-
ycompound concentration (C0 ¼ 0.03 mM) and temperature
(T ¼ 15  C) in Fig. 5B (dotted line). For this particular choice of
parameter values, the model predicts that bacterial concen-
tration falls approximately 2 log units within 12 h, and
increases thereafter, eventually rising above the initial
bacteria concentration at a residence time of around two days.
As expected, bacteria removal increases monotonically with
increasing initial peroxycompound concentration, as illus-
trated for a fixed residence time (s ¼ 5 days) and temperature
(T ¼ 15  C) in Fig. 5C (dotted line). Using the activation energies
for the intrinsic disinfection rate constant reported in Section
3.3, the model predicts similar trends over the range of
temperatures (13e17  C) typically measured in Avalon shallow
groundwater (compare the family of curves in Fig. 5B and C).
In an ideal scenario, by the time a fluid parcel of shallow
groundwater is discharged to the ocean, both its bacterial
concentration and PAA residual would be very small. The
dependence of PAA residual on initial peroxycompound
concentration and residence time can be estimated from Eq.
(6), by setting the left hand side equal to a fixed residual
concentration, Cres and solving for residence time s:
1

1 
b
sres ¼ C
C
1 C0 (14)
a res 0
Curves of constant peroxycompound residual are plotted in
Fig. 5A (dashed curves). As expected, peroxycompound
residual declines with increasing residence time for a fixed C0.
Interestingly, the curve for Cres ¼ 1 mM roughly demarcates the
transition from bacteria disinfection to regrowth (compare
solid and dotted lines in Fig. 5A).
5.2. Residence time of fluid parcels
The Lagrangian model presented above reveals that both the
bacterial concentration and PAA residual were sensitive to the
residence time of water parcels in shallow groundwater, and
in this section we describe some of the physical processes that
can affect this key parameter. Here, residence time has
precisely the same meaning as in estuarine systems: “how
long a parcel, starting from a specified location within
a waterbody, will remain in the waterbody before exiting”
(Monsen et al., 2002). Provided that material diffusion can be
neglected (a key assumption in the Lagrangian approach
adopted here) (Deleersnijder et al., 2001), the residence time

Fig. 5 e Model predictions for EC concentration and PAA residual for the in-situ disinfection of sewage contaminated shallow
groundwater with PAA. Panel A: Curves of constant EC log reduction predicted by Eq. (13) (solid lines labeled with numbers
ranging from L1 to L9), and curves of constant peroxycompound residual predicted by Eq. (14) (dashed curves, labeled with
numbers ranging from 0.2 to 1 mM). Panel B: Change in EC concentration with residence time predicted by Eq. (13) for
C0 [ 0.03 mM and the ambient groundwater temperatures shown. (C) Change in EC concentration with increasing initial
PAA concentration predicted by Eq. (13) for a fixed residence time of s [ 5 days and the ambient groundwater temperatures
shown. Parameter values used to generate these curves are listed in Table 2.
5650 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

Table 2 e Inputs parameters used to produce the disinfection contour plot (Fig. 5), residence time model (Fig. 6).
Constant Name Value Source
(b
1)
1
a k* fitting parameter 0.0093 mM min Wagner et al.
b k* fitting parameter 1.42 Wagner et al.
kd EC disinfection rate (T ¼ 13,15,17  C) 3.3, 3.7, 4.1 mM
1 min
1 Measured
ki EC natural decay rate 0.1 h
1 Estimated
kf EC filtration rate 0 min
1 Estimated
mg Max growth rate 0.005 min
1 Surbeck et al.
Ks Growth rate saturation constant 0.07 Surbeck et al.
DOCT0 Initial DOC Concentration mg L
1 Measured
M Mass of carbon per mole PAA 25 g Calculated
A Wave amplitude 1m Estimated
Kp Hydraulic conductivity 5  10
4 m s
1 Calculated via KozenyeCarman
l Wave number 0.05 m
1 Estimated

associated with the movement of a fluid parcel from point A to
B in Fig. 4 can be written explicitly as follows:
ZL
dx
sc ¼
v (15)e(17) can be combined to yield the following estimate for the
0
where, v is the velocity experienced by the fluid particle as it
moves toward the ocean. Here we have subscripted the resi-
dence time calculated from Eq. (15) (sc) to distinguish it from
the actual residence time of a water parcel included in our
disinfection model above (s). Fluid parcel velocity has contri-
butions from tidal pumping (vt), wave set-up (vw), meteoric
groundwater flow (vm), seasonal evapotranspiration (vs), and
density driven flow (vd) (Burnett et al., 2003b; Michael
et al., 2005):
v ¼ vt þ vw þ vm þ vs þ vd
Given that tidal pumping appears to dominate the discharge
of shallow groundwater to Avalon Bay (Boehm et al., 2009b),
here we focus on the tidal pumping and meteoric terms in Eq.
(16), vt and vm. Flow fields generated by tidal pumping can be
approximated from the following expression (Nielsen, 1990):
vt ðx; tÞ ¼ Kp Ale
lx ðcos½ut
lx
sin½ut
lxÞ
with the falling tide) passes point A in Fig. 4, which is mathe-
matically equivalent to assigning the value p to the quantity
(ut
lx) in Eq. (17). After invoking this simplification and
allowing for the possibility of non-zero meteoric flow, Eqs.
(15)
characteristic residence time of a fluid parcel released at a set-
back distance x ¼ L from the beach:
 
log
AlKp
elL vm
log
AlKp
vm
sc ¼ ; vm > 0 (18a)
vm
1  lL
sc ¼ e
1 ; vm ¼ 0 (18b)
Al2 Kp
For the range of parameter values typical of the field site in
Avalon Bay (see Table 2), the residence times predicted by Eq.
(18) vary over one hundred thousand fold, from approximately
(16)
30 mine1000 days (Fig. 6). This variability in residence time
derives, in part, from the non-linear dependence of residence
time on both set-back distance and meteoric flow (Eq. (18a)).
Furthermore, if studies of residence times in estuaries are any
guide, the residence time sc of water parcels in shallow
groundwater is best characterized by a probability distribu-
tion, not a single value. Studies in estuarine systems have
noted that fluid parcel residence times tend to follow proba-
(17) bility distributions characterized by long tails, implying that

Nielsen derived Eq. (17) from the Boussenesq equation after

invoking a number of simplifying assumptions, including
a homogeneous unconfined aquifer of hydraulic conductivity
KP, a vertical beach face, and a one-dimensional flow field
(parallel to the x-axis, see Fig. 4) characterized by a wave number
l, and forced by a single harmonic tide with amplitude A and
angular frequency u. The wave number is defined as l ¼ 2p/x0,
where x0 represents the inland distance over which tidal fluc-
tuations in the shallow groundwater are significant. Because
the flow field predicted by Eq. (17) is tidally periodic and spatially
variable, the residence time of a fluid parcel will depend not only
on where in the aquifer it is released (referred to here as the set-
back distance, x ¼ L, see Fig. 4) but also on when in the tidal cycle
that release occurs; a very similar phenomenon has been
described for the residence time distributions in coastal estu-
aries (Monsen et al., 2002; Oliveira and Baptista, 1997). Despite
these complications, a characteristic residence time can be Fig. 6 e Characteristic shallow groundwater residence
estimated from Eqs. (15)e(17) by releasing the fluid parcel at the times predicted by Eq. (18) for various set-back distances
precise moment when the recessional tide wave (associated and meteoric flow velocities.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
Table 3 e Inputs Parameters used in the uncertainty analysis (Eq. (19)).
Constant Name
k* PAA decay rate
kd EC disinfection rate (T ¼ 15  C)
ki EC natural decay rate
mg Max growth rate
Ks Growth rate saturation constant
DOCT0 Initial DOC concentration
C0 Initial peroxycompounds
s Residence time
a minority of fluid parcels spend a very long time in the
estuary (Oliveira and Baptista, 1997).
6. Uncertainty analysis and practical
implications
One advantage of the analytical in-situ disinfection model
derived earlier (Eq. (13)) is that the uncertainty associated with
different independent variables can be assessed quantita-
tively using the Law of Uncertainty (Taylor and Kuyatt, 1993):
X
p  2
vS
u2 ðSÞ ¼ u2 ðXi Þ
i¼1
vXi
In this equation, u2(S ) and u2(Xi) represents the variance of the
dependent and independent variables, respectively, the
dependent variable is the log-transformed bacteria concen-
tration (S ¼ log(N )), the independent variables Xi include all
variables appearing on the right hand side of Eq. (13) (i.e., C0,
Ks, mg, DOCT0, k*, kd, ki, or s), vS/vXi is the sensitivity of the
dependent variable to change in a particular independent
variable (computed analytically from Eq. (13)), and the
summation is taken over all independent variables (P ¼ 8). The
uncertainty (or variance) associated with each independent
variable was estimated from the relative uncertainty UR(Xi)
and magnitude jXi j values listed in Table 3:
uðXi Þ ¼ UR ðXi ÞjXi j
The form of the Law of Uncertainty adopted here assumes
that independent variables do not co-vary, which is reason-
able for most combinations of independent variables included
in this analysis. When applied to Eq. (13), the Law of Uncer-
tainty reveals that 99% of the variance in the log-transformed
bacteria concentration can be attributed to variance in just
three independent variables: residence time (s) (90%),
maximum growth rate (mg) (8%), and the inactivation rate (ki)
(1%). These results imply that the prediction (and optimiza-
tion) of in-situ disinfection will depend strongly on the resi-
dence time of shallow groundwater which, in turn, depends
non-linearly on the injection well set-back distance (L) and
physical characteristics of the shallow groundwater system
that can vary in time and space (l, A, KP, vm) (see Eq. (18)).
Given the very approximate nature of the analysis that led to
Eq. (18), experimental characterization of shallow ground-
water residence times would be a fruitful topic for further
investigation.
5651
Value Estimated
uncertainty
1.35 L$mmol
1$min
1 0.2
3.7 L$mmol
1$min
1 0.2
0.1 h
1 0.2
0.005 min
1 0.2
0.07 0.2
0.006 g/ L
1 0.2
0.03 mmol 0.2
2 day 1
The disinfection model’s sensitivity to residence time is,
in part, a consequence of the fact that fecal bacteria can
grow in the environment, and thus dramatically different
disinfection outcomes (e.g., from a net reduction in bacteria
concentrations to a net increase in bacteria) can be caused
by slight changes in residence time of water parcels in the
shallow groundwater. Given that most recreational water-
borne illnesses are caused by human viruses that cannot
grow outside their host (Schoen et al., 2011), it is possible
that in-situ disinfection of sewage contaminated shallow
groundwater would reduce shoreline concentrations of
human pathogens (and hence lower recreational waterborne
illness rates), even if it did not substantially reduce fecal
(19)
indicator bacteria concentrations. Indeed, the environ-
mental growth of EC and ENT can lead to a decoupling
between fecal indicator bacteria and human pathogens in
recreational waters (Litton et al., 2010), and potentially
nullify epidemiological relationships upon which current
fecal indicator bacteria criteria are based (Colford et al.,
2007). In light of these and other concerns the U.S Environ-
mental Protection Agency is evaluating and possibly revising
the current water quality criteria for marine recreational
beaches (Boehm et al., 2009a).
7. Conclusions
(20)
 Aeration alone and aeration with brine did not significantly
reduce EC and ENT concentrations in mixtures of raw
sewage and shallow groundwater after 6 h of exposure,
while 4e5 mg L
1 of PAA and 4 mg L
1 Oxone achieved >3
log reduction of EC and ENT after 15 min of exposure.
 Oxone disinfection is enhanced at higher salinities, most
likely due to the formation of secondary oxidants (e.g.,
chlorine and bromine) that make this disinfectant inap-
propriate for marine applications.
 PAA disinfection of fecal bacteria in shallow groundwater in
coastal settings depends non-linearly on residence time,
and the “ideal” disinfection outcome (low bacterial
concentration and low PAA residual) is achieved over
a relatively narrow window of residence times and initial
disinfection concentrations.
 By analogy to surface estuaries, the residence time of water
parcels in shallow groundwater under the influence of
marine tides (i.e., subterranean estuaries) is likely to exhibit
broad (e.g., logenormal) probability distributions with long
5652 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
tails, and depend sensitively on local precipitation, meteoric
flow, and tidal variability.
 Uncertainty calculations suggest 99% of the uncertainty
associated with the log reduction of bacteria is caused by
just three independent variables: residence time (s) (90%),
maximum growth rate (mg) (8%), and the inactivation rate
(ki) (1%).
 Given these results, further research into the residence time
of water in shallow groundwater is needed before an in-situ
disinfection schemes can be successfully designed and
implemented.
Acknowledgments
The authors thank R. Litton, L. Ho, and J. Monroe for assis-
tance with the experiments, and C. Wagner and P. Woolson,
and the City of Avalon staff for the use of City Hall for the
disinfection studies. Funding was provided by the City of
Avalon and State Water Resources Control Board Clean Bea-
ches Initiative, under Agreement 07-582-550.This is publica-
tion 66 of the Urban Water Research Center, University of
California, Irvine.
references
Abarnou, A., Miossec, L., 1992. Chlorinated waters discharged to
the marine-environment chemistry and environmental-
impact e an overview. Science of the Total Environment 126
(1e2), 173e197.
Allonier, A.S., Khalanski, M., Camel, V., Bermond, A., 1999.
Characterization of chlorination by-products in cooling
effluents of coastal nuclear power stations. Marine Pollution
Bulletin 38 (12), 1232e1241.
Anipsitakis, G.P., Dionysiou, D.D., 2003. Degradation of organic
contaminants in water with sulfate radicals generated by the
conjunction of peroxymonosulfate with cobalt.
Environmental Science & Technology 37 (20), 4790e4797.
Antonelli, M., Mezzanotte, V., Panouilleres, M., 2009. Assessment
of peracetic acid disinfected effluents by microbiotests.
Environmental Science & Technology 43 (17), 6579e6584.
Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D.,
Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and
shorter period variability of surf zone water quality at
Huntington Beach, California. Environmental Science &
Technology 36 (18), 3885e3892.
Boehm, A.B., Fuhrman, J.A., Mrse, R.D., Grant, S.B., 2003. Tiered
approach for identification of a human fecal pollution source
at a recreational beach: case study at Avalon Cay, Catalina
Island, California. Environmental Science & Technology 37 (4),
673e680.
Boehm, A.B., Shellenbarger, G.G., Paytan, A., 2004. Groundwater
discharge: potential association with fecal indicator bacteria
in the surf zone. Environmental Science & Technology 38 (13),
3558e3566.
Boehm, A.B., Ashbolt, N.J., Colford Jr., J.M., Dunbar, L.E., Fleming, L.E.,
Gold, M.A., Hansel, J.A., Hunter, P.R., Ichida, A.M., McGee, C.D.,
2009a. A sea change ahead for recreational water quality criteria.
Journal of Water and Health 7 (1), 9e20.
Boehm, A.B., Yamahara, K.M., Love, D.C., Peterson, B.M.,
McNeill, K., Nelson, K.L., 2009b. Covariation and
photoinactivation of traditional and novel indicator
organisms and human viruses at a sewage-impacted marine
beach. Environmental Science & Technology 43 (21),
8046e8052.
Booth, R., Lester, J., 1995. The potential formation of halogenated
by-products during peracetic acid treatment of final sewage
effluent. Water Research 29 (7), 1793e1801.
Burnett, W.C., Bokuniewicz, H., Huettel, M., Moore, W.S.,
Taniguchi, M., 2003a. Groundwater and pore water inputs to
the coastal zone. Biogeochemistry 66 (1), 3e33.
Burnett, W.C., Bokuniewicz, H., Huettel, M., Moore, W.S.,
Taniguchi, M., 2003b. Groundwater and pore water inputs to
the coastal zone. Biogeochemistry 66 (1e2), 3e33.
Chick, H., 1908. An investigation of the laws of disinfection.
Journal of Hygiene 8 (01), 92e158.
Colford Jr., J.M., Wade, T.J., Schiff, K.C., Wright, C.C., Griffith, J.F.,
Sandhu, S.K., Burns, S., Sobsey, M., Lovelace, G., Weisberg, S.B.,
2007. Water quality indicators and the risk of illness at beaches
with nonpoint sources of fecal contamination. Epidemiology
18 (1), 27.
Cristiani, P., 2005. Solutions to fouling in power station
condensers. Applied Thermal Engineering 25 (16), 2630e2640.
de Lafontaine, Y., Despatie, S.P., Wiley, C., 2008. Effectiveness and
potential toxicological impact of the PERACLEAN (R) ocean
ballast water treatment technology. Ecotoxicology and
Environmental Safety 71 (2), 355e369.
Deleersnijder, E., Campin, J.M., Delhez, E.J.M., 2001. The concept
of age in marine modelling I. Theory and preliminary model
results. Journal of Marine Systems 28 (3e4), 229e267.
Dell’Erba, A., Falsanisi, D., Liberti, L., Notarnicola, M., Santoro, D.,
2004. Disinfecting behaviour of peracetic acid for municipal
wastewater reuse. Desalination 168, 435e442.
Dell’Erba, A., Falsanisi, D., Liberti, L., Notarnicola, M., Santoro, D.,
2007. Disinfection by-products formation during wastewater
disinfection with peracetic acid1. Desalination 215 (1e3),
177e186.
Grant, S.B., Sanders, B.F., 2010. Beach boundary layer:
a framework for addressing recreational water quality
impairment at enclosed beaches. Environmental Science &
Technology, 4818e4824.
Guo, G.M., Lin, F.K., 2009. The bromination kinetics of phenolic
compounds in aqueous solution. Journal of Hazardous
Materials 170 (2e3), 645e651.
Hom, L.W., 1972. Kinetics of chlorine disinfection in an ecosystem.
Journal of the Sanitary Engineering Division 98 (1), 183e194.
Kitis, M., 2004. Disinfection of wastewater with peracetic acid:
a review. Environment International 30 (1), 47e55.
Lazarova, V., Janex, M., Fiksdal, L., Oberg, C., Barcina, I.,
Pommepuy, M., 1998. Advanced wastewater disinfection
technologies: short and long term efficiency. Water Science
and Technology 38 (12), 109e117.
Lefevre, F., Audic, J., Ferrand, F., 1992. Peracetic acid disinfection
of secondary effluents discharged off coastal seawater. Water
Science and Technology 25 (12), 155e164.
Levenspiel, O., 1980. The monod equation e a revisit and
a generalization to product inhibition situations.
Biotechnology and Bioengineering 22 (8), 1671e1687.
Liberti, L., Lopez, A., Notarnicola, M., 1999. Disinfection with
peracetic acid for domestic sewage re use in agriculture.
Water and Environment Journal 13 (4), 262e269.
Lipp, E.K., Farrah, S.A., Rose, J.B., 2001. Assessment and impact of
microbial fecal pollution and human enteric pathogens in
a coastal community. Marine Pollution Bulletin 42 (4),
286e293.
Litton, R.M., Ahn, J.H., Sercu, B., Holden, P.A., Sedlak, D.L., Grant, S.
B., 2010. Evaluation of chemical, molecular, and traditional
markers of fecal contamination in an effluent dominated urban
stream. Environmental Science & Technology.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
Manheim, F.T., Krantz, D.E., Bratton, J.F., 2004. Studying ground
water under Delmarva coastal bays using electrical resistivity.
Ground Water 42 (7), 1052e1068.
McIlvin, M.R., Altabet, M.A., 2005. Chemical conversion of nitrate
and nitrite to nitrous oxide for nitrogen and oxygen isotopic
analysis in freshwater and seawater. Analytical Chemistry 77
(17), 5589e5595.
Michael, H.A., Mulligan, A.E., Harvey, C.F., 2005. Seasonal
oscillations in water exchange between aquifers and the
coastal ocean. Nature 436 (7054), 1145e1148.
Mika, K.B., Imamura, G., Chang, C., Conway, V., Fernandez, G.,
Griffith, J.F., Kampalath, R.A., Lee, C.M., Lin, C.C., Moreno, R.,
Thompson, S., Whitman, R.L., Jay, J.A., 2009. Pilot- and bench-
scale testing of faecal indicator bacteria survival in marine
beach sand near point sources. Journal of Applied
Microbiology 107 (1), 72e84.
Monarca, S., Feretti, D., Collivignarelli, C., Guzzella, L., Zerbini, I.,
Bertanza, G., Pedrazzani, R., 2000. The influence of different
disinfectants on mutagenicity and toxicity of urban
wastewater. Water Research 34 (17), 4261e4269.
Monsen, N.E., Cloern, J.E., Lucas, L.V., Monismith, S.G., 2002. A
comment on the use of flushing time, residence time, and age
as transport time scales. Limnology and Oceanography 47 (5),
1545e1553.
Moore, W.S., 1999. The subterranean estuary: a reaction zone of
ground water and sea water. Marine Chemistry 65 (1e2), 111e125.
Nielsen, P., 1990. Tidal dynamics of the water-table in beaches.
Water Resources Research 26 (9), 2127e2134.
Oliveira, A., Baptista, A.M., 1997. Diagnostic modeling of
residence times in estuaries. Water Resources Research 33 (8),
1935e1946.
Sanchez-Ruiz, C., MartÌnez-Royano, S., Tejero-MonzÛn, I., 1995.
An evaluation of the efficiency and impact of raw wastewater
disinfection with peracetic acid prior to ocean discharge.
Water Science and Technology 32 (7), 159e166.
Santoro, D., Gehr, R., Bartrand, T.A., Liberti, L., Notarnicola, M.,
Dell’Erba, A., Falsanisi, D., Haas, C.N., 2007. Wastewater
disinfection by peracetic acid: assessment of models for
5653
tracking residual measurements and inactivation. Water
Environment Research 79 (7), 775e787.
Santos, I.R., Niencheski, F., Burnett, W., Peterson, R., Chanton, J.,
Andrade, C.F.F., Milani, I.B., Schmidt, A., Knoeller, K., 2008.
Tracing anthropogenically driven groundwater discharge into
a coastal lagoon from southern Brazil. Journal of Hydrology
353 (3e4), 275e293.
Schoen, M.E., Soller, J.A., Ashbolt, N.J., 2011. Evaluating the
importance of faecal sources in human-impacted waters.
Water Research.
Stampi, S., De Luca, G., Onorato, M., Ambrogiani, E., Zanetti, F.,
2002. Peracetic acid as an alternative wastewater disinfectant
to chlorine dioxide. Journal of Applied Microbiology 93 (5),
725e731.
Swarzenski, P., Burnett, W., Greenwood, W., Herut, B., Peterson, R.,
Dimova, N., Shalem, Y., Yechieli, Y., Weinstein, Y., 2006.
Combined time-series resistivity and geochemical tracer
techniques to examine submarine groundwater discharge at
Dor Beach, Israel. Geophysical Research Letters 33 (24), L24405.
Taniguchi, M., Burnett, W.C., Cable, J.E., Turner, J.V., 2002.
Investigation of submarine groundwater discharge.
Hydrological Processes 16 (11), 2115e2129.
Taylor, B.N., Kuyatt, C.E., Standards N.I.o. and Technology, 1993.
Guidelines for Evaluating and Expressing the Uncertainty of
NIST Measurement Results, Citeseer.
USEPA, 2003. Guidelines Establishing Test Procedures for the
Analysis of Pollutants; Analytical Methods for Biological
Pollutants in Ambient Water, Final Rule US Federal Register,
40 CFR 68, 139, FRL-7529-7.
Wagner, M., Brumelis, D., Gehr, R., 2002. Disinfection of
wastewater by hydrogen peroxide or peracetic acid:
development of procedures for measurement of residual
disinfectant and application to a physicochemically treated
municipal effluent. Water Environment Research 74 (1),
33e50.
Watch F.a.W., 2008. Overloaded and Underfunded: Northwestern
California’s Aging Infrstructure Needs A Clean Water Trust
Fund. Food and Water Watch.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

In situ disinfection of sewage contaminated shallow

groundwater: A feasibility study

Morgan M. Bailey a, William J. Cooper b, Stanley B. Grant a,b,*

a
Department of Chemical Engineering and Material Sciences, University of California, Henry Samueli School of Engineering,
University of California, Irvine, CA 92697, USA
b
Department of Civil and Environmental Engineering, and Urban Water Research Center, Henry Samueli School of Engineering,
University of California, Irvine, CA 92697, USA

article info abstract

Article history: Sewage-contaminated shallow groundwater is a potential cause of beach closures and
Received 18 June 2011 water quality impairment in marine coastal communities. In this study we set out to
Received in revised form evaluate the feasibility of several strategies for disinfecting sewage-contaminated shallow
9 August 2011 groundwater before it reaches the coastline. The disinfection rates of Escherichia coli (EC)
Accepted 14 August 2011 and enterococci bacteria (ENT) were measured in mixtures of raw sewage and brackish
Available online 22 August 2011 shallow groundwater collected from a coastal community in southern California. Different
disinfection strategies were explored, ranging from benign (aeration alone, and aeration
Keywords: with addition of brine) to aggressive (chemical disinfectants peracetic acid (PAA) or per-
Disinfection oxymonosulfate (Oxone)). Aeration alone and aeration with brine did not significantly
Shallow groundwater reduce the concentration of EC and ENT after 6 h of exposure, while 4e5 mg L
1 of PAA or
Sewage Oxone achieved >3 log reduction after 15 min of exposure. Oxone disinfection was more
Remediation rapid at higher salinities, most likely due to the formation of secondary oxidants (e.g.,
Water quality bromine and chlorine) that make this disinfectant inappropriate for marine applications.
Peroxymonosulfate Using a Lagrangian modeling framework, we identify several factors that could influence
Peracetic acid the performance of in-situ disinfection with PAA, including the potential for bacterial
Oxone regrowth, and the non-linear dependence of disinfection rate upon the residence time of
Brine water in the shallow groundwater. The data and analysis presented in this paper provide
a framework for evaluating the feasibility of in-situ disinfection of shallow groundwater,
and elucidate several topics that warrant further investigation.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction identified as a source of fecal pollution at some coastal bea-

Beach closures and advisories are generally caused by the
coastal discharge of land-side sources of fecal pollution such
as urban runoff or sewage effluent (Boehm et al., 2002; Grant
and Sanders, 2010). However, contamination of shallow
groundwater by aging sewage collection systems and failing
onsite sewage treatment and disposal systems has also been
ches (Lipp et al., 2001). The latter typically involves a two-step
process: (1) shallow groundwater is first contaminated by the
surface or subsurface release of sewage, and (2) the shallow
groundwater is then discharged to coastal waters under the
influence of land hydraulic gradients, tidal pumping, and/or
current induced pressure gradients (Burnett et al., 2003b).
The coastal discharge of sewage-contaminated shallow

* Corresponding author. Tel.: þ1 949 824 8277; fax: þ1 949 824 2541.
E-mail addresses: mbailey@uci.edu (M.M. Bailey), sbgrant@uci.edu (S.B. Grant).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.020
5642 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
groundwater can impact near-shore water quality by
vectoring fecal-oral pathogens (or their indicators) directly
from sources of human sewage to receiving waters (Boehm
et al., 2003). The discharge may also deliver nutrients that
promote the survival, and perhaps growth, of fecal bacteria in
near-shore waters (Boehm et al., 2004).
To date, most studies on shallow groundwater discharge
focus on its quantification using isotopic (McIlvin and Altabet,
2005), geophysical (Manheim et al., 2004; Santos et al., 2008;
Swarzenski et al., 2006), or hydrological methods (Burnett
et al., 2003a; Michael et al., 2005; Taniguchi et al., 2002), but
to our knowledge there are no published reports on the very
practical and important question of how to remediate shallow
groundwater once it becomes contaminated with sewage.
Given the advancing age and poor condition of sewage
collection systems and onsite sewage treatment and disposal
systems in many coastal areas of the U.S. (Food and Water
Watch, 2008), a cost-effective strategy for remediating
sewage-derived fecal pollution in brackish shallow ground-
water is urgently needed for the sustainable use of coastal
beaches and bays.
While the best remediation strategy should always be to
identify and eliminate the source(s) of sewage responsible for
groundwater contamination, logistical and economic
constraints may make such action infeasible in the near term,
in which case interim solutions are needed. Further, even if
a source of sewage contamination is identified and repaired, it
may take some time before pathogens and indicator bacteria
decay to acceptable concentrations, particularly when the
sediments in question are subsurface and thus not exposed to
sunlight (Mika et al., 2009). Subsurface injection of disinfec-
tant (or “in-situ disinfection”) is one obvious interim strategy
for remediating sewage-contaminated shallow groundwater,
but several issues need to be resolved before such an approach
can be implemented in practice. (1) Disinfectants can react
with organic matter and/or trace anions (e.g., bromide and
chloride ions) in brackish coastal groundwater to produce
toxic disinfection by-products, and their discharge to coastal
zones could pose a health risk to bathers along the shoreline
and negatively impact sensitive marine ecosystems (Monarca
et al., 2000). (2) The shallow groundwater adjacent to marine
coastal areas can be thought of as “subterranean estuaries”, in
which fresh (meteoric) waters mix with seawater before dis-
charging to the ocean (Moore, 1999). Consequently, the flow in
shallow groundwater can vary substantially in space and
time, potentially affecting the disinfection contact times
water parcels experience before discharging to coastal waters.
(3) Sewage constituents, such as pathogens and their indica-
tors, may not mix fully over the vertical dimension of the
shallow groundwater, and therefore disinfection strategies
may need to be tailored to target sewage constituents where
they are located, for example near the top of the water table.
(4) Planning and implementation of in-situ disinfection would
greatly benefit from quantitative “design criteria” that account
for the rate at which the target organisms decay with time
upon exposure to disinfectant, the decay in disinfectant with
time, the time scale over which water parcels reside in the
shallow groundwater before discharge, and ecological
processes that promote the subsurface regrowth of fecal
bacteria.
In this paper we present laboratory data and modeling
efforts intended to, at least preliminarily, address the issues
raised above, using as a test case Avalon Bay, Catalina Island,
California, where chronic fecal contamination of near-shore
waters has been linked to sewage contamination of shallow
groundwater by leaking sewage collection systems beneath
the City (Boehm et al., 2003, 2009b). The goals of this study are
to: (1) test three strategies for inactivating sewage-associated
fecal bacteria from shallow groundwater, including aeration
alone, aeration with addition of brine, and aeration with
addition of either peracetic acid (PAA) or peroxymonosulfate
(commercially known as Oxone); (2) characterize the salinity
dependence and disinfection by-product formation potential
of the latter two disinfectants; (3) evaluate several models for
PAA disinfection; and (4) develop a Lagrangian modeling
framework for evaluating, and potentially designing, in-situ
disinfection of sewage-contaminated coastal shallow ground
waters.
2. Materials and methods
2.1. Choice of disinfectant
Bench-scale experiments were carried out to measure the
decay of sewage-associated fecal indicator bacteria in shallow
groundwater upon exposure to: (1) aeration alone, (2) aeration
with addition of brine, (3) aeration with addition of PAA (35%
solution, CAS-79-21-0, Pfaltz & Bauer, CT), and (4) aeration
with addition of Oxone (43% dry weight, CAS-37222-66-5, Alfa
Aesar, Ward Hill, MA). The impact of aeration on bacterial die-
off was evaluated on the premise that, if efficacious, injection
of ambient air into the subsurface would be a straightforward,
economical, and environmentally benign approach for reme-
diating sewage-contaminated shallow groundwater. The
choice of brine as a candidate disinfectant was motivated by
the fact that its injection into shallow groundwater is unlikely
to cause environmental harm, and because the field site under
consideration, Avalon Bay, has a desalination plant from
which brine is produced as a waste product. PAA is a prom-
ising disinfectant for this application, because it produces
little or no toxic disinfection by-products when mixed with
seawater (it degrades into acetic acid, vinegar) (Kitis, 2004), is
a strong biocide in marine waters as evidenced by its use as an
antifouling agent in cooling water systems for coastal power
plants, and is regarded as relatively safe for discharge to
sensitive marine waters (Sanchez-Ruiz et al., 1995) including
the highly regulated Italian Lagoon of Venice (Cristiani, 2005).
To our knowledge, Oxone has not been tested as a biocide in
coastal marine settings, but its application to this particular
problem was motivated by the fact that, upon addition to
water, it generates hydroxyl radicals and sulfate ions
(Anipsitakis and Dionysiou, 2003); the former should accel-
erate the die-off of bacteria and viruses while the latter is
already present at high concentrations in marine waters.
Many commonly used disinfectants (e.g., ozone, chlorine gas,
sodium hypochlorite) were excluded from consideration for
logistical reasons and, although highly effective biocides, they
would likely react with trace anions and organics in sewage-
contaminated brackish groundwater to form toxic
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
disinfection by-products (Abarnou and Miossec, 1992; Allonier
et al., 1999).
2.2. Experimental characterization of disinfection kinetics
Raw (untreated) sewage was diluted 1:100 using various
mixtures (with various final salinities) of three source waters: (1)
relatively fresh (salinity ca. 1) water from an inland groundwater
well, (2) Avalon Bay water (salinity ca. 32), and (3) desalination
plant brine (salinity ca. 45). All source waters were aseptically
collected from the City of Avalon using sterile polypropylene
bottles and mixed with raw sewage collected from influent to
the City of Avalon wastewater treatment plant. Source waters
were not sterilized prior to mixing with sewage, and therefore
bacteria populations present in the final mixtures of sewage and
source waters probably included (minor) contributions from the
latter. Bench-scale disinfection experiments commenced
within 6 h of sample collection and were carried out in 4 L Nal-
gene polypropylene reaction vessels maintained in the dark and
partially submerged in a temperature controlled and recircu-
lating water bath at 15  1  C. Contents of the reactor vessel were
continuously aerated using a frit through which ambient air was
driven by an aquarium pump. To capture a range of salinities
and disinfectant concentrations, a matrix design was adopted in
which over 45 separate disinfection experiments (including 10
no-disinfectant-added control experiments) were conducted at
five different disinfectant concentrations (ranging from 0 to
7.3 mg L
1 for Oxone, and 0e8 mg L
1 for PAA) and four different
salinities (0, 15, 32, 45). The concentration range adopted for PAA
was based on previously published studies with this disinfec-
tant (Kitis, 2004; Santoro et al., 2007). To our knowledge, this is
the first study to evaluate Oxone disinfection in saline mixtures,
and thus the concentration range adopted for this disinfectant
was based on pilot studies (not reported) with this disinfectant.
Each batch reactor was sampled seven times over a period of
1e6 h (depending on experiment). This sampling schedule was
chosen to resolve disinfection over a single ebb tide, which
represents a theoretical minimum time a fluid parcel would be
in contact with disinfectant before discharging to Avalon Bay,
assuming that the discharge of shallow groundwater was under
tidal control. Samples were extracted from the reactor using
either a sterile syringe or pipet, analyzed for pH and conduc-
tivity, quenched by addition of approximately 0.4 mL of 0.1 N
sodium thiosulfate (CAS-7772-98-7, Mallinckrodt Chemicals),
immediately diluted either 1:10 or 1:100 in sterile deionized
water (Hardy Scientific, California), and enumerated for Escher-
ichia coli (EC) and enterococci bacteria (ENT) using Colilert-18 and
Enterolert defined substrate tests implemented in a 97-Well
Quanti-Tray format (IDEXX Laboratories, Maine). The adoption
of IDEXX Colilert-18 and Enterolert was motivated based on the
fact that both tests are approved by the U.S. Environmental
Protection Agency for enumerating EC and ENT bacteria in
ambient waters (USEPA, 2003) and, more to the point, are used
by the Los Angeles Department of Health Services in their
routine monitoring of recreational beach water quality in
Avalon Bay, and as a basis for management decisions regarding,
for example, the posting of Avalon beaches as unfit for swim-
ming. For a subset of the disinfection experiments, dissolved
organic carbon (DOC) and total organic carbon (TOC) concen-
trations were measured: (1) on the source water (groundwater or
5643
bay water) prior to addition of sewage and disinfectant; (2) on the
source water plus sewage prior to the addition of disinfectant;
and (3) at the end of the disinfection experiment. Samples
analyzed for DOC and TOC were collected in 500 mL amber glass
bottles and acidified with 2 mL of hydrochloric acid, and then
analyzed by TestAmerica using Standard Method 5310B.
2.3. Disinfection rate constants
Two different disinfection rate constants are reported in this
study: (1) an effective first-order disinfection rate constant k0d ,
and (2) an intrinsic disinfection rate constant kd. Values of the
effective first-order rate constant k0d were estimated by
regressing log-transformed bacteria concentration against
time based on Chick’s Law (Chick, 1908):
NðtÞ
ln ¼
k0d t (1)
N0
In Eq. (1), N(t) represents the concentration of bacteria in the
reaction vessel at any time t and N0 represents the bacterial
concentration present in the mixture at the start of the
disinfection experiment. The regression excluded any bacte-
rial measurements in the initial lag period and bacterial
measurements that fell below the lower-limit of detection for
the Colilert and Enterolert assays (in most cases, the lower
limit of detection was 10 most probable number (MPN) per
100 mL of sample).
In the case of PAA disinfection, an intrinsic disinfection
rate constant was also calculated, which represents the
susceptibility of bacteria to disinfectant independent of
disinfectant concentration:
kd ¼ k0d =C0 (2)
The variable C0 represents the initial (molar) concentration of
peroxycompounds which, in the case of commercial prepa-
rations of PAA, includes both PAA and hydrogen peroxide
(Wagner et al., 2002).
2.4. Activation energy for PAA disinfection
Using the reactor set-up described above, four separate
disinfection experiments were carried out at five tempera-
tures (T ¼ 5, 10, 15, 20 and 30  C) and a fixed concentration of
PAA (6 mg L
1) to determine the temperature sensitivity of EC
and ENT disinfection by PAA. Intrinsic disinfection rate
constants kd calculated from these experiments (see Section
2.3) were used to determine activation energies for the PAA
disinfection of EC and ENT, based on an Arrhenius plot of
ln(kd) against 1/T.
3. Disinfection results and discussion
3.1. Measurements of pH, TOC, and DOC
Over aeration of the reactor vessel caused the pH to increase
slightly from 7.9  0.2 to 8.3  0.2. No other systematic pH
trends were observed across the different experiments and
different source waters tested. The average DOC concentra-
tion in groundwater ranged from 1.6  0.14 mg L
1 (N ¼ 7,
5644 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
foreshore well, salinity 20) to 2.9  0.2 mg L
1 (N ¼ 4, inland
well, salinity 1). The TOC/DOC ratio was near unity in water
collected from the inland well (1.1 0.1, N ¼ 4) implying that
most of the organic carbon was dissolved. Raw sewage
collected from the Avalon Sewage Treatment Plant had much
higher levels of DOC (25 mg L
1, N ¼ 1). However, sewage
contributed only background levels of DOC (<0.25 mg L
1) to
the final 1:100 mixtures of sewage and source waters used in
the disinfection experiments. For disinfection experiments
carried out with PAA, the largest source of DOC was frequently
the PAA solution itself, which contained upwards of
280 mg L
1 of carbon as acetic acid. The acetic acid was
present as an equilibrium component of the commercial PAA
preparation, and generated as a breakdown product of PAA.
3.2. Experimental characterization of disinfection
kinetics
First-order effective disinfection coefficients ðk0d Þ estimated
from the 45 separate disinfection experiments are contoured
using Delaunay triangulation (Igor Pro v 6.10, Lake Oswego,
Oregon) against disinfectant concentration and salinity in
Fig. 1. These results are described in the sections below. A more
detailed analysis of k0d values, and an evaluation of different
disinfection models, is presented later in the paper (Section 4).
3.2.1. Aeration alone and aeration with addition of brine
We started by evaluating two environmental benign
approaches for removing fecal indicator bacteria from sewage-
contaminated shallow groundwater; namely, injection of
ambient air and/or injection of brine produced from a local
Fig. 1 e Contour plots of the effective first-order rate constants k0d [minL1] for PAA disinfection of EC (panel A) or ENT (panel
B), and Oxone disinfection of EC (panel C) or ENT (panel D).
desalination plant. Aeration alone and aeration with brine did
not significantly reduce EC and ENT concentrations after 6 h of
exposure; i.e., within the resolution of these experiments
k0d ¼ 0. Thus these two strategies are unlikely to be effective
against indicator bacteria in shallow groundwater over the time
scale of a single ebb tide.
3.2.2. Aeration with addition of PAA and oxone
EC and ENT concentrations were reduced by more than 1000
fold (3 log units) after 15 min of exposure to 4e5 mg L
1 of PAA.
The effective disinfection rate constants calculated for these
experiments (Fig. 1A and B): (1) increased monotonically with

1
PAA dose up to the maximum rate ðk0d ¼ 0:2 min Þ resolvable
with our experimental set-up; (2) do not depend, at least not
dramatically, on the salinity of the disinfection mixture; and
(3) are larger for EC than ENT at a fixed PAA dose. The last
observation is consistent with the results reported in Stampi
et al. (2002). Averaged across all experiments (and excluding
any experiments where the effective disinfection rate
constant exceeded 0.2 min
1), the average intrinsic rate
constants for PAA disinfection are kENT d ¼ 2:3  0:57 and

1
1
d ¼ 3:7  0:2 mM min : The estimate for kd is similar to
kEC EC
intrinsic rate constants estimated from previously published
data for PAA disinfection of EC in secondary settled effluent
(Dell’Erba et al., 2004) and PAA disinfection of fecal coliform
(FC) in secondary-treated sewage effluent (Wagner et al., 2002)
(Table 1). The fact that the intrinsic rate constants for PAA
disinfection of EC and FC are similar across these three studies
is notable, given that EC is a subset of FC, and the very
different initial bacterial concentrations (and presumably
sewage content) associated with the different source waters
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5645

Table 1 e Intrinsic disinfection rates and parameters from multiple studies.

Source Value kd (mM
1 min
1)a %H2 O2=%PAAb Temp ( C) N0 c Solutiond

FC EC ENT

This study NA 3.7  0.2 2.3  0.57 0.183 15 104 GW, BW,
Br, TPI
Dell’Erba et al. NA 3.65  2.41 NA 1.53 NR 102 TPE
Wagner et al. 1.04  1.26 NA NA 1.54 NR 105 TPE

NR-not reported. NA-not applicable.

a Intrinsic disinfection rate.
b Ratio of percent mass of PAA and hydrogen peroxide in PAA equilibrium solution.
c Magnitude of initial bacteria concentration.
d Sources waters for disinfection studies (GW-groundwater, BW-way water, Br-brine, TPI-treatment plant influent, TPE-treatment plant
effluent).

used in these studies, ranging from 102 to 105 bacteria per
100 mL (Table 1).
EC and ENT concentrations were reduced by more than 3
log units after 15 min of exposure to 4 mg L
1 Oxone. However,
unlike PAA, the disinfection rate depends on both Oxone dose
and solution salinity (Fig. 1C and D). At low Oxone dose
(<2 mg L
1), the effective disinfection rate increases mono-
tonically with Oxone dose, and exhibits no obvious salinity
dependence (i.e., the contour lines are near vertical in this
region of the plot). At higher Oxone dose, the effective disin-
fection rate increased monotonically with salinity, and
exhibits no obvious dose dependence (i.e., the contour lines
were near horizontal in this region of the plot). One possible
explanation for this pattern is that Oxone may oxidize anions
in the shallow groundwater and brine to yield, for example,
the secondary oxidants chlorine and bromine.
To explore this idea, a set of control experiments were
carried out in which either PAA or Oxone was added to an
aqueous solution consisting of 35 g L
1 NaCl to mimic the
background salinity of seawater and varying concentrations
(0e70 mg L
1) of KBr to mimic the presence of trace anions.
The formation of oxidized forms of chloride and bromide ions
(e.g., chlorine and bromine) was monitored by measuring
absorbance of the indicator dye N,N-diethyl-p-phenylenedi-
amine (DPD) (Standard Methods, 4500-CL G). Fig. 2A shows
DPD spectra measured after addition of 10 mg L
1 of either
Oxone (solid lines) or PAA (dashed lines) to distilled water
(Solution 1), an aqueous solution consisting of 35 g L
1 NaCl
(Solution 2), or an aqueous solution consisting of 35 g L
1 NaCl
and 70 mg L
1 KBr (Solution 3). In this set of experiments, the
disinfectant was allowed to react in the solution for 5 min,
whereupon DPD was added, and 1 min later the DPD absor-
bance spectrum was measured. Referring to Fig. 2A, the
absorbance of DPD increased in the order Solution
1 < Solution 2 < Solution 3, consistent with the idea that
Oxone oxidizes both chloride and bromide ions to form
secondary oxidants. DPD absorbance did not increase when
PAA was allowed to react for 5 min in Solution 2 and increased
only slightly when PAA was allowed to react for 5 min in
Solution 3. These latter results are consistent with the findings

Fig. 2 e Panel A: DPD absorbance spectra measured after 10 mg LL1 of either Oxone (solid lines) or PAA (dashed lines) is
allowed to react for 5 min in: Solution 1 (DI water), Solution 2 (DI water and 35 g LL1 NaCl), or Solution 3 (DI water, 35 g LL1
NaCl, and 70 mg LL1 KBr). Panel B: DPD absorbance at 515 nm after 10 mg LL1 of either Oxone (solid lines) or PAA (dashed
lines) are allowed to react for 5 min in Solution 2 with KBr concentrations shown. Panel C: the kinetics of oxidant formation
by 10 mg LL1 of either Oxone (solid line) or PAA (dashed line) in Solution 3.
5646 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
of Booth and Lester (1995), who report that PAA could not
oxidize chloride ion to hypochlorous acid, but could oxidize
bromide ion to hypobromous acid. In the presence of Oxone,
DPD absorbance increased with increasing KBr concentration
in Solution 2 (Fig. 1B), and the oxidation kinetics in Solution 3
are relatively rapidly (<5 min of exposure time, Fig. 1C). In the
presence of PAA, on the other hand, DPD absorbance
increased only slightly with increasing KBr concentration in
Solution 2 (Fig. 2B), and oxidization kinetics in Solution 3 are
relatively slow (>150 min exposure time) (Fig. 2C).
Collectively, the results presented above are consistent with
the idea that Oxone quickly oxidized trace anions in brackish
shallow groundwater to form secondary oxidants that act
synergistically with Oxone to enhance disinfection rates.
However, the reaction of Oxone with trace anions could also
increase the toxicity of the shallow groundwater, by producing
compounds that are carcinogenic (e.g., bromateion) and/or by
producing secondary oxidants (e.g., bromine and chlorine) that
subsequently react with organic compounds to form toxic
disinfection by-products (e.g., trihalomethanes) (Guo and Lin,
2009). For these reasons, and despite its obvious utility as
a disinfectant, Oxone should not be used for in-situ disinfection
of sewage contaminated shallow groundwater in settings, like
Avalon Bay, where the groundwater is under marine influence.
PAA, on the other hand, does not appear to react strongly with
chloride ion, or quickly with bromide ion, and thus is less likely
to produce toxic disinfection by-products; a conclusion sup-
ported by several published studies (Liberti et al., 1999; Dell’Erba
et al., 2007; Kitis, 2004). Although PAA does not appear to
generate the quantity and spectrum of disinfection by-products
associated with many disinfectants, some researchers have
raised concern about the potential environmental impact of
releasing disinfected effluents that contain a PAA residual
(Antonelli et al., 2009; de Lafontaine et al., 2008), which can be
toxic to crustaceans and microorganisms (Antonelli et al., 2009;
de Lafontaine et al., 2008). However, Lafontaine et al. (2008) note
that, because PAA breaks down rapidly in seawater, any
potential impacts would be localized around the region where
disinfection residual is discharged to the environment. Thus, if
PAA is used for in-situ disinfection of sewage contaminated
shallow groundwater, care should be taken to minimize the
PAA residual discharged to coastal waters.
3.3. Activation energy for PAA disinfection
Based on measurement of intrinsic disinfection rates over
a range of temperatures, from 5 to 30  C, the following acti-
vation energies were estimated for PAA disinfection of EC and
ENT: 37.5  7.8 kJ mol
1 and 38.0  9.3 kJ mol
1, respectively.
These activation energies are used later in the paper to esti-
mate intrinsic disinfection rates for PAA over a range of
temperatures relevant to the shallow groundwater in Avalon
Bay (see Section 5.1).
4. Modeling PAA disinfection kinetics
Of the strategies evaluated above, PAA disinfection appears
the most viable, given that it is both a potent biocide and less
likely to form toxic disinfection by-products. In this section we
evaluate several published models for PAA disinfection, all of
which are special cases of Hom’s Law (Hom, 1972):
dN
¼
kd Ntm Cn (3)
dt
where, N, C, kd, and t represent bacterial concentration,
disinfectant concentration, intrinsic disinfection rate
constant, and time, respectively. Chick’s Law (Eqs. (1) And (2)
of this paper) corresponds to a choice of exponent values
m ¼ 0, n ¼ 1, and a constant disinfectant concentration, C ¼ C0.
Wagner et al. (2002) suggested that, during the disinfection
process, the concentration of peroxycompounds in commer-
cial PAA mixtures decays with time in accordance with the
following second-order rate law:
dC
¼
k C2 (4)
dt
The rate constant k* depends on the initial peroxycompound
concentration C0 (Wagner et al., 2002):
k ¼ aC
b
0 (5)
Where the pre-factor and power-law exponents are given by
a ¼ 0.0093 mM(b
1) min
1 and b ¼ 1.420. In their analysis of PAA
disinfection, Santoro et al. (2007) suggest that peroxycompounds
exhibit zero-order decay; however, zero-order decay models
predict negative concentration in finite time, and therefore will
not be considered further here. Combining Eqs. (4) and (5), and
solving the differential equation, yields the following prediction
for disinfectant concentration as a function of time:
1
CðtÞ ¼ (6)
1=C0 þ k t
Given this time-dependence for the disinfectant concentra-
tion, Hom’s Law becomes:
dN
n
¼
kd Ntm ½1=C0 þ k t (7)
dt
Wagner et al. (2002) solved Eq. (7) for two different choices of
the exponents n and m: (1) n ¼ 1 and m ¼ 0 (no-tailing model),
and (2) n ¼ 1 and m ¼
1 (tailing model). The tailing model
provided a better empirical fit to the PAA disinfection data,
although their intrinsic disinfection rate constant kd varied
with the initial disinfectant concentration, and a new fitting
parameter (the time t at which N(t) ¼ N0) was introduced. Here
we opt for the more parsimonious no-tailing model, for which
an exact solution can be derived:

kd =k
N ¼ N0 ðk C0 t þ 1Þ (8)
PAA disinfection data collected in this study were tested
against Chick’s Model (Eq. (1)) and the no-tailing model (Eq. (8))
in Fig. 3. The data are plotted so that the intrinsic disinfection
rate constant kd can be estimated directly from the slope b of
the best-fit line: kd ¼ 2.303b (Chick’s Law) or kd ¼ b (no-tailing
model). Also shown in Fig. 3 are two estimates of model
performance, including the Pearson’s r2 correlation between
logN/N0 and the x-axis (either C0t or log½k C0 t þ 1=k ), and the
root mean square error (RMSE) between modeled and
measured values of bacterial log reduction. In general, for
a given choice of fecal bacteria group (either ENT or EC) the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5647

Fig. 3 e PAA disinfection data for EC (panel A and B) and ENT (panel C and D) fit to either Chick’s model (Eq. (1), panel A and C)
or Wagner et al.’s no-tailing model (Eq. (8), panel B and D). Values for kd are mML1 minL1 and units for RMSE are logN/N0.

two disinfection models have similar r2 and RMSE values
(compare panels A and C with panels B and D in Fig. 3). Both
models are a better description of ENT disinfection (r2 ¼ 0.68 to
0.72, RMSE ¼ 0.32e0.37) than EC disinfection (r2 ¼ 0.52 to 0.55,
RMSE ¼ 0.6e0.62). The fact that ENT exhibited less dispersion
around the no-tailing model may reflect less variability
(compared to EC) in the disinfection resistance of enterococci
bacteria populations present in the sewage and source waters.
Intrinsic rate constants estimated from the slope b are also
similar for the two models. Chick’s Law yields intrinsic rate
constants for EC and ENT of 3.0 and 1.9 mM
1 min
1, respec-
tively (panels A and C). The no-tailing model yields intrinsic
rate constants for EC and ENT of 3.3 and 3.5 mM
1 min
1,
respectively (panels B and D). The intrinsic disinfection rate
for EC is also similar to values estimated by averaging rate
constants obtained from our individual experiments (see
Section 3), and from data reported in other studies of PAA
disinfection (Table 1). In summary, Chick’s Law and the no-
tailing model were both reasonably good predictors of bacte-
rial decay caused by PAA disinfection, and both yield values of
the intrinsic disinfection rate constant kd that were consistent
across models, and across studies. Relative to the modeling
effort described in the next section, the primary benefit of the
no-tailing model is that it accounts for the decay in disinfec-
tant concentration that will inevitably occur following injec-
tion of PAA into the subsurface.
5. Lagrangian model of in situ disinfection
In this section we develop a quantitative model that accounts
for the physical, chemical, and biological factors that might
influence the in-situ disinfection of sewage contaminated
shallow groundwater with PAA. The model is developed in
two stages. First, a Lagrangian framework is used to predict
the concentration of both sewage constituents (fecal indicator
bacteria) and PAA residual in a parcel of shallow groundwater
as it travels from the point of injection to the point where it is
discharged to the coastal ocean. Second, an analytical model
is derived for fluid parcel residence time in shallow
5648 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

groundwater, for the situation (relevant to Avalon Bay) where
shallow groundwater discharge is dominated by tidal pump-
ing and meteoric flow.
5.1. Lagrangian framework for modeling disinfection
kinetics
Fig. 4 illustrates the physical transport processes that might
affect fecal bacteria concentration in a parcel of groundwater
as it moves from the point where it first encounters subsur-
face injected disinfectant (point A) to the point where it exits
the shallow groundwater and is discharged to the coastal
ocean (point B). As the parcel moves from A to B, the
concentration of bacteria in the fluid parcel changes with time
due to disinfection, non-disinfection related die-off, regrowth,
and attachment to the porous matrix (filtration). Mass balance
over a fluid parcel as it travels from A to B yields the following
rate expression for the concentration of bacteria (N, bacteria
L
3), where s [min
1] represents the time a sewage contami-
nated water parcel has been in contact with disinfectant
(referred to here as “residence time”), C is the disinfectant
concentration [mM], and the rate constants for disinfection,
inactivation, filtration, and growth are kd [mM
1 min
1], ki
[min
1], kf [min
1], and mg [min
1]:
dN
1

¼
kd N½1=C0 þ k s
ki þ kf N þ mg N
ds
In formulating Eq. (9), we adopted the no-tailing version of
Hom’s Law described earlier (Eq. (7) with m ¼ 0 and n ¼ 1)
and assume that inactivation, filtration, and regrowth of
bacteria all follow first-order kinetics. While we did not
(e.g., Lazarova et al., 1998; Lefevre et al., 1992). Here we use
the Monod equation (Levenspiel, 1980) to model the
dependence of growth rate mg on dissolved organic carbon
(DOCT):
mg;max DOCT ðsÞ
mg ðsÞ ¼ (10)
Ks þ DOCT ðsÞ
where, mg,max and Ks represent the maximum growth rate and
saturation constant, respectively. Based on the measurements
of DOC presented in Section 3.1, potential sources of DOC
include ambient groundwater (DOCGW), sewage (DOCSewage),
and acetic acid associated with the PAA mixture, including
acetic acid present as an equilibrium component of commer-
cial PAA preparations (DOCAA,0), and acetic acid formed by the
decomposition of PAA during disinfection (DOCA(s)), where the
latter can be estimated from the loss of peroxycompound
concentration with time:
DOCAA ðsÞ ¼ 0:82MðC0
CðsÞÞ (11)
Eq. (11) assumes stoichiometric conversion of PAA to acetic
acid, taking into account the molar fraction (0.82) of the per-
oxycompound concentration that is PAA and the weight of
carbon associated with every mole of acetic acid, M ¼ 25 g of
carbon per mole. Combining Eqs. (10) and (11), we have the
following prediction for the total DOC available for growth of
fecal indicator bacteria:
(9) DOCT ðsÞ ¼ DOCGW þ DOCSewage þ DOCAA;0 þ 0:82MðC0
CðsÞÞ: (12)
Combining Eqs. (9)e(12) and solving the resulting differential
equation, yields the following formula for the concentration of
bacteria in a parcel of shallow groundwater as a function of
residence time (s):

 

kf ðC0 M þ Ks Þ þ C0 Mki
C0 Mmg;max þ DOCT0 kf þ ki
mg;max þ ki Ks 
a
N0 ðDOCT0 þ Ks Þ exp
s
C0 M þ DOCT0 þ Ks
NðsÞ ¼ kd =k a
(13a)
ðC0 k s þ 1Þ ½C0 k sðMC0 þ DOCT0 þ Ks Þ þ DOCT0 þ Ks 

observe regrowth in the disinfection experiments presented

earlier, it is a well-known that the acetic acid in commercial Ks mg M
a¼ (13b)
PAA mixtures can serve as a carbon source for the growth k ðC0 M þ DOCT0 þ Ks Þ2
of heterotrophic bacteria, including fecal indicator bacteria

Fig. 4 e A conceptual model for the in-situ disinfection of sewage-contaminated shallow groundwater in coastal marine
environments.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3 5649

Fig. 5A presents the bacterial log reduction predicted by Eq.
(13) (solid curves) as a function of initial PAA concentration
(vertical axis) and residence time (horizontal axis). This graph
was generated using the parameter values listed in Table 2,
which are based on the experimental measurements pre-
sented earlier and field conditions relevant to the Avalon Bay
field site, including a shallow groundwater temperature of
15  C. For the range of peroxycompound concentrations
considered in Fig. 5A, the dependence of bacteria concentra-
tion on residence time exhibits two patterns. At small resi-
dence times, very little of the PAA has been converted to acetic
acid, disinfection dominates, and bacteria concentration
declines rapidly with increasing residence time. At longer
residence times, PAA has been mostly converted to acetic
acid, regrowth of bacteria dominates, and bacterial concen-
tration increases with increasing residence time. This decay/
regrowth pattern is illustrated for a single initial perox-
ycompound concentration (C0 ¼ 0.03 mM) and temperature
(T ¼ 15  C) in Fig. 5B (dotted line). For this particular choice of
parameter values, the model predicts that bacterial concen-
tration falls approximately 2 log units within 12 h, and
increases thereafter, eventually rising above the initial
bacteria concentration at a residence time of around two days.
As expected, bacteria removal increases monotonically with
increasing initial peroxycompound concentration, as illus-
trated for a fixed residence time (s ¼ 5 days) and temperature
(T ¼ 15  C) in Fig. 5C (dotted line). Using the activation energies
for the intrinsic disinfection rate constant reported in Section
3.3, the model predicts similar trends over the range of
temperatures (13e17  C) typically measured in Avalon shallow
groundwater (compare the family of curves in Fig. 5B and C).
In an ideal scenario, by the time a fluid parcel of shallow
groundwater is discharged to the ocean, both its bacterial
concentration and PAA residual would be very small. The
dependence of PAA residual on initial peroxycompound
concentration and residence time can be estimated from Eq.
(6), by setting the left hand side equal to a fixed residual
concentration, Cres and solving for residence time s:
1

1 
b
sres ¼ C
C
1 C0 (14)
a res 0
Curves of constant peroxycompound residual are plotted in
Fig. 5A (dashed curves). As expected, peroxycompound
residual declines with increasing residence time for a fixed C0.
Interestingly, the curve for Cres ¼ 1 mM roughly demarcates the
transition from bacteria disinfection to regrowth (compare
solid and dotted lines in Fig. 5A).
5.2. Residence time of fluid parcels
The Lagrangian model presented above reveals that both the
bacterial concentration and PAA residual were sensitive to the
residence time of water parcels in shallow groundwater, and
in this section we describe some of the physical processes that
can affect this key parameter. Here, residence time has
precisely the same meaning as in estuarine systems: “how
long a parcel, starting from a specified location within
a waterbody, will remain in the waterbody before exiting”
(Monsen et al., 2002). Provided that material diffusion can be
neglected (a key assumption in the Lagrangian approach
adopted here) (Deleersnijder et al., 2001), the residence time

Fig. 5 e Model predictions for EC concentration and PAA residual for the in-situ disinfection of sewage contaminated shallow
groundwater with PAA. Panel A: Curves of constant EC log reduction predicted by Eq. (13) (solid lines labeled with numbers
ranging from L1 to L9), and curves of constant peroxycompound residual predicted by Eq. (14) (dashed curves, labeled with
numbers ranging from 0.2 to 1 mM). Panel B: Change in EC concentration with residence time predicted by Eq. (13) for
C0 [ 0.03 mM and the ambient groundwater temperatures shown. (C) Change in EC concentration with increasing initial
PAA concentration predicted by Eq. (13) for a fixed residence time of s [ 5 days and the ambient groundwater temperatures
shown. Parameter values used to generate these curves are listed in Table 2.
5650 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3

Table 2 e Inputs parameters used to produce the disinfection contour plot (Fig. 5), residence time model (Fig. 6).
Constant Name Value Source
(b
1)
1
a k* fitting parameter 0.0093 mM min Wagner et al.
b k* fitting parameter 1.42 Wagner et al.
kd EC disinfection rate (T ¼ 13,15,17  C) 3.3, 3.7, 4.1 mM
1 min
1 Measured
ki EC natural decay rate 0.1 h
1 Estimated
kf EC filtration rate 0 min
1 Estimated
mg Max growth rate 0.005 min
1 Surbeck et al.
Ks Growth rate saturation constant 0.07 Surbeck et al.
DOCT0 Initial DOC Concentration mg L
1 Measured
M Mass of carbon per mole PAA 25 g Calculated
A Wave amplitude 1m Estimated
Kp Hydraulic conductivity 5  10
4 m s
1 Calculated via KozenyeCarman
l Wave number 0.05 m
1 Estimated

associated with the movement of a fluid parcel from point A to
B in Fig. 4 can be written explicitly as follows:
ZL
dx
sc ¼
v (15)e(17) can be combined to yield the following estimate for the
0
where, v is the velocity experienced by the fluid particle as it
moves toward the ocean. Here we have subscripted the resi-
dence time calculated from Eq. (15) (sc) to distinguish it from
the actual residence time of a water parcel included in our
disinfection model above (s). Fluid parcel velocity has contri-
butions from tidal pumping (vt), wave set-up (vw), meteoric
groundwater flow (vm), seasonal evapotranspiration (vs), and
density driven flow (vd) (Burnett et al., 2003b; Michael
et al., 2005):
v ¼ vt þ vw þ vm þ vs þ vd
Given that tidal pumping appears to dominate the discharge
of shallow groundwater to Avalon Bay (Boehm et al., 2009b),
here we focus on the tidal pumping and meteoric terms in Eq.
(16), vt and vm. Flow fields generated by tidal pumping can be
approximated from the following expression (Nielsen, 1990):
vt ðx; tÞ ¼ Kp Ale
lx ðcos½ut
lx
sin½ut
lxÞ
with the falling tide) passes point A in Fig. 4, which is mathe-
matically equivalent to assigning the value p to the quantity
(ut
lx) in Eq. (17). After invoking this simplification and
allowing for the possibility of non-zero meteoric flow, Eqs.
(15)
characteristic residence time of a fluid parcel released at a set-
back distance x ¼ L from the beach:
 
log
AlKp
elL vm
log
AlKp
vm
sc ¼ ; vm > 0 (18a)
vm
1  lL
sc ¼ e
1 ; vm ¼ 0 (18b)
Al2 Kp
For the range of parameter values typical of the field site in
Avalon Bay (see Table 2), the residence times predicted by Eq.
(18) vary over one hundred thousand fold, from approximately
(16)
30 mine1000 days (Fig. 6). This variability in residence time
derives, in part, from the non-linear dependence of residence
time on both set-back distance and meteoric flow (Eq. (18a)).
Furthermore, if studies of residence times in estuaries are any
guide, the residence time sc of water parcels in shallow
groundwater is best characterized by a probability distribu-
tion, not a single value. Studies in estuarine systems have
noted that fluid parcel residence times tend to follow proba-
(17) bility distributions characterized by long tails, implying that

Nielsen derived Eq. (17) from the Boussenesq equation after

invoking a number of simplifying assumptions, including
a homogeneous unconfined aquifer of hydraulic conductivity
KP, a vertical beach face, and a one-dimensional flow field
(parallel to the x-axis, see Fig. 4) characterized by a wave number
l, and forced by a single harmonic tide with amplitude A and
angular frequency u. The wave number is defined as l ¼ 2p/x0,
where x0 represents the inland distance over which tidal fluc-
tuations in the shallow groundwater are significant. Because
the flow field predicted by Eq. (17) is tidally periodic and spatially
variable, the residence time of a fluid parcel will depend not only
on where in the aquifer it is released (referred to here as the set-
back distance, x ¼ L, see Fig. 4) but also on when in the tidal cycle
that release occurs; a very similar phenomenon has been
described for the residence time distributions in coastal estu-
aries (Monsen et al., 2002; Oliveira and Baptista, 1997). Despite
these complications, a characteristic residence time can be Fig. 6 e Characteristic shallow groundwater residence
estimated from Eqs. (15)e(17) by releasing the fluid parcel at the times predicted by Eq. (18) for various set-back distances
precise moment when the recessional tide wave (associated and meteoric flow velocities.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
Table 3 e Inputs Parameters used in the uncertainty analysis (Eq. (19)).
Constant Name
k* PAA decay rate
kd EC disinfection rate (T ¼ 15  C)
ki EC natural decay rate
mg Max growth rate
Ks Growth rate saturation constant
DOCT0 Initial DOC concentration
C0 Initial peroxycompounds
s Residence time
a minority of fluid parcels spend a very long time in the
estuary (Oliveira and Baptista, 1997).
6. Uncertainty analysis and practical
implications
One advantage of the analytical in-situ disinfection model
derived earlier (Eq. (13)) is that the uncertainty associated with
different independent variables can be assessed quantita-
tively using the Law of Uncertainty (Taylor and Kuyatt, 1993):
X
p  2
vS
u2 ðSÞ ¼ u2 ðXi Þ
i¼1
vXi
In this equation, u2(S ) and u2(Xi) represents the variance of the
dependent and independent variables, respectively, the
dependent variable is the log-transformed bacteria concen-
tration (S ¼ log(N )), the independent variables Xi include all
variables appearing on the right hand side of Eq. (13) (i.e., C0,
Ks, mg, DOCT0, k*, kd, ki, or s), vS/vXi is the sensitivity of the
dependent variable to change in a particular independent
variable (computed analytically from Eq. (13)), and the
summation is taken over all independent variables (P ¼ 8). The
uncertainty (or variance) associated with each independent
variable was estimated from the relative uncertainty UR(Xi)
and magnitude jXi j values listed in Table 3:
uðXi Þ ¼ UR ðXi ÞjXi j
The form of the Law of Uncertainty adopted here assumes
that independent variables do not co-vary, which is reason-
able for most combinations of independent variables included
in this analysis. When applied to Eq. (13), the Law of Uncer-
tainty reveals that 99% of the variance in the log-transformed
bacteria concentration can be attributed to variance in just
three independent variables: residence time (s) (90%),
maximum growth rate (mg) (8%), and the inactivation rate (ki)
(1%). These results imply that the prediction (and optimiza-
tion) of in-situ disinfection will depend strongly on the resi-
dence time of shallow groundwater which, in turn, depends
non-linearly on the injection well set-back distance (L) and
physical characteristics of the shallow groundwater system
that can vary in time and space (l, A, KP, vm) (see Eq. (18)).
Given the very approximate nature of the analysis that led to
Eq. (18), experimental characterization of shallow ground-
water residence times would be a fruitful topic for further
investigation.
5651
Value Estimated
uncertainty
1.35 L$mmol
1$min
1 0.2
3.7 L$mmol
1$min
1 0.2
0.1 h
1 0.2
0.005 min
1 0.2
0.07 0.2
0.006 g/ L
1 0.2
0.03 mmol 0.2
2 day 1
The disinfection model’s sensitivity to residence time is,
in part, a consequence of the fact that fecal bacteria can
grow in the environment, and thus dramatically different
disinfection outcomes (e.g., from a net reduction in bacteria
concentrations to a net increase in bacteria) can be caused
by slight changes in residence time of water parcels in the
shallow groundwater. Given that most recreational water-
borne illnesses are caused by human viruses that cannot
grow outside their host (Schoen et al., 2011), it is possible
that in-situ disinfection of sewage contaminated shallow
groundwater would reduce shoreline concentrations of
human pathogens (and hence lower recreational waterborne
illness rates), even if it did not substantially reduce fecal
(19)
indicator bacteria concentrations. Indeed, the environ-
mental growth of EC and ENT can lead to a decoupling
between fecal indicator bacteria and human pathogens in
recreational waters (Litton et al., 2010), and potentially
nullify epidemiological relationships upon which current
fecal indicator bacteria criteria are based (Colford et al.,
2007). In light of these and other concerns the U.S Environ-
mental Protection Agency is evaluating and possibly revising
the current water quality criteria for marine recreational
beaches (Boehm et al., 2009a).
7. Conclusions
(20)
 Aeration alone and aeration with brine did not significantly
reduce EC and ENT concentrations in mixtures of raw
sewage and shallow groundwater after 6 h of exposure,
while 4e5 mg L
1 of PAA and 4 mg L
1 Oxone achieved >3
log reduction of EC and ENT after 15 min of exposure.
 Oxone disinfection is enhanced at higher salinities, most
likely due to the formation of secondary oxidants (e.g.,
chlorine and bromine) that make this disinfectant inap-
propriate for marine applications.
 PAA disinfection of fecal bacteria in shallow groundwater in
coastal settings depends non-linearly on residence time,
and the “ideal” disinfection outcome (low bacterial
concentration and low PAA residual) is achieved over
a relatively narrow window of residence times and initial
disinfection concentrations.
 By analogy to surface estuaries, the residence time of water
parcels in shallow groundwater under the influence of
marine tides (i.e., subterranean estuaries) is likely to exhibit
broad (e.g., logenormal) probability distributions with long
5652 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
tails, and depend sensitively on local precipitation, meteoric
flow, and tidal variability.
 Uncertainty calculations suggest 99% of the uncertainty
associated with the log reduction of bacteria is caused by
just three independent variables: residence time (s) (90%),
maximum growth rate (mg) (8%), and the inactivation rate
(ki) (1%).
 Given these results, further research into the residence time
of water in shallow groundwater is needed before an in-situ
disinfection schemes can be successfully designed and
implemented.
Acknowledgments
The authors thank R. Litton, L. Ho, and J. Monroe for assis-
tance with the experiments, and C. Wagner and P. Woolson,
and the City of Avalon staff for the use of City Hall for the
disinfection studies. Funding was provided by the City of
Avalon and State Water Resources Control Board Clean Bea-
ches Initiative, under Agreement 07-582-550.This is publica-
tion 66 of the Urban Water Research Center, University of
California, Irvine.
references
Abarnou, A., Miossec, L., 1992. Chlorinated waters discharged to
the marine-environment chemistry and environmental-
impact e an overview. Science of the Total Environment 126
(1e2), 173e197.
Allonier, A.S., Khalanski, M., Camel, V., Bermond, A., 1999.
Characterization of chlorination by-products in cooling
effluents of coastal nuclear power stations. Marine Pollution
Bulletin 38 (12), 1232e1241.
Anipsitakis, G.P., Dionysiou, D.D., 2003. Degradation of organic
contaminants in water with sulfate radicals generated by the
conjunction of peroxymonosulfate with cobalt.
Environmental Science & Technology 37 (20), 4790e4797.
Antonelli, M., Mezzanotte, V., Panouilleres, M., 2009. Assessment
of peracetic acid disinfected effluents by microbiotests.
Environmental Science & Technology 43 (17), 6579e6584.
Boehm, A.B., Grant, S.B., Kim, J.H., Mowbray, S.L., McGee, C.D.,
Clark, C.D., Foley, D.M., Wellman, D.E., 2002. Decadal and
shorter period variability of surf zone water quality at
Huntington Beach, California. Environmental Science &
Technology 36 (18), 3885e3892.
Boehm, A.B., Fuhrman, J.A., Mrse, R.D., Grant, S.B., 2003. Tiered
approach for identification of a human fecal pollution source
at a recreational beach: case study at Avalon Cay, Catalina
Island, California. Environmental Science & Technology 37 (4),
673e680.
Boehm, A.B., Shellenbarger, G.G., Paytan, A., 2004. Groundwater
discharge: potential association with fecal indicator bacteria
in the surf zone. Environmental Science & Technology 38 (13),
3558e3566.
Boehm, A.B., Ashbolt, N.J., Colford Jr., J.M., Dunbar, L.E., Fleming, L.E.,
Gold, M.A., Hansel, J.A., Hunter, P.R., Ichida, A.M., McGee, C.D.,
2009a. A sea change ahead for recreational water quality criteria.
Journal of Water and Health 7 (1), 9e20.
Boehm, A.B., Yamahara, K.M., Love, D.C., Peterson, B.M.,
McNeill, K., Nelson, K.L., 2009b. Covariation and
photoinactivation of traditional and novel indicator
organisms and human viruses at a sewage-impacted marine
beach. Environmental Science & Technology 43 (21),
8046e8052.
Booth, R., Lester, J., 1995. The potential formation of halogenated
by-products during peracetic acid treatment of final sewage
effluent. Water Research 29 (7), 1793e1801.
Burnett, W.C., Bokuniewicz, H., Huettel, M., Moore, W.S.,
Taniguchi, M., 2003a. Groundwater and pore water inputs to
the coastal zone. Biogeochemistry 66 (1), 3e33.
Burnett, W.C., Bokuniewicz, H., Huettel, M., Moore, W.S.,
Taniguchi, M., 2003b. Groundwater and pore water inputs to
the coastal zone. Biogeochemistry 66 (1e2), 3e33.
Chick, H., 1908. An investigation of the laws of disinfection.
Journal of Hygiene 8 (01), 92e158.
Colford Jr., J.M., Wade, T.J., Schiff, K.C., Wright, C.C., Griffith, J.F.,
Sandhu, S.K., Burns, S., Sobsey, M., Lovelace, G., Weisberg, S.B.,
2007. Water quality indicators and the risk of illness at beaches
with nonpoint sources of fecal contamination. Epidemiology
18 (1), 27.
Cristiani, P., 2005. Solutions to fouling in power station
condensers. Applied Thermal Engineering 25 (16), 2630e2640.
de Lafontaine, Y., Despatie, S.P., Wiley, C., 2008. Effectiveness and
potential toxicological impact of the PERACLEAN (R) ocean
ballast water treatment technology. Ecotoxicology and
Environmental Safety 71 (2), 355e369.
Deleersnijder, E., Campin, J.M., Delhez, E.J.M., 2001. The concept
of age in marine modelling I. Theory and preliminary model
results. Journal of Marine Systems 28 (3e4), 229e267.
Dell’Erba, A., Falsanisi, D., Liberti, L., Notarnicola, M., Santoro, D.,
2004. Disinfecting behaviour of peracetic acid for municipal
wastewater reuse. Desalination 168, 435e442.
Dell’Erba, A., Falsanisi, D., Liberti, L., Notarnicola, M., Santoro, D.,
2007. Disinfection by-products formation during wastewater
disinfection with peracetic acid1. Desalination 215 (1e3),
177e186.
Grant, S.B., Sanders, B.F., 2010. Beach boundary layer:
a framework for addressing recreational water quality
impairment at enclosed beaches. Environmental Science &
Technology, 4818e4824.
Guo, G.M., Lin, F.K., 2009. The bromination kinetics of phenolic
compounds in aqueous solution. Journal of Hazardous
Materials 170 (2e3), 645e651.
Hom, L.W., 1972. Kinetics of chlorine disinfection in an ecosystem.
Journal of the Sanitary Engineering Division 98 (1), 183e194.
Kitis, M., 2004. Disinfection of wastewater with peracetic acid:
a review. Environment International 30 (1), 47e55.
Lazarova, V., Janex, M., Fiksdal, L., Oberg, C., Barcina, I.,
Pommepuy, M., 1998. Advanced wastewater disinfection
technologies: short and long term efficiency. Water Science
and Technology 38 (12), 109e117.
Lefevre, F., Audic, J., Ferrand, F., 1992. Peracetic acid disinfection
of secondary effluents discharged off coastal seawater. Water
Science and Technology 25 (12), 155e164.
Levenspiel, O., 1980. The monod equation e a revisit and
a generalization to product inhibition situations.
Biotechnology and Bioengineering 22 (8), 1671e1687.
Liberti, L., Lopez, A., Notarnicola, M., 1999. Disinfection with
peracetic acid for domestic sewage re use in agriculture.
Water and Environment Journal 13 (4), 262e269.
Lipp, E.K., Farrah, S.A., Rose, J.B., 2001. Assessment and impact of
microbial fecal pollution and human enteric pathogens in
a coastal community. Marine Pollution Bulletin 42 (4),
286e293.
Litton, R.M., Ahn, J.H., Sercu, B., Holden, P.A., Sedlak, D.L., Grant, S.
B., 2010. Evaluation of chemical, molecular, and traditional
markers of fecal contamination in an effluent dominated urban
stream. Environmental Science & Technology.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 4 1 e5 6 5 3
Manheim, F.T., Krantz, D.E., Bratton, J.F., 2004. Studying ground
water under Delmarva coastal bays using electrical resistivity.
Ground Water 42 (7), 1052e1068.
McIlvin, M.R., Altabet, M.A., 2005. Chemical conversion of nitrate
and nitrite to nitrous oxide for nitrogen and oxygen isotopic
analysis in freshwater and seawater. Analytical Chemistry 77
(17), 5589e5595.
Michael, H.A., Mulligan, A.E., Harvey, C.F., 2005. Seasonal
oscillations in water exchange between aquifers and the
coastal ocean. Nature 436 (7054), 1145e1148.
Mika, K.B., Imamura, G., Chang, C., Conway, V., Fernandez, G.,
Griffith, J.F., Kampalath, R.A., Lee, C.M., Lin, C.C., Moreno, R.,
Thompson, S., Whitman, R.L., Jay, J.A., 2009. Pilot- and bench-
scale testing of faecal indicator bacteria survival in marine
beach sand near point sources. Journal of Applied
Microbiology 107 (1), 72e84.
Monarca, S., Feretti, D., Collivignarelli, C., Guzzella, L., Zerbini, I.,
Bertanza, G., Pedrazzani, R., 2000. The influence of different
disinfectants on mutagenicity and toxicity of urban
wastewater. Water Research 34 (17), 4261e4269.
Monsen, N.E., Cloern, J.E., Lucas, L.V., Monismith, S.G., 2002. A
comment on the use of flushing time, residence time, and age
as transport time scales. Limnology and Oceanography 47 (5),
1545e1553.
Moore, W.S., 1999. The subterranean estuary: a reaction zone of
ground water and sea water. Marine Chemistry 65 (1e2), 111e125.
Nielsen, P., 1990. Tidal dynamics of the water-table in beaches.
Water Resources Research 26 (9), 2127e2134.
Oliveira, A., Baptista, A.M., 1997. Diagnostic modeling of
residence times in estuaries. Water Resources Research 33 (8),
1935e1946.
Sanchez-Ruiz, C., MartÌnez-Royano, S., Tejero-MonzÛn, I., 1995.
An evaluation of the efficiency and impact of raw wastewater
disinfection with peracetic acid prior to ocean discharge.
Water Science and Technology 32 (7), 159e166.
Santoro, D., Gehr, R., Bartrand, T.A., Liberti, L., Notarnicola, M.,
Dell’Erba, A., Falsanisi, D., Haas, C.N., 2007. Wastewater
disinfection by peracetic acid: assessment of models for
5653
tracking residual measurements and inactivation. Water
Environment Research 79 (7), 775e787.
Santos, I.R., Niencheski, F., Burnett, W., Peterson, R., Chanton, J.,
Andrade, C.F.F., Milani, I.B., Schmidt, A., Knoeller, K., 2008.
Tracing anthropogenically driven groundwater discharge into
a coastal lagoon from southern Brazil. Journal of Hydrology
353 (3e4), 275e293.
Schoen, M.E., Soller, J.A., Ashbolt, N.J., 2011. Evaluating the
importance of faecal sources in human-impacted waters.
Water Research.
Stampi, S., De Luca, G., Onorato, M., Ambrogiani, E., Zanetti, F.,
2002. Peracetic acid as an alternative wastewater disinfectant
to chlorine dioxide. Journal of Applied Microbiology 93 (5),
725e731.
Swarzenski, P., Burnett, W., Greenwood, W., Herut, B., Peterson, R.,
Dimova, N., Shalem, Y., Yechieli, Y., Weinstein, Y., 2006.
Combined time-series resistivity and geochemical tracer
techniques to examine submarine groundwater discharge at
Dor Beach, Israel. Geophysical Research Letters 33 (24), L24405.
Taniguchi, M., Burnett, W.C., Cable, J.E., Turner, J.V., 2002.
Investigation of submarine groundwater discharge.
Hydrological Processes 16 (11), 2115e2129.
Taylor, B.N., Kuyatt, C.E., Standards N.I.o. and Technology, 1993.
Guidelines for Evaluating and Expressing the Uncertainty of
NIST Measurement Results, Citeseer.
USEPA, 2003. Guidelines Establishing Test Procedures for the
Analysis of Pollutants; Analytical Methods for Biological
Pollutants in Ambient Water, Final Rule US Federal Register,
40 CFR 68, 139, FRL-7529-7.
Wagner, M., Brumelis, D., Gehr, R., 2002. Disinfection of
wastewater by hydrogen peroxide or peracetic acid:
development of procedures for measurement of residual
disinfectant and application to a physicochemically treated
municipal effluent. Water Environment Research 74 (1),
33e50.
Watch F.a.W., 2008. Overloaded and Underfunded: Northwestern
California’s Aging Infrstructure Needs A Clean Water Trust
Fund. Food and Water Watch.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Effect of some parameters on the rate of the catalysed

decomposition of hydrogen peroxide by
iron(III)-nitrilotriacetate in water

Joseph De Laat a,*, Yen Hai Dao a, Nasma Hamdi El Najjar a, Claude Daou b
a
Université de Poitiers, Laboratoire de Chimie et Microbiologie de l’Eau (CNRS UMR 6008), Ecole Nationale Supérieure d’Ingénieurs de
Poitiers, 1, rue Marcel Doré, 86 022 Poitiers Cedex, France
b
Université Saint-Esprit de Kaslik (USEK), Faculté des Sciences, Jounieh, Lebanon

article info abstract

Article history: The decomposition rate of H2O2 by iron(III)-nitrilotriacetate complexes (FeIIINTA) has been
Received 24 May 2011 investigated over a large range of experimental conditions: 3 < pH < 11, [Fe(III)]T,0:
Received in revised form 0.05e1 mM; [NTA]T,0/[Fe(III)]T,0 molar ratios : 1e250; [H2O2]0: 1 mMe4 M) and concentrations


12 August 2011 of HO radical scavengers: 0e53 mM. Spectrophotometric analyses revealed that reactions
Accepted 15 August 2011 of H2O2 with FeIIINTA (1 mM) at neutral pH immediately lead to the formation of inter-
Available online 24 August 2011 mediates (presumably peroxocomplexes of FeIIINTA) which absorb light in the region
350e600 nm where FeIIINTA and H2O2 do not absorb. Kinetic experiments showed that the
Keywords: decomposition rates of H2O2 were first-order with respect to H2O2 and that the apparent
Fenton reaction first-order rate constants were found to be proportional to the total concentration of
Aminopolycarboxylate ligands FeIIINTA complexes, were at a maximum at pH 7.95  0.10 and depend on the [NTA]T,0/
Kinetics [Fe(III)]T,0 and [H2O2]0/[Fe(III)]T,0 molar ratios. The addition of increasing concentrations of
Effect of pH tert-butanol or sodium bicarbonate significantly decreased the decomposition rate of H2O2,


Tert-butanol suggesting the involvement of HO radicals in the decomposition of H2O2. The decompo-
III
Bicarbonate ion sition of H2O2 by Fe NTA at neutral pH was accompanied by a production of dioxygen and
by the oxidation of NTA. The degradation of the organic ligand during the course of the
reaction led to a progressive decomplexation of FeIIINTA followed by a subsequent
precipitation of iron(III) oxyhydroxides and by a significant decrease in the catalytic activity
of Fe(III) species for the decomposition of H2O2.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction a biological treatment can be used to achieve a partial oxida-

Biological processes are widely used to treat industrial
wastewater because they are very efficient and much more
cost-effective than physical and chemical processes.
However, the effectiveness of biological treatments can be
limited when wastewaters contain non-biodegradable or toxic
substances. In these cases, a chemical oxidation step prior to
tion of toxic and/or bio-recalcitrant organic pollutants into
biodegradable by-products (Pera-Titus et al., 2004). Chemical
oxidation can be performed by using advanced oxidation
processes (AOPs) which involve the generation of highly
reactive hydroxyl radicals. Among the AOPs, the Fenton and
Fenton-like reagents (Fe(II)/H2O2 or Fe(III)/H2O2, pH  3) have
been implemented on a number of industrial wastewater

* Corresponding author. Tel.: þ33 5 49 45 39 21; fax: þ33 5 49 45 37 68.

E-mail address: joseph.de.laat@univ-poitiers.fr (J. De Laat).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.028
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4 5655

Table 1 e Simplified reaction model of the FeIINTA/H2O2 and of the FeIIINTA/H2O2 systems.
Reactions* Rate or equilibrium constant

Acid-base equilibrium reactions

1 H2O2 $ HO 2 þ H
þ
pKa ¼ 10.7 (Buxton et al., 1998)
3-
NTA ¼ L ¼ N(CH2COO-)3 pKa1 ¼ 1.79 (Sanchiz et al., 1999)
2 H3L $ H2L þ Hþ pKa2 ¼ 2.31 (Sanchiz et al., 1999)
H2L $ HL2- þ Hþ pKa3 ¼ 9.37 (Sanchiz et al., 1999)
HL2- $ L3- þ Hþ
HO $ O  þ Hþ pKa ¼ 11.9 (Buxton et al., 1998)



3
HO2 $ O2 þ Hþ pKa ¼ 4.8 (Bielski et al., 1985)



4

Reactions involving FeIINTA or FeIIINTA complexes

5 FeðIIÞ þ NTA %FeII NTA log K for Fe2þ þ L3 % FeL :
log K ¼ 7.47 (Demmink and Beenackers, 1997)
log K ¼ 8.8 (Kurimura et al., 1968)
log K ¼ 10.6 (MINEQLþ database)
6 FeðIIIÞ þ NTA %FeIII NTA log K for Fe3þ þ L3 % FeL:
log K ¼ 15.09 (Sanchiz et al., 1999)
log K ¼ 18.6 (MINEQLþ database)
log K ¼ 15.90 (Kurimura et al., 1968)
log K ¼ 15.90 (Motekaitis and Martell, 1994)
FeIINTA þ H2O2 / FeIIINTA þ HO þ HO k ¼ 9.7 103 M1 s1 (Gilbert and Jeff, 1988)


7
k ¼ 1.84 104 M1 s1 (Borggaard et al., 1971)
8 FeIIINTA þ H2O2 $ {FeIIINTAH2O2}
{FeIIINTAH2O2} / FeIINTA þ HO2/O2



9
FeIINTA þ HO2/O2 / FeIIINTA þ H2O2



10
FeIIINTA þ HO2/O2 / FeIINTA þ O2



11
HO þ FeIINTA / FeIIINTA þ HO 5.0 109 M1 s1 at pH ¼ 6.2 (Lati and Meyerstein, 1978)


12
2.3 109 M1 s1 (Cabelli and Bielski, 1990)
HO þ FeIIINTA / Products k ¼ 1.6 108 M1 s1 at pH z2 (Sharma and Sahul, 1982)


13

Other reactions
HO þ H2O2 / HO2 þ H2O 2.7 107 M1 s1 (Buxton et al., 1998)



14a
HO þ HO 
2 / O2 þ H2O 7.5 109 M1 s1 (Buxton et al., 1998)



14b
HO þ NTA / Products k ¼ 6.1 107 M1 s1 at pH ¼ 2.0; k ¼ 5.5 108 M1 s1


15
at pH ¼ 6.0; k ¼ 4.2 109 M1 s1 at pH ¼ 10 (Sahul and Sharma, 1987)
k ¼ 7.5 108 M1 s1 at pH ¼ 4.0; k ¼ 2.5 109 M1 s1
at pH ¼ 9.0 (Lati and Meyerstein, 1978)
k ¼ 2.1 109 M1 s1 at pH w0 (Borggaard, 1972)
HO þ tert-Butanol / Products þ H2O 6.0 108 M1 s1 (Buxton et al., 1998)


16
HO þ HCO 
3 / CO3 þ H2O 8.6 106 M1 s1 (Buxton et al., 1998)



17a
 
HO þ CO2- / 3 þ HO 3.9 108 M1 s1 (Buxton et al., 1998)



17b 3 CO
CO3 þ H2O2 / HO2 þ HCO

8 105M1 s1 (Behar et al., 1970)



18 3
4.3 105M1 s1 (Draganic et al., 1991)

*For simplicity, the terms FeIINTA and FeIIINTA will be used hereafter to represent all the forms of iron(II) and iron(III)-nitrilotriacetate
complexes, unless specifically stated otherwise. Several reactions are not chemically balanced.

treatment plants (Bautista et al., 2008). The main benefits of
the Fenton and Fenton-like reactions are the use of environ-
mentally friendly and low cost reagents. However, these
processes have also some limitations. They must be operated
at low pH (pH z 3) in order to prevent the precipitation of
ferric oxyhydroxides and in the absence of high concentra-
tions of chloride or sulfate ions because of the formation of
inactive chloro or sulfato-iron(III)-complexes (De Laat and Le,
2005, 2006).
To overcome these drawbacks, the addition of organic and
inorganic iron-chelating agents has been used to increase the
solubility of iron species at neutral pH and therefore to enhance
the efficiency of the homogenous Fenton-like oxidation
processes (Sun and Pignatello, 1992; 1993; Li et al., 2007; Lee and
Sedlak, 2009; Rastogi et al., 2009), or of the heterogenous
systems involving ion bearing minerals (Xue et al., 2009) or
granular zero-valent iron (Keenan and Sedlak, 2008; Lee et al.,
2008). The efficiency of 50 iron(III)-chelates for the decomposi-
tion of H2O2 and of 2,4-dichlorophenoxyacetic at pH 6 have been
examined by Sun and Pignatello (1992) and nitrilotriacetic acid
(NTA) was found to be one of the most active chelates. The iron-
NTA/H2O2 process was also found to be effective for the
degradation of tetrachloroethene in contaminated soils
(Howsawkeng et al., 2001; Ndjou’ou et al., 2006). Kim and Kong
(2001) showed that the FeIIINTA/H2O2 system degraded more
efficiently 1-hexanol and carbon tetrachloride at pH 9 than at
pH 3 and suggested that the degradations of 1-hexanol and


carbon tetrachloride are initiated by hydroxyl radical (HO ) and
by superoxide anion radical (HO2/O2), respectively.





The nature of the reactive oxidant species (HO radicals or/
and high-valent-oxoiron species) generated by the reaction of
H2O2 with free and complexed Fe(II) and Fe(III) species
remains a controversial issue (Walling et al., 1975; Yamazaki
and Piette, 1991; Bossmann et al., 1998; Bamnolker et al.,
5656 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
1991; Pignatello et al., 2006). It is generally accepted that the
reaction of H2O2 with free Fe2þ at low pH yields hydroxyl
radicals whereas other oxidants such as the ferryl ion may
be produced at neutral pH (Gallard et al., 1998; Rivas et al.,
2001; Hug and Leupin, 2003; Keenan and Sedlak, 2008;
Katsoyiannis et al., 2008). In the case of systems involving
FeIINTA and FeIIINTA complexes, a recent competitive kinetic
study showed that the degradation of three probe compounds
(atrazine, fenuron, parachlorobenzoic acid) by FeIINTA/H2O2,
FeIIINTA/H2O2 and FeIINTA/O2 at neutral pH could be attrib-


uted to HO radicals (Dao and De Laat, 2011). This was also
confirmed by a decrease of the rates of degradation of the


probe compounds in the presence of HO scavengers (tert-
butanol and bicarbonate ion). These data do not exclude the
formation of other oxidant species concurrently with the


formation of HO radicals during the decomposition of H2O2 by
FeIINTA and FeIIINTA complexes.
Previous works on the FeIINTA/H2O2 and FeIIINTA/H2O2
systems showed that the rates of decomposition of H2O2 and
of degradation of organic solutes were at a maximum at pH z
8 (Tachiev et al., 2000; Kim and Kong, 2001; Dao and De Laat,
2011) and decreased in the presence of increasing concentra-
tions of hydroxyl radical scavengers such as tert-butanol and
bicarbonate ions (Dao and De Laat, 2011).
On the basis of the generally accepted mechanisms of the
Fenton and Fenton-like oxidation processes, Table 1 reports
the main reactions which can be proposed for the catalytic
decomposition of hydrogen peroxide by iron(II)-NTA and
iron(III)-NTA complexes. As shown in Table 1, most of the rate
constants of elementary reactions which may be involved in
the FeIINTA/H2O2 and FeIIINTA/H2O2 systems are unknown or


not well known such as rate constants for the reactions of HO
with the various FeIINTA and FeIIINTA complexes (reactions 12
and 13 in Table 1) and with the various acid-base forms of NTA
(reactions 4 and 15 in Table 1). In addition, it has been
demonstrated that the formation of iron(III)-peroxocomplexes
represents the first step of the decomposition of H2O2 by free
Fe(III) (Gallard et al., 1999) and by iron(III) complexes of eth-
ylenediamine tetraacetate (EDTA). In the case of the
FeIIIEDTA/H2O2 system, the reaction of H2O2 with FeIIIEDTA
leads to the formation of purple peroxocomplexes and several
studies have been conducted in order to determine equilib-
rium and rate constants for the formation of FeIIIEDTA per-
oxocomplexes (Walling et al., 1970; Francis et al., 1985). The
formation of similar peroxocomplexes with FeIIINTA
complexes has never been reported in literature. Table 1 also
indicates that NTA would be consumed during the course of
the reaction (reaction 15) and that this degradation would lead
to a progressive decomplexation of iron(III) at neutral pH and
therefore to a precipitation of ferric oxyhydroxides.
Because of the large uncertainties on the rate constants of
the chemical reaction model presented in Table 1, the rates of
decomposition of H2O2 by FeIINTA and FeIIINTA complexes
can not be predicted by computer simulations. In addition, the
effects of various parameters on the decomposition rates of
H2O2 are not well documented. Therefore, the main objective
of the present work was to investigate the decomposition
rates of H2O2 by FeIINTA and FeIIINTA complexes over a wide
range of experimental conditions. Experiments were con-
ducted at 25  C and the following parameters have been
investigated: pH, initial concentrations of Fe(III) and H2O2,
[NTA]T,0/[FeIII]T,0 and [H2O2]0/[Fe(III)]T,0 molar ratios, concen-
trations of tert-butanol and bicarbonate ions.
2. Materials and methods
2.1. Reagents and preparation of solutions
All chemicals were reagent grade and were used without
additional purification. All aqueous solutions were prepared
using 18 MU Milli-Q water from a Millipore system. Glassware
was washed with perchloric acid (pH z 2) and rinsed before
use.
Stock solutions of Fe(II) and of Fe(III) were prepared by
dissolving the required amounts of iron(II) perchlorate
(Fe(ClO4)2, 98%) and of iron(III) perchlorate (Fe(ClO4)3, 9 H2O) in
0.1 M HClO4. Stock solutions of NTA (from 5 to 20 mM) were
prepared from nitrilotriacetic acid.
Solutions of FeIINTA were prepared by mixing the required
volumes of stock solutions of Fe(II) and of NTA in deoxygen-
ated water to prevent spontaneous oxidation of Fe(II). The
solutions were kept under dinitrogen gas atmosphere and the
pH adjusted to the desired value with NaOH 1 M or 0.1 M.
Solutions of FeIIINTA ([NTA]0/[Fe(III)]0 > 1 mol mol1) were
prepared by mixing appropriate volumes of stock solutions of
Fe(III) (10 mM) and of NTA. Under these conditions, Fe(III) was
present only as FeIIINTA complexes at the beginning of the
reaction (Motekaitis and Martell, 1994). The pH was then
adjusted with NaOH 1 M or 0.1 M. All the FeIINTA and FeIIINTA
solutions were freshly prepared each time before use and the
concentrations of iron determined just before addition of
H2O2.
2.2. Reaction conditions
All kinetic experiments were carried out using a batch reactor.
The batch reactor consisted of a 1.3-L cylindrical double-wall
jacketed reactor to circulate thermostated water with an
external circulating pump connected to a thermostated water
bath (25.0  0.5  C) (Section S2.2 in Supplementary material).
The reactor was covered by a black plastic film to protect the
aqueous solution from ambient light. The reactor was filled
with 1-L of solution containing FeIINTA or FeIIINTA. The
solution was mixed using a magnetic stirrer at nearly 800 rpm
during all the course of the reaction. Initial pH was adjusted to
the desired value using HClO4 or NaOH. During the course of
the reaction, the pH was kept constant using a pH transmitter
(OPM 223, Endress þ Hauser) and a peristaltic pump (Gilson
Minipuls 3) for the injection of NaOH 1 M (liquid flow rate :
10e15 mL/h). The reaction was initiated by adding H2O2 into
the reactor. 1, 2 or 5 mL aliquots were withdrawn at selected
time intervals and immediately analyzed for H2O2.
2.3. Analytical methods
The concentration of H2O2 in stock solutions of H2O2 was
determined by iodometric titration. Concentrations of H2O2 in
solutions containing Fe(II) or Fe(III) were determined spec-
trophotometrically using the TiCl4 method (Eisenberg, 1943)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
and a molar absorption coefficient of 724 M1 cm1 for the
titanium peroxocomplex. The concentrations of Fe(II) or of
Fe(III) (after reduction of Fe(III) by hydroxylamine hydrochlo-
ride) were measured by the o-phenanthroline colorimetric
method and by using a molar extinction coefficient of
1.105 104 M1 cm1 at 510 nm for the Fe(II)ephenanthroline
complex (Tamura et al., 1974).
Absorption spectra were measured with a SAFAS DES 190
double beam spectrophotometer and pH measurements were
made with a Meter Lab PHM 240 pH meter calibrated with
standard buffers.
COD (chemical oxygen demand) was analyzed following the
closed reflux - titrimetric method with potassium dichromate
as oxidant using a TR320 thermoreactor (Merck, Germany). TOC
and TN (total organic carbon and total nitrogen) were measured
by using a carbon analyzer (Shimadzue TOC 5000).
3. Results
3.1. Decomposition rate of H2O2 by FeIINTA at pH 7
Ferrous species play a key role on the overall rate of decom-
position of H2O2 and of degradation of organic pollutants by
the Fenton and Fenton-like reactions because the reaction of


H2O2 with Fe(II) represents the unique source of HO radicals.
Limited data exist on the rate of decomposition of H2O2 by
FeIINTA (reaction 7 in Table 1). A few papers indicate that
FeIINTA complexes are readily oxidized by H2O2 (Borggaard
et al., 1971; Gilbert and Jeff, 1988; Rush and Koppenol, 1988)
at circumneutral pH.
To illustrate the fast decomposition rate of H2O2 by
FeIINTA, Fig. 1 presents typical decomposition profiles of H2O2
obtained after introduction of 200, 500, 700 or 930 mM of H2O2
into an oxygen-free aqueous solution of FeIINTA ([FeII]T,0 ¼
200 mM; [NTA]T,0 ¼ 500 mM) at pH 7.0  0.1). The data showed
a very fast decomposition of H2O2 during the first 20 s of the
reaction. This initial stage was then followed by a slow
decomposition of H2O2 which corresponds to the catalyzed
1000
First stage : very fast decomposition
800
Second stage : slow decomposition
600
400
200
0 the first step of the decomposition mechanism of H2O2 by
0 200 400 600
Time (s)
Fig. 1 e Decomposition of H2O2 by FeIINTA
([Fe(II)]T,0 [ 200 mM, [NTA]T,0 [ 0.5 mM, [H2O2]0 [ 200, 500,
700 or 930 mM, pH [ 7.0 ± 0.1, [O2]0 < 0.1 mg/L).
5657
decomposition of H2O2 by the FeIIINTA complexes formed
from the oxidation of the initial FeIINTA complexes by H2O2.
For this second stage of the reaction, the decomposition of
H2O2 by FeIIINTA was first-order with respect to H2O2
concentration, with an apparent first-order rate constant
equal to (1.83  0.05) 103 s1 (Section S3.1.1 in Supporting
Material).
The completion of the initial stage of the reaction within
a reaction time of less than 20 s indicates that the second-
order rate constants for the reaction of H2O2 with FeIINTA
complexes are greater than 103 M1 s1, in agreement with
literature (Borggaard et al., 1971; Gilbert and Jeff, 1988). The
rate constants for the reaction of H2O2 with FeIINTA could not
be determined under the experimental conditions used in the
present study.
At the end of the initial fast phase of the reaction, the
consumption of H2O2 was nearly 1.0e1.1 mol of H2O2/mol of
FeIINTA (for [H2O2]0 ¼ 500, 700 or 930 mM). By assuming that the
reaction of H2O2 with FeIINTA leads to the formation of HO


radicals (reaction 7 in Table 1), the consumption of 1.0e1.1 mol
of H2O2/mol of FeIINTA obtained in the presence of an excess


of H2O2 suggests that the HO radicals formed by the initial
attack of H2O2 with FeIINTA do not contribute to the oxidation
of FeIINTA but are trapped by NTA.
To examine the effect of the dose of H2O2 and of the
[NTA]T,0/[Fe(II)]T,0 ratio on the overall stoichiometry between
H2O2 and FeIINTA, an other set of experiments was performed
by using FeIINTA in excess ([Fe(II)]T,0 ¼ 1.0 mM;
0.1 < [H2O2]0 < 0.7 mM, pH ¼ 7.0  0.1, [O2] < 0.1 mg O2/L) and
with [NTA]0/[Fe(II)]T,0 ratios ranging from 0 to 10 mol/mol (0,
0.5, 1.1, 4 and 10 mol/mol) (Section S3.1.2 in Supporting
Material). In the absence of NTA (FeII/H2O2 system), the overall
stoichiometry was nearly equal to 1.85  0.05 mol Fe2þ/mol
H2O2, in agreement with the expected value of 2 mol of H2O2/
mole of Fe(II) (reactions 7 and 12 in Table 1; Gallard et al., 1998).
In the presence of NTA, the measured consumptions of H2O2
were equal to 1.5 mol H2O2/mole Fe(II) for [NTA]T,0/
[Fe(II)]T,0 ¼ 0.5 mol/mol (z50% of Fe(II) present as FeIINTA),
1.3 mol Fe(II)/mol H2O2 for [NTA]T,0/[Fe(II)]T,0 ¼ 1.1 mol/mol
and 1.0 mol Fe(II)/mol H2O2 for [NTA]0/[Fe(II)]T,0 ¼ 4 and


10 mol/mol. These data demonstrate that NTA acts as an HO
radical scavenger when NTA is in large excess.
3.2. Decomposition rate of H2O2 by FeIIINTA
3.2.1. Spectrophotometric study
In our previous spectrophotometric studies on the Fenton-like
reaction (FeIII/H2O2, pH < 3), we showed that the addition of
H2O2 into a solution of Fe(III) immediately produces iron(III)-
peroxocomplexes which absorb in the region 350e600 nm
(Gallard et al., 1999) and that their unimolecular decomposi-
tion represents the rate limiting step in the regeneration of the
ferrous ion (De Laat and Gallard, 1999). It is also generally
accepted that the formation of peroxocomplexes represents
800
iron(III)-aminopolycarboxylate complexes (Walling et al.,
1970; Francis et al., 1985; Tachiev et al., 2000). Contrary to
the reaction of H2O2 with FeIIIEDTA complexes which leads to
the formation of purple intermediates (Walling et al., 1970), no
significant change of color was observed upon addition of
5658
H2O2 to FeIIINTA solutions. However, spectrophotometric
analyses conducted in the present work demonstrate for the
first time that the reaction of H2O2 (0.5e4 M) with FeIIINTA
([Fe(III)]T,0 ¼ 0.89 mM; [NTA]0/([Fe(III)]T,0 ¼ 1.2 mol/mol) leads
to the formation of species which absorb UV/visible light and
in particular in the region 350e600 nm where FeIIINTA
complexes and H2O2 do not absorb (Fig. 2). Analyses also
showed that the rise in the absorption occurred immediately
after the addition of H2O2 and was pH dependant. As it has
been previously demonstrated for uncomplexed iron(III)
species at pH < 3 (Gallard et al., 1999), the change in the UV/
visible absorbance upon the addition of H2O2 can be attributed
to the formation of complexes between H2O2 and FeIIINTA.
Unfortunately, the complexation constants for the formation
of peroxocomplexes of FeIIINTA could not be determined in
the present work, presumably because the solutions contain
several iron(III) species.
3.2.2. Effects of the total concentration of Fe(III) on the
decomposition rate of H2O2
Fig. 3a presents typical data obtained for the decomposition
of H2O2 by FeIIINTA in the presence of increasing concentra-
tions of Fe(III) (from 0.05 to 0.5 mM) at pH 8.0  0.1 and for
[NTA]T,0/[Fe(III)]T,0 ¼ 5 mol/mol. Similar data were also
obtained at pH 7.0 and 9.0 (Fig. S3.5 in Supplementary Mate-
rial). Under the conditions used, the rate of decomposition of
H2O2 was first-order with respect to the concentration of H2O2:
d½H2 O2 
 ¼ kapp ½H2 O2 
dt
where kapp is the pseudo-first-order rate constant for the
overall depletion of H2O2. The values of kapp have been
reported in Table S2 and in Fig. 3b. Fig. 3b shows that the
measured first-order rate constants (kapp) were found to
increase linearly with the total concentration of FeIII. These
3 decomposition of the various FeIIINTA peroxocomplexes in
III
[Fe ]T,0 = 0.83 mM, [NTA]T,0 = 1.0 mM
2.5
Absorbance (1-cm cell)
pH = 7.0
[H2O2]0 :
2 2.43 M
1.5
1
0.5
0
300 350 400
Wavelength (nm)
Fig. 2 e Absorption spectra of solutions of FeIIINTA in the
absence of H2O2 and in the presence of H2O2 recorded
immediately after introduction of H2O2 in the
spectrophotometric cell ([Fe(III)]T,0 [ 0.83 mM,
[NTA]T,0 [ 1.0 mM, [H2O2]0 [ 0, 0.42, 0.83, 2.43 and 4.0 M,
pH0 [ 7.0 ± 0.1, path length of the cell [ 10 mm).
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
data suggest that the rate of decomposition of H2O2 can be
described by a second-order kinetic expression:
d½H2 O2 
 ¼ kd ½FeðIIIÞT;0 ½H2 O2  (2)
dt
where kd is the second-order rate constant for the overall rate
of decomposition of H2O2 by FeIIINTA complexes. Under the
conditions used in the present work ([H2O2]0 ¼ 5 mM;
[Fe(III)]T,0 < 0.5 mM; [NTA]T,0/[Fe(III)]T,0 ¼ 5 mol/mol), the
mean values of kd show that the reaction was faster at pH 8.0
(kd ¼ 27.1  1.6 M1 s1) than at pH 7.0 (kd ¼ 16.1  0.7 M1 s1)
or pH 9.0 (kd ¼ 21.4  1.7 M1 s1).
3.2.3. Effect of pH
To confirm the effect of pH on the rate of decomposition of
H2O2 by FeIIINTA, a series of kinetic experiments has been
carried out with pH values ranging from 2.7 to 10.5 and with
initial concentrations of H2O2, FeIII and NTA equal to 1 mM,
0.2 mM and 1 mM, respectively. For all the pH studied, the
decay of H2O2 was first-order with respect to H2O2. The plots of
the apparent first-order rate constants (kapp) (Section S3.2.3 in
Supplementary Material) and of the corresponding apparent
second-order rate constants (kd ¼ kapp/[Fe(III)]T,0) as a function
of pH demonstrate that the rate of decomposition of H2O2 was
maximal at pH ¼ 7.95  0.10 (Fig. 3c). At this pH and for the
conditions used in this work ([Fe(III)]T,0 ¼ 0.05e0.5 mM;
[NTA]T,0/([Fe(III)]T,0 ¼ 5 mol/mol), the second-order rate
constant for the reaction of H2O2 with FeIIINTA is equal to
(1) kd ¼ 28  2 M1 s1.
The existence of an optimal pH for the decomposition of
H2O2 by FeIIINTA complexes is in agreement with the data
obtained by Francis et al. (1985), Tachiev et al. (2000) and Kim
and Kong. (2001). As previously discussed by Tachiev et al.
(2000), the effects of pH on the initial rates of decomposi-
tion of H2O2 have been attributed to the stability constants
for the formation of the various peroxocomplexes from the
different acid-base forms of FeIIINTA complexes (four
FeIIINTA complexes, Fig. 3d) and to the unimolecular rates of
water. The increase in the rate of decomposition of H2O2
when the pH increased from to 2.7 to 8.0 may also be
attributed to the dissociation of H2O2 (pKa ¼ 11.7) and the
3.99 M
decrease in the reaction rates at pH > 8 to changes in the
speciation of Fe(III).
0.83 M
0.41 M
3.2.4. Effects of [NTA]T,0/[Fe(III)]T,0 molar ratio
0.00 M
Fig. 4 presents the effects of the [NTA]T,0/[Fe(III)]T,0 molar ratios
(2.5 < [NTA]T,0/[FeIII]T,0 < 100 mol/mol) on the apparent first-
order rate constants of decomposition of H2O2 obtained at
pH ¼ 7.00  0.05 and 8.00  0.05. All the kinetic constants were
determined for decomposition yields of H2O2 which did not
450 500 exceed 50%. The data showed that the effects depend on the
initial concentrations of reactants and on pH. For initial
concentrations of Fe(III) and H2O2 equal to 0.2 mM and 5 mM,
respectively, the rate of decomposition of H2O2 was not
significantly affected at pH 7.0 by the [NTA]T,0/[Fe(III)]T,0 ratio
for ratios less than 12.5 mol/mol (mean values of kapp z 2.86
103 s1) and decreased when the [NTA]T,0/[Fe(III)]T,0 molar
ratio increased from 12.5 mol/mol (kapp z 2.86 103 s1) to
100 mol/mol (kapp z 1.4 103 s1). For an initial concentration of
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4 5659

Fig. 3 e Decomposition of H2O2 by FeIIINTA a) First-order plots for the decomposition of H2O2 at pH [ 8.0 and with various
initial concentrations of FeIIINTA; b) Plots of the first-order rate constant (kapp) as a function of the concentration of FeIIINTA
at pH 7.0, 8.0 and 9.0; c) Change of the first-order rate constant with pH and d) Distribution of FeIIINTA complexes as
a function of pH ([Fe(III)]T,0 [ 0.05e0.5 mM; [NTA]T,0/[Fe(III)]T,0 [ 5.0 mol/mol, [H2O2]0 [ 5.0 mM, 25  C).

H2O2 of 40 mM, the pseudo-first-order rate constants increased
from (2.7  0.1) 103 s1 to (3.2  0.1) 103 s1 when the [NTA]T,0/
[Fe(III)]T,0 molar ratio increased from 2.5 to 10 mol/mol. An
increase was also observed for the series of experiments con-
ducted at pH 8.0 ([Fe(III)]T,0 ¼ 0.05 mM, [H2O2]0 ¼ 3 mM) (Fig. 4).
Several assumptions can be made to explain the observed
effects of [NTA]T,0/[Fe(III)]T,0 molar ratio on the rate of
decomposition of H2O2. Increasing [NTA]T,0/[Fe(III)]T,0 molar
ratios would maintain Fe(III) in the form of FeIIINTA complexes
during the course of the reaction, especially when high [H2O2]0/
[Fe(III)]T,0 molar ratios are used. Increasing [NTA]T,0/
[Fe(III)]T,0 molar ratios also affect the distribution of FeIIINTA
complexes by increasing the proportion of the FeIII(NTA)2
complex. Therefore, the change in the distribution of FeIIINTA
complexes may affect the reaction rates. In addition, at
[NTA]T,0/[Fe(III)]T,0 higher than 2 mol/mol, NTA is mainly


present as free NTA in solution and may act as an HO radical
scavenger.
3.2.5. Effects of [H2O2]0/[Fe(III)]T,0 molar ratio
Fig. 5 presents the effects of the initial concentration of H2O2
(0.7e50 mM) on the rate of decomposition of H2O2. Rate
constants were determined from 36 experiments carried out
at pH 7.0 and with initial concentrations of Fe(III) and NTA
equal to 0.2 mM and 0.5 mM, respectively. Under our condi-
tions, Fe(III) is only present as FeIIINTA complexes at the
beginning of the reaction and the [H2O2]0/[FeIIINTA]T,0 molar
ratios vary from 3.5 to 250 mol/mol. For all experiments, the
initial rates of decomposition of H2O2 ([H2O2]/[H2O2]0 < 0.5)
could be accurately described by a first-order kinetic law with
respect to H2O2 (Section S3.2.5 in Supporting Material). The
change of the first-order rate constant (kapp) with the
concentration H2O2 shows an increase of kapp (from 1.2
103 s1 to (3.2  0.1) 103 s1 when the concentration of H2O2
increased from 0.7 to 10 mM (Fig. 5a). These data can be
explained by the fact that an increase of the concentration of
H2O2 would increase the concentration of the peroxo FeIIINTA
complexes (reaction 7 in Table 1) and therefore the decom-
position rate of H2O2. For initial concentrations of H2O2 higher
than 25e30 mM, a small decrease in the values of kapp could be
observed because kapp decreased from (3.1  0.1) 103 s1 to
(2.6  0.1) 103 s1 ([H2O2]0 ¼ 50 mM). As will be seen below
(Section 3.3), the decrease of the kapp values observed at high
[H2O2]0/[FeIIINTA]T,0 can probably be attributed to a degrada-
tion of the FeIIINTA complexes during the first 240 s of the
reaction. It should also be noted that the initial rates of
decomposition of H2O2 increased from 8 107 M s1 to 1.3
104 M s1 when the initial concentration of H2O2 increased
from 0.7 mM to 50 mM (Fig. 5b).
3.2.6. Effects of hydroxyl radical scavengers
As depicted in Fig. 6, the decomposition rate of H2O2 decreased
in the presence of increasing concentrations of hydroxyl
radical scavengers such as tert-butanol and bicarbonate ion at
pH 8.3. These data are in agreement with those obtained in
a previous work dealing with the effects of tert-butanol and
5660 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4

a 0.005
kapp (s )

III
-1

[H2O2]0 = 5 mM, [Fe ]T,0 = 0.2 mM; pH = 7.0

0.003

0.001
0.005
III
kapp (s )

[H2O2]0 = 40 mM; [Fe ]T,0 = 0.2 mM; pH = 7.0

-1

0.003

0.001
0.002
kapp (s )
-1

0.001
III
[H2O2]0 = 5 mM, [Fe ]T,0 = 0.05 mM; pH = 8.0

0.000
0 20 40 60 80 100
III
[NTA]T,0 / [Fe ]T,0 (mol/mol)

b 100 III
[Fe ]T,0 = 0.2 mM, pH 7.0
80 III
Fe (OH)NTA
III
60 Fe (NTA)2
%

40
20
III
Fe (OH)2 NTA
0 ([Fe(III)]T,0 [ 0.2 mM, [NTA]T,0 [ 0.5 mM, pH [ 7.00 ± 0.05,
0 20 40 60 80
[NTA]T,0 / [Fe(III)]T,0 (mol/mol)
Fig. 4 e Effects of the [NTA]T,0/[Fe(III)]T,0 molar ratio a) on
the pseudo-first-order rate constant of decomposition of
H2O2 at pH 7.0 and 8.0 and b) on the distribution of the
FeIIINTA.


bicarbonate ion on the degradation of HO probes (Dao and De
Laat, 2011).
In the case of H2O2, the overall rate of decomposition of
H2O2 by FeIIINTA complexes depends on the rate of the initi-
ation step of decomposition of H2O2 by FeIIINTA (reactions
8e10 in Table 1) and of the rates of all the secondary reactions
that consume H2O2 or that can affect the steady-state
concentration of FeIINTA. As for the FeIII/H2O2 system, the
steady-state concentration of ferrous species in the FeIIINTA/
H2O2 system would play a key role in the overall rate of
decomposition of H2O2. FeIINTA complexes can be formed by
the decomposition of peroxocomplexes of FeIIINTA (reaction 9
in Table 1) and by reduction of FeIIINTA complexes by HO2/O2



radicals (reaction 11 in Table 1). The latter can be generated
from the decomposition of peroxocomplexes of FeIIINTA
Fig. 5 e Effects of the [H2O2]0/[Fe(III)]T,0 molar ratio a) on the
pseudo-first-order rate constant of decomposition of H2O2
and b) on the initial rate of decomposition of H2O2
25  C).
100


(reaction 9 in Table 1), the oxidation of H2O2 by HO radicals
(reaction 14a and 14b in Table 1) and from the decomposition


of peroxyl radicals formed from the oxidation of NTA by HO
radicals.
The decrease in the decomposition rate of H2O2 in the
presence of tert-butanol can be explained by the fact that
increasing concentrations of tert-butanol would decrease the


steady-state concentrations of HO radicals and therefore


would decrease the rate of decomposition of H2O2 by HO
radicals (reactions 14a and 14b in Table 1) and the rate of
formation of FeIINTA complexes by reduction of FeIIINTA by
HO2/O2 radicals (reaction 11 in Table 1).



The inhibiting effect of tert-butanol and bicarbonate ion for
the oxidation of probe compound would depend on the rela-


tive rates of consumption of HO radicals by all the species


present in solution. These relative rates of consumption of HO
radicals are proportional to the term ki Ci where ki represents


the second-order rate constant for the reaction of HO with
a solute i (Table 1) and Ci, the concentration of the solute i.
Under the conditions used for the experiments depicted on
Fig. 6 ([NTA]0 ¼ 0.5 mM, [H2O2]0 ¼ 5 mM), the scavenging terms
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
Fig. 6 e Pseudo-first-order plots of H2O2 decomposition as
a function of the concentration of a) bicarbonate ion and
b) tert-butanol. c) Comparison of the pseudo-first-order rate
constants of decomposition of H2O2 as a function of the
value of the term ki Ci ([Fe(III)]T,0 [ 0.2 mM,
[NTA]T,0 [ 0.5 mM, [H2O2]0 [ 5 mM, pH [ 8.30 ± 0.05,
25  C).
for NTA (ki z 109 M1 s1) and H2O2 (ki z 3 107 M1 s1 at
neutral pH) are roughly equal to 5 105 s1 and 1.5 105 s1,
respectively. The kinetic data obtained with tert-butanol
showed a nearly complete inhibition of the decomposition
rate of H2O2 at concentrations of tert-butanol higher than
5 mM (kapp z 5.7 103 s1 for [tert-butanol] ¼ 0 mM; kapp z 3.5
104 s1 for 10 mM < [tert-butanol] < 50 mM, Fig. 6a). These
5661


data are consistent with the fact that almost all the HO radi-
cals would be scavenged by tert-butanol at concentrations of
tert-butanol higher than 5 mM (ki Ci > 3 106 s1 for [tert-
butanol] > 5 mM).
In the case of bicarbonate ion, the data in Fig. 6b demon-
strate that the inhibiting effect of bicarbonate ion is less
important than the one to tert-butanol. This finding is not very


surprising because the reaction of HO radicals with bicar-
bonate and carbonate ions leads to the formation of carbonate
radicals (reactions 17a and 17b in Table 1) which are reactive
toward H2O2 (reaction 18 in Table 1) and probably also toward
NTA.
3.2.7. Effects of chloride, sulfate and phosphate ions
Other kinetic experiments showed that the rate of decompo-
sition of H2O2 by FeIIINTA at pH 8.3 ([Fe(III)]0 ¼ 0.2 mM,
[NTA]0 ¼ 0.5 mM, [H2O2]0 ¼ 5 mM) was not affected by chloride
ions (100 mM) and by sulfate ions (33 mM) whereas the pres-
ence of phosphate ions was found to decrease the decompo-
sition rate of H2O2 (kapp z 5.7 103 s1, 4.97 103 and 2.7 103
for [PO3-
4 ]T ¼ 0, 5 and 30 mM). The kinetic data obtained in the
presence of phosphate ions might be explained by the
formation of less reactive phosphate iron(III)-complexes but
spectrophotometric analyses did not reveal a change of the
UV/Visible spectra of FeIIINTA solutions ([H2O2] ¼ 0 mM) in the
presence of phosphate ions (0e30 mM) as well as in the
presence of chloride or sulfate ions.
3.2.8. Effect of dissolved oxygen
The concentration of dissolved oxygen increased during the
decomposition of H2O2 by FeIIINTA complexes. Under our
experimental conditions ([Fe(III)]T,0 ¼ 0.2 mM, [NTA]T,0 ¼
0.5 mM, [H2O2]0 ¼ 5 mM, pH ¼ 7 or 8.3), the production of
dissolved oxygen at the beginning of the reaction was nearly
equal to 0.25 mol O2/mol of H2O2 decomposed at pH 7 and 8.3
in the absence of tert-butanol and was roughly equal to
0.12 mol O2/mole of H2O2 decomposed in the presence of tert-
butanol ([tert-butanol] > 0.5 mM, pH 8.3). These values repre-
sent net productions of dissolved oxygen because dissolved
oxygen can be produced through various reactions (i.e. reac-
tion 11 in Table 1) and consumed by organic radicals to
produce organic peroxyl radicals.
3.3. Degradation of the FeIIINTA complexes


It is known that NTA degradation by HO radicals leads to the
formation of iminodiacetic acid (IDA), glycine, oxalic acid,
ammonia and carbon dioxide (Chen et al., 1995). A decrease in
the efficiency of the FeIIINTA/H2O2 process is therefore
expected during the course of the reaction because the
formation of weaker chelating by-products would lead to
a progressive decomplexation and precipitation of Fe(III) after
the complete depletion of free NTA in solution. In addition,
complexes of Fe(III) with IDA are less active than FeIIINTA
complexes for the decomposition of H2O2 (Fig S3.18). HPLC
analyses for the determination of NTA and its by-products
could not be done during this work. To illustrate the fate of
NTA and FeIIINTA complexes, a series of flasks containing
Fe(III) (1 mM) and NTA (3 mM) were treated with H2O2 doses
ranging from 0 to 50 mM at pH 7. After a reaction time of 24 h
5662 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4

1.0 to 15 mM. Increasing doses of H2O2 led to a progressive

decrease in the concentrations of COD and of TOC which
demonstrate an oxidation of NTA. For the highest dose tested
0.8
(50 mM), COD and TOC removals reached 50% and 40%,
respectively. TN loss was less than 5% indicating that the
0.6 degradation of free and complexed NTA by the FeIIINTA/H2O2
[C] / [C]0

system did not lead to significant amounts of N2 or of N2O. The

0.4
COD
TOC
0.2
TN
Absorbance at 260 nm
Total dissolved iron
0.0
0 10 20
H2O2 consumed (mM)
Fig. 7 e Degradation of the ligand during the
decomposition of H2O2 by FeIIINTA: Normalized values for
the UV absorbance at 260 nm and for the concentrations of
COD ([COD]0 [ 432 mg O2/L), TOC ([TOC]0 [ 216 mg C/L), TN
([TN]0 [ 42 mg N/L) and of total dissolved concentration of
iron ([Fe(III)]T,0 [ 1.0 mM) as a function of the consumption
of H2O2 (Batch experiments, 24-h reaction time,
[Fe(III)]T,0 [ 1 mM, [NTA]T,0 [ 3 mM, [H2O2]0 [ 0e50 mM,
H2O2 removal after 24-h reaction time : >97%, Initial
pH [ 7.0, Final pH > 5.5e6.8).
in the dark, the solutions were filtered through 0.45 mm filters
to eliminate the precipitate of iron(III) oxyhydroxides. The
filtered samples were then analyzed by UV/Visible absorption
spectroscopy and the residual concentrations of H2O2, dis-
solved iron, COD, TOC and TN were determined. The initial
concentrations of COD, TOC and TN were equal to 432 mg
O2/L, 216 mg C/L and 42 mg N/L, in agreement with the theo-
retical values.
After a reaction time of 24 h, analyses showed that H2O2
was completely depleted in samples treated with H2O2 doses
less than 10 mM and that the H2O2 removals were nearly equal
to 97% for all samples treated with H2O2 doses higher or equal
decrease in the concentrations of DOC and TOC was accom-
panied by a decrease in the concentration of dissolved iron in
solution which is consistent with the formation of ferric
oxyhydroxide precipitates and with the decrease in the UV
absorbance of the solutions at the wavelength of 260 nm.
Under the conditions used in the present work ([Fe(III)]T,0 ¼
30 40 50
1 mM, [NTA]T,0 ¼ 3 mM, initial pH ¼ 7.0), the decrease of the UV
absorbance and of the concentration of dissolved iron
appeared at H2O2 doses higher than 5 mM. The decreases in
the UV absorbance at 260 nm and in the total concentration of
Fe(III) were nearly equal to 10, 30 and 70% for H2O2 doses of 10,
20 and 30 mM, respectively (Fig. 7).
To illustrate the effects of the degradation of the ligand on
the activity of the FeIIINTA for the decomposition of H2O2,
Fig. 8 presents decomposition curves of H2O2 obtained after
three consecutive additions of H2O2 (introduction of 10 mM of
H2O2 at reaction times 0, 30 min and 2 h). As expected, the rate
of decomposition of H2O2 and the total concentration of dis-
solved iron slowed down after each addition of H2O2 (Fig. 8).
The initial rates of decomposition of H2O2 were equal to 92, 44
and 2 mM s1 after the first, the second and the third addition
of H2O2, respectively. These data demonstrate that the
FeIIINTA/H2O2 is not a true catalytic system because of the
progressive degradation of the ligand during the course of the
reaction. To avoid the decrease in the catalytic activity of
FeIIINTA, NTA must be continuously added to the reactor at
a feed rate that maintains FeIII in the form of soluble FeIIINTA
complexes with a minimum concentration of uncomplexed
NTA in solution.
4. Conclusions

1
The data obtained in the present work showed that the cata-
III
lyzed decomposition of H2O2 by FeIIINTA complexes repre-
0.8
H2 O2 (Fe /H2 O2 , pH 7 sents the rate limiting step for the decomposition of H2O2 by
without NTA) FeIINTA complexes. Spectrophotometric studies reveal that
addition of H2O2 to FeIIINTA solutions immediately leads to
[C]t / [C] 0

0.6
H2O2 the formation of transient species (probably peroxocomplexes
nd rd
of FeIIINTA) which absorb light in the region 350e600 nm. All
0.4 2 injection 3 injection the kinetic data show that the initial rates of decomposition of
H2O2 by FeIIINTA followed a pseudo-first-order kinetic law
Total dissolved iron
0.2 with respect to H2O2 concentration. Apparent first-order
kinetic rate constants were found to be at a maximum at pH
z 8 and to depend on the initial concentration of Fe(III) and on
0
the molar ratios of reactants ([NTA]T,0/[Fe(III)]T,0 and [H2O2]0/
0 3000 6000 9000 12000 15000 18000
[Fe(III)]T,0). These data suggest that the kinetic modeling of the
Time (s)
rate of decomposition of H2O2 by FeIIINTA over a wide range of
Fig. 8 e Changes of the catalytic activity of FeIIINTA for the experimental conditions should be rather complicated and
decomposition of H2O2 and of the total concentration of that additional research is required prior the validation of
dissolved iron measured after 3 consecutive injections of a reaction kinetic model in order to determine rate constants
10 mM H2O2 at reaction times 0, 30 and 120 min or equilibrium constants of many elementary reactions
([Fe(III)]T,0 [ 1 mM, [NTA]T,0 [ 3 mM, pH [ 7.0 ± 0.1). involved in the FeIINTA/H2O2 or FeIIINTA/H2O2 systems.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
In addition, the data confirmed that the FeIIINTA/H2O2
system is not a true catalytic system because the progressive
degradation of the organic ligand during the course of the
reaction leads to the formation of much less active precipi-
tates of oxyhydroxides at neutral pH. These data also suggest
that the use of the FeIIINTA/H2O2 system for the oxidation of
pollutants in industrial wastewater must be operated at pH z
8 and in continuous flow reactors with a continuous intro-
duction of NTA to avoid precipitation of Fe(III) and of H2O2. For
each application, the injection doses of reactants (NTA and
H2O2) must be optimised in order to minimise the scavenging
effects of an excess of NTA and H2O2 toward hydroxyl radicals
and to minimise the consumption of reactants.
Appendix. Supplementary material
Supplementary data related to this article can be found online
at doi:10.1016/j.quascirev.2011.08.009.
references
Bamnolker, H., Cohen, H., Meyerstein, D., 1991. Reactions of low
valent transition-metal complexes with hydrogen peroxide.
Are they “Fenton-like” or not? 3. the case of Fe(II)
[N(CH2CO2)3](H2O)2. Free Radical Research Communications 15
(4), 231e241.
Bautista, P., Mohedano, A.F., Casas, J.A., Zazo, J.A., Rodriguez, J.J.,
2008. An overview of the application of Fenton oxidation to
industrial wastewaters treatment. Journal of Chemical
Technology and Biotechnology 83, 1323e1338.
Behar, D., Czapski, G., Duchovny, I., 1970. Carbonate radical in
flash photolysis and pulse radiolysis of aqueous carbonate
solutions. Journal of Physical and Chemical 74, 2206e2210.
Bielski, B.H.J., Cabelli, D.E., Arudi, R.L., Ross, A.B., 1985. Reactivity
of HO2/O2e radicals in aqueous solution. Journal of Physical
and Chemical Reference Data 14 (4), 1041e1100.
Borggaard, O.K., 1972. Polarographic determination of diffusion
coefficients of hydrogen peroxide and iron chelates and rate
constants of hydroxyl radical reactions. Acta Chemica
Scandinavica 26 (8), 3393e3394.
Borggaard, O.K., Farver, O., Andersen, V., 1971. Polarographic
study of the rate of oxidation of iron(II) chelates by hydrogen
peroxide. Acta Chemica Scandinavica 25 (9), 3541e3543.
Bossmann, S.H., Oliveros, E., Göb, S., Siegwart, S., Dahlen, E.P.,
Payawan, A.L., Straub, M., Wörner, M.B., Braun, A., 1998. A new
evidence against hydroxyl radicals as reactive intermediates
in the thermal and photochemically enhanced Fenton
reactions. Journal of Physical and Chemical A 102, 5542e5550.
Buxton, G.U., Greenstock, C.L., Helman, W.P., Ross, A.B., 1988.
Critical review of rate constants for reactions of hydrated
electrons, hydrogen atoms and hydroxyl radicals (OH$/O$) in
aqueous solution. Journal of Physical and Chemical Reference
Data 17 (2), 513e886.
Cabelli, D.E., Bielski, B.H.J., 1990. In: Packer, L., Glazer, A.N. (Eds.),
The Use of Polyaminocarboxylates (EDTA, DTPA, etc.) as Metal
Chelators in methods in Enzymology, Oxygen Radicals in
Biological systems. Academic Press, San Diego, CA, pp. 116e120.
Chen, D., Martell, A.E., McManus, D., 1995. Studies on the
mechanism of chelate degradation in iron-based, liquid redox
H2S removal process. Canadian Journal of Chemistry 73,
264e274.
5663
Dao, Y.H., De Laat, J., 2011. Hydroxyl radical involvement in the
decomposition of hydrogen peroxide by ferrous and ferric-
nitrilotriacetate complexes at neutral pH. Water Research 45,
3309e3317.
De Laat, J., Gallard, H., 1999. Catalytic decomposition of hydrogen
peroxide by Fe(II) in homogenous aqueous solution:
mechanism and kinetic modelling. Environmental Science
Technololgy 33, 2726e2732.
De Laat, J., Le, T.G., 2005. Kinetics and modelling of the Fenton-like
reaction in the presence of sulfate in acidic aqueous solutions.
Environmental Science and Technololgy 39, 1811e1818.
De Laat, J., Le, T.G., 2006. Effects of chloride ions on the iron(III)-
catalyzed decomposition of hydrogen peroxide and the
efficiency of the Fenton-like oxidation process. Applied
Catalysis B: Environmental 66, 137e146.
Demmink, J.F., Beenackers, A.A.C.M., 1997. Oxidation of ferrous
nitrilotriacetic acid with oxygen: a model for oxygen mass
transfer parallel to reaction kinetics. Industrial and
Engineering Chemistry Research 36, 1989e2005.
Draganic, Z.D., Negron-Mendoza, A., Sehested, K., Vujosevic, S.I.,
Navarro-Gonzales, R., Albarran-Sanchez, M.G., Draganic, I.G.,
1991. Radiolysis of aqueous solutions of ammonium
bicarbonate over a large dose range. Radiation Physics and
Chemistry 38, 317e321.
Eisenberg, G.M., 1943. Colorimetric determination of hydrogen
peroxide. Industrial and Engineering Chemistry 15, 327e328.
Francis, K.C., Cummins, D., Oakes, J., 1985. Kinetic and structural
investigation s of [FeIII(edta)] e [edta ¼ ethylenediamine-tetra-
acetate(4-)] catalysed decomposition of hydrogen peroxide.
Journal of the Chemistry Society Dalton 3, 493e501.
Gallard, H., De Laat, J., Legube, B., 1998. Influence du pH sur la
vitesse d’oxydation de composés organiques par FeII/H2O2.
Mécanismes réactionnels et modélisation. New Journal of
Chemistry, 263e268.
Gallard, H., De Laat, J., Legube, B., 1999. Spectrophotometric study
of the formation of iron(III)-hydroperoxy-complexes in
homogeneous aqueous solutions. Water Research 33,
2929e2936.
Gilbert, B.C., Jeff, M., 1988. In: Rice- Evans, C., Dormandy, T. (Eds.),
Free Radicals: Chemistry, Pathology and Medicine, pp. 25e49.
Howsawkeng, J., Watts, R.J., Washington, D.L., Teel, A.L., Hess, T.
F., Crawford, R.L., 2001. Evidence for simultaneous abiotic-
biotic oxidations in a microbial-Fenton’s system.
Environmental Science and Technology 35, 2961e2966.
Hug, S.J., Leupin, O., 2003. Iron-catalyzed oxidation of arsenic(III)
by oxygen and by hydrogen peroxide: pH-dependent
formation of oxidants in the Fenton reaction. Environmental
Science and Technology 37, 2734e2742.
Katsoyiannis, I.A., Ruettimann, T., Hug, S.J., 2008. pH dependence
of Fenton reagent generation and As(III) oxidation and
removal by corrosion of zero valent iron in aerated water.
Environmental Science and Technology 42, 7424e7430.
Keenan, C.R., Sedlak, D.L., 2008. Ligand-enhanced reactive
oxidant generation by nanoparticulate zero-valent iron and
oxygen. Environmental Science and Technology 42,
6936e6941.
Kim, Y.-S., Kong, S.-H., 2001. Treatment of 1-hexanol and carbon
tetrachloride in modified Fenton’s reagent using iron(III)-NTA
complex. Journal of Korean Industry and Engineering
Chemistry 12, 787e792.
Kurimura, Y., Ochiai, R., Matsuura, N., 1968. Oxygen oxidation of
ferrous ions induced by chelation. Bulletin of the Chemical
Society of Japan 41, 2234e2239.
Lati, J., Meyerstein, D., 1978. Oxidation of first-row bivalent
transition-metal complexes containing ethylenediaminetetra-
acetate and nitrilotriacetate ligands by free radicals:
a pulse-radiolysis study. Journal of the Chemical Society,
1105e1118.
5664 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 5 4 e5 6 6 4
Lee, C., Keenan, C.R., Sedlak, D.L., 2008. Polyoxometalate
oxidation of organic compounds by nanoparticulate zero-
valent iron and ferrous ion in the presence of oxygen.
Environmental Science and Technology 42, 4921e4926.
Lee, C., Sedlak, D.L., 2009. A novel homogeneous Fenton-like
system with Fe(III)-phosphotungstate for oxidation of organic
compounds at neutral pH values. Journal Molecular Catalysis,
A: Chemical 311, 1e6.
Li, Y.C., Bachas, L.G., Bhattacharyya, D., 2007. Selected chloro-
organic detoxifications by polychelate (poly(acrylic
acid)) and citrate-based Fenton reaction at neutral pH
environment. Industrial and Engineering Chemistry Research
46, 7984e7992.
Motekaitis, R.J., Martell, A.E., 1994. The iron(III) and iron(II)
complexes of nitrilotriacetic acid. Journal of Coordination.
Chemistry 31, 67e78.
Ndjou’ou, A.-C., Bou-Nasr, J., Cassidy, D., 2006. Effect of Fenton
reaction dose on coexisting chemical and microbial oxidation
in soil. Environ. Science and Technology 40, 2778e2783.
Pera-Titus, M., Garcıa-Molina, V., Baños, M.A., Giménez, J.,
Esplugas, S., 2004. Degradation of chlorophenols by means of
advanced oxidation processes: a general review. Applied
Catalysis B: Environmental 47, 219e256.
Pignatello, J., Oliveros, E., MacKay, A., 2006. Advanced oxidation
process for organic contaminant destruction based on the
Fenton reaction and related chemistry. Critical Reviews in
Environmental Science and Technology 36, 1e84.
Rastogi, A., Al-Abed, S.R., Dionysiou, D.D., 2009. Effect of
inorganic, synthetic and naturally chelating agents of Fe(II)
mediated advanced oxidation of chlorophenols. Water
Research 43, 684e694.
Rivas, F.J., Beltran, F.J., Frades, J., Buxeda, P., 2001. Oxidation of
p-hydroxybenzoic acid by Fenton’s reagent. Water Research
35, 387e396.
Rush, J.D., Koppenol, W.H., 1988. Reactions of FeIInta and FeIIedda
with hydrogen peroxide. Journal of American Chemical
Society 110 (15), 4957e4963.
Sahul, K., Sharma, B.K., 1987. Gamma radiolysis of nitrilotriacetic
acid (NTA) in aqueous solutions. Journal of Radioanalytical
and Nuclear Chemistry 109, 321e327.
Sanchiz, J., Esparza, P., Dominguez, S., Brito, F., Mederosa, S.,
1999. Solution studies of complexes of iron(III) with
iminodiacetic, alkyl-substituted iminodiacetic and
nitrilotriacetic acids by potentiometry and cyclic
voltammetry. Inorganica Chimica Acta 291, 158e165.
Sharma, B.K., Sahul, K., 1982. Co-60 gamma radiolysis of iron(III)-
nitrilotriacetate in aqueous solutions. Radiation Physics and
Chemistry 20, 341e346.
Sun, Y., Pignatello, J.J., 1992. Chemical treatment of pesticide
wastes. Evaluation of Fe(III) chelates for catalytic hydrogen
peroxide oxidation of 2,4-D at circumneutral pH. Journal of
Agricultural Food Chemistry 40, 322e327.
Sun, Y., Pignatello, J.J., 1993. Activation of hydrogen peroxide by
iron(III) chelates for abiotic degradation of herbicides and
insecticides in water. Journal of Agricultural Food Chemistry
41, 308e312.
Tachiev, G., Roth, J.A., Bowers, A.R., 2000. Kinetics of hydrogen
peroxide decomposition with complexed and “free” iron
catalysts. International Journal of Chemical Kinetics 32,
24e35.
Tamura, H., Goto, K., Yotsuyanagi, T., Nagayama, M., 1974.
Spectrophotometric determination of iron(II) with 1.10-
phenanthroline in the presence of large amounts of iron(III).
Talanta 21, 314e318.
Walling, C., Kurz, M., Schugar, H.J., 1970. The iron(III)-
ethylenediaminetetraacetic acid-peroxide system. Inorganic
Chemistry 9, 4931e4937.
Walling, C., Partch, R.E., Weil, T., 1975. Kinetics of the
decomposition of hydrogen peroxide catalyzed by ferric
ethylenediaminetetraacetate complex. Proceedings of the
National Academy Sciences of the United States of America
72, 140e142.
Xue, X., Hanna, K., Despas, C., Wu, F., Deng, N., 2009. Effect of
chelating agent on the oxidation of PCP in the magnetite/H2O2
system at neutral pH. Journal of Molecular Catalysis A:
Chemical 311, 29e35.
Yamazaki, I., Piette, L., 1991. EPR spin-trapping study on the
oxidizing species formed in the reaction of the ferrous ion
with hydrogen peroxide. Journal of American Chemical
Society 113, 7588e7593.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Operational aspects of the desulfurization process of energy

gases mimics in biotrickling filters5

Marc Fortuny a,1, Xavier Gamisans b, Marc A. Deshusses c, Javier Lafuente a,

Carles Casas a, David Gabriel a,*
a
Department of Chemical Engineering, Universitat Autònoma de Barcelona, Edifici Q, Campus de Bellaterra, 08193 Bellaterra, Spain
b
Department of Mining Engineering and Natural Resources, Universitat Politècnica de Catalunya, Manresa, Spain
c
Department of Civil and Environmental Engineering, Duke University, Durham, NC, USA

article info abstract

Article history: Biological removal of reduced sulfur compounds in energy-rich gases is an increasingly
Received 29 April 2011 adopted alternative to conventional physicochemical processes, because of economical
Received in revised form and environmental benefits. A lab-scale biotrickling filter reactor for the treatment of high-
13 August 2011 H2S-loaded gases was developed and previously proven to effectively treat H2S concen-
Accepted 16 August 2011 trations up to 12,000 ppmv at gas contact times between 167 and 180 s. In the present work,
Available online 24 August 2011 a detailed study on selected operational aspects affecting this system was carried out with
the objective to optimize performance. The start-up phase was studied at an inlet H2S
Keywords: concentration of 1000 ppmv (loading of 28 g H2S m
3 h
1) and inoculation with sludge from
Desulfurization a municipal wastewater treatment plant. After reactor startup, the inlet H2S concentration
Hydrogen sulfide was doubled and the influence of different key process parameters was tested. Results
Biotrickling filter showed that there was a significant reduction of the removal efficiency at gas contact times
Startup below 120 s. Also, mass transfer was found to be the main factor limiting H2S elimination,
Process performance whereas performance was not influenced by the bacterial colonization of the packed
column after the initial startup. The effect of gas supply shutdowns for up to 5 days was
shown to be irrelevant on process performance if the trickling liquid recirculation was kept
on. Also, the trickling liquid velocity was investigated and found to influence sulfate
production through a better use of the supplied dissolved oxygen. Finally, short-term pH
changes revealed that the system was quite insensitive to a pH drop, but was markedly
affected by a pH increase, affecting both the biological activity and the removal of H2S.
Altogether, the results presented and discussed herein provide new insight and operational
data on H2S removal from energy gases in biotrickling filters.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction compounds, such as reduced sulfur compounds (RSC) (Ross

et al., 1996; Tchobanoglous et al., 2003). Among those RSC,
Energy rich off-gases such as biogas are sometimes not used hydrogen sulfide (H2S) is one of the most commonly reported
for electric power generation due to the presence of corrosive impurities. H2S concentrations in biogas can range from 0.1 to

5
We dedicate this article to the memory of Carles Casas Alvero, a valued colleague, proficient researcher and skilled educator who
passed away on July 20, 2010.
* Corresponding author. Tel.: þ34 935811587; fax: þ34 935812013.
E-mail address: david.gabriel@uab.cat (D. Gabriel).
1
Present address: Aeris Tecnologies Ambientals S.L., Edifici Eureka, Parc de Recerca de la UAB, 08193 Barcelona, Spain.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.029
5666 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
extremely high values of 2% v/v (1000e20,000 ppmv), whereas
the specifications for the maximum content of H2S in typical
biogas-burning engines are in the range of 0.02e0.05% v/v
(200e500 ppmv).
At present, biogas energy recovery is becoming more and
more interesting due to increasing environmental and
economical constraints associated to fossil fuels. Further-
more, an increasing number of solid and liquid wastes
management facilities (biomethanation plants) are being
installed with biogas production as the main economical
benefit.
So far, biological sulfide removal has usually been applied
to odor control (Yang and Allen, 1994; Devinny et al., 1999;
Gabriel and Deshusses, 2003; González-Sánchez et al., 2008).
However, the growing interest in biological treatment alter-
natives has lead to an increasing number of studies in the
recent years where these techniques are applied to the treat-
ment of highly-loaded off-gases (Buisman et al., 1989; Bailón,
2005; van den Bosch et al., 2007; Fortuny et al., 2008).
After absorption, treatment and biodegradation of H2S in
bioreactors occur according to the following overall reactions
(Kuenen, 1975):
H2 S4HS
þ Hþ ðnon
biologicalÞ
HS
þ 0:5O2 /S0 þ OH
ðbiologicalÞ

þ
HS þ 2O2 /SO2

4 þH ðbiologicalÞ
Also, depending on the redox conditions, further oxidation
to sulfate can take place if sulfide is limited but elemental
sulfur (S0) is present (Kuenen, 1975):
þ packing was HD-QPAC (Lantec Products Inc., Agoura Hills,
S0 þ 3=2O2 þ H2 O/SO2

4 þ 2H ðbiologicalÞ
All four reactions above result in pH changes, as do parallel
abiotic reactions such as oxidation of sulfide to thiosulfate (Eq.
(5)), which in turn can also be biologically oxidized to sulfate
(Eq. (6)).
þ
2H2 S þ 2O2 /S2 O2

3 þ H2 O þ 2H ðnon
biologicalÞ
þ needed. Aerobic conditions in the liquid phase were ensured
S2 O2
2

3 þ 2O2 þ H2 O/2SO4 þ 2H ðbiologicalÞ
Also, abiotic polysulfide formation and oxidation under
alkaline conditions as described by van den Bosch et al. (2007)
and González-Sánchez et al. (2008) may occur. These highlight
the complex relationships between oxygen availability, pH
and sulfide oxidation processes; a better understanding of the
effects of these parameters is required for improved system
design.
In a preliminary study (Fortuny et al., 2008), the technical
feasibility of using a single lab-scale biotrickling filter for the
treatment of off-gases containing high concentrations of H2S
was demonstrated. Preliminary results on the system
robustness when exposed to short-term perturbations, and
the relationship between the choice of packing type and the
reactor long-term performance were discussed. Also,
the inoculation procedure and start-up phase were studied.
Furthermore, Montebello et al. (2010) studied the
reactor performance when exposed to increased inlet H2S
concentrations when validating an integrated analyzer for on-
line process monitoring consisting of a Flow Injection
Analyzer coupled to a Continuous Flow Analyzer with
a previous Gas-Diffusion step (FIA/GD-CFA). The results
showed that a slow drop in H2S removal was caused by
progressive sulfide accumulation in the liquid phase. That
system was operated at high H2S loadings (162 g H2S m
3 h
1)
and the performance was limited by the biological oxidation
of H2S.
As a follow up, this study was directed toward optimizing
the start-up phase using different inoculation methods and an
improved reactor design. A second objective was to acquire
a deeper knowledge of the influence of key process parame-
ters when operating the system at less extreme conditions
(lower H2S loading rates or inlet concentrations) than those
previously tested (Fortuny et al., 2008). Thus, short-term
experiments targeting the gas empty bed residence time
(EBRT), the trickling liquid velocity (TLV), operating pH and
H2S supply shutdowns were carried out and the response of
the bioreactor was determined. Also, the influence of the H2S
loading rate and operating conditions changes on the pH and
biological activity were investigated.
(1)
2. Materials and methods
(2)
2.1. Experimental setup
(3)
A lab-scale prototype reactor described in details elsewhere
(Fortuny et al., 2010) was used for this study. In short, the
biotrickling filter had an inner diameter of 7.1 cm, a packed
bed height of 50 cm and a total liquid volume of 2 L. The
(4)
CA, USA) with a 4  4 mm (0.1600  0.1600 ) grid opening cut to
tightly fit inside the reactor. Except for the startup period (see
Section 2.3), the bioreactor was continuously operated at an
inlet H2S concentration of 2000 ppmv (corresponding to
a loading of 56 g H2S m
3 h
1), an EBRT of 180 s, an average
liquid hydraulic retention time (HRT) of 51  6 h and a TLV of
(5) 3.8 m h
1 (liquid flow of 255 ml min
1). A pH range of 6e6.5
was maintained by automated addition of NaOH 1 M as
(6)
by continuous air addition at an O2/H2S supplied ratio of 23.6
(v v
1) through a diffuser located in an oxygenation
compartment installed in the recycle line (Fortuny et al., 2010).
The conditions were only altered during the start-up phase
(see Section 2.3) and for short-term exposures to higher
loading rates (LR) up to 400 g H2S m
3 h
1 and for biomass
sampling episodes.
Metered amounts of H2S, N2 and air using digital mass flow
controllers (Bronkhorst, The Netherlands) were used to
simulate a controlled biogas inflow. Mineral medium (MM) as
described elsewhere (Fortuny et al., 2010) and a solution of
NaHCO3 (21 g L
1) as inorganic carbon source were continu-
ously fed at a rate of 0.8 and 0.4 L day
1, respectively.
2.2. Analytical methods
Continuous monitoring of outlet H2S concentration was per-
formed using an electrochemical H2S sensor (Sure-cell, Euro-
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
Gas Management Services LTD, UK) calibrated up to
300  2 ppmv. In order to measure higher H2S concentrations,
a mass flow controller (Bronkhorst, The Netherlands) was
used to precisely and continuously dilute the analyte flow
with air. Dilution ratios between 1:20 and 1:5 were used.
On-line liquid phase monitoring included pH and oxidation-
reduction potential (ORP) (PH 28, Crison Instruments, Spain) and
dissolved oxygen (DO) (oxi340i, WTW, Germany) measure-
ments. Also, daily samples of the purge flow were taken for ionic
sulfur species and total inorganic carbon (TIC) analysis, using an
ICS-1000 Ion Chromatography system with an IonPac AS9-HC
column (Dionex Corporation) and a TIC-TOC 1020 analyzer (IO
Analytical) respectively. Biomass concentration in the liquid
phase was measured as mg N L
1 according to van den Bosch
et al. (2007). S0 concentration in the liquid recirculation was
also measured according to Goehring and Helbing (1949).
2.3. Inoculation and start-up
Reactor inoculation was carried out using aerobic sludge from
a local municipal wastewater treatment plant (MWWTP)
diluted 1:1 with MM, thus having a final volatile suspended
solids (VSS) concentration of 1.9 g L
1. During the start-up
phase the inlet H2S concentration was set to 1000 ppmv
(28 g H2S m
3 h
1), the pH setpoint to 6.5e7 in order to match
the pH of the original inoculum, and the O2/H2S supplied ratio
to 15.7 (v v
1). During the first four days no new MM was
supplied, though the NaHCO3 was fed to the reactor to avoid
carbon limitation. 10% of the liquid volume needed to be
removed twice (on the second and third days) in order to keep
the liquid volume constant. Once the start-up phase was over
(after 5 days), the operating conditions were set as previously
described (see Section 2.1).
2.4. Specific experiments
The maximum elimination capacity (EC) and the effect of
reduced EBRTs were assessed 1 and 12 months after reactor
7.5 100
90
7.0
80
6.5 70
60
6.0
RE (%)
pH
50
5.5
40
5.0 30
20
4.5
10
4.0 0 0
0 1 2 3
Fig. 1 e On-line monitoring of the pH, H2S inlet and outlet concentrations and removal efficiency (RE) during the start-up
phase of the bioreactor.
5667
inoculation in order to asses the evolution of the bioreactor
treatment capacity over time. The EBRT was decreased step-
wise hourly from 180 s down to 30 s through gas flow increases
at a constant inlet H2S concentration of 2000 ppmv and an O2/
H2S supplied ratio of 23.6 (v v
1). This corresponds to LR
increases from 55 to 334 g H2S m
3 h
1. Also, the effect of
a stepwise increase of the trickling velocity at a constant EBRT
(180 s), LR of 84 g H2S m
3 h
1 (3000 ppmv), HRT of 10 h and O2/
H2S supply ratio of 23.6 (v v
1) was determined. The TLV was
increased stepwise from 0.52 to 20 m h
1 every 48 h (i.e., about
5 HRT) and the reactor response was monitored.
To test the effect of gas supply shutdowns, the biogas mimic
supply was stopped while the air flow, liquid recirculation, the
purge and the make-up water flows were all kept constant.
Since inorganic carbon in a full-scale system would be nor-
mally supplied via the gas, the HCO
3 supply in the make-up
water was also discontinued while the gas supply was stopped.
Finally, short-term, large pH variations were introduced in
the system, using either HCl or NaOH 1 M and the pH
controller, to assess the impact of pH shocks. Initially, a pH
drop down to 2.5 was imposed and kept constant for a period
of 34 h prior to resuming normal pH conditions (pH 6e6.5).
Afterward, a pH increase up to 9.5 was imposed and main-
tained for 24 h before returning the pH to its normal setpoint.
In both cases, a shorter HRT of 19  1 h was used in order to
shorten the reactor response time.
3. Results and discussion
3.1. Inoculation and start-up
After 1 h of operation at 1000 ppmv inlet concentration, H2S
was already detected in the outlet gas stream. This illustrates
the low sorption capacity of the system, even when working at
constant pH of 7 (Fig. 1). This rapid breakthrough pattern is
consistent with the high pKa and relatively unfavorable gas-
liquid partition of H2S (pKa1 ¼ 6.9; pKa2 ¼ 12.8;
2000
1800
1600
1400
[H2 S] (ppmv )
RE
[H2 S]in 1200
[H2 S]out
1000
pH
800
600
400
200
4 5 6 7 8 9 10
Time (days)
5668
dimensionless H ¼ 0.39). Even so, some biological activity
coupled to absorption of H2S into the fresh NaHCO3 solution
allowed RE to stay within 55e65% during the first day of
operation. The average DO concentration was 0.4 mg L
1 (data
not shown) which could have been limiting, and thus was
increased to 2 mg L
1 on day 2 by increasing the O2/H2S
supplied ratio from 15.7 to 23.6 (v v
1). After the first day, the
RE progressively increased; H2S outlet concentrations were
below the detection limit of the H2S sensor (30 ppmv, corre-
sponding to RE > 97%) already by day 2. In parallel,
a progressive shift in the sulfur species composition in the
liquid phase, i.e. from primarily reduced S species (H2S (aq) and
HS
) to more oxidized ones (S2 O2
2

3 and SO4 ) occurred due to
increasing biological activity as well as to some chemical
oxidation (Steudel, 2004). As shown in Fig. 2, there was an
initial accumulation and subsequent depletion of thiosulfate
and inorganic carbon during the first two days, concurrent
with a progressive accumulation of sulfate in the liquid phase.
This adds further evidence that hydrogen sulfide sorption and
chemical oxidation (Eq. (5)) predominated during the first two
days, but were rapidly surpassed by biological oxidation (Eqs.
2, 3 and 6), which became the main mechanism for removal
from day 3 onwards. The TIC profile was also in agreement
with the above explanation. For the first five days, the
carbonate supply was kept constant at 0.9  0.1 g C-NaHCO3
g
1 SeH2S with minimum purging of the trickling liquid (see
methods). Thus, despite of the increased CO2 stripping
because of the increase of the O2/H2S supplied ratio on day 2,
the TIC decrease observed after day 2 (Fig. 2) was most likely
linked to an increase in the biological uptake due to onset of
H2S removal.
Thus, a very short start-up phase of only 2 days when
based on H2S gas concentrations or about 5 days when
considering liquid phase concentrations was obtained.
Moreover, even after doubling the LR up to 56 g H2S m
3 h
1
(i.e., inlet H2S of 2000 ppmv) on day six, the system perfor-
mance in terms of RE remained high (Fig. 1). A start-up phase
of 3e5 days is a much shorter than observed in previous
experiments (Fortuny et al., 2008). Such a fast start-up was
50
TIC
45 SO42-
2-
S2 O3
40
TIC, S-S2 O3 2- (mg L )
-1
35
30
25
20
15
10
5 100
0 0
0 2 4 6 8 10 12
Time (days)
Fig. 2 e Sulfate, thiosulfate and total inorganic carbon
profiles during the first 20 days of operation. Day zero
concentrations correspond to the inoculum (MWWTP
sludge diluted 1:1 with MM) concentrations. The arrow
indicates the beginning of the liquid phase renewal.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
attributed to two main factors. First, the pH control ensured
a constant operation at a pH between 6.5 and 7.0 (Fig. 1), which
was the same pH as that of the original inoculum. The pH
control also avoided any pH increase resulting from the
addition of NaHCO3 in the early stages of the system opera-
tion, when there was little sulfate production to balance the
pH. Second, heavy inoculation with the MWWTP sludge may
have played an important role in the startup process. It has
been previously described (Prado et al., 2005), that a high
biomass concentration, like in MWWTP sludge, may facilitate
biofilm formation onto a new packing material.
Overall, this results show that specific sulfide oxidizing
cultures are not needed in order to start-up a biological sulfide
treatment system despite what has been reported by other
authors (Koe and Yang, 2000; Sercu et al., 2004; Duan et al.,
2006). Sludge from MWWTP works well as an inoculum
because of its high microbial diversity (Maestre et al., 2010)
and adding a high biomass concentration in the biotrickling
filter under suitable conditions ensures that sulfide oxidizing
organisms will rapidly establish a thriving community.
3.2. Maximum EC and effect of the EBRT
Fig. 3 shows real time data for the first determination of the
maximum EC carried out after one month of operation, while
Fig. 4 (run 1) reports the elimination capacities obtained at the
end of each step change as a function of the EBRT and the LR.
The results show that even after decreasing the EBRT to 120 s
(i.e., a LR of 84 g H2S m
3 h
1) the reactor was able to maintain
an average RE of 97.7  0.3%. At an EBRT of 90 s, the RE only
dropped to 88.6  0.5%. However, when the EBRT was
decreased further, there was and important effect on H2S
removal. The RE reached an average value of 39.7  0.9% at an
EBRT of 30 s (i.e., a LR of 334 g H2S m
3 h
1).
Examination of Figs. 3 and 4 shows that there was practi-
cally no RE reduction at LR up to 84 g H2S m
3 h
1 (or EBRT of
120 s). This would allow significant reduction of the EBRT
without any effect on H2S removal at 2000 ppmv inlet
concentration. Another important observation is the rapid
recovery after returning the system to its original EBRT of 180 s
(at time 288 min). The small ORP drop during the experiment
1000
(
50 mV, Fig. 3) provides information on the possible accu-
900 mulation of sulfide in the liquid. Montebello et al. (2010)
800 showed that ORP values around
50 mV corresponded to
S-SO4 2- (mg L )
very low sulfide liquid concentrations (<1 mg L
1) whereas
-1
700
600 significant sulfide accumulation caused ORP to drop below
500
200 mV. Thus, the
50 mV drop suggests that there was not
accumulation of sulfide in the liquid phase, indicating that
400
sulfide was oxidized rather than being absorbed. It is worth
300
mentioning that low concentrations of thiosulfate were
200
detected in the liquid phase, but the amount corresponded to
less than 1% of the total amount of H2S removed. Thus, abiotic
oxidation was negligible, even during the high loading
14 16 18 20
periods.
The absence of sulfide accumulation during the high-load
periods (LR up to 334 g H2S m
3 h
1) indicates that the rate
limiting step was mainly mass transfer. This is an interesting
finding which contrasts with previous results (Montebello
et al., 2010) in which the same reactor became kinetically
limited rather than mass-transfer limited when exposed to LR
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4 5669

100 200 25

90 180
0
80
160
70 -25
140
60

EBRT (s)

ORP (mV)
RE (%)

RE 120 -50
50 EBRT
ORP
100 -75
40
80
30 -100
60
20
-125
10 40

0 20 -150
-200 -150 -100 -50 0 50 100 150 200 250 300 350 400

Time (min)
Fig. 3 e Removal efficiency (RE), redox potential (ORP) and empty bed gas residence time (EBRT) during run 1 of the EBRT
experiment. The dashed line indicates beginning of experiment.

above 162 g H2S m
3 h
1 by only increasing the inlet concen- contact, therefore it is likely that increased bacterial
tration at a constant EBRT. Therefore, the system becomes coverage of the packing occurred after a year of operation,
kinetically limited when high LR are achieved by an increasing which resulted in a greater interfacial surface-to-volume ratio
concentration, but is mass transfer limited when high LR are inside the reactor. Another possible contributing factor is that
achieved by reducing the EBRT. slight accumulation of (bio)solids onto the clean, open-
A maximum EC of 126 g H2S m
3 h
1 was achieved (Fig. 4B), structure of the packing material could cause a slight gas
which compares favorably to other reported maximum EC velocity increase, which in turn would increase pollutant
(Koe and Yang, 2000; Sercu et al., 2004; Duan et al., 2006). This mass transfer. The former hypothesis seems more likely than
is remarkable value for a bioreactor that had been operating the latter, given the fact that wetting in biotrickling filters is
for only a month. It demonstrates that reactor design, inocu- often incomplete (Kim and Deshusses, 2008), which probably
lation and operating conditions during startup were close to extends the time required for complete bacteria colonization
optimum. of the packing.
The same determination of EC was repeated after one year
of reactor operation (run 2). A 14% increase in the maximum 3.3. Effect of the trickling liquid velocity
EC was observed, reaching an average value of
144  4 g H2S m
3 h
1 (Fig. 4B). Since no biological limitation The TLV is an important parameter for the attachment (and
was observed in run 1, the increase in the maximum EC of run shear) of biomass onto the packing material, for proper gas-
2 must be related to an improved H2S mass transfer. Pollutant liquid mass transfer and for S0 flushing in case of accumula-
transfer occurs through the gas-liquid and gas-biofilm tion. As shown in Fig. 5, no difference in the performance was

100 200
A B
175
80
150
EC (g H2 S m -3 h -1 )

60 125

100
40 Run 1
75
Run 2 Run 1
50 Run 2
20 EC = Load
25

0 0
0 20 40 60 80 100 120 140 160 180 200 0 50 100 150 200 250 300 350 400
EBRT (s) LR (g H2 S m-3 h-1 )

Fig. 4 e Results for both experimental runs of the EBRT experiment. A) Removal efficiency (RE) versus applied empty bed
residence time (EBRT). B) Elimination capacity (EC) versus applied loading rate (LR).
5670 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4

130 110 20
100 18

sulfate production (g S-SO42-m h-1)

120
RE

RE (%), O2 load (g O2 m-3h-1)

Trickling liquid velocity (m h-1)

90

-3
Trickling liquid velocity 16
110
80 Average O2 load
14
100 70 S-SO42-
12
60
90 10
50

80
8
40
6
30
70
20 4
60
10 2

50 0 0
94 96 98 100 102 104 106 108
Time (days)
Fig. 5 e Variation of the trickling liquid velocity (TLV), removal efficiency (RE), sulfate production and oxygen load supplied
to the packed bed during the 14 days experiment.

observed when changing the TLV within the tested range shown). These results indicate that increasing the TLV, in the
(0.51e19 m h
1) at a LR of 84 g H2S m
3 h
1. However, one short run (hours or days), may not be an efficient measure to
effect of changing the TLV was on oxygen transfer, since scour the S0 accumulated on the packing material. Once S0
changing the TLV altered the liquid retention time in the accumulates, it forms solid aggregates on the packing that are
oxygenation compartment located in the recycle line (Fortuny not easily removed. Further research is needed to establish
et al., 2010). Both the DO in the liquid phase (data not shown) optimum TLV or flushing methods to prevent S0 accumulation
and thus the oxygen load supplied to the packed bed (in terms onto the packing.
of g DO m
3 h
1actually available and supplied via the recir-
culation liquid) were increased with increases in the TLV 3.4. Effect of intermittent H2S supply
(Fig. 5). At TLVs lower than the standard conditions
(TLV < 3.8 m h
1), the DO in the recirculation liquid increased Starvation periods were implemented to simulate industrial
up to 4.5 mg L
1, even if the DO load supplied to the packed operation with short-term shutdowns. The gas supply was
bed was significantly reduced (79% and 20% reduction for discontinued for 2.5 and 5 days and the response of the reactor
TLV ¼ 0.51 and 2.04 m h
1, respectively, compared to the was monitored. Only a very small and brief reduction of the RE
oxygen load under standard conditions). Conversely, TLVs was detected during the first hours after resuming the gas
greater than 3.8 m h
1 which led to a DO reduction to an supply. The maximum RE reduction was 4% after the 2.5-day
average concentration of 2.5 mg/L, corresponded to an oxygen stop (results not shown) and 7% after the 5-day shutdown
load increase in proportion to the TLV. Consequently, raising (Fig. 6A) and a steady removal of over 99% was reestablished
the TLV caused a net increase in the O2 availability and within 4 h.
resulted in increased sulfate production (Fig. 5). Montebello Shortly after the H2S supply was resumed (day 86.7, Fig. 6A)
et al. (2010) already pointed out to the existence of DO gradi- the ORP dropped to about
150 mV, which indicated a slight
ents throughout the bed which could be partially reduced by sulfide accumulation, in agreement with the transient drop in
increasing the TLV due to a larger penetration of DO RE. However, ORP returned to pre-shutdown values
throughout the bed depth when liquid and gas flows operate (between  50 mV) within less than 24 h, indicating absence of
in counter-current mode. It is relevant to stress that this dissolved sulfide and presence of oxygen (DO 0.5e2 mg L
1).
occurred without any change in the air gas flow rate or oxygen The ORP pattern after resuming H2S indicated a temporary lag
amount supplied to the system and thus, from an operational of biological activity followed by a complete recovery. If the
point of view, such increased sulfate production is an inter- culture had been severely affected by the starvation, greater
esting strategy as it could be used to reduce S0 accumulation accumulation of sulfide and an ORP drop below
350 mV
without the need for further dilution of the raw gas. would have been observed (Montebello et al., 2010).
The increase in TLV above typical values for biological Interestingly, a sulfur mass balance during the starvation
systems (1e12 m h
1 see Kim and Deshusses, 2008) had been period reveals that actual sulfate concentrations were 30e40%
expected to cause a direct effect on (bio)solids shear and wash higher than expected from simple dilution by the make-up
out. Surprisingly, monitoring of the S0 and biomass concen- water (Fig. 6B). Because no dissolved sulfide had accumu-
trations in the recirculation liquid did not show a relationship lated, the excess sulfate concentration measured indicates
between S0 or biomass concentration and TLV (data not that biological activity did not completely stop during the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4 5671

8.0 200
A
100 7.5
150
7.0

LR (g H 2 S m -3 h -1 ), RE (%)
100
80
6.5
50

ORP (mV)
6.0
60

pH
5.5 0

5.0
40 -50
4.5
LR -100
20 RE 4.0
pH -150
ORP 3.5

0 3.0 -200
80 81 82 83 84 85 86 87 88 89 90

Time (days)
800
B
100 LR
700
SO4 2- measured

SO4 2- expected 600

80
LR (g H 2 S m -3 h -1 )

SO 4 2- (mg S L )
-1
500
60
400

40 300

200
20
100

0 0
80 81 82 83 84 85 86 87 88 89 90
Time (days)
Fig. 6 e Performance of the reactor during and after the 5-day H2S starvation period. The gas stream was discontinued, but
the liquid feed was maintained. A) pH, redox potential (ORP), removal efficiency (RE) and inlet load. B) Inlet load, measured
sulfate concentration and expected sulfate concentration (from washout calculation.

starvation period and that oxidation of accumulated S0 inside
the reactor occurred, as previously reported (Fortuny et al.,
2010). The source of oxygen for the aerobic oxidation of S0
was both via the liquid feed, and possibly diffusion of atmo-
spheric O2 into the biotrickling filter.
These results show that H2S shutdowns up to 5 days
have no, or very little impact on the long-term reactor
operation. Proper control of pH during starvation events is
likely to be important for prompt recovery. It is also prob-
able that biological oxidation of accumulated biological S0
during the starvation contributed to the fast recovery of the
biotrickling filter, and could possibly allow longer periods of
starvation without a greater impact on the reactor
performance.
3.5. Effect of short-term pH changes
Complete biological oxidation of sulfide leads to protons
production and thus pH changes. However, it is not known to
what extent, pH variations need to be controlled to ensure
stable bioreactor operation. Since a pH control failure (lack or
excess of control actuation) is relatively plausible during
industrial operation, the reactor response to large but fast pH
variations was studied.
On day 135 the reactor pH was forced to drop to 2.5 for
a period of 34 h. Later, after 60 h of operation at normal
conditions (i.e., pH 6e6.5), the pH was increased to 9.5 for
a 24 h period. As shown in Fig. 7A, the pH drop did not affect
the RE of H2S. However, a sulfur mass balance indicated that
5672 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
Fig. 7 e Reactor response to pH drop and pH increase episodes. A) Redox potential (ORP), pH and removal efficiency (RE). B)
Sulfate and thiosulfate concentration in the liquid phase.
a 25% reduction in the sulfate production occurred after
decreasing the pH (Fig. 7B), probably indicating a reduction in
the biological activity. A possibility is that under these
conditions, sulfide chemical oxidation to thiosulfate, which is
unstable at low pH and chemically reacts to produce S0
contributed to maintain the high RE. This is consistent with
the sulfur speciation shown in Fig. 7B, in which neither thio-
sulfate (chemically converted to S0) nor other sulfurous ionic
species accumulated. Inorganic polysulfides were not
measured since they only occur at pH above 6 (Steudel, 2004).
Sulfate production increased as soon as the pH was returned
to its original value and continued to increase throughout the
period at which the pH was kept at 6e6.5 (Fig. 7B). It is worth
noticing that sulfate production rates greater than the sulfur
load ðS
SO2
4 =S
H2 Sremoved > 100%Þ were encountered,
probably due to the oxidation of accumulated S0.
The high pH test was conducted next. Immediately after
increasing the pH to 9.5 for 24 h, a sharp drop in the ORP
occurred indicating an accumulation of sulfide (Fig. 7A) while
a shift in the sulfur mass balance from sulfate to thiosulfate
was observed (Fig. 7B). These reveal that a significant slow-
down of the biological activity as well as a shift in the
metabolism occurred as a result of the pH change. However, at
high pH, both physical absorption of H2S and chemical
oxidation of dissolved sulfide to thiosulfate (van den Bosch
et al., 2008) and possibly (poly)sulfide are favored. Therefore,
although the biological oxidation was affected, the removal of
H2S only slowly dropped to about 95% at the end of the high pH
step. When the pH was returned to 6e6.5, an important
transient phase of very low RE (<20%) was observed (Fig. 7A).
This was attributed to H2S stripping and is consistent with the
concurrent increase in ORP (i.e., decrease in dissolved sulfide)
during this short phase. For the next two days, biological
activity was probably severely inhibited and the primary
means for H2S removal was likely sorption and chemical
oxidation to thiosulfate. Thiosulfate concentration slowly
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
decreased because of dilution rather than to biological
oxidation since biological oxidation would have led to
a sulfate concentration increase. It was not until day 143, i.e.,
close to 3 days after the perturbation, that biological oxidation
to sulfate became again the primary removal means, as
shown by the ORP increase (Fig. 7A) and sulfate yields (Fig. 7B).
The H2S RE reached over 99% indicating that the reactor had
recovered and was operating normally about 3 days after the
pH spike.
This demonstrates that the bioreactor was much more
susceptible to high pH exposure than to low pH changes. Such
behavior can be explained by the fact that acidification is the
natural consequence of sulfide oxidation and, therefore, most
sulfide oxidizers are more tolerant to acidic conditions than to
alkaline conditions (Brüser et al., 2004; Syed et al., 2006). Also,
a low pH results in a decrease of the dissolved sulfurous
species concentration (H2S by stripping and polysulfides and
thiosulfate by chemical destabilization), whereas a high pH
leads to accumulation of dissolved sulfurous species (sulfides
and polysulfides), which can negatively affect the activity of
the sulfide degraders.
4. Conclusions
A detailed study of the operational aspects affecting the bio-
logical sweetening of energy gases mimics was carried out in
a lab-scale biotrickling filter. Results showed that:
B Detailed monitoring of the process with on-line H2S gas, ORP
and pH sensors combined with off-line analysis of sulfurous
species dissolved in the trickling liquid provided a detailed
understanding of the phenomena involved during the
treatment of H2S in the biotrickling filter.
B Inoculation of the biotrickling filter with MWWTP sludge led
to a very fast (3 days) start-up therefore showing that,
provided certain conditions, it is not necessary to obtain
a specific culture of sulfide oxidizers to inoculate an H2S
degrading system.
B The EBRT (so reactor volume) can be importantly reduced
without significantly affecting the removal of H2S while
treating 2000 ppmv H2S. However, excessive EBRT reduction
lead to an important RE drop due to mass transfer limitation
and not due to biological limitation as observed in our
previously research when the LR was increased by
increasing the H2S inlet concentration.
B Gas supply shutdowns for up to 5 days had little effect on
the biotrickling filter performance after resuming normal
operation. Possibly, controlling pH, addition of make-up
water coupled with biological oxidation of S0 present in
the biotrickling filter helped maintain a healthy biological
activity during the perturbation.
B A high TLV is recommended because it favors sulfate
production through a better use of the oxygen supplied.
B A short (34 h) but wide (down to 2.5) pH drop was much less
aggressive to the biological activity and overall reactor
performance than a short (24 h) but also wide (up to 9.5) pH
rise. Even so, the robustness of the system allowed for a full
recovery after about 48 h of normal operation.
5673
Acknowledgments
The Spanish government (MEC) provided financial support
through the project CICYT CTM2009-14338-C03-01. The
Department of Chemical Engineering at UAB (Universitat
Autònoma de Barcelona) is a unit of Biochemical Engineering
of the Xarxa de Referència en Biotecnologia de Catalunya
(XRB), Generalitat de Catalunya. The study sponsors were not
involved in: study design; collection, analysis, and interpre-
tation of data; writing of the report; nor in the decision to
submit the paper for publication.
references
Bailón, L., 2005. Development of a biotrickling filter for the
removal of H2S from biogas. In: Proceedings of the II
International Congress on techniques for Air Pollution
Control. La Coruña, Spain, pp. 143e148.
Brüser, T., Lens, P.N.L., Trüper, H.G., 2004. The biological sulfur
cycle. In: Lens, P., Hulshoff Pol, L. (Eds.), Environmental
Technologies to Treat Sulfur Pollution. Principles and
Engineering. IWA Publishing, London, UK, pp. 47e86.
Buisman, C.J.N., Post, R., Ijspeert, P., Geraats, G., Lettinga, G., 1989.
Biotechnological process for sulphide removal with sulphur
reclamation. Acta Biotechnologica 9, 255e267.
Devinny, J.S., Deshusses, M.A., Webster, T.S., 1999. Biofiltration
for Air Pollution Control. CRC-Lewis Publishers, Boca Raton,
Florida, USA.
Duan, H., Koe, LC.C., Yan, R., Chen, X., 2006. Biological treatment
of H2S using pellet activated carbon as a carrier of
microorganisms in a biofilter. Water Research 40, 2629e2636.
Fortuny, M., Baeza, J.A., Deshusses, M.A., Gamisans, X., Casas, C.,
Lafuente, J., Gabriel, D., 2008. Biological sweetening of energy
gases mimics in biotrickling filters. Chemosphere 71, 10e17.
Fortuny, M., Guisasola, A., Casas, C., Gamisans, X., Lafuente, J.,
Gabriel, D., 2010. Oxidation of biologically produced elemental
sulfur under neutrophilic conditions. Journal of Chemical
Technology and Biotechnology 85 (3), 378e386.
Gabriel, D., Deshusses, M.A., 2003. Retrofitting existing chemical
scrubbers to biotrickling filters for H2S emission control.
Proceedings of the National Academy of Sciences of U.S.A 100
(11), 6308e6312.
González-Sánchez, A., Revah, S., Deshusses, M.A., 2008. Alkaline
biofiltration of H2S odors. Environmental Science &
Technology 42 (19), 7398e7404.
Goehring, M., Helbing, W., 1949. Journal of Analytical Chemistry
V, 129e346.
Kim, S., Deshusses, M.A., 2008. Determination of mass transfer
coefficients for packing materials used in biofilters and
biotrickling filters for air pollution control - 2: development of
mass transfer coefficients correlations. Chemical Engineering
Science 63 (4), 856e861.
Koe, L.C.C., Yang, F., 2000. A bioscrubber for hydrogen sulphide
removal. Water Science and Technology 41 (6), 141e145.
Kuenen, J.G., 1975. Colourless sulphur bacteria and their role in
the sulphur cycle. Plant and Soil 43, 49e76.
Maestre, J.P., Rovira, R., Álvarez Hornos, J., Fortuny, M.,
Lafuente, J., Gamisans, X., Gabriel, D., 2010. Bacterial
community analysis of a gas-phase biotrickiling filter for
biogas mimics desulfurization through the rRNA approach.
Chemosphere 80, 872e880.
Montebello, A.M., Baeza, M., Lafuente, F.J., Gabriel, D., 2010.
Monitoring and performance of a desulphurization
5674 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 6 5 e5 6 7 4
biotrickling filter with an integrated continuous gas/
liquid flow analyser. Chemical Engineering Journal 165,
500e507.
Prado, Ó.J., Veiga, M.C., Kennes, C., 2005. Treatment of gas-phase
methanol in conventional biofilters packed with lava rock.
Water Research 39, 2385e2393.
Ross, C.C., Drake, T.J., Walsh, J.L., 1996. Handbook of Biogas
Utilization, second ed. SERBEP, c/o General Bioenergy,
Florence, AL.
Sercu, B., Núñez, D., van Lagenhove, H., Aroca, G., Verstraete, W.,
2004. Operational and microbiological aspects of a
bioaugmented two-stage biotrickling filter removing hydrogen
sulfide and dimethyl sulfide. Biotechnology and
Bioengineering 90 (2), 259e269.
Steudel, R., 2004. The chemical sulfur cycle. In: Lens, P.N.L.,
Hulshoff Pol, L. (Eds.), Environmental Technologies to Treat
Sulfur Pollution. Principles and engineering. IWA Publishing,
London, UK, pp. 1e31.
Syed, M., Soreanu, G., Falletta, P., Béland, M., 2006. Removal of
hydrogen sulphide from gas streams using biological processes -
a review. Canadian Biosystems Engineering 48, 2.1e2.14.
Tchobanoglous, G., Burton, F.L., Stensel, H.D., 2003. Wastewater
Engineering. Treatment and Reuse, fourth ed. McGraw-Hill
Companies, New York, NY, USA, pp. 1505e1532.
van den Bosch, P.L.F., van Beusekom, O.C., Buisman, C.J.N.,
Janssen, A.J.H., 2007. Sulfide oxidation at halo-alkaline
conditions in a fed batch bioreactor. Biotechnology and
Bioengineering 97 (5), 1053e1063.
van den Bosch, P.L.F., Sorokin, D.Y., Buisman, C.J.N., Janssen, A.J.
H., 2008. The effect of pH on thiosulfate formation in a
biotechnological process for the removal of hydrogen sulfide
from gas streams. Environmental Science and Technology 42
(7), 2637e2642.
Yang, Y., Allen, E.R., 1994. Biofiltration control of hydrogen
sulfide.1. Design and operational parameters. Journal of Air
and Waste Management 44, 863e868.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

The rheological behaviour of anaerobic digested sludge

J.C. Baudez a,b,*, F. Markis b, N. Eshtiaghi b, P. Slatter b

a
Cemagref, UR TSCF, F-03150 Montoldre, France
b
Rheology and Materials Processing Centre, Dept. of Chemical Engineering, RMIT University, Victoria 3001, Australia

article info abstract

Article history: Producing biogas energy from the anaerobic digestion of wastewater sludge is one of the
Received 20 May 2011 most challenging tasks facing engineers, because they are dealing with vast quantities of
Received in revised form fundamentally scientifically poorly understood and unpredictable materials; while
3 August 2011 digesters need constant flow properties to operate efficiently. An accurate estimate of
Accepted 19 August 2011 sludge rheological properties is required for the design and efficient operation of digestion,
Available online 1 September 2011 including mixing and pumping. In this paper, we have determined the rheological
behaviour of digested sludge at different concentrations, and highlighted common
Keywords: features. At low shear stress, digested sludge behaves as a linear viscoelastic solid, but
Digested sludge shear banding can occur and modify the apparent behaviour. At very high shear stress, the
Bingham model behaviour fits well to the Bingham model. Finally, we show that the rheological behaviour
HerscheleBulkley model of digested sludge is qualitatively the same at different solids concentrations, and depends
Shear banding only on the yield stress and Bingham viscosity, both parameters being closely linked to the
Viscoelasticity solids concentration.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
Renewable energy is said to be one of the pillars of sustainable
management. Biogas from the anaerobic digestion of sewage
sludge can provide a clean, easily controlled source of
renewable energy from sewage sludge, replacing fossil fuels.
However, an accurate estimate of sludge rheological proper-
ties is required for the design and efficient operation of the
pumping systems which surround anaerobic digesters
(Slatter, 1997, 2003). Indeed Tarp and Melbinger (1967) showed
the significant advantages of recycling and recirculating
digested sludge to mix it with excess sludge, among them an
increase in biogas production (Sperry, 1959). The mixture can
be concentrated to a much higher solid content than would be
possible for the excess sludge alone, and recirculation also
facilitates improved mixing efficiency over mechanical stir-
ring. However, the flow rate in the recirculation circuits has to
be very large (Appels et al., 2008) and rheology is needed to
calculate head losses and pumping power (Slatter, 2001).
Except for the work of Monteiro (1997) who showed that
anaerobic digestion induces a decrease of the rheological
characteristics of sludge, most investigations on sludge
rheology were focused on activated sludge. No reliable data, at
high shear rate (within recirculation pipes), can be found in
the literature for digested sludge while at low shear rate
(within the digester), results are scarce and not always usable.
Most of these were obtained by applying shear rate ramps that
gave distinct peaks in the flow curve (for example, Ayol, 2006),
but Baudez (2006) clearly established that these peaks in the
flow curves were principally instrument artefacts, and not
material characteristics. However, the work of Ayol et al.
(2006) pointed out that with very dilute sludge, the Ostwald
model, i.e. a power-law model with no yield stress, gave the
best fit.
From a physical perspective, digested sludge appears to be
a stable suspension with low settling rates (Namer and
Ganczarczyk, 1993) and low surface charge (Forster, 2002),
implying that interactions are more steric than electrostatic.

* Corresponding author. Cemagref, UR TSCF, F-03150 Montoldre, France.

E-mail address: jean-christophe.baudez@cemagref.fr (J.C. Baudez).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.035
5676 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0
The most important constituents in digested sludge are lipo-
polysaccharides (Forster, 1983) which are amphiphile lipids
with both hydrophilic and hydrophobic heads. These mole-
cules displayed a very intriguing rheological behaviour
(Muñoz et al, 2000), showing linear viscoelasticity, non-
Newtonian viscous flow and shear banding (Miller and
Rothstein, 2007).
In this paper, our intention is to establish the basic char-
acteristics of the rheological behaviour of digested sludge,
with the objective of industrial applications in digester mix-
ing, pumping and pipe flows, meaning that we will focus on
short-term behaviour. Short-term behaviour means we will
not focus our research on eventual thixotropic effects. As
predicted by the literature on amphiphile rheology, we show
that digested sludge exhibits linear viscoelastic behaviour at
low shear stresses, followed by shear banding phenomena at
intermediate stresses, and finally a non-Newtonian fluid
behaviour with a yield stress, modelled by a HerscheleBulkley
model at intermediate shear rates and by a Bingham model at
very high shear rates. We also highlight the fact that the
rheological behaviour is qualitatively the same at different
solid concentrations, allowing us to define a master curve for
which the dimensionless parameters are the yield stress and
the Bingham viscosity.
2. Material and methods
The digested sludge was sampled at the Mount Martha waste
water treatment plant (Melbourne, Victoria, Australia) at the
outlet of the digester number 1. Its initial solid concentration
was at 18.5 g L
1 and was also gently concentrated to 25.5, 32
and 49 g L
1 by using a Buchner vacuum. Sludge samples were
stored at 4  C for 30 days before experiments, in order to
reduce temporal variability. Indeed, even after anaerobic
digestion, sludge may not be fully stabilised and organic
changes may still occur. By storing the sludge sample for such
an extended period, the potential for composition changes is
reduced; and we can assume that we used exactly the same
material throughout all our experiments.
Rheological measurements were performed with a DSR200
instrument from Rheometric Scientific, connected to
a temperature controlled water bath. The rheometer was
equipped with a cup and bob geometry (inner diameter:
29 mm, outer diameter: 32 mm, length: 44 mm). Temperature
was kept at 25  C. To avoid evaporation, sludge was covered
with a thin film of immiscible Newtonian oil.
Before each measurement, sludge was presheared for
10 min at a shear rate of 1000 s
1 then left at rest for 10 min.
This procedure allowed us to erase material memory and to
have reproducible measurements. Then, different tests were
performed:
 Shear stress sweep, by applying a linear ramp of increasing
stress over time. In this test, we changed the time of rest
between preshear and shear, from 1 to 60 min in order to
investigate structural changes occurring during rest;
 Creep test, by applying constant shear stress and measuring
the corresponding shear strain, at different shear stresses in
the linear viscoelastic regime and above;
 Decreasing stress ramp to determine the flow curve, starting
at a high stress corresponding to a shear rate of approxi-
mately 1000 s
1 or lower for the less concentrated sludge (to
avoid turbulent conditions).
3. Results and discussion
Starting from rest, the shear stress sweep first elicits a linear
viscoelastic response from the digested sludge up to a critical
shear stress s0 above which the material apparently starts to
flow (Fig. 1). In the linear viscoelastic region, the behaviour is
modelled by a generalised Kelvin-Voigt model, with a wide
relaxation time spectrum modelled by a stretched
exponential:
1


gðtÞ ¼ s$ $ 1
exp
ðltÞm (1)
G
where g represents the strain, s the stress and l ¼ G=m with G
and m the usual parameters of a Kelvin-Voigt model.
Assuming that the sludge is flowing in its liquid regime
above the critical shear stress following a HerscheleBulkley
model (Monteiro, 1997), the additional strain can be expressed
as:
Zt Zt  1=n Z t  1=n
s
s0 a$x
a$t0
gðtÞ ¼ _
gdu ¼ ¼ dx
K K
t0 t0 t0
a$n
¼ ðt
t0 Þ1þ1=n (2)
ðn þ 1ÞK
where a is the slope of the shear stress ramp and t0 the time
such that the shear stress equals the yield stress of the Her-
scheleBulkley model s0 ¼ a$t0 .
Thus, the total strain, which predicts the experimental
data (Fig. 1), can be expressed as
1


gðtÞ ¼ s$ $ 1
exp
ðltÞn þ b$ðt
t0 Þ1þ1=n (3)
G
with b ¼ an=ðn þ 1Þ$K
Applying this to the experimental data gave a flow behav-
iour index for the HerscheleBulkley model, n, greater than 1
(Fig. 1), meaning that the digested sludge could apparently be
a shear-thickening liquid above s0 , which is unusual.
Creep tests confirmed a change in the behaviour above s0 .
Below s0 , the strain slowly increased with time, while above s0 ,
the increase is faster (Fig. 2), both following a power-law with
time. However, even for stresses higher than s0 (Fig. 2) the
shear strain follows a power-law with a power-law index less
than 1, indicating that the shear rate is a decreasing function
of time: there is no steady state and so, sludge is restructuring
and not flowing (otherwise, the shear rate would have been
constant over time for a constant shear stress). The value s0
cannot therefore be considered as a classical yield stress
above which digested sludge flows in its liquid regime. These
power-law relationships between strain and time are in fact
a consequence of a structural relaxation process which occurs
during creep (Baudez, 2008).
When the time of rest between the preshear and the stress
sweep increases, the behaviour is globally the same, with first
a linear viscoelastic behaviour (Fig. 3) but the critical shear
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0 5677

Fig. 1 e Strain-stress behaviour of the 4.9% digested

sludge. The dashed lined corresponds to the model of (1)
with G [ 0.62 Pa, l [ 3.3.10L7 sL1, m [ 0.34, b [ 0.35 sL2, Fig. 3 e Strain-stress behaviour when a stress sweep is
t0 [ 315.5 s, corresponding to a stress equals to 2.13 Pa and applied after different time of rest.
n [ 1.30.

representative of the true material behaviour but derives from

stress, s0 decreases with increase of the time of rest, the global
elasticity decreases, the mean relaxation time (inverse of l)
increases and the strain corresponding to s0 decreases (Fig. 4):
the longer the time of rest, the smaller the linear viscoelastic
range.
At rest, the digested sludge structure became weaker and
weaker (decrease of s0 and elasticity) but concurrently the
relaxation time increased, indicating an evolution from
a viscoelastic material towards a more elastic solid (the
decrease of m is faster than the decrease of G). Since this is
physically impossible, this observed apparent behaviour is not
erroneous interpretation of raw data.
Above s0 , experimental results showed the viscosity is
globally decreasing, which is inconsistent with the apparent
shear-thickening behaviour noted earlier, but oscillations of
viscosity regarding shear stress are reported (Fig. 5). These
oscillations indicated local minima in the flow curve where
apparent shear rate occasionally decreased while shear stress
increased.
If we assume the relationship between local shear rate and
local shear stress is monotonic, then we can write:
g_ ¼ f ðsÞ where f is the inverse function of the behaviour
law.
In a Couette geometry, the shear rate can be expressed as:

ZR2 ZR
vðuÞ g_ local f ðslocal Þ
g_ local ¼
r$ 5u ¼ dr5u ¼ dr (4)
vr r r
R1 R1

Fig. 2 e Creep test below, above and equal to the critical

shear stress. Here, the critical stress is 2.5 Pa for the 4.9%
sludge. The insert is a focus on the strain at the highest
strain at longer time, following a power-law with an index Fig. 4 e Evolution of the Kelvin-Voigt model parameters as
smaller than 1. a function of the time of rest.
5678
Fig. 5 e Stress-viscosity variations highlighted oscillations
with the 4.9% digested sludge.
where u is the angular velocity, R1 the inner radius and R
R1
the thickness of the sheared region. The maximum value of
R
R1 is R2
R1 , where R2 is the outer radius.
The apparent shear rate is calculated from the measured
angular velocity, the only raw data measured by the rheom-
eter. The shear rate given by the rheometer is calculated with
the assumption of a full shear within the gap. So, if the
apparent shear rate decreased, it means the angular velocity
decreased. However, because f is a monotonic function, this
decrease of u is rather the consequence of a decrease of the
effective gap R
R1 , implying that shear banding has occurred
during the measurement.
Such behaviour (shear banding and viscoelastic behaviour)
has to be taken into account in digester design and operation,
because shear banding means that there is coexistence of
both sheared and unsheared zones in the digester, these last
being useless, unmixed, dead zones.
60
Test 1
50 Test 2
Test 3
Shear stress [Pa]
40
30
20
10
0 0
1 10 100
Shear rate [s-1]
Fig. 6 e Repeatability of the measurements.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0
Fig. 7 e Flow curves regarding the concentration of the
digested sludge.
According to Moller et al. (2008), the width of the flowing
band can be directly related to the macroscopically imposed
shear rate. At high shear rates, the whole gap is sheared and
when the applied stress is much higher than s0 , the sludge
flows normally, with no apparent perturbation effects,
allowing us to have achieve reproducible measurements
(Fig. 6) with the corresponding smooth classical shape of the
flow curve.
As expected, the higher the concentration, the thicker the
sludge (Fig. 7) but depending on the shear rate range, different
well-known models can be used to describe the rheological
behaviour of digested sludge. At high shear rates, a basic
Bingham model is sufficient (Fig. 8) while at low and inter-
mediate shear rates, HerscheleBulkley and power-law models
are more appropriate (Fig. 9). They all represent the same
material but can only be used in a specific range of validity,
regarding the complexity of the process to be modelled. Thus,
70
60
50
Shear stress [Pa]
40
30
20
10
0
500 1000 1500
1000
Shear rate [s-1]
Fig. 8 e At high shear rates, the rheological behaviour can
be basically modelled with a Bingham plastic model.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0 5679

Fig. 9 e At low and intermediate shear rates, the

HerscheleBulkley model or power-law model are the most Fig. 10 e Dimensionless flow curves of the digested sludge
suitable. The dashed line represents the power-law model at different concentrations.
s[2:05,g_ 0:45 .

s K a0 s
s ¼ sc þ K$g_ n þ a0 $g5
_ ¼ 1 þ $g_ n þ $g5
_ ¼ 1 þ b$Gn þ G
for pumping where shear rates are very high, a Bingham  n s c sc sc sc
(6)
a0 K sc
model would be appropriate since it deals with simple char- G ¼ $g; _ b¼ $
sc s c a0
acteristics, i.e. a yield stress and a constant rheogram slope
above it.
From a more general point of view, in the liquid regime we In such a dimensionless form, all the flow curves are
can summarize the rheological behaviour of digested sludge similar, independent of solids concentration (Fig. 10). From
as a shear-thinning yield stress fluid with a plateau viscosity a physical point of view, this result means that there is some
at high shear rates: similarity of the network of interactions within the sludge at
different concentrations, which is at the origin of the simi-



s ¼ sc þ h g_ $g_ with h g_ / a0 larity of its macroscopic behaviour. In such suspensions,
_
g/N
interactions can be classified into two main groups (Baudez
and Coussot, 2001): hydrodynamic interactions (between
Moreover, at low and intermediate shear stresses,
solid particles and surrounding fluid, here basically
s ¼ sc þ K$g_ n ¼ sc þ hðgÞ$
_ g5hð
_ _
gÞzK$ g_ n
1
Thus, for the sake of simplicity, we define the rheological
behaviour of digested sludge as follows:
 
s ¼ sc þ K$g_ n
1 þ a0 $g_ (5)

On our range of data, i.e. below 1000 s
1, this model was
pffiffiffiffiffiffiffiffiffiffiffi
successful. However, if g_ << 1
n K=a0 , the HerscheleBulkley
model is sufficient to model the behaviour, which corresponds
to a shear rate smaller than 565 s
1 for the most concentrated
sludge and smaller than 145 s
1 for the less concentrated
sludge as shown below (Table 1).
Eq. (5) can also be expressed as:

Table 1 e Shear rate above which the HerscheleBulkley

model is not suitable.
Concentration [%] Limit shear rate [s
1]

1.85 145 Fig. 11 e Evolution of the yield stress and the Bingham
2.56 280
viscosity regarding the concentration. The parameters of
3.17 470
4.89 565
the Eqs. (7) and (8) are respectively a [ 0.19 Pa, f0 [ 1.17%,
m [ 1.89 and m0 [ 0.0018 Pa s, b [ 0.604.
5680 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 7 5 e5 6 8 0
represented by the Bingham viscosity) and non-
hydrodynamic interactions (between solid particles, basi-
cally represented by the yield stress). Increasing the concen-
tration doesn’t change the nature of these interactions, but
rather modifies their relative intensity. The dimensionless
form smoothed these differences because both kinds of
interactions in this form will approach unity.
On our range of concentrations, yield stress and Bingham
viscosity increase with the solid concentration (Fig. 11)
respectively following a power-law and an exponential law of
the following form, which is in agreement with the literature,
both for the yield stress (Baudez, 2008) and the Bingham
viscosity (Sanin, 2002):
sc ¼ a$ðf
f0 Þm
k2 ¼ m0 $expðb$fÞ
where f0 is the lowest concentration below which there is no
yield stress, m is related to the fractal dimension of sludge flocs
(Baudez, 2008) and m0 is the viscosity of the liquid medium.
We found that the value m0 is twice that of pure water, which
can be explained by the large amount of dissolved matter
present, which may increase the supernatant viscosity.
4. Conclusion
In this paper, we have shown that digested sludge is a shear-
thinning yield stress fluid, presenting flow instabilities at low
shear rates, manifesting as shear banding. At low shear stress,
below the yield stress, digested sludge behaved as a visco-
elastic solid. When the applied stress is increased, above
a critical shear strain, which decreases with the restructuring,
shear banding appears. Then, at higher stresses, digested
sludge behaves like a yield stress fluid and can be modelled
using both the HerscheleBulkley and Bingham plastic models
over a wide range of shear rates.
This behaviour was similar at different concentrations and
yield stress followed a power-law with the concentration
while the Bingham viscosity followed an exponential law with
concentration.
By reducing the rheological parameters with the yield
stress and the Bingham viscosity, which have to be measured
separately, a master curve was obtained. This result means
that the rheological behaviour of the digested sludge at any
concentration can be deduced from this master curve.
However, further work has to be done on shear banding.
This behaviour will have to be taken into account in digester
design and process operations, in order to avoid dead zones in
the digester.
Acknowledgements
The authors acknowledge the Cemagref-RMIT agreement for
our collaboration.
references
Appels, L., Baeyens, J., Degrève, J., Dewil, R., December 2008.
Principles and potential of the anaerobic digestion of
waste-activated sludge. Prog. Energy Combust. Sci. 34 (6),
755e781.
Ayol, A., Filibeli, A., Dentel, S.K., 2006. Evaluation of conditioning
(7) responses of thermophilic-mesophilic anaerobically and
mesophilic aerobically digested biosolids using rheological
properties. Water Sci. Technol. 54 (5), 23e31.
(8)
Baudez, J.C., 2006. About peak and loop in sludge rheogram. J.
Environ. Manage. 78, 232e239.
Baudez, J.C., 2008. Physical aging and thixotropy in sludge
rheology. Appl. Rheology 18 (13495), 1e8.
Baudez, J.C., Coussot, P., 2001. Rheology of aging, concentrated,
polymeric suspensions e Application to pasty sewage sludges.
J. Rheol. 45 (5), 1123e1139.
Forster, C.F., 1983. Bound water in sewage sludge and its
relationship to sludge surfaces and sludge viscosities. J. Chem.
Tech. Biol. 33B, 76e84.
Forster, C.F., 2002. The rheological and physico-chemical
characteristics of sewage sludge. Enzym. Microb. Tech. 30 (3),
340e345.
Miller, E., Rothstein, J.P., 2007. Transient evolution of shear
banding in wormlike micelle solutions. J. Non-Newtonian
Fluid Mech. 143, 22e37.
Moller, P.C.F., Rodts, S., Michels, M.A.J., Bonn, D., 2008. Shear
banding and yield stress in soft glassy materials. Phys. Rev. E
77, 041507.
Monteiro, P.S., 1997. The influence of the anaerobic digestion
process on the sewage sludges rheological behaviour. Water
Sci. Technol. 36 (11), 61e67.
Munoz, J., Alfaro, M.C., 2000. Rheological and phase behaviour of
amphiphilic lipids. Grasas y aceites 51, 6e25.
Namer, J.J., Ganczarczyk, L., 1993. Settling properties of digested
sludge particle aggregates. Water Res. 27, 1285e1294.
Sanin, F.D., 2002. Effect of solution physical chemistry on the
rheological properties of activated sludge. Water SA 28,
207e212.
Slatter, P., 1997. The rheological characterisation of sludges.
Water Sci. Technol. 36 (11), 9e18.
Slatter, P., 2001. Sludge pipeline design. J. Water Sci. Technol. 44
(10), 115e120.
Slatter, P., 2003. Pipeline transport of thickened sludges. Water 21,
56e57.
Sperry, W.A., 1959. Gas recirculation at Aurora, Illinois. Sewage
Ind. Waste 31 (6).
Tropey, W.N., Melbinger, N.R., 1967. Reduction of digested sludge
volume by controlled recirculation. J. Water Pollut. Control
Fed. 39 (9), 1464e1474.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Survival of environmental and clinical strains

of methicillin-resistant Staphylococcus aureus [MRSA]
in marine and fresh waters

Emily Levin-Edens a, Natasha Bonilla b, J. Scott Meschke a, Marilyn C. Roberts a,*

a
Department of Environmental and Occupational Health Sciences 357234, School of Public Health, 1959 NE Pacific St, University of
Washington, Seattle, WA 98195-7234, USA
b
Department of Biology, University of Puerto Rico, San Juan, Puerto Rico

article info abstract

Article history: Recent studies have found variable levels of methicillin-resistant Staphylococcus aureus
Received 26 April 2011 [MRSA] in marine water from temperate and warmer climates suggesting that temperature
Received in revised form may play a role in survival of MRSA in the environment. The aim of the study was to
19 August 2011 compare the survival of clinical and environmental MRSA and MSSA strains in fresh and
Accepted 20 August 2011 marine water incubated at 13
C and 20
C over 14 days. Seven different MRSA strains and
Available online 31 August 2011 the MSSA ATCC 25923 were tested. Individual strains were diluted in sterile saline to a 0.5
McFarland standard (108 cfu/ml), serially diluted in duplicate to a final concentration of
Keywords: 105 cfu/ml in pooled filter-sterilized marine or fresh water and incubated at 13
C or 20
C in
MRSA die-off the dark. The results of this study found that temperature and salinity are important
Ambient and cold water tempera- factors in MRSA and MSSA survival; the decay rate was w28% higher at 20
C versus 13
C
ture and w34e44% higher in fresh water versus marine water. There was no statistical
Salt and fresh water difference between environmental and clinical MRSA strain survival [P ¼ 0.138]. The study
found that MRSA/MSSA survival was significantly longer in marine water at 13
C typical of
the Pacific Northwest, which may have important implications for recreational beach
visitors in colder climates.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction previously healthy populations have been reported in prison

Over the past decade community-acquired methicillin-resis-
tant Staphylococcus aureus [CA-MRSA] has emerged as a major
cause of disease in the general population with no health care
exposure or known classical risk factors for methicillin-
resistant S. aureus [MRSA] infections. CA-MRSA causes skin
and soft tissue infections, pneumonia and can lead to death
(Bartlett, 2008; King et al., 2006) and the morbidity and
mortality rate per 100,000 people is estimated to be 4.6 and 0.5,
respectively (Klevens et al., 2007). Outbreaks of CA-MRSA in
populations, religious communities, sports teams, and mili-
tary training camps (Coronado et al., 2007; David et al., 2008;
Morrison-Rodriguez et al., 2010; Romano et al., 2006).
Previously, contact with seawater has been associated
with a four-fold increase in risk of infection from S. aureus
(Charoenca and Fujioka, 1995). More recently, MRSA strains
have been isolated and characterized from marine water and
intertidal sand samples from five of ten Pacific Northwest
[PNW] marine beaches (Soge et al., 2009), while S. aureus and
MRSA have also been isolated from Florida (Abdelzaher et al.,

* Corresponding author. Tel.: þ1 206 543 8001; fax: þ1 206 543 3873.
E-mail address: marilynr@u.washington.edu (M.C. Roberts).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.037
5682 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6

2010), Hawaii (Tice et al., 2010) and California (Goodwin and

Table 1 e MRSA strain characterization.
Pobuda, 2009) marine water and sand. The sources of S.
aureus and MRSA contamination in marine environments Isolate Straina collection Source mecAb SCCmec
date typec
have yet to be characterized, though both potential point and
non-point sources have been suggested. Two studies have Clinical
shown that within the first 15 min of water immersion, the ATCC 25923 MSSA 1949 Clinical e NAd
average person sheds 105e106 cfu S. aureus (Elmir et al., 2007; MS361 MRSA 1993 Clinical D II
MS1053 MRSA 1993 Clinical D I
Plano et al., 2011), suggesting that bathers are a potential
VA # 6 MRSA 2009 Clinical D IV
source of both S. aureus and MRSA contamination. MRSA has
Environmental
been isolated in wastewater prior to treatment (Börjesson 9e48 MRSA 2008 Marine water D IV
et al., 2009) while S. aureus has been isolated from urban L1 MRSA 2010 Marine water D IV
runoff from high density residential areas suggesting that G1 MRSA 2010 Fresh water D IV
combined sewer overflows [CSO] during storm events and M1 MRSA 2010 Fresh water D IV
urban runoff may contribute to MRSA and S. aureus loads at Clinical strain ATCC 25923 was obtained from the American Type
marine and fresh water recreational beaches (Selvakumar Culture Collection. Clinical strain VA #6 and the environmental
and Borst, 2006). strains 9-48, LI, GI, and MI were collected in Seattle, WA.
In addition to the source of MRSA and S. aureus, the a Methicillin-susceptible Staphylococcus aureus [MSSA]; Methicillin-
survivability of MRSA and S. aureus in beach environments is resistant Staphylococcus aureus [MRSA].
b mecA gene codes for an altered penicillin binding protein and all
also an important factor that may impact the concentration
MRSA isolates carry this genes.
level found in the beach environments. One study has c staphylococcal cassette chromosome mec elements.
previously investigated MRSA die-off kinetics in marine and d NA: not applicable.
fresh water using community- and hospital-acquired clinical
strains, isolated from the hospital, and found an average of
7.4 days and 3.53 days for a 1-log reduction of MRSA from
marine and fresh water, respectively, at ambient tempera- 2.2. Marine and stream water preparation
ture which we estimate was between 20 and 22
C (Tolba
et al., 2008). The authors found no statistical significance Marine and fresh water samples were collected in sterile 1-L
differences between survival of CA-MRSA and hospital- Nalgene bottles from two Seattle marine beaches on the Puget
acquired methicillin-resistant S. aureus [HA-MRSA] strains. Sound. Fresh water samples were collected from streams
However, there is little research concerning the persistence from surrounding upland drainage areas that traverse
of MRSA and S. aureus at lower temperatures (<20
C), that through the beach to the marine receiving waters. After
are normally found in Pacific Northwest marine beaches collection, the fresh and marine water samples were respec-
during summer. Furthermore, there is no information tively pooled and filter-sterilized through a 0.22 mM filter (Pall
comparing survival of environmental vs and clinical MRSA Corporation, Port Washington, NY, USA). Sterility of each
strains. water type was verified by plating 100 mL onto Brain Heart
There has been limited research on survival of MRSA at the Infusion (BHI) agar (Difco Laboratories, Div. Becton Dickinson
temperature of 13
C which is common at Pacific Northwest & Co., Sparks, MD, USA) in triplicate and incubated at 36.5
C
marine beaches during summer when the beaches are most for 24 h. The pooled fresh was had a salinity of <1 ppt and pH
frequented. The objective of this study was to compare 7.43 and the pooled marine water had a salinity of 29 ppt and
survival of both environmental and clinical MRSA strains and pH 7.77.
a control MSSA ATCC 25923 strain for an assessment of MRSA
survival in the Pacific Northwest beach environment [13
C]
compared to warmer environments [20
C]. 2.3. Fresh and marine water inoculation

Overnight cultures from a Brucella agar (Difco Laboratories,

2. Materials and methods
2.1. Strain description
Seven MRSA strains [four environmental and three clinical]
representing SCCmec type I [1 isolate], II [1 isolate], and IV [5
isolates] and MSSA ATCC 25923 were tested in this study
(Table 1). The environmental strains were isolated from
regional Pacific Northwest beaches in 2008 and 2010, and
the clinical strains were isolated from skin and wound
infections from hospitalized patients in 1945, 1993 and 2009.
The presence of the mecA gene and the staphylococcal
cassette chromosome mec [SCCmec] elements type were
confirmed for each MRSA isolate by PCR (Soge et al., 2009)
(Table 1).
Sparks, MD, USA) supplemented with 5% sterile sheep blood
for each strain were diluted in 0.85% NaCl to a 0.5 McFarland
standard (108 cfu/ml) and serially diluted to approximately
105 cfu/ml in 50 ml of pooled, filter-sterilized marine or fresh
water in a sterile conical tube. Marine and fresh water
microcosms were incubated at ambient temperatures
(20
C  2
C) or 13
C  0.5
C in the dark. The starting
concentration for each water microcosm was determined on
day 0 by plating a 10-fold dilution series onto Brain Heart
Infusion agar [BHI] (Difco Laboratories, Sparks, MD, USA). BHI
plates were incubated at 36.5
C for 24 h and colony counts
(cfu/ml) determined. Daily colony counts were determined at
days 1e3 (13
C) or days 1e5 (20
C) as described for day 0 and
then determined every second day for 14 days (Fig. 1). Prior to
plating for cfu/ml determination each microcosm was
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6 5683

Fig. 1 e For each time point, spread plate counts were averaged and log-transformed. A linear regression was calculated for
each microcosm to assess the time (days) needed for a 1-log reduction. A. Survival of environmental MRSA isolates in
marine and fresh water at 13
C and 20
C. B. Survival of clinical MRSA isolates in marine and fresh water at 13
C and 20
C.

vortexed for approximately 30 s to disrupt adhered cells from
the side of the tube and cell aggregates.
2.4. Statistical analysis
For each microcosm, the average log-transformed spread
plate counts were analyzed by linear regression in GraphPad
Prism 5 (GraphPad Software, Inc, San Diego, CA, USA) to
calculate the time elapse for a 1-log removal [T90]. The results
of the seven MRSA strains were averaged to determine the 1-
log reduction (T90) of the MRSA strains. The significance of
association between water type (marine vs fresh), tempera-
ture and strain classification (environmental vs. clinical) on
survival was modeled by multivariable linear regression in
Stata11 I.C (StataCorp LP, College Station, TX, USA). P < 0.05
was considered statistically significant.
3. Results and discussion
The survival of the eight strains over 14 days is illustrated in
Fig. 1. The cfu/ml plate counts were log-transformed and
plated by day. On average, the seven MRSA strains in marine
water at 20
C and 13
C, it took 7.89  1.62 days and
10.97  3.47 days, respectively, for a 1-log reduction (T90). The
survival of 7.89  1.62 days in marine water at 20
C in the
current study is very similar to 7.4 days previously reported by
Tolba et al. (2008). Survival of all strains was reduced in fresh
water with an average of 2.71  0.58 days and 4.84  1.11 days
for a 1-log removal at 20
C and 13
C, respectively. For the
ATCC 25923 MSSA strain, the T90 at 20
C and 13
C was 3.70
days and 13.29 days, respectively in marine water and 2.65
days and 4.32 days, respectively in fresh water (Fig. 1). The
average survival results of 2.71  0.58, in the current study for
fresh water at 20
C, was in the range of 3.5 days previously
reported by Tolba et al. (2008). Overall the decay rate was
w28% higher at 20
C versus 13
C and w34e44% higher in
fresh water versus marine water. Table 2 describes individual
strain correlation coefficient (R2), T90, die-off rate (k) and
illustrates the variability found between strains. Unfortu-
nately individual results were not available in the Tolba et al.
(2008) for comparison.
A multivariable linear regression model found that
temperature [20
C vs 13
C] and water types [fresh vs marine]
were significantly [P < 0.001] associated with MRSA and MSSA
survival (Table 3). The source of the strain [environmental vs
clinical] was not significantly associated with MRSA survival
[P ¼ 0.138] when adjusted for temperature, time and water
type even though the strains were collected between 1993 and
2010 when the ATCC 25923 MSSA strain was excluded from
the analysis. When the multivariable linear regression model
included the ATCC 25923 MSSA strain, the source of the strain
became significantly associated with survival [P ¼ 0.011],
suggesting that the ATCC 25923 MSSA strain is different from
the other isolates (Table 3). This difference may be because the
ATCC 25923 MSSA strain was isolated in 1949 and passed in
the laboratory while the other isolates were collected more
recently and had limited laboratory passage, or that there are
genetic differences such as the lack of a SCCmec cassette. The
MRSA strains reduction rate, in the absence of organic mate-
rial and light, followed first-order kinetics ( y ¼ mx þ b), which
were previously found (Tolba et al., 2008).
There are several limitations to this study that may over or
underestimate MRSA survival times, because survival was
5684 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6

Table 2 e Regression results, die-off rate (k), and T90 removal time for MRSA strains in fresh and marine water.
Isolate Marine water Fresh water
1 c
Temp ( C)

R 2 a
T90 (days) b
k (day ) R 2 a
T90 (days)b k (day1)c

ATCC 25923 20 0.89 3.70 0.27 0.89 2.65 0.38

13 0.78 13.29 0.08 0.87 4.32 0.23
MS361 20 0.71 4.91 0.20 0.99 2.15 0.47
13 0.75 5.41 0.18 0.81 4.13 0.24
MS1053 20 0.82 7.55 0.13 0.92 2.25 0.44
13 0.76 13.64 0.07 0.93 6.84 0.15
Va#6 20 0.87 8.18 0.12 0.90 2.43 0.41
13 0.67 13.36 0.07 0.94 5.77 0.17
9e48 20 0.80 8.04 0.12 0.98 2.36 0.42
13 0.79 15.43d 0.06 0.96 4.57 0.22
L1 20 0.80 9.90 0.10 0.87 2.92 0.34
13 0.83 8.74 0.11 0.94 3.97 0.25
G1 20 0.82 9.38 0.11 0.86 3.78 0.26
13 0.82 11.11 0.09 0.96 4.89 0.20
M1 20 0.76 7.30 0.14 0.89 3.06 0.33
13 0.90 9.07 0.11 0.98 3.73 0.27
T90 Average  SDe 20 7.89  1.62 days 2.71  0.58 days
13 10.97  3.47 days 4.84  1.11 days

a Regression correlation coefficient.

b Time needed for a 1-log reduction.
c Die-off rate: k ¼ 1/T90
d Extrapolated beyond data range.
e Does not include the ATCC 25923 MSSA strain.

analyzed only in sterile microcosms in the dark. Intrinsic
variables such as survival mechanisms and extrinsic factors
such as nutritional sources, UV light, and antagonistic
microflora were not examined. The study by Fuijoka and
Unutoa (2006) showed that UV light and supplemented
organic material such as sewage reduced or prolonged
survival time, respectively, compared to survival time
measured in the absence of these variables while without
supplemented organic material, the proportion of culturable
S. aureus populations declines relative to the total population
in a seawater microcosm over time (Masmoudi et al., 2010).
The potential role of algae wrack in the survival of MRSA and
S. aureus is currently unknown, however data from Great
Lakes studies show algal mats as reservoirs for Escherichia coli,
Enterococcus, Shigella, Campylobacter, and Salmonella (Ishii et al.,
2006; Olapade et al., 2006; Verhougstraete et al., 2010). More
recently marine wrack was shown to promote persistence of
fecal indicator bacteria in laboratory marine microcosms
(Imamura et al., 2011). Sand has also been identified as
a protective from the bactericidal effects of UV light and dry
sand near the high-tide line has been found to be a significant
reservoir of pathogenic bacteria (Abdelzaher et al., 2010;
Beversdorf et al., 2007; Bonilla et al., 2007).
The current results are the first to examine MRSA survival
in 13
C marine and fresh water. This is essential because the
stability of MRSA in recreational beach environments is
important for an accurate exposure assessment and ulti-
mately risk characterization of human MRSA infection. This

Table 3 e Multivariable linear regression results for predictor variables water type, temperature and isolate type on MRSA
survival.
ATCC 25923 Excluded ATCC 25923 Included

95% CIb 95% CIb

ba Lower Upper P-valuec ba Lower Upper P-valuec

Waterd 0.88 0.70 1.07 <0.001 e0.47 0.75 1.10 <0.001

Tempe 0.41 0.59 0.23 <0.001 0.47 0.64 0.30 <0.001
Typef 0.14 0.36 0.05 0.138 0.22 0.39 0.05 0.011

a Coefficient.
b CI: confidence interval.
c P < 0.05 considered significant.
d Water: marine vs fresh.
e Temperature: 13
C vs 20
C.
f Type: environmental vs clinical; Model correlation coefficient including and excluding ATCC 25923 MSSA strain: R2 ¼ 0.57.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6
study found that there was an increased survival of MRSA and
MSSA strains in marine vs fresh water. Survival was longer at
13
C vs 20
C suggesting that there may be differences in
survival of MRSA at recreational beaches in the Pacific
Northwest and those of subtropical and tropical climates. All
four of the environmental strains were isolated in the Seattle
area and it is unknown whether environmental MRSA strains
isolated in warmer water (25
C) would respond the same as
the strains in the current study.
4. Conclusions
 The study found that there was an increased survival of
MRSA/MSSA strains in marine vs fresh water and survival
was higher at 13
C vs 20
C in fresh and marine waters
suggesting that there may be major differences between S.
aureus survival at recreational beaches in the Pacific
Northwest vs those of subtropical and tropical climates.
 Temperature [13
C vs 20
C] and salinity [marine water vs
fresh water] were significantly [P < 0.001] associated with
MRSA/MSSA survival.
 MRSA survival rate was not significantly associated with the
isolate source [environmental vs clinical].
Acknowledgements
The work was funded in part by NIH National Institute of
Environmental Health and Science (NIEHS) 5 R25 ES016150-03
REVISED. Ms. Bonilla was supported by an NIH Minority
Access to Research Careers (MARC) grant (# 5T34GM07821).
references
Abdelzaher, A.M., Wright, M.E., Ortega, C., Solo-Gabriele, H.M.,
Miller, G., Elmir, S., Newman, X., Shih, P., Bonilla, J.A.,
Bonilla, T.D., Palmer, C.J., Scott, T., Lukasik, J., Harwood, V.J.,
McQuaig, S., Sinigalliano, C., Gidley, M., Plano, L.R., Zhu, X.,
Wang, J.D., Fleming, L.E., 2010. Presence of pathogens and
indicator microbes at a non-point source subtropical
recreational marine beach. Appl. Environ. Microbiol. 76 (3),
724e732.
Bartlett, J.G., 2008. Methicillin-resistant Staphylococcus aureus
infections. Top. HIV Med. 16 (5), 151e155.
Beversdorf, L.J., Bornstein-Forst, S.M., McLellan, S.L., 2007. The
potential for beach sand to serve as a reservoir for Escherichia
coli and the physical influences on cell die-off. J. Appl.
Microbiol. 102 (5), 1372e1381.
Börjesson, S., Matussek, A., Melin, S., Löfgren, S., Lindgren, P.E.,
2009. Methicillin-resistant Staphylococcus aureus (MRSA) in
municipal wastewater: an uncharted threat? J. Appl.
Microbiol. 108 (4), 1244e1251.
Bonilla, T.D., Nowosielski, K., Cuvelier, M., Hartz, A., Green, M.,
Nwadiuto, E., McCorquodale, D.S., Fleisher, J.M., Rogerson, A.,
2007. Prevalence and distribution of fecal indicator organisms
in South Florida beach sand and preliminary assessment of
health effects associated with beach sand exposure. Mar.
Pollut. Bull. 54 (9), 1472e1482.
5685
Charoenca, N., Fujioka, R.S., 1995. Assessment of Staphylococcus
bacteria in Hawaii recreational waters. Water. Sci. Technol. 31
(5e6), 11e17.
Coronado, F., Nicholas, J.A., Wallace, B.J., Kohlerschmidt, D.J.,
Musser, K., Schoonmaker-Bopp, D.J., Zimmerman, S.M., Boller, A.
R., Jernigan, D.B., Kacica, M.A., 2007. Community-acquired
methicillin-resistant Staphylococcus aureus skin infections in
a religious community. Epidemiol. Infect. 135 (3), 492e501.
David, M.Z., Mennella, C., Mansour, M., Boyle-Vavra, S., Daum, R.S.,
2008. Predominance of methicillin-resistant Staphylococcus
aureus among pathogens causing skin and soft tissue infections
in a large urban jail: risk factors and recurrence rates. J. Clin.
Microbiol. 46 (10), 3222e3227.
Elmir, S.M., Wright, M.E., Abdelzaher, A., Solo-Gabriele, H.M.,
Fleming, L.E., Miller, G., Rybolowik, M., Peter Shih, M.T.,
Pilla, S.P., Cooper, J.A., Quaye, E.A., 2007. Quantitative
evaluation of bacteria released by bathers in marine water.
Water. Res. 41 (18), 3e10.
Fuijoka, R.S., Unutoa, T.M., 2006. Comparative stability and
growth requirements of S. aureus and faecal indicator bacteria
in seawater. Water. Sci. Technol. 54 (3), 169e175.
Goodwin, K.D., Pobuda, M., 2009. Performance of CHROMagar
Staph aureus and CHROMagar MRSA for detection of
Staphylococcus aureus in seawater and beach sand e
Comparison of culture, agglutination, and molecular analysis.
Water. Res. 43, 4802e4811.
Imamura, G.J., Thompson, R.S., Boehm, A.B., Jay, J.A., 2011. Wrack
promotes the persistence of fecal indicator bacteria in marine
sands and seawater. FEMS. Microbiol. Ecol. 77, 40e49.
Ishii, S., Yan, T., Shively, D.A., Byappanahalli, M.N., Whitman, R.L.,
2006. Cladorphora (Chlorophyta) spp. harbor human bacterial
pathogens in nearshore water of Lake Michigan. Appl. Environ.
Microbiol. 72 (7), 4545e4553.
King, M.D., Humphrey, B.J., Wang, Y.F., Kourbatova, E.V., Ray, S.
M., Blumberg, H.M., 2006. Emergence of community-acquired
methicillin-resistant Staphylococcus aureus USA 300 clone as
the predominant cause of skin and soft-tissue infections. Ann.
Intern. Med. 144 (5), 309e317.
Klevens, R.M., Morrison, M.A., Nadle, J., Petit, S., Gersham, K.,
Ray, S., Harrison, L.H., Lynfield, R., Dumyati, G., Townes, J.M.,
Craig, A.S., Zell, E.R., Fosheim, G.E., McDougal, L.K., Carey, R.B.,
Fridken, S.K., 2007. Invasive methicillin-resistant Staphylococcus
aureus infections in the United States. JAMA 298 (15),
1763e1767.
Masmoudi, S., Denis, M., Maalej, S., 2010. Inactivation of the gene
katA and sodA affects the transient entry into the viable but non-
culturable response of Staphylococcus aureus in natural seawater
at low temperatures. Mar. Pollut. Bull. 60 (12), 2209e2214.
Morrison-Rodriguez, S.M., Pacha, L.A., Patrick, J.E., Jordan, N.N.,
2010. Community-associated methicillin-resistant
Staphylococcus aureus infections at an Army training
installation. Epidemiol. Infect. 138 (5), 721e729.
Olapade, O.A., Depas, M.M., Jensen, E.T., McLellan, S.L., 2006.
Microbial communities and fecal indicator bacteria associated
with Cladophora mats on beach sites along Lake Michigan
shores. Appl. Environ. Microbiol. 72 (3), 1932e1938.
Plano, L.R.W., Garza, A.C., Shibata, T., Elmir, S.M., Kish, J.,
Sinigalliano, C.D., Gidley, M.L., Miller, G., Withum, K., Fleming, L.
E., Solo-Gabriele, H.M., 2011. Shedding of Staphylococcus aureus
and methicillin-resistant Staphylococcus aureus from adult and
pediatric bathers in marine waters. BMC Microbiol 11 (1), 5.
Romano, R., Lu, D., Holtom, P., 2006. Outbreak of community-
acquired methicillin-resistant Staphylococcus aureus skin
infections among a collegiate football team. J. Athl Train. 41
(2), 141e145.
Selvakumar, A., Borst, M., 2006. Variation of microorganism
concentrations in urban runoff with land use and seasons. J.
Water Health 4 (1), 109e124.
5686 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 1 e5 6 8 6
Soge, O.O., Meschke, J.S., No, D.B., Roberts, M.C., 2009.
Characterization of methicillin-resistant Staphylococcus aureus
and methicillin-resistant coagulase-negative Staphylococcus
spp. isolated from US west coast public marine beaches. J.
Antimicrob. Chemother. 64 (6), 1148e1155.
Tice, A.D., Pombo, D., Hui, J., Kurano, M., Bankowski, M.J.,
Seifried, S.E., 2010. Quantification of Staphylococcus aureus
in seawater using CHROMagar SA. Hawaii. Med. J. 69 (1),
8e12.
Tolba, O., Loughrey, A., Goldsmith, C.E., Millar, B.C., Rooney, P.J.,
Moore, J.E., 2008. Survival of epidemic strains of healthcare (HA-
MRSA) and community-associated (CA-MRSA) methicillin-
resistant Staphylococcus aureus (MRSA) in river-, sea- and
swimming pool water. Int. J. Hyg. Environ. Health. 211 (3e4),
398e402.
Verhougstraete, M.P., Byappanahalli, M.N., Rose, J.B., Whitman, R.
L., 2010. Cladophora in the Great Lakes: impacts on beach water
quality and human health. Water. Sci. Technol. 62 (1), 68e76.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

NDMA formation kinetics from three pharmaceuticals in four

water matrices

Ruqiao Shen*, Susan A. Andrews

Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, Ontario, Canada M5S 1A4

article info abstract

Article history: N, N-nitrosodimethylamine (NDMA) is an emerging disinfection by-product (DBP) that has
Received 17 May 2011 been widely detected in many drinking water systems and commonly associated with the
Received in revised form chloramine disinfection process. Some amine-based pharmaceuticals have been demon-
11 August 2011 strated to form NDMA during chloramination, but studies regarding the reaction kinetics
Accepted 20 August 2011 are largely lacking. This study investigates the NDMA formation kinetics from ranitidine,
Available online 27 August 2011 chlorphenamine, and doxylamine under practical chloramine disinfection conditions. The
formation profile was monitored in both lab-grade water and real water matrices, and
Keywords: a statistical model is proposed to describe and predict the NDMA formation from selected
NDMA pharmaceuticals in various water matrices. The results indicate the significant impact of
Ranitidine water matrix components and reaction time on the NDMA formation from selected
Chlorphenamine pharmaceuticals, and provide fresh insights on the estimation of ultimate NDMA forma-
Doxylamine tion potential from pharmaceutical precursors.
Chloramination ª 2011 Elsevier Ltd. All rights reserved.
Kinetics

1. Introduction
N, N-nitrosodimethylamine (NDMA) is a member of N-nitro-
samines found in food, beer, cured meats, rubber products,
tobacco smoke and more recently, drinking water. There is
growing concern regarding the health effects associated with
exposure to nitrosamines because of their potential carcino-
genicity (EPA IRIS, 1993). The occurrence of NDMA in finished
drinking water has been commonly associated with the
application of chloramine as a final disinfectant. Recent
surveys in Canada and the U.S. have revealed occurrence of
NDMA in many chloraminated drinking water systems with
concentration up to 630 ng/L (Blute et al., 2010; Charrois et al.,
2007). The widespread detections of NDMA in source water
and treated drinking water have spurred local governments
and agencies to take actions. The Ontario Ministry of the
Environment (MOE) has established a maximum acceptable
level of 9 ng/L for NDMA (MOE, 2003), and the California
Department of Health Services has implemented an NDMA
notification level of 10 ng/L (OEHHA, 2006). The USEPA has
placed NDMA together with other four nitrosamines on the
latest drinking water contaminant candidate list 3 (CCL3)
(USEPA, 2009). More recently, Health Canada has proposed
a maximum acceptable concentration for NDMA of 40 ng/L in
drinking water (Health Canada, 2010).
A number of research efforts have been invested in iden-
tifying potential NDMA precursors relevant to drinking water.
Theoretically, any amine compounds containing dimethyl-
amine (DMA) groups may react with chloramine to form
NDMA. Typical precursors found in source water include
some tertiary and quaternary amines (Kemper et al., 2010;
Mitch et al., 2003; Mitch and Schreiber, 2008), and fractions
of natural organic matter (NOM) (Chen and Valentine, 2007;
Dotson et al., 2007; Gerecke and Sedlak, 2003; Mitch and

* Corresponding author. Tel.: þ1 4169783141; fax: þ1 4169783674.

E-mail addresses: shenruqiao@yahoo.com.cn (R. Shen), sandrews@civ.utoronto.ca (S.A. Andrews).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.034
5688 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4
Sedlak, 2004). Some chemicals used in water treatment
processes may also contribute to NDMA formation, such as
certain amine-based polymers and anion exchange resins
(Kohut and Andrews, 2003; Mitch and Sedlak, 2004; Najm and
Trussell, 2001; Wilczak et al., 2003). More recently, dime-
thylsulfamide, a degradation product of the fungicide toly-
fluanide, was newly identified as an NDMA precursor during
ozonation (Schmidt and Brauch, 2008). Pharmaceuticals first
came to attention as potential NDMA precursors when rani-
tidine was demonstrated to convert into NDMA at high
conversion rate during chloramination (Sacher et al., 2008;
Schmidt et al., 2006). Krasner (2009) has suggested that
amine-based pharmaceuticals and their breakdown products
might be part of the NDMA precursor pool. A recent study by
the authors has demonstrated the formation of nitrosamines
from twenty amine-based pharmaceuticals and personal care
products (PPCPs) upon chloramine disinfection (Shen and
Andrews, 2011).
Up to now, studies on NDMA formation via pharmaceuti-
cals have been mostly conducted using lab-grade water.
Specifically, data regarding the reaction kinetics in real water
matrices are largely lacking. Krasner et al. (2010) have inves-
tigated the NDMA formation over time from ranitidine under
different pH and temperature, but only conducted the exper-
iments in deionized water. Due to the lack of knowledge about
the reactivity and chemistry, it is difficult to predict NDMA
formation from pharmaceuticals using traditional kinetic
models. In the literature, some kinetic models have been
developed for the prediction of NDMA formation from DMA
(Choi and Valentine, 2002; Kim and Clevenger, 2007) and from
NOM (Chen and Valentine, 2006); however, these models use
comparable concentrations of precursors and chloramines,
and thus might not apply to pharmaceuticals which are
usually present at trace levels in the source water and are at
much lower concentrations relative to chloramine concen-
trations in real samples.
This study demonstrates the NDMA formation kinetics
from three amine-based pharmaceuticals in four different
water matrices, and proposes a statistical model to describe
and predict the NDMA molar conversion from selected phar-
maceuticals during chloramination.
2. Materials and methods
Three pharmaceuticals were selected to determine their
NDMA formation potential (NDMA-FP) over time, including
chlorphenamine, doxylamine, and ranitidine (Fig. 1). Stock
solutions of pharmaceuticals were prepared in methanol and
stored at 4  C until use. NDMA (reagent grade) and deuterated
NDMA (d6-NDMA, 98 atom %D) were used as standard and
internal standard, respectively. All chemicals were purchased
from SigmaeAldrich Canada (Oakville, Ontario).
Experiments were carried out under the Simulated Distri-
bution System (SDS) conditions (pH ¼ 7.0  0.1; 21  C; Cl2: N
mass ratio ¼ 4.2:1; chloramine dosage ¼ 2.5  0.2 mg/L after
satisfying 24 hr chloramine demand). Further details con-
cerning the experimental procedure and NDMA analysis have
been described in Shen and Andrews (2011). NDMA formation
from each pharmaceutical was monitored for up to 144 hr
Fig. 1 e Structures of selected pharmaceuticals.
based on preliminary results, depending on the compound
and water matrix. At each time point samples were prepared
in duplicate, together with one blank control to account for
the potential background interference. Error bars in all the
kinetic graphs represent the maximum and minimum values
in the formation potential tests under the same reaction
conditions (n ¼ 2).
Experiments were conducted in four water matrices; the
water sources and basic water quality parameters are
summarized in Table 1. Lake and river water samples were
taken from the influent of two drinking water treatment
plants in June and October, 2010, respectively. The pH was
determined using a pH meter (Model 8015, VWR Scientific Inc.,
Mississauga, Ontario). Alkalinity was measured based on an
end-point titration, according to Standard Method 2320B
(APHA, 2005). The total organic carbon (TOC) was analyzed
with an Aurora 1030 TOC analyzer (O.I. Analytical, College
Station, Texas). The ultraviolet absorbance at 254 nm (UV254)
was determined by a CE3055 Reflectance Spectrophotometer
(Cecil Instruments Ltd., Cambridge, England). The specific UV
absorbance (SUVA) is calculated by normalizing the UV254 to
the TOC.
3. Results and discussion
3.1. Formation kinetics in MQ water
Kinetic experiments in MQ (Milli-Q, Ultra Pure Water System,
MilliPore, Etobicoke, Ontario) water were conducted for rani-
tidine, chlorphenamine, and doxylamine at two concentra-
tion levels (5 and 25 nM), as shown in Fig. 2. The markers in the
figure are the measured NDMA molar conversion values, and
the lines are model-estimated results. Details about the model
development and estimation will be discussed in Section 3.3.
NDMA formation via the three pharmaceuticals followed
similar pattern over time. Generally, an initial lag period was
observed, followed by a fast increase in NDMA concentration;
the molar conversion then gradually leveled off and eventu-
ally reached a plateau (maximum molar conversion). More-
over, the formation kinetic behavior was observed to be
relatively independent of the initial pharmaceutical concen-
tration, except that NDMA formation from doxylamine in MQ
water showed a more significant difference after 24 hr than
did the other pharmaceuticals. Given the large excess of
chloramine relative to the pharmaceuticals (mg/L vs. lower mg/
L), availability of chloramine was not a limiting factor at the
concentration range of pharmaceuticals tested. These results
also support observations made in an earlier study where
the NDMA molar conversion at 24 hr for 20 selected
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4
Table 1 e Water matrix source and quality.
Matrix Source
Milli-Q (MQ) Ultra Pure Water System
(MilliPore, Etobicoke, Ontario)
Tap water (TAP) Toronto, Ontario
Lake water (LW) Lake Ontario, Ajax, Ontario
River water (RW) Otonabee River, Peterborough, Ontario
pharmaceuticals was found to be independent of their initial
concentrations (Shen and Andrews, 2011).
3.2. Formation kinetics in different water matrices
Kinetic experiments were also performed using real water
samples dosed with selected pharmaceuticals. The same
shape of NDMA formation curve was observed in real water
matrices as in MQ water except that the initial lag phase was
longer, especially for tests performed using lake or river water
(Fig. 3). Similarly to Fig. 2, the markers are the measured
NDMA molar conversion values, and the lines are the model-
estimated results that will be discussed in detail in 3.3.
Although the NDMA formation kinetics was shown to be
unique to each water matrix tested, the kinetic behavior was
relatively independent of the initial pharmaceutical concen-
tration within a given water matrix, further confirming that
observation in MQ water.
The different NDMA formation profiles were likely influ-
enced by the water matrix components, rather than by added
reagents, since the pH of the water samples was controlled
with a phosphate buffer and the same chloramine dosage was
applied to all samples. Both bromide and NOM have been
shown to influence NDMA formation, with bromide being
reported to either catalyze NDMA formation (Mitch et al., 2003;
Valentine et al., 2005) or have an inhibitory effect (Chen et al.,
2010). However, bromide levels in the water sources that were
tested are typically much lower than those for studies that
have reported these effects, so the differences in the observed
formation profiles were thought to be due to some aspect of
the NOM. Since bromide is in higher concentration and so may
be more of a concern in coastal waters due to saltwater
intrusion, the potential impact from bromide was considered
Fig. 2 e NDMA molar conversion over time for ranitidine, chlorphenamine, and doxylamine in MQ water (SDS conditions).
5689
pH Alkalinity TOC UV254 SUVA
(mg/L) (mg/L) (cm1) (L,m/mg)
7.5  0.1 1.8  0.3 0.0 0.000 0.000
7.1  0.1 88.5  2.8 2.1  0.1 0.021  0.001 1.01  0.05
8.0  0.1 94.6  1.5 2.3  0.2 0.024  0.002 1.08  0.12
7.8  0.1 86.5  1.2 6.2  0.5 0.143  0.002 2.32  0.17
to be outside the scope of the present tests but would be of
interest for future study.
NOM may affect the NDMA formation in two ways. The
influence of NOM’s competition for chloramine was consid-
ered to be minimal due to the small observed chloramine
decay (data not shown) and the large excess of chloramine
relative to the pharmaceuticals (mg/L vs. lower mg/L) at the
end of the kinetic experiment. On the other hand, NOM may
interact with the pharmaceuticals and then inhibit the reac-
tion to form NDMA, and it is these interactions that were
thought to better explain the observed results. NOM compo-
nents can be at least partially described by the samples’ TOC
and SUVA values. It was observed that water with higher TOC
and SUVA levels tended to have a longer initial lag phase; all
three pharmaceuticals exhibited their longest initial lag
period in river water samples. However, while tap and lake
water samples had similar TOC and SUVA values, ranitidine
showed a longer initial lag phase in lake water samples. This
suggests that some specific NOM fractions or moieties might
be more relevant than would be indicated by simple bulk
measurements of water quality, such as TOC and SUVA.
Previous studies have demonstrated that aromatic amines
undergo reversible covalent binding with carbonyls and
quinones in soil humic substances in the environment (Parris,
1980; Thorn et al., 1996; Weber et al., 1996). Therefore, it is
possible that certain fractions or functional groups in NOM
may interact with these amine-based pharmaceuticals and
thus hinder their initial contact with chloramine species. As
the binding is reversible and chloramine is in large excess,
eventually the NDMA conversion from pharmaceuticals can
still reach the maximum level given enough reaction time.
Currently, although no direct spectroscopic evidence exists
for the NOM-pharmaceutical binding in aqueous phase, this
5690 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4

Fig. 3 e NDMA molar conversion over time for (a) ranitidine; (b) chlorphenamine; and (c) doxylamine in different water
matrices (SDS conditions).

theory is indirectly supported by some literature investigating
the removal of pharmaceuticals during coagulation/floccula-
tion process, where the removal of pharmaceuticals was likely
due to the sorption onto particulate organic matter and co-
removed through the settling process (Ballard and Mackay,
2005; Stackelberg et al., 2007; Vieno et al., 2006; Westerhoff
et al., 2005). Stackelberg et al. (2007) also detected the target
pharmaceuticals in the dried solids of settled sludge. In
addition, De Ridder et al. (2011) observed enhanced removal of
some positively charged pharmaceuticals using granular
activated carbon preloaded with NOM. They attributed the
enhancement to the electrostatic attraction since the surface
of NOM is usually negatively charged due to abundant
carboxyl groups. In the current study, the selected amine
pharmaceuticals are positively charged at neutral pH, there-
fore the possible electrostatic attraction may also lead to the
formation of NOM-pharmaceutical complexes. Future studies
are needed to further investigate the role of NOM components
in the conversion of pharmaceuticals into NDMA, and alter-
native methods for the characterization of NOM components
will be helpful, such as the application of size-exclusion
chromatography with organic carbon detection (Huber et al.,
2011).
3.3. Kinetic model
The NDMA formation curves for the pharmaceuticals in this
study have a sigmoidal shape that resembles the typical shape

Table 2 e Model parameter estimation and model verification.

Compound Concentration Matrix Parameter Estimation Model Verification
a a 1 a 2
q Lag (hr) k (hr ) R Model-predicted Measured
conversion @ 24 h conversion @ 24 hb

Ranitidine 5 nM MQ 0.912 (0.045) 6.3 (0.5) 0.225 (0.043) 0.992 91.2% 85.2% (0.8%)
TAP 0.902 (0.045) 6.7 (0.8) 0.169 (0.045) 0.986 90.1% 83.4% (8.1%)
LW 0.729 (0.032) 13.3 (0.7) 0.251 (0.057) 0.997 72.7% 64.1% (3.6%)
RW 0.822 (0.056) 20.8 (1.8) 0.086 (0.026) 0.984 53.6% 51.4% (4.9%)
25 nM MQ 0.906 (0.045) 4.6 (0.5) 0.313 (0.084) 0.991 90.6% 82.7% (2.4%)
TAP 0.847 (0.039) 6.5 (0.7) 0.177 (0.043) 0.988 84.6% 88.4% (5.9%)
LW 0.769 (0.018) 13.1 (0.4) 0.252 (0.028) 0.999 76.8% 70.1% (4.8%)
RW 0.841 (0.044) 21.9 (1.4) 0.083 (0.021) 0.990 50.4% 43.2% (7.1%)
Chlorphenamine 5 nM MQ 0.027 (0.003) 7.0 (1.3) 0.143 (0.050) 0.974 2.7% 2.9% (0.2%)
TAP 0.023 (0.003) 10.8 (3.0) 0.065 (0.024) 0.953 2.0% 2.0% (0.8%)
RW 0.037 (0.002) 38.4 (3.4) 0.039 (0.007) 0.990 0.8% 1.0% (0.1%)
25 nM MQ 0.033 (0.002) 8.7 (0.6) 0.234 (0.079) 0.994 3.3% 1.8% (0.1%)
TAP 0.030 (0.002) 19.4 (1.8) 0.070 (0.015) 0.988 2.0% 1.5% (0.1%)
RW 0.043 (0.002) 39.9 (1.9) 0.056 (0.009) 0.996 0.5% 0.5% (0.02%)
Doxylamine 5 nM MQ 0.062 (0.005) 16.6 (2.2) 0.068 (0.017) 0.975 4.7% 3.8% (0.1%)
TAP 0.068 (0.006) 48.8 (5.3) 0.027 (0.005) 0.993 1.2% 2.5% (0.2%)
RW 0.059 (0.003) 67.0 (4.3) 0.026 (0.006) 0.990 0.4% 1.1% (0.03%)
25 nM MQ 0.106 (0.007) 22.0 (2.0) 0.069 (0.018) 0.985 6.1% 4.2% (0.1%)
TAP 0.092 (0.006) 46.1 (4.1) 0.030 (0.005) 0.992 1.6% 3.2% (0.3%)
RW 0.060 (0.002) 51.1 (1.8) 0.051 (0.011) 0.996 0.2% 0.5% (0.04%)

a Numbers in the bracket represent the 95% confidence interval of each model parameter.
b Numbers in the bracket represent standard deviation from multiple tests (n ¼ 3).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4
of doseeresponse curves. A standard doseeresponse curve
can be defined by a four-parameter logistic function,
ab
y¼þ
1 þ 10c$ðdxÞ
where y is the response caused by certain dose of pharma-
ceuticals (x); a and b are the maximum and baseline response,
respectively; c is the slope of the curve; and d is the dose which
provokes a response halfway between the baseline and
maximum (Motulsky and Christopoulos, 2003). Accordingly,
the following model was proposed to describe the reaction
kinetics for NDMA formation from selected pharmaceuticals,
q
Y¼
1 þ 10k$ðLagtÞ
where Y is the NDMA molar conversion at given reaction time
(t); q is the ultimate NDMA molar conversion, i.e., the maximum
molar conversion obtained at the plateau during kinetic testing;
k is the pseudo-first order reaction rate constant; Lag is the time
required to achieve 50% of the ultimate molar conversion, and
thus is associated with the length of initial lag phase observed.
Comparing this model with the four-parameter logistic func-
tion, the parameter b was set to zero because any possible
NDMA in the background and any potential NDMA formed from
Fig. 4 e Linear correlation between (a) Lag and TOC; (b) Lag and SUVA; (c) k and TOC; (d) k and SUVA for three
pharmaceuticals (SDS conditions; [Pharmaceutical] [ 5 and 25 nM; error bars represent the 95% confidence interval for
estimated model parameters).
5691
the matrix components were accounted for by the blank control
samples. It is noted that the proposed model does not pass
through the point of (0, 0), although, once background NDMA
has been subtracted, there should be zero molar conversion at
the beginning. Since the doseeresponse model is based on log
(drug dose), it is always positive on the x-axis, and thus does not
go through the point of (0, 0). Therefore, the proposed model
was arbitrarily set to be:
8
< 0 ðt ¼ 0Þ
Y¼ q
: ðt > 0Þ
1 þ 10k$ðLagtÞ
The formation curve was fitted using GraphPad Prism 5
software, and the estimated model parameters for each
compound in different matrices are summarized in the
Parameter Estimation section of Table 2.
The proposed model fit the experimental data very well,
with correlation coefficients (R2) higher than 0.95 in all cases,
and predicted accurately all three phases of the NDMA
formation curve, as shown previously in Figs. 2 and 3. It is
worth noting that the model requires data capturing all three
phases of the NDMA formation curve in order to acquire
reliable model parameters. For datasets lacking the plateau
data, the model will arbitrarily assume the last point as the
5692 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4

plateau. In this study, the NDMA formation curve for raniti-
dine in LW samples did not achieve the plateau within the
24hr of the experiment (Fig. 3a). Therefore, the calculated q
may underestimate the ultimate NDMA molar conversion for
ranitidine in LW samples.
The estimated model parameters each well reflected the
different aspects of the NDMA formation profiles that were
observed in the different water matrices. Generally, the
matrix had a minor impact on the ultimate NDMA molar
conversion for ranitidine (q ¼ 84.1  6.7%), chlorphenamine
(q ¼ 3.2  0.7%), and doxylamine (q ¼ 7.5  2.0%). Instead, the
matrix components had a more profound impact on the initial
lag phase (Lag) and the pseudo-first order rate constant (k). As
summarized in Fig. 4, the Lag value is positively correlated
with both TOC and SUVA values for all three pharmaceuticals;
k value is negatively correlated with TOC and SUVA values for
ranitidine and chlorphenamine, but not well related for dox-
ylamine. The estimated model parameters and these corre-
lations support the theory that water matrix components can
affect NDMA formation from selected pharmaceuticals by
inhibiting the initial reaction with chloramine and/or slowing
down subsequent reactions.
Chen and Westerhoff (2010) recently found NDMA-FP very
difficult to predict based upon bulk water quality measurements
such as DOC or UVA254. Results from the current study have
indicated that knowledge of the reaction kinetics is essential in
the prediction of NDMA formation. While typical water quality
measurements like TOC and SUVA can have significant impact
on the reaction kinetics, they are not directly associated with the
ultimate NDMA molar conversion, and thus are not appropriate
for directly predicting NDMA-FP using empirical models.
Moreover, knowledge of the reaction time employed is specifi-
cally crucial for compounds that react slowly with chloramine,
such as doxylamine. Recently, more and more utilities have
shown interest in conducting NDMA-FP tests. Because there is
no standard protocol at the moment, many are considering
adopting typical disinfection by-products formation potential
tests (Summers et al., 1996) and have applied a 24 hr incubation
time from a practical viewpoint; however, the NDMA formed
after 24 hr from some compounds may only represent a small
portion of their ultimate formation potential, especially in real
water matrices where the initial reaction could be significantly
inhibited. For example, the NDMA molar conversion at 24 hr for
doxylamine in TAP and RW samples only accounted for less
than 20% of its ultimate NDMA molar conversion. The results
have suggested that typical bench-scale NDMA-FP tests may
underestimate the ultimate NDMA-FP for some precursors. For
water systems with higher water age, prolonged NDMA forma-
tion in the outreaches of the distribution system might be
a potential risk and should be taken into consideration.

Fig. 5 e Linear correlation between the model-predicted and the independently measured NDMA molar conversion at 24 h
for (a) ranitidine; (b) chlorphenamine; (c) doxylamine; and (d) three compounds together (SDS conditions; data from four
matrices (MQ, TAP, LW, and RW) and two concentration levels ([Pharmaceutical] [ 5 and 25 nM) were included; the slope
was reported as “the best fit value ± the standard error” at 95% confidence level).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4
The model was verified by comparing the 24hr NDMA-FP
predicted using the estimated model with that measured
from independent 24 hr formation potential tests, as
summarized in the last two columns of Table 2 (Model Verifi-
cation). Linear regression was applied between the measured
and predicted NDMA molar conversion for each compound
individually and for three compounds all together (Fig. 5); the
Student’s t-test was then conducted to determine whether the
slope of each regression line differed significantly from 1.0.
This goodness of fit test has suggested that there is significant
correlation between the measured and model-predicted
molar conversion (F-test, 95% confidence level), except for
chlorphenamine, although even this correlation was deter-
mined to be significant at 90% confidence level. The t-test has
indicated that the slope of the regression line for ranitidine
and chlorphenamine did not differ from 1.0 ( p-value of 99.9%
and 15.9%, respectively; 95% confidence level); yet the slope
was determined to be smaller than 1.0 for doxylamine and
three compounds altogether ( p-value of 3.3% and 0.1%,
respectively; 95% confidence level). In general, however, the
model-predicted molar conversion was within the 95% confi-
dence interval of the measured value.
4. Conclusions
NDMA formation kinetics from ranitidine, chlorphenamine,
and doxylamine during chloramination was determined in
four water matrices. The NDMA conversion over time fol-
lowed a general three-phase formation curve: an initial lag
phase was observed, followed by a fast increase in NDMA
formation, and eventually a plateau was reached that repre-
sented the ultimate NDMA molar conversion. The NDMA
formation profile was relatively independent of the initial
pharmaceutical concentration in the same matrix. Water
matrix components affected the NDMA conversion rates,
most likely by inhibiting their initial contact with chloramine
and slowing down the reaction, while they had less impact on
the ultimate NDMA molar conversion.
A three-parameter kinetic model was proposed to describe
the NDMA formation over time during chloramination. The
model accurately reflected all the three significant character-
istics of the NDMA formation curve, and was able to predict the
NDMA molar conversion from the selected pharmaceuticals to
within the 95% confidence interval of the measured values.
The model needs to be further verified using different potential
precursors, water matrices, and reaction conditions. Bulk
water quality measurements such as TOC and SUVA were
found to correlate better with model parameters Lag and k than
with the ultimate NDMA molar conversion (q), indicating
interactions between the pharmaceuticals and NOM that
might impact NDMA formation are not limited to those based
on the general organic character or aromatic nature of either
substance. Alternative methods are needed to better charac-
terize the matrix components in order to further investigate
their impact on NDMA formation from pharmaceuticals.
Knowledge about the formation kinetics is essential in the
prediction of NDMA formation from pharmaceuticals. Short-
term NDMA-FP tests (24 h), although practical, may underes-
timate the contribution of certain slow-reacting precursors.
5693
Thus, prolonged tests (perhaps >4 days for substances exam-
ined in this study) are required to determine the ultimate NDMA-
FP, especially in distribution systems with long water age.
Acknowledgment
This research was supported by the Canadian Water Network,
the Natural Sciences and Engineering Research Council of
Canada, and the Ontario Research Fund. Special thanks are
dedicated to Richard Jones and John Armour in the water
treatment plants for their assistance in water sampling.
references
American Public Health Association (APHA), American Water
Works Association (AWWA), Water Environment Federation
(WEF), 2005. In: Eaton, A.D., Clesceri, L.S., Rice, E.W.,
Greenberg, A.E. (Eds.), Standard Methods for the Examination
of Water and Wastewater, 21st ed. Washington, DC.
Ballard, B.D., Mackay, A.A., 2005. Estimating the removal of
anthropogenic organic chemicals from raw drinking water by
coagulation flocculation. Journal of Environmental
Engineering 131, 108e118.
Blute, N., Russell, C., Chowdhury, Z., Wu, X., Via, S., 2010.
Nitrosamine occurrence in the U.S. e analysis and
interpretation of UCMR2 data. In: Proceedings of the AWWA
Water Quality Technology Conference Savannah, GA,
November 14-18, 2010.
Charrois, J.W.A., Boyd, J.M., Froese, K.L., Hrudey, S.E., 2007.
Occurrence of N-nitrosamines in Alberta public drinking-
water distribution systems. Journal of Environmental
Engineering and Science 6, 103e114.
Chen, B., Westerhoff, P., 2010. Predicting disinfection by-product
formation potential in water. Water Research 44, 3755e3762.
Chen, Z., Valentine, R.L., 2006. Modeling the formation of
N-Nitrosodimethylamine (NDMA) from the reaction of natural
organic matter (NOM) with monochloramine. Environmental
Science & Technology 40, 7290e7297.
Chen, Z., Valentine, R.L., 2007. Formation of
N-Nitrosodimethylamine (NDMA) from humic substances in
natural water. Environmental Science & Technology 41,
6059e6065.
Chen, Z., Yang, L., Zhai, X., Zhao, S., Li, A., Shen, J., 2010. N-
nitrosamine formation during chlorination/chloramination of
bromide-containing water. Water Science and Technology
Water Supply 10, 462e471.
Choi, J., Valentine, R.L., 2002. A kinetic model of N-
nitrosodimethylamine (NDMA) formation during water
chlorination/chloramination. Water Science and Technology
46 (3), 65e71.
De Ridder, D.J., Verliefde, A.R.D., Heijman, S.G.J., Verberk, Q.J.C.,
Rietveld, L.C., van der Aa, L.T.J., Amy, G.L., van Dijk, J.C., 2011.
Influence of natural organic matter on equilibrium adsorption
of neutral and charged pharmaceuticals onto activated
carbon. Water Science and Technology 63, 416e423.
Dotson, A., Westerhoff, P., Krasner, S.W., 2007. Nitrosamine
formation from natural organic matter isolates and sunlight
photolysis of nitrosamines. In: Proceedings of AWWA Annual
Conference and Exposition Toronto, ON, Canada, June 24e28,
2007.
EPA Integrated Risk Information System (IRIS), 1993.
N-Nitrosodimethylamine; CASRN 62-75-9. www.epa.gov/iris/
subst/0045.htm.
5694 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 8 7 e5 6 9 4
Gerecke, A.C., Sedlak, D.L., 2003. Precursors of
N-nitrosodimethylamine (NDMA) in natural waters.
Environmental Science & Technology 37, 1331e1336.
Health Canada, 2010. Guideline Technical Document on
N-Nitrosodimethylamine (NDMA) in Drinking Water for Public
Comment. http://www.hc-sc.gc.ca/ewh-semt/consult/_2010/
ndma/draft-ebauche-eng.php#a3.
Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation of
aquatic humic and non-humic matter with size-exclusion
chromatography e organic carbon detection e organic nitrogen
detection (LC-OCD-OND). Water Research 45, 879e885.
Kemper, J.M., Walse, S.S., Mitch, W.A., 2010. Quaternary amines
as nitrosamine precursors: a role for consumer products?
Environmental Science & Technology 44, 1224e1231.
Kim, J., Clevenger, T.E., 2007. Prediction of N-
nitrosodimethylamine (NDMA) formation as a disinfection by-
product. Journal of Hazardous Materials 145, 270e276.
Kohut, K.D., Andrews, S.A., 2003. Polyelectrolyte age and N-
nitrosodimethylamine formation in drinking water treatment.
Water Quality Research Journal of Canada 38 (4), 719e735.
Krasner, S.W., 2009. The formation and control of emerging
disinfection by-products of health concern. Philosophical
Transactions of the Royal Society A 367, 4077e4095.
Krasner, S.W., Dale, M.S., Lee, C.F.T., Garcia, E.A., Wong, T.M.,
Mitch, W., Von Gunten, U., 2010. Difference in reactivity and
chemistry of NDMA precursors from treated wastewater and
from polyamine polymers. In: Proceedings of the AWWA
Water Quality Technology Conference Savannah, GA,
November 14e18, 2010.
Mitch, W.A., Schreiber, I.M., 2008. Degradation of tertiary
alkylamines during chlorination/chloramination: implications
for formation of aldehydes, nitriles, halonitroalkanes, and
nitrosamines. Environmental Science & Technology 42,
4811e4817.
Mitch, W.A., Sedlak, D.L., 2004. Characterization and fate of
N-nitrosodimethylamine precursors in municipal wastewater
treatment plants. Environmental Science & Technology 38,
1445e1454.
Mitch, W.A., Sharp, J.O., Trussell, R.R., Valentine, R.L., Alvarez-
Cohen, L., Sedlak, D.L., 2003. N-nitrosodimethylamine (NDMA)
as a drinking water contaminant: a review. Environmental
Engineering Science 20 (5), 389e404.
MOE, 2003. Ontario Regulation 268/03 made Under the Safe
Drinking Water Act, 2002. http://www.e-laws.gov.on.ca/html/
source/regs/english/2003/elaws_src_regs_r03268_e.htm June
25, 2003.
Motulsky, H.J., Christopoulos, A., 2003. Fitting Models to Biological
Data using Linear and Nonlinear Regression. A Practical Guide
to Curve Fitting. GraphPad Software Inc., San Diego, CA. www.
graphpad.com.
Najm, I., Trussell, R.R., 2001. NDMA formation in water and
wastewater. Journal of the American Water Works
Association 93 (2), 92e99.
Office of Environmental Health Hazard Assessment (OEHHA),
2006. Public Health Goal for N-nitrosodimethylamine and
Cadmium in Drinking Water. http://www.oehha.org/water/
phg/cadndma122206.html.
Parris, G.E., 1980. Covalent binding of aromatic amines to
humates. 1. reactions with carbonyls and quinones.
Environmental Science & Technology 14, 1099e1106.
Sacher, F., Schmidt, C.K., Lee, C., von Gunten, U., 2008. Strategies
for Minimizing Nitrosamine Formation during Disinfection.
Water Research Foundation, Denver, C.O. AwwaRF Report
91209, August 15, 2008.
Schmidt, C.K., Brauch, H.J., 2008. N, N-dimethylsulfamide as
precursor for N-Nitrosodimethylamine (NDMA) formation
upon ozonation and its fate during drinking water treatment.
Environmental Science & Technology 42, 6340e6346.
Schmidt, C.K., Sacher, F., Brauch, H.J., 2006. Strategies for
minimizing formation of NDMA and other nitrosamines
during disinfection of drinking water. In: Proceedings of the
AWWA Water Quality Technology Conference Denvor, CO,
November 5e9, 2006.
Shen, R., Andrews, S.A., 2011. Demonstration of 20
pharmaceuticals and personal care products (PPCPs) as
nitrosamine precursors during chloramine disinfection. Water
Research 45, 944e952.
Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D.,
Lippincott, R.L., 2007. Efficiency of conventional drinking-
water-treatment processes in removal of pharmaceuticals and
other organic compounds. Science of the Total Environment
377, 255e272.
Summers, R.S., Hooper, S.M., Shukairy, H.M., Solarik, G., Owen, D.
, 1996. Assessing DBP yield: uniform formation conditions.
Journal of the American Water Works Association 88 (6),
80e93.
Thorn, K.A., Pettigrew, P.J., Goldenberg, W.S., Weber, E.J., 1996.
Covalent binding of aniline to humic substances. 2. 15N NMR
studies of nucleophilic addition reactions. Environmental
Science & Technology 30, 2764e2775.
USEPA, 2009. Contaminant Candidate List 3 (CCL3). http://water.
epa.gov/scitech/drinkingwater/dws/ccl/ccl3.cfm.
Valentine, R.L., Choi, J., Chen, Z., Barrett, S.E., Hwang, C., Guo, Y.C.
, Wehner, M., Fitzsimmons, S., Andrews, S.A., Werker, A.G.,
Brubacher, C., Kohut, K., 2005. Factors affecting the Formation
of NDMA in Water and Occurrence, Denver, CO, pp. 81e90.
Vieno, N., Tuhkanen, T., Kronberg, L., 2006. Removal of
pharmaceuticals in drinking water treatment: effect of
chemical coagulation. Environmental Technology 27,
183e192.
Weber, E.J., Spidle, D.L., Thorn, K.A., 1996. Covalent binding of
aniline to humic substances. 1. kinetic studies. Environmental
Science & Technology 30, 2755e2763.
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. Environmental Science & Technology 39,
6649e6663.
Wilczak, A., Assadi-Rad, A., Lai, H.H., Hoover, L.L., Smith, J.F.,
Berger, R., Rodigari, F., Beland, J.W., Lazzelle, L.J.,
Kincannon, E.G., Baker, H., Heaney, C.T., 2003. Formation of
NDMA in chloraminated water coagulated with DADMAC
cationic polymer. Journal of the American Water Works
Association 95 (9), 94e106.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Fate of N-nitrosodimethylamine, trihalomethane and

haloacetic acid precursors in tertiary treatment including
biofiltration

Maria José Farré*, Julien Reungoat, Francois Xavier Argaud, Maxime Rattier, Jürg Keller,
Wolfgang Gernjak
The University of Queensland, Advanced Water Management Centre (AWMC), Qld 4072, Australia

article info abstract

Article history: The presence of disinfection by-products (DBPs) such as trihalomethanes (THMs), halo-
Received 5 May 2011 acetic acids (HAAs) and N-nitrosamines in water is of great concern due to their adverse
Received in revised form effects on human health. In this work, the removal of N-nitrosodimethylamine (NDMA),
16 August 2011 total THM and five HAA precursors from secondary effluent by biological activated carbon
Accepted 20 August 2011 (BAC) is investigated at full and pilot scale. In the pilot plant two filter media, sand and
Available online 30 August 2011 granular activated carbon, are tested. In addition, we evaluate the influence of ozonation
prior to BAC filtration on its performance. Among the bulk of NDMA precursors, the fate of
Keywords: four pharmaceuticals containing a dimethylamino moiety in the chemical structure are
Adsorption individually investigated. Both NDMA formation potential and each of the studied phar-
Biological activated carbon maceuticals are dramatically reduced by the BAC even in the absence of main ozonation
Disinfection by-product prior to the filtration. The low removal of NDMA precursors at the sand filtration in
NDMA comparison to the removal of NDMA precursors at the BAC suggests that adsorption may
Sand filtration play an important role on the removal of NDMA precursors by BAC. Contrary, the
Ozone precursors for THM and HAA formation are reduced in both sand filtration and BAC indi-
cating that the precursors for the formation of these DBPs are to some extent
biodegradable.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction even if it is not yet included in the drinking water regulation,

Disinfection by-products in water are of great concern to
public health since bladder and colorectal cancers have been
associated with exposure to them in drinking water, and
experimental evidence suggests that exposure also occurs
through inhalation and dermal absorption (Villanueva et al.,
2007). The US Environmental Protection Agency (USEPA)
allows a maximum of 80 mg/L of total THMs (TTHMs) and
60 mg/L of five HAAs in drinking water based on its current
regulation guidelines (Richardson et al., 2007). For NDMA,
the USEPA classifies it in the group B2, which includes
compounds that are probably carcinogenic to humans (EPA,
2008). In Australia, NDMA has been included in the draft
Australian Drinking Water Guidelines at a maximum
concentration of 100 ng/L (ADWG, 2010). Moreover, NDMA was
recently identified as one of the DBPs with the great-
estpotential impact on public health (Hebert et al., 2010).
While THMs and HAAs are mainly formed when water is
disinfected with chlorine (Richardson et al., 2007), NDMA has
been related to the presence of chloramines, specifically

* Corresponding author. Tel.: þ61 7 33463233; fax: þ61 7 33654726.

E-mail address: m.farre@awmc.uq.edu.au (M.J. Farré).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.033
5696 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
dichloramines generated during the disinfection process
(Schreiber and Mitch, 2006).
Population increases, particularly in cities, and scarce
water resources have increased the demand for use of highly
treated municipal wastewater as a source of potable water
(Shannon et al., 2008). Studying the fate of DBP precursors
during secondary effluent treatment is of importance as an
increasing number of municipal water treatment plants are
engaged in the practice of potable reuse of treated wastewa-
ters. A common method to measure the precursors of DBP in
water is by means of formation potential tests. In these tests,
chlorine or chloramines are added to a buffered sample at
relatively high concentrations and kept reacting for at least
seven days to achieve the maximum formation of the specific
DBPs (Greenberg et al., 1992; Mitch et al., 2003).
While THM and HAA precursors are difficult to charac-
terize because different fractions of NOM may generate
these DBPs upon disinfection (Xie, 2004), some particular
NDMA precursors may be easier to monitor because the
specific dimethylamino moiety is required to generate the
nitrosamine upon chloramination. Traditionally, dimethyl-
amine (DMA) was the first NDMA precursor considered
during water chloramination since direct reaction between
this molecule and chloramine produces the carcinogen.
Dimethylamine is not only the typical catabolic product of
proteins in animals and plants (implying a typical concen-
tration in urine of 40 mg/L) (Tricker et al., 1994) but also
amongst the most frequently produced amines by the
chemical industry. Recently, it has also been shown that
other pharmaceuticals and personal care products with
substituted amino groups can serve as NDMA precursors
during chloramine disinfection (Lee et al., 2007; Kemper
et al., 2010; Shen and Andrews, 2010). In addition, natural
organic matter is another typical source of NDMA
(Westerhoff and Mash, 2002; Chen and Valentine, 2006, 2007,
2008). Wastewater treatment plants (WWTPs) remove 90% of
dissolved organic nitrogen (DON) and the typical concen-
tration of DON remaining after treatment at WWTPs is in
the range from 1 to 3 mgN/L (Pehlivanoglu-Mantas and
Sedlak, 2006). This remaining percentage consists of either
difficult-to-remove DON species, or DON produced during
biological treatment.
Finding technologies and processes to reduce DBP precur-
sors in water is thus of great interest. Adsorption on activated
carbon (AC) is one of the proven methods for removing natural
organic matter (NOM). However, the capacity for adsorption is
limited and replacement and disposal is costly. AC with active
biomass established on its surface is called biological activated
carbon (BAC). A BAC filter consists of a fixed bed of granular AC
supporting the growth of bacteria attached on the surface. This
technology has been used for many years in drinking water
treatment, usually after ozonation, and has proven to signifi-
cantly remove NOM, pharmaceuticals and personal care
products and ozonation by-products as well as odour and taste
compounds (e.g. geosmin and 2-methylisoborneol) (Simpson,
2008; Reungoat et al., 2011).
Although some work has already been published showing
the efficiency of BAC to remove DBP precursors in drinking
water (Simpson, 2008), to our knowledge, there are no reports
regarding the degradation of THM, HAA and N-nitrosamine
precursors in secondary treated effluent using ozone/BAC. In
the present study, we investigated the removal of NDMA, HAA
and THM precursors in full and pilot-scale biofilters treating
the effluent from a municipal WWTP. Additionally, the fate
along the treatment train of specific pharmaceuticals con-
taining tertiary amines (hence suspected to be NDMA
precursors) is presented.
2. Materials and methods
2.1. Chemicals
All chemicals used for chemical analysis were of analytical
grade and commercially available. NDMA (5000 mg/mL in
methanol) had a purity of >99.9% and was obtained from
Supelco. Deuterated d6-NDMA and d14-NDPA (N-nitro-
sodipropylamine) were used as surrogate and internal stan-
dard, respectively (1000 mg/mL in dichloromethane, >98.9%,
supplied by Accustandard and Ultra Scientific, respectively).
For the NDMA formation potential test, ammonium chloride
(TraceSELECT, 99.9% purity), sodium hydroxide (SigmaUl-
tra, 98%, pellets) and sodium hypochlorite solution (reagent
grade, available chlorine 4%) were used. Potassium dihy-
drogenphosphate (KH2PO4, Fluka, puriss. p.a., 99.5%) and
disodiumhydrogenphosphate (Na2HPO4$2H2O, Fluka, puriss.
p.a., 99.5%) were used to prepare pH buffer solutions. To
quench the chloramines solution, sodium sulphite (Fluka,
puriss. p.a., 98.0%) was employed. Commercial DPD test kits
(Hach) were used for the analysis of free and total chlorine (DPD
Total and Free Chlorine Reagent, test tube vials 2105545 and
2105645). To standardize the chlorine solution, sodium thio-
sulphate (SigmaUltra, 99.5%), potassium dichromate (Sig-
maUltra, 99.5%), acetic acid (ReagentPlus, 99%), soluble
starch (ACS reagent) and potassium iodide (ReagentPlus, 99%)
were used. For solid phase extraction (SPE), EPA commercial
charcoal optimized for NDMA analysis (Restek) was used. HPLC
grade dichloromethane, methanol and water were used for
conditioning and cleaning the SPE cartridges. Anhydrous
sodium sulphate, granular 10e60 mesh from Mallinckrodt was
used to remove water from the extracts. Finally 99% decane
(SigmaeAldrich) was used as keeper in the final concentration
step. Chemical standards of, venlafaxine hydrochloride, rox-
ythromycin, tramadol hydrochloride and doxylamine were
purchased from SigmaeAldrich (Steinheim, Germany) at
analytical grade (99%).
2.2. Full scale reclamation plant
The South Caboolture Water Reclamation Plant was designed
to reduce riverine pollution from the 40,000 population
equivalent wastewater treatment plant and to provide recy-
cled water to industry and community consumers (van
Leeuwen et al., 2003; Reungoat et al., 2010). The treatment
process as detailed in Fig. 1a incorporates biological denitri-
fication, pre-ozonation, coagulation/flocculation/dissolved air
flotation-sand filtration (DAFF), main ozonation, granular
activated carbon (GAC) filtration and final ozonation for
disinfection. The GAC was replaced in March 2008, 19 months
before the first sampling campaign, and the filter had treated
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
Fig. 1 e South Caboolture Water Reclamation Plant with sampling locations S (a) and pilot-scale biofilters (b). WWTP [
wastewater treatment plant, HRT [ hydraulic residence time, SRT [ sludge residence time. ----> refers to pilot scale
system.
approximately 45,000 bed volumes by that time. The proper-
ties of the GAC “Acticarb BAC GA1000N” (Activated Carbon
Technologies Pty Ltd, Australia) are detailed in Table SI 1. The
empty bed contact time (EBCT) is 18 min.19 months of oper-
ation have been shown to be sufficient for the development of
bacteria on the AC filters (Simpson, 2008), as confirmed by its
oxygen consumption (Reungoat et al., 2011). Therefore, the
full-scale filter is assumed to be biologically active. However,
in order to simplify the reading of the paper so as not confuse
with the pilot plants, we have kept the AC nomenclature for
the full scale unit in the manuscript.
2.3. Pilot-scale biofilters
Three pilot-scale biofilters (Fig. 1b) were set up in December
2006 at the South Caboolture Water Reclamation Plant,
Australia (Fig. 1a) parallel to the full scale AC. The biofilters are
3 m high and 22.5 cm internal diameter PVC columns; they
consist of 80  1 cm filtering bed height supported by a 20 cm
layer of gravel at the bottom, the top of the columns are filled
with water. One column contains sand as the filtering medium
and the other two are filled with the same GAC as the full-scale
filter but with a slightly lower particle size. Details on the
filtering media can be found in the supplementary information
(Table SI 1). The filters were continuously fed with water from
the main stream of the reclamation plant; BAC 1 received pre-
ozonated, whereas BAC 2 and SAND received water after main
ozonation. Pre-ozonated water refers here and in the rest of the
5697
manuscript to the denitrified effluent after pre-ozonation and
dissolve air flotation and filtration. A prior study showed that
the ozone dose added in the pre-ozonation is very low relative
to the dissolved organic carbon (DOC) concentration at this
stage (0.1 mgO3 mg1DOC) and does not lead to any significant
removal of DOC or organic micropollutants (Reungoat et al.,
2010). To support biological activity in the BAC filters,
compressed air and later 90% oxygen gas was bubbled in the
water column above the filtering bed to increase the dissolved
O2 concentration. The EBCT was controlled by adjusting the
effluent flow rate at the bottom of the columns and set at
60 min in all filters. The top layer of each filter bed (SAND and
BAC filters) was stirred weekly while withdrawn from above
the filter, to avoid clogging of the columns. This operation
removed some of the biomass from the top of the filter;
however no backwash of the entire filter was performed. A
previous study showed that biological activity had developed
on the filtering media and dissolved organic removal had
reached a steady state by June 2007 (Pipe-Martin et al., 2010).
2.4. Sample collection
During the first sampling campaign (October 2009), two sets of
time proportional 24-h composite samples were collected to
quantify NDMA formation potential. Sampling points are
shown in Fig. 1a. As the flow rates in the reclamation plant (due
to the presence of storage tanks) and in the pilot scale filters
were constant at the time of sampling, representative samples
5698 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
were collected using continuous pumping at 7 ml/min. Samples
were collected into glass bottles pre-washed with MilliQ water
and HPLC grade methanol and rinsed with the water sampled
directly before sampling. The samples were protected from
light and refrigerated during collection. 1 L of each composite
sample was then transferred into an amber glass bottle and
transported on ice to the laboratory for nitrosamines quantifi-
cation and NDMA formation potential test.
During the second sampling campaign (July 2010), three
sets of grab samples were collected to quantify instantaneous
concentrations and formation potential of THMs and HAAs.
Because the WWTP and the balance tank are located upstream
of the reclamation plant, variations in the water quality were
not expected to occur within the time of sampling. For THMs
and HAAs quantification, 200 mL of sample were collected in 2
amber glass bottles supplied by Queensland Health Forensic
and Scientific Services (QHFSS). In order to quench any chlo-
rine residual, 200 mg of ammonium chloride were added to
each bottle before sample collection. These samples were
transported on ice to QHFSS for direct analysis. To perform the
formation potential test, 1 L of sample was collected in amber
glass bottles that had been soaked in soap overnight, then
soaked in 10% nitric acid solution overnight, finally rinsed
with MilliQ water and HPLC grade methanol and dried in an
oven at 105  C. This sample was transported on ice to the
laboratory. During all sampling procedures, particular care
was taken to avoid stripping of the volatiles DBPs by slowly
filling the bottles and leaving no head space.
Simultaneously to other sample collection, 100 mL of
sample were collected in MilliQ rinsed plastic bottles for DOC
and nutrient concentration determination.
2.5. Analytical methods
2.5.1. Dissolved oxygen
Dissolved oxygen (DO) concentration was measured with an
YSI 6562 Dissolved Oxygen Probe connected to an YSI MDS 650
multi-parameter display system. An YSI 6560 conductivity
and temperature probe connected to the same multi-
parameter display system allowed to simultaneously correct
the DO concentration value and display it directly as
a concentration.
2.5.2. Dissolved organic carbon and nutrients
Prior to analysis, samples were filtered through a 0.45 mm PTFE
membrane. The DOC was measured as non-purgeable organic
carbon (NPOC) with an Analytik Jena multi N/C 3100 instru-
ment. For each sample, 2e3 replicates were measured, giving
a relative standard deviation of less than 3%. Ammonia,
nitrite, total NOx and phosphate were measured on a Lachat
flow injection analyzer as per the Lachat QuickChem method
31-107-06-1-A. Dissolved organic nitrogen (DON) was calcu-
lated to be the difference between total Kjeldahl nitrogen
(TKN) and NH4eN nitrogen. TKN was measured using the
Lachat QuickChem method 10-107-06-2-D.
2.5.3. Nitrosamines
2.5.3.1. Nitrosamines quantification. The method used for
N-nitrosamines was based on EPA Method 251 (Munch and
Bassett, 2004). N-nitrosodimethylamine (NDMA), N-nitro-
sodiethylamine (NDEA), N-nitrosomorpholine (NMOR), N-
nitrosopiperidine (N-Pip), N-nirosodibuthylamine (NDBA)
were included in the analysis. The details of the analysis are
published elsewhere (Farré et al., 2011). In short, water is
passed through a carbon solid phase extraction cartridge and
the N-nitrosamines are eluted off with dichloromethane. The
extracts are concentrated by evaporation under nitrogen to
1 mL and analysed by capillary GC-mass spectrometer in
positive chemical ionisation (PCI) mode with anhydrous
ammonia as the chemical ionisation gas (Finnigan Trace G.C.
Ultra and Finnigan Trace DSQ Mass Spectrometer). The
detection limit for the technique used was 5 ng/L for NDMA,
10 ng/L for NDEA and NMOR, and 20 ng/L for N-Pip and NDBA.
2.5.3.2. NDMA formation potential (FP) test. The NDMA FP
test follows closely the procedure described as nitrosamine
precursor test by Mitch et al. (2003) and is described elsewhere
(Farré et al., 2011). A concentration of 2 mM (140 mg/L Cl2) was
used to determine the NDMA FP of the selected samples to
ensure that all precursors are reacting with chloramines to
generate NDMA. The test was performed at pH 6.8 which was
achieved by adding 700 mg/L KH2PO4 and 880 mg/L Na2H-
PO4$2H2O to the water sample (10 mM phosphate buffer). All
experiments were performed in 1 L amber glass bottles which
were stored at room temperature (23  2  C) in the dark for
seven days. On the seventh day, the residual chloramine
concentration was quenched with 2.5 g/L sodium sulphite
(added in solid phase to the sample) to prevent further NDMA
formation. The samples were then analysed for NDMA.
2.5.4. Trihalomethanes and haloacetic acids
2.5.4.1. Trihalomethanes quantification. Chloroform (TCM),
bromodichloromethane (BDCM), dibromochlorometane
(DBCM) and bromoform (TBM) were quantified by QHFSS by
purging the volatile organic directly from the aqueous sample
and subjecting the volatilized component to gas
chromatography-mass spectrometry, in accordance with
USEPA method 524.2 revision 4.1 (Munch, 1995). This was
achieved using a using a Shimadzu QP2010 equipped with
a purge and trap system (Tekmar Velocity). A column ZB-624
(20 m length  0.18 mm inner diameter  1.0 mm film thick-
ness) was used for separation. Injection was done 1/1 split at
200  C. The carrier gas employed was helium at 42 mL/min
and 118 kPa. Initial temperature of the oven was 40  C, held for
2 min and then increased to 200  C at a rate of 10  C/min rate.
The mass spectrometer operating conditions were: ion source
and interface line temperatures 200  C and 230  C; Scan
35e300 amu at 1111 amu/sec. The LOQ is 1 mg/L for all
analytes.
2.5.4.2. Haloacetic acids quantification. Five HAAs (abbrevia-
tion 5HAAs; monochloroacetic acid-MCAA, dichloroacetic
acid-DCAA, trichloroacetic acid-TCAA, bromochloroacetic
acid-BCAA, monobromoacetic acid-MBAA and dibromoacetic
acid-DBAA) were extracted from aqueous samples by
portioning into methyl tert-butyl ether (MtBE) after adding
sulphuric acid and sodium sulphate following USEPA method
number 552.3 (USEPA., 2003).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4 5699

The analysis was carried out by QHFSS using a gas chro-
matography coupled with an electron capture detector (GC-
ECD) at 300  C (Shimadzu GC2010). A column DB-1701 (15 m
length  0.32 mm inner diameter  0.25 mm film thickness)
was used for identification and a second column DB5 (30 m
length  0.25 mm inner diameter  0.25 mm film thickness)
was used for confirmation. 1 mL of sample was splitless
injected at 250  C. The carrier gas employed was hydrogen at
68.1 mL/min and 45.3 kPa. Initial temperature of the oven was
50  C, held for 1 min and then increased to 75  C at a rate of
5  C/min rate. The temperature was finally increased to 280  C
at a 40  C/min. The LOQ is 10 mg L1 for MCAA, DCAA and
TCAA and 5 mg L1 for BCAA, MBAA and DBAA.
2.5.4.3. THMs and HAAs formation potential test. The
formation potential test was performed following Standard
Methods for the Examination of Water and Wastewater
(Greenberg et al., 1992). Chlorine demand of each sample was
determined with a 4 h test prior to starting the 7 days
formation potential tests to calculate the chlorine dose
necessary to ensure a residual of free chlorine at the end of the
experiment. The 7 days test was performed in 1 L amber glass
bottles washed following the same procedure as for the
sample collection and filled completely to avoid the presence
of a head space. Bottles were kept in the dark at a constant
temperature of 23  2  C. At the end of the 7 days, the residual
free chlorine was measured to ensure chlorine did not limit
the reaction. Then 200 mL were collected in each of 2 amber
glass bottles supplied by QHFSS for THMs and HAAs quanti-
fication. The residual chlorine was quenched by adding
sodium sulfite (0.1 mL of 100 g/L solution per 25 mL of sample)
and ammonium chloride (0.2 mL of 50 g/L solution per 250 mL
of sample) in the THMs and HAAs bottles respectively, as
recommended by the standard method. The bottles were sent
immediately to QHFSS for analysis. A blank sample (MilliQ
water) was included in each test batch for quality control
purposes, they were all below LOQ for all THMs and HAAs.
and a BAC filter after main ozonation (BAC 2). NDMA and
NMOR were detected beyond LOQ across the plant (i.e., 5 ng/L
for NDMA and 10 ng/L for NMOR).
3.1.2. NDMA formation potential
The same samples were subjected to a NDMA formation
potential test for 7 days. Fig. 2 shows the result of the NDMA
formation potential across the full-scale plant and the pilot
plant. No other N-nitrosamines, considered in this work, were
observed to be formed above the LOQ during the formation
potential tests. The NDMA formation potential measured at
the influent of the reclamation plant was 423  55 ng/L and
remained in the same range after denitrification confirming
that this treatment does not affect NDMA precursors (Mitch
and Sedlak, 2004). The NDMA formation potential of the
secondary effluent used in South Caboolture Water Recla-
mation Plant was found to be similar to other domestic
wastewater treatment plants in South East Queensland (Farré
et al., 2011) and in other countries (Pehlivanoglu-Mantas and
Sedlak, 2006) supporting that no effluents with high NDMA
formation potential were discharged to this specific WWTP.
Pre-ozonation (2 mgO3/L) and DAFF reduced the NDMA
formation potential by around 20% each bringing the
concentration down to 260  31 ng/L. The main ozonation (5
mgO3/L with 15 min contact time) was the most effective step
of the full scale treatment, reducing the NDMA formation
potential by another 66% to levels below 100 ng/L (this corre-
sponds to the 78% of total NDMA precursors removal from the
beginning of the treatment as shown in Fig. 2). This data
follows the trends observed by Lee et al. (2007) when
measuring the effect of ozone treatment on NDMA precursors
in natural waters. That study reported a NDMA formation
potential reduction of 32e94% by applying up to 40 mM
(1.9 mgO3/L) ozone, depending on the natural water and
500 100

2.5.5. Micropollutants 400 80

% NDMA FP removal
Doxylamine, roxithromycin, tramadol and venlafaxine were
NDMA FP (ng/L)

quantified according to the method described in Reungoat 300 60

et al. (2011) in the sampling campaigns done for NDMA
precursors and in three additional sampling campaigns. The
200 40
method consisted of solid phase extraction (SPE), elution,
concentration, and analysis by liquid chromatography
coupled with tandem mass spectrometry (LC/MS-MS). 100 20

0 0
nt on on F 1 on on 2 C nt
3. Results and discussion lue ati ati AF AC ati ati AC st A fflue
y eff trific ozon ost D ost B ozon f iltr ost B po
a l e
r
da eni pre
- p p ain t san
d p fin
on st d post
sec po st m pos
po
3.1. Nitrosamines and NDMA formation potential
Fig. 2 e Bar charts correspond to NDMA precursors
3.1.1. Nitrosamines measured by NDMA formation potential test (FP) across
NDMA, NDEA, NMOR, N-Pip and NDBA were analysed in all South Caboolture Water Reclamation Plant and pilot-scale
the samples taken from South Caboolture Water Reclamation biofilters. Error bars correspond to the standard deviation
Plant. Sampling points along the treatment train are indicated of two independent sampling campaigns (n [ 2). Dot
in Fig. 1. The selected N-nitrosamines were also analysed in points correspond to the percentage of NDMA precursor’s
the effluent of the three pilot scale columns, which are BAC removal relative to the raw water. Striped bars represent
filter placed after pre-ozonation and dissolve air flotation and results from the pilot plants while plain bars are results
filtration (BAC 1), sand filtration after main ozonation (SAND) from the full-scale plant.
5700 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
oxidation conditions. In South Caboolture Water Reclamation
Plant, the activated carbon filter reduced the NDMA precur-
sors further, leaving a concentration of NDMA precursors in
the final effluent of 53  6 ng/L.
BAC 1 (pilot plant fed with water post DAFF) was able to
reduce the NDMA formation potential by more than 85%,
which is better than the main ozonation alone, and the
effluent concentration was 48  4 ng/L, similar to the effluent
of the main plant. The SAND filter was not able to reduce the
NDMA FP after main ozonation but the BAC 2 decreased
NDMA FP down to 39  9 ng/L, which is only slightly better
than BAC 1. Ammonia, nitrate, DOC and dissolved organic
nitrogen (DON) concentrations from the same sampling
campaign are shown in Figure SI 1 of the supporting infor-
mation. Overall, the average DON concentration measured
across the treatment scheme decreased form an initial value
of 1.6 mg/L to 1.0 mg/L after main ozonation with a further
reduction to 0.8 mg/L after AC (18 min EBCT), while the
concentration after BAC 1 (60 min EBCT) was 0.6 mg/L showing
that this treatment is more effective than main ozonation plus
AC at the full scale. From the residual 1.0 mg/L DON left after
main ozonation, 0.15 mg/L could be degraded at the sand
filtration while 0.5 mg/L could be removed at BAC 2 leaving
a final 0.5 mg/L of DON when ozone plus BAC 2 are employed.
Regarding DOC the same trend was observed. The initial
concentration of DOC at the plant was 11 mg/L and this value
was reduced to 6.5 mg/L after the main ozonation and to
4.7 mg/L after AC at the full-scale plant. BAC 1 was able to
decrease DOC concentration to 4.7 mg/L even without
the need of adding ozone in the system. From the residual
6.5 mg/L DOC left after main ozonation, 1.8 mg/L could be
degraded at the sand filtration while 3.4 mg/L could be
removed at BAC 2 leaving a final 3.1 mg/L of DOC when ozone
plus BAC 2 are employed. It has been reported that NDMA
formation potential cannot be predicted by simple measure-
ments such as DOC because of the specific nature of NDMA
precursors that is a very minor and specific fraction of DOC
(Pehlivanoglu-Mantas and Sedlak, 2008). However, our data
across the plant showed that this measurement may act as
a good indicator of the NDMA formation potential for this
specific scenario since the coefficient of regression when
plotting these two parameters was higher than 0.9. For DON
also a linear relation (R2 ¼ 0.8) could be observed even though
the fitting was lower in this case. This data is plotted in the
Figure SI 2 of the supporting information. The fraction of DOC
removal in the BAC 1 (i.e., 56%) is not as high as the fraction of
NDMA formation potential removal (85%); hence, it appears
that the DOC removed at the BAC 1 contains a higher
concentration of NDMA precursors, which means that the
NDMA precursors are preferentially adsorbed or degraded in
the BAC compared to DOC.
3.1.3. NDMA model precursors
NDMA precursors contain a tertiary amino group with two
methyl substituents except for dimethylamine which is the
only secondary amine that can act as NDMA precursor.
Dimethylamine concentrations in primary wastewater efflu-
ents are typically in the range of 20e80 mg/L. Since dimethyl-
amine is degraded by bacteria, levels in secondary wastewater
effluents are lower (<10 mg/L) (Mitch et al., 2003; Mitch and
Sedlak, 2004). This low concentration commonly present in
secondary effluents could not explain the total concentration
of NDMA precursors measured since the molar yield of
dimethylamine conversion to NDMA is lower than 0.5% (Mitch
et al., 2003). Although we did not measure dimethylamine in
the samples, we considered other tertiary amines present in
the secondary effluent to investigate the fate of specific NDMA
precursors and along the treatment train. To this aim, four
pharmaceuticals known to be NDMA precursors were ana-
lysed for and found at low mg/L levels in the secondary effluent
used as influent in South Caboolture reclamation plant. Dox-
ylamine, with a conversion yield to NDMA during chlorami-
nation of 8.0e9.8% (Shen and Andrews, 2010), was found in
the influent to the reclamation plant at an average concen-
tration of 289 ng/L. Roxithromycin, tramadol and venlafaxine
were also found in the source water of the treatment plant at
average concentrations of 240, 1300 and 1330 ng/L, respec-
tively. However, the yield of conversion for these pharma-
ceuticals is around 0.5% (Shen and Andrews, 2010). The
chemical structures of these NDMA precursors are plotted in
Fig. 3 in conjunction with the logD and pKa values. The fate of
the pharmaceuticals in conjunction with the NDMA formation
potential is plotted in Fig. 4. Even though the presence of these
four pharmaceuticals could only account for less than 4% of
the NDMA precursors present in the inlet of South Caboolture
Water Reclamation Plant based on the conversion yields
published by Shen and Andrews (2010), they may be used as
indicators of the fate of pharmaceuticals containing the
dimethylamino moiety in the chemical structure across BAC.
The high removal of both NDMA formation potential and
the suspected pharmaceutical NDMA precursors observed in
BAC 1 could be due to biodegradation activity of the biomass
attached on the surface of activated carbon, adsorption of the
compounds on the surface of the activated carbon or the
combined effects of adsorption and biodegradation. Some
authors have also hypothesized that the biodegradation
continuously regenerates adsorption sites by degrading
adsorbed molecules (Herzberg et al., 2003; Simpson, 2008).
This effect is called bioregeneration. In a previous article
Reungoat et al. (2011) observed that the concentration of
pharmaceuticals and personal care products (PPCP) removed
in the same BAC 1 used in the present study remained
constant over more than two years of continuous operation.
Therefore, they suggest that the removal of organic matter
and PPCPs observed in the BAC filters was due to biodegra-
dation (or adsorption followed by biodegradation) rather than
adsorption alone as adsorption efficiency would typically
decrease over time. When evaluating the fate of the NDMA
precursors selected for this study we observed that the four
pharmaceuticals, which have a logD(pH7)<2 and a pKa>8,
remained at similar concentration before and after the SAND
filter but decreased significantly after BAC 1 or BAC 2. The low
removal of NDMA precursors during SAND filtration in
comparison to the BACs suggests that adsorption may play
a significant role in the removal of NDMA precursors. Hydro-
phobic interactions are the dominant mechanism in activated
carbon adsorption of neutral organic compounds (Yoon et al.,
2003). However, for charged molecules the charge interactions
are of high importance. Activated carbon is usually negatively
charged at pH typical for drinking water treatment, hence,
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4 5701

Fig. 3 e Pharmaceuticals known to be NDMA precursors detected in the treatment plant. Log D is calculated using Advanced
Chemistry Development (ACD/Labs) Software V11.02 (ª 1994e2011 ACD/Labs). pKa from Reungoat et al. (2011).

increasing the adsorption of positively charged compounds as while in BAC 1 and BAC 2 this number increased to 9.6 and
the ones studied in this work. In accordance, Westerhoff et al. 9.9 mg/L, respectively. This oxygen consumption in both BAC
(2008), previously reported a good removal by adsorption on 1 and BAC 2 confirms that there is higher biological activity in
PAC of protonated compounds. Nevertheless, activated the BAC filters compared to the SAND filter which may explain
carbon typically has a surface area of several hundred square
meters per gram due to its high porosity but most of this 10000 10000
surface is not accessible to bacteria as it is located in micro- doxylamine
roxithromycin
pores with a diameter smaller than 2 nm (Reungoat et al., 1000
tramadol
1000
venlafaxine
2011). Nonetheless, the external surface of the activated NDMA FP
NDMA FP (ng/L)

carbon grains is much rougher and uneven than the surface of 100 100
sand grains and therefore potentially provides more sites for
ng/L

the bacteria to attach. Therefore, both a higher population of

10 10
bacteria attached to the surface and a higher adsorption
capacity could explain the enhanced removal of NDMA
1 1
precursors observed at BAC in comparison to sand filtration
for doxylamine, roxithromycin, tramadol and venlafaxine. It
0.1 0.1
cannot be excluded that extracellular polymeric substances nt tion tion 1 nt
F n on 2 C
lu e a a AF AC atio ati AC st A fflue
such as polysaccharides produced by bacteria attaching to the y eff trific ozon ost D ost B ozon filtr ost B p o l e
r a
da deni pre- p p in
AN
D p fin
activated carbon grains could adsorb the pharmaceuticals on st ma
sec po st ost S
po p
with medium sorption potential in comparison to the GAC
without biological activity (Zhang et al., 2010). In order to gain Fig. 4 e Fate of specific NDMA precursors across South
insight into understanding the hypothesis of biodegradation Caboolture Water Reclamation Plant and pilot-scale
of NDMA precursors as mechanism for their removal, the biofilters, error bars correspond to standard deviation of
dissolved oxygen was measured in the inlet and outlet of the five independent sampling campaigns (n [ 5). The dot
three pilot scale filters (i.e., BAC 1, SAND filter and BAC 2). The points correspond to the average NDMA formation
average uptake of oxygen in the SAND filter was 2.7 mg/L potential.
5702 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4

the higher removal of NDMA precursors observed. However, across the full-scale plant and the pilot plant filters in
adsorption cannot be completely discarded and more studies conjunction with 5HAAs, TTHMs and DOC data.
are being undertaken to distinguish biodegradation and As seen in Fig. 5, monohalogenated acids were not formed
adsorption effects in the BAC filters. in the formation potential test. Among the HAAs generated
during the tests, DCAA was measured at the highest concen-
tration (154  32 mg/L) followed by TCAA (143  17 mg/L) and
3.2. Other disinfection by-products
BCAA (30  6 mg/L). The formation of chlorine-containing
HAAs was significantly reduced after BAC 1 following the
3.2.1. Trihalomethanes and haloacetic acids
trend observed for NDMA precursors (76  36 mg/L and
TTHMs and five (5HAAs) are regulated DBPs in both drinking
60  36 mg/L for DCAA and TCAA, respectively), however this
and recycled water. Therefore, we also evaluate the fate of
behavior was not observed with bromine-containing HAAs
those DBP precursors across Caboolture Water Reclamation
since the concentration of BCAA and DBAA remained
Plant and the pilot plants. To this aim, TTHMs and five HAAs
unchanged by BAC 1. The same trend was observed for THMs.
and their precursors were analysed in three different
TCM formation (224  24 mg/L after DAFF) was reduced
sampling campaigns. Sampling points included at the full-
significantly in the BAC 1 (103  36 mg/L) but bromine-
scale plant were: post DAFF, post main ozonation, post AC
containing DBP formation remained constant along the
and final effluent. The DBPs were also analysed at the three
plant. At the BAC 1 the organic matter is reduced from around
pilot scale columns. No HAAs were measured above the LOQ
7 mg/L to 5 mg/L as seen in Fig. 5, either by adsorption or
for any of the sampling points during the different sampling
biodegradation. Nevertheless, the ion content remains
campaigns. On the other hand low concentrations of THMs
constant as shown by the conductivity data presented in the
were measured across the treatment train but the TTHM
supportive information (Table SI 3). Since we could not
concentration was always below 11 mg/L (see Table SI 2 of the
measure bromate formation either above the LOD (i.e., 10 mg/L)
supporting information).
across the treatment plant we assumed the oxidation of Br to
BrO 
3 by ozone was minimal. Therefore all Br was available to
3.2.2. Trihalomethane and haloacetic acid precursors
be oxidised to HOBr by HOCl during the formation potential
The same samples were subjected to DBP formation potential
test. The rate constant of bromide with HOCl to generate HOBr
for 7 days, as previously described. Fig. 5 shows the result of
is 1.5  103 M1 s1 (Kumar and Margerum, 1987) and the rate
HAA and THM formation potential of the selected samples
constant of THMs formation is in the range of 0.01 and
200 8 400
0.03 M1 s1 (Gallard and Von Gunten, 2002). It is known that
DCAA once formed, bromine reacts about 10 times faster than
TCAA
BCAA
DBAA
350 chlorine with natural organic matter since the activities of
150 6
individual HAA (µg/L)

DOC
5HAAs
electrophilic substitution for electron release to stabilize car-
300
5HAAs (µg/L)

bocation are more favourable for the Br atom due to its higher
DOC (mg/L)

100 4 250 electron density and smaller bond strength relative to the Cl
atom (Westerhoff et al., 2004; Hua et al., 2006). Hence, the
200 formation of Br-DBPs is limited by the initial Br concentration
50 2 whereas the Cl-DBPs would be limited by the organic matter.
150
Therefore, when organic matter decreases along the treat-
0 0 100 ment train, the formation of chlorine-containing DBPs is
FF C1 ati
on
ati
on C2 AC reduced while the formation of bromine-containing DBPs
DA BA on iltr BA po
st
st st oz df st
po po in an po remains constant.
a s
st
m st
po po
Main ozonation removes the precursors for TCAA and TCM
300 8 400 while biofiltration decreases also the concentration of DCAA
TCM
BDCM precursors. The increase of DBCM observed by others (Chen
250 DBCM 350
TBM et al., 2009) is also seen to a small degree in our data as the
individual THM (µg/L)

DOC 6
TTHMs concentration of this DBP increases from 11 mg/L to 15 mg/L
200 300
TTHMs (µg/L)

from after DAFF to after ozonation. Liang and Singer (2003)

DOC (mg/L)

150 4 250 have suggested that bromide is more reactive with aliphatic
precursors, such as hydrophilic organic material, than with
100 200
aromatic precursors, such as hydrophobic organic material.
2
Hence, the organic matter becoming more hydrophilic after
50 150
ozonation may explain the increase of the formation of this
0 0 100 specific DBP. The concentration of DCAA, TCAA and TCM
FF C1 ati
on ion C2 AC precursors were further reduced in the sand filtration to 24, 18
DA BA on rat BA st
st st oz filt st po
po po t d o
po
s
st
san
p
and 22 mg/L, respectively showing that biodegradation indeed
po
plays a role in their reduction since adsorption is considered
Fig. 5 e HAAs and THMs precursors and DOC across South minimal in the SAND filter. Further in the treatment, the
Caboolture Water Reclamation Plant and pilot-scale formation of these DBPs is reduced in BAC 2 and AC. In
biofilters, error bars correspond to standard deviation of general, the reduction of TTHM and 5HAA precursors was
three independent sampling campaigns (n [ 3). proportional to the decay of DOC across the plant with
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
R2 > 0.9, as seen in Figure SI 3 of the supporting information,
confirming that this simple parameter is a good indicator to
estimate the precursors for those specific DBPs in treated
secondary effluents.
4. Conclusions
Both, the bulk of NDMA precursors and individual pharma-
ceuticals that may form NDMA are dramatically reduced by
the BAC filters even without preceding main ozonation. DOC
removed by the BAC contains a higher concentration of NDMA
precursors than the bulk DOC, which means that the NDMA
precursors are preferentially adsorbed or degraded in the BAC.
The percentage of removal in the BACs is unaffected by a main
ozonation step when considering long filtration times
(EBCT ¼ 60 min). On the contrary, under similar conditions,
sand filtration does not significantly remove NDMA precur-
sors. Since adsorption is considered minimal on the sand
filtration while biodegradation may potentially take place and
due to the different removal observed at the BACs and sand
filter, adsorption is deemed likely to be important for the
observed removal NDMA precursors during BAC filtration.
However, more research is necessary to fully understand the
removal mechanisms in order to quantify the adsorption
versus biodegradation processes.
THM and HAA precursors follow the trend of NDMA
precursors since they are also considerably reduced by the
BACs. However, the precursors for these specific DBPs are also
removed in the sand filter. Hence, both adsorption and
biodegradation may play a significant role on their removal.
The formation potential for brominated THMs and HAAs
remains constant across the different treatment steps due to
increase on the bromine/DOC ratio along the filtration steps.
Acknowledgements
The authors want to specifically acknowledge Urban Water
Security Research Alliance for funding the “NDMA formation
potential project” and “The Enhanced Treatment Project” and
Unity Water and their staff for giving access to the plant.
Thanks to Miss Hollie King for correcting the English.
Appendix. Supplementary material
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.watres.2011.08.033.
references
Australian Drinking Water Guideline, 2010. Public Draft
Document for Consultation. http://www.nhmrc.gov.au/
guidelines/consult/consultations/draft_adwg_guidelines.htm.
Chen, Z., Valentine, R.L., 2006. Modelling the formation of
N-nitrosodimethylamine (NDMA) from the reaction of natural
5703
organic matter (NOM) with monochloramine. Environmental
Science and Technology 40 (23), 7290e7297.
Chen, Z., Valentine, R.L., 2007. Formation of
N-nitrosodimethylamine (NDMA) from humic substances in
natural water. Environmental Science and Technology 41 (17),
6059e6065.
Chen, Z., Valentine, R.L., 2008. The influence of the pre-oxidation
of natural organic matter on the formation of
N-nitrosodimethylamine (NDMA). Environmental Science and
Technology 42 (14), 5062e5067.
Chen, K.C., Wang, Y.H., Chang, Y.H., 2009. Using catalytic
ozonation and biofiltration to decrease the formation of
disinfection by-products. Desalination 249, 929e935.
Environmental Protection Agency, EPA, 2008. Emerging
Contaminant- N-Nitrosodimethylamine (NDMA) Fact Sheet.
Farré, M., Keller, J., Holling, N., Poussade, Y., Gernjak, W., 2011.
Occurrence of NDMA precursors in wastewater treatment
plant effluent and their fate during UF-RO membrane
treatment. Water Science and Technology 63, 605e612.
Gallard, H., Von Gunten, U., 2002. Chlorination of natural organic
matter: kinetics of chlorination and of THM formation. Water
Research 36, 65e74.
Greenberg, A., Clesceri, L., Deaton, A., 1992. Standard Methods for
the Examination of Water and Wastewater. Water
Environment Federation, Washington DC, USA.
Hebert, A., Forestier, D., Lenes, D., Benanou, D., Jacob, S., Arfi, C.,
Lambolez, L., Levi, Y., 2010. Innovative method for prioritizing
emerging disinfection by-products (DBPs) in drinking water on
the basis of their potential impact on public health. Water
Research 44, 3147e3165.
Herzberg, M., Dosoretz, C.G., Tarre, S., Green, M., 2003. Patchy
biofilm coverage can explain the potential advantage of BGAC
reactors. Environmental Science & Technology 37, 4274e4280.
Hua, G., Reckhow, D.A., Kim, J., 2006. Effect of bromide and iodide
ions on the formation and speciation of disinfection
byproducts during chlorination. Environmental Science and
Technology 40, 3050e3056.
Kemper, J.M., Walse, S.S., Mitch, W.A., 2010. Quaternary amines
as nitrosamine precursors: a role for consumer products?
Environmental Science and Technology 44, 1224e1231.
Kumar, K., Margerum, D.W., 1987. Kinetics and mechanism of
general-acid-assisted oxidation of bromide by hypochlorite
and hypochlorous acid. Inorganic Chemistry 26, 2706e2711.
Lee, C., Schmidt, C., Yoon, J., Von Gunten, U., 2007. Oxidation of N-
nitrosodimethylamine (NDMA) precursors with ozone and
chlorine dioxide: kinetics and effect on NDMA formation
potential. Environmental Science and Technology 41, 2056e2063.
Liang, L., Singer, P.C., 2003. Factors influencing the formation and
relative distribution of haloacetic acids and trihalomethanes
in drinking water. Environmental Science and Technology 37,
2920e2928.
Mitch, W.A., Gerecke, A.C., Sedlak, D.L., 2003. A N-
Nitrosodimethylamine (NDMA) precursor analysis for
chlorination of water and wastewater. Water Research 37,
3733e3741.
Mitch, W.A., Sedlak, D.L., 2004. Characterization and fate of N-
Nitrosodimethylamine precursors in municipal wastewater
treatment plants. Environmental Science and Technology 38,
1445e1454.
Munch, J.W., 1995. Measurement of Purgeable Organic
Compounds in Water by Capillary Column Gas
Chromatography/Mass Spectrometry Cincinnati, OH.
Munch, J.W., Bassett, M.V., 2004. Method Development for the
Analysis of N-Nitrosodimethylamine and other N-
Nitrosamines in Drinking Water at Low Nanogram/Liter
Concentrations using Solid-Phase Extraction and Gas
Chromatography with Chemical Ionization Tandem Mass
Spectrometry Method 521. EPA/600/R-05/054.
5704 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 6 9 5 e5 7 0 4
Pehlivanoglu-Mantas, E., Sedlak, D.L., 2006. The fate of
wastewater-derived NDMA precursors in the aquatic
environment. Water Research 40, 1287e1293.
Pehlivanoglu-Mantas, E., Sedlak, D.L., 2008. Measurement of
dissolved organic nitrogen forms in wastewater effluents:
concentrations, size distribution and NDMA formation
potential. Water Research 42, 3890e3898.
Pipe-Martin, C., Reungoat, J., Keller, J., 2010. Dissolved Organic
Carbon Removal by Biological Treatment. Water Quality
Research Australia, Adelaı̈de (Australia).
Reungoat, J., Macova, M., Escher, B.I., Carswell, S., Mueller, J.F.,
Keller, J., 2010. Removal of micropollutants and reduction of
biological activity in a full scale reclamation plant using
ozonation and activated carbon filtration. Water Research 44,
625e637.
Reungoat, J., Escher, B.I., Macova, M., Keller, J., 2011. Biofiltration
of wastewater treatment plant effluent: effective removal of
pharmaceuticals and personal care products and reduction of
toxicity. Water Research 45, 2751e2762.
Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R.,
DeMarini, D.M., 2007. Occurrence, genotoxicity, and
carcinogenicity of regulated and emerging disinfection
by-products in drinking water: a review and roadmap for
research. Mutation Research - Reviews in Mutation Research
636, 178e242.
Schreiber, I.M., Mitch, W.A., 2006. Nitrosamine formation
pathway revisited: the importance of chloramine speciation
and dissolved oxygen. Environmental Science and Technology
40, 6007e6014.
Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G.,
Mariñas, B.J., Mayes, A.M., 2008. Science and technology for
water purification in the coming decades. Nature 452, 301e310.
Shen, R., Andrews, S.A., 2010. Demonstration of 20
pharmaceuticals and personal care products (PPCPs) as
nitrosamine precursors during chloramine disinfection. Water
Research 45, 944e952.
Simpson, D.R., 2008. Biofilm processes in biologically active
carbon water purification. Water Research 42, 2839e2848.
Tricker, A.R., Pfundstein, B., Preussmann, R., 1994. Nitosable
secondary amines: exogenous and endogenous exposure and
nitrosation in vivo. ACS Symposium Series 553, 93e101.
USEPA., 2003. Method 552.3. Determination of Haloacetic Acids
and Dalapon in Drinking Water by Liquideliquid
Microextraction, Derivatization, and Gas Chromatography
with Electron Capture Detection Cincinnati, OH.
van Leeuwen, J., Pipe-Martin, C., Lehmann, R.M., 2003. Water
reclamation at South Caboolture, Queensland, Australia.
Ozone: Science & Engineering 25, 107e120.
Villanueva, C.M., Cantor, K.P., Grimalt, J.O., Malats, N.,
Silverman, D., Tardon, A., Garcia-Closas, R., Serra, C.,
Carrato, A., Castaño -Vinyals, G., Marcos, R., Rothman, N.,
Real, F.X., Dosemeci, M., Kogevinas, M., 2007. Bladder cancer
and exposure to water disinfection by-products through
ingestion, bathing, showering, and swimming in pools.
American Journal of Epidemiology 165, 148e156.
Westerhoff, P., Chao, P., Mash, H., 2004. Reactivity of natural
organic matter with aqueous chlorine and bromine. Water
Research 38, 1502e1513.
Westerhoff, P., Mash, H., 2002. Dissolved organic nitrogen in
drinking water supplies: a review. Journal of Water Supply:
Research and Technology - AQUA 51 (8), 415e448.
Westerhoff, P., Yoon, J., Snyder, S., Wert, E., 2008. Fate of
endocrine-disruptor, pharmaceutical, and personal care
product chemicals during simulated drinking water treatment
processes. Environmental Science & Technology 39,
6649e6663.
Xie, Y.F., 2004. Disinfection by-Products in Drinking Water
Formation, Analysis and Control. Lewis Publishers, Florida.
Yoon, Y., Westerhoff, P., Snyder, S.A., Esparza, M., 2003. HPLC-
fluorescence detection and adsorption of bisphenol A, 17b-
estradiol, and 17a-ethynyl estradiol on powdered activated
carbon. Water Research 37, 3530e3537.
Zhang, Z., Wang, L., Shao, L., 2010. Study on Relationship between
Characteristics of DOC and Removal Performance by BAC
Filter 2010 4th International Conference on Bioinformatics and
Biomedical Engineering, iCBBE 2010.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Continuous combined Fenton’s oxidation and biodegradation

for the treatment of pentachlorophenol-contaminated water

Julio A. Zimbron 1, Kenneth F. Reardon*

Department of Chemical and Biological Engineering, 100 Glover Building, Colorado State University, Fort Collins, CO 80523-1370, USA

article info abstract

Article history: Pentachlorophenol (PCP) was studied as a model recalcitrant compound for a sequential
Received 5 July 2011 chemical oxidation and biodegradation treatment, in a continuous laboratory-scale system
Received in revised form that combined a Fenton’s chemical reactor and a packed-bed bioreactor.
22 August 2011 PCP degradation and dechlorination were observed in the Fenton’s reactor at a resi-
Accepted 23 August 2011 dence time of 1.5 h, although no reduction of total organic carbon (TOC) was observed. Both
Available online 31 August 2011 PCP degradation and dechlorination were strongly dependent on the H2O2 dose to the
chemical reactor. The PCP degradation intermediates tetrachlorohydroquinone and
Keywords: dichloromaleic acid were identified in this reactor. Further treatment of the Fenton’s
Biodegradation reactor effluent with a packed-bed bioreactor (operating at a residence time of 5.5 h)
Fenton’s reaction resulted in partial biodegradation of PCP degradation intermediates and reduction in TOC,
Hydroxyl radical although no further reduction of PCP or dechlorination was achieved in the bioreactor.
Kinetics model Increased residence time in the bioreactor had no significant impact on degradation of
Pentachlorophenol TOC. Recycle of the effluent from the bioreactor to the chemical reactor increased the TOC
degradation, but not the extent of the PCP degradation or dechlorination.
A mathematical model of the combined Fenton’s oxidation and biodegradation system
supported the experimental results. While the model over-predicted the PCP and TOC
degradation in the combined system, it adequately predicted the sensitivity of these
parameters to different H2O2 doses and recycle rates. The model indicated that high recycle
rates would improve TOC degradation.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction et al., 1994). Specialized cultures can achieve faster degrada-

Pentachlorophenol is a common water contaminant, released
into the environment from wood treating and biocide
formulation operations. It often occurs in mixtures with other
contaminants, constituting the most recalcitrant fraction of
such mixtures (Graves and Joyce, 1994).
PCP biodegradation has been reported both aerobically and
anaerobically, although rates are too slow for practical treat-
ment, owing to the need to induce specific enzymes (Hale
tion rates (Puhakka and Melin, 1996), but might not be resilient
when exposed to environmental disturbances (Scott and Ollis,
1995).
Due to this reported recalcitrance of PCP and other
compounds to biodegradation, there has been increased
interest in alternative treatment technologies. Advanced
oxidation processes (AOPs) rely on the generation of hydroxyl
free radicals ($OH), a highly reactive chemical species
(Venkatadri and Peters, 1993). Complete mineralization of

* Corresponding author. Tel.: þ1 970 491 6505; fax: þ1 970 491 7369.
E-mail address: kenneth.reardon@colostate.edu (K.F. Reardon).
1
Present address: Department of Civil and Environmental Engineering, Colorado State University, Fort Collins, CO 80523-1320, USA.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.038
5706 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
List of symbols and abbreviations
a recycle rate, equal to the ratio of the flow rate of
the effluent from the bioreactor that is recycled
back to the chemical reactor with respect to the
inlet flow rate to the combined system
sBR residence time (time units) in the bioreactor
sChRx residence time (time units) in the chemical
(Fenton’s) reactor
mi microbial growth rate (1/h)
mi,max Monod’s maximum microbial growth rate (1/h)
$OH hydroxyl free radical
DCMA dichloromaleic acid
Fe(II) ferrous iron
Fe(III) ferric iron
H2O2 hydrogen peroxide
ki Monod’s half saturation microbial degradation
constant for compound i (mol/L)
kj second-order kinetic constant for the Fenton’s
system reaction j (L/mol$s)
mi,j stoichiometric coefficient for reactant i in Fenton’s
reaction j PCP pentachlorophenol
recalcitrant compounds can be achieved using these tech-
nologies (Scott and Ollis, 1995), although long treatment times
and strong doses of the oxidizing reagents are required
(Esplugas et al., 2004; Marco et al., 1997; Pera-Titus et al., 2004).
Often, the chemically-oxidized intermediates are less recal-
citrant than the parent compound (Comninellis et al., 2008).
Since biological processes are typically less expensive than
chemical processes, addition of a biodegradation stage can
improve the economy of the overall process, particularly for
low concentration wastewater (Pera-Titus et al., 2004).
Examples of this combined treatment have been presented
(Pera-Titus et al., 2004; Marco et al., 1997; Scott and Ollis, 1995),
but the mechanistic information required for the combined
reactor system analysis and design remains scarce
(Comninellis et al., 2008; Esplugas et al., 2004; Mantzavinos
and Psillakis, 2004). As a result of the complexity of the
processes and the lack of mechanistic data, process integra-
tion has typically been experimentally evaluated on a case-by-
case basis. For chemical oxidation alone, a widely accepted
idea is that kinetics modeling requires a complete reaction
pathway for the organic substrate (Duesterberg and Waite,
2006; Kang et al., 2002; Rivas et al., 2001). Although simpli-
fied kinetics models have provided significant insight into
reactor design (Esplugas et al., 2004), they typically are based
on first-order reactions. Such simplified analysis has the
limitation of neglecting competitive effects of the chemically
oxidized by-products on the target parent compound, an
essential feature of AOPs that in practice precludes complete
treatment of target contaminants.
The purpose of this work was to study the combination of
Fenton’s oxidation (the combination of H2O2 and ferrous
iron) and biodegradation of PCP-contaminated water.
Experiments were conducted in a continuous stirred-tank
reactor (CSTR) in which Fenton’s degradation of PCP
occurred. The effluent from this reactor was continuously
Ri,j rate law for Fenton’s reaction j (in which i is
a reaction member)
ratei,Bio rate of reaction of chemical species i under
biodegradation treatment
ratei,Ch rate of reaction of chemical species i under the
Fenton’s reaction system
Si,BioRx concentration (mol/L) of the chemical species i in
the bioreactor
SiChRx concentration (mol/L) of the chemical species i in
the chemical reactor
Si,FbioRx concentration (mol/L) of the chemical species i in
the feed to the bioreactor (which in the combined
model was equal to the concentration of this
species i n the chemical reactor)
Si,FChRx concentration (mol/L) of the chemical species i in
the feed to the chemical reactor
TCHQ tetrachlorohydroquinone
TOC total organic carbon
XT packed bed biomass content (mg)
YX/Si yield of biomass to carbon substrate consumption
(as TOC) in mg of biomass/mol of total organic
carbon
treated in a second reactor system that included a packed-
bed bioreactor to degrade the PCP intermediates. The
effects of H2O2 dose, residence time in the bioreactor, and
recycle of the bioreactor effluent back to the chemical reactor
on the combined system performance were studied. A
mathematical model was developed for the combined
system, based on a kinetics model simplified using a lumping
approach, for both chemical oxidation and biodegradation.
In this approach, a non-PCP fraction was defined to account
for $OH scavenging effects of the by-products on the chem-
ical oxidation of PCP, and as the biodegradable fraction on
the biodegradation process. The model was used to test the
consistency of the experimental data and provide insights
into reactor design and operating strategies (i.e., the recycle
of effluent from the second stage bioreactor back to the
chemical reactor).
2. Kinetics modeling
2.1. Fenton’s oxidation kinetics model
A mathematical model was developed previously for the
chemical oxidation kinetics and validated using data from
batch experiments (Zimbron and Reardon, 2009). This Fen-
ton’s kinetics model included 11 reactions between inorganic
species involved in the Fenton’s system (i.e., H2O2, Fe(II)/
Fe(III) and initiation and propagation reactions involving
radicals $OH, $O2H/$O 2 ). The mass balances were rewritten
for the CSTR used in the combined Fenton’s-biodegradation
system.
The termination reaction for PCP with $OH (calculated
using the competitive kinetics method) was included with
a value of 4.4  1009 L/mol$s. The scavenging effects of the PCP
by-products on the degradation of PCP (and the rest of the
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
reaction system) were included as a lumped reaction with
a kinetic constant (Zimbron and Reardon, 2009). This lumping
approach has been successfully applied to estimate the
degradation of contaminants by different AOPs, including
photo-Fenton’s reactions (Zepp and Scholtzhauer, 1979),
ultraviolet radiation (Haag and Hoigne, 1985), and soil slurries
(Huling et al., 1998).
Using the above mentioned set of reactions, the reaction
rate for each reactant or product i in the Fenton’s reaction
system is assumed to follow second-order kinetics, general-
ized as:
X
n X
n
ratei;Ch ¼ mi;j Ri;j ¼ mi;j kj Si;jChRx Sh;jChRx
j¼1 j¼1
in which Si,jChRx indicates the molar concentration of chemical
species, i. For each specific Fenton’s reaction j, Ri,j and mi,j are
the rate law and the stoichiometric coefficient of reactant i
(negative for reactants, positive for products), respectively, kj
is the reaction rate constant, and the subscript h indicates
other chemical species involved in that particular reaction
rate (e.g., $OH).
The resulting mass balance for the chemical reactor
(including a recirculation stream from the bioreactor to the
chemical reactor) is:
Si;FChRx þ ratei;Ch sChRx þ aSi;BioRx
Si;ChRx ¼
1þa
in which Si is the molar concentration of the reactive species i,
the subscripts ChRx and FChRx indicate the chemical reactor
and feed stream to this reactor, ratei is the rate of reaction of
the chemical species i (mol/L$s), sChRx is the residence time (s)
in the chemical reactor, and a is the ratio of recycle flow rate
from the bioreactor to the chemical reactor to the total flow
rate to the system.
2.2. Biodegradation kinetics model
Kinetics data from the observed biodegradation of PCP by-
products was incorporated into a biodegradation kinetics
model. The mass balances for PCP and TOC were calculated,
based on the non-PCP TOC fraction (the difference of PCP
concentrations in the feed and the chemical reactor,
multiplied by 6, the PCP stoichiometric carbon molecular
content).
Monod-type biodegradation kinetic parameters were
obtained based on the non-PCP TOC as substrate (at different
feed concentrations), as supported by experimental data. The
Monod model for the rate of biodegradation of substrate i,
ratei,Bio, depends on the total concentration of biomass con-
tained in the bioreactor (XT, assumed constant for an immo-
bilized biomass reactor over the period of the experiment), the
yield of biomass to substrate i (YX/Si), and the molar concen-
tration of substrate (Si,BioRx):
XT mi;max Si;BioRx XT
ratei;Bio ¼ mi ¼
YX=Si ki þ Si;BioRx YX=Si
The combined model assumed that chemical oxidation and
biodegradation were mutually exclusive. Furthermore, upon
recycle, biomass and excess nutrients incorporated into the
5707
bioreactor stream were assumed not to affect the degradation
of PCP by Fenton’s reaction. These assumptions were verified
in batch experiments (data not shown).
The mass balance for the bioreactor is given by:
ratei;Bio sBioRx
Si;BioRx ¼ Si;ChRx þ (4)
1þa
in which sBioRx indicates the residence time (in time units) in
the bioreactor. The combined reactor model results in
a system of coupled algebraic equations that was solved using
Engineering Equation Solver (F-chart Software).
This model represents a large simplification of the
combined system, for both the chemical oxidation process
(1)
and the biodegradation process. However, the mass balance
approach to model development has the potential to provide
insights into the study of the process and chemical reactor
design. The need for such simplified, yet mechanistic models
for process integration has been highlighted by others
(Esplugas et al., 2004; Mantzavinos and Psillakis, 2004).
3. Material and methods
3.1. Chemicals
Pentachlorophenol (99%), the extraction surrogate (dibro-
(2)
mophenol, 95%) and the internal standard (dibromo-
benzene, 98%) were purchased from Sigma. FeSO4$7H2O
(99%þ) was purchased from Baxter. H2O2 (non-stabilized,
31.4%), sodium chloride (99.9%) and sulfuric acid (conc.)
were all purchased from Fisher. Solvents (chloroform and
ethyl acetate) for the analysis of PCP and organic PCP by-
products were pesticide-grade (Fisher). Deionized and
dechlorinated water was used for PCP solution and plate
media preparation, while water used for preparation of all
other reactants was Reagent Type I Grade water (Nanopure,
18 mU conductivity).
3.2. Continuous system apparatus
The continuous combined Fenton’s and biodegradation
system consisted of two CSTRs in series (2 L Bioflo C-30, New
Brunswick Scientific) (Fig. 1). In the first reactor (with a liquid
volume of 250 mL and residence time of 1.5 h), Fenton’s
treatment of the PCP-contaminated water was performed at
pH 3.5, within the reported optimum range for Fenton’s
treatment (for example Venkatadri and Peters, 1993; Zepp
and Scholtzhauer, 1979). A Markson controller with
a combination pH electrode and H2SO4 0.1 N were used to
control pH at 3.5. Ferrous iron and H2O2 were dosed to ach-
ieve specific concentrations as described in the next section.
The chemical reactor and reactant feeds were covered with
aluminum foil.
The effluent from the Fenton’s reactor was fed to the
bioreactor at pH 7, adding supplementary nutrient solution
(3) (phosphates, trace minerals and nitrogen; Section 3.3). pH
control was achieved with NaOH 0.2 N solution and an Omega
pH/ORP controller with a combination pH electrode. The
bioreactor residence time was either 5.5 or 10 h, achieved by
adjusting the bioreactor liquid volume (either 750 or 1500 mL,
5708 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
Fig. 1 e Continuous combined reactor system for Fenton’s
oxidation and biodegradation of PCP-contaminated water.
respectively). In some reactor configurations, the effluent
from the bioreactor was recycled to the chemical reactor.
The low organic carbon concentration in the feed stream
(due to the low PCP solubility) resulted in oligotrophic (low
nutrient) conditions, under which suspended cell cultures
might be washed out of the reactor. Thus, a packed-bed
column (glass, 3.8 cm OD, 14.5 cm long) filled with 38.8 g of
Celite biocatalyst carrier R-635 (Manville) was provided, to
which the contents of the main bioreactor vessel were recir-
culated (in an upflow configuration) at a rate 20 times higher
than the feed to the stirred tank to ensure that the liquid
contents were well mixed. The bioreactor was initially inoc-
ulated with 3 mL of Fort Collins wastewater treatment plant
aerobic activated sludge.
The reactor tubing consisted of chemically resistant PTFE
when possible, or Norprene (Masterflex). All other parts in the
system were stainless steel or glass. No PCP sorption or
degradation was detected on these materials or the Celite
biomass support through batch 24 h sorption experiments.
3.3. Reactants to the continuous system
The PCP solution was prepared with deionized and dechlori-
nated water adjusted to pH 10 with 0.8 mL of 0.1 M NaOH to
facilitate PCP dissolution. PCP was added to the reported
solubility at neutral pH (14 ppm). The flow rate of this solution
to the reactor system was 3.5 L/day.
The flow rate of the H2O2 feed solution reactant was 85 mL/
day, equivalent to 2.5% of the total flow rate to the combined
system.At this flow rate, the H2O2 feed of 880 mM and 1800 mM
solutions yielded the low and high doses of [H2O2] ¼ 220 and
[H2O2] ¼ 370 mM to the chemical reactor, respectively.
The iron feed solution was prepared by dissolving 2.3 g of
FeSO4 $7H2O in water, with 6.4 mL of concentrated H2SO4 to
avoid Fe(II) oxidation, and diluted to 1.0 L to a final concen-
tration of [Fe(II)] ¼ 8300 mM. The flow rate of this reactant was
85 mL/day, resulting in an actual dose of [Fe(II)] ¼ 200 mM to the
chemical reactor.
The supplementary nutrient medium was the PAS mineral
salts medium (see Section 3.5), modified by replacing calcium
chloride with calcium sulfate to eliminate background chlo-
ride concentrations. Nitrogen in the supplementary nutrient
medium was stoichiometric to carbon in the PCP solution at
a 1:5 N:C molar ratio (Shuler and Kargi, 1992). This nutrient
solution was fed to the bioreactor at a rate of 85 mL/day
(approximately 2.5% of the total flow rate to the system).
3.4. Analytical methods
Concentrations of PCP and the identified PCP degradation by-
products tetrachlorohydroquinone and dichloromaleic acid
were determined with a HP 5890 Series II GC-MS, after solvent
extraction. Ferrous iron, chloride ion, and H2O2 analyses were
done spectrophotometrically, using the 1,10-phenanthroline,
thiocyanate (Hach Method 20635-00), and titanium sulfate
methods, respectively (APHA et al., 1980). Additional details
about these analyses are available from a previous report
(Zimbron and Reardon, 2009).
Dissolved oxygen was measured with a temperature-
compensated oxygen electrode (Phoenix Electrode), and
a Cole Parmer 01971-00 analyzer. Purgeable total organic
carbon (TOC) was analyzed with a Dohrmann DC-80 TOC
Analyzer (detection limit lower than 1 mg/L).
3.5. Biomass analysis
A protein quantitative assay was used to measure biomass
indirectly, since iron precipitates precluded direct estimation
by optical density (600 nm). Samples were collected and
frozen until analysis. Samples were centrifuged in 250-mL
Teflon bottles for 25 min at 13200g. The biomass was
resuspended with phosphate 0.1 M pH 7 buffer and centri-
fuged again twice, then reconstituted to 3.0e10.0 mL. The
biomass concentrate was lysed with a sonicator (UPXL, Heath
Systems) for 20 min on ice and analyzed for protein using the
micro-BCA assay (Pierce), comparing absorbance (562 nm) to
that of bovine serum albumin standards.
The ability of the bioreactor effluent biomass to degrade
PCP was tested by plating on an agar medium with PCP as the
only carbon source (PCP medium) at a concentration of 0.5 mg/
L (to prevent inhibitory effects).This medium included 7.5 g of
Bacto-Agar with 38.5 mL PAS concentrate (13.8 g NH4Cl, 10.97 g
KH2PO4 and 28.39 g K2HPO4 diluted to 500 mL) and 5 mL of PAS
100 salts solution (0.15 g CaCl2 $2H2O, 0. 5 g FeSO4 $7H2O, 2.5 g
MnSO4 $H2O, 9.75 g MgSO4, 2 drops H2SO4 diluted to 500 mL),
diluted to 500 mL (Bedard et al., 1986) and autoclaved.
The ability of the biomass to grow on a readily available
carbon source was tested by plating on a non-selective
medium, Trypticase Soy Broth (TSB), diluted to 1:4 to simu-
late an oligotrophic (low nutrient) environment.
3.6. Experiments
In addition to the basic sequential chemical oxidation and
biodegradation reactor configuration, the effect of recycling
part of the bioreactor effluent back to the chemical reactor
was evaluated, at the low residence time of 5.5 h in the
bioreactor, for both oxidation strengths. Two recycle rates
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
(0.2 and 0.4) were used for this purpose. This recycle rate (a) is
defined as the ratio of the bioreactor effluent flow rate sent
back to the chemical reactor (Fig. 1) to the total inlet flow to the
system.
After a new operating condition was established, the
system ran for 48 h to achieve steady state before sampling.
This represents 32 and 9 residence times for the chemical
reactor and the bioreactor, respectively. A set of samples at
a H2O2 dose of 220 mM taken at 36 h yielded similar PCP and
TOC results to those at 48 h, suggesting that 36 h was sufficient
for the system to achieve steady state.
4. Results and discussion
4.1. Experimental results
4.1.1. Hydrogen peroxide consumption in side reactions
Control experiments without Fe(II) in the continuous Fenton’s
reactor showed that H2O2 consumption occurred in that
reactor in the absence of Fe(II), and that neither significant PCP
degradation nor dechlorination resulted. The measured H2O2
concentrations were 29% and 30% lower than the mass
balance around the chemical reactor at H2O2 doses of 178 and
406 mM, respectively, despite the lack of measured losses in the
feed flask. These losses likely occurred at iron precipitate-
coated reactor surfaces (due to the long term operation). The
reactivities of such surfaces toward H2O2 have been docu-
mented (Watts et al., 1997). For these reasons, all reported H2O2
doses tested in this work were corrected to 30% lower than the
actual dose, as estimates of the H2O2 available for Fenton’s
mechanisms (i.e., resulting in PCP degradation). These are
referred hereafter as the effective hydrogen peroxide doses.
4.1.2. Biomass characterization
The system operated for several months without recycle at
constant conditions (Fe(II) dose of 200 mM and an effective
H2O2 of 260 mM) to establish a stable microbial community. At
these conditions, the PCP degradation (with respect to the feed
concentration) achieved in the chemical reactor was 65% and
the TOC degradation achieved in the bioreactor was 13%. A
bioreactor effluent aliquot was analyzed for protein, yielding
a concentration of 0.41 mg/L (c.v. ¼ 11%). This effluent con-
tained 1.3  1005 CFU, obtained by plating on ¼-strength TSB
medium (after 48 h of incubation).
Upon plating the bioreactor effluent on PCP medium and
incubating for one week, no colonies developed, although
growth in the non-selective TSB medium occurred at 2 days,
showing the lack of PCP degraders within this microbial
population.
Two samples of five pellets each were taken from the
packed bed and analyzed for dry and organic matter (by drying
at 35  C and incinerating). The estimated organic content of
the pellets was 0.41% (0.13%), or 160 mg of biomass (dry
weight) for the entire packed column.
Three pellets from the column were crushed, sonicated,
and analyzed for protein. The organic-free dry weight was
estimated by incineration. The resulting approximate protein
content of biomass (organic matter) in the bioreactor was 69%,
within the reported range of 40%e70% (Shuler and Kargi,
5709
1992). Biomass and protein measurements in the column
and the bioreactor effluent confirmed that more than 99% of
the biomass present in the bioreactor was located in the
packed bed, where most of the microbial activity occurred.
4.1.3. Effect of hydrogen peroxide dose
PCP degradation achieved in the chemical reactor (with
a residence time of 1.5 h without recycle) and the resulting
chloride production under a constant dose of Fe(II) of 200 mM
and variable doses of hydrogen peroxide is shown in Fig. 2.
Also shown are the model-predicted Fenton’s-driven degra-
dation of PCP and dechlorination in the chemical reactor at
these conditions. The model over-predicted PCP degradation
in the continuous system, although it achieved considerably
more accurate predictions for batch experiments data
(Zimbron and Reardon, 2009). The lowest experimental PCP
dimensionless concentrations (approximately 45%) were
higher than the lowest model-predicted PCP dimensionless
concentration (25%, corresponding to H2O2 doses of 200 mM or
higher).
The lack of agreement between experimental and model-
predicted PCP degradation is attributed to side reactions of
free radicals at reactive surfaces in the continuous system.
These side reactions consumed H2O2, but as shown by the no-
Fe(II) control experiment with a high dose of H2O2, H2O2-
scavenging reactions did not yield significant PCP degrada-
tion. Detailed evaluation of the nature of the interactions of
active surfaces with Fenton’s reactive species (i.e., free radi-
cals) was beyond the scope of this work.
Both PCP degradation and Cl release depended strongly
on the H2O2 dose, up to a concentration of 200 mM (Fig. 2). The
mean observed-to-theoretical chloride concentration ratio
Fig. 2 e Pentachlorophenol and chloride concentrations
achieved at different effective H2O2 doses in the
continuous chemical reactor. [Fe(II)] [ 200 mM. Error bars
represent the standard deviation of duplicate samples.
Filled (C) and open (B) circles represent dimensionless
concentrations of PCP in the chemical reactor (with respect
to the feed concentration) and measured Cl-
concentrations, respectively. The open triangle symbol is
the result of a control experiment with a high dose of H2O2
and no Fe(II). The solid and dotted line represent the
model-predicted PCP degradation and chloride production
based on complete PCP dechlorination in the chemical
reactor, respectively.
5710 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4

Fig. 3 e Dimensionless concentrations of PCP (bottom) and TOC (top) at different hydrogen peroxide doses. Bars in white,
light grey, and dark grey represent concentrations in the feed, chemical reactor, and bioreactor, respectively. Residence
times in the chemical and bioreactor were 1.5 and 5.5 h, respectively. Error bars represent the standard deviation of
duplicate samples.

(assuming complete dechlorination of the degraded PCP) was
84% (9%). Previous work on electrochemical oxidation of PCP
in soil suspension reported about 80% dechlorination (Hanna
et al., 2005), while studies on ozonation and UV/H2O2 have
reported lower partial dechlorination (in the order of 30e60%)
(Hirvonen et al., 2000).
The PCP and TOC concentrations in the feed, chemical
reactor, and bioreactor are presented in Fig. 3. The average
TOC concentration of the saturated PCP feed was 4.7 mg/L
(0.43), slightly higher than the theoretical TOC concentration
of saturated PCP solution (4.0 ppm) (probably due to calibra-
tion error). Triplicate feed water blanks yielded values of 0.208
(0.07) mg/L, showing that the feed water contained very low
levels of extraneous carbon. The chloride (at the same three
points along the experimental apparatus) and protein
concentrations in the bioreactor effluent are shown in Fig. 4.
In agreement with the plating test results, Figs. 3 and 4
indicate that microbial growth (based on protein measure-
ments) was supported by the partial degradation of the non-
PCP TOC as a carbon source. In addition to the lack of PCP
biodegradation (beyond the level achieved by chemical
oxidation), no further dechlorination occurred at the biore-
actor. As indicated by the TOC reduction observed in the
bioreactor, the bioreactor microbial population had a prefer-
ence for non-chlorinated intermediates over chlorinated ones,
consistent with previous findings for PCP treated by electro-
Fenton’s (Hanna et al., 2005).
No PCP losses in the chemical reactor (or the bioreactor)
occurred in control experiments without H2O2. Cell wash-out
due to the lack of nutrients at these conditions might have
caused the non-zero protein concentration in the bioreactor
effluent.

Fig. 4 e Concentrations of chloride (bottom) and protein (top) at different hydrogen peroxide doses. Bars in white, light grey,
and dark grey represent concentrations in the feed, chemical reactor, and bioreactor, respectively. Residence times in the
chemical and bioreactor were 1.5 and 5.5 h, respectively. Error bars represent the standard deviation of duplicate samples.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4 5711

Tetrachlorohydroquinone (TCHQ) and dichloromaleic acid
(DCMA) were identified as PCP intermediates of Fenton’s
oxidation in the continuous system, consistent with
a previous batch reactor study (Zimbron and Reardon, 2009)
and in agreement with previous reports of AOP treatment of
PCP (Benitez et al., 2003; Hong and Zeng, 2002; Mills and
Hoffman, 1993; Wong and Crosby, 1981; Zhao et al., 2006).
DCMA was observed as a PCP by-product (and that of TCHQ
and tetrachlorocatechol) under treatment with different AOPs
(Hanna et al., 2005; Hong and Zeng, 2002; Sen Gupta et al.,
2002; Shen et al., 2009; Wong and Crosby, 1981). The concen-
trations of TCHQ and DCMA in the chemical reactor and
bioreactor are shown in Fig. 5. Attempts to identify other PCP
intermediates were not successful.
Owing to partial degradation of PCP in the Fenton’s system,
the non-PCP TOC fraction was less than 55% of the PCP-
saturated solution equivalent of 330 mM TOC. At a residence
time of 5.5 h in the bioreactor, the maximum TOC reductions
were about 14%. Further increases of the retention time in the
bioreactor (with reactor conditions in the chemical reactor
held constant) at the same effective H2O2 doses of 0, 150 mM,
and 260 mM did not significantly increase TOC degradation
(nor protein production) (95% confidence level, data not
shown).
Low biodegradation rates are typically observed at high
substrate or at very low biomass concentrations (conditions
that lead to zero-order biodegradation rates). These potential
causes seem unlikely in this system, due to the low soluble
TOC (limited by PCP solubility) and the large amount of
biomass present in the bioreactor column. A better explana-
tion is that a simple Monod-type biodegradation model might
not adequately describe the complex mixture kinetics of the
Fenton’s-treated PCP. The different biodegradability of the
observed intermediates (Fig. 5) supports this. The observed
lack of dechlorination, and reported toxicity of highly chlori-
nated compounds (such as PCP) (Pera-Titus et al., 2004), might
explain the observed lack of sensitivity to extended residence
time and the low extent of biodegradation of non-PCP TOC.
Analysis of the biomass in the effluent indicated a yield of
7.5 g of biomass (dry weight)/mol TOC consumed. This
measured yield was only 30% of the theoretical maximum of
25.5 g biomass dry weight/mol TOC, obtained assuming
a theoretical biomass formula of CH2N0.25O0.5 (Shuler and
Kargi, 1992). The remainder (70%) of the reduction of TOC in
the bioreactor is an order of magnitude estimate of the
mineralization extent to CO2 (rather than incorporation into
biomass).
The TCHQ yields upon PCP degradation in the continuous
chemical reactor were 3.0 mol%, at both H2O2 doses (150 mM
and 260 mM). The DCMA yields measured during continuous
operation of the chemical reactor were 3.8 and 3.3 mol% at low
and high H2O2 dose, respectively. These yields for TCHQ and
DCMA were consistent with batch experiments results
(Zimbron and Reardon, 2009).
4.1.4. Effect of recycle
The effect of recycling the effluent from the bioreactor back to
the chemical reactor for further treatment was tested at
a residence time of 1.5 h in the chemical reactor and 5.5 h in
the bioreactor. For this, 20% and 40% of the total flow rate to
the system was recycled from the bioreactor back to the
chemical reactor (a ¼ 0.2 and a ¼ 0.4, respectively). These
experiments were conducted at both low (150 mM) and high
(260 mM) effective H2O2 doses. Two-way ANOVA indicated that
PCP and chloride concentrations in the effluent were not
affected by the recycle level (20% or 40%), compared with no
recycle (confidence level ¼ 95%). At the low hydrogen peroxide
dose there were no significant effects of recycle (at both
recycle levels tested) on TOC biodegradation. In contrast, the
recycle rate effects (for both recycle ratios of 0.2 and 0.4) on the
TOC biodegradation were significant at the high H2O2 dose
(Fig. 6). At this high H2O2 dose, recycle had the effect of
increasing the TOC reduction that occurred at the biodegra-
dation stage with respect to no recycle. This improved TOC
degradation upon recycling was confirmed by protein analysis
that indicated increased biomass production (data not shown).

Fig. 5 e Concentration of two observed PCP intermediates in the combined system. Tetrachlorohydroquinone (TCHQ)
concentrations (bottom) and dichloromaleic acid (DCMA) concentrations (top) at different hydrogen peroxide doses. Bars in
white, light grey, and dark grey represent concentrations in the feed (not present), chemical reactor, and bioreactor,
respectively. Residence times in the chemical and bioreactor were 1.5 and 5.5 h, respectively. Error bars represent the
standard deviation of duplicate samples.
5712 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4

Fig. 6 e Dimensionless concentrations of PCP (bottom) and TOC (top) at three different recycle ratios (0, 0.2, and 0.4) and
a high hydrogen peroxide effective dose of 260 mM. Bars in white, light grey, and dark grey represent concentrations in the
feed, chemical reactor, and bioreactor, respectively. Residence times in the chemical and bioreactor were 1.5 and 5.5 h,
respectively. Error bars represent the standard deviation of duplicate samples (except for feed concentrations, for which
error bars represent the standard deviation of 10 measurements).

4.2. Kinetics modeling results
The biodegradation data presented in Figs. 3 and 6, in addition
to data obtained at the longer bioreactor residence time (11 h,
for a total of 6 data points) were used to estimate the Monod
biodegradation kinetics of the non-PCP TOC. The yield coef-
ficient was estimated as 5.2 (2.2) g/L of protein produced per
mol/L of TOC consumed, based on the measured concentra-
tions of protein in the effluent from the bioreactor. The line-
arized Monod equation (Shuler and Kargi, 1992) was used,
resulting in the Monod values of ks ¼ 250 mM TOC and
mmax ¼ 3.1  103 h1 (with a correlation coefficient of 0.74).
The model-predicted TOC and PCP concentrations at
different recycle ratios (0, 0.4, and 2.0) were estimated for two
different scenarios:
a) Variable biodegradation residence time, with constant
residence time in the chemical reactor (1.5 h) and fixed
H2O2 and ferrous iron doses (300 mM and 200 mM,
respectively).
b) Variable H2O2 dose, with constant residence times in both
chemical and biodegradation reactors (1.5 h and 5.5,
respectively), and Fe(II) ¼ 200 mM).
back to the chemical reactor (in agreement with the experi-
mental results).
Model-predicted PCP and TOC dimensionless concentra-
tions at a variable H2O2 dose, [Fe(II)] ¼ 200 mM, recycle ratios of
0.4 and 2.0, and constant residence time in each reactor (1.5 h
in the chemical reactor and 5.5 h in the bioreactor) are shown
in Fig. 8. At the three tested recycle rates (0, 0.4, and 2), the PCP
degradation was not sensitive to additional H2O2 doses above
200 mM. Recycle rates of 0.4 and 2.0 achieve small increases in
the degradation of PCP in the combined system. The observed
lack of sensitivity of PCP degradation achieved by the experi-
mental combined system at recycle rates of 0.2 and 0.4 can be
explained in terms of the lack of sensitivity of the PCP and
chloride analysis to detect the expected small changes.
Model-predicted TOC and PCP degradation in the combi-
ned system increased with higher H2O2 doses (at constant

At constant oxidizer dose and variable biodegradation

residence time, the model-predicted PCP dimensionless
concentrations (Fig. 7) were limited by the degradation of PCP
achieved in the chemical reactor, because Fenton’s reaction
was the only mechanism to degrade PCP in the combined
system. Increases in recycle rate (from 0 to 0.4 and 2) achieved
only small increases in the degradation of PCP in the chemical
reactor. This is in agreement with the observed lack of Fig. 7 e Model-predicted dimensionless degradation of PCP
sensitivity of PCP degradation in the experimental combined and TOC as a function of residence time in the bioreactor.
system at recycle rates of 0.2 and 0.4. In contrast, the model- The residence time in the chemical reactor was 1.5 h
predicted degradation of TOC under the same conditions [H2O2] [ 300 mM and [Fe(II)] [ 200 mM. Thicker lines
was more sensitive to the recycle of the bioreactor effluent represent model solutions at lower recycle rates.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
Fig. 8 e Model-predicted dimensionless degradation of
TOC and PCP in the combined system at variable hydrogen
peroxide dose. The PCP degradation scale (right) is
reversed to better separate the data. Residence time in the
chemical reactor was 1.5 h, and 5.5 h in the bioreactor.
[Fe(II)] [ 200 mM. Thicker lines represent model solutions
at lower recycle rates.
residence time), as shown in Fig. 8. However, the magnitude of
the increases was larger for TOC than for PCP degradation. This
higher sensitivity of the model-predicted TOC degradation
over that of PCP degradation can be explained in terms of the
biodegradation kinetics constants for TOC and the scavenging
effects of the PCP by-products on the chemical oxidation of
PCP. The two limiting cases for the Monod-type kinetics are
zero- and first-order kinetics, which occur under high and low
substrate concentrations (compared to ki), respectively. The
TOC degradation would be sensitive to increases in PCP
degradation achieved in the chemical reactor only under
a regime of first-order degradation kinetics (i.e., at low
substrate concentrations) as determined by the magnitude of
the ratio of the Monod’s constants mi,max and ki (which deter-
mines the value of the first-order constant).
5. Conclusions
 The combined system achieved both PCP and TOC degra-
dation. All of the PCP degradation (which was correlated
with dechlorination) occurred in the chemical reactor, while
all of the TOC degradation occurred at the bioreactor. TOC
biodegradation only occurred upon Fenton’s oxidation of
PCP in the chemical reactor.
 Plating tests and bioreactor performance indicated that the
microbial population could not grow on PCP as a single carbon
source. During bioreactor operation, this population partially
mineralized the non-PCP fraction of the TOC without further
significant dechlorination of the Fenton’s reactor effluent.
The observed Fenton’s oxidation intermediates differed in
their biodegradability: TCHQ was completely biodegraded,
while DCMA was only partially biodegraded.
 Increased bioreactor residence time did not yield higher
TOC biodegradation extents, possibly because of (a)
a limited amount of biodegradable intermediates (with
respect to the non-PCP TOC) and/or (b) the partial biode-
gradability of these intermediates.
5713
 Recycling of the waste from the bioreactor back to the
chemical reactor proved useful in achieving additional TOC
degradation, but achieved only marginal additional PCP
degradation.
 The model developed in this work was useful to explain
some of the key experimental findings: the sensitivity of PCP
degradation to H2O2 doses up to 200 mM, the lack of the
system sensitivity to increases in bioreactor residence time,
and the increased sensitivity of TOC degradation to recy-
cling effects, compared to PCP degradation. This supports
the idea that within the model limitations, the lumped
chemical approach presented in this work is a useful tool for
design and study of these combined treatment systems.
Acknowledgments
This research was possible by grants from Consejo Nacional
de Ciencia y Tecnologı́a (Mexico), grant number 5P42ES05949-
05 from the National Institute for Environmental Health
Sciences, the Colorado Institute for Research in Biotech-
nology, and the Colorado Section of the American Water
Research Association.
references
APHA, AWWA and WPCF, 1980. Standard Methods for the
Examination of Water and Wastewater.
Bedard, D., Unterman, R., Bopp, L., Brenna, M., Haberl, M.,
Johnson, C., 1986. Rapid assay for screening and
characterizing microorganisms for the ability to degrade
polychlorinated biphenyls. Applied and Environmental
Microbiology 51 (4), 761e768.
Benitez, F.J., Acero, J.L., Real, F.J., Garcia, J., 2003. Kinetics of
photodegradatio and ozonaton of pentachlorophenol.
Chemosphere 51, 651e662.
Comninellis, C., Kapalka, A., Malato, S., Parsons, S.A., Poulios, I.,
Mantzavinos, D., 2008. Perspective: advanced oxidation
processes for water treatment: advances and trends for R&D.
Journal of Chemical Technology and Biotechnology 83, 769e776.
Duesterberg, C., Waite, D., 2006. Process optimization of fenton
oxidation using kinetic modeling. Environmental Science and
Technology 40, 4189e4195.
Esplugas, S., Contreras, S., Ollis, D., 2004. Engineering aspects of
the integration of chemical and biological oxidation: simple
mechanistic models for the oxidation treatment. Journal of
Environmental Engineering 130 (9), 967e974.
Graves, J.W., Joyce, T., 1994. A critical review of the ability of
biological treatment systems to remove chlorinated organics
discharged by the paper industry. Water S.A 20 (2), 155e159.
Haag, W., Hoigne, J., 1985. Photo-sensitized oxidation in natural
water via. OH radicals. Chemosphere 14 (11/12), 1659e1671.
Hale, D., Reineke, W., Wiegel, J., 1994. In: Chaudry, R. (Ed.),
Biological Degradation and Bioremediation of Toxic
Chemicals. Dioscorides Press, Portland, OR, USA, pp. 74e91.
Hanna, K., Chiron, S., Oturan, M., 2005. Coupling enhanced water
solubilization with cyclodextrin to indirect electrochemical
treatment for pentachlorophenol contaminated soil
remediation. Water Research 39, 2763e2773.
Hirvonen, A., Trapido, M., Hentunen, J., Tarhanen, J., 2000.
Formation of hydroxylated and dimeric intermediates during
oxidation of chlorinated phenols in aqueous solution.
Chemosphere 41, 1211e1218.
5714 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 0 5 e5 7 1 4
Hong, A., Zeng, Y., 2002. Degradation of pentachlorophenol by
ozonation and biodegradability of intermediates. Water
Research 36, 4243e4254.
Huling, S., Arnold, R., Sierka, R., Miller, M., 1998. Measurement of
hydroxyl radical activity in a soil slurry using the spin trap
alpha-(4-pyridyl-1-oxide)-N-tert-butylnitrone. Environmental
Science and Technology 32, 3436e3441.
Kang, N., Lee, D.S., Yoon, J., 2002. Kinetic modeling of Fenton
oxidation of phenol and monochlorophenols. Chemosphere
47, 915e924.
Mantzavinos, D., Psillakis, E., 2004. Review: enhancement of
biodegradability of industrial wastewaters by chemical
oxidation pre-treatment. J Chem Technol Biotechnol 79,
431e454.
Marco, A., Esplugas, S., Saum, G., 1997. How and Why combine
chemical and biological processes for wastewater treatment.
Water Science and Technology 35 (4), 321e327.
Mills, G., Hoffman, M., 1993. Photocatalytic degradation of
pentachlorophenol on TiO2 Particles: identification of
intermediates and mechanism of reaction. Environmental
Science and Technology 27, 1681e1689.
Pera-Titus, M., Garcia-Molina, V., Baños, M., Giménez, J.,
Esplugas, S., 2004. Degradation of chlorophenols by means of
advanced oxidation processes: a general review. Applied
Catalysis B: Environmental 47, 219e256.
Puhakka, J., Melin, E., 1996. In: Crawford, R., Crawford, D. (Eds.),
Bioremediation: Principles and Applications. Cambridge
University Press, Cambridge, MA, USA, pp. 254e299.
Rivas, F., Beltrán, F.J., Frades, J., Buxeda, P., 2001. Oxidation of
p-hydroxybenzoic acid by Fenton’s reagent. Water Research
35 (2), 387e396.
Scott, J., Ollis, D., 1995. Integration of chemical and biological
oxidation processes for water treatment: review and
recommendations. Environmental Progress 14, 88e103.
Sen Gupta, S., Stadler, M., Noser, C., Ghosh, A., Steinhoff, B.,
Lenoir, D., Howitz, C., Schramm, K., Collins, T., 2002. Rapid
total destruction of chlorophenols by activated hydrogen
peroxide. Science 296, 326e328.
Shen, X., Zhu, L., Liu, G., Tang, H., Liub, S., Lib, W., 2009.
Photocatalytic removal of pentachlorophenol by means of an
enzyme-like molecular imprinted photocatalyst and
inhibition of the generation of highly toxic intermediates. New
Journal of Chemistry 33, 2278e2285.
Shuler, M., Kargi, S., 1992. Principles of Biochemical Engineering.
Venkatadri, R., Peters, R., 1993. Chemical oxidation technologies:
ultraviolet light/hydrogen peroxide, Fenton’s reagent, and
titanium dioxide-assisted photocatalysis. Hazardous Waste
and Hazardous Materials 10, 107e149.
Watts, R., Jones, A., Cheng, P., Kelly, A., 1997. Mineral-catalyzed
Fenton-like oxidation of sorbed chlorobenzenes. Water
Environment Research 69 (3), 269e275.
Wong, A.S., Crosby, D.G., 1981. Journal of Agricultural and Food
Chemistry 29, 125.
Zepp, R., Scholtzhauer, P., 1979. In: Leber, P.W.J.a.P. (Ed.),
Polynuclear Aromatic Hydrocarbons. Ann Arbor Science
Publishers, Ann Arbor, pp. 141e156.
Zhao, L., Yu, Z., Peng, P.A., Huang, W., Feng, S., Zhou, H., 2006.
Oxidation kinetics of pentachlorophenol by manganese
dioxide. Environmental Toxicology & Chemistry 25,
2912e2919.
Zimbron, J., Reardon, K., 2009. Fenton’s degradation of
pentachlorophenol. Water Research 43, 1831e1840.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Spatio-temporal variation in trihalomethanes in New South

Wales5

Richard J. Summerhayes a,e,*, Geoffrey G. Morgan a,d, Douglas Lincoln a,

Howard P. Edwards a, Arul Earnest b, Md. Bayzidur Rahman c, Paul Byleveld f,
Christine T. Cowie g, John R. Beard a
a
University Centre for Rural Health, Northern Rivers, University of Sydney, PO Box 3074, Lismore, NSW 2480, Australia
b
Centre for Quantitative Medicine, Duke-NUS Graduate Medical School, Singapore
c
School of Public Health and Community Medicine, University of New South Wales, Sydney, NSW 2052, Australia
d
North Coast Area Health Service, NSW, Australia
e
School of Health and Human Sciences, Southern Cross University, Lismore, NSW 2480, Australia
f
Water Unit, NSW Department of Health, PO Box 798, Gladesville, NSW 2111, Australia
g
Respiratory & Environmental Epidemiology, Woolcock Institute of Medical Research, 431, Glebe Point Road, Glebe, NSW 2037, Australia

article info abstract

Article history: Aim: This paper describes the spatio-temporal variation of trihalomethanes in drinking
Received 16 December 2010 water in New South Wales, Australia from 1997 to 2007
Received in revised form Method: We obtained data on trihalomethanes (THMs) from two metropolitan and 13 rural
22 August 2011 water utilities and conducted a descriptive analysis of the spatial and temporal trends in
Accepted 24 August 2011 THMs and the influence of season and drought.
Available online 1 September 2011 Results: Concetrations of monthly THMs in the two metropolitan water utilities of Sydney/
Illawarra (mean 66.8 mg/L) and Hunter (mean 62.7 mg/L) were similar compared to the
Keywords: considerable variation between rural water utilities (range in mean THMs: 14.5e330.7 mg/L).
Trihalomethanes Chloroform was the predominate THM in two-thirds of the rural water utilities. Higher
Chloroform concentrations of THMs were found in chlorinated water distribution systems compared to
Bromodichloromethane chloraminated systems, and in distribution systems sourced from surface water compared
Drought to ground water or mixed surface and ground water. Ground water sourced supplies had
Disinfection by-products a greater proportion of brominated THMs than surface water sourced supplies. There was
Australia substantial variation in concentration of THMs between seasons and between periods of
drought or no drought. There was a moderate correlation between heavy rainfall and
elevated concentrations of THMs.
Conclusion: There is considerable spatial and temporal variation in THMs amongst New
South Wales water utilities and these variations are likely related to water source, treat-
ment processes, catchments, drought and seasonal factors.
ª 2011 Elsevier Ltd. All rights reserved.

Abbreviations: THMs, trihalomethanes; THM4, total THM; NSW, New South Wales; DBP, disinfection by products; BDCM, bromodi-
chloromethane; DBCM, dibromochloromethane.
5
Institution where this work was performed: University Centre for Rural Health, Northern Rivers, University of Sydney, Lismore, NSW
2480, Australia.
* Corresponding author. Tel.: þ61 419249037.
E-mail addresses: summerhayes.richard@gmail.com (R.J. Summerhayes), geoff.morgan@ncahs.health.nsw.gov.au (G.G. Morgan),
H.Edwards@massey.ac.nz (H.P. Edwards), arul_earnest@hotmail.com (A. Earnest), bayzid@unsw.edu.au (Md.B. Rahman), pbyle@doh.-
health.nsw.gov.au (P. Byleveld), christinec@health.usyd.edu.au (C.T. Cowie).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.045
5716 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
1. Introduction
The provision of safe drinking water through disinfection to
remove water-borne microbiological pathogens causing
typhoid, cholera and gastroenteritis is one of the major
achievements of public (Galal-Gorchev, 1996). An unexpected
consequence of disinfection is the interaction of the disin-
fectant with natural organic matter (NOM) in the source water
forming a range of chemicals collectively called disinfection
by-products (DBPs) (Rook, 1974).
DBP formation is not well understood but is influenced by
the presence of bromide, iodide, pH, temperature, seasonal
and climatic factors. Operational factors in water treatment
also play a role in DBP formation including residency time
within the distribution system, filtration methods used in
removal of NOM, and disinfectant type and dose (Chowdhury
et al., 2009). Trihalomethanes (THMs) were the first DBPs
discovered in 1974 and are a group of organic compounds
formed through reactions between methane (CH4) derivatives
in NOM and chlorine or chloramines (Garrido and Fonseca,
2010). Since their discovery, more than 600 different DBPs
have been reported, with THMs being the most frequently
detected compounds (Richardson et al., 2007).
In 1976, the US National Cancer Institute declared chloro-
form a carcinogen in animals and a suspected carcinogen in
humans (National Cancer Institute, 1976). Exposure to DBPs is
associated with bladder, rectal and colon cancer, ‘suggestive
of a causal inference’ (Hrudey, 2009). While some reproductive
outcomes such as small for gestational age and pre-term
births have also been associated with DBP exposure, the
current evidence is inconclusive (Grellier et al., 2010). In Aus-
tralia, DBPs are not regulated but a guideline value of 250 mg/L
for total THM is recommended and action to reduce THMs is
encouraged, while not compromising disinfection as exposure
to non-disinfected water poses substantially greater health
risk than exposure to low level THMs (NHMRC, 2004). There is
limited published data describing THMs or other disinfection
by-products in Australia. A survey of several DBPs during
1994-95 in 16 cities throughout Australia found that some
Australian drinking water supplies had high THM concentra-
tions (up to 191 ug/L) (Simpson and Hayes, 1998).
In New South Wales (NSW), Australia’s most populated
state, more than 5 million residents (approximately 80% of the
population) use public drinking water as their usual source of
drinking water (Centre for Epidemiology and Research, 2002).
The two largest public water suppliers provide drinking water
to the Sydney/Illawarra (4.2 million people) and Hunter
(516,000 people) metropolitan areas. In rural areas, local water
utilities (largely through local government authorities)
provide drinking water. While the NSW Health Department
recommends that all water utilities collect monthly THM
samples (NSW Health, 2000), only the large metropolitan
utilities of Sydney/Illawarra and the Hunter, and a small
number of rural water utilities conduct regular THM
monitoring.
This paper describes THM concentrations throughout New
South Wales, Australia, covering large metropolitan water
utilities, as well as small to medium size rural water utilities.
Our study assesses geographic differences and temporal
trends in monthly THM data from metropolitan utilities over
several years, and rural utilities for at least one year. During
our study period, much of NSW experienced one of the longest
droughts on record lasting nearly a decade from 1997 to 2006
(Bond et al., 2008) and we also investigated the influence of
drought on THM’s.
2. Methodology
2.1. Water utility data
We obtained data on THMs for the periods 1998 to 2004 for
Sydney/Illawarra region, 1997 to 2004 for Hunter region and
various time periods between 1997 and 2007 for the rural
water utilities summarized in Table 1.
2.1.1. Sydney/Illawarra water utility data
The Sydney Water Corporation (SWC) supplies water from five
surface catchments to the Sydney/Illawarra metropolitan
region, covering an area of 12,700 km2. Sydney/Illawarra has
a three level hierarchical structure, with 14 delivery systems
(average area 241 km2) supplied with surface water treated at
nine water filtration plants. Each delivery system contains
from one to six distribution systems (average area 84.3 km2)
which are treated either by chloramination or chlorination.
Rechlorination occurs within the distribution system. Water is
stored in 180 water supply zones (average area 16.4 km2)
which supply water to homes within the distribution systems.
Limited data was also available on a range of water quality
factors and other DBP’s including: trichloracetonitrile (5
months in 1998 covering 18% of supply zones); six haloacetic
acids (HAAs) (two periods of 12 and 7 months covering 69% of
supply zones); and three HAAs (bromoacetic-, dibromoacetic-
and tribromoacetic acids, 5 months covering all supply zones).
Due to the limited duration of sampling for these DBP’s, and
the lack of comparable data from other water supplies in
NSW, we have not reported these results in detail in the paper,
but summary statistics are provided in the Supplementary
Material AeC.
There are some 3000 THM water sampling sites within the
Sydney/Illawarra distribution systems. Monthly monitoring is
generally conducted at these sites on a three to six-monthly
rotational cycle with a minimum of 3e6 sites operating
within each distribution system (SWC, 2002). Monthly THM
data from all available monitoring sites within each supply
zone was averaged to obtain zone/month THM concentrations
for each month in the study period.
2.1.2. Hunter water utility data
The Hunter Water Corporation supplies the Hunter metro-
politan region via nine water distribution systems covering an
area of 5400 km2. Six distribution systems received surface
water and three received ground water for most of the study
period although prior to 2003 blending of surface and ground
waters from different sources occurred at various time
periods for three distribution systems normally supplied by
surface water. After 2003 two distribution systems, previously
supplied by ground water, were continuously augmented with
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6 5717

Table 1 e Characteristics of NSW metropolitan and rural water utilities included in the study.
Water Utility Water Disinfectant Treatment Process Approx. Number of Study Period
Source Process (Catchment characteristics) pop. (2002) samples (months of data)

Sydney/ Surface Chlorinateda Filtration, flocculation, 600,000 5341 Jan 1998eDec 2004 (84)
Illawarra Chloraminated coagulation, lime. 3,600,000
Hunterb Surface Chlorinated Filtration, coagulation, 452,000 863 Jan 1997eDec 2004 (96)
flocculation, sedimentation,
powdered activated
carbon (PAC)
Ground Chlorinated Aeration, lime, coagulation, 60,000
filtration
Rural
Utilities
1 Surface Chlorinated Filtration, coagulation, 4000 28 Feb 2003eNov 2003 (9)
sedimentation, PAC
2 Surface Chlorinated Filtration, flocculation, 21,000 156 Nov 1997eDec 2005 (89)
sedimentation,
3 Surface Chlorinated Filtration, coagulation, 160,000 72 Nov 2003eJan 2004 (8)
flocculation, sedimentation
4 Surface Chlorinated Filtration 26,000 314 Jan 2001eMay 2007 (62)
5 Surface Chlorinated No treatment 65,000 508 Jan 2001eDec 2004 (54)
6 Surface Chlorinated Lime, aeration 2000 61 Jan 2001eMay 2006 (61)
blended
7 Surface Chlorinated Coagulation, flocculation, 20,000 30 Jul 2001eJan 2004 (30)
sedimentation
8ac Ground Chloraminated Direct filtration, aeration, 4500 1378 Dec 2000eMay 2006 (63)
coagulation, flocculation,
dissolved air flotation
8bc Surface Chlorinated Filtration, coagulation, 11,000 660 Dec 2000eMay 2006 (66)
flocculation, sedimentation
9 Surface Chlorinated Direct filtration 21,000 85 Jan 2000eOct 2006 (80)
10 Surface Chlorinated Direct filtration, coagulation, 53,000 43 Jun 2001eAug 2005 (40)
flocculation, sedimentation
11 Surface Chlorinated Filtration, dissolved air 74,000 1610 Jan 2000-Apr 2006 (61)
flotation, flocculation
12 Surface Chlorinated Filtration, flocculation 8000 43 FebeDec 2003 (11)
13 Surface Chlorinated Filtration, coagulation, 135,000 159 Mar 1999eJan 2006 (86)
flocculation, sedimentation

a In Sydney/Illawarra, 4 of the 33 distribution systems changed disinfection from chlorination to chloramination in June 2003.
b In Hunter, two distribution systems using ground water were augmented with ground and surface waters from other sources after 2003, and
prior to 2003, some systems supplied with surface water were occasionally augmented with blended ground and surface waters from other
sources.
c Water utility 8 includes a chloraminated distribution system with ground water supply (8a) and a chlorinated distribution system with
a surface supply. Data for 8a should be treated with caution as majority of values were below <5 mg/L detection limit.

blended surface and ground water from different sources
(HWC, 2004). A fixed sampling site is located within each
distribution system towards the extremities of the systems,
providing distribution/month THM values.
2.1.3. Rural water utility data
We surveyed all 106 utilities in rural NSW requesting infor-
mation and data on DBPs and water quality parameters.
Ninety-four rural water utilities (89%) responded to the survey
and 27 (26%) indicated they collected some data on THMs. We
received data on THMs from 25 (24%) rural water utilities of
which 13 (12%) had sufficient THM data to be included in our
analysis. Five of these rural utilities supplied additional
monitoring data on a range of other water quality parameters,
however these data covered limited durations and were
insufficient to assess seasonal trends and geographic differ-
ences and in keeping with the objectives of the paper are not
reported in detail, although they are briefly summarised in the
Supplementary Material A.
The size and complexity of rural water utilities varies, and
THM sampling varied from one-off short term sampling
periods to routine quarterly or monthly sampling covering
various time periods between 1997 and 2007 (see Table 1). All
sampling was conducted at random sites throughout the
distribution systems. Water utility #6 provided data only on
THM4 from sample points within the distribution system. A
separate one-month survey in utility #6 including post-
treatment samples indicated that chloroform comprised
approximately 86% and BDCM 13% of the THM4 concentration
entering the distribution system (these survey data not
included in the final analysis but are provided in
Supplementary Material D). Rural utility #8 includes a distri-
bution system which is chloraminated and mainly sourced
from ground water, and a chlorinated distribution system
5718 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
sourced from surface water. We designated the two distribu-
tion systems as #8a and #8b respectively for reporting. The
source water, disinfection, treatment types, number of THM
samples and sampling period for each of the water utilities
included in our analysis are shown in Table 1.
We included only those rural water utilities with monthly
monitoring data covering all four seasons in at least one year
(DecembereFebruary ¼ summer, MarcheMay ¼ autumn,
JuneeAugust ¼ winter and SeptembereNovember ¼ spring) so
that seasonal variation could be assessed. For each rural water
utility with sufficient data to be included in the study we
averaged available monthly THM concentrations to obtain
utility/month concentrations.
2.1.4. Trihalomethane monitoring
Water samples were analysed for THMs by Sydney Water
Corporation laboratories using Standard Methods for the
Examination of Water and Waste Water (20th Ed.), Method
6200B (modified) and USEPA Method 8260 (modified) (APHA-
AWWA-WPCF, 1998; USEPA, 2008); in the Hunter Water
Corporation based on the USEPA standard Method 8260
(USEPA, 2008) and in rural water utilities using the Standard
Methods for the Examination of Water and Waste Water (18th
ed) Method 6232 (APHA-AWWA-WEF, 1992). Further details of
the analytical techniques are provided in the Supplementary
Material E.
Multiple values below the detection limit were recorded
within each water utility over the study period. The limit of
detection for individual THMs was usually <1 mg/L, with
occasional higher limits of detection reported. We used
a similar approach to other studies and substituted samples
flagged as below the detection limit with two-thirds of the
detection limit value, which is the approximate mean of a log-
normal distribution (Whitaker et al., 2003). Bromoform was
found to be below the detection limit more than 75% of the
time across most utilities (metropolitan and rural), and
therefore we examined bromoform only as part of total
brominated THMs (BrTHM) and the sum of the four trihalo-
methanes (THM4) (percentage of observations below detec-
tion values provided in Supplementary material F).
2.2. Rainfall data
Data on rainfall was obtained from the Sydney Catchment
Authority, and the Australian Bureau of Meteorology website
(Bureau Of Meteorology, 2010). Data on monthly drought
status for NSW districts containing the water utilities was
obtained from the NSW Department of Primary Industries,
with each month classified as ‘no drought’, ‘marginal’ or
‘drought’. Drought status is defined by a number of criteria
including monthly rainfall, temperature, frost and evapora-
tion, poor pasture biomass, soil moisture and livestock.
Marginal conditions include rainfall for the previous 3 months
within or below average for the three month rainfall decile
and surface water supplies less than 50% of normal for the
time of year. Drought conditions include rainfall for the
previous 6 months within or below average for the 6-month
rainfall decile and surface water supplies less than 30% of
normal for the time of year (NSW Department of Primary
Industries, 2010).
2.3. Statistical analysis
We produced a range of descriptive statistics using log
transformed THM concentrations due to the skewed THM
distribution and back-transformed the results for reporting.
Differences in the means for season and drought periods were
assessed using KruskaleWallis tests. Correlations between
THMs and rainfall lagged up to one month were conducted
using the Spearman correlation coefficient. The selection of
a one month lag is based on a 30e40 day residency time for
Prospect Reservoir in the Sydney/Illawarra supply (Hamilton
et al., 1995). Data management was conducted in Excel
(version 2003) and Access (version 2003) and all statistical
analyses were conducted using SAS version 9.1.3 (SAS Insti-
tute Inc., Cary North Carolina USA).
3. Results
3.1. Spatial variation
Table 2 summarises descriptive statistics for individual THMs
and THM4 concentrations in treated water for all the selected
NSW water utilities. The mean THM4 concentrations in Syd-
ney/Illawarra (66.8 mg/L) and Hunter are similar (62.7 mg/l).
Water treated by chlorination in Sydney/Illawarra had higher
mean THM4 concentrations (81.1 mg/L) than water treated by
chloramination (50.8 mg/L). The Prospect South delivery
system in Sydney/Illawarra switched from chlorination to
chloramination in July 2003 and the mean monthly chloro-
form concentration decreased substantially from 59.5 mg/L
during the chlorination period (1998 to mid-2003) to 23.0 mg/L
( p < 0.001) during the chloramination period (mid-
2003e2004). In the Hunter, chlorinated surface water had
higher mean THM4 concentrations (74.4 mg/L) than chlori-
nated ground water (36.1 mg/L). From 2003, Hunter augmented
two ground supplies and three surface supplies continuously
with blended ground and surface waters in response to
drought conditions. The mean THM4 in the ground water
supplies increased from 17.9 to 40.6 mg/L with augmentation
and from 66.4 to 74.2 mg/L in the surface supplies.
The majority of the rural water utilities reporting in the
study use chlorination, one system used chloramination
together with chlorination. Rural NSW water utilities gener-
ally use surface water, with some using ground water or
a blend of ground and surface waters. During the study period
there was large variability in the monthly mean THM4
concentration in treated water between rural water utilities.
The THM4 concentration of the rural utility with the
minimum (#8a: 15 mg/L) and maximum (#6: 331 mg/L) monthly
concentrations were substantially different compared to the
remaining 11 utilities which ranged from of 63.7 mg/L to
189.1 mg/L, although this trimmed range still represents a 3
fold difference in monthly THM4 concentrations.
Chloroform is the main component of total THM in chlo-
rinated (66%) and chloraminated (59%) surface water in Syd-
ney/Illawarra. Chloroform is also the main component of total
THM in chlorinated surface water in the Hunter (60%) and in
most rural utilities with the exception of utilities #2 (29%), #3
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6 5719

Table 2 e Concentrations of trihalomethanes in treated drinking water in water utilities in New South Wales.
Water Utility N Mean THM4 Chloroform BDCM DBCM

Min P95 Max Mean Min Max Mean Min Max Mean Min Max

Metropolitan
Sydney/Illawarra 563 66.8 22.7 114.6 196.7 40.1 4.5 162.5 16.5 3.0 46.0 7.1 0.7 27.3
Chloraminated 278 50.8 22.7 81.3 199.0 28.3 9.0 88.0 14.2 3.0 39.0 5.8 0.7 31.0
Chlorinated 340 81.1 25.0 127.0 196.7 50.7 4.5 162.5 18.6 5.7 46.0 8.1 0.8 27.3
Hunter 72 62.7 10.1 105.2 114.6 35.2 0.7 96.3 15.2 0.7 31.2 9.1 0.9 24.9
Ground 18 36.1 10.1 74.3 74.3 16.0 0.7 46.8 8.3 0.7 20.8 5.9 2.8 9.2
Blended 30 72.6 14.7 106.4 134.0 33.4 1.0 78.3 20.7 0.7 34.3 15.2 3.7 27.0
Surface 48 74.4 30.1 105.2 131.0 44.5 20.5 96.3 18.1 3.1 30.0 10.0 0.9 24.0
Rural Utilities
1a 9 189.1 67.5 290.7 290.7 104.8 29.7 190.1 48.8 22.5 94.1 19.8 4.7 76.3
2 89 114.9 21.0 231.0 346.0 33.6 2.0 172.0 32.5 4.5 104.7 32.7 2.8 128.5
3 8 81.7 65.4 102.3 102.3 29.5 13.3 47.9 24.9 20.6 31.2 16.6 6.9 24.7
4 62 106.1 21.1 170.2 245.4 45.0 16.8 108.8 26.8 0.7 56.7 16.0 0.7 56.7
5 54 73.4 23.8 120.1 122.3 48.2 13.9 90.0 16.5 6.3 24.2 5.7 2.7 10.3
6a 61 330.7 63.0 501.0 586.0 na na na na na na na na na
7 30 87.5 22.7 147.3 213.7 72.8 3.0 200.0 11.9 3.0 20.0 1.4 1.0 2.0
8a 63 14.5 13.1 17.8 23.6 4.4 3.3 11.2 3.4 3.3 5.4 3.3 2.5 4.1
8b 66 102.2 44.9 145.4 221.2 63.9 22.2 159.2 25.6 8.2 51.3 9.1 3.3 25.3
9 80 99.3 39.0 157.9 180.0 46.8 12.0 112.8 33.4 12.0 57.0 17.9 8.0 56.0
10 42 63.8 13.3 86.3 120.0 34.1 3.3 68.0 16.2 3.3 39.0 9.5 1.0 32.0
11 61 63.7 18.4 102.3 128.9 33.0 5.7 86.6 16.6 3.8 32.2 9.2 3.3 32.0
12a 11 138.5 52.3 286.0 286.0 87.1 17.3 155.0 31.3 13.3 78.0 12.3 1.3 48.0
13 86 94.9 2.7 152.7 229.0 54.5 0.7 110.0 26.7 0.7 76.0 11.1 0.7 64.0

All values in mg/L ¼ micrograms per litre, N ¼ number of observations, Min ¼ minimum, P95 ¼ 95th percentile, Max ¼ maximum, na ¼ not
available.
Sydney/Illawarra mean: mean of annual zone means.
Hunter mean: mean of annual distribution system means.
Rural water utilities mean: mean of monthly utility means.
Water utility 8 includes a chloraminated distribution system with ground water supply (8a) and a chlorinated distribution system with a surface
supply. Data for 8a should be treated with caution as majority of values were below <5 mg/L detection limit.
Water utility 6 only has data for THM4.
a Utilities with 95th percentile THM4 concentrations above the Australian National Drinking Water Guideline value of 250 mg/L.

(36%) and in the ground water sources in the Hunter (56%) and
#8a (30%) which had higher proportions of brominated THMs.
3.2. Rainfall
Fig. 1 illustrates the temporal variation in Sydney/Illawarra
chloroform and brominated THM concentrations and mean
monthly rainfall. Prior to October 1998 brominated THM
concentrations were generally higher than chloroform, then
decreased until mid 1999 and have since remained relatively
constant. Chloroform concentrations decreased steadily from
mid-1998 but showed considerable variability, and from late
2003 concentrations were similar or lower than brominated
species. Several peaks and troughs in chloroform concentra-
tion coincide with heavy rainfall events. We found a moderate
correlation between overall monthly THM4 and rainfall (lag
one month, r ¼ 0.35, p < 0.01) for the Sydney/Illawarra, with
higher correlations within some of the five Sydney/Illawarra
catchments (Blue Mountains/Cascade catchment: r ¼ 0.57,
p < 0.001; Upper Nepean/Illawarra catchment r ¼ 0.47,
p < 0.001).
Fig. 2 shows the temporal variation in chloroform, bromi-
nated THMs and rainfall in the Hunter region. Chloroform
concentrations are generally higher than brominated THMs,
with occasional large peaks in chloroform and brominated
THMs from 2001. We found no correlation in the Hunter region
between rainfall (lag one month) and mean THM concentra-
tion in distribution systems supplied by ground water, and
a moderate correlation in distribution systems supplied by
surface waters (r ¼ 0.59, p < 0.01).
We generally found moderate correlations between rainfall
and THM4 concentrations in the rural utilities located in sub-
tropical coastal floodplains including utilities #11 (r ¼ 0.30,
p < 0.02), #8b (r ¼ 0.58, p < 0.001) and #5 (0.67, p < 0.001). We
generally found little correlation with mean monthly rainfall
(lag 1 month) and THM in the other rural utilities (data not
shown).
3.3. Drought
We examined the difference in the concentration in THMs in
periods of drought compared to marginal and no drought and
the results are summarised in Table 3. Sydney/Illawarra was
in drought for 18 months (21.4%) of the 84 month study period
and continuously in marginal drought from September 2002
until mid-2007. There was a consistent significant decrease in
mean THM4 concentrations during drought months
compared to non-drought (THM4: 52.6ug/L compared to
5720 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6

90 120.0
BrTHM Chloroform Rainfall
80
Mean Monthly Chloroform and Brominated THM

100.0

Mean Monthly Catchment Rainfall (mm)

70

60 80.0
Concentration (ug/L)

50
60.0
40

30 40.0

20
20.0
10

0 0.0

Fig. 1 e Temporal variation in Sydney/Illawarra monthly chloroform and brominated THM concentration and monthly
rainfall.

75.9ug/L, p < 0.001), and this was reflected in chloroform and
BDCM, while DBCM concentration increased. The majority of
sampling for HAAs in Sydney/Illawarra occurred in a non-
drought period, however three HAAs (bromoacetic-;
dichloroacetic- and trichloroacetic acids) showed significant
decline during drought compared to no drought consistent
with reductions in THMs, except DBCM (see Supplementary
Material C).
Hunter was in drought for 19 months of the 96 month study
period and only the blended distribution systems showed
a significant THM4 decrease during drought compared to no
drought (THM4: 69.8 mg/L compared to 46.4 mg/L, p < 0.05), re-
flected in chloroform, BDCM and DBCM. THM concentrations
in rural water utilities during drought periods compared to
non-drought periods were extremely variable. Five rural water
utilities that were in drought continuously for 12 months or
longer experienced significantly reduced chloroform concen-
trations compared to periods of no drought (#2, #4, #9, #10 and
#13), and these same utilities experienced significantly
increased DBCM concentrations.
3.4. Seasonal variation
We found considerable seasonal variability in mean THM4
concentrations between locations and these results are sum-
marised in Table 4. In Sydney/Illawarra there was a significant

100 450
chloroform BrTHM Rainfall
90 400

80
Mean Monthly Catchment Rainfall (mm)
Mean Monthly Chloroform and Brominated THM

350

70
300
60
Concentration (ug/L)

250
50
200
40
150
30

100
20

10 50

0 0

Fig. 2 e Temporal Variation in Hunter monthly chloroform and brominated THM (BrTHM) concentration and monthly
rainfall (mm).
Table 3 e Mean monthly trihalomethane concentrations during drought compared to no drought and marginal drought, NSW water utilities.
Water Utility Time (months THM4 Chloroform BDCM DBCM

No Marginal Drought No Marginal Drought No Marginal Drought No Marginal Drought No Marginal Drought
Drought n (%) n (%) Drought Mean Mean Drought Mean Mean Drought Mean Mean Drought Mean Mean
n (%) Mean Mean Mean Mean

Sydney/Illawarra 50 (59.5) 16 (19.1) 18 (21.4) 75.9 51.1 52.6** 49.0 27.5 27.6** 17.7 13.7 14.1** 6.6 7.4 8.0**
Chloramination 59.4 37.0 36.5** 36.0 17.4 16.5** 16.1 11.3 11.4** 5.3 6.3 6.6**

w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
Chlorination 87.2 66.7 67.8** 57.8 38.7 37.9** 18.8 16.2 16.7** 7.5 8.7 9.4**
Hunter 60 (62.5) 17 (17.7) 19 (19.8) 58.4 73.6 66.4* 33.2 41.6 35.4 13.9 18.1 17.0 8.2 10.7 10.6
Ground 31.5 56.1 50.2** 12.5 22.9 20.7** 6.8 16.2 14.4** 5.9 11.6 9.8**
Blended 69.8 70.6 46.4* 36.7 35.4 19.0** 20.0 19.8 11.6* 11.5 12.1 9.9
Surface 72.3 84.2 80.6* 46.4 54.5 47.5 16.1 18.1 20.0** 8.4 9.5 11.2*
Rural Utilities
2 26 (27.1) 19 (19.8) 51 (53.1) 149.9 97.6 113.5* 55.6 36.8 25.3* 32.6 32.1 30.9 32.4 23.2 36.6*
3 1 (12.5) 4 (50.0) 3 (37.5) 67.4 86.7 88.8 14.3 30.8 39.4 22.7 26.0 26.7 20.2 21.9 17.2
4 10 (16.1) 5 (8.10) 47 (75.8) 151.2 134.5 103.8* 92.3 92.9 41.0* 21.8 30.2 31.2 5.8 11.6 22.2*
5 24 (44.4) 16 (29.6) 14 (25.9) 74.8 81.3 82.6 49.2 56.9 57.7 17.7 17.0 17.6 7.2 6.7 7.2
6 23 (37.7) 20 (32.8) 18 (29.5) 287.5 379.3 331.9** a a a a a a a a a
7 12 (40.0) 6 (20.0) 12 (40.0) 86.0 89.2 88.1 74.3 74.3 70.6 9.1 12.0 14.8** 1.3 1.6 1.5
8b 24 (36.4) 7 (10.6) 35 (53.0) 97.1 108.8 105.9 56.8 70.9 68.9 26.8 26.3 25.3 10.1 8.3 8.7
9 29 (36.2) 16 (20.0) 35 (43.8) 100.4 97.3 100.3 54.8 40.5 43.7* 32.5 34.8 33.6 12.7 20.2 21.1**
10 10 (25.0) 7 (17.5) 23 (57.5) 63.7 56.9 67.5 48.3 29.5 30.6* 9.5 15.4 19.7* 2.6 8.7 13.0**
11 38 (62.3) 15 (24.6) 8 (13.1) 63.4 69.0 71.9 33.5 39.7 34.0 16.9 18.5 18.6 9.6 8.8 14.8
12 - (0.0) 1 (09.0) 10 (91.0) na 157.9 143.1 na 135.0 94.1 na 20.3 33.8 na 1.3 14.1
13 47 (54.7) 22 (25.6) 17 (19.8) 83.9 112.7 104.5* 50.7 67.5 48.5* 22.9 33.0 31.3* 9.4 10.4 18.7*

* Significant at p < 0.05 level.

** Significant at p < 0.001 level, na e no data were available Means are reported in mg/L.
Water utility #12 was in marginal or drought only.
Water utilities #1 and #8a not shown as utility #1 was continuously in drought during the study period and water utility #8a values were below threshold and showed no variation between the means.
a Water utility #6 reported THM4 data only.

5721
5722 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6

Table 4 e Seasonal variation in the mean trihalomethane concentration in water utilities in New South Wales.
Summer Mean THM4 Autumn Mean Winter Mean Spring Mean Summer-Winter Difference (%)

Sydney/Illawarra 69.9 72.6 64.3 60.3 þ5.6 (9.0)

Chloramination 54.8 58.0 54.0 52.0 þ0.8 (1.5)
Chlorination 80.0a,byy 79.3a,byy 71.9 69.8 þ8.1 (11.3)
Hunter 60.8 66.8 60.5 62.5 þ0.3 (0.5)
Ground 34.2 39.6 39.5 39.8 5.3 (15.5)
Blended 43.1a,cy 74.8 66.7 63.1 23.6 (54.8)
Surface 75.4 80.6 77.4 71.5 þ2.0 (3.0)
Rural Utilities
1 233.7 183.6 165.4 189.3 þ68.3 (41.3)
2 146.0ay 112.3 90.9 104.8 þ55.1 (60.6)
3 82.0 66.2 90.2 79.7 8.2 (9.1)
4 110.8 122.5a,y 87.4 86.0 þ23.4 (26.8)
5 89.7ayy 85.5ay 46.2 62.0 þ43.5 (94.2)
6 340.9 313.8 324.3 345.1 þ16.6 (5.1)
7 80.5 96.9 77.1 97.8 þ3.4 (4.4)
8b 124.1ayy 107.0ay 79.8 91.3 þ44.3 (56.5)
9 97.6 102.2 99.5 97.7 1.9 (1.9)
10 64.1 70.7 60.1 64.2 þ4.0 (6.7)
11 72.1ay 67.3 48.7 58.1 þ23.4 (48.0)
12 212.5 129.2 89.5 126.1 þ123.0 (137.4)
13 93.7 92.0 97.3 94.7 3.6 (3.7)

Means and difference between summer and winter means are reported in mg/L.
y significant at the p < 0.05 level.
yy significant at the 0.001 level.
Water utility 8a values were below threshold and showed no variation between the means.
a Difference in means compared to winter.
b Difference in means compared to spring.
c Difference in means compared to autumn.

increase ( p < 0.001) in mean THM4 concentration in chlori-
nated water during summer (mean 80.0 mg/L) compared to
winter (mean 71.9 mg/L) and spring (69.8 mg/L), but there was no
difference in chloraminated distribution systems. In the
Hunter THM4 concentrations in blended water increased
significantly ( p < 0.05) by 23.6 mg/L in winter compared to
summer, while there was generally a small increase in surface
and ground water concentrations. There was substantial
variation in the seasonality of THM4 concentrations between
rural water utilities. Ten of the 13 rural water utilities had
higher mean THM4 concentrations in summer compared to
winter ranging in increases of 4e123 mg/L, and this increase
was significant ( p < 0.05) in four utilities.
4. Discussion
4.1. Spatial variation
We found the concentration of THM4 and speciation to be
similar for the Sydney/Illawarra (THM4 ¼ 67 mg/L, 60% chlo-
roform) and Hunter (THM4 ¼ 63 mg/L, 56% chloroform)
metropolitan water utilities.
We found considerable variation in THM concentrations
and speciation between rural water utilities, with mean
monthly THM4 concentrations ranging from 15 to 331 mg/L
(29e86% chloroform). The THM4 concentration of the rural
utility with the minimum (#8a: 15 mg/L) and maximum (#6:
331 mg/L) in the range were substantially different compared to
the remaining 11 utilities and suggests that the factors influ-
encing THM4 concentrations in utilities #8a and #6 are
substantially different to the other 11 rural utilities. Differ-
ences in geography and catchment characteristics may
contribute to variation in THM concentrations throughout
NSW. Utilities in the semi-arid floodplains in the far-west of
the State (#1, #2 and #12) had higher THM4 concentrations
(range of means: 114e189 mg/L) compared to other water
utilities (range of means: 15e106 mg/L) with the exception of
water utility #6 (mean 331 mg/L) which is in a sub-tropical
coastal floodplain area. The 95th percentile for THM4
concentration for three rural water supplies was above the
Australian guideline value of 250 mg/L (Table 2) suggesting
these utilities need to implement strategies to reduce THM
concentrations (NHMRC, 2004).
The Australian guideline value for THM4 of 250ug/L is
higher than a number of other developed countries including
United Kingdom, Japan (100 mg/L), USA (80 mg/L) France
(30 mg/L) and Germany (10m/L) (Rizzo et al., 2005). The large
variation in mean THM4 in rural NSW is similar to that re-
ported in a survey of DBPs in North Virginia, USA prior to the
introduction of the 1979 interim Disinfection By-Product Rule
of <100 mg/L (USEPA, 1979) which found mean THM4
concentrations of 249 mg/L (range 40e531 mg/L) in 1975e76 and
173 mg/L (range 43e889 mg/L) in 1976e77 (Hoehn and Randall,
1979). In a study in Turkey, where THMs are not regulated,
mean THM4 concentrations of 159 mg/L and 129 mg/L were
reported (Rizzo et al., 2005). Similar large variation has also
been found in rural communities in Alberta, Canada during
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
Autumn 2000 with mean THM3 (excluding bromoform)
ranging from 44m/L to 210m/L (Charrois et al., 2004).
THM4 levels in NSW utilities were generally higher than
levels found in countries where THMs are regulated. A 1988-89
survey reported median THM4 value of 39 mg/L for 35 utilities
across the USA (Krasner et al., 1989). A survey of 113 systems
in North Carolina in 2004-05 reported an average THM4 of
40.8 mg/L. In the United Kingdom, a study of data from 1992 to
1996 reported mean THM4 concentrations of 46 mg/L (Keegan
et al., 2001).
Chloroform generally comprised the majority of the total
THM in treated surface water in NSW (52%e86%). The domi-
nation of brominated THMs in treated ground water in the
Hunter (56%) and rural utility #8a (70%) is likely due to the
coastal aquifers having higher levels of bromide compared to
the surface supplies (Kampioti and Stephanou, 2002). The
surface water supplies of utility #3 (64% brominated THMs)
lies in a coastal floodplain area which may be affected by
saltwater intrusion; while utility #2 (71% brominated THMs)
lies in an ancient sea-bed which has been noted to have high
bromide concentrations (Richardson, 2005).
Chloraminated systems generally had lower THM4
concentrations (utility #8a mean THM4 14.5 mg/L, Sydney
mean THM4 50.8 mg/L) than chlorinated systems. A change
treatment from chlorination to chloramination in a delivery
system in Sydney resulted in a 159% reduction in THM4.
Bougeard et al. (2010) in a 2008 study in the United Kingdom
reported a 92% reduction in THM4 with a shift in disinfection
from chlorination to chloramination.
A major strength of this study lies in the large number of
observations covering a lengthy time period, incorporating
periods of long-term drought, drought-free periods and heavy
rainfall events. While we were only able to access THM data on
a small proportion of the total number of rural NSW water util-
ities the surveyed utilities cover a wide range of geographic and
climatic regions encountered across the Australian continent.
4.2. Seasonal variation
We found substantial seasonal variation in THM4 concentra-
tions in chlorinated water in Sydney/Illawarra and in many of
the rural NSW water utilities, with generally higher concen-
trations in summer/autumn and lower concentrations in
winter, especially in semi-arid locations in the far-west of NSW
(utilities #1, #2 and #12). The ground water supplied systems in
the Hunter showed little THM variation between seasons and
these results are consistent with overseas studies. A Canadian
survey of THMs showed a more than two-fold variation during
summer and winter in chlorinated systems (62.5 mg/L
compared to 33.5 mg/L) and chloraminated systems (32.8 mg/L
compared to 13.7 mg/L), however in ground water supplies there
was little variation between seasons (Williams et al., 1995).
Another study in China in 2003 found THM levels to be 50 mg/L
in autumn and around 10 mg/L in spring in chloraminated
waters and noted variations in different organic matter
concentrations and in the dynamics of algae/plankton
production in the different seasons (Chen et al., 2008).
Summer is considered the most challenging period for
treating water and maintaining water quality and the
5723
magnitude of seasonal fluctuations may be dependent on the
water source (McGuire and Meadow, 1988).
4.3. Rainfall
Several peaks and troughs in chloroform concentration coin-
cide with heavy rainfall events in Sydney/Illawarra. One such
event was the rains in mid 1998 that broke a drought lasting
from 1992 to 1998 and resulted in large inflows of contami-
nants and organic matter into the Sydney/Illawarra catch-
ment (Cox et al., 2003). The moderate correlations found
between rainfall and THM reported in our results are consis-
tent with the correlations found in a North Virginia, USA study
(Hoehn and Randall, 1979). Heavy drought breaking rainfall is
associated with high turbidity, increased inflows, soil leaching
of NOM and microbiological contamination from degraded
catchments (Stein, 2000).
4.4. Drought
Droughts lasting several months to several years are a regular
feature of the Australian environment. Much of NSW experi-
enced one of the longest droughts on record lasting nearly
a decade from 1997 to 2006 (Bond et al., 2008). Sydney and parts
of NSW experienced drought from 1992 to 98 with a significant
wet period in mid-1998, then entered into drought again from
2002 until early 2011. During this period the Sydney catchment
storage (2 million megalitres) fell from 91% in 2000 to 32.5% in
mid-2007, in some rural catchments water levels fell below
30%, whereas the Hunter maintained a storage capacity above
60% by augmenting its water supply from 2002 with ground
and surface water from additional sources in response to
drought conditions (HWC, 2004). Inflows into major river
systems in NSW were some of the lowest on record and some
floodplains and wetlands had not been flooded during the
decade of drought (Murphy and Timbal, 2008).
While the effects of drought were variable, Sydney/Ill-
awarra and five of the 13 rural utilities experienced consistent
decrease in chloroform and an increase in DBCM during
drought periods. The effect of drought was also evident in
both chlorinated and chloraminated water as illustrated by
decreased chloroform concentrations in the Prospect South
distribution system during periods of drought when the
system was chlorinated and when it changed to being chlor-
aminated. The blended water distributions systems in the
Hunter showed decreased THM concentrations during
drought periods, while the ground and surface water systems
showed increased THM concentration.
Our findings in NSW are broadly similar to a UK study that
also that found decreases in chloroform concentration in
water supplies during drought (Whitaker et al., 2003). A study
in Greece showed a lower mean chloroform concentration
during an intense drought period (2.27 mg/L) compared to no
drought (10.22 mg/L) and an increased mean DBCM concen-
tration (drought 13.76 mg/L; no drought 1.38 mg/L) from
increased bromide through saltwater intrusion and ground
water augmentation (Kampioti and Stephanou, 2002). A North
Virginia, UAS study reported a large variation in THMs during
a period of no drought from 1975 to 76 (mean THM4 249 mg/L)
compared to severe drought from June 1976 to November 1977
5724 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
(mean THM4 173 mg/L). The mean concentration of chloroform
decreased from 246 to 152 mg/L in the drought compared to no
drought while a small increase in BDCM from 19 to 21 mg/L was
reported (Hoehn and Randall, 1979). Drought can cause long-
term variations in DBP formation and the 1997-98 US Infor-
mation Collection rule survey found that long-term seasonal
factors such as drought, flood, hurricanes may account for
seasonal variation in water quality both in the source water
and in the distribution system although this was not assessed
in detail (McGuire et al., 2002).
Our results support previous findings that seasonal and
long-term effects of drought can have complex effects on the
nature and presence of NOM resulting in either an increase or
decrease in NOM which may in turn affect the nature and
concentration of THMs, suggesting that the effects of drought
may be catchment specific (Williamson et al., 1999). Algal
blooms, common in Australian waters may also play a large
role during warmer months and periods of low rainfall. Algal
blooms have been correlated with THM levels and regarded as
an important precursor for DBP production (Hoehn and
Randall, 1979; Chen et al., 2008). Although we had no data on
algae levels, algal blooms occurred in several rural water
utilities in coastal floodplain/estuarine areas (utilities #3, #6,
#13) and sub-tropical coastal areas (#11, #8b) during warmer
months and periods of low rainfall and drought. The
concentration of chloroform in all these surface supplied
distribution systems was higher during drought periods. Rural
utility #6 suffered a number of major algal blooms during the
study period and recorded the highest concentrations of
THM4 amongst the study locations.
A decrease in organic matter can occur during drought due
to reduced run-off from catchments and inflows, which can
lead to lower turbidity and settling of sediments and organic
matter (Bond et al., 2008; Murphy and Timbal, 2008). Degra-
dation of organic matter over time during lengthy dry periods
and slower travel time of water compared to wet events can
also reduce organics in the water (Personal communication
Sydney Catchment Authority 2011). Drought can also influ-
ence ground water quality due to declines in aquifer levels and
increased salinity causing bromide concentrations to rise
which can affect THM speciation (Krasner et al., 1994).
While there is much speculation on the effects of climate
change on drinking water quality (Bates et al., 2008), increases
in the frequency and extent of drought affected areas and
flooding from drought breaking rains are expected in south
eastern Australia (ABS, 2008; Dore, 2005). Water treatment
authorities respond to the higher risk of microbial contami-
nation of the raw source water during or shortly after heavy
rainfall by using higher disinfectant concentrations which in
turn can promote THM formation (Eikebrokk et al., 2004).
Changes in climatic factors related to climate change are
likely to create additional operational challenges for water
utilities in Australia, particularly in rural areas of NSW with
limited resources (Hurst et al., 2004; Soh et al., 2008).
5. Conclusion
We found considerable variation in THM concentration
between rural water utilities in NSW with some utilities
experiencing elevated concentrations above current national
THM4 guideline values indicating that these utilities require
action to reduce THM’s, while not compromising disinfection.
We obtained THM data from a small proportion of rural NSW
water utilities and our results suggests that elevated THM4
concentrations may occur in water utilities across rural NSW,
and rural Australia. THM concentrations in the chloraminated
systems in Sydney and the one rural utility were generally
lower than the chlorinated water supply systems.
We found considerable variation in THM’s between and
within utilities associated with extended drought periods
experienced during our study, with some utilities showing
decreases in chloroform and BDCM and increases in DBCM.
While we generally found higher concentration of THM during
the warmer compared to cooler months the overarching
influence of drought on THMs makes it difficult to identify the
principal drivers. Unfortunately our data lacked detailed data
on NOM, algae and catchment characteristics and inflows
which may help understand the possible drivers for these
variations.
The Australian Drinking Water Guidelines are currently
under review and at this time no change has been proposed to
the guideline value of 250 mg/L for THM4 (NHMRC, 2009). The
Guidelines note that a high concentration of THM4 is a good
indicator that other DBPs may be present. However data on
the occurrence, nature and concentrations of other DBPs are
scarce due to the lack of comprehensive survey data in NSW
and Australia. We recommend improved monitoring and
reporting of DBP’s and the collection of good data on catch-
ment characteristics and treatment to enable the assessment
of contributing factors to elevated concentrations of DBPs,
especially in rural Australia.
Competing financial interests
None identified.
Acknowledgements
The authors would like to acknowledge the following organi-
sations and individuals for their work on this manuscript:
Dr Mark Angles, Dr Peter Cox, Dr Vicky Whiffin, Mr David
Holland from Sydney Water Corporation and Adam Lovell
from Water Services Association of Australia for expert advice
on the Sydney/Illawarra water utility.
Mr Bruce Cole and Ms Pam O’Donoghue from Hunter Water
Corporation for expert advice on the Hunter water utility,
Mr Peter Littlejohns from the Sydney Catchment Authority
for expert advice on drought effects in Sydney catchments in
NSW.
We also thank Dr Nel Glass and Dr Stephen Kermode from
Southern Cross University, Ms Therese Dunn and Mr Paul
Houlder for their valuable contributions to this manuscript.
We wish to acknowledge the support from the Australian
Research Council Linage Grant (LP0348628) and the Network
for Spatially Integrated Social Science. This work is part of
a PhD thesis by Richard Summerhayes.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
Appendix. Supplementary material
Supplementary data associated with this article can be found
in the online version, at doi:10.1016/j.watres.2011.08.045.
references
APHA-AWWA-WEF, 1992. Standard Methods for the Examination
of Water and Waste Water, American Water Works
Association. Water Environment Federation, Washington DC.
APHA-AWWA-WPCF, 1998. Standard Methods for the
Examination of Water and Waste Water, P. 1134, American
Water Works Association. Water Pollution Control Federation,
Washington DC.
ABS, 2008. Year Book Australia, 2007. Australian Bureau of
Statistics (ABS), Australian Government Publishers, Canberra,
Australia.
Bates, B.C., Kundzewicz, Z.W., Wu, S., Palutikof, J.P. (Eds.), 2008.
Climate Change and Water. IPCC Secretariat, Geneva.
Bond, N.R., Lake, P.S., Arthington, A.H., 2008. The impacts of
drought on freshwater ecosystems: an Australian perspective.
Hydrobiologia 600, 3e16.
Bureau Of Meteorology, 2010. Climate Data Online. Australian
Government, Canberra. URL: http://www.bom.gov.au/climate/
data/ (last accessed December 2010).
Bougeard, C.M.M., Goslan, E.H., Jefferson, B., Parsons, S.A., 2010.
Comparison of the disinfection by-product formation
potential of treated waters exposed to chlorine and
monochloramine. Water Research 44, 729e740.
Centre for Epidemiology and Research, 2002. NSW Health Survey
2002. NSW Department of Health, Sydney.
Charrois, J.W.A., Graham, D., Hrudey, S.E., Froese, K.L., 2004.
Disinfection By-Products in small Alberta community
drinking-water supplies. Journal of Toxicology and
Environmental Health, Part A 67, 1797e1803.
Chen, C., Zhang, X.-J., Zhu, L.-X., Liu, J., He, W.-J., Han, H.-D., 2008.
Disinfection by-products and their precursors in a water
treatment plant in North China: seasonal changes and
fraction analysis. Science of the Total Environment 397 (1e3),
140e147.
Chowdhury, S., Champagne, P., McLellan, P.J., 2009. Models for
predicting disinfection byproduct (DBP) formation in drinking
waters: a chronological review. Science of the Total
Environment 407 (14), 4189e4206.
Cox, P., Fisher, I., Kastl, G., Jegatheesan, V., Warnecke, M.,
Angles, M., Bustamante, H., Chiffings, T., Hawkins, P., 2003.
Sydney 1998: lessons from a drinking water crisis. Journal of
the American Water Works Association 95 (5), 147e161.
Dore, M.H.I., 2005. Climate change and changes in global
precipitation patterns: what do we know? Environment
International 31 (8), 1167e1181.
Eikebrokk, B., Vogt, R.D., Liltved, H., 2004. NOM increase in
Northern European source waters: discussion of possible
causes and impacts on coagulation/contact filtration
processes. Water Science and Water Technology: Water
Supply 44 (4), 47e54.
Galal-Gorchev, H., 1996. Chlorine in water disinfection. Pure and
Applied Chemistry 68 (9), 1731e1736.
Garrido, S.E., Fonseca, M.G., 2010. Speciation and Kinetics of
trihalomethane formation in drinking water in Mexico.
Ground Water Monitoring & Remediation 30 (1), 77e84.
Grellier, J., Bennett, J., Patelarou, E., Smith, R., Toledano, M.,
Rushton, L., Briggs, D., Nieuwenhuijsen, M., 2010. Exposure to
disinfection by-products and Adverse birth outcomes related
5725
to Fetal Growth and Prematurity e A Systematic review and
Meta-analysis. Epidemiology 21 (3), 300e313.
Hamilton, D.P., Schladow, G., Fisher, I.H., 1995. Controlling the
indirect effects of flow diversions on water quality in an
Australian reservoir. Environment International 21 (5), 583e590.
Hoehn, R.C., Randall, C.W., 1979. Viruses, organics, and other
health-related Constituents of the Occoquan reservoir: Part I
trihalomethanes, Pesticides, and Metals. Criteria and
Standards Div., USEPA, Washington D.C.
Hrudey, S.E., 2009. Chlorination disinfection by-products, public
health risk tradeoffs and me. Water Research 43 (8), 2057e2092.
HWC, 2004. Hunter Water Catchment Report 2003/4. Hunter
Water Corporation (HWC), Newcastle Australia.
Hurst, A.M., Edwards, M.J., Chipps, M., Jefferson, B., Parsons, S.A.,
2004. The impact of rainstorm events on coagulation and
clarifier performance in potable water treatment. Science of
the Total Environment 321 (1e3), 219e230.
Kampioti, A.A., Stephanou, E.G., 2002. The impact of bromide on
the formation of neutral and acidic disinfection by-products
(DBPs) in Mediterranean chlorinated drinking water. Water
Research 36 (10), 2596e2606.
Keegan, T., Whitaker, H., Nieuwenhuijsen, M.J., Toledano, M.B.,
Elliott, P., et al., 2001. Use of routinely collected data on
trihalomethane in drinking water for epidemiological purposes.
Occupational & Environmental Medicine 58 (7), 447e452.
Krasner, S.W., McGuire, M.J., Jacangelo, J.G., et al., 1989. The
occurrence of disinfection by-products in US drinking water.
Journal of American Water Works Association 81, 41e53.
Krasner, S.W., Sclimenti, M.J., Means, E.G., 1994. Quality
degradation: implication for DBP formation. Journal of
American Water Works Association 86 (6), 34e47.
McGuire, M.J., Meadow, R.G., 1988. AWWARF trihalomethane
survey. Journal of American Water Works Association 80, 61e68.
McGuire, M.J., McLain, J.L., Obolensky, A., 2002. Information
Collection Rule Data Analysis. AWWA Research Foundation,
USA.
Murphy, B.F., Timbal, B., 2008. A review of recent climate
variability and climate change in southeastern Australia.
International Journal of Climatology 28, 859e879.
National Cancer Institute, 1976. Report on the Carcinogenesis
Bioassay of Chloroform. National Cancer Institute, Bethesda, MD.
NHMRC, 2004. Australian Drinking Water Guidelines 6. National
Health & Medical Research Council. Australian Government,
Canberra.
NHMRC, 2009. Draft Australian Drinking Water Guidelines.
National Health & Medical Research Council (NHMRC),
Canberra Australia.
NSW Department of Primary Industries, 2010. Drought. NSW
Department of Primary Industries. NSW Government, Orange.
URL: <http://www.dpi.nsw.gov.au/agriculture/emergency/
drought/situation/drought-maps> (Last accessed December
2010).
NSW Health, 2000. Drinking Water Monitoring Program. NSW
Health Department, pp. 1e39.
Richardson, S., 2005. New disinfection by-product issues:
emerging DBPs and alternative routes of exposure. Global
NEST Journal 7 (1), 43e60.
Richardson, S.D., Plewa, M.J., Wagner, E.D., Schoeny, R.,
DeMarini, D.M., 2007. Occurrence, genotoxicity, and
carcinogenicity of regulated and emerging disinfection by-
products in drinking water: a review and roadmap for
research. Mutation Research 636, 178e242.
Rizzo, L., Selcuk, H., Nikolaou, A., Belgiorno, V., Bekbolet, M.,
Meric, S., 2005. Formation of chlorinated organics in drinking
water of Istanbul (Turkey) and Salerno (Italy). Global NEST
Journal 7 (1), 95e105.
Rook, J.J., 1974. Formation of haloforms during chlorination of
natural water. Water Treatment Exam 23, 234e243.
5726 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 1 5 e5 7 2 6
Simpson, K.L., Hayes, K.P., 1998. Drinking water disinfection by-
products: an Australian perspective. Water Research 32,
1522e1528.
Stein, P., 2000. The great Sydney water crisis of 1998. Water, Air
and Soil Pollution 123, 419e436.
Soh, Y., Roddick, F., van Leeuwen, J., 2008. The future of water in
Australia: the potential effects of climate change and ozone
depletion on Australian water quality, quantity and
treatability. Environmentalist 28, 158e165.
SWC, 2002. Sydney Water Annual Report. Sydney Water
Corporation (SWC), Sydney, Australia.
USEPA, 1979. National Interim Primary Drinking Water
Regulations; Control of Trihalomethanes in Drinking Water.
Environmental Protection Agency, U.S.
USEPA, 2008. Test Methods for Evaluating Solid Waste, Physical.
Chemical Methods: SW-846. Environmental Protection
Agency, U.S.
Whitaker, H., Nieuwenhuijsen, M.J., Best, N., Fawell, J., Gowers, A.,
Elliot, P., 2003. Description of trihalomethane levels in three UK
water suppliers. Journal of Exposure Analysis and
Environmental Epidemiology 13, 17e23.
Williams, D.T., LeBel, G.L., Benoit, F.M., 1995. A National Survey of
Chlorinated Disinfection By-Products in Canadian Drinking
Water. Health Canada, Ottawa.
Williamson, C.E., Morris, D.P., Pace, M.L., Olson, O.G., 1999.
Dissolved organic Carbon and Nutrients as Regulators of Lake
ecosystems: resurrection of a more Integrated Paradigm.
Limnology and Oceanography 44 (3), 795e803.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

X-ray absorption and photoelectron spectroscopic study

of the association of As(III) with nanoparticulate FeS
and FeS-coated sand

Young-Soo Han a,1, Hoon Y. Jeong b, Avery H. Demond a, Kim F. Hayes a,*
a
Department of Civil and Environmental Engineering, University of Michigan, Ann Arbor, MI 48109, United States
b
Department of Geological Sciences, Pusan National University, Busan 609-735, South Korea

article info abstract

Article history: Iron sulfide (FeS) has been demonstrated to have a high removal capacity for arsenic (As) in
Received 7 February 2011 reducing environments. However, FeS may be present as a coating, rather than in nano-
Received in revised form particulate form, in both natural and engineered systems. Frequently, the removal capacity
12 August 2011 of coatings may be different than that of nanoparticulates in batch systems. To assess the
Accepted 15 August 2011 differences in removal mechanisms between nanoparticulate FeS and FeS present as
Available online 26 August 2011 a coating, the solid phase products from the reaction of As(III) with FeS-coated sand and
with suspensions of nanoparticulate (NP) FeS were determined using x-ray absorption
Keywords: spectroscopy and x-ray photoelectron spectroscopy. In reaction with NP FeS at pH 5, As(III)
Arsenic was reduced to As(II) to form realgar (AsS), while at pH 9, As(III) adsorbed as an As(III)
Sorption thioarsenite species. In contrast, in the FeS-coated sand system, As(III) formed the solid
Mackinawite phase orpiment (As2S3) at pH 5, but adsorbed as an As(III) arsenite species at pH 9. These
Redox different solid reaction products are attributed to differences in FeS concentration and the
XAS resultant redox (pe) differences in the FeS-coated sand system versus suspensions of NP
XPS FeS. These results point to the importance of accounting for differences in concentration
Coatings and redox when making inferences for coatings based on batch suspension studies.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction material, develops an FeS coating under reducing conditions

The high removal capacity of mackinawite (FeS) for arsenic in
reducing environments has been demonstrated in nano-
particulate suspensions (Wolthers et al., 2005; Gallegos et al.,
2007). However, FeS often exists as a coating, both in natural
and engineered environments. For example, mackinawite is
found as a coating on the surfaces of minerals comprising
soils and sediments (Rickard and Morse, 2005). Zero valent
iron (ZVI), emplaced as a permeable reactive barrier (PRB)
in the field (Beak and Wilkin, 2009). Recently, a procedure was
developed to coat nanoparticulate (NP) FeS on a natural sand
to create a PRB material for treating arsenic-contaminated
groundwater (Han et al., 2011). One consequence of having
FeS as a coating is that its effective concentration (g-FeS/L) is
much lower than in batch reactor studies due to the practical
limit of the amount of solid that can be added (e.g., 1e10 g/L
solid suspensions results in 1e10 g/L for NP FeS compared to
4e40 mg/L for FeS-coated sand). Gallegos et al. (2008) recently

* Corresponding author. 1351 Beal Avenue, Department of Civil and Environmental Engineering, MI, United States. Tel.: þ1 734 763 9661;
fax: þ1 734 763 2275.
E-mail addresses: youngsoohan@lbl.gov (Y.-S. Han), hjeong@pusan.ac.kr (H.Y. Jeong), averyd@umich.edu (A.H. Demond), ford@
umich.edu (K.F. Hayes).
1
Present address: Lawrence Berkeley National Laboratory, 1 Cyclotron Rd, Berkeley, CA, United States. Tel.: þ1 510 486 6950.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.026
5728 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5
measured and modeled the redox potential of varying
concentrations of mackinawite suspensions (from 0.01 to 10 g-
FeS/L) and illustrated that higher concentrations of mack-
inawite resulted in a lower redox potential. This decrease in
redox potential with increasing concentrations of FeS was
attributed to a surface-mediated reduction reaction with
mackinawite. Given the importance of redox (and pH) in
controlling the principal arsenic solid phases that may form in
AseFeeS systems (O’Day et al., 2004), and the possible
differences in redox conditions of NP FeS versus FeS coatings,
this study was undertaken to assess whether the mechanisms
of uptake of As(III) established from studies in NP FeS
suspensions are representative of those occurring in FeS-
coated sand systems. In order to successfully predict the
performance of FeS coatings, the difference in uptake mech-
anisms of such coatings relative to NP suspensions needs to be
evaluated.
Using suspensions of synthetic or isolated sulfide minerals,
laboratory studies have previously established mechanisms
of arsenic removal by various iron sulfides such as troilite,
pyrite and mackinawite (Moore et al., 1988; Farquhar et al.,
2002; Bostick and Fendorf, 2003; Wolthers et al., 2005;
Gallegos et al., 2007; Jeong et al., 2010). As(III) sorption on
troilite and pyrite surfaces was characterized as a FeAsS-like
surface precipitation (Bostick and Fendorf, 2003). Farquhar
et al. (2002) demonstrated that mackinawite is more efficient
than other iron-oxide phases or pyrite in removing As(III), by
adsorption of outer-sphere surface complexes and the
formation of poorly crystalline arsenic sulfide precipitates. A
disordered NP mackinawite, reacted within 1 h after forma-
tion, removed 0.012 mol As(III)/mol FeS at neutral pH by the
formation of outer-sphere surface complexes on the mack-
inawite surface (Wolthers et al., 2005). In another study,
nanoparticulate mackinawite that had been aged three days
showed an As(III) removal capacity of 0.16, 0.018 and 0.004 mol
As(III)/mol FeS at pH 5, 7, and 9, respectively (Han et al., 2011).
The precipitation of realgar (AsS) was found to be responsible
for the high As(III) removal observed at pH 5, while the
formation of a thioarsenite surface species were identified at
pH 9, based on x-ray absorption spectroscopy and thermo-
dynamic calculations (Gallegos et al., 2007, 2008), and later
supported by x-ray photoelectron spectroscopy and high
resolution electron transmission microscopy (Renock et al.,
2009). Yet, with the exception of the work reported in
Gallegos et al. (2008), the redox state in those laboratory
studies was not specifically controlled or reported.
Even though the association of As(III) with NP mackinawite
has been intensively studied by several research groups
(Farquhar et al., 2002; Wolthers et al., 2005; Gallegos et al.,
2007; Jeong et al., 2010), the association of As(III) with FeS
coatings has not been studied. Yet, such coatings result in
different As(III) uptake (Han et al., 2011). Therefore, to assess
the potential impact caused by FeS attachment as a coating,
this study characterized the solid phase reaction products of
As(III) reacted with NP FeS and FeS-coated sand using x-ray
absorption spectroscopy (XAS) and x-ray photoelectron spec-
troscopy (XPS). This comparison provides an opportunity to
evaluate if laboratory-generated results with suspensions of
NP FeS reasonably represent the As(III) uptake reactions of
FeS-coated sand, and by implication, FeS coatings on other
minerals found in natural or engineered anoxic
environments.
2. Materials and methods
2.1. Preparing FeS-coated sand
Wedron 510 silica sand (Wedron Silica Co., Wedron, IL) with
a geometric mean size of 0.15e0.22 mm was used as the
substrate for the FeS coating. A batch of sand was washed
several times with Milli-Q water, soaked in Milli-Q water
overnight with mild shaking, and then dried at ambient
temperature (around 25  C). FeS was synthesized inside an
anaerobic chamber maintained at a 5% H2/95% N2 atmosphere
by mixing 2.0 L of a 0.57 M FeCl2 (Fisher Chemical) with 1.2 L of
1.1 M Na2S solution (Butler and Hayes, 1998). As soon as the
two chemicals were mixed, a black nanoparticulate suspen-
sion of mackinawite particles was obtained. The precipitated
particles were stirred for three days, rinsed with distilled
water to remove excessive salt and sulfide ions, and then
separated from the supernatant using centrifugation.
To coat the sand with FeS, freeze-dried mackinawite
particles were resuspended to form a 2 g/L FeS suspension.
The suspension was adjusted to a pH value of 5.5 and mixed
with the sand for three days using an end-over-end rotator, at
which point a clear supernatant was obtained. The FeS-coated
sand was separated from the solution, dried in an anaerobic
glovebox and stored in an air-tight container in the glovebox.
The amount of FeS coating was determined to be 1.42  105 M
FeS/g sand (1.24  103 g FeS/g sand) using an acid-extraction
method. More details of the FeS coating methodology, and
physical and chemical characterization can be found in Han
et al. (2011).
2.2. Spectroscopy sample preparation
X-ray absorption spectroscopy (XAS) and x-ray photoelectron
spectroscopy (XPS) were used to determine the solid phase
arsenic oxidation state and the relative proportions of
different As species in As(III)-reacted NP FeS and FeS-coated
sand samples. For the reaction of As(III) with FeS, 300 mL of
a 5 g/L NP FeS suspension was placed in 400 mL glass reactors
that allowed for the continuous measurement of pH in
a closed system. The pH of a 5 g/L suspension was adjusted to
pH 5 or 9 using HCl, and an aliquot of a 1.33 M NaAsO2 stock
solution was added to achieve a concentration of 1.33  102 M
As(III). The pH was monitored over a two-day equilibrium
period and adjusted as necessary with acid to maintain the pH
at 5 or 9. For the reaction of FeS-coated sand with As(III), 416 g/
L of FeS-coated sand (equivalent to 0.5 g/L FeS based on
1.24  103 g FeS/g sand) were reacted with a 1.33  103 M
As(III) solution in 50 mL polypropylene tubes and mixed by an
end-over-end rotating mixer for two days. Since the pH of this
system could not be easily monitored continuously given the
smaller volume reactor, multiple samples were prepared and
a sample that gave a pH of 5 or 9 after two days of equilibration
was selected for the spectroscopic analysis. The redox
potential for the pH 5 samples was measured using an
oxidation-reduction potential (ORP) combination platinum
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5
electrode with a Ag/AgCl reference (ColeeParmer). The
measured potentials (in mV) were corrected for the standard
hydrogen electrode (SHE) and temperature and converted to
pe by multiplying by 0.0169.
The samples for XAS analyses were filtered using 0.22 mm
nylon filters and the wet filtered paste of particles was
transferred into air-tight, crimp-sealed serum vials and then
shipped to Stanford Synchrotron Radiation Lightsource (SSRL)
(Stanford, CA) for analysis. For the XPS analysis, the samples
were filtered, freeze-dried, crimp-sealed and stored in an
anaerobic chamber until the analysis.
2.3. X-ray absorption spectroscopy (XAS)
At SSRL, right before the XAS analysis, the wet paste samples
were loaded, inside an anaerobic glove box, onto sample
holders wrapped with a double layer of Kapton tape. XAS
spectra were collected at SSRL on beamline 10-2 (3 GeV,
w100 mA of maximum current) using an unfocused beam
with a beam size of 2.0  20.0 mm at the sample holder.
Arsenic K-edge XAS spectra were obtained using a Si(220)
double-crystal monochromator with a 13-element solid-state
Ge-array fluorescence detector or Lytle detector. The sample
chamber was continuously purged with He gas to avoid
potential oxidation. Based on a comparison of spectra, no
oxidation occurred during the data collection. XAS spectra
were also collected for reference compounds such as metallic
arsenic (As(0)), amorphous AsS, amorphous As2S3 dissolved
As(III), and dissolved As(V). All arsenic reference compounds
were purchased from Alfa Aesar (Lancaster, UK).
The XAS spectra were analyzed using the program package
SixPACK (Webb, 2002). Individual spectra were first averaged,
and the background absorbance was subsequently removed
by a linear fit through the pre-edge region. X-ray absorption
near-edge structure (XANES) spectra (e.g., 11,86011,890 eV)
were obtained by normalizing the fluorescence signal to the
edge jump height. The absorption edges (i.e., inflection ener-
gies) of XANES spectra were determined to compare the
oxidation state of arsenic between the samples and reference
compounds. Extended x-ray absorption fine structure (EXAFS)
spectra were also obtained by fitting a quadratic spline func-
tion above the edge. EXAFS spectra were normalized using
a Victoreen polynomial function and then transformed from
energy (eV) to k space (Å1) using E0 ¼ 11,885 eV. The resultant
EXAFS functions (c(k)) were weighted by k3 to amplify the
higher k region, and Fourier-transformed to produce radial
structural functions (RSF) in R space over k ¼ 3.5e11.5 Å1.
Structural parameters were obtained by fitting k3-weighted
EXAFS functions with the phase and amplitude functions
derived from FEFF 8 (Ankudinov et al., 1998). The amplitude-
reduction factor (S2o ¼ 0.92) was optimized from the fitting of
the reference compound spectra and kept constant for all
EXAFS analysis. The Debye-Waller factors (s2) were also fixed
based on the similarity between the sample spectra and the
reference compound spectra or the optimization among the
sample spectra to reduce the degrees of freedom during
the fitting. Coordination number (N ), interatomic distance (R),
and energy shift (DE0) were allowed to vary. The optimal fitting
was obtained by minimizing the goodness of fit parameter (Rf).
5729
2.4. X-ray photoelectron spectroscopy (XPS)
Kratos Axis Ultra X-ray photoelectron spectrometer was used
to examine the chemical composition and oxidation state of
As species sorbed on NP FeS and FeS-coated sand. The refer-
ence compounds used for the XPS analyses were As(0),
arsenic(II) sulfide, arsenic(III) sulfide, NaAsO2 and Na2HA-
sO4$7H2O, all purchased from Alfa Aesar (Lancaster, UK). The
reference compounds and As(III) reacted samples were
mounted on a sample bar in an anaerobic glove box and
transferred using an air-tight container filled with 95% N2/5%
H2 gas mixture to minimize the oxygen contact with the
sample surface. The Al-Ka line (1486.6 eV) was used as the
radiation source. Survey spectra were obtained using analyzer
pass energies of 160 eV. Narrow XPS peaks were obtained
primarily with a pass energy of 20 eV, but for the As(III)-
reacted FeS-coated sand samples, the higher pass energy of
160 eV was needed due to a low As loading. Energies were
corrected for charging effects using the reference peak of
adventitious carbon C ls with a binding energy of 284.6 eV.
Raw spectra were smoothed using a linear base line for 1s
peaks and a Shirley base line for 2p peaks, respectively, and
then fitted using a GaussianeLorentzian function. To estimate
the standard deviation of each of the component’s contribu-
tion to the overall XPS spectrum in the fitting procedure,
Monte-Carlo analysis (CasaXPS, Casa Software Ltd., UK) was
used. This program applies artificial noise to a spectrum and
8.8
(j)
Normalized absorbance
(i)
6.8
(h)
(g)
4.8 (f)
(e)
2.8 (d)
(c)
(b)
0.8 (a)
-1.2
11860 11870 11880 11890
E (eV)
Fig. 1 e Arsenic K-edge XANES spectra of (a) As(0) (grey),
(b) FeAsS (green), (c) AsS (blue), (d) As2S3 (pink), (e) 5 g/L NP
FeS reacted with 1.33 3 10L2 M As(III) for 2 days at pH 5,
(f) 416 g/L FeS-coated sand reacted with 1.33 3 10L3 M
As(III) for 2 days at pH 5, (g) 5 g/L NP FeS reacted with
1.33 3 10L2 M As(III) for 2 days at pH 9, (h) 416 g/L FeS-
coated sand reacted with 1.33 3 10L3 M As(III) for 2 days at
pH 9, (i) dissolved NaAsO2 (yellow) and (j) dissolved
Na2HAsO4$7H2O (red). The absorption edges correspond to
the first derivative maxima of XANES spectra. (For
interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this
article.)
5730 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5

calculates an error matrix to give the variance of each fit based
on the fitting constraints used.
3. Results and discussion
3.1. Arsenic loading
The supernatants of the samples used for the spectroscopic
analyses were analyzed using ICP-MS to calculate the arsenic
loading on the solid phase. NP FeS removed 99% (at pH 5) and
20% (at pH 9) of the initial 1.33  102 M As(III) present in solu-
tion, and FeS-coated sand removed 63% (at pH 5) and 23% (at pH
9) of the initial 1.33  103 M As(III). These data demonstrate
that substantially higher As(III) sequestration occurs at pH 5
where the solubility of arsenic sulfides is lower than that of
mackinawite. In terms of removal capacity (mg-As(III)/g-FeS),
the FeS-coated sand had approximately 70% and 400% of the
capacity of the NP FeS at pH 5 and pH 9, respectively, deter-
mined based on the results reported in Han et al. (2011).
3.2. XAS analysis
The XAS spectra were subjected to both XANES and EXAFS
analyses. While the oxidation state of arsenic can be obtained
from XANES analysis, structural parameters such as inter-
atomic distance (R) and coordination number (N ) on the near
coordination environment around arsenic are gleaned from
EXAFS analysis. In XANES spectra (Fig. 1), the absorption edges
(i.e., inflection energies) of the samples are compared with
those of reference compounds. While higher absorption edge
energy is indicative of a higher oxidation state of arsenic, lower
absorption edge energy corresponds to a lower oxidation state.
The absorption edge energies of As(0), arsenopyrite, AsS, As2S3,
dissolved As(III), and dissolved As(V) were 11866.7, 11867.0,
11868.1, 11869.0, 11870.9, and 11874.4 eV, respectively. At pH 5,
the NP FeS system reacted with As(III) had an absorption edge
energy of 11868.4 eV, only slightly higher than that of AsS,
indicating that the dominant oxidation state of As in the NP FeS
sample was þII. The As(III)-reacted FeS-coated sand had an
absorption energy of 11869.1 eV, close to that of As2S3, sug-
gesting the formation of As2S3. At pH 9, the NP FeS reacted with
As(III) had an absorption energy of 11868.06 eV, slightly lower
than that of As2S3 but much lower than that of dissolved As(III),
indicating the possible formation of thioarsenite species. The
As(III)-reacted FeS-coated sand at pH 9 had an absorption
energy of 11870.87 eV, close to that of dissolved As(III), sug-
gesting the surface complexation of arsenite species.
The EXAFS spectra and corresponding Fourier transforms
of the experimental samples and reference compounds are
compared in Fig. 2. The fitting results of the EXAFS analysis are
shown in Table 1. For the NP FeS sample at pH 5, the first
coordination shell around As is characterized by the AseS
interaction, with the coordination number (NAseS) of 2.1 at
a distance of 2.27 Å, in good agreement with that of the AsS
referemce compound (NAseS ¼ 2.0 at 2.26 Å). Also, the second
coordination shell for the NP FeS sample is characterized by
the AseAs interaction at 3.49 Å, with a coordination number
(NAseAs ¼ 0.95) twice that of the AsS reference compound
(NAseAs ¼ 0.41 at 3.50 Å). Compared with the AsS reference
compound, the NP FeS sample had a stronger second shell
feature and a slightly higher absorption energy, indicating that
another As phase, in addition to AsS, may form in the NP FeS
system. Previously, Bostick et al. (2003) proposed a surface

Fig. 2 e k3-weighted arsenic K-edge EXAFS spectra (k3c(k)) and their FT transforms for (a) As(0) (grey), (b) FeAsS (green),
(c) AsS (blue), (d) As2S3 (pink), (e) 5 g/L NP FeS reacted with 1.33 3 10L2 M As(III) for 2 days at pH 5, (f) 416 g/L FeS-coated sand
reacted with 1.33 3 10L3 M As(III) for 2 days at pH 5, (g) 5 g/L NP FeS reacted with 1.33 3 10L2 M As(III) for 2 days at pH 9, (h)
416 g/L FeS-coated sand reacted with 1.33 3 10L3 M As(III) for 2 days at pH 9, (i) dissolved NaAsO2 (yellow) and (j) dissolved
Na2HAsO4$7H2O (red). Solid lines are the experimental data; dashed lines are the numerical fits. The peak positions in the
Fourier transform (FT) are uncorrected for phase shift as indicated by the x-axis notation of R D D(Å). (For interpretation of
the references to colour in this figure legend, the reader is referred to the web version of this article.)
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5 5731

Table 1 e EXAFS fit results for nanoparticulate FeS and FeS-coated sand reacted with As(III) at pH 5 and 9, and As model
compounds.
Pair EXAFS fita Crystallographic data Reference
2 2 b
N R(Å) s (Å ) N R(Å)

As(0) AseAs 1.1 2.50 0.0058 3 2.5 O‘Day et al., 2004

AseAs 3 3.13
DE0 ¼ 6.95 eV, Rf ¼ 0.065

AsS AseS 2.0 2.26 0.003 2 2.24 Farquhar et al., 2002

AseAs 1 2.57
AseAs 0.41 3.50 0.006 2.5 3.44e3.51
AseS 1 3.41e3.52
DE0 ¼ 9.80 eV, Rf ¼ 0.061

As2S3 AseS 3.0 2.28 0.0045 3 2.24e2.31 Farquhar et al., 2002

AseAs 0.37 3.54 0.006 1 3.19
AseS 3 3.22e3.57
AseAs 2.5 3.52e3.64
DE0 ¼ 7.75 eV, Rf ¼ 0.047

pH 5 NP-FeS AseS 2.1 2.27 0.003 This study

AseAs 0.95 3.49 0.006
DE0 ¼ 8.24 eV, Rf ¼ 0.0412

pH 9 NP-FeS As(III)-O 0.69 1.77 0.0045 This study

AseS 1.96 2.25 0.0045
AseAs (As(0)) 1.10 2.55 0.008
AseAs (adsorbed) 1.73 3.50 0.006
DE0 ¼ -8.84 eV, Rf ¼ 0.0205

pH 5 FeS coated-sand AseS 2.9 2.26 0.0045 This study

AseAs 0.31 3.66 0.006
DE0 ¼ 10.09 eV, Rf ¼ 0.0974

pH 9 FeS coated-sand As(III)eO 2.88 1.78 0.0045 This study

DE0 ¼ 2.78 eV, Rf ¼ 0.2119

As(III)aq As(III)eO 3.0 1.76 0.0045 3c 1.78c Wolthers et al., 2005

DE0 ¼ 7.90 eV, Rf ¼ 0.069

As(V)aq As(V)eO 4.0 1.69 0.0025 4c 1.69c Yamauchi and Fowler, 1994
DE0 ¼ 5.01 eV, Rf ¼ 0.024

N: coordination number; R: interatomic distance; DE0: energy shift; Rf: goodness of fit parameter.
a The amplitude-reduction factor (S2o) was set at 0.92.
b The Debye-Waller factors (s2) were fixed during the numerical fit.
c Structural data were obtained from EXAFS analysis.

precipitate in the form of trimeric arsenic sulfide for As(III)
sorption by PbS and ZnS. The AseAs bonding distance (w3.6 Å)
in their study is close to the value of 3.50 Å observed here. Thus,
the formation of surface precipitates as thioarsenites may
explain the observed differences of the As coordination in the
AsS reference compound and in the NP FeS system.
The first coordination shell of the FeS-coated sand system
at pH 5 is characterized by the AseS interaction with the
coordination number (NAseS) of 2.9 at a distance of 2.26 Å,
consistent with the coordination chemistry of As2S3
(NAseS ¼ 3.0 at 2.28 Å). Unlike the NP FeS system, both the FeS-
coated sand system and the As2S3 reference compound show
very weak second coordination shells (NAseAs ¼ 0.31 at 3.66 Å
for the FeS-coated sand system and NAseAs ¼ 0.37 at 3.54 Å for
the As2S3 reference compound). Taken together, the forma-
tion of As2S3 is mainly responsible for As(III) uptake in the FeS-
coated sand system.
At pH 9, the NP FeS system reacted with 1.33  102 M
As(III) exhibited different characteristics in its EXAFS
spectrum compared to the pH 5 system. The best fit resulted
from the inclusion of an AseS coordination shell (NAseS of 1.96
at a distance of 2.25 Å) and an AseAs coordination shell
(NAseAs of 1.73 at a distance of 3.50 Å), consistent with the
presence of surface precipitates or thioarsenites surface
clusters (Jeong et al., 2010). A minor contribution of AseAs
bonding with NAseAs of 1.10 at a distance of 2.55 Å also
improved the fit, indicating the possible formation of As(0)
(Jeong et al., 2010). In contrast, the FeS-coated sand sample at
pH 9 reacted with 1.33  102 M As(III) showed a first coordi-
nation shell comprised of As(III)-O with NAseO of 2.88 at
a distance of 1.78 Å, suggesting the surface complexation of
arsenite species in the FeS-coated sand system at pH 9.
3.3. XPS analysis
Fig. 3 shows the As 3d spectra of arsenic reference compounds
and NP FeS and FeS-coated sand samples reacted with As(III)
(See Table 2 for fitting results). Each surface species in the
5732 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5

Table 2 e XPS fit results nanoparticulate FeS reacted with

As(III) at pH 5 and 9, and As model compounds using
a low pass energy of 20 eV illustrated in Fig. 3.
Percent peak area of each component

AseAs As(II)eS As(III)eS As(III)eO As(V)eO

BE 41.8 42.8 43.2 43.0 44.5

(1) (1) (1) (1) (1)
(g) e e e 15.76 84.24
(f) e e e 89.52 10.48
(e) 19.78 e e 56.62 29.22
(d) e 100.0 e e e
(c) e e 87.28 e 12.72
(b) e 100.0 e e e
(a) 100.0 e e e e

BE: Binding energy (FWHM).

compounds reveal that each compound consisted of at least

Fig. 3 e XPS spectra of As 3d peaks for (a) As(0), (b) AsS,
(c) As2S3, (d) 5 g/L NP FeS reacted with 1.33 3 10L2 M As(III)
for 2 days at pH 5, (e) 5 g/L NP FeS reacted with
1.33 3 10L2 M As(III) for 2 days at pH 9, (f) NaAsO2 salt and
(g) Na2HAsO4$7H2O salt using a low pass energy of 20 eV
(a, b, c, f and g are model compounds, and d and e are
samples. The spectrum of sample (e) is enlarged for better
viewing.).
As(3d ) spectrum is fitted with a doublet representing the spin-
orbit splitting of the As 3d5/2 and As 3d3/2 peaks. A higher
binding energy is indicative of a higher oxidation state of
arsenic and a lower binding energy corresponds to a lower
oxidation state. The positions of the As 3d5/2 peaks for (a) As(0),
(b) arsenic(II) sulfide, (c) arsenic(III) sulfide, (f) NaAsO2 and (g)
Na2HAsO4$7H2O were determined to be 41.8, 42.8, 43.1, 43.5,
and 44.5 eV, respectively. These peaks for the model
83% of the expected dominant oxidation state, with minor
contributions of other oxidation states of arsenic. The full-
width-at-half-maximum (FWHM) values for all the model
compounds were constrained to be 1.0 eV for low pass energy
scans and 2.2 eV for the high pass energy scans. The binding
energies determined by the As 3d5/2 peak positions and FWHM
values were used to determine the predominant As oxidation
states of the NP FeS samples reacted with 1.33  102 M As(III)
and FeS-coated sand reacted with 1.33  103 M As(III) at pH 5
and 9.
At pH 5, the As 3d5/2 peaks of NP FeS with 1.33  102 M
As(III) were sharp, indicating that one prominent arsenic solid
species precipitated as a result of the reaction between NP FeS
and As(III) at pH 5 (Fig. 3 (d)). The peaks were well fitted with
a doublet of As 3d5/2 and As 3d3/2 at binding energies of 42.8
and 43.5 eV, respectively. These peak positions indicate that
the As oxidation state is primarily As(II) and that realgar is
formed. This is consistent with the XAS analysis above (Table
1) and previous thermodynamic modeling of a similar system
(Gallegos, 2007; Gallegos et al., 2008). As no AsS diffraction
patterns were observed using XRD (Figure S-1 in
Supplementary Material), the precipitated AsS is likely in an
amorphous form. Renock et al. (2009) also reported a realgar-
like precipitate as the primary product formed by As reac-
tion with mackinawite under similar experimental condi-
tions. However, their arsenic XPS peaks showed much broader
features, representing a mixture of different species of various
arsenic oxidation states, whereas in this study, the peak
represents primarily a single contribution of As(II)-S.
At pH 9, the intensity of the As 3d5/2 peaks is much smaller
than that at pH 5 because only 20% out of the 1.33  102 M As(III)
was solid-phase associated, compared with almost 100% at pH
5. The low solid phase arsenic resulted in a weak, broad peak,
but with the center of the As 3d5/2 peak shifted to a higher
binding energy and extending throughout the binding energy
ranges of As(II)-S, As(III)-S and As(III)-O, possibly indicating the
presence of a mixture of various thioarsenite species. Inter-
preted in this manner, the result is consistent with the EXAFS
analysis.
The XPS spectra for As(III) reacted with FeS-coated sand at
pH 5 and 9 are shown in Fig. 4, with the peak fitting parameters
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5 5733

results indicate the presence of primarily As(III)-O on the FeS-

coated sand surface, suggesting arsenite adsorption. These
results support those obtained with XAS. The broadening of
the peak in the low and high binding region of the spectrum is
(b)
likely caused by the use of higher pass energy of 160 eV. This is
Relative Intensity

confirmed by the measurements presented in Figure S-2

(Supplementary Material), which show that the FWHM
values increase with increasing pass energy. However, the
highest peak positions still were located at the same binding
energy. Consequently, the identification of the peak position
using the higher pass energy for the samples with weak
element intensity appears to be valid when used in this
(a) qualitative manner to verify the peak maximum.

3.4. Comparison of NP FeS and FeS-coated sand systems

reacted with As(III)
50 49 48 47 46 45 44 43 42 41 40 39 38 37
Binding Energy (eV) NP FeS and FeS-coated sand reacted with As(III) at pH 5
resulted primarily in the precipitation of arsenic sulfides but
As(5+): As(V)-O As(3+): As(III)-O
different oxidation states were detected in each of the arsenic
As(3+): As(III)-S As(0): As-As
Base line Fitted envelope solid phases. The XANES and EXAFS analyses indicated the
Measured point formation of AsS and a thioarsenite surface precipitate for
As(III) uptake in the FeS system at pH 5, while the formation of
Fig. 4 e XPS spectra of As 3d peaks for FeS-coated sand As2S3 occurred in the FeS-coated sand system at pH 5. The XPS
reacted with As(III) reacted with 1.33 3 10L3 M As(III) for results also support the formation of these different arsenic
2 days at (a) pH 5 and (b) pH 9 using a high pass energy sulfide solids in each system.
of 160 eV. In NP FeS systems, more reduced conditions favor realgar
precipitation, while more oxidizing conditions favor orpiment
precipitation. To assess the redox conditions in the NP FeS and
Fe-coated sand, the redox potential was measured. In the
systems consisting of 5 g FeS/L reacted with 1.33  102 M
given in Table 3. Since the As(III) loading on the FeS-coated
As(III) and 416 g FeS-coated sand/L (0.5 g-FeS/L) reacted with
sand after reaction with As(III) was too low for sufficient
1.33  103 M As(III), both at pH 5, the redox potential was
spectral peak quality for quantitative analysis, a pass energy
measured to be 326 mV and 246 mV, respectively. The pe
of 160 eV was used instead of the more widely used pass
values calculated from these measured redox potentials are
energy of 20 eV. A pass energy of 160 eV is sometimes used to
2.15 for the NP FeS system and 0.91 for the FeS-coated sand
identify an oxidation state of an element of interest (Su and
system. A thermodynamic simulation of the FeeAseSeH2O
Puls, 2008). In this study, the XPS As 3d spectrum obtained
batch system predicted amorphous As2S3 precipitation at pe
using a pass energy of 160 eV was used to qualitatively
values ranging from 1.0 to around 2.0, and AsS precipitation
compare the peak position of the As(III) reacted FeS-coated
at pe values ranging from 2.0 to 5.0 (Gallegos et al., 2008).
sand samples. Fig. 4 shows the XPS analysis of the FeS-
The different redox conditions were postulated to result from
coated sand sample reacted with As(III) at pH 5 and indi-
FeS surface redox reactions and the different total amounts of
cates that the solid phase reaction product is primarily As(III)-
FeS in each system. The measured pe values in the present
S, consistent with the formation of orpiment. At pH 9, the
study match well with the thermodynamically-predicted
arsenic sulfide species and those identified by XPS and XAS.
In both the NP FeS and FeS-coated sand systems, the
primary As(III) uptake process is precipitation at pH 5, while at
Table 3 e EXAFS fit results for nanoparticulate FeS and pH 9, uptake is controlled by adsorption and/or surface
FeS-coated sand reacted with As(III) at pH 5 and 9, and As precipitation reactions. However, the type of surface species
model compounds using a high pass energy of 160 eV
formed is different. In the NP FeS system, thioarsenite surface
illustrated in Fig. 4.
species formed, but in the FeS-coated sand system, arsenite
Percent peak area of each component
species were detected. Our previous study of FeS-coated sand
AseAs As(III)eS As(III)eO As(V)eO showed that a small amount of an iron oxyhydroxide phase
formed in the FeS-coated sand system due to the more
BE 41.8 43.2 43.5 44.5
(2.2) (2.2) (2.2) (2.2)
oxidizing conditions that prevailed in this system (Han et al.,
(b) 11.37 e 82.63 e 2011). The presence of this phase may also explain why arse-
(a) 31.05 48.63 e 20.31 nite surface complexes form on FeS-coated sand whereas thi-
oarsenite species form on NP FeS and why there is greater
BE: Binding energy (FWHM).
adsorption of As(III) by FeS-coated sand at pH 9 (Han et al., 2011).
5734 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5
3.5. Implications of As speciation differences in NP FeS
and FeS-coated sand
The different species that form when As(III) is reacted with
FeS-coated sand compared to with NP FeS are thought to
result from the different redox conditions: (1) the higher
redox potential of the FeS-coated sand system at pH 5 results
in the precipitation of orpiment rather than realgar as in the
NP FeS system, and (2) the higher redox potential of the FeS-
coated sand system at pH 9 leads to the formation of a small
amount of iron oxyhydroxide solid and the adsorption of
arsenite rather than the adsorption of thiosarsenite as in the
NP FeS system. The overall implication of these results is
that the reaction products of arsenic with FeS in natural
systems may be different from those determined using
concentrated laboratory batch systems of NP FeS, in which
artificially enhanced reducing conditions may result from
concentrated NP FeS suspensions. This is a reminder that
both pH and pe conditions need to be assessed in AseFeeS
systems if results are to be applied more generally. Given
that the extent of As removal by FeS-coated sand is similar to
that by NP FeS, despite the differences in uptake mecha-
nisms, these results support that FeS-coated sand may be
a suitable sorbent material for in situ removal of arsenic in
PRB applications where the emplacement of NP FeS is
impractical.
4. Conclusions
 XAS and XPS analyses showed differences in the solid phase
species that form during the reaction of As(III) with NP FeS
and FeS-coated sand.
 At pH 5, the reaction of As(III) with NP FeS results primarily
in the precipitation of realgar but the reaction with FeS-
coated sand results in orpiment precipitation.
 At pH 9, As(III) is removed through adsorption or surface
precipitation of thioarsenite species by NP FeS but through
the adsorption of arsenite surface species by FeS-coated
sand.
 The differences in As speciation in the NP FeS system
compared to the FeS-coated sand system are attributed to
the lower redox potentials in the NP FeS suspensions.
 The mechanisms of As uptake determined in batch reactor
systems with concentrated NP FeS may not be indicative of
the uptake mechanisms that occur in systems in which FeS
is present as a coating.
Acknowledgements
The authors gratefully acknowledge the assistance of Tom
Yavaraski (Department of Civil and Environmental Engi-
neering, University of Michigan) for his help in developing
the analytical methods for As and Fe analyses, and Udo
Becker and Devon Renock (Department of Geological
Sciences, University of Michigan) for their guidance on the
XPS data collection and analysis. Portions of this research
were carried out at the Stanford Synchrotron Radiation
Lightsource, a Directorate of SLAC National Accelerator
Laboratory and an Office of Science User Facility operated for
the U.S. Department of Energy Office of Science by Stanford
University. The SSRL Structural Molecular Biology Program is
supported by the DOE Office of Biological and Environmental
Research, and by the National Institutes of Health, National
Center for Research Resources, Biomedical Technology
Program (P41RR001209). This research was supported by the
Strategic Environmental Research and Development Program
(SERDP) under Department of Defense, Department of Army,
Contract Number W912HQ-04-C-0035. This paper has not
been subject to agency review; it, therefore, does not neces-
sarily reflect the sponsor’s view, and no official endorsement
should be inferred.
Appendix. Supplementary Material
Supplementary material associated with this article can be
found, in the online version, at doi:10.1016/j.watres.2011.08.026.
references
Ankudinov, A.L., Ravel, B., Rehr, J.J., Conradson, S.D., 1998. Real-
space multiple-scattering calculation and interpretation of x-
ray-absorption near-edge structure. Physical Review B-
Condensed Matter 58 (12), 7565e7576.
Beak, D.G., Wilkin, R.T., 2009. Performance of a zerovalent iron
reactive barrier for the treatment of arsenic in groundwater:
Part 2. Geochemical modeling and solid phase studies. Journal
of Contaminant Hydrology 106 (1e2), 15e28.
Bostick, B.C., Fendorf, S., 2003. Arsenite sorption on troilite (FeS)
and pyrite (FeS2). Geochimica et Cosmochimica Acta 67 (5),
909e921.
Bostick, B.C., Fendorf, S., Manning, B.A., 2003. Arsenite adsorption
on galena (PbS) and sphalerite (ZnS). Geochimica et
Cosmochimica Acta 67 (5), 895e907.
Butler, E.C., Hayes, K.F., 1998. Effects of solution composition and
pH on the reductive dechlorination of hexachloroethane by
iron sulfide. Environmental Science and Technology 32 (9),
1276e1284.
Farquhar, M.L., Charnock, J.M., Livens, F.R., Vaughan, D.J., 2002.
Mechanisms of arsenic uptake from aqueous solution by
interaction with goethite, lepidocrocite, mackinawite, and
pyrite: an X-ray absorption spectroscopy study.
Environmental Science and Technology 36 (8), 1757e1762.
Gallegos, T.J., 2007. Sequestration of As(III) by synthetic
mackinawite under anoxic conditions. Ph.D. Dissertation, The
University of Michigan, Ann Arbor, MI.
Gallegos, T.J., Hyun, S.P., Hayes, K.F., 2007. Spectroscopic
investigation of the uptake of arsenite from solution by
synthetic mackinawite. Environmental Science and
Technology 41, 7781e7786.
Gallegos, T.J., Han, Y.S., Hayes, K.F., 2008. Model predictions of
realgar precipitation by reaction of As(III) with synthetic
mackinawite under anoxic conditions. Environmental Science
and Technology 42, 9338e9343.
Han, Y.-S., Gallegos, T.J., Demond, A.H., Hayes, K.F., 2011. FeS-
coated sand for removal of arsenic(III) under anaerobic
conditions in permeable reactive barriers. Water Research 45,
593e604.
Jeong, H.Y., Han, Y.S., Hayes, K.F., 2010. X-ray absorption and x-
ray photoelectron spectroscopic study of arsenic mobilization
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 2 7 e5 7 3 5
during mackinawite (FeS) oxidation. Environmental Science
and Technology 44 (3), 955e961.
Moore, J.N., Ficklin, W.H., Johns, C., 1988. Partitioning of arsenic
and metals in reducing sulfidic sediments. Environmental
Science and Technology 22 (4), 432e437.
O’Day, P.A., Vlassopoulos, D., Root, R., Rivera, N., 2004. The influence
of sulfur and iron on dissolved arsenic concentrations in the
shallow subsurface under changing redox conditions.
Proceedings of the National Academy of Sciences of the United
States of America 101 (38), 13703e13708.
Renock, D., Gallegos, T., Utsunomiya, S., Hayes, K., Ewing, R.C.,
Becker, U., 2009. Chemical and structural characterization of
As immobilization by nanoparticles of mackinawite (FeSm).
Chemical Geology 268 (1e2), 116e125.
Rickard, D., Morse, J.W., 2005. Acid volatile sulfide (AVS). Marine
Chemistry 97 (3e4), 141e197.
5735
Su, C.M., Puls, R.W., 2008. Arsenate and arsenite sorption on
magnetite: relations to groundwater arsenic treatment using
zerovalent iron and natural attenuation. Water Air and Soil
Pollution 193 (1e4), 65e78.
Webb, S.M., 2002. Sam’s Interface for XAS Package (SixPACK).
Stanford Synchrotron Radiation Laboratory, Menlo Park, CA.
Wolthers, M., Charlet, L., Van der Weijden, C.H., Van der Linde, P.
R., Rickard, D., 2005. Arsenic mobility in the ambient sulfidic
environment: sorption of arsenic(V) and arsenic(III) onto
disordered mackinawite. Geochimica et Cosmochimica Acta
69, 3483e3492.
Yamauchi, H., Fowler, B.A., 1994. Toxicity and metabolism of
inorganic and methylated arsenicals. In: Advances in
Environmental Science and Technology, Arsenic in the
Environment, Part II. Human Health and Ecosystem Effects.
Wiley, New York, pp. 35e53.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Defluoridation of water via electrically controlled anion

exchange by polyaniline modified electrode reactor

Hao Cui, Qin Li, Yan Qian, Rong Tang, Hao An, Jianping Zhai*
State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing 210093, PR China

article info abstract

Article history: A polyaniline (PANI) modified electrode reactor was designed for fluoride removal from
Received 13 June 2011 aqueous solutions. The innovative concept behind the reactor design is that the uptake and
Received in revised form elute of fluoride could be well controlled by modulating the potential of the PANI film. The
24 August 2011 maximum fluoride removal capacity of PANI is more than 20 mg/g at a positive voltage
Accepted 25 August 2011 based on the electrically controlled anion-exchange mechanism. The results of batch tests
Available online 1 September 2011 showed that terminal potential values had a major impact on fluoride removal by this
PANI, with optimal removal occurring at 1.5 V. The fluoride removal capacity (qe) increased
Keywords: rapidly within 5 min and reached equilibrium within 10 min, which indicated a rapid
Fluoride removal velocity of fluoride by PANI under this condition. The applicability of defluor-
Electrochemical idation using the PANI reactor to treat fluoride-contaminated tap water was also tested
Anion exchange through flow cell breakthrough studies. At initial fluoride concentrations of 5 mg/L and
Polyaniline 10 mg/L, the breakthrough capacities were 20.08 mg/g and 19.24 mg/g, respectively.
Water treatment Moreover, during the first half of the period before the breakthrough point, the fluoride
concentration of the treated solution was below the WHO’s recommended levels (1.5 mg/L).
The results of the five consecutive treatment-regeneration studies also showed that the
PANI films could be reused. Taken together, these results implied that the electrically
controlled anion exchange by the PANI-modified electrode reactor may be an effective
technique for the removal of fluoride from water.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction industrial wastewater. Precipitation treatment usually reduces

Fluoride contamination in drinking water is a serious problem
in several parts of East Asia and South America (Guo et al.,
2010; Armienta and Segovia, 2008; Wang and Cheng, 2001).
The acceptable safe limit of fluoride recommended by the
World Health Organization (WHO) is 1.5 mg/L (World Health
Organization, 2004).
Current methods used to remove fluoride from water can be
divided into three categories, precipitation, membrane tech-
niques and adsorption. Precipitation of fluoride with calcium
and aluminum salts has been used to remove fluoride from
high concentrations of fluoride to 2 mg/L (Fan et al., 2003).
Membrane techniques such as reverse osmosis (Ndiayea et al.,
2005), nanofiltration (Diawara, 2008), donnan dialysis (Durmaz
et al., 2005), electrocoagulation (Zhu et al., 2007) and electro-
dialysis (Kabay et al., 2008) have been developed to effectively
remove the fluoride from water. Adsorption is a cost-effective
and extensively used method that has recently received
a great deal of attention. Accordingly, many studies have been
conducted to identify effective, low-cost adsorbents (Chen
et al., 2011; Sivasankara et al., 2010). However, there is still
a great demand for identification of environmentally friendly,

* Corresponding author. Tel./fax: þ86 25 8359 2903.

E-mail address: jpzhai@nju.edu.cn (J. Zhai).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.049
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4 5737

simple, and low-cost technologies for the removal of fluoride In this study, a polyaniline (PANI) modified electrode
from drinking water. reactor was designed and evaluated as an electrically
In recent years, there has been considerable interest in switched ion exchanger for the removal of fluoride from
conducting polymers owing to their extensive potential aqueous solution. A series of batch tests were also conducted
applications in areas such as microelectronics, composite to identify the key parameters for electrochemical defluor-
materials, catalysts, and chemical sensing. In addition, con- idation by PANI (such as pH and terminal potential), and flow
ducting polymers used as anion exchanger materials have experiments were employed to optimize future reactor
shown new potential applications in water and wastewater design.
treatment. Use of adsorbents such as polyaniline has been
reported for the removal of heavy metal ions such as Hg(II)
(Wang et al., 2009) and Cr(VI) from water (Reza, 2006). 2. Experimental section
Recently, studies have been conducted to investigate the
adsorption of fluoride onto polyanile (PANI) and polypyrrole 2.1. Chemicals
(PPy) (Karthikeyan et al., 2009a,b). Karthikeyan et al. found
that fluoride ions could be removed from aqueous solutions by Fluoride (NaF, >99%) was purchased from Fluka. Aniline was
PANI via doping of the polymer matrix. However, batch tests purified by distillation under reduced pressure, kept in the
showed that the adsorption capacity of pure PANI for fluoride dark, and stored in a refrigerator until use. Ultrapure water
was relatively low, being 0.78 mg/g at pH 7.0 when a 50 mg/ (18.25 MU cm) was employed to prepare all solutions used in
50 ml dose was used to treat water with initial fluoride ion this study, and the solution pH was adjusted by the addition of
concentrations of 2e10 mg/l. HCl and NaOH (guaranteed grade). Other chemicals were of
One of the most effective approaches to enhancement of analytical grade and used without further purification.
the defluoridation capacities of adsorbents such as polyaniline Tap water contaminated with fluoride was prepared by the
is development of PANI-based composites. For example, the addition of NaF to original tap water to obtain final
alumina composites of PANI and PPy exhibited an appreciable F concentration of 5 and 10 mg L1. The pH was adjusted to
capacity for the removal of fluoride ions from water, with 5.0 by the addition of HCl. The water quality of the original tap
defluoridation capacities of 6.6 and 8.0 mg/g, respectively water is shown in Table 1.
(Karthikeyan et al., 2009c). Chitosan composites of PANI and
PPy were also found to enhance the amounts of fluoride ions 2.2. Preparation of the working electrodes
adsorbed to 5.9 and 6.7 mg/g, respectively (Karthikeyan et al.,
2011). Polyaniline-tamarind seed biomaterial powders were Electrochemical removal of fluoride was conducted in
fabricated for adsorption of fluoride (1e10 mg/L solutions) and a 0.10 mol/L NaCl supporting electrolyte solution. Accordingly,
found to have an adsorption capacity of 4.8 mg/g (Subramanian the electrochemical polymerization of aniline was conducted
and Ramalakshmi, 2010). In another study, a novel nano- in an HCl medium (200 mL, 1.0 M) containing 0.2 M distilled
composite with magnetic properties combining both poly- aniline. Two pieces of indium tin oxide (ITO) conductive glass
pyrrole and Fe3O4 was prepared and used for defluoridation. were employed as the anode and cathode. Polyaniline was
The fluoride adsorption capacity of the PPy/Fe3O4 nano- deposited onto the anode at 2.0 V for 5 min. After synthesis,
composites was found to range from 17.63 to 22.31 mg/g. the PANI films were washed with 0.1 M HCl solution to remove
However, use of conducting polymers as adsorbents in
batch-test mode could not fully utilize their remarkable
advantage of electroactivity. Therefore, a novel technique,
Table 1 e Quality of original tap water.
electrically switched ion exchange (ESIX), was recently inves-
tigated and applied for the removal of toxic ions from aqueous Item Value
solutions. In the ESIX technique, the electrochemical oxidation Cations
or reduction of the electroactive species is conducted by Naþ 20 mg/L
applying an anodic potential to the film concurrently with ion Kþ 3.5 mg/L
uptake and elution. In a previous study, Weidlich et al. (2005) Ca2þ 15.6 mg/L
used the ion exchanger ability of PPy to develop an electro- Mg2þ 2.4 mg/L
Zn2þ 0.06 mg/L
chemically switchable ion exchanger for softening drinking
Al3þ 0.01 mg/L
water. Lin et al. (2006) investigated the perchlorate removal
NHþ4 2 mg/L
process based on ESIX in a conventional three-electrode system Anions
using polypyrrole deposited carbon nanotubes as the working Fluoride 0.2 mg/L
electrode. They demonstrated that the redox switching of pol- Chloride 38 mg/L
ypyrrole was accompanied by the exchange of perchlorate ions Nitrate 1.2 mg/L
into or out of the polymer. We previously reported the electri- Sulfate 15 mg/L
Phosphate N.D.a
cally controlled anion-exchange process for the removal and
pH 7.2
release of perchlorate and chromium using poly(aniline-co-o- Electric conductivity 252 mS/cm
aminophenol) (Zhang et al., 2009, 2010). Taken together, the Free chlorine 0.3
results of these previous studies indicate the potential for Combined chlorine 0.05
development of a green process for fluoride removal from water
a N.D., not detected.
using polyaniline-based materials.
5738 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4

the unreacted aniline, after which they were rinsed with

ultrapure water, and finally dried using a hairdryer. TG/DTA
analysis showed that the average weight of PANI on the ITO
conductive glass was about 15  0.5 mg/cm2.

2.3. Electrode reactor and electrochemical experiments

procedure

Two series of defluoridation experiments, batch tests and flow

experiments were performed. The batch tests were conducted
in reservoir tank containing 500 mL of fluoride solution, and
0.5 mL aliquots of the solution were sampled from the reser-
voir tank at intervals of 1 or 2 min. The effects of pH (pH 4e9)
and terminal potential (0.5, 1, 1.5, 2.5, 3 V) on electrochemical
defluoridation by PANI were investigated. Following treat-
ment, the PANI films were regenerated at a negative voltage of
1.0 V in 100 mL of HCl solution (0.10 M) for 5 min.
The polyaniline (PANI) modified electrochemical defluor-
idation reactor was employed for the flow experiments.
During the flow experiments, potentiostatic anion exchange
at a terminal voltage of 1.5 V was applied. Tap water
contaminated with fluoride was passed through the flow cell
of the electrode reactor once. The treated tap water flowing
from the outlet was then sampled for chemical analysis. All
runs of both the batch test and flow experiments were con-
ducted at a velocity of 10 mL/min while maintaining room
temperature at 25  C.
Fig. 1 shows the schematic view of the experimental setup
and the structure of the electrochemical defluoridation
reactor. Two pieces of indium tin oxide (ITO) conductive glass
modified with and without PANI film were employed to act as
the anode and cathode, respectively. An insulated 5 mm thick
silicon rubber spacer was inserted between the anode and
cathode, and the effective surface area of each electrode was
3  5 cm2. Therefore, the working volume of the flow cell was
7.5 mL. A DC voltage-stabilized power supply (purchased from
Shanghai Liyou Electrification Co., Ltd., China) was connected
to both the anode and cathode to control the potential. The Fig. 1 e A schematic view of the experimental setup of the
fluoride solution was fed into the flow cell from bottom to PANI-modified electrolytic defluoridation indium tin oxide
head using a syringe pump (Baoding Longer Precision Pump (ITO) conductive glass slide electrode for flow experiments
Co., Ltd., China). (a), and structure of the electrochemical defluoride flow
cell (b).
2.4. Analysis

The fluoride concentrations were evaluated by ion chroma-

tography (IC) using a Dionex ICS-2000 ion chromatography
with a conductivity detector and a 25 mL sample loop. The
eluent was a 30 mM potassium hydroxide buffer solution and
the total separation time was 10 min. The typical experi-
mental error was lower than 5% for all experimental results.
The XPS measurements were conducted on a Thermo ESCA-
LAB 250 spectrometer with an Al Ka X-ray source (1486.6 eV). XPS
analyses of three different samples (raw PANI, PANI after treat-
ment and regenerated PANI) were conducted to demonstrate
the electrically controlled anion exchange of the PANI films. All
the binding energies were referenced to C1s neutral carbon peak
at 284.6 eV.
To assess the possible risk of aniline release from the
electrode, the concentration of aniline in the effluent following
treatment of fluoride-spiked tap water was analyzed by high-
performance liquid chromatography (HPLC) as previously
described (Tanaka et al., 2009).
3. Results and discussion
3.1. Electrically controlled anion-exchange mechanism
Fig. 2A shows three images of PANI films, PANI synthesized in
the HCl solution in the oxidized state, PANI doped in a solution
containing fluoride at pH 5 (PANI-F), and de-doped PANI in the
reduced state (PANI-R). It is well known that polyaniline can
exist in three different discrete oxidation states at the molec-
ular level, both in the doped and undoped forms (Chiang and
Macdiarmid, 1986). Accordingly, the color of the polyaniline
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4
Fig. 2 e Images of PANI films (A), XPS spectra of survey
scan (B) and XPS spectra of N1s (C) for PANI films: original
PANI film (a); PANI film after electrochemical fluoride
removal (b); PANI film after regeneration (c).
5739
film will change with the potential of the polyaniline film and
the pH value of the electrolyte (Yuan et al., 1995). As shown in
Fig. 2A, the raw deep green PANI film (Fig. 2A-a) turned violet
blue after fluoride uptake (Fig. 2A-b) and turned light green
after regeneration (Fig. 2A-c).
To confirm the presence of fluoride ions in the doped PANI,
XPS was used to characterize the three samples. Curves aec in
Fig. 2B are the XPS spectra for PANI, PANI-F and PANI-R,
respectively. It is obvious that, in the PANI film after fluoride
removal (curve b), the F1s peak appeared with the decrease of
the Cl2p peak when compared to the raw PANI film (curve a).
Moreover, the F1s peak disappeared after regeneration of the
PANI film (curve c). The contents of F1s, Cl1s and the three
major species of N1s are listed in Table 2. Based on the data
shown in Table 2, the Cl content in PANI films decreased from
4.68% to 0.15% after fluoride uptake, while the F content
increased from 0 to 2.96%. The level of F dropped after
regeneration of the PANI film. These results suggest that the
uptake and elute of F ions can be controlled by modulating
the potential of the PANI film.
To further explain the anion-exchange mechanism, this
study focused on the molecular conformational change of
nitrogen atoms. The XPS core level spectra of N1s for the
three samples are presented in Fig. 2C. The broad peaks in
the spectra indicate the existence of several different struc-
tures in the PANI films. Since the nitrogen atoms in PANI film
could be classified into three groups, nitride (]Ne), amine
(eNHe), and doped imine (eNHþe), the N1s spectra are
reasonably deconvoluted into three Gaussian peaks at
399.72, >400.00 and>402.00 eV, respectively (Chen et al.,
2002). The nitride/imine ratio of the original PANI film ob-
tained in HCl was 0.72, which is smaller than the theoretical
value of 1 for the 50% intrinsically oxidized emeraldine.
These findings imply that some imine species were con-
verted into amine ones in the HCl solution. For the PANI film
after fluoride removal, the ratio increased to 2.99, indicating
that PANI was less protonated (Kang et al., 1998). The results
also revealed that the binding energy of various N1s
increased to some extents after fluoride removal. These
findings could be attributed to the morphological changes in
amine groups and the secondary doping of F ions on PANI
chains, i.e. changing from compact coils to expanded coils
(Reghu et al., 1993), which resulted in longer bond length and
higher bond energy. Another interesting observation is that
the nitride/imine ratio decreased to 0.26 after regeneration,
indicating that the PANI film had been regenerated well and
could be used repeatedly.
Based on the XPS results, we proposed an ESIX mechanism
of defluoridation on PANI films. The schematic illustration of
the polymerization of PANI and fluoride ion intake and elution
with the oxidation and reduction of the polyaniline film is
shown in Fig. 3. First, polymerization of free aniline monomers
occurred in HCl solution, during which time Cl was doped
into the PANI film. Second, in the electrochemical anion-
exchange process, the loss of electrons from PANI chains was
caused by morphological changes in the nitrogen atoms;
therefore, the secondary doping of PANI by F ions occurred at
a suitable anodic voltage. A portion of the Cl ions in the PANI
film was then eluted into the solution via anion exchange, and
F ions were extracted from the solution by secondary doping,
5740 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4

Table 2 e Distribution of Cl, F, and N species on PANI at different stages.a

Element Cl1s F1s -N] eNHe eNHþ$e

PANI Peak BE/eV 199.47 685.99 399.39 400.48 402.38 403.72

Atom % 4.36 N.D.c 0.93 5.56 0.94 0.36
PANI-F Peak BE/eV 199.73 685.99 399.57 400.83 402.46 403.98
Atom % 0.15 2.96 2.06 5.55 0.49 0.2
PANI-Rb Peak BE/eV 200.45 685.99 399.39 400.57 402.43 403.75
Atom % 2.74 N.D.c 0.43 6.94 1.28 0.37

a BE, binding energy.

b PANI-R, PANI regenerated.
c N.D., not detected.

thereby decreasing the fluoride concentration. Finally, the
used PANI film could be regenerated in HCl solution at
a certain negative voltage for F ion dedoping, which enabled
the PANI electrode to be used repeatedly. A similar process
was observed in a study of perchlorate removal by PPy (Lin
et al., 2006).
3.2. Effect of terminal potential on fluoride removal
Time-course changes in fluoride uptake by PANI films during
potentiostatic batch tests conducted different terminal
potentials are shown in Fig. 4. In this study, the fluoride
removal capacity (qe) can be calculated by:
   

F 0 F e  V
qe ¼
wS
where [F]0 is the initial fluoride concentration (mg/L), and
[F]e is the equilibrium fluoride concentration (mg/L), V is the
solution volume (0.1 L), w is the average weight of PANI on the
ITO conductive glass (15  0.5 mg/cm2), and S is the effective
surface area of each electrode (3  5 cm2).
In the present study, the removal of fluoride by adsorption
in the PANI film was 0.21 mg/g without terminal potential,
which was similar to the results of Karthikeyan’s study
(Karthikeyan et al., 2009a). As shown in Fig. 4, the external
voltage resulted in a significantly enlarged fluoride removal
capacity (qe) of the PANI films based on the electrically
controlled anion-exchange mechanism. These results
showed that terminal potential values had a major impact on
fluoride removal by PANI. The optimal removal was observed
around 1.5 V at pH 6.0 and the amount of the fluoride removal
reached 20.49 mg/g, which was higher than that observed for
adsorbents. The fluoride removal capacity (qe) increased
rapidly within 5 min and then continued to increase at lower
speeds until equilibrium was reached at 10 min. These find-
ings indicated a rapid removal of fluoride by PANI under
these conditions. Both lower and higher potential leads to
a slower reaction velocity and smaller fluoride removal
capacity. Within the range of 0.5 Ve1.5 V, the increase in
potential means the higher the current density, which leads
to a higher fluoride removal capacity. However, due to the
overoxidation reaction of PANI films, the terminal potentials
above 2.0 V results in decreased the fluoride removal
efficiency.
3.3. Effect of pH on fluoride removal
For quantitative analysis of the fluoride removal capability of
the PANI-modified electrode, a series of potentiostatic elec-
trochemical experiments were performed at pH 5e9 in
a solution containing 50 mg/L fluoride and 0.1 M NaCl. The
effects of pH on the electrochemical fluoride removal capac-
ities are presented in Fig. 5. The solution pH values had
a significant impact on the fluoride removal by the PANI film,
with apparent inhibition in the conditions at pH 7. For
example, the final fluoride removal rate was more than 90% at

Fig. 3 e Schematic illustration of the polymerization of PANI and fluoride ion intake and elution with the oxidation and
reduction of PANI film.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4 5741

3.4. Treatment of fluoride-contaminated tap water

20

Fluoride retention in the defloration reactor at a potential of

1.5 V was investigated under different fluoride concentrations
15 (5 and 10 mg/L). The characteristic shape and position of the
breakthrough curves (BTC) on the time axis for different
concentrations are shown in Fig. 6. As shown in Fig. 6, the
q e (m g /g )

10 values of Ct/C0 were very low and the breakthrough curves

0V
0.5 V
1.0 V
5
1.5 V
2.5 V
3.5 V
0 the increase in initial fluoride concentration caused the PANI
0 5 10 15
Time (min)
Fig. 4 e Effect of terminal potential on fluoride uptake by
PANI films during potentiostatic treatment (C0 [ 10 mg/L;
I [ 0.1 M NaCl; pH [ 6.0; volume of solution, 500 ml). The
amount of PANI was approximated to 225 mg for each test.
pH 4e6, but decreased abruptly to 48.8% at pH 7 and to less
than 25% at pH 8 and 9.
The reason for the aforementioned phenomenon is that
the electrochemical activity of PANI under acidic solutions is
much better than under alkaline conditions. It is because of
there is a much higher doped imine/nitride ratio (i.e. eNHþ$e/
]Ne ratio) in the structure of the polyaniline chain. PANI
contains nitrogen atoms that are easy to protonate, and the
protonation of eNHþ$e/]N, which are both pH sensitive, is
simultaneously achieved at a pKa in the range of 5.5e8 (Slim
et al., 2008). Because the concentrations of Hþ played an
important role in the doping process, the treatment of
fluoride-contaminated water should be conducted under
acidic conditions.
20
were smooth in the beginning. During this period, the fluoride
concentration of treated tap water was below the WHO rec-
ommended level of 1.0 mg/L. As more fluoride-contaminated
tap water passed through the flow cell, the BTC became
sharper, rapidly reaching the breakthrough point. A sharp BTC
was observed at higher feed fluoride concentrations because
20 films to become saturated with the fluoride ion earlier in the
experiment. At initial concentrations of 5 mg/L and 10 mg/L,
the breakthrough capacities were 20.08 mg/g and 19.24 mg/g,
respectively, which indicates high fluoride removal by the
reactor.
To evaluate the reuse value of the PANI films, the consec-
utive treatment-regeneration process was conducted three
times. HCl solution at 0.1 mol/L of was used as the desorbing
agent. Fig. 7 shows the characteristic shape and position of
BTC for fresh PANI film, first regenerated PANI film and second
regenerated PANI film. The fluoride removal capacity of fresh
PANI film was slightly higher than that of the other films,
while the fluoride removal capacity of the first and second
regenerated samples were similar. Specifically, the break-
through capacities of the fresh PANI, first regenerated PANI
and second regenerated PANI were 20.07 mg/g, 18.79 mg/g and
18.35 mg/g, respectively.
In addition, the defluoridation capacity of PANI for five
treatment-regeneration cycles is illustrated in Table S-1
(Support Information), and all tests were conducted three
1.0
C0 = 10 mg/L
C0 = 5 mg/L
0.8
0.015
0.6
0.010
C t/ C 0

Ct /C0

15
0.4
qe (mg/g)

0.005

10 0.2
0.000
0 5 10 15 20 25
0.0 Time (min)
5
pH = 4 pH = 5 0 20 40 60 80 100 120 140 160 180
pH = 6 pH = 7
pH = 8 pH = 9
Time (min)
0
0 2 4 6 8 10 12 14 Fig. 6 e Fluoride breakthrough continuous flow
experiments at initial fluoride concentrations of 5 and
Time (min)
10 mg/L (I [ 0.1 M NaCl; pH [ 6.0; flow rate, 10 ml/min;
Fig. 5 e Effect of pH on fluoride removal during terminal potential, 1.5 V). The broken line (Ce/C0 [ 1)
potentiostatic treatment (C0 [ 10 mg/L; I [ 0.1 M NaCl; represents the breakthrough point. The inset shows
volume of solution, 100 ml; terminal potential, 1.5 V). details during the first 25 min.
5742 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4

1.0 used for water with an initial fluoride concentration ranging

0.06 from 5 to 35 mg/L, pH 5e7 and 25  C, which is similar to the
conditions that PANI-modified electrode reactors are usually
0.8 used under. However, the applied voltages in these systems are
0.04
much higher than the voltage used in the PANI-modified elec-
C /C

trode reactor developed in this study, which was 5e30 V

0.6 0.02
(Emamjomeh and Sivakumar, 2009; Khatibikamal et al., 2010; Lo
C t /C 0

et al., 2007). These findings indicated that the ESIX technique

0.4
0.00 reduces energy use.
0 10 20 30 40 50 60 70
Several conducting polymers and their composites were
Time (min)
recently investigated for the ability to adsorb fluoride. The
0.2 fresh PANI adsorption capacity (Qe) and specific conditions such as equi-
first regeneration
second regeneration
librium time (te), initial concentration (C0), optimum pH,
temperature (T ) and regeneration method of these polymers
0.0 are listed in Table 3. Because of the special dopingededoping
0 20 40 60 80 100 capacity of conducting polymers such as PANI and PPy under
Time (min) acidic conditions, the optimum pH for fluoride removal by
these adsorbents was less than 7, which is similar to the results
Fig. 7 e Comparison of fluoride removal using fresh PANI,
of the present study. When compared to these conducting
first regenerated sample and second regenerated sample
polymer adsorbents, the ESIX technique using the PANI elec-
in continuous flow experiments (C0 [ 10 mg/L; I [ 0.1 M
trode was also a good method for fluoride removal. By adding
NaCl; pH [ 6.0; flow rate, 10 ml/min; terminal potential,
a certain terminal voltage to the PANI-modified electrode, the
1.5 V). The inset shows the details during the first 70 min.
defluoridation capability of the purified PANI was significantly
enhanced from 0.78 to about 20 mg/g. These findings indicated
that the ESIX technique can effectively enlarge the anion-
times to eliminate the error range. The results revealed that exchange capability of the conducting polymers. Further
the defluoridation capacity of PANI was still 89% at the fifth studies using conducting polymer composites such as the PPy/
cycle, which suggests that the PANI-modified electrode Fe3O4 nanocomposite as electrode materials to improve fluo-
possessed the potential for regeneration and reuse. ride removal efficiency are still necessary.

3.5. Comparison of ESIX technique with other methods 3.6. Safety and durability assessment

Using this method of fluoride removal via electrically controlled
anion exchange, the uptake and elute of fluoride could be easily
controlled by modulating the potential of the PANI film. It is
interesting to note that the removal of fluoride is a rapid process,
and the PANI electrode was easily regenerated. Since the PANI-
modified electrode reactor removes fluoride by electrochemi-
cally controlled ion exchange, it is necessary to compare it with
other electrochemical methods, such as electrocoagulation. In
the electrocoagulation unit, aluminum electrodes are usually
applied to produce aluminum ions as a coagulant for the
removal of fluoride. These electrocoagulation units are usually
The concentration of aniline in the treated fluoride-spiked tap
water was analyzed by HPLC after successive reac-
tioneregeneration cycles and no aniline was detected in the
effluent during five adsorptioneregeneration cycles. However,
it is important to note that there is still a risk of release of
aniline in future applications of this method. Furthermore,
this risk may increase as the service time and number of
adsorptioneregeneration cycles increases, and some small
PANI particles may even enter the effluent. To avoid such
a risk, additional studies are needed to design more stable
PANI-based composites for electrode coatings or to combine

Table 3 e Adsorption capacity of other typical adsorbents.

Adsorbent Qe, mg/g te, min C0, mg/L Optimum pH T,  C Regeneration Reference

PANI 0.78 5 4 3 30 0.1 M NH4OH Karthikeyan et al. (2009a)

PPy 6.37 10 10 <7 30 Unknown Karthikeyan et al. (2009b)
PANI/alumina 5.8 20 10 3 30 Unknown Karthikeyan et al. (2009c)
PPy/alumina 6.7 20 10 3 30 Unknown
PANI/chitosan 5.5 10 10 3e4 30 Unknown Karthikeyan et al. (2011)
PPy/chitosan 5.8 10 10 3e4 30 Unknown
PANI/TS 10.7 30 10 unknown 30 Unknown Subramanian
and Ramalakshmi (2010)
PPy/Fe3O4 22.31 20 100 6.5 25 2 M HCl Bhaumik et al. (2011)
nanocomposite
PANI-modified 20 10 10 6 25 Negative voltage This study
electrode
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4
this reactor with other techniques such as membrane tech-
niques in practical application. Overall, this defluoridation
method via electrically controlled anion exchange by a poly-
aniline modified electrode reactor provides an alternative
technique for the removal of fluoride from aqueous solutions.
4. Conclusions
Fluoride removal via electrically controlled anion exchange by
a polyaniline modified electrode reactor was investigated in
this study. Based on the results of XPS measurements of the
PANI films and ion chromatography analysis of the solutions
before and after fluoride removal, the fluoride ion uptake and
elution was well controlled by modulation of the potential of
the PANI-modified electrode. Batch tests and flow experiments
showed that the PANI-modified electrode reactor had good
fluoride removal capability (about 20 mg/g) in acidic solutions
at a positive voltage of 1.5 V, giving a high removal efficiency for
trace fluoride ions in contaminated tap water. Moreover, the
doped fluoride on the PANI film could be effectively eluted at
a negative voltage of 1.0 V in 0.10 M HCl solution, and the
regenerated PANI films still had high fluoride uptake and elute
efficiencies in the three-time consecutive treatment-regener-
ation studies. Therefore, the results of this study suggested
that electrically controlled anion exchange by a polyaniline
modified electrode reactor has great potential for the removal
of fluoride from drinking water. Accordingly, further studies
investigating application in this area are warranted.
Acknowledgment
This work was supported by the Natural Science Foundation
of China (Grants 51008154), foundation of State Key Labora-
tory of Pollution Control and Resource Reuse of China, and the
Scientific Research Foundation of Graduate School of Nanjing
University (Grants 2010CL07).
Appendix. Supplementary data
Supplementary data related to this article can be found online
at doi:10.1016/j.watres.2011.08.049.
references
Armienta, M.A., Segovia, N., 2008. Arsenic and fluoride in the
groundwater of Mexico. Environmental Geochemistry and
Health 30, 345e353.
Bhaumik, M., Leswifi, T.Y., Maity, A., Srinivasu, V.V., Onyango, M.S.,
2011. Removal of fluoride from aqueous solution by polypyrrole/
Fe3O4 magnetic nanocomposite. Journal of Hazardous Materials
186, 150e159.
Chen, L., Wang, T.J., Wu, H.X., Jin, Y., Zhang, Y., Dou, X.M., 2011.
Optimization of a FeeAleCe nano-adsorbent granulation
process that used spray coating in a fluidized bed for fluoride
removal from drinking water. Powder Technology 206, 291e296.
5743
Chen, Y.J., Kang, E.T., Neoh, K.G., 2002. Electroless polymerization
of aniline on platinum and palladium surfaces. Applied
Surface Science 185, 267e276.
Chiang, J.C., Macdiarmid, A.G., 1986. Polyaniline e protonic acid
doping of the emeraldine form to the metallic regime.
Synthetic Metals 13, 193e205.
Diawara, C.K., 2008. Nanofiltration process efficiency in water
desalination. Separation and Purification Reviews 37, 303e325.
Durmaz, F., Kara, H., Cengeloglu, Y., Ersoz, M., 2005. Fluoride
removal by Donnan dialysis with anion exchange membranes.
Desalination 177, 51e57.
Emamjomeh, M.M., Sivakumar, M., 2009. Fluoride removal by
a continuous flow electrocoagulation reactor. Journal of
Environmental Management 90, 1204e1212.
Fan, X., Parker, D.J., Smith, M.D., 2003. Adsorption kinetics of
fluoride on low cost materials. Water Research 37, 4929e4937.
Guo, Q.H., Wang, Y.X., Guo, Q.S., 2010. Hydrogeochemical genesis
of groundwaters with abnormal fluoride concentrations from
Zhongxiang City, Hubei Province, central China.
Environmental Earth Sciences 60, 633e642.
Kabay, N., Arar, O., Samatya, S., Yuksel, U., Yuksel, M., 2008.
Separation of fluoride from aqueous solution by
electrodialysis: effect of process parameters and other ionic
species. Journal of Hazardous Materials 153, 107e113.
Kang, E.T., Neoh, K.G., Tan, K.L., 1998. Polyaniline: a polymer with
many interesting intrinsic redox states. Progress in Polymer
Science 23, 277e324.
Karthikeyan, M., Kumar, K.K.S., Elango, K.P., 2011. Batch sorption
studies on the removal of fluoride ions from water using
eco-friendly conducting polymer/bio-polymer composites.
Desalination 267, 49e56.
Karthikeyan, M., Satheeshkumar, K.K., Elango, K.R., 2009a.
Defluoridation of water via doping of polyanilines. Journal of
Hazardous Materials 163, 1026e1032.
Karthikeyan, M., Satheeshkumar, K.K., Elango, K.P., 2009b.
Removal of fluoride ions from aqueous solution by conducting
polypyrrole. Journal of Hazardous Materials 167, 300e305.
Karthikeyan, M., Satheeshkumar, K.K., Elango, K.P., 2009c.
Conducting polymer/alumina composites as viable adsorbent
for the removal of fluoride ions from aqueous solution. Journal
of Fluorine Chemistry 130, 894e901.
Khatibikamal, V., Torabian, A., Janpoor, F., Hoshyaripour, G.,
2010. Fluoride removal from industrial wastewater using
electrocoagulation and its adsorption kinetics. Journal of
Hazardous Materials 179, 276e280.
Lin, Y.H., Cui, X.L., Bontha, J., 2006. Electrically controlled anion
exchange based on polypyrrole and carbon nanotubes
nanocomposite for perchlorate removal. Environmental
Science & Technology 40, 4004e4009.
Lo, S.L., Hu, C.Y., Kuan, W.H., 2007. Simulation the kinetics of fluoride
removal by electrocoagulation (EC) process using aluminum
electrodes. Journal of Hazardous Materials 145, 180e185.
Ndiayea, P.I., Moulin, P., Dominguez, L., Millet, J.C., Charbit, F.,
2005. Removal of fluoride from electronic industrial effluent by
RO membrane separation. Desalination 173, 25e32.
Reghu, M., Cao, Y., Moses, D., Heeger, A.J., 1993. Counterion-
induced processibility of polyaniline e transport at the metal-
insulator boundary. Physical Review B 47, 1758e1764.
Reza, A., 2006. Application of polyaniline and its composites for
adsorption/recovery of chromium (VI) from aqueous
solutions. Acta Chimica Slovenica 53, 88e94.
Slim, C., Ktari, N., Cakara, D., Kanoufi, F., Combellas, C., 2008.
Polyaniline films based ultramicroelectrodes sensitive to pH.
Journal of Electroanalytical Chemistry 612, 53e62.
Sivasankara, V., Ramachandramoorthy, T., Chandramohan, A.,
2010. Fluoride removal from water using activated and MnO2-
coated Tamarind Fruit (Tamarindus indica) shell: batch and
column studies. Journal of Hazardous Materials 177, 719e729.
5744 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 3 6 e5 7 4 4
Subramanian, E., Ramalakshmi, R. Dhana, 2010. Pristine,
purified and polyaniline-coated tamarind seed (Tamarindus
indica) biomaterial powders for defluoridation: synergism
and enhancement in fluoride-adsorption by polyaniline
coating. Journal of Scientific and Industrial Research 69,
621e628.
Tanaka, T., Hachiyanagi, H., Yamamoto, N., Iijima, T., Kido, Y.,
Uyeda, M., Takahama, K., 2009. Biodegradation of endocrine-
disrupting chemical aniline by microorganisms. Journal of
Health Science 55, 625e630.
Wang, G.X., Cheng, G.D., 2001. Fluoride distribution in water and
the governing factors of environment in arid north-west
China. Journal of Arid Environments 49, 601e614.
Wang, J., Deng, B.L., Chen, H., Wang, X.R., Zheng, J.J., 2009.
Removal of aqueous Hg (II) by polyaniline: sorption
characteristics and mechanisms. Environmental Science &
Technology 43, 5223e5228.
Weidlich, C., Mangold, K.-M., Jüttner, K., 2005. Continuous ion
exchange process based on polypyrrole as an
electrochemically switchable ion exchanger. Electrochimica
Acta 50, 5247e5254.
World Health Organization (WHO), 2004. Guidelines for
Drinkingwater Quality, third ed., vol. 1. WHO, Geneva.
Yuan, R.K., Gu, Z.P., Yuan, H., 1995. Studies of controllable color-
change properties of polyaniline film. Synthetic Metals 69,
233e234.
Zhang, Y., Li, Q., Tang, R., Hu, Q.C., Sun, L., Zhai, J.P., 2009.
Electrocatalytic reduction of chromium by poly(aniline-co-o-
aminophenol): an efficient and recyclable way to remove
Cr(VI) in wastewater. Applied Catalysis B: Environmental 92,
351e356.
Zhang, Y., Mu, S.L., Deng, B.L., Zheng, J.Z., 2010. Electrochemical
removal and release of perchlorate using poly(aniline-co-o-
aminophenol). Journal of Electroanalytical Chemistry 641,
1e6.
Zhu, J., Zhao, H., Ni, J., 2007. Fluoride distribution in
electrocoagulation defluoridation process. Separation and
Purification Technology 56, 184e191.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Optimization of capacity and kinetics for a novel bio-based

arsenic sorbent, TiO2-impregnated chitosan bead

Sarah M. Miller a, Matthew L. Spaulding a, Julie B. Zimmerman a,b,*

a
Department of Chemical and Environmental Engineering, Yale University, United States
b
School of Forestry and Environmental Studies, Yale University, United States

article info abstract

Article history: The optimization of TiO2-impregnated chitosan beads (TICB) as an arsenic adsorbent is
Received 2 May 2011 investigated to maximize the capacity and kinetics of arsenic removal. It has been previ-
Received in revised form ously reported that TICB can 1) remove arsenite, 2) remove arsenate, and 3) oxidize arsenite
24 August 2011 to arsenate in the presence of UV light and oxygen. Herein, it is reported that adsorption
Accepted 25 August 2011 capacity for TICB is controlled by solution pH and TiO2 loading within the bead and
Available online 1 September 2011 enhanced with exposure to UV light. Solution pH is found to be a critical parameter,
whereby arsenate is effectively removed below pH 7.25 and arsenite is effectively removed
Keywords: below pH 9.2. A model to predict TICB capacity, based on TiO2 loading and solution pH, is
Arsenic presented for arsenite, arsenate, and total arsenic in the presence of UV light. The rate of
Water removal is increased with reductions in bead size and with exposure to UV light. Phosphate
Chitosan is found to be a direct competitor with arsenate for adsorption sites on TICB, but other
TiO2 relevant common background groundwater ions do not compete with arsenate for
Bio-based adsorption sites. TICB can be regenerated with weak NaOH and maintain full adsorption
Sustainable capacity for at least three adsorption/desorption cycles.
ª 2011 Published by Elsevier Ltd.

1. Introduction Arsenic in groundwater primarily exists as arsenite (As(III))

Over 120 million people in Bangladesh and India rely upon
groundwater with elevated arsenic levels as their primary
source of drinking water (Ratnaike, 2003). This “mass
poisoning” has resulted in many negative health outcomes,
including toxicity to the liver, skin, kidney, and cardiovascular
system, as well as multiple cancers (Agros et al., 2010). As
community scale infrastructure for water treatment is not
likely in the near future, point-of-use technologies, most
relying on adsorption, have been advocated for arsenic removal
(Petrusevski et al., 2008). Despite promising lab results for many
of these technologies, no technology has adequately demon-
strated long-term effectiveness and sustainability in the field
(Petrusevski et al., 2008).
and arsenate (As(V)) (Bhattacharya et al., 2007). Arsenite is up
to 60 times more toxic than arsenate (Ratnaike, 2003; Tien
et al., 2004). Because arsenite is uncharged at environmen-
tally relevant pH, it is also more difficult to remove than
arsenate, which is negatively charged at environmentally
relevant pH. Metal oxide sorption technologies, usually based
on Ti oxides, Fe oxides, or Al oxides, have the unique ability to
remove both As(III) and As(V). This is a major advantage
relative to technologies like ion exchange, which can only
remove As(V) and require a pre-oxidation step to achieve total
arsenic removal.
The synthesis of TiO2-impregnated chitosan bead (TICB),
a bio-based adsorbent with promising arsenic removal
capacity, has been previously reported (Miller and

* Corresponding author. Department of Chemical and Environmental Engineering, Yale University, United States. Tel.: þ1 203 432 9703.
E-mail address: Julie.Zimmerman@Yale.edu (J.B. Zimmerman).
0043-1354/$ e see front matter ª 2011 Published by Elsevier Ltd.
doi:10.1016/j.watres.2011.08.040
5746 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4
Zimmerman, 2010). Previously reported results were for TICB
comprised of 70% chitosan and 30% TiO2 by weight (Miller and
Zimmerman, 2010). Chitosan, a derivative of chitin, has
a positive environmental and economic profile because it is
renewable, biodegradable (Muzzarelli and Muzzarelli, 2005),
and can be isolated from the millions of tons of shellfish waste
generated globally per year (Gerente et al., 2007). TiO2 is
a nontoxic nanopowder (Deebar et al., 2009) that has shown
promise as filtration media in a point-of-entry arsenic
removal system (Bang et al., 2005). Like pure TiO2, TICB has
demonstrated removal of both arsenite and arsenate and has
demonstrated the ability to oxidize arsenite to arsenate in the
presence of UV light (Miller and Zimmerman, 2010). Although
arsenic removal by TICB is less than TiO2 on a sorbent weight
basis, TICB exceeds TiO2 sorption on a sorbent surface area
basis (Dutta et al., 2004; Bang et al., 2005; Miller and
Zimmerman, 2010), and has the added advantage of self-
separation.
The objective of this study was to optimize the arsenic
adsorption capacity and kinetics of TICB and develop
a predictive model to guide system conditions for imple-
mentation. Parameters affecting capacity and kinetics,
including solution pH, bead surface area, bead size, TiO2
loading in the bead, UV irradiation, and background ions in
the water matrix, are investigated. Optimization of the useful
lifetime of the TICB adsorbent is also considered by examining
the potential for regeneration and reuse of TICB.
2. Materials and methods
2.1. Standards and reagents
Experiments were conducted with either As(III) (arsenite) or
As(V) (arsenate) as indicated. Stock solutions of As(III) and
As(V) were prepared by dissolving NaAsO2 (Sigma Aldrich)
and Na2HAsO.47H2O (Fisher Scientific) in deionized water,
respectively. Stock solutions (10,000 mg/L) and appropriate
dilutions were prepared daily, immediately before use. Chi-
tosan was purchased from TCI America. TiO2 (anatase nano-
powder, 99.7% trace metals basis, <25 nm particle size) was
purchased from Sigma Aldrich. All other reagents were of
standard laboratory grade. HNO3 (Fisher Scientific, trace
metal grade) and HCl solutions were prepared from concen-
trated stock solutions; NaOH solutions were prepared from
pellets.
2.2. Bead preparation and characterization
TiO2-impregnated chitosan beads were prepared as reported
in (Miller and Zimmerman, 2010). In brief, chitosan was dis-
solved in 0.1 M HCl (1 g/60 mL). TiO2 (0.4242 g TiO2/1 g chitosan)
was added and mixed with a magnetic stir bar until
a homogenous solution was achieved. Unless otherwise
noted, TICB is 30% TiO2 on mass basis. Syringes with 18G1
needles were filled with the homogenous solution and loaded
into a syringe pump; bead size was adjusted by varying the
needle gage (19G1, 22G1). The syringe pump discharged the
homogenous solution into 0.1 M NaOH (20 mL solution/100 mL
0.1 M NaOH), resulting in beads. Beads were rinsed in
deionized water until filtrate reached pH 6. After drying
for > 18 h, beads were collected and stored at room temper-
ature in the dark.
BET surface area analysis was performed by Micromeritics
Analytical Services (Norcross, Georgia). Samples were
weighed at room temperature after degassing at 45  C for 16 h.
Gas adsorption analysis was conducted with krypton gas
using a TriStar II 3020 surface area and porosity system.
Bead porosity was measured with Hg intrusion by Micro-
meritics Analytical Services (Norcross, Georgia).
Bead diameter was measured with a Marathon Electronic
Digital Micrometer (0e25 mm). Reported diameter values are
the average and standard deviation of 10 randomly selected
beads from the same batch.
2.3. Adsorption experiments
Duplicate samples were prepared in 50 mL polypropylene
falcon tubes into which 40 mL arsenic solution (either arsenite
or arsenate) and the specified amount of TICB was added.
Batch adsorption experiments were conducted in
a shaking incubator (VWR 1575R), where temperature was
maintained at 25  C and samples were agitated at 150 rpm.
Unless otherwise indicated, experiments were conducted for
>185 h (Miller and Zimmerman, 2010). Irradiated samples
were continuously exposed to an 8 W, 365 nm lamp (UVP,
UVL-28 EL Series 8) (Ferguson et al., 2005) suspended 1.5 feet
from samples in a closed incubator.
In identical experimental conditions to the batch experi-
ments, kinetics of As(III) and As(V) were measured in the
presence and absence of UV light. These studies were con-
ducted by sacrificing duplicate samples at specified time
intervals for up to 317 h.
pH experiments were conducted in the presence and
absence of UV light, where [As]0 ¼ 500 mg/L pH adjustments
were made with 16 M HNO3 or 1 M NaOH.
Synthetic groundwater was prepared based on a procedure
by (Leupin and Hug, 2005). MgCO3 (112.7 mg), CaCO3 (367 mg),
and KH2PO4 (11.0 mg) were added to 900 mL of deionized water
and stirred, resulting in pH 9.06. Under rapid mixing, CO2 gas
was bubbled through this solution for 42 min, resulting in pH
5.08. A 10 mL solution of Na2SiO.39H2O (205.9 mg) dissolved in
deionized water was added to the bulk solution and stirred,
resulting in pH 5.39. Compressed air (house air filtered
through granular activated carbon) was bubbled through the
solution for 15 min, resulting in pH 6.93. 40 mL deionized
water and 50 mL of freshly prepared 10 ppm As(III) stock
solution was added to the solution to achieve 500 mg As (III)/L
initial concentration and pH 7.19.
Groundwater was collected from a tubewell at Amhiribad
High School in West Bengal, India on December 17, 2010. The
tubewell was flushed for 5 min prior to collection. Water
composition was analyzed by EnviroCheck Laboratory in West
Bengal and is as follows: 62 ppb As, 0.83 ppm Fe, 0.37 ppm Mn,
0.66 ppm P, 30.66 ppm Si, 7.5 ppm SO2- 4 , 33.18 ppm Mg,
101.0 ppm Ca, 408.2 ppm HCO 3 , 48.04 ppm Cl, 2.78 ppm K, and
14.44 ppm Na. Tubewell water was transported to Yale
University where batch experiments and analyses were per-
formed as previously described.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4
For desorption/resorption experiments, TICB (25 mg) was
saturated with As(III) or As(V), as indicated, in a batch
experiment without the presence of UV light, where [As]0
varied from 100e10,000 mg/L as indicated. Samples were
removed after > 185 h in the shaking incubator (25  C,
150 rpm), and the aqueous solution was immediately dec-
anted from the falcon tube and measured for arsenic
concentration. Deionized water (10 mL) was added to the
falcon tube to rinse the beads. After gentle agitation, the rinse
water was decanted, and the beads, contained in uncapped
falcon tubes, were allowed to dry in the fumehood for 24 h.
Beads were then reweighed and transferred to a new 50 mL
polypropylene falcon tube. 40 mL of NaOH (0.07 M, 0.67 M,
1.67 M) was added to these falcon tubes and placed in the
shaking incubator, where temperature was maintained at
25  C and samples were agitated at 150 rpm for 24 h. NaOH
was then decanted and measured for arsenic concentration.
Deionized water (10 mL) was added to the falcon tube to rinse
the beads. After gentle agitation, the rinse water was dec-
anted and the beads, contained in uncapped falcon tubes,
were allowed to dry in the fumehood for 24 h. This process
was repeated for a total of three cycles.
2.4. Analyte concentration
To measure arsenic concentration, all samples were diluted
to a maximum of 100 mg As/L and acidified with 16 M HNO3 to
reach a final concentration of 1% HNO3. Samples with neat
TiO2 nanopowder were filtered through a 0.45 mm filter prior
to dilution. Measurement of As, Mg, Ca, Si, and P was con-
ducted on a Perkin Elmer DRC-e ICP-MS. An internal
Germanium standard was used, and quality control stan-
dards were analyzed every 10 samples to verify instrument
performance. Three readings for each sample were per-
formed and an average and standard deviation for each
sample was reported. Detection limit was determined to be
0.25 mg As/L. Arsenic speciation was measured using HPLC-
ICP-MS (Perkin Elmer DRC-e) following published proce-
dures (Neubauer et al., 2004).
3. Results and discussion
3.1. Equilibrium adsorption capacity of TICB for
dissolved arsenic
3.1.1. pH
Solution pH is a critical variable in metal oxide chemistorptive
processes. Fig. 1 shows data for arsenite and arsenate removal
by TICB without UV irradiation as a function of pH, where
arsenite and arsenate both exist. These results, obtained by
deliberately altering solution pH, are consistent with
a hydroxide exchange mechanism. Above pH 7.25, the pzc of
TICB (Miller and Zimmerman, 2010), electrostatic repulsion
between arsenate and hydroxyl groups on the bead surface
prevents chemisorption. Neuturally charged arsenite is effec-
tively sorbed up to its pKa, 9.2, where it becomes a negative
oxyanion. Differences between arsenate and arsenite removal
at pH > 9.2 can be attributed to the fact that arsenate (H2AsO2 4 )
is more negatively charged than arsenite (H2AsO 3 ) in this pH
5747
100
As(III)
80 As(V)
% As removal
60
40
20
0
4 6 8 10 12
Final pH
Fig. 1 e Role of final solution pH in arsenite and arsenate
removal by TICB, where data for pH < 4.4 has been
omitted because of bead dissolution. Experiments
conducted in absence of UV light, where [As]0 [ 500 mg/L.
range and, therefore, experiences more repulsion with net
negatively charged TICB (Miller and Zimmerman, 2010).
3.1.2. Bead size
Surface area measurements were performed across adjustable
parameters in the synthesis and preparation of TICB, including
bead size (Fig. 2a). TICB surface area increases as bead size
decreases and with exposure to UV light. Because UV exposure
enhances surface area (Miller and Zimmerman, 2010), surface
area differences across bead size are more apparent after
exposure to UV light. Given that materials with high surface
area can achieve high adsorption capacities per unit mass,
these results suggest that an optimized design of TICB might
incorporate diameter reduction and UV irradiation.
To maximize efficacy, that is removal of As(III) and As(V),
there is an inherent tradeoff in that smaller particles (with
higher surface area) are more effective but require post-
treatment filtration. As such, there is likely an optimal bead
size range that balances removal efficiency while maintaining
density-separation. Fig. 2b shows the percent As(III) and As(V)
arsenic removal across a range of bead sizes, where the final
pH is <7.6 for As(III) and <7.2 for As(V), and where the dosing
of TiO2: As ratio remains constant to isolate the impact of bead
size on efficacy. The largest bead size tested (937 um)
demonstrates similar removal capacities to neat TiO2 nano-
powder. That is, bead size does not have an effect on the
removal efficiency in terms of capacity for a given set of
system conditions (i.e., arsenic oxidation state; irradiation).
This is an unexpected result because the surface area of
neat TiO2 powder is two orders of magnitude greater than that
of TICB, and the TiO2 in TICB is bound to a chitosan matrix
such that it may not all be available for arsenic removal. This
suggests that TICB, in terms of capacity, acts like TiO2 powder
given sufficient exposure time. However, this also suggests
a need to assess the system kinetics, a critical factor for actual
implementation in a field setting.
Fig. 2 also illustrates improved performance (i.e., sorption
of both As(III) and As(V)) for UV-exposed TICB relative to UV-
unexposed TICB, indicating that UV light is a variable that
may be employed to enhance adsorption capacity.
5748 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

a b
0.8
100

BET surface area (m2/g)

0.6 80

% As Removal
60
0.4

40
0.2
20

0
0
0 200 400 600 800 1000
0 200 400 600 800 1000
bead diameter (µm) bead diameter (µm)
As(III); 0 h UV irradiation
220 h UV irradiation As(III); 220 h UV irradiation
0 h UV irradiation As(V); 0 h UV irradiation
As(V); 220 h UV irradiation

Fig. 2 e Relationship between bead size (where all beads are 30% TiO2 by weight and have been exposed to UV irradiation as
noted) and a) BET surface area and b) % arsenic removal from [As]0 [ 1000 mg/L.

3.1.3. TiO2 loading conditions are included in Supplementary Information (SI.1).

Chitosan beads impregnated with increasing amounts of TiO2 This suggests that all TiO2 in the bead contributes to arsenic
were synthesized, where it was empirically determined that adsorption, regardless of the mass fraction of TiO2 in the
46% TiO2 by mass was the maximum amount of TiO2 that can bead.
be incorporated into a homogenous bead formulation. As To confirm that all TiO2 present provides adsorption
shown in Fig. 3a, the surface area of these beads increases capacity in TICB, regardless of relative TiO2 mass fraction in
with increasing concentration of TiO2 in the bead. the bead, neat TiO2 nanopowder, equivalent in mass to the
These beads were tested for As(III) and As(V) removal, in amount of TiO2 impregnated in the beads, was tested across
both the presence and absence of UV light to determine the the same experimental conditions. For removal of both As(III)
relationship between TiO2 loading and functional perfor- and As(V), in the presence or absence of UV light, arsenic
mance. Arsenic removal increases as TiO2 loading increases removal by TICB and TiO2 for a given mass of TiO2 are nearly
for all experimental conditions, as illustrated with the identical (Fig. 3 and SI.1). This supports the finding that all of
representative plot in Fig. 3b; results from other experimental the TiO2 impregnated in the bead, not just the TiO2 on the

a 0.8 b 100
BET surface area (m2/g)

80
% As removal

0.6
60
0.4
40
0.2
20

0 0
0 0.05 0.1 0.15 0 0.05 0.1 0.15
mmol TiO2 /25 mg TICB mmol TiO2/ 25 mg TICB
As(III)_UV
As(V)_UV
As(III)_UV, TiO2 powder
As(V)_UV, TiO2 powder

Fig. 3 e Relationship between mmol TiO2 incorporated into TICB (875 mm) and a) BET surface area and b) % arsenic removal
from [As]0 [ 1000 mg/L in the presence of UV light.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4 5749

bead surface, is forming chemical complexes with the arsenic

Table 1 e % removal from batch experiment where
in solution, given sufficient time to equilibrate. [As(V)]0 [ 1000 mg/L and samples were analyzed at
time [ 240 h. UV absent samples were spiked with HNO3
3.1.4. UV light to match pH drop of UV present samples. Values reported
Beads of identical size were exposed to different durations of are averages of four replicates.
UV irradiation. As shown in Fig. 4, there is a positive rela- Final pH (<7) As removal (mg/g)
tionship between surface area and UV exposure, and SEM Experimental observation
images of these beads (Miller and Zimmerman, 2010) reveal
UV absent 6.88  0.04 1054  74
changes in surface morphology after exposure to UV light. UV present 6.42  0.03 1397  11
This may be a result of a photooxidative process occurring at
the bead surface. Although UV irradiation affects the surface
of TICB, porosity measurements of beads with varying expo-
binuclear complex forms between TiO2 and both arsenite and
sures to UV light are relatively constant and suggest that UV
arsenate (Jegadeesan et al., 2010; Jing et al., 2005, 2009; Pena
irradiation does not affect pores >3 nm within the dehydrated
et al., 2006) (Fig. 5b). A similar complex, shown in Fig. 5c,
bead interior (Webb, 2001).
where one of the TiO2 coordinating sites is replaced with a UV-
The presence of UV light enhances arsenate sorption on
induced COOH group may form.
TICB (Miller and Zimmerman, 2010). Because the pH of all
It has been previously reported that TICB can oxidize As(III)
samples, including controls, is lowered with UV irradiation
to As(V) in the presence of UV light under system conditions of
(SI.2) and because lower solution pH is associated with greater
365 nm and sufficient oxygen (Miller and Zimmerman, 2010).
arsenic removal (Fig. 1) (Miller and Zimmerman, 2010), pH was
Experiments conducted in sunlight confirm that sunlight can
investigated as a potentially confounding variable. When
also induce TICB’s photooxidative process. As(III) samples
controlling for the pH fluctuations caused by irradiation,
exposed to sunlight showed 100% conversion to As(V) within 8
enhanced arsenic removal for samples exposed to UV light is
days, whereas samples not exposed to sunlight showed <16%
still observed. As shown in Table 1, more As(V) is removed
conversion to As(V) over the course of 12 days (SI.3). As
when UV is present than when UV is absent, where final pH is
<7. Based on a model developed to predict TICB performance
in the presence of UV light (Section 3.3), the 7% difference in
final pH would account for a 15% difference in final removal
capacities, far less than the actual 28% difference observed
between 1054 mg/g and 1397 mg/g, suggesting that UV light
does in fact enhance sorption independent of its impact on
solution pH.
UV irradiation can result in scission of linkages in the
chitosan backbone, producing carboxyl groups that do not
significantly affect the chemical structure of chitosan (Zubieta
et al., 2008). One likely oxidation product of chitosan, where
a carboxy group is formed in the C6 position, is shown in
Fig. 5a (Ahmed et al., 2003). The addition of new carboxyl
groups may provide additional coordinating sites for arsenic
oxyanions. Previous studies have reported that a bidentate,

0.8 100
BET surface area (m2/g)

80
0.6
% Porosity

60
0.4
40

0.2
20

0 0
0 100 200 300 Fig. 5 e Molecular structures relevant to the TICB-As
h UV irradiation pretreatment system. a) Potential scheme for oxidation of chitosan
(Ahmed et al., 2003). b) Bidentate binuclear complexation of
Surface area Porosity
arsenate by TiO2 (Jegadeesan et al., 2010; Jing et al., 2005;
Fig. 4 e Relationship between UV irradiation duration and Jing et al., 2009; Pena et al., 2006). c) Potential bidentate
BET surface area and porosity, where all beads are binuclear complexation of arsenate by TiO2 and oxidation
w875 mm and 30% TiO2 by weight. product of chitosan.
5750 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4
decentralized treatment is the intended application, this is
a significant result because sunlight is a free, non-resource
intensive, and readily available source of UV light.
3.1.5. Groundwater
Adsorbent behavior in a deionized water matrix may not
be predictive of behavior in a natural water matrix, where
background ions and pH can control adsorption processes
(Schwartzenbach et al., 2003). For arsenic chemisorption
by TiO2, in particular, researchers have reported that phos-
phate and silicate can act as competitive background ions
(Jing et al., 2009; Mohan and Pittman Jr., 2007; Viraraghavan
and Chowdhury, 2007). In a synthetic groundwater matrix,
designed to model water from Bangladesh (Leupin and Hug,
2005), removal of As, Ca, Mg, P, and Si with TICB was tested
(Fig. 6). These batch TICB adsorption tests were conducted in
the presence and absence of UV light. In all cases, concentra-
tions of Ca, Mg, and Si remained constant, and the concen-
tration of As and P decreased. In the UV irradiated system,
percentages of arsenic and of phosphate removed, with a range
of TICB dosing, were nearly identical, suggesting that phos-
phate and arsenate are direct competitors for sorption sites.
TICB performance was also tested in a natural arsenic-laden
groundwater, collected from a tubewell in West Bengal, India.
As with the synthetic groundwater, TICB removed nearly
identical percentages of arsenic and phosphate in natural
groundwater (SI.4). This suggests the TICB intended for use in
the field will require enough TiO2 (either through increased
mass per bead or an increased number of beads) to account for
removal of both P and As.
3.1.6. Capacity for arsenic desorption and resorption
From ease of use and sustainability perspectives, an adsor-
bent that can be simply regenerated and reused minimizes
resource consumption and offers economic and environ-
mental advantages. TICB was investigated for regeneration
and reuse capacity over several adsorption/desorption cycles.
A variety of solvents in which chitosan is insoluble, including
base and some acids (Pillai et al., 2009), were evaluated for
arsenic desorption from TICB. These regeneration solvents
were evaluated at various concentrations to minimize the
a UV absent
100
As
Ca
80
% analyte removal
Mg
P P
60
Si
40
20
0 0
0 0.02 0.04 0.06
g TICB/ 40 mL groundwater
Fig. 6 e Ion removal in synthetic groundwater matrix by TICB where a) UV is absent, and b) UV is present.
amount of resources required for regeneration and to mini-
mize hazard associated with aqueous arsenic-laden waste
after regeneration.
Of potential desorbing solvents screened, NaOH was the
most effective. Kinetics of As(III)- and As(V)- desorption from
saturated TICB were tested for a range of NaOH concentra-
tions (SI.5). In all cases, >50% of arsenic adsorbed was des-
orbed within 2 h. Of the NaOH concentrations tested (0.07 M,
0.67 M, 1.67 M), 0.07 M NaOH demonstrated the most rapid and
most complete desorption for both As(III) and As(V).
For three cycles of adsorption/desorption, TICB achieved
equilibrium with a non-buffered arsenic solution of either
10,000 mg As(III)/L or 10,000 mg As(V)/L. During desorption, As-
saturated beads were equilibrated with NaOH (0.07 M, 0.67 M,
1.67 M). For all NaOH concentrations, > 60% of the arsenic
adsorbed, whether As(III) or As(V), is desorbed; with 0.07 M
NaOH, >85% of As(III) adsorbed and >78% of As(V) adsorbed
was desorbed each of the three cycles. A representative plot of
these adsorption/desorption cycles is shown in Fig. 7.
In addition to investigating bead reuse and resorption from
10,000 mg/L arsenic solutions, other concentrations of greater
relevance to source waters were tested. Resorption isotherms
for up to three cycles, where initial As concentration ranged
from 100 mg/L to 10,000 mg/L, are shown in Fig. 8. These
isotherms show sustained and possibly improved resorption
performance for As(III) but slightly varied resorption behavior
for As(V), where the final solution pH can account for the
changes in the observed amounts of As(V) resorbed.
3.2. Kinetics of adsorption
Consistent with other literature reports (Bang et al., 2005; Pena
et al., 2005), equilibrium between arsenic and TiO2 nano-
powder in our batch system was reached rapidly, within 2 h
(SI.6). Equilibrium between our “standard” TICB system (w875
um in diameter, 30% TiO2 loading by mass) and arsenic,
however, was not reached for at least seven days. This was the
case for As(III) and As(V), in the absence or presence of UV
light. The adsorption kinetics of TICB systems with varying
bead size and UV exposure (identical to those reported in
Section 3.1.2 (Fig. 2) and Section 3.1.4 (Fig. 4), respectively)
b UV present
100
As
Ca
80
% analyte removal
Mg
60
Si
40
20
0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
g TICB/ 40 mL groundwater
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4 5751

100

80

% As(III) Removal
60

40

20

0
0 100 200 300 400
Time (h)
TICB ∅ ~875 μm
TICB ∅ ~745 μm
TiO2∅ ~0.025 μm
Fig. 7 e Mass of As(III) sorbed onto TICB from a 10,000 mg/L
Fig. 9 e % As(III) removal in the absence of UV light for neat
stock solution and subsequently desorbed from TICB with
TiO2 nanopowder, 744 um TICB and 875 mm TICB, where
0.07 M NaOH for three cycles.
the mass of TiO2 per sample is equal and [As]0 [ 500 mg/L.

were evaluated to identify potential strategies to minimize

this kinetic limitation which would likely be unacceptable in
a field situation. These experiments were conducted without
buffering to assess the kinetics of the system without inter-
ference from buffering ions. However, because pH can influ-
ence the rate of arsenic adsorption by TiO2 (Dutta et al., 2004),
solution pH was closely monitored and was found to remain
relatively constant throughout kinetic experiments.
3.2.1. TICB size
The relationship between bead size and rate of removal was
investigated. Kinetics of arsenic adsorption by TICB with
diameter of 875 um and 744 um as well as neat TiO2 nano-
powder are shown in Fig. 9. In these experiments, the mass of
TiO2 in the system is constant for the two TICB sizes and the
neat powder. Results indicate that a slight reduction (<15%) in
the size of standard TICB can significantly increase the rate of
arsenic removal, so much so that these slightly smaller TICB
directly mimic neat TiO2 powder.
Water and free molecules can travel within chitosan,
which forms hydrogels in aqueous solution (Berger et al.,
2004a, 2004b). Hydration of a hydrogel requires diffusion of
water molecules into the polymer network, relaxation of

a As(III) b As(V)
3000
3000
µg As resorbed/g TICB

2500
µg As resorbed/g TICB

2500

2000 2000

1500 1500

1000 1000

500 500

0 0
0 5000 10000 0 2000 4000 6000 8000 10000

[As]e (µg/L) [As]e (µg/L)

Sorption cycle 1 Sorption cycle 1
Sorption cycle 2 Sorption cycle 2
Sorption cycle 3 Sorption cycle 3

Fig. 8 e Resorption isotherms over three adsorption/desorption cycles with 0.07 M NaOH where for a) As(III), where final
pH [ 9.27, 8.85, 8.88 for sorption cycles 1, 2, and 3 respectively, and b) As(V), where final pH [ 8.74, 7.46, 7.72 for sorption
cycles 1, 2, and 3, respectively.
5752 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4

Table 2 e Kinetic calculations for TICB at 25  C, Orpm, model 10% TiO2

where [As]0 [ 10,000 mg/L and bead diameter w875 mm. 1600 model 30% TiO2
qe, exp (mg/g) Pseudo 1st 10% TiO2; R2=0.87

µ g As sorbed / g TICB
30% TiO2; R2=0.94
qe, calc k R2 1200
(mg/g) (1/hour)

As(III) removal UV absent 2861.6 2892.9 9.30  103 0.904

UV present 4708.4 3102.6 1.39  102 0.868 800
As(V) removal UV absent 1940.8 2540.2 7.54  103 0.886
UV present 3807.9 3827.6 1.56  102 0.946
400

polymer chains, and expansion of the polymer network into

the surrounding bulk water medium (Okano, 1998). Enhanced
kinetics observed for smaller beads reflects the fact that less
diffusion is required for dissolved arsenic oxyanions to reach
potential TiO2 binding sites when bead diameter is reduced.
3.2.2. UV light
Several kinetic models, including power function, pseudo first
order and pseudo second order, were tested to fit experi-
mental data for arsenic sorption by TICB, models previously
reported for similar metal-sorbent systems with photooxida-
tion (Jing et al., 2009; Ofomaja et al., 2010). For all systems
examined, that is regardless of arsenic species and the pres-
ence of UV light, the pseudo first order model best predicted
the equilibrium sorption capacity, and best fit parameters for
the system at 25  C are shown in Table 2. Not only does UV
irradiation result in higher equilibrium capacities, but UV
irradiation also results in higher rates of arsenic removal. The
increased rate of arsenite and arsenate removal in the pres-
ence of UV light can be attributed to the availability of more
arsenic binding sites as described above.
Identical kinetic testing was conducted outdoors with
fluctuating temperature, both with and without exposure to
sunlight; a pseudo first order model also best predicts equi-
librium sorption capacity for these data, and sunlight resul-
ted in higher removal capacities as well as rates of removal
(SI.7).
3.3. Toward a predictive TICB sorption capacity model
Data presented in Section 3.1 indicate that arsenic removal in
a given TICB system is directed by the solution pH and by the
TiO2 content in the bead and is enhanced in the presence of
0
4 6 8 10 12
pH
Fig. 10 e Arsenic removal (mg As/g TICB) across pH for
beads of varying TiO2 loading. Experimental and model
data are compared for beads loaded with 10% TiO2 by
weight and 30% TiO2 by weight.
UV light. A semi-empirical model, based on the logistic func-
tion, was developed to predict TICB sorption capacity from
solution pH and TiO2 content in the bead. Based on our
mechanistic understanding that TICB-As complexes form
through hydroxide exchange and can only form at pH values
where electrostatic repulsion does not occur, a logistic func-
tion was chosen. The logistic function depicts the sigmoidal
shape of the arsenic removal data across pH, analogous to the
shape of a titration curve across pH. Fitting parameters were
derived from batch equilibrium experiments between arsenic
([As]0 ¼ 1000 mg/L) and TICB (30% TiO2 by weight, 875 mm in
diameter), across the pH range 4e11. This model, shown in
Table 3, was optimized for three experimental conditions:
As(III), As(V), and AsUV(III) þ AsUV(V).
The different adsorption capacities for the experimental
conditions are accounted for in parameter a. Different inflec-
tion points observed for As(III) and As(V) across pH (Fig. 1) are
a result of repulsive charges between negatively charged TICB
and negative arsenic oxyanions. These inflection points are
determined by the relevant pKas for As(III), As(V), and
AsUV(III) þ AsUV(V), where all arsenic exists as As(V). Experi-
mental data is compared to this model for the experimental
condition AsUV(III) þ AsUV(V) in Fig. 10; R2 values for TiO2
compositions of 10% and 30% by weight are 0.87 and 0.94,
respectively.

Table 3 e Semi-empirical model to predict TICB

performance for pH range 4e11 and % TiO2 (by mass)
range 0e46. Results apply to reference experiment, where 4. Conclusions
equilibrium is reached after 220 h between 25 mg TICB
and 40 mL 1000 mg As/L solution. The optimization of TiO2-impregnated chitosan bead, TICB, as
a an arsenic sorbent was performed. As TiO2 loading increases,
mg As=g TICB ¼
1 þ ð20:8=%TiO2 Þ  e^ðpH  pKa Þ TICB surface area increases, and arsenic removal capacity of
Experimental conditions a pka TICB loaded with a given amount of TiO2 is similar to the
arsenic removal capacity of the neat TiO2 nanopowder
As(III) 599 9.2
equivalence. pH is a critical variable in the TICB sorption
As(V) 880 6.98
system, where As(III) removal is greatest below 9.2 and As(V)
AsUV(III) þ AsUV(V) 1658 6.98
removal is greatest below 7.25. Sorption capacity by TICB is
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4
determined to be a function of pH and % TiO2 loading, and
a model was developed to predict arsenic sorption, for given
experimental conditions, based on these two variables. Across
pH 4e11 in the presence of UV light, this model predicts
adsorption capacity with R2 values of 0.87 and 0.94 for 10%
TiO2 beads and 30% TiO2 beads, respectively.
Reduction in bead diameter does not influence sorption
capacity, given sufficient equilibrium time. That is, the largest
TICB (800 mme940 mm) results in arsenic removal that is
similar to that of an equivalent dose of neat TiO2 nanopowder.
However, bead diameter reductions increase bead surface
area and increase rate of arsenic removal. Exposure to UV light
also increases bead surface area and the rate of arsenic
removal but does not affect the porosity of the bead. Long
detention times are not practical for point-of-use systems,
and future work will aim to minimize detention times through
bead diameter reductions and incorporation of UV irradiation.
Laboratory tests were performed to predict performance of
TICB in a field setting. In synthetic and natural groundwater,
phosphate is found to be a direct competitor for arsenic
sorption sites on TICB. Other ions tested, including Ca, Mg,
and Si, do not compete for sorption sites on TICB. In the
presence of sunlight and TICB, dissolved As(III) is oxidized to
As(V) within 8 days. Finally, TICB can be regenerated with
weak NaOH, and TICB can be reused without losing effec-
tiveness for at least three cycles in batch experiments.
Acknowledgments
We are grateful to Jamila Yamani for her laboratory assistance
and for her support of this project. Funding was provided by
the National Science Foundation Graduate Research Fellow-
ship and National Science Foundation Environmental
Sustainability (CBET-0932060).
Appendix. Supplementary material
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.040.
references
Agros, M., Kalra, T., Rathouz, P.J., Chen, Y., Pierce, B., Farvez, F.,
Islam, T., Ahmed, A., Rakibuz-Zaman, R., Hasan, R., Sarwar, G.,
Slavkovich, V., van Geen, A., Graziano, J., Ahsan, H., 2010.
Arsenic exposure from drinking water, and all-cause and
chronic-disease mortalities in Bangladesh (HEALS):
a prospective cohort study. The Lancet 376 (9737), 252e258.
Ahmed, G.A.-W., Khairou, K.S., Hassan, R.M., 2003. Kinetics and
mechanism of oxidation of chitosan polysaccharide by
permanganate ion in aqueous perchlorate solutions. Journal
of Chemical Research S, 182e183.
Bang, S., Patel, M., Lippincott, L., Meng, X., 2005. Removal of
arsenic from groundwater by granular titanium dioxide
adsorbent. Chemosphere 60, 389e397.
Berger, J., Reist, M., Mayer, J.M., Felt, O., Gurny, R., 2004a.
Structure and interactions in chitosan hydrogels formed by
5753
complexation or aggregation for biomedical applications.
European Journal of Pharmaceutics and Biopharmaceutics 57
(1), 35e52.
Berger, J., Reist, M., Mayer, J.M., Felt, O., Peppas, N.A., Gurny, R.,
2004b. Structure and interactions in covalently and ionically
crosslinked chitosan hydrogels for biomedical applications.
European Journal of Pharmaceutics and Biopharmaceutics 57
(1), 19e34.
Bhattacharya, P., Mukherjee, A.B., Bundschuh, J., Zevenhoven, R.,
Loeppert, R.H. (Eds.), 2007. Arsenic in Soil and Groundwater
Environment. Elsevier, Amsterdam.
Deebar, N., Irfan, A., Ishtiaq, Q.A., 2009. Evaluation of the
adsorption potential of titanium dioxide nanoparticles for
arsenic removal. Journal of Environmental Sciences 21,
402e408.
Dutta, P.K., Ray, A.K., Sharma, V.K., Millero, F.J., 2004. Adsorption
of arsenate and arsenite on titanium dioxide suspensions.
Journal of Colloid and Interface Science 278, 270e275.
Ferguson, M.A., Hoffmann, M.R., Hering, J.G., 2005. TiO2-
photocatalyzed As(III) oxidation in aqueous suspensions:
reaction kinetics and effects of adsorption. Environmental
Science & Technology 39, 1880e1886.
Gerente, C., Lee, V.K.C., Le Cloirec, P., McKay, G., 2007. Application
of chitosan for the removal of metals from wastewaters by
adsorption - mechanisms and models review. Critical Reviews
in Environmental Science and Technology 37 (1), 41e127.
Jegadeesan, G., Al-Abed, S.R., Sundaram, V., Choi, H., Scheckel, K.G.
, Dionysiou, D.D., 2010. Arsenic sorption on TiO2 nanoparticles:
size and crystallanity effects. Water Research 44, 965e973.
Jing, C., Liu, S., Patel, M., Meng, X., 2005. Arsenic leachability in
water treatment adsorbents. Environmental Science &
Technology 39, 5481e5487.
Jing, C., Meng, X., Calvache, E., Jiang, G., 2009. Remediation of
organic and inorganic arsenic contaminated groundwater
using a nanocrystalline TiO2-based adsorbent. Environmental
Pollution 157, 2514e2519.
Leupin, O.X., Hug, S.J., 2005. Oxidation and removal of arsenic (III)
from aerated groundwater by filtration through sand and
zero-valent iron. Water Research 39, 1729e1740.
Miller, S.M., Zimmerman, J.B., 2010. Novel, bio-based, photoactive
arsenic sorbent: TiO2-impregnated chitosan bead. Water
Research 44, 5722e5729.
Mohan, D., Pittman Jr., C.U., 2007. Arsenic removal from water/
wastewater using adsorbents - A critical review. Journal of
Hazardous Materials 142, 1e53.
Muzzarelli, R.A.A., Muzzarelli, C., 2005. Polysaccharides 1:
Structure, Characterization and Use, pp. 151e209.
Neubauer, K.R., Reuter, W., Perrone, P., Grosser, Z., 2004. In:
Services, P.L.a.A. (Ed.), Simultaneous Arsenic and Chromium
Speciation by HPLC/ICP-MS in Environmental Waters.
PerkinElmer, Shelton, CT.
Ofomaja, A.E., Baidoo, E.B., Modise, S.J., 2010. Kinetic and pseudo-
second-order modeling of lead biosorption onto pine cone
powder. Industrial & Engineering Chemistry Research 49,
2562e2572.
Okano, T. (Ed.), 1998. Biorelated Polymers and Gels. Academic
Press, San Diego.
Pena, M., Meng, X., Korfiatis, G.P., Jing, C., 2006. Adsorption
mechanism of arsenic on nanocrystalline titanium dioxide.
Environmental Science & Technology 40, 1257e1262.
Pena, M.E., Korfiatis, G.P., Patel, M., Lippincott, L., Meng, X., 2005.
Adsorption of As(V) and As(III) by nanocrystalline titanium
dioxide. Water Research 39, 2327e2337.
Petrusevski, B., Sharma, S., van der Meer, W.G., Kruis, F.,
Khan, M., Barua, M., Schippers, J.C., 2008. Four years of
development and field-testing of IHE arsenic removal family
filter in rural Bangladesh. Water Science and Technology 58
(1), 53e58.
5754 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 4 5 e5 7 5 4
Pillai, C.K.S., Paul, W., Sharma, C.P., 2009. Chitin and chitosan
polymers: Chemistry, solubility, and fiber formation. Progress
in Polymer Science 34, 641e678.
Ratnaike, R.N., 2003. Acute and chronic arsenic toxicity.
Postgraduate Medical Journal 79, 391e396.
Schwartzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003.
Environmental Organic Chemistry. John Wiley and Sons, Inc.,
Hoboken, New Jersey.
Tien, V.N., Chaudhary, D.S., Ngo, H.H., Vigneswaran, S., 2004.
Arsenic in water: concerns and treatment technologies. Journal
of Industrial and Engineering Chemistry 10 (3), 337e348.
Viraraghavan, T., Chowdhury, P., 2007. Recent trends in arsenic
removal technologies. Progress in Environmental Science and
Technology, 692e694.
Webb, P.A., 2001. An Introduction to the Physical Characterization
of Materials by Mercury Intrusion Porosimetry with Emphasis
on Reduction and Presentation of Experimental Data Norcross,
Georgia.
Zubieta, C.E., Messina, P.V., Luengo, C., Dennehy, M., Pieroni, O.,
Schulz, P.C., 2008. Reactive dyes remotion by porous TiO2-
chitosan materials. Journal of Hazardous Materials 152,
765e777.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

One year environmental surveillance of rotavirus specie

A (RVA) genotypes in circulation after the introduction
of the Rotarix
vaccine in Rio de Janeiro, Brazil

Tulio Machado Fumian a,*, José Paulo Gagliardi Leite a, Tatiana Lundgreen Rose a,
Tatiana Prado b, Marize Pereira Miagostovich a
a
Laboratory of Comparative and Environmental Virology, Oswaldo Cruz Institute, Oswaldo Cruz Foundation (Fiocruz), Av. Brasil 4.365,
Manguinhos, CEP 21040-360, Rio de Janeiro (RJ), Brazil
b
Laboratory of Technological Development in Virology, Oswaldo Cruz Institute, Oswaldo Cruz Foundation (Fiocruz), Av. Brasil 4.365,
Manguinhos, CEP 21040-360, Rio de Janeiro (RJ), Brazil

article info abstract

Article history: Rotavirus specie A (RVA) infection is the leading cause of severe acute diarrhea among
Received 26 May 2011 young children worldwide. To reduce this major RVA health impact, the Rotarix
vaccine
Received in revised form (GlaxoSmithKline, Rixensart, Belgium) was introduced in the Brazilian Expanded Immu-
24 August 2011 nization Program in March 2006 and became available to the entire birth cohort. The aim of
Accepted 25 August 2011 this study was to evaluate the spread of RVA in the environment after the introduction of
Available online 1 September 2011 Rotarix
in Brazil. For this purpose, a Wastewater Treatment Plant (WTP) in Rio de Janeiro
was monitored for one year to detect, characterize and discriminate RVA genotypes and
Keywords: identify possible circulation of vaccine strains. Using TaqMan
quantitative PCR (qPCR),
Rotavirus A genotypes RVA was detected in 100% (mean viral loads from 2.40
105 to 1.16
107 genome copies
Rotarix
vaccine (GC)/L) of sewage influent samples and 71% (mean viral loads from 1.35
103 to
Wastewater 1.64
105 GC/L) of sewage effluent samples. The most prevalent RVA genotypes were P[4],
Wastewater treatment plant P[6] and G2, based on VP4 and VP7 classification. Direct nucleotide sequencing (NSP4
fragment) and restriction enzyme digestion (NSP3) analysis did not detect RVA vaccine-like
strains from the sewage samples. These data on RVA detection, quantification and
molecular characterization highlight the importance of environmental monitoring as a tool
to study RVA epidemiology in the surrounding human population and may be useful on
ongoing vaccine monitoring programs, since sewage may be a good screening option for
a rapid and economical overview of the circulating genotypes.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction developed countries, this virus remains a common cause of

Rotavirus specie A (RVA) is the main etiological agent of viral
gastroenteritis in infants throughout the world and is associ-
ated with significant mortality in developing countries, where
over 600,000 deaths occur annually (Parashar et al., 2006). In
morbidity with significant economic burden (Charles et al.,
2006; Parashar et al., 2006).
RVA belongs to the Reoviridae family, Rotavirus genus, and
possesses a double-stranded RNA (dsRNA) genome with 11
segments that encode six structural (VP) and six non-

* Corresponding author. Tel.: þ55 21 25621875; fax: 55 21 25621851.

E-mail address: tuliomf@ioc.fiocruz.br (T.M. Fumian).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.039
5756 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
structural proteins (NSP) (Estes and Kapikian, 2007). A widely
used binary classification scheme has been established based
on the two genes that codify the outer capsid proteins, VP4
and VP7, defining G (from VP7, glycoprotein) and P (from VP4,
protease-cleaved protein) genotypes. Currently 27 G and 35 P
genotypes are recognized (Abe et al., 2009; Solberg et al., 2009;
Ursu et al., 2009; Matthijnssens et al., 2011); however, only five
RVA G genotypes (G1eG4 and G9) and two P genotypes (P[8]
and P[4]) are prevalent worldwide (Santos and Hoshino, 2005;
Ursu et al., 2009).
RVA virions are shed in extremely high concentrations (up
to 1010 virus/g) in the stool of infected children with acute
gastroenteritis and can persist in the environment for long
periods of time (Carter, 2005; Bosch et al., 2008). The features
of the virions, including stability in aqueous environments
and resistance to water treatment, may facilitate their trans-
mission to humans via contaminated water (Ansari et al.,
1991; Espinosa et al., 2008). Direct sewage discharge into
environmental waters such as lagoons, rivers, beaches and
coastal waters represents a public health problem mainly in
developing countries. These contaminated waters have been
broadly linked to the causation of several waterborne gastro-
enteritis outbreaks (Kukkula et al., 1997; Villena et al., 2003a;
Schmid et al., 2005; Godoy et al., 2006). Despite the difficulty
of determining the proportion of gastroenteritis cases due to
contaminated water, it has been suggested that a significant
percentage of the cases are related to the quality of the water
(Bosch et al., 2008).
As RVA is one of the most important causes of mortality in
infants worldwide, two equally safe and efficacious live oral
rotavirus vaccines, G1P[8] RVA vaccine (RV1 e Rotarix
,
GlaxoSmithKline, Rixensart, Belgium) and a pentavalent G1-
G4 and P[8] RVA vaccine (RV5 e RotaTeq
, Merck and Co.,
Whitehouse Station, NJ, USA), were developed and are
licensed for use in more than 100 countries worldwide (Jiang
et al., 2010). The first one, RV1, was included in the Brazilian
Expanded Immunization Program (PNI) in March 2006 and
became available to the entire birth cohort.
The impact of this vaccine on the circulating RVA geno-
types is unknown and difficult to predict, so continuous
genotype surveillance is needed to identify the effects of the
vaccine program on circulating strains, particularly on geno-
type prevalence and the emergence of uncommon strains.
The monitoring of the viruses circulating in sewage from
a wastewater treatment plant (WTP) has been described as an
appropriate model to understand the spread of RVA in the
population served by the WTP, as influents may contain
viruses shed from patients with sporadic or asymptomatic
cases (Haramoto et al., 2006; Bosch et al., 2008).
The main goal of this study was to evaluate the spread of
RVA in the environment following the introduction of the
Rotarix
vaccine in Brazil. For this purpose, a WTP located in
Rio de Janeiro was monitored for one year to detect, quantify
and characterize RVA genotypes and to investigate the
possible presence of the vaccine strain in sewage samples.
RVA genomes were investigated in samples collected from
raw and treated sewage using Taqman
quantitative PCR
(qPCR), and the P (VP4) and G (VP7) genotypes were charac-
terized by nested PCR in a multiplex reaction (Gentsch et al.,
1992; Gouvea et al., 1994). Protocols based on direct
nucleotide sequencing (NSP4) and restriction enzyme diges-
tion (NSP3) analysis (Rose et al., 2010) were applied to
discriminate between wild-type and vaccine strains.
2. Materials and methods
2.1. Wastewater treatment plant (WTP) sample
collection
Sewage samples were collected from an urban WTP located in
the metropolitan area of Rio de Janeiro, Brazil. The WTP
receives sewage from around 1.5 million inhabitants leaving
in both the central and north zone of the city and is one of the
largest in Brazil. Sewage treatment employs a secondary
treatment (aerobic process: activated sludge) with an inflow
mean of 2500 L s1. Initial sewage treatment is composed of
grid separation and primary sedimentation (five primary
settling tanks with a volume of 7700 m3 each). There are four
aeration tanks in parallel (volume: 11,500 m3 per tank) with
a capacity to treat 625 L s1 of effluent. Secondary sedimen-
tation is performed in four secondary settling tanks (volume:
8800 m3 per tank) with no chlorination before effluents are
discharged into the water environment.
A total of 48 sewage samples were collected bi-monthly (15
day interval) from August 2009 to July 2010, 24 of them were
collected from raw sewage (influent) and 24 from the final
treated sewage (effluent). At each sampling point, 50 ml of
sewage was collected in sterile plastic bottles, kept at 4  C and
transported to the laboratory for immediate analysis.
2.2. Virus concentration
Viruses were concentrated using the ultracentrifugation
method as described by Pina et al. (1998). To avoid false nega-
tive results and to evaluate the presence of inhibitors, sewage
samples were inoculated with 500 ml of an internal control
(bacteriophage PP7) before the concentration assay, and the
extracted RNA was diluted 10-fold (Fumian et al., 2010).
2.3. Nucleic acid extraction, reverse transcription (RT)
and quantitative PCR (qPCR)
The viral dsRNA was extracted by the glass powder method
(Boom et al., 1990), and the synthesis of cDNA was carried out
by reverse transcription using a random primer (PdN6 e 50
A260 units e Amersham Biosciences, Chalfont St Giles,
Buckinghamshire, UK). Multiplex qPCR to detect RVA and PP7
was performed as described previously (Fumian et al., 2010)
using primers described by Zeng et al. (2008) and Rajal et al.
(2007). RVA primers and probe were designed to target
a highly conserved region of the non-structural protein 3
(NSP3), and the PP7 primers to amplify a region of the PP7
replicase gene. Both were synthesized by Applied Biosystems
(CA, USA). qPCR was carried out using an ABI PRISM 7500
Sequence Detection System (Applied Biosystems, CA, USA).
A standard curve (SC; 107, 105, 103 and 101 copies per reaction)
was generated using 10-fold serial dilutions of a pCR2.1 vector
(Invitrogen, USA) containing either the RVA NSP3 gene or the
PP7 replicase gene.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
2.4. VP4 and VP7 nested PCR amplification
Nested PCR was used for molecular characterization of RVA
genotypes G and P, and it partially amplified VP7 and VP4
segments, respectively. In the first-round, RT-PCR was per-
formed with VP7 and VP4 consensus primers 9con1e9con2 (Das
et al., 1994) and 4con2e4con3 (Gentsch et al., 1992), respec-
tively. Following the first-round, RVA G genotype classification
was performed using specific primers for genotypes G1eG4, G5
and G9 (Das et al., 1994; Gouvea et al., 1994), and P genotype
classification was carried out using primers for genotypes P[4],
P[6] and P[8]-P[10], described by Gentsch et al. (1992).
2.5. RVA molecular characterization
To discriminate RVA wild-type G1P[8] from the vaccine strain,
the NSP4 gene was amplified according to the protocol
described by Cunliffe et al. (1998). Two nucleotide mutations
after the first initiation ATG codon at positions 100 and 134
were observed when the NSP4 nucleotide sequence of the
Rotarix
vaccine (patent number: PCT/EP2004/009725) was
compared to sequences available in GenBank including
reference strains. These two nucleotide shifts were used to
classify RVA (data not shown).
The NSP4 PCR amplicons were purified and sequenced
using an ABI Prism BigDye Terminator Cycle Sequencing
Ready Reaction Kit
and an ABI Prism 3730 Genetic Analyzer
(Applied Biosystems, Foster City, CA, USA). The chromato-
grams were analyzed using BioEdit
(Hall, 1999). A phyloge-
netic dendrogram was constructed by the neighbor-joining
method using a matrix of genetic distances established under
the Kimura-two parameter model (Felsenstein, 1993) using
MEGA V. 4.0 (Tamura et al., 2007). The robustness of each node
was assessed by bootstrap analysis using 2000 pseudo-
replicates. RVA NSP4 isolated from sewage samples was
classified according to the most recent full genome-based
classification proposed by Matthijnssens et al. (2011).
2.6. Cloning and restriction endonuclease analysis
To characterize vaccine strains, another protocol based on
BspHI restriction endonuclease analysis of the NSP3 gene was
Fig. 1 e Monthly distribution of rotavirus specie A (RVA) in influent (A) and effluent (B) samples from a WTP in Brazil. The
plots show the geometric monthly mean values (GC/L). The upper and lower bars show the standard deviations of the mean
values.
5757
performed as previously described (Rose et al., 2010). Prior to
endonuclease restriction analysis, PCR amplicons generated
by NSP3 amplification (Matthijnssens et al., 2006) were cloned
into the PCR4-TOPO vector (Invitrogen, USA) following the
manufacturer’s recommendations.
2.7. Statistical analysis
The total frequency of detection obtained in WTP, in both
influent and effluent samples, using qPCR assay was
compared by using a chi-square test and Fisher’s exact test at
a significance level of 0.05. The same statistical analysis was
performed to determine significant differences between VP4
and VP7 PCR detection in all of the 48 samples collected.
Analysis of Variance (ANOVA) was performed to determine
differences in mean levels of RVA, present in influent
samples, during the four seasons (summer, fall, spring and
winter), and a paired t-test was performed to verify differ-
ences between the mean levels of RVA in influent and effluent
samples throughout the study. Statistical analyses were per-
formed using GraphPad Prism
software version 5.
3. Results
3.1. Rotavirus A detection and quantification
The RVA genome levels and genotypes were determined in
a one-year monitoring study from influent and effluent
streams at a WTP located in Rio de Janeiro city, Brazil. Using
qPCR, 41 out of 48 (85%) of the samples were positive, corre-
sponding 100% (24/24) of influent and 71% (17/24) of effluent
samples. The difference in the total frequencies of RVA
detection (qPCR) in WTP was significant between influent and
effluent samples ( p ¼ 0.0042, Chi-square; p ¼ 0.0047, Fisher).
Fig. 1 shows the monthly distribution of RVA genome copies
(GC/L) and the standard deviation. For sewage influent, RVA
concentrations ranged from 2.40
104 to 1.16
107 GC/L, and
in effluent, positive sample concentrations, ranged from
1.35
103 to 1.64
105 GC/L. The differences in mean levels of
RVA present in influent samples throughout the seasons
were not significant (ANOVA/ NewmaneKeuls Multiple
5758 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3

Comparison Test), on the other hand, difference in the mean in GenBank. The phylogenetic analysis of this segment clas-
levels of RVA was significant between influent and effluent sified the samples within two distinct genotypes: E1 genotype,
samples ( p ¼ 0.041, Paired t-test). with a single sequence (RJ-VA-550) and E2 genotype, in which
Table 1 summarizes the results obtained from the genomic 10 sequences clustered. Among the sequences that clustered
amplification protocols used for detection, quantification in genotype E2, the sequence obtained from RJ-VA-575 sample
(qPCR) and molecular characterization of RVA genes (RT-PCR clustered in a separate branch of the tree (Fig. 2). None of the
for NSP4, VP4 and VP7). The RVA genotype G2 was detected in 11 NSP4 sequences showed the two nucleotide mutations as
100% (24/24) of influent samples, and the genotypes P[4] and P in the vaccine pattern.
[6] were detected in 33% (8/24) and 25% (6/24), respectively. RVA NSP4 nucleotide sequences obtained in the present
Effluent samples showed a lower RVA detection rate, with study were deposited at the National Center for Biotechnology
genotypes G2 and P[4] being detected in 25% (6/24) and 4% (1/ Information (GenBank, http://www.ncbi.nlm.nih.gov/) under
24) of samples, respectively. Genotypes G1 and P[8] were not the accession numbers: RVA/Env-wt/BRA/RJ-VA-500/2009/
identified in the samples tested. The difference in the GXP[X]: JF731369; RVA/Env-wt/BRA/RJ-VA-515/2009/GXP[X]:
frequency of RVA detection using VP4 and VP7 PCR was JF731370; RVA/Env-wt/BRA/RJ-VA-518/2009/GXP[X]: JF731371;
significant ( p ¼ 0.002, Chi-square; p ¼ 0.004, Fisher). RVA/Env-wt/BRA/RJ-VA-521/2009/GXP[X]: JF731372; RVA/Env-
RT-PCR based on the NSP4 gene was able to detect viruses wt/BRA/RJ-VA-523/2009/GXP[X]: JF731373; RVA/Env-wt/BRA/
in 92% (22/24) and 21% (5/24) of influent and effluent samples, RJ-VA-524/2009/GXP[X]: JF731374; RVA/Env-wt/BRA/RJ-VA-
respectively. No evidence of inhibitors was observed as 526/2009/GXP[X]: JF731375; RVA/Env-wt/BRA/RJ-VA-527/2009/
bacteriophage PP7, inoculated as an internal control in all 48 GXP[X]: JF731376; RVA/Env-wt/BRA/RJ-VA-547/2009/GXP[X]:
sewage samples, was detected in 100% of samples tested using JF731377; RVA/Env-wt/BRA/RJ-VA-550/2009/GXP[X]: JF731378;
a multiplex qPCR. PP7 viral titers recovered ranged from and RVA/Env-wt/BRA/RJ-VA-575/2010/GXP[X]: JF731379.
3.4
105 to 1.6
104 GC per 500 ml of PP7 suspension. In order to accurately determine the presence of the
vaccine components in the environment, six influent samples
with high viral loads (1.7
107e3.7
105 GC/L) were subjected
3.2. Rotavirus A strain characterization
to PCR amplification of the region of the genome encoding
NSP3, and the resulting products were cloned. Forty-seven
The sequence of gene segment 10 (encoding NSP4) of Brazilian
colonies were screened for the appropriate banding pattern
waste samples were compared with NSP4 segments available
after BspHI restriction endonuclease analysis, and none of
them demonstrated the vaccine pattern. A Rotarix
NSP3
amplicon was analyzed in the same reaction as a positive
control.
Table 1 e Rotavirus specie A (RVA) detection from influent
(24) and effluent (24) sewage samples by quantitative
(qPCR) and qualitative (NSP4, VP4, VP7) PCR protocols for
genotyping. 4. Discussion
Year Month Sewage influent Sewage effluent
In this study, an environmental approach was used to eval-
qPCR NSP4 VP4 VP7 qPCR NSP4 VP4 VP7
uate the circulation of RVA genotypes in the city of Rio de
2009 Aug þ   G2     Janeiro, Brazil, which has the second largest population in the
þ þ P[4] G2 þ    country. Samples from a large WTP were analyzed using
Sep þ þ P[4] G2 þ    a concentration method (ultracentrifugation) and molecular
þ þ  G2 þ   
techniques to detect, quantify and characterize the detected
Oct þ þ P[6] G2 þ   G2
viruses (Pina et al., 1998; Fumian et al., 2010). This type of
þ þ P[4] G2 þ þ P[4] G2
Nov þ þ P[6] G2 þ þ  G2 approach has been extensively employed to obtain informa-
þ þ P[4] G2 þ    tion on circulating viruses in populations throughout the
Dec þ þ P[4] G2 þ    world, independently of single reported cases or outbreaks,
þ þ P[6] G2     and to assess virus circulation causing asymptomatic infec-
2010 Jan þ þ P[6] G2 þ    tions (Bosch et al., 2008; Gajardo et al., 1995; Haramoto et al.,
þ þ  G2 þ   
2006; Clemente-Casares et al., 2009; Fumian et al., 2010;
Feb þ þ P[6] G2 þ   
Kamel et al., 2010; Prado et al., 2011). Besides revealing the
þ þ  G2    
Mar þ þ  G2     predominant genotypes circulating in Rio de Janeiro, this
þ þ  G2 þ þ  G2 monitoring strategy also aimed to investigate the presence of
Apr þ þ  G2 þ þ  G2 the attenuated G1P[8] RVA vaccine Rotarix
in the
þ þ  G2 þ   G2 environment.
May þ þ P[4] G2     The combination of virus concentration and molecular
þ þ  G2 þ   
detection methods was successfully employed, and the
Jun þ   G2 þ   
þ þ P[4] G2    
results showed a high level of RVA contamination in sewage
Jul þ þ P[6] G2     samples. The high recovery rate of RVA (47%) from sewage
þ þ P[4] G2 þ    samples (Fumian et al., 2010), using the ultracentrifugation
method, was fundamental for the success in RVA recovering.
þ: Positive; and : negative.
Another study using this ultracentrifugation method to
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3 5759

Fig. 2 e Phylogenetic dendrogram based on partial NSP4 nucleotide sequences of rotavirus A strains isolated from sewage
samples in this study. All sequences obtained from GenBank are named according to Matthijnssens et al. (2011), and G and P
genotypes are indicated at the right. The Brazilian environmental samples are marked with a filled diamond (influent
samples) and an unfilled diamond (effluent samples). The scale bar at the bottom of the tree indicates distance. Bootstrap
values (2000 replicates) are shown at the branch nodes and values lower than 50% are not shown.

recover RVA from domestic sewage and polluted water river
samples demonstrated a high percentage of positive samples:
67% and 83% (Rodrı́guez-Dı́az et al., 2009). Lower RVA detec-
tion rates have been observed when membrane-active
charged filtration was used as a concentration method asso-
ciated with organic or inorganic elution (Ferreira et al., 2009;
Kamel et al., 2010).
A pattern of seasonality of RVA-induced gastroenteritis has
been demonstrated in Latin American countries, including
Brazil, based on a higher incidence of infection occurring in
winter months (Kane et al., 2004; Carvalho-Costa et al., 2011).
However, this differential distribution was not observed by
the analysis of sewage samples during the monitoring period,
suggesting a high level of virus shedding occurring throughout
the year.
The average reduction of 2 logarithms in viral load
observed in effluent samples demonstrates that WTPs play an
important role in reducing environmental contamination.
However, as demonstrated in this study, the persistence of
such viruses in treated effluents and in other studies from
different regions, highlights the importance of evaluating the
efficiency of different types of treatments used by WTPs in
viral load reduction (Bofill-Mas et al., 2006; Haramoto et al.,
2006; da Silva et al., 2007; Meleg et al., 2008; La Rosa et al., 2010).
Despite the difficulties in associating virus infection
to contact with contaminated water, the environmental
dissemination of RVA, demonstrated by the high prevalence
and concentration in the treated or untreated sewage samples,
poses a risk to human health that must be considered and
evaluated. Although the detection of nucleic acid does not
5760 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
directly indicate the presence of infectious viruses, it is
strongly suggestive of an infectious particle (Girones et al.,
2010). Different studies have demonstrated that signals
generated after RT-PCR amplification of viral genomes corre-
lated well with infectivity or that a great part of viral nucleic
acid recovered from environmental samples corresponded to
infectious virus particles (Bhattacharya et al., 2004; Espinosa
et al., 2008; Barrella et al., 2009).
The results obtained in this study regarding RVA dissemi-
nation, along with other studies conducted in developing
countries, indicate RVA as a possible viral indicator of human
fecal contamination in environmental samples, at least in
countries where there is a high RVA prevalence (Ferreira et al.,
2009; Miagostovich et al., 2008; Rodrı́guez-Dı́az et al., 2009;
Prado et al., 2011; Sdiri-Loulizi et al., 2010).
Data concerning virus genotyping provide significant
epidemiological information necessary for the introduction
and ongoing monitoring of vaccination programs (Villena
et al., 2003b; Pinto et al., 2007; Bosch et al., 2008). The NSP4
segment analysis showed samples clustered with two geno-
types. E1 sequence was close related to NSP4 from genotypes
G1P[8], G3P[8] and G12P[6] and was probably associated with P
[6] genotype detected. Within E2 genotype, nine samples
formed a monophyletic group, and one sequence (RJ-VA-575)
clustered in another group, including Brazilian G2P[4], isolated
in 2008. The high detection frequency of E2 genotype, with G2
and P[4], is in agreement with trends described by
Matthijnssens et al. (2011), where strains with a G2 and P[4]
genotype presented an E2 profile and strains with a G1, G3,
G12, P[6] and P[8] genotypes demonstrated an E1 profile.
The prevalence of G2 and P[4] genotypes in sewage samples
is in agreement with the results obtained in a previous survey
using clinical samples from acute infantile gastroenteritis
cases in the municipality of Rio de Janeiro after Rotarix

introduction (Carvalho-Costa et al., 2009, 2011). RVA P[6]
genotype detected in a lower prevalence than P[4] in sewage
sample, reflects results obtained from clinical samples,
showing that, in Brazil, the major circulation of G2P[4] and in
a slight ratio, G2P[6] genotype. Data from the Laboratory of
Comparative and Environmental Virology (LVCA), a Brazilian
Regional Reference Laboratory for Rotaviruses, from 2009 to
2010, showed a higher percentage (78%) of RVA G2P[4] circu-
lating in Rio de Janeiro when compared with other genotypes
characterized as G4P[8] (6%); G9P[X] (6%); G2P[6], G2P[X] and
G1P[X] (3%) (data not published). Although an increasing
prevalence of genotype G2P[4] has also been reported in
countries that have not established Rotarix
vaccination
programs (Ferrera et al., 2007; Antunes et al., 2009), it is
important to note that in a smaller WTP sewage monitoring
program also conducted in Rio de Janeiro in 2005, before RVA
vaccine introduction, G1 and P[8] were the most prevalent
RVA genotypes detected (Ferreira et al., 2009). This change in
the RVA genotypes prevalence profile could be explained by
a natural genotypic fluctuation, although the role of the
Rotarix
vaccine introduction cannot be ruled out (Gómez
et al., 2011). In Australia, where both vaccine types are used,
it was observed a higher prevalence of G2P[4] genotypes in
states that used exclusively Rotarix
vaccine when compared
with states that used Rotateq
, showing a higher prevalence
of G3P[8] strains (Kirkwood et al., 2011). Another important
aspect regarding the RVA vaccine is that Rotarix
prevents
around 90% of severe gastroenteritis cases caused by G1P[8],
as well as other partially heterotypic strains; however, it is less
effective (45%) in preventing diarrhea caused by fully hetero-
typic G2P[4] strains (Ruiz-Palacios et al., 2006). Linhares et al.
(2008), when evaluating Rotarix
efficacy against rotavirus
gastroenteritis in a phase III study performed in Latin Amer-
ican infants, demonstrated a vaccine efficacy of 82% and 40%
against G1P[8] and G2P[4], respectively.
Genotypes G1 and P[8] were not found in these sewage
samples, showing that these genotypes are no longer circu-
lating or are circulating at a very low level, reinforcing data
obtained via surveillance of clinical specimens (Carvalho-
Costa et al., 2009, 2011). The high shedding of RVA antigens
(up to 1010 virus/g) from naturally infected individuals may
restrict the detection of the vaccine strain that would be less
prevalent in the environment. In a study conducted in South
Africa during 2003e2004 to evaluate the safety, reac-
togenicity and immunogenicity of the Rotarix
vaccine, virus
shedding was observed in healthy infants, ranging from 31%
to 46% depending on the vaccination regimen used (Steele
et al., 2010).
The cloning of the gene NSP3 followed by restriction
enzyme analysis was an alternative attempt to increase the
probability of vaccine strain detection, as cloning of the PCR
products enables detection of genotypes that are at lower
abundance in the environment. This methodology based on
NSP3 gene amplification followed by BspHI digestion was
previously described for discrimination of the Rotarix

vaccine (Rose et al., 2010).
Epidemiological and laboratory surveillances to assess
vaccine effectiveness and vaccine impact are currently
significant concerns (WHO, 2008). As Brazil was the first Latin
American country to introduce universal rotavirus vaccina-
tion, the evaluation of vaccine performance to examine
possible changing strain patterns of RVA in circulation is
a priority in this country. Sentinel RVA surveillance in selected
pediatric settings has been recommended as part of the
immunization program in Latin America (Carvalho-Costa
et al., 2009). Environmental surveillance, as conducted in
this study by investigating RVA in sewage samples, could be
an alternative approach to support clinical monitoring of RVA
infection. This kind of surveillance would allow continuous
investigation of the genotypes circulating in the WTP service
area, providing an overview of the prevalent genotypes and
possibly discriminating between RVA vaccine and wild
strains.
5. Conclusion
(1) The high circulation of RVA in the population was
measurable by environmental surveillance coupled with
appropriate molecular tools.
(2) Wastewater surveillance demonstrated that genotypes G2
and P[4] were the most prevalent, reflecting a natural
fluctuation of RVA genotypes or a consequence of Rotarix

vaccine introduction or even both.
(3) This is the first study concerning RVA detection and
discrimination between vaccine and wild-type strains
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
from environmental samples carried out in a WTP and
may assist clinical epidemiological studies that will be
essential in the post-vaccination era.
Acknowledgements
This work was financially sponsored by the National Council
for Scientific and Technological Development (CNPq e PRO-
SUL 490292/2008-9; CNPq e PAPES V) and by CGVAM/Ministry
of Health, Brazil. The authors thank the staff of PDTIS DNA
Sequencing Platform at FIOCRUZ (RPT01A) for technical
support in sequencing reactions and the WTP staff for
supplying the samples, under the agreement between Fiocruz
and the Water Company of Rio de Janeiro state (CEDAE). This
research study is under the scope of the activities of Fiocruz as
a collaborating center of PAHO/WHO of Public and Environ-
mental Health.
references
Abe, M., Ito, N., Morikawa, S., Takasu, M., Murase, T.,
Kawashima, T., Kawai, Y., Kohara, J., Sugiyama, M., 2009.
Molecular epidemiology of rotaviruses among healthy calves
in Japan: isolation of a novel bovine rotavirus bearing new P
and G genotypes. Virus Res. 144, 250e257.
Ansari, S.A., Springthorpe, V.S., Sattar, S.A., 1991. Survival and
vehicular spread of human rotaviruses: possible relation to
seasonality of outbreaks. Rev. Infect. Dis. 13, 448e461.
Antunes, H., Afonso, A., Iturriza, M., Martinho, I., Ribeiro, C.,
Rocha, S., Magalhães, C., Carvalho, L., Branca, F., Gray, J., 2009.
G2P[4] the most prevalent rotavirus genotype in 2007 winter
season in an European non-vaccinated population. J. Clin.
Virol. 45, 76e78.
Barrella, K.M., Garrafa, P., Monezi, T.A., Hársi, C.M., Salvi, C.,
Violante, P.A.B.C., Mehnert, D.U., 2009. Longitudinal study on
occurrence of adenoviruses and hepatitis A virus in raw
domestic sewage in the city of Limeira, São Paulo. Braz. J.
Microbiol. 40, 102e107.
Bhattacharya, S.S., Kulka, M., Lampel, K.A., Cebula, T.A.,
Goswami, B.B., 2004. Use of reverse transcription and PCR to
discriminate between infectious and non-infectious hepatitis
A virus. J. Virol. Methods 116, 181e187.
Bofill-Mas, S., Albinana-Gimenez, N., Clemente-Casares, P.,
Hundesa, A., Rodriguez-Manzano, J., Allard, A., Calvo, M.,
Girones, R., 2006. Quantification and stability of human
adenoviruses and polyomavirus JCPyV in wastewater
matrices. Appl. Environ. Microbiol. 72, 7894e7896.
Boom, R., Sol, C.J., Salimans, M.M., Jansen, C.L., Wertheim-van
Dillen, P.M., van der Noordaa, J., 1990. Rapid and simple
method for purification of nucleic acids. J. Clin. Microbiol. 28,
495e503.
Bosch, A., Guix, S., Sano, D., Pintó, R.M., 2008. New tools for the
study and direct surveillance of viral pathogens in water. Curr.
Opin. Biotechnol. 19, 295e301.
Carter, M.J., 2005. Enterically infecting viruses: pathogenicity,
transmission and significance for food and waterborne
infection. J. Appl. Microbiol. 98, 1354e1380.
Carvalho-Costa, F.A., Araújo, I.T., Santos de Assis, R.M., Fialho, A.
M., de Assis Martins, C.M., Bóia, M.N., Leite, J.P., 2009.
Rotavirus genotype distribution after vaccine introduction, Rio
5761
de Janeiro, Brazil. Emerg. Infect. Dis. 15, 95e97.
Carvalho-Costa, F.A., Volotão Ede, M., de Assis, R.M., Fialho, A.M.,
de Andrade Jda, S., Rocha, L.N., Tort, L.F., da Silva, M.F.,
Gómez, M.M., de Souza, P.M., Leite, J.P., 2011. Laboratory-
based rotavirus surveillance during the introduction of
a vaccination program, Brazil, 2005e2009. Pediatr. Infect. Dis.
J. 30, S35eS41.
Charles, M.D., Holman, R.C., Curns, A.T., Parashar, U.D., Glass, R.I.,
Bresee, J.S., 2006. Hospitalizations associated with rotavirus
gastroenteritis in the United States, 1993e2002. Pediatr. Infect.
Dis. J. 25, 489e493.
Clemente-Casares, P., Rodriguez-Manzano, J., Girones, R., 2009.
Hepatitis E virus genotype 3 and sporadically also genotype 1
circulate in the population of Catalonia, Spain. J. Water Health
7, 664e673.
Cunliffe, N.A., Kilgore, P.E., Bresee, J.S., Steele, A.D., Luo, N.,
Hart, C.A., Glass, R.I., 1998. Epidemiology of rotavirus
diarrhoea in Africa: a review to assess the need for rotavirus
immunization. Bull. World Health Organ. 76, 525e537.
da Silva, A.K., Le Saux, J.C., Parnaudeau, S., Pommepuy, M.,
Elimelech, M., Le Guyader, F.S., 2007. Evaluation of removal of
noroviruses during wastewater treatment, using real-time
reverse transcription-PCR: different behaviors of genogroups I
and II. Appl. Environ. Microbiol. 73, 7891e7897.
Das, B.K., Gentsch, J.R., Cicirello, H.G., Woods, P.A., Gupta, A.,
Ramachandran, M., Kumar, R., Bhan, M.K., Glass, R.I., 1994.
Characterization of rotavirus strains from newborns in New
Delhi, India. J. Clin. Microbiol. 32, 1820e1822.
Espinosa, A.C., Mazari-Hiriart, M., Espinosa, R., Maruri-Avidal, L.,
Méndez, E., Arias, C.F., 2008. Infectivity and genome
persistence of rotavirus and astrovirus in groundwater and
surface water. Water Res. 42, 2618e2628.
Estes, M., Kapikian, A.Z., 2007. Rotaviruses. In: Fields Virology,
fifth ed. Lippincott Williams &Wilkins, Philadelphia.
Felsenstein, J., 1993. Phylogeny Interference Package, Version 3.
5. Department of Genetics, University of Washington,
Seattle, USA.
Ferreira, F.F., Guimarães, F.R., Fumian, T.M., Victoria, M.,
Vieira, C.B., Luz, S., Shubo, T., Leite, J.P., Miagostovich, M.P.,
2009. Environmental dissemination of group A rotavirus:
P-type, G-type and subgroup characterization. Water Sci.
Technol. 60, 633e642.
Ferrera, A., Quan, D., Espinoza, F., 2007. Increased prevalence of
genotype G2P(4) among children with rotavirus-associated
gastroenteritis in Honduras. In: 17th European Congress of
Clinical Microbiology and Infectious Diseases ICC, Munich,
Germany.
Fumian, T.M., Leite, J.P., Castello, A.A., Gaggero, A., Caillou, M.S.,
Miagostovich, M.P., 2010. Detection of rotavirus A in sewage
samples using multiplex qPCR and an evaluation of the
ultracentrifugation and adsorption-elution methods for virus
concentration. J. Virol. Methods 170, 42e46.
Gajardo, R., Bouchriti, N., Pinto, R.M., Bosch, A., 1995. Genotyping
of rotaviruses isolated from sewage. Appl. Environ. Microbiol.
61, 3460e3462.
Gentsch, J.R., Glass, R.I., Woods, P., Gouvea, V., Gorziglia, M.,
Flores, J., Das, B.K., Bhan, M.K., 1992. Identification of group A
rotavirus gene 4 types by polymerase chain reaction. J. Clin.
Microbiol. 30, 1365e1373.
Girones, R., Ferrús, M.A., Alonso, J.L., Rodriguez-Manzano, J.,
Calgua, B., Corrêa Ade, A., Hundesa, A., Carratala, A., Bofill-
Mas, S., 2010. Molecular detection of pathogens in waterethe
pros and cons of molecular techniques. Water Res. 44,
4325e4339.
Godoy, P., Nuı́n, C., Alsedà, M., Llovet, T., Mazana, R.,
Domı́nguez, A., 2006. Waterborne outbreak of gastroenteritis
caused by Norovirus transmitted through drinking water. Rev.
Clin. Esp. 206, 435e437.
5762 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
Gómez, M.M., de Mendonça, M.C., Volotão Ede, M., Tort, L.F., da
Silva, M.F., Cristina, J., Leite, J.P., 2011. Rotavirus A genotype P
[4]G2: genetic diversity and reassortment events among
strains circulating in Brazil between 2005 and 2009. J. Med.
Virol. 83, 1093e1106.
Gouvea, V., de Castro, L., Timenetsky, M.C., Greenberg, H.,
Santos, N., 1994. Rotavirus serotype G5 associated with
diarrhea in Brazilian children. J. Clin. Microbiol. 32, 1408e1409.
Hall, T.A., 1999. BioEdit: a user-friendly biological sequence
alignment editor and analysis program for Windows 95/98/NT.
Nucl. Acids Symp. Ser. 41, 95e98.
Haramoto, E., Katayama, H., Oguma, K., Yamashita, H., Tajima, A.,
Nakajima, H., Ohgaki, S., 2006. Seasonal profiles of human
noroviruses and indicator bacteria in a wastewater treatment
plant in Tokyo, Japan. Water Sci. Technol. 54, 301e308.
Jiang, V., Jiang, B., Tate, J., Parashar, U.D., Patel, M.M., 2010.
Performance of rotavirus vaccines in developed and
developing countries. Hum. Vaccin. 6, 532e542.
Kamel, A.H., Ali, M.A., El-Nady, H.G., Aho, S., Pothier, P., Belliot, G.,
2010. Evidence of the co-circulation of enteric viruses in sewage
and in the population of Greater Cairo. J. Appl. Microbiol. 108,
1620e1629.
Kane, E.M., Turcios, R.M., Arvay, M.L., Garcia, S., Bresee, J.S.,
Glass, R.I., 2004. The epidemiology of rotavirus diarrhea in
Latin America. Anticipating rotavirus vaccines. Rev. Panam.
Salud. Publica 16, 371e377.
Kirkwood, C.D., Boniface, K., Barnes, G.L., Bishop, R.F., 2011.
Distribution of rotavirus genotypes after introduction of rotavirus
vaccines, Rotarix
and RotaTeq
, into the National
Immunization Program of Australia. Pediatr. Infect. Dis. J. 30,
48e53.
Kukkula, M., Arstila, P., Klossner, M.-L., Maunula, L., von
Bonsdorff, C.-H., Jaatinen, P., 1997. Waterborne outbreak of
viral gastro-enteritis. Scand. J. Infect. Dis 29, 415e418.
La Rosa, G., Pourshaban, M., Iaconelli, M., Muscillo, M., 2010.
Quantitative real-time PCR of enteric viruses in influent and
effluent samples from wastewater treatment plants in Italy.
Ann. Ist. Super. Sanita. 46, 266e273.
Linhares, A.C., Velázquez, F.R., Pérez-Schael, I., Sáez-Llorens, X.,
Abate, H., Espinoza, F., López, P., Macı́as-Parra, M., Ortega-
Barrı́a, E., Rivera-Medina, D.M., Rivera, L., Pavı́a-Ruz, N.,
Nuñez, E., Damaso, S., Ruiz-Palacios, G.M., De Vos, B.,
O’Ryan, M., Gillard, P., Bouckenooghe, A., 2008. Human
Rotavirus Vaccine Study Group. Efficacy and safety of an oral
live attenuated human rotavirus vaccine against rotavirus
gastroenteritis during the first 2 years of life in Latin American
infants: a randomised, double-blind, placebo-controlled phase
III study. Lancet 371, 1181e1189.
Matthijnssens, J., Rahman, M., Martella, V., Xuelei, Y., De Vos, S.,
De Leener, K., Ciarlet, M., Buonavoglia, C., Van Ranst, M., 2006.
Full genomic analysis of human rotavirus strain B4106 and
lapine rotavirus strain 30/96 provides evidence for
interspecies transmission. J. Virol. 80, 3801e3810.
Matthijnssens, J., Ciarlet, M., McDonald, S.M., Attoui, H.,
Bányai, K., Brister, J.R., Buesa, J., Esona, M.D., Estes, M.K.,
Gentsch, J.R., Iturriza-Gómara, M., Johne, R., Kirkwood, C.D.,
Martella, V., Mertens, P.P., Nakagomi, O., Parreño, V.,
Rahman, M., Ruggeri, F.M., Saif, L.J., Santos, N., Steyer, A.,
Taniguchi, K., Patton, J.T., Desselberger, U., Van Ranst, M.,
2011. Uniformity of rotavirus strain nomenclature proposed
by the Rotavirus Classification Working Group (RCWG). Arch.
Virol. 156, 1397e1413.
Meleg, E., Bányai, K., Martella, V., Jiang, B., Kocsis, B., Kisfali, P.,
Melegh, B., Szucs, G., 2008. Detection and quantification of
group C rotaviruses in communal sewage. Appl. Environ.
Microbiol. 74, 3394e3399.
Miagostovich, M.P., Ferreira, F.F., Guimaraes, F.R., Fumian, T.M.,
Diniz-Mendes, L., Luz, S.L., Silva, L.A., Leite, J.P., 2008. Molecular
detection and characterization of gastroenteritis viruses
occurring naturally in the stream waters of Manaus, central
Amazonia, Brazil. Appl. Environ. Microbiol. 74, 375e382.
Parashar, U.D., Gibson, C.J., Bresse, J.S., Glass, R.I., 2006. Rotavirus
and severe childhood diarrhea. Emerg. Infect. Dis. 12, 304e306.
Pina, S., Jofre, J., Emerson, S.U., Purcell, R.H., Girones, R., 1998.
Characterization of a strain of infectious hepatitis E virus
isolated from sewage in an area where hepatitis E is not
endemic. Appl. Environ. Microbiol. 64, 4485e4488.
Pinto, R.M., Alegre, D., Dominguez, A., El Senousy, W.M.,
Sanchez, G., Villena, C., Costafreda, M.I., Aragones, L.,
Bosch, A., 2007. Hepatitis A vı́rus in urban sewage from two
Mediterranean countries. Epidemiol. Infect. 135, 270e273.
Prado, T., Silva, D.M., Guilayn, W.C., Rose, T.L., Gaspar, A.M.,
Miagostovich, M.P., 2011. Quantification and molecular
characterization of enteric viruses detected in effluents from two
hospital wastewater treatment plants. Water Res. 45, 1287e1297.
Rajal, V.B., McSwain, B.S., Thompson, D.E., Leutenegger, C.M.,
Kildare, B.J., Wuertz, S., 2007. Validation of hollow fiber
ultrafiltration and real-time PCR using bacteriophage PP7 as
surrogate for the quantification of viruses from water
samples. Water Res. 41, 1411e1422.
Rodrı́guez-Dı́az, J., Querales, L., Caraballo, L., Vizzi, E., Liprandi, F.,
Takiff, H., Betancourt, W.Q., 2009. Detection and
characterization of waterborne gastroenteritis viruses in
urban sewage and sewage-polluted river waters in Caracas,
Venezuela. Appl. Environ. Microbiol. 75, 387e394.
Rose, T.L., Miagostovich, M.P., Leite, J.P., 2010. Rotavirus A
genotype G1P[8]: a novel method to distinguish wild-type
strains from the Rotarix vaccine strain. Mem. Inst. Oswaldo
Cruz 105, 1068e1072.
Ruiz-Palacios, G.M., Pérez-Schael, I., Velázquez, F.R., Abate, H.,
Breuer, T., Clemens, S.C., Cheuvart, B., Espinoza, F., Gillard, P.,
Innis, B.L., Cervantes, Y., Linhares, A.C., López, P., Macı́as-
Parra, M., Ortega-Barrı́a, E., Richardson, V., Rivera-Medina, D.
M., Rivera, L., Salinas, B., Pavı́a-Ruz, N., Salmerón, J.,
Rüttimann, R., Tinoco, J.C., Rubio, P., Nuñez, E., Guerrero, M.L.,
Yarzábal, J.P., Damaso, S., Tornieporth, N., Sáez-Llorens, X.,
Vergara, R.F., Vesikari, T., Bouckenooghe, A., Clemens, R., De
Vos, B., O’Ryan, M., 2006. Human rotavirus vaccine study
group. Safety and efficacy of an attenuated vaccine against
severe rotavirus gastroenteritis. N. Engl. J. Med. 354, 11e22.
Santos, N., Hoshino, Y., 2005. Global distribution of rotavirus
serotypes/genotypes and its implication for the development
and implementation of an effective rotavirus vaccine. Rev.
Med. Virol. 15, 29e56.
Schmid, D., Lederer, I., Much, P., Pichler, A.M., Allerberger, F., 2005.
Outbreak of norovirus infection associated with contaminated
flood water, Salzburg, 2005. Euro. Surveill. 16, 10e16.
Sdiri-Loulizi, K., Hassine, M., Aouni, Z., Gharbi-Khelifi, H.,
Chouchane, S., Sakly, N., Neji-Guédiche, M., Pothier, P.,
Aouni, M., Ambert-Balay, K., 2010. Detection and molecular
characterization of enteric viruses in environmental
samples in Monastir, Tunisia between January 2003 and
April 2007. J. Appl. Microbiol. 109, 1093e1104.
Solberg, O.D., Hasing, M.E., Trueba, G., Eisenberg, J.N., 2009.
Characterization of novel VP7, VP4, and VP6 genotypes of
a previously untypeable group A rotavirus. Virology 385,
58e67.
Steele, A.D., Reynders, J., Scholtz, F., Bos, P., de Beer, M.C.,
Tumbo, J., Van der Merwe, C.F., Delem, A., De Vos, B., 2010.
Comparison of 2 different regimens for reactogenicity, safety,
and immunogenicity of the live attenuated oral rotavirus
vaccine RIX4414 coadministered with oral polio vaccine in
South African infants. J. Infect. Dis. 202, S93e100.
Tamura, K., Dudley, J., Nei, M., Kumar, S., 2007. MEGA4: molecular
evolutionary genetics analysis (MEGA) software version 4.0.
Mol. Biol. Evol. 24, 1596e1599.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 5 5 e5 7 6 3
Ursu, K., Kisfali, P., Rigo, D., Ivanics, E., Erdelyi, K., Dan, A.,
Melegh, B., Martella, V., Bányai, K., 2009. Molecular analysis of
the VP7 gene of pheasant rotaviruses identifies a new
genotype, designated G23. Arch. Virol. 154, 1365e1369.
Villena, C., El-Senousy, W.M., Abad, F.X., Pinto, R.M., Bosch, A.,
2003a. Group A rotavirus in sewage samples from Barcelona
and Cairo: emergence of unusual genotypes. Appl. Environ.
Microbiol. 69, 3919e3923.
Villena, C., Gabrieli, R., Pinto, R.M., Guix, S., Donia, D.,
Buonomo, E., Palombi, L., Cenko, F., Bino, S., Bosch, A.,
5763
Divizia, M., 2003b. A large infantile gastroenteritis outbreak in
Albania caused by multiple emerging rotavirus genotypes.
Epidemiol. Infect. 131, 1105e1110.
WHO, 2008. Generic Protocol for Monitoring Impact of Rotavirus
Vaccination on Gastroenteritis Disease Burden and Viral Strains.
Available at:. WHO www.who.int/vaccines-documents/.
Zeng, S.Q., Halkosalo, A., Salminen, M., Szakal, E.D., Puustinen, L.,
Vesikari, T., 2008. One-step quantitative RT-PCR for the
detection of rotavirus in acute gastroenteritis. J. Virol.
Methods 153, 238e240.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Managed aquifer recharge of treated wastewater: Water

quality changes resulting from infiltration through
the vadose zone

Elise Bekele a,*, Simon Toze b,c, Bradley Patterson a,d, Simon Higginson e
a
CSIRO Water for a Healthy Country Flagship, CSIRO Centre for Environment and Life Sciences, Private Bag No 5, PO Wembley,
Western Australia 6913, Australia
b
CSIRO Water for a Healthy Country Flagship, Ecosciences Precinct, 41 Boggo Road, Dutton Park, QLD 4102, Australia
c
School of Population Health, University of Queensland, Herston Road, Herston, QLD 4006, Australia
d
School of Biomedical, Biomolecular and Chemical Sciences, University of Western Australia, Crawley, WA 6009, Australia
e
Water Corporation of Western Australia, PO Box 100, Leederville, WA 6902, Australia

article info abstract

Article history:
Received 8 June 2011
Received in revised form
23 August 2011
Accepted 27 August 2011
Available online 3 September 2011
Keywords:
Managed aquifer recharge
Wastewater infiltration
Natural attenuation processes
Secondary treated wastewater was infiltrated through a 9 m-thick calcareous vadose zone
during a 39 month managed aquifer recharge (MAR) field trial to determine potential
improvements in the recycled water quality. The water quality improvements of the
recycled water were based on changes in the chemistry and microbiology of (i) the recycled
water prior to infiltration relative to (ii) groundwater immediately down-gradient from the
infiltration gallery. Changes in the average concentrations of several constituents in the
recycled water were identified with reductions of 30% for phosphorous, 66% for fluoride,
62% for iron and 51% for total organic carbon when the secondary treated wastewater was
infiltrated at an applied rate of 17.5 L per minute with a residence time of approximately
four days in the vadose zone and less than two days in the aquifer. Reductions were also
noted for oxazepam and temazepam among the pharmaceuticals tested and for a range of
microbial pathogens, but reductions were harder to quantify as their magnitudes varied
over time. Total nitrogen and carbamazepine persisted in groundwater down-gradient
from the infiltration galleries. Infiltration does potentially offer a range of water quality
improvements over direct injection to the water table without passage through the
unsaturated zone; however, additional treatment options for the non-potable water may
still need to be considered, depending on the receiving environment or the end use of the
recovered water.
Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.

1. Introduction either ponds, basins or shallow buried trenches. There is

An essential design question for proponents of managed
aquifer recharge (MAR) schemes using recycled water in
shallow aquifers is whether it is more beneficial to recharge
the aquifer via direct well injection or to use infiltration with
a range of different types of MAR as described in Dillon (2005).
Ideally the design of a MAR scheme in an urban setting should
have minimal surface footprint and minimal exposure
potential for the community. Well injection offers advantages
such as no evaporative loss, algae or mosquitoes, and no loss

* Corresponding author. Tel.: þ61 0 8 9333 6718; fax: þ61 0 8 9333 6211.
E-mail address: Elise.Bekele@csiro.au (E. Bekele).
0043-1354/$ e see front matter Crown Copyright ª 2011 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.058
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
of prime land (Pyne, 2006). Well injection, however, can suffer
from clogging issues relating to the initial quality of the
recycled water and loses the benefit of potential treatment
processes provided by the unsaturated zone. In comparison,
surface infiltration offers potential treatment during the
migration of recycled water through the vadose zone, which
can result in improved water quality via biological, chemical
and physical processes before reaching the water table but
can have the drawback of having a larger surface footprint and
potential for exposed water, none of which is ideal in an urban
environment. Improvements to water quality using infiltra-
tion have been demonstrated to reduce organic matter
(Quanrud et al., 2003; Vanderzalm et al., 2010), trace organic
compounds (Montgomery-Brown et al., 2003), nitrogen (Zhang
et al., 2005) and bacteria (Schafer et al., 1998; Toze et al., 2004).
However, a rigorous determination of the magnitudes of
concentration reductions for a range of different chemical and
biological contaminants in treated wastewater has not been
conducted previously in MAR studies using unsaturated,
calcareous sands such as the Spearwood sands of Swan
Coastal Plain of Western Australia.
The aim of this study was to determine the changes in
recycled water quality after infiltrating vertically through
a 9 m-thick vadose zone and migrating laterally through 2.3 m
of aquifer using infiltration galleries as the MAR method
during a 39 month MAR field trial. The focus of this work is on
MAR in calcareous sand and limestone due to the prevalence
of these deposits, which comprise unconfined aquifers in the
Perth metropolitan region of Western Australia where there is
also keen interest in enhancing the role of MAR (Scatena and
Williamson, 1999; Smith and Pollock, 2010). The evaluation of
water quality was based on measured concentrations of major
ions, nutrients, trace metals, organic carbon, pharmaceutical
compounds (carbamazepine, diazepam, oxazepam,
phenytoin, temazepam) and numbers of faecal indicator
microorganisms and selected enteric pathogens (thermoto-
lerant coliforms, enterococci, bacteriophage, adenovirus)
before and after recycled water passed through the
subsurface.
2. Materials and methods
2.1. Facilities
The source of secondary treated wastewater was the Subiaco
Wastewater Treatment Plant in Western Australia. After
passage through a multi-media filtration system (AMIAD),
the treated wastewater was pumped to the recharge site
(Bekele et al., 2009). Two infiltration galleries were used at the
CSIRO Centre for Environment and Life Sciences in Floreat,
Western Australia for a pilot-scale investigation of MAR
(Bekele et al., 2009). The east gallery was filled with 10 mm
graded and washed granite gravel; the west gallery contained
a series of modular polypropylene tanks, referred to as the
Atlantis system by the manufacturer. The dimensions of
each tank were 685 mm  408 mm  450 mm (length by width
by height) and the tanks have a modular design so that they
can be positioned in the trench and clipped together. The
construction of each module resembles a milk crate that has
5765
large void spaces supported by matrix. In this study, each
gallery trench was 25 m  1 m  0.5 m (length by width by
height) and the width of the Atlantis gallery was 816 mm,
consisting of two tank modules positioned side-by-side. Each
gallery was buried to a depth of 1 m below ground and covered
with 0.5 m of sediment backfill composed of Spearwood Dune
sand from digging the trench (Bekele et al., 2009). The recharge
site was located in a grassy paddock where sheep were
allowed to graze. The infiltration galleries operated almost
continuously for 39 months and infiltrated a total of 36.7 ML of
treated wastewater to the aquifer supplied at a daily constant
rate of 17.5 L per minute (Bekele et al., 2009).
2.2. Local geology and hydraulic considerations
The geology of the site consisted of a 7 m-thick top layer of
Spearwood Dune sand, overlying the Tamala Limestone
aquifer. The Tamala Limestone is a calcareous aeolinite which
has been weathered to produce the overlying Spearwood
sands (Tapsell et al., 2003). The aquifer extends to a depth of
31 m below ground and is underlain by sediments from
a regional aquitard. Regional investigations of the Tamala
Limestone reveal high porosity zones from fractures and
cavities (Davidson, 1995).
The sand mineralogy from 0 to 7.5 m below ground was
predominantly quartz (>80%), underlain by a more heteroge-
neous section from 7.5 to 11.6 m below ground with an
average composition of quartz (60%), calcite (30%), microcline
(5%) and anorthite (5%) from XRD analysis. Further mineral
phase characterization using AutoGeoSEM (Robinson et al.,
2000) on a sand sample revealed aluminum and iron oxides.
The aluminum oxides are silicate weathering products (clay
minerals) deposited as coatings on sand grains, which are
colored yellow by the presence of hydrated iron oxides
(Bastian, 1996).
The depth to the water table below the galleries varied
seasonally between 10 and 11 m. The regional groundwater
flow direction was from east to west. For experimental
purposes, an artificial hydraulic gradient was produced by
continual pumping from a well located 50 m west of the
infiltration galleries. The natural hydraulic gradient coincided
with the imposed gradient. A series of monitoring wells were
installed with slotted intervals positioned at different depths
below ground as shown in Fig. 1.
The migration rate of the infiltrated recycled water through
the vadose zone was previously determined based on
a bromide tracer experiment (Bekele et al., 2009). From this
data, a minimum travel time of 3.7 days was estimated
through the unsaturated zone. The travel time to BH1 after
passage through the vadose zone was estimated to be an
additional 0.5 day based on a comparison of the electrical
conductivity of the recycled water relative to groundwater
from BH1, indicating that the substantial time for processes to
occur was in the vadose zone compared to the aquifer.
2.3. Description of monitoring wells
To assess the potential infiltration benefits, improvements in
quality of the recycled water immediately down-gradient
from the infiltration galleries were compared to the quality
5766 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2

general accordance with ASTM D 4448-01 Standard Guide for

Sampling Ground-Water Monitoring Wells (2001). Wastewater
samples were collected from the discharge chamber in the
infiltration galleries.
Water samples were analyzed for major ions, nutrients,
trace metals, dissolved and total organic carbon, trace
organics and faecal indicator microorganisms and pathogens.
Further details regarding water sampling, analytical proce-
dures and detection limits for the water chemistry are given in
Bekele et al. (2009). Faecal indicator microbes E. coli and
enterococci were detected using membrane filtration and
selective media while the presence of enteric pathogens were
determined from a concentrate of 40 L using PCR as described
in Toze and Bekele (2009). Sampling water from the galleries
and the wells for different chemicals and microbes was con-
ducted during the first 25 months of the MAR trial at different
time intervals, but generally on a weekly, fortnightly or
monthly basis.
Physical water parameters measurements were taken
before water sampling occurred. These included pH, dissolved
oxygen, electrical conductivity, temperature, oxidation-
reduction potential and turbidity measurements using
a Troll 9000 multisensory meter housed within a flow cell.

2.5. Criteria to assess changes in recycled water quality

Fig. 1 e The Floreat MAR site showing the positions of the
two galleries, the monitoring wells (BH1, BH2 and BH5) and
the recovery well in map view (a), and relative to the
maximum height of the water table in cross-section (b).
of the infiltrated recycled water. Monitoring wells close to the
infiltration galleries (BH1, BH2 and BH5) were selected to
evaluate the benefits to water quality of infiltrating recycled
water through the vadose zone. Three background ground-
water monitoring wells located hydraulically up-gradient of
the infiltration galleries (BGRND1, BGRND2 and BGRND3) were
also monitored to provide an assessment of changes in
groundwater quality due to infiltration. Monitoring well
details are given in Table 1.
2.4. Collection and analysis of water samples
Details of the groundwater sampling procedure are given in
Bekele et al. (2009). Groundwater sampling procedures were in
To assess the water quality changes as a result of the migra-
tion through the vadose zone, the quality of water samples
collected from the shallow monitoring wells immediately
down-gradient from the infiltration galleries were compared
to the quality of the recycled water prior to infiltration.
Sampling of the vadose zone water directly below the infil-
tration galleries and immediately above the water table was
not undertaken. Thus, only gross changes after passage over
the entire vadose zone could be quantified as changes during
infiltration could not be determined. The significance of the
effects of migration through the subsurface on recycled water
quality was assessed using Student’s two-tailed t-test
(unpaired) applied to datasets of measured nutrients, inor-
ganic compounds, pharmaceutical and microbial pathogens,
comparing the mean concentrations in recycled water prior to
infiltration relative to the mean concentrations in recharged
water sampled from BH1. For the Student’s t-test, the null
hypothesis was that water quality was unchanged despite
passage through the vadose zone. Microsoft Excel was used to

Table 1 e Monitoring well details.

Well Ground elevation Distance and direction Total depth Screened depth interval below the
(m AHDa) from west gallery below ground (m) maximum height of the water table (m)

BH1 12.97 2.3 m west 12.01 0.0e2.0

BH2 13.01 2.5 m west 12.04 0.0e2.0
BH5b 12.86 8.8 m east 10.81 0.0e1.0
BGRND1 15.94 185 m northeast 15.00 0.0e2.05
BGRND2 12.00 75 m east 12.23 2.23e3.23
BGRND3 12.00 75 m east 20.11 10.11e11.11

a Australian Height Datum.

b BH5 is located up-gradient relative to the groundwater flow direction and a distance of 4 m east of the east gallery.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2 5767

calculate the probability (P) and the null hypothesis was
rejected if P < 0.05.
3. Results and discussion
3.1. Characterization of recycled water and ambient
groundwater
To understand what water quality changes were potentially
achievable during passage of the recycled water through the
vadose zone, it was also necessary to characterise and
compare the ambient groundwater. Both the ambient
groundwater and the recycled water were consistently
aerobic. The average water temperature, pH, electrical
conductivity, dissolved oxygen concentrations and sulfate
concentrations of the ambient groundwater were similar to
that of the recycled water (Table 2). The concentration of total
dissolved solids was highly variable in the recycled water and
the range of variation overlapped with that of the ambient
groundwater. The major ions defining the water types were
CaeNaeCleHCO3 for ambient groundwater and NaeCleHCO3
for recycled water.
The recycled water and ambient groundwater had low
concentrations of several inorganic species, namely arsenic,
boron, cadmium, cobalt, chromium, copper, mercury,
manganese, molybdenum, nickel, lead, selenium, uranium,
vanadium and zinc. The impact of infiltration on these inor-
ganic chemicals was not investigated as their measured
concentrations were either very low or were below detection
limits.
3.2. MAR impacts on groundwater quality near the
galleries
Site groundwater monitoring was undertaken to confirm that
the shallow groundwater collected closest to the infiltration
galleries did represent infiltrated recycled water and not
ambient groundwater. Average concentrations of potassium,
chloride and sodium in the recycled water were similar to the
average concentrations from water samples collected from
the monitoring well BH1 (Table 3). The recycled water had an
average potassium concentration that was greater than four-
fold more enriched in potassium compared to the ambient
groundwater and therefore was an effective tracer for the
infiltrated recycled water as demonstrated by others in
previous studies (Rueedi et al., 2009; Wolf et al., 2004). The
recharged water quality sampled from monitoring wells BH1
and BH2 was indistinguishable from the recycled water prior
to infiltration, but substantially different from ambient
groundwater on the basis of potassium and chloride concen-
trations (Fig. 2).
Water samples collected from the monitoring well BH5,
4 m up-gradient from the galleries, showed a greater propor-
tion of ambient groundwater mixed with recycled water
compared to the water samples from the monitoring wells
down-gradient of the galleries. A theoretical line of mixing
between the average concentrations of chloride and potas-
sium in ambient groundwater and that of recycled water was
generated (Fig. 2). During the first three months of infiltration,
the composition of groundwater from BH5 was predominantly
ambient groundwater (80%); thereafter, the composition of
groundwater from BH5 had a higher proportion of recycled
water relative to ambient groundwater. Since water sampled
from well BH1 did not show temporal transition and it was
frequently monitored, water collected from BH1 was used for
evaluating influent water quality improvements as a result of
its migration through the vadose zone.
3.3. Water quality changes during infiltration
3.3.1. Inorganic chemicals
The recycled water prior to infiltration had higher average
concentrations of most species of nitrogen, phosphorous, and
fluoride compared to ambient groundwater (Table 3).
Recharged water sampled down-gradient from the galleries
from BH1 had average concentrations for these chemicals that
were generally between those for ambient groundwater and
recycled water or within one standard deviation of these two
water sources.
To assess improvements due to infiltration, total nitrogen
concentrations and a relatively non-reactive analyte (i.e.
Table 3 e Mean concentrations (mg/L) of selected
chemicals in recycled water and groundwater sampled
down-gradient from the galleries from BH1 compared to
ambient groundwater. Standard deviations are provided
in parentheses.
Analtye Recycled Ambient BH1
water groundwater

Chloride 245 (62) 162 (27) 248 (50)

Table 2 e Mean values of water quality parameters of the Potassium 22.9 (3.5) 4.96 (0.83) 21.7 (4.1)
recharge water that were similar to ambient groundwater Sodium 194 (44) 92.6 (15) 194 (42)
with standard deviations in parentheses. Ammonia as Nitrogen 0.64 (0.85) <0.01 0.037 (0.046)
Nitrate as Nitrogen 2.16 (1.41) 0.16 (0.26) 3.90 (1.61)
Parameter Recycled Ambient
Total Kjeldahl Nitrogen 1.86 (0.98) 0.066 (0.030) 0.95 (0.51)
(Unit of Measure) water Groundwater
Total Nitrogen 4.27 (1.9) 0.302 (0.36) 4.93 (1.8)
Water temperature ( C) 24 (4) 22 (1) Soluble reactive 6.31 (3.32) 0.0126 (0.007) 2.19 (1.53)
pH 7.33 (1.11) 7.04 (0.88) phosphorus as P
Electrical Conductivity (mS m1) 143 (53) 124 (63) Total organic carbon 9.98 (3.8) 2.66 (2.72) 6.42 (3.57)
Eh (mV-SHE) 385 (184) 321 (189) Fluoride 0.73 (0.2) 0.15 (0.1) 0.25 (0.05)
Dissolved Oxygen (mg L1) 2.15 (1.82) 4.02 (2.56) Iron 0.14 (0.20) 0.44 (0.72) 0.053 (0.054)
Sulfate as S (mg L1) 64.1 (8) 64.5 (18) Calcium 28.6 (9.49) 98.8 (23.5) 60.4 (7.28)
Total Dissolved Solids (mg L1) 755 (179) 644 (25) Aluminum 0.018 (0.011) 0.22 (0.48) 0.045 (0.048)
5768
30
25
Potassium (mg/L)
20
15
10
5 BH5, samples during first 3 months
0
0 100 200
Chloride (mg/L)
Fig. 2 e Cross plot showing concentrations of chloride and
potassium in ambient groundwater sampled from the
background monitoring wells, recycled water and in water
recovered from the monitoring wells (BH1, BH2 and BH5). A
mixing line is shown between the average composition of
recycled water and ambient groundwater.
potassium) were plotted (Fig. 3). No improvement in total
nitrogen was demonstrated as shown by the cluster of data for
BH1 in Fig. 3, which overlaps with that of recycled water and
indicates a higher range of total nitrogen in contrast to
ambient groundwater (<1.2 mg/L). Migration of recycled water
through the vadose zone and aquifer to reach BH1 did not
improve total nitrogen concentrations (P ¼ 0.24, Student’s t-
test).
Although passage through the vadose zone did not signif-
icantly alter the concentration of total nitrogen in the recycled
water, there were changes to different nitrogen species
(Table 3). The mean concentration of total Kjeldahl nitrogen in
the recharged water sampled from BH1 was 49% lower relative
to the mean concentrations in recycled water prior to infiltra-
tion by (P < 0.01, Student’s t-test), whereas the mean concen-
tration of ammonia was 94% lower (P < 0 0.01). In comparison,
the mean concentration of nitrate was 77% higher (P < 0.001) in
the recharged water sampled from BH1 relative to the mean
concentrations in recycled water prior to infiltration. This data
suggests that nitrification was occurring through the vadose
10
9 14
Total Nitrogen (mg/L)
8 BH1
7
6 10
5
4
3 6
2
1
0 2
0 5 10 15 20
Potassium (mg/L)
Fig. 3 e Cross plot showing concentrations of total nitrogen
and potassium in recycled water, ambient groundwater
and recharged water sampled down-gradient from the
galleries from monitoring well BH1.
w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
zone with nitrate production from the ammonium and total
Kjeldahl nitrogen, and provides a plausible explanation for the
lack of change in total nitrogen concentrations.
As conditions in the vadose zone were not conducive for
reducing nitrate concentrations, potential treatment options
to consider include manipulation of the oxygen content to
drive denitrification and the addition of carbon amendments
Ambient groundwater to promote nitrate removal within the aquifer (Patterson et al.,
BH1
BH2 2004, 2011).
BH5, samples after first 3 months
Migration of the recycled water through the vadose zone
Recycled water and 2.3 m of aquifer aided removal of phosphorous (P < 0.0001,
300 400 500 Student’s t-test). Further investigation revealed a time-
dependence for phosphorous removal. During the first 124
days of infiltration, phosphorous in recharged water sampled
from the monitoring well BH1 down-gradient from the
galleries remained low, similar to ambient groundwater
(<0.02 mg/L), whereas the average phosphorous in the
influent recycled water during this initial period was 10 mg/L
(Fig. 4). After 271 days of infiltration, the average phosphorous
concentration in the influent recycled water was 4.7 mg/L and
more variable than previously (standard deviation of 2.1 mg/
L). In comparison, the concentration of phosphorous in
recharged water collected from BH1 reached a maximum on
day 271 and then stabilized with a mean concentration of
3.2 mg/L (standard deviation of 0.68 mg/L).
Comparison of average phosphorous concentrations in
BH1 and the influent recycled water reveals a reduction of 31%
after 271 days of infiltration.
Adsorption and precipitation are reported to be the main
causes of phosphorous retention in calcareous sands and soils
(von Wandruszka, 2006; Whelan, 1988; Whelan and Barrow,
1984). However, calcareous sands have a limited capacity to
store phosphorous, thus caution may be warranted in antici-
pating consistent levels of phosphorous removal, particularly
at sites recharging higher volumes of recycled water and after
prolonged recharge. Desorption and remobilization of phos-
phorous should be considered in dealing with long-term MAR
systems. Previous studies have demonstrated the transient
nature of phosphorous retention in carbonate soils receiving
wastewater. Examples include the release of sorbed phos-
phate caused by dissolution of carbonate, e.g. from the acid-
ification of leachate during nitrification of ammonium
Ambient groundwater
12 Recycled water
Phosphorous (mg/L)
8
4
25 30 35
Ambient groundwater 0
BH1 0 100 200 300 400 500 600 700
Recycled water Time since start of infiltration (days)
Fig. 4 e Temporal changes in concentrations of
phosphorus in recycled water, ambient groundwater and
water sampled down-gradient from the galleries from
monitoring well BH1.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
(Whelan, 1988), and soil saturated with phosphate becoming
a source of phosphate if concentrations in the wastewater
recharge decrease (Lin and Banin, 2005).
Migration through the subsurface reduced fluoride
concentrations in the infiltrated water (P < 0.0001, Student’s t-
test). The comparison of average fluoride concentrations in
BH1 and the influent recycled water reveals a reduction of 66%
(Table 3). Similar to the reduction in phosphorous, reductions
in fluoride concentrations were likely due to adsorption and
precipitation in calcite-bearing layers (Turner et al., 2005).
Adsorption of fluorite was more likely as both the recycled
water and ambient groundwater were under-saturated with
respect to fluorite.
Concentrations of soluble iron in the ambient groundwater
sampled from the three background wells were quite variable
(mean of 0.44 mg/L; standard deviation of 0.72 mg/L) making it
difficult to interpret changes to recycled water quality from
passage through the subsurface. Infiltration of the recycled
water sustained aerobic conditions favorable for the removal
of soluble iron near the gallery. This may explain the nearly
three-fold difference between the average concentration of
soluble iron in water sampled from well BH1 and recycled
water prior to infiltration (P ¼ 0.013, Student’s t-test).
Comparison of average iron concentrations in BH1 and the
influent recycled water reveals a reduction of 62%.
While the pH of the recycled water varied over time during
the trial, the pH of the groundwater at BH1 was consistently
higher than the recycled water. During infiltration, the
average pH of water sampled from well BH1 was 7.61 (stan-
dard deviation of 0.80), whereas the average pH of recycled
water prior to infiltration was 7.33 (standard deviation of 1.11;
Table 2). The increase in pH of the recycled water as it infil-
trated through the calcareous sand and limestone was likely
due to calcite dissolution in the vadose zone.
Calcium and aluminium concentrations in water collected
from well BH1 were higher relative to their concentrations in
the recycled water; however ambient groundwater had the
highest concentrations of these ions (Table 3). The average
calcium concentration in the water sampled from BH1 was
twice the average concentration in the recycled water prior to
infiltration. The recycled water was under-saturated with
respect to calcite whereas ambient groundwater was in
equilibrium with calcite; hence conditions were favourable for
calcite dissolution thereby increasing calcium in water
sampled down-gradient from the galleries. The average
concentration of aluminium in groundwater sampled from
well BH1 was also twice the average aluminium concentration
in the recycled water. It is likely that the mobility of
aluminium is linked to dissolution of calcium carbonate
coated with aluminium oxides.
3.3.2. Organic carbon
Migration of recycled water through the subsurface to reach
well BH1 reduced the total organic carbon concentrations
(P < 0.01, Student’s t-test). Comparison of average TOC
concentrations in BH1 and the influent recycled water reveals
a reduction of 51% after 400 days of infiltration (Fig. 5). Prob-
able mechanisms for the removal of organic carbon included
filtration, adsorption and biodegradation (Drewes et al., 2003;
Fox et al., 2001; Vanderzalm et al., 2010).
5769
30
Ambient groundwater
BH1
Total Organic Carbon (mg/L)
25 Recycled water
20
15
10
5
0
0 100 200 300 400 500 600 700
Time since start of infiltration (days)
Fig. 5 e Temporal changes in concentrations of total
organic carbon in recycled water, ambient groundwater
and down-gradient from the galleries in water sampled
from monitoring well BH1.
3.3.3. Pharmaceuticals
All of the pharmaceuticals tested revealed concentrations in
ambient groundwater that were below detection limits (i.e.
<0.05 mg/L for carbamazepine; <0.01 mg/L for diazepam,
oxazepam, phenytoin and temazepam). In comparison, diaz-
epam and phenytoin were the only pharmaceuticals below
the detection limit in the recycled water (Table 4). Migration of
recycled water through the vadose zone produced reductions
in oxazepam and temazepam (P < 0.01, Student’s t-test);
however, concentrations of these pharmaceuticals in water
collected from BH1 were highly variable. Additional treatment
may be required to produce consistent levels of concentration
reduction for oxazepam and temazepam, depending on the
end use of the extracted water and the potential for large
volumes to be consumed intentionally or unintentionally by
the community. Sorption and biodegradation are common
mechanism for the removal of pharmaceuticals under aerobic
conditions in the vadose zone (Amy and Drewes, 2007; Conn
et al., 2010; Scheytt et al., 2006; Tiehm et al., 2011), but
further investigation of removal mechanisms for the phar-
maceuticals was not undertaken in this study.
An unexpected observation was that concentrations of car-
bamazepine in the water collected from well BH1 were consis-
tently higher (80% higher on average; standard deviation of 37%)
than carbamazepine detected in the recycled water. The higher
carbamazepine concentrations from BH1 relative to recycled
water may be due to changes in concentration in the recycled
water over time, which varied between 0.13 and 0.33 mg/L
(average of 0.21 mg/L, Table 4). In column experiments using
aquifer sediment and recycled water from the MAR project and
under fully saturated, aerobic conditions, no sorption or
degradation were observed for carbamazepine and oxazepam
(Patterson et al., 2009). The persistence of carbamazepine in the
environment has been demonstrated previously and it has been
suggested that it could be used as a potential anthropogenic
marker in aquatic environments (Clara et al., 2004).
These results may be important for wastewater reuse near
wetlands or ecologically-sensitive areas where endocrine-
disrupting chemicals could impact negatively on aquatic
species in the receiving environment (Lister et al., 2009;
Saaristo et al., 2010; Sulleabhain et al., 2009).
5770 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
Table 4 e Mean concentrations (mg/L) of pharmaceuticals
in recycled water, water sampled from BH1, and ambient
groundwater. Standard deviations are given in
parentheses.
Analtye Recycled water Ambient BH1
(n ¼ 25) groundwater (n ¼ 16)
(n ¼ 33)
Carbamazepine 0.21 (0.067) <0.05 0.38 (0.089)
Diazepam <0.1 <0.1 <0.1
Oxazepam 0.31 (0.090) <0.1 0.21 (0.14)
Phenytoin <0.1 <0.1 <0.1
Temazepam 0.31 (0.084) <0.1 0.17 (0.12)
3.3.4. Faecal indicator microorganisms and enteric pathogens
It is worth noting that thermotolerant coliforms and entero-
cocci were detected in ambient groundwater from wells up-
gradient from the recharge site in the sheep paddock. With
regard to thermotolerant coliforms, 13% of ambient ground-
water samples tested (n ¼ 38) and 19% of water samples from
BH1 (n ¼ 27) had numbers greater than 10 cfu 100 mL1. It is
hypothesized that surface contamination was most likely due
to excreta from the grazing sheep at the recharge site as the
source of these microbes and not the treated wastewater
being recharged to the aquifer.
Despite the additional source of faecal indicator micro-
organisms from sheep excreta, migration of recycled water
from the galleries through the vadose zone to well BH1
produced a reduction in thermotolerant coliforms
(P < 0.0001, Student’s t-test). Thermotolerant coliforms were
routinely detected in the recycled water with 80% of the
samples tested (n ¼ 46), exceeding 10 colony forming units
(cfu) 100 mL1.
High numbers of enterococci were present in the recycled
water with 79% of the water samples tested (n ¼ 48) exceeding
100 cfu 100 mL1, whereas 19% of ambient water samples
(n ¼ 42) exceeded 100 cfu 100 mL1. Despite the high counts in
the recycled water, there was some reduction in enterococci
numbers comparing recycled water and water from BH1 with
only 28% of samples from the BH1 (n ¼ 29) exceeding 100 cfu
100 mL1 (P < 0.001, Student’s t-test).
Reductions in microbial pathogen numbers in the recycled
water were shown by fewer detections of adenovirus and Fþ
bacteriophage after migration through the subsurface.
Adenovirus was detected in 68% of the samples of the recycled
water (n ¼ 19) prior to infiltration but in only 6% of the samples
from BH1 (n ¼ 18) and in none of the wells further down
gradient. Adenovirus was not detected in any of the ambient
groundwater samples. Fþ bacteriophage, commonly used as
a surrogate enteric virus, were detected in 94% of recycled
water samples tested (n ¼ 36). In comparison only 4% of the
water samples from BH1 (n ¼ 24), and 6% of the ambient
groundwater samples (n ¼ 33) were positive for this
bacteriophage.
Reductions in microbial pathogens in treated wastewater
recharged to the Tamala Limestone aquifer over time were
demonstrated using survival experiments conducted with
selected faecal indicators (Cryptosporidium, adenovirus, rota-
virus, coxsackievirus, MS2, E. coli, S. enterica and E. faecalis) in
in-situ diffusion chambers (Toze et al., 2010). The reductions in
microbial pathogens were attributed to a combination of
physical removal processes during filtration and the activity of
indigenous groundwater microorganisms (Gordon et al., 2002;
Toze and Hanna, 2002). The aquifer has an active treatment
capacity to remove pathogens (Toze and Bekele, 2009), but
a longer period of aquifer residence may be needed to allow for
more inactivation of microbial pathogens (Toze et al., 2010).
4. Conclusions
 Water quality benefits were achieved by infiltrating
secondary treated wastewater through weathered calcar-
eous sands and limestone of the vadose zone in an urban
area using infiltration galleries.
 Reductions in the average concentrations of several
constituents in the recycled water before and after MAR
were identified as follows: 30% for phosphorous, 66% for
fluoride, 62% for iron and 51% for total organic carbon.
Phosphorous removal declined over time, implying the
“maximum” P adsorption capacity was reached. Reductions
in the average concentrations of oxazepam and temazepam
and the numbers of thermotolerant coliforms and entero-
cocci in the infiltrated water were also identified, but not as
easily quantified.
 The removal rates determined for chemical and biological
species are particular to the MAR conditions in this study.
The impact of a thicker unsaturated zone, anoxic conditions
and different flow rates and contact times were not inves-
tigated and would give rise to different removal rates.
 The aerobic conditions present in the vadose zone were not
conducive for denitrification to reduce nitrate concentra-
tions in the water recharged to the aquifer, revealing that
secondary treated wastewater recharged via infiltration
cannot rely entirely upon processes in the vadose zone for
nitrate removal.
 Geochemical conditions were favorable for calcite dissolu-
tion, suggesting that porosity in the calcareous vadose zone
may increase over time and could cause faster breakthrough
of contaminants to the water table.
 Future MAR in this aquifer type should consider the effects
of prolonged recharge, higher rates of recharge and
concentration-dependency of adsorption and precipitation
reactions (e.g. controlling phosphorous mobility) to deter-
mine the long-term sustainability of recharging recycled
water to a calcareous aquifer in urban environments.
Acknowledgements
This research was funded by the Western Australian
Government through the Water Foundation (gs1), the Water
Corporation of WA and the CSIRO Water for a Healthy
Country Flagship Program. Chemical analyses were
conducted by the Chemistry Centre of Western Australia.
The authors would like to thank Mr. Mark Shackleton and
Mr. Sebit Gama from the CSIRO for their assistance with
water sampling and microbial analyses. Drs Joanne
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
Vanderzalm and Grant Douglas (CSIRO) are gratefully
acknowledged for discussions on the geochemistry. The
content of this paper benefited from internal reviews by Drs
Declan Page and Warish Ahmed at the Commonwealth
Scientific and Industrial Research Organization.
references
Amy, G., Drewes, J., 2007. Soil aquifer treatment (SAT) as a natural
and sustainable wastewater reclamation/reuse technology:
fate of wastewater effluent organic matter (EfOM) and trace
organic compounds. Environmental Monitoring and
Assessment 129 (1e3), 19e26.
ASTM, 2001. Standard Guide for Sampling Ground-water
Monitoring Wells D 4448-01. edited. American Society for
Testing and Materials (ASTM).
Bastian, L., 1996. Residual soil mineralogy and dune subdivision,
Swan coastal plain, Western Australia. Australian Journal of
Earth Sciences 43 (1), 31e44.
Bekele, E., Toze, S., Patterson, B., Devine, B., Higginson, S.,
Fegg, W., Vanderzalm, J., 2009. Design and operation of
infiltration galleries and water quality guidelines - Chapter 1.
In: Toze, S., Bekele, E. (Eds.), Determining the Requirements
for Managed Aquifer Recharge in Western Australia (A Report
to the Water Foundation).
Clara, M., Strenn, B., Kreuzinger, N., 2004. Carbamazepine as
a possible anthropogenic marker in the aquatic environment:
investigations on the behaviour of Carbamazepine in
wastewater treatment and during groundwater infiltration.
Water Research 38 (4), 947e954.
Conn, K.E., Siegrist, R.L., Barber, L.B., Meyer, M.T., 2010. Fate of
trace organic compounds during vadose zone soil treatment in
an onsite wastewater system. Environmental Toxicology and
Chemistry 29 (2), 285e293.
Davidson, W.A., 1995. Hydrogeology and groundwater resources
of the Perth region, Western Australia, Bulletin. Western
Australia Geological Survey 142.
Dillon, P., 2005. Future management of aquifer recharge.
Hydrogeology Journal 13 (1), 313e316.
Drewes, J.E., Reinhard, M., Fox, P., 2003. Comparing microfiltration-
reverse osmosis and soil-aquifer treatment for indirect potable
reuse of water. Water Research 37 (15), 3612e3621.
Fox, P., Narayanaswamy, K., Genz, A., Drewes, J., 2001. Water
quality transformations during soil aquifer treatment at the
Mesa Northwest water reclamation plant, USA. Water Science
and Technology 43 (10), 343e350.
Gordon, C., Wall, K., Toze, S., O’Hara, G., 2002. Influence of
Conditions on the Survival of Enteric Viruses and Indicator
Organisms in Groundwater. In: Paper Presented at
Proceedings of the 4th International Symposium on Artificial
Recharge of Groundwater Isar-4-Management of Aquifer
Recharge for Sustainability. Balkema Publishers, Adelaide SA.
Lin, C.Y., Banin, A., 2005. Effect of long-term effluent recharge on
phosphate sorption by soils in a wastewater reclamation
plant. Water Air and Soil Pollution 164 (1e4), 257e273.
Lister, A., Regan, C., Van Zwol, J., Van Der Kraak, G., 2009.
Inhibition of egg production in zebrafish by fluoxetine and
municipal effluents: a mechanistic evaluation. Aquatic
Toxicology 95 (4), 320e329.
Montgomery-Brown, J., Drewes, J., Fox, P., Reinhard, M., 2003.
Behavior of alkylphenol polyethoxylate metabolites during
soil aquifer treatment. Water Research 37 (15), 3672e3681.
Patterson, B., Pearce, J., Spadek, T., Linge, K., Busetti, F.,
Shackleton, M., Furness, A., Blair, P., Heitz, A., 2009. Fate of
trace organics using laboratory column experiments - chapter
5771
2. In: Toze, S., Bekele, E. (Eds.), Determining the Requirements
for Managed Aquifer Recharge in Western Australia (A Report
to the Water Foundation).
Patterson, B.M., Grassi, M.E., Robertson, B.S., Davis, G.B., Smith, A.
J., McKinley, A.J., 2004. Use of polymer mats in series for
sequential reactive barrier remediation of ammonium-
contaminated groundwater: field evaluation. Environmental
Science & Technology 38 (24), 6846e6854.
Patterson, B.M., Shackleton, M., Furness, A.J., Bekele, E., Pearce, J.,
Linge, K.L., Busetti, F., Spadek, T., Toze, S., 2011. Behaviour and
fate of nine recycled water trace organics during managed
aquifer recharge in an aerobic aquifer. Journal of Contaminant
Hydrology 122 (1e4), 53e62.
Pyne, R.D.G., 2006. Groundwater Recharge and Wells: a Guide to
Aquifer Storage Recovery, second ed. CRC Press, Boca Raton,
Florida, 401 pp.
Quanrud, D., Hafer, J., Karpiscak, M., Zhang, H., Lansey, K.,
Arnold, R., 2003. Fate of organics during soil-aquifer
treatment: sustainability of removals in the field. Water
Research 37 (14), 3401e3411.
Robinson, B.W., Hitchen, G.J., Verrall, M.R. 2000, The
AutoGeoSEM: A programmable fully-automatic SEM for rapid
grain-counting and heavy mineral characterisation in
exploration, paper presented at International Mineralogical
Association Commission on Ore mineralogy, Proceeding of the
Modern Approaches to Ore and Environmental mineralogy,
Short Course, p. 71e74.
Rueedi, J., Cronin, A.A., Morris, B.L., 2009. Estimation of sewer leakage
to urban groundwater using depth-specific hydrochemistry.
Water and Environment Journal 23 (2), 134e144.
Saaristo, M., Craft, J.A., Lehtonen, K.K., Lindstrom, K., 2010. An
endocrine disrupting chemical changes courtship and parental
care in the sand goby. Aquatic Toxicology 97 (4), 285e292.
Scatena, M.C., Williamson, D.R. 1999, A potential role for artificial
recharge within the Perth region: a pre-feasibility study.
Center for Groundwater Studies report No. 84.
Schafer, A., Ustohal, P., Harms, H., Stauffer, F., Dracos, T.,
Zehnder, A., 1998. Transport of bacteria in unsaturated porous
media. Journal of Contaminant Hydrology 33 (1e2), 149e169.
Scheytt, T.J., Mersmann, P., Heberer, T., 2006. Mobility of
pharmaceuticals carbamazepine, diclofenac, ibuprofen, and
propyphenazone in miscible-displacement experiments.
Journal of Contaminant Hydrology 83 (1e2), 53e69.
Smith, A.J., Pollock, D.W., 2010. Artificial Recharge Potential of the
Perth Region Superficial Aquifer: Lake Preston to Moore River.
CSIRO (Water for a Healthy Country National Research
Flagship Report).
Sulleabhain, C.O., Gill, L.W., Misstear, B.D.R., Johnston, P.M., 2009.
Fate of endocrine-disrupting chemicals in percolating
domestic wastewater effluent. Water and Environment
Journal 23 (2), 110e118.
Tapsell, P., Newsome, D., Bastian, L., 2003. Origin of yellow sand
from Tamala limestone on the Swan Coastal Plain, Western
Australia. Australian Journal of Earth Sciences 50 (3), 331e342.
Tiehm, A., Schmidt, N., Stieber, M., Sacher, F., Wolf, L., Hoetzl, H.,
2011. Biodegradation of pharmaceutical compounds and their
occurrence in the Jordan Valley. Water Resources
Management 25 (4), 1195e1203.
Toze, S., Bekele, E., 2009. Determining Requirements for Managed
Aquifer Recharge in Western Australia. A Report to the Water
Foundation. CSIRO, Canberra, AU. http://www.csiro.au/files/
files/py7j.pdf.
Toze, S., Bekele, E., Page, D., Sidhu, J., Shackleton, M., 2010. Use of
static Quantitative Microbial Risk Assessment to determine
pathogen risks in an unconfined carbonate aquifer used for
Managed Aquifer Recharge. Water Research 44 (4), 1038e1049.
Toze, S., Hanna, J., 2002. The Survival Potential of Enteric
Pathogens in a Reclaimed Water ASR Project. In: Paper
5772 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 6 4 e5 7 7 2
Presented at Proceedings of the 4th International Symposium
on Artificial Recharge of Groundwater ISAR-4-Management of
Aquifer Recharge for Sustainability. Balkema Publishers,
Adelaide, SA, pp. 139e142.
Toze, S., Hanna, J., Smith, T., Edmonds, L., McCrow, A., 2004.
Determination of water quality improvements due to the
artificial recharge of treated effluent. In: Steenvoorden, J.,
Endreny, T. (Eds.), International Symposium on Wastewater
Re-Use and Groundwater Quality. Int Assoc Hydrological
Sciences, Wallingford, UK, pp. 53e60.
Turner, B., Binning, P., Stipp, S., 2005. Fluoride removal by calcite:
evidence for fluorite precipitation and surface adsorption.
Environmental Science & Technology 39 (24), 9561e9568.
Vanderzalm, J., Page, D., Barry, K., Dillon, P., 2010. A comparison
of the geochemical response to different managed aquifer
recharge operations for injection of urban stormwater in
a carbonate aquifer. Applied Geochemistry 25 (9), 1350e1360.
von Wandruszka, R., 2006. Phosphorus retention in
calcareous soils and the effect of organic matter on its
mobility. Geochemical Transactions 7 (6). doi:10.1186/1467-
4866-7-6.
Whelan, B.R., 1988. Disposal of septic tank effluent in calcareous
sand. Journal of Environmental Quality 17 (2), 272e277.
Whelan, B.R., Barrow, N.J., 1984. The movement of septic-tank
effluent through sandy soils near Perth .2. Movement of
phosphorus. Australian Journal of Soil Research 22 (3),
293e302.
Wolf, L., Held, I., Eiswirth, M., Hotzl, H., 2004. Impact of leaky
sewers on groundwater quality. Acta Hydrochimica Et
Hydrobiologica 32 (4e5), 361e373.
Zhang, J., Huang, X., Liu, C., Shi, H., Hu, H., 2005. Nitrogen removal
enhanced by intermittent operation in a subsurface
wastewater infiltration system. Ecological Engineering 25 (4),
419e428.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Removal of mercury (II) by dithiocarbamate surface

functionalized magnetite particles: Application to synthetic
and natural spiked waters

P. Figueira a, C.B. Lopes b,*, A.L. Daniel-da-Silva a, E. Pereira b, A.C. Duarte b, T. Trindade a
a
CICECO & Department of Chemistry, University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal
b
CESAM & Department of Chemistry, University of Aveiro, Campus de Santiago, 3810-193 Aveiro, Portugal

article info abstract

Article history: In order to take advantage of the high affinity between mercury and sulphur, magnetite
Received 24 March 2011 (Fe3O4) particles functionalized with dithiocarbamate groups (CS

2 ), were synthesized to be
Received in revised form used as a new type of sorbent to remove Hg (II) from synthetic and natural spiked waters.
22 July 2011 The effectiveness of this type of sorbent was studied, and its potential as cleanup agent for
Accepted 27 August 2011 contaminated waters was assessed.
Available online 3 September 2011 Batch stirred tank experiments were carried out by contacting a volume of solution with
known amounts of functionalized Fe3O4 particles, in order to study the effect of sorbent
Keywords: dose, salinity, and the kinetics and the equilibrium of this unit operation. A complete Hg (II)
Mercury removal (ca. 99.8%) was attained with 6 mg/L of magnetic particles for an initial metal
Magnetite particles concentration of 50 mg/L. It was confirmed that highly complex matrices, such as seawater
Dithiocarbamate functionalization (ca. 99%) and river water (ca. 97%), do not affect the removal capacity of the functionalized
Isotherms magnetic particles. Concerning isotherms, no significant differences were observed
Water remediation between two- and three-parameter models (P ¼ 0.05%); however, Sips isotherm provided
the lowest values of SS and Sx/y, predicting a maximum sorption capacity of 206 mg/g, in
the range of experimental conditions under study. The solid loadings measured in this
essay surmount the majority of the values found in literature for other type of sorbents.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction
The increasing awareness for the effects of pollution, has led
to strict environmental regulations, such as the creation of
a list of 33 priority hazardous substances by European
Parliament and the Council of the European Union (EU 2008),
and the instigation of the Member States to implement the
necessary actions aiming at a progressive reduction of
pollutants, and ceasing or phasing out emissions and
discharges of priority hazardous substances, as considered in
the Framework Directive (EU 2000).
Mercury and its compounds are one of the most dangerous
contaminants in the environment, threatening the human
health and natural ecosystems. They are included in the list of
priority hazardous substances and consequently, the removal
of Hg and its compounds, particularly, from aquatic systems is
a major goal of wastewater treatment and cleanup technolo-
gies. Conventional techniques for Hg removal from aqueous
solutions include sulphate or hydrazine precipitation, ion-
exchange, liquideliquid extraction, adsorption and solid
phase extraction via activated carbon adsorption (Starvin and
Rao, 2004).

* Corresponding author. Tel.: þ351 234 370 721; fax: þ351 234 370 084.
E-mail address: claudia.b.lopes@ua.pt (C.B. Lopes).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.057
5774 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4
Nomenclature
C Concentration of Hg (II) in bulk solution, mg/L
V Volume of solution, L
M Dry weight of Fe3O4/SiO2/NH/CS
2 particles, mg
q Amount of Hg (II) in the particles, mg/g
k1 First-order rate constant, h
1
k2 Second-order rate constant, g/mg h
kL Langmuir constant, L/g
qm Maximum loading of Fe3O4/SiO2/NH/CS
2 particles,
mg/g
kF Freundlich parameter, (mg1
1/n$L1/n/g)
n Freundlich parameter
Nt Total number of binding sites, mg/g
a Sips constant
m Heterogeneous index
kRP Redlich-Peterson constant, L/g
The high toxicity of Hg to organisms is generally attributed
to the blockage of the enzyme binding sites and interference
in protein synthesis (Starvin and Rao, 2004), due to the high
affinity of Hg to bind with any molecule that has sulphur or
a sulphurehydrogen combination in its structure. A prom-
ising strategy to achieve high-performance materials able to
capture Hg (II) from aqueous solutions is to take advantage of
its high affinity to sulphur, developing new materials with
sulphur-containing functional groups such as diverse thio-
lates. This approach has been investigated by Antochshuk
et al. (2003) and materials with high surface area have been
used as an insoluble matrix for the attachment of sulphur-
containing groups (Antochshuk et al., 2003). Several mate-
rials such as mesoporous silica (Antochshuk et al., 2003;
Venkatesan et al., 2003; Mattigod et al., 1999; Mercier and
Pinnavaia, 1998; Feng et al., 1997; Mureseanu et al., 2010),
silica gel (Venkatesan et al., 2002), activated carbon (Starvin
and Rao, 2004) and organoceramic composites (Nam et al.,
2003) have been used as matrices for the attachment of
sulphur-containing groups, such as thiol (Mattigod et al., 1999;
Mercier and Pinnavaia, 1998; Nam et al., 2003), dithiocarba-
mate (Venkatesan et al., 2003, 2002), 1-(2-thiazolylazo)-2-
naphthol (Starvin and Rao, 2004), mercaptopropylsilane
(Feng et al., 1997), 1-furoyl thiourea urea (Mureseanu et al.,
2010) and benzoythiourea (Antochshuk et al., 2003). Those
materials have surface coverage of ligands typically between
2.5  10
4 and 3.7  10
3 mol/g, and some of them exhibit very
high Hg (II) loading capacities (Antochshuk et al., 2003).
We have been interested on magnetic particles, in partic-
ular iron oxides (e.g. Fe3O4), to develop new materials for
environmental and bio-applications (Daniel-da-Silva et al.,
2007). Iron oxides such as maghemite and magnetite offer
convenient magnetic properties, low toxicity and price, high
surface to volume ratios, and possibility for surface chemical
modification (Girginova et al., 2010). Keeping in mind the
interesting properties of magnetic magnetite and the higher
values of the stability constants reported for Hg-
dithiocarbamate (Dtc) complexes (e.g. 1.2  1038 for
Hg(Et2Dtc)2 (Venkatesan et al., 2002)), we have recently re-
ported the synthesis and the preliminary application of silica
aRP Redlich-Peterson constant, (L/mg)b
b Redlich-Peterson exponent
RL Separation factor
ARE Average relative error
SS Sum of squares
CV-AFS Cold vapour atomic fluorescence spectroscopy
DF Degrees of freedom
RE Relative Error
RSD Relative Standard Deviation
Sx/y Standard Deviation of Residues
Subscripts
0 initial condition of experiment
t intermediate condition of the experiment at
a certain time
e equilibrium condition of experiment
coated Fe3O4 particles functionalized with Dtc groups for the
decontamination of synthetic waters with realistic Hg (II)
levels (Girginova et al., 2010). However, this work also raised
a number of important issues related with the performance of
these materials as colloidal adsorbents in natural waters,
particularly in strong salinity conditions.
We wish to report here our research on the removal of Hg
(II) at trace levels found in natural waters, and the effect of
sorbent dose and salinity on the sorption efficiency. The
results of this study on the sorption equilibrium and kinetics
of Hg (II) onto the functionalized magnetic particles (MPs)
were fitted to well-known kinetic and equilibrium equations.
For the assessment of the real effectiveness of the developed
sorbent, the functionalized magnetic NPs were applied in two
natural waters (seawater and river water).
2. Material and methods
2.1. Chemicals
All chemicals used in this work were of analytical reagent
grade, and obtained from commercial chemical suppliers and
were used without further purification. The certified standard
stock solution of mercury (II) nitrate was purchased from
Merck (1000  2 mg/L).
2.2. Adsorbent material
2.2.1. Synthesis
Fe3O4/SiO2/NH/CS
2 magnetic particles were investigated for
remediation of Hg (II) contaminated waters. The synthesis of
these magnetic particles includes three distinct steps: (i) the
synthesis of magnetite particles (Fe3O4), and their encapsu-
lation in amorphous silica shells (Fe3O4/SiO2), (ii) further
modification with 3-aminopropyltriethoxysilane (Fe3O4/SiO2/
NH2) and (iii) the grafting of dithiocarbamate groups at the
surface of amine modified silica coated magnetite (Fe3O4/SiO2/
NH/CS
2 ).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4
The first two steps of the synthesis were performed exactly
as described by Girginova et al. (Girginova et al., 2010), while
the procedure adopted for the grafting of dithiocarbamates
groups was slightly different in order to graft a higher amount
of particles. This procedure was performed as follows: the
amine modified silica coated magnetite (50 mg) was added to
15 mL 2-propanol under mechanic stirring for 1 h. After,
0.75 mL 1 M NaOH, and 0.12 mL CS2 were added and the
suspension was mechanically stirred for 4 h. The powder was
then collected magnetically from the suspension formed,
washed with 2-propanol and dried at room temperature.
2.2.2. Structural and chemical characterization
Fourier Transform Infrared (FT-IR) spectra of Fe3O4/SiO2/NH/
CS
2 magnetic particles were recorded using a spectrometer
Mattson 7000 with 256 scans and 4 cm
1 resolution, using
a horizontal attenuated total reflectance (ATR) cell. Elemental
analysis for carbon, nitrogen, hydrogen and sulphur were
performed using a LECO CHNS-932 elemental analyzer. The
crystalline phase of the particles was identified by X-ray
powder diffraction of the dried samples using a Philips X’Pert
X-ray diffractometer equipped with a Cu Ka monochromatic
radiation source. Transmission electron microscopy (TEM)
was performed using a transmission electron microscope JEOL
200CX operating at 300 kV. The specific surface area of the
magnetic particles was determined with nitrogen adsorption
BET measurements performed with a Gemini Micromeritics
instrument.
2.3. Batch adsorption experiments
The ability of Fe3O4/SiO2/NH/CS
2 particles to capture Hg (II)
from water was evaluated by contacting the particles with
a Hg (II) solution for a required period of time. The sorption
experiments were carried out in 500 mL batch reactors at
295  1 K, under mechanical stirring. Hg (II) solutions were
prepared daily by diluting the corresponding standard solu-
tion, in high purity water (18 MUcm), to the desired initial
concentration (50 mg/L), with a subsequent adjustment of pH
to 7 with 0.1 M NaOH. The initial concentration of 50 mg/L was
selected, as this is the current limit value for Hg discharges
from industrial sectors other than the chloro-alkali electrol-
ysis industry.
All glassware used in these experiments was acid-washed
prior to use with HNO3 25%, 12 h, and ultra-pure water.
Experiments were performed to evaluate the time profile
and the effects of sorbent concentration and salinity on the
sorption of Hg (II) onto the functionalized magnetic particles.
Accurately weighed amounts of Fe3O4/SiO2/NH/CS
2 particles
were added to Hg (II) solutions, in glass batch reactors, which
were immediately placed in an ultrasonic bath for ca. 10 s, for
dispersing the magnetic particles. This time was considered
the starting point for each experiment. For every experiment,
the Hg (II) solution was continuously stirred, using a glass rod,
and samples were withdrawn for analysis at several sampling
times. Each sample was analyzed for the Hg (II) concentration
after magnetic separation of the particles using a NdFeB
magnet (1.48 T), and posterior pH adjustment (lower than 2)
with concentrate HNO3 (Hg free, Merck). Mercury analyses
were performed by cold vapour atomic fluorescence
5775
spectroscopy (CV-AFS), on a flow-injection-cold vapour
atomic fluorescence spectrometer (Hydride/vapour generator
PS Analytical Model 10.003, coupled to a PS Analytical Model
10.023 Merlin atomic fluorescence spectrometer; PS Analyt-
ical, Orpington, Kent, England) and using SnCl2 as reducing
agent. An Hg (II) solution in the absence of magnetic particles
was run as a control experiment.
The time profile of Hg (II) sorption onto Fe3O4/SiO2/NH/CS
2
particles and the effect of sorbent dose, were determined by
using four different amounts of the functionalized particles,
namely 0.124, 0.256, 0.501 and 3.063 mg, which correspond to
a sorbent dose of 0.2, 0.5, 1 and 6 mg/L.
In order to study the effects of salinity changes on the
sorption capacity of Fe3O4/SiO2/NH/CS
2 , 6 mg/L of particles
were added to three different matrices spiked with 50 mg Hg
(II)/L: ultra-pure water, 3 g/L NaCl solution and seawater
collected from about 1 nautical mile from the Portuguese
coast. In the following sections, those matrices will be deno-
ted in terms of salinity (0% of salinity, 10% of salinity and 100%
of salinity, respectively).
The efficiency of Hg (II) removal by Fe3O4/SiO2/NH/CS
2
particles from different types of natural waters was also
assessed by adding 6 mg/L of particles to river water collected
from Vouga River (near the drinking water treatment plant)
and spiked with 50 mg Hg (II)/L.
Isotherms were obtained varying the amount of Fe3O4/
SiO2/NH/CS
2 particles from 0.124 to 3.063 mg, for an initial
Hg (II) concentration of 50 mg/L and a temperature of
295  1 K.
The amount of Hg (II) sorbed per unit of particles, at time t,
qt (mg/g) was estimated from the mass balance between
initial Hg (II) concentration and concentration at time t in
solution,
V
qt ¼ ðC0
Ct Þ  (1)
M
where V is the volume of solution (L) and M is the dry weight of
Fe3O4/SiO2/NH/CS
2 particles (mg), C0 (mg/L) is the initial Hg (II)
concentration and Ct (mg/L) is its concentration at time t.
The results were also compared by removal percentage,
which at equilibrium time is defined by:
ðC0
Ce Þ
Removal% ¼  100 (2)
C0
where, Ce (mg/L) is the equilibrium Hg (II) concentration in the
solution.
2.4. Kinetic and equilibrium models
Different theoretical models were applied to experimental
data in order to find a model which adequately describes
kinetic and equilibrium data.
2.4.1. Kinetic models
The chemical kinetics gives relevant information about reac-
tion pathways and rates, along with time to reach equilibrium.
Moreover, the physical and chemical features of the sorbent
have great impact on both sorption kinetics and sorption
mechanism (Hasan and Srivastava, 2009).
5776 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4
In order to evaluate the differences in the kinetic rates and
to describe the kinetic removal of Hg (II) ions onto Fe3O4/SiO2/
NH/CS
2 particles in three matrices tested (ultra-pure water,
NaCl solution and seawater), two simple and widely applied
models, the pseudo-first- and pseudo-second-order models
were used to describe the removal process. Numerous studies
in literature have pointed out that one of those two models is
able to describe the majority of the sorption studies (Wang
and Chen, 2009; Ho and McKay, 1999).
2.4.1.1. Pseudo-first-order equation or Lagergren model. The
first-order equation, firstly applied by Lagergren (Lagergren,
1898), is mathematically expressed by:
dqt

¼ k1 qe
qt
dt
where k1 (h
1) is the rate constant of pseudo-first-order and qe
(mg/g) is the amount of solute sorbed per gram of sorbent at
equilibrium. After integration and application of the boundary
condition qt¼0 at t¼0, Equation (3) becomes


qt ¼ qe 1
e
k1 t
2.4.1.2. Pseudo-second-order equation. The pseudo-second-
order equation may be also applicable and, in contrast with
the previous model, usually correlates the behaviour over the
whole range of sorption (Ho and McKay, 1999). The kinetic rate
equation is expressed as:
dqt
2
¼ k2 qe
qt
dt
where k2 (g/mg h) is the rate constant of pseudo-second-order.
By applying the boundary conditions t¼0 to t¼t and qt¼0 to
qt¼qe, the integrated form of Eq. (5) is:
k2 q2e t
qt ¼
1 þ k2 qe t
2.4.2. Equilibrium models
Sorption equilibrium provides fundamental data for evalu-
ating the applicability of the sorption process. Equilibrium
isotherms give the equilibrium relationships between sorbent
and sorbate, i.e. the quantity of solute sorbed and remaining in
solution at a given temperature, at equilibrium. The equation
parameters and the underlying thermodynamic assumptions
of the equilibrium models often provide some insight into the
sorption mechanism, the surface properties and the capacity
and affinity of the sorbent (Ho et al., 2002). In this study, two-
parameter isotherms (Langmuir and Freundlich) and three-
parameter isotherms (Sips and Redlich-Peterson), in their
nonlinear form were chosen to fit the experimental data.
2.4.2.1. Freundlich isotherm. This empirical model developed
by Freundlich in 1906 (Freundlich, 1906), can be applied to
multilayer sorption as well as non ideal sorption on hetero-
geneous surfaces and is represented by the following
equation:
qe ¼ kF Ce1=n
where kF (mg(1
1/n) L(1/n)/g) and n are the Freundlich parame-
ters. n is usually between 1 and 10, which points out
favourable adsorption, and is related to the non-linearity of
the model.
2.4.2.2. Langmuir isotherm. This two-parameter model
developed by Langmuir in 1916, relating the amount of gas
sorbed on a surface to the pressure of the gas is probably one
of the best known and widely used sorption isotherm (Ho
et al., 2002). This model assumes that exist a fixed number
of accessible sites available on the sorbent surface, all of
them with the same energy and that the sorption is revers-
ible. The Langmuir model is represented by the following
equation:
qm kL Ce
qe ¼ (8)
1 þ kL Ce
(3)
where kL (L/mg) and qm (mg/g) are the Langmuir sorption
equilibrium constant related to the energy of sorption and the
maximum sorption capacity corresponding to complete
monolayer coverage, respectively.
2.4.2.3. Sips or Langmuir-Freundlich isotherm. As the own
(4)
name indicates, this three-parameter isotherm is a composite
of the former isotherms. At high sorbate concentrations, it
predicts a monolayer sorption capacity, characteristic of the
Langmuir isotherm and at low sorbate concentrations it
reduces to a Freundlich isotherm (Ho et al., 2002). The Sips
equation can be describe as follows:
Nt aCm
qe ¼ e
(9)
(5) 1 þ aCm e
where Nt (mg/g) is the total number of binding sites, a is
related to the median binding affinity (k), since a¼km and m is
the heterogeneous index, which varies from 1 (homogeneous
material) to 0 (heterogeneous material for m < 1) (Umpleby
et al., 2001).
(6)
2.4.2.4. Redlich-Peterson isotherm. This three-parameter
isotherm, which also incorporates characteristics of both the
Langmuir and Freundlich isotherms, can be represented as
follows:
kRP Ce
qe ¼ (10)
1 þ aRP Cbe
where kRP (L/g) and aRP ((L/mg)b) are the Redlich-Peterson
constants and b is the Redlich-Peterson exponent.
2.4.3. Error analysis
The parameters of the kinetic and equilibrium models here
considered were obtained by nonlinear regression analysis
using GraphPad Prism 5 program, which uses the least-squares
as fitting method and the method of Marquardt and Levenberg,
which blends two other methods, the method of linear descent
and the method of Gauss-Newton for adjusting the variables.
In order to confirm which model presents the best fit to
experimental data, the adjusted R squared (R0 2), the sum of
squares (SS) and the standard deviation of residues (Sx/y) were
analyzed. The standard error of the best fit parameters, the
(7)
average relative error (ARE) of the fittings and the relative error
(RE) between the experimental qe value and the models’ esti-
mation value are also presented. The different statistical
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4 5777

parameters were determined using the following mathemat-

ical expressions:

 n
1
R02 ¼ 1
1
R2  (11)
n
p
1

(311)
3000
P
^ 2
yi
y i

(440)
R2 ¼ 1
P
2 (12)

(511)
yi
yi

Counts (a.u.)

(400)
(220)
2000

(422)
(111)
X
^ 2
SS ¼ yi
y i (13)
1000

(P
^ 2 )1=
yi
y i 2
Sx=y ¼ (14)
n
2 0
20 30 40 50 60 70
2 θ (°)
 ^ 
P jyi
y i j
yi Fig. 1 e X-ray powder diffraction patterns of Fe3O4 and TEM
ARE ¼  100 (15) image of silica coated Fe3O4 particles. The Miller indices
n
corresponding to the most intense reflection peaks of
^ Fe3O4 are indicated in brackets.
jyi
y i j
RE
 100 (16)
yi

where n is the sample size, p is the number of adjustable
parameters in the model and yi are the experimental or
observed values, yi is the mean of the observed data and y^i are
the modelled or predicted values. The absolute error for each
experimental point from the kinetics and equilibrium exper-
iments can be found in the complementary file.
groups on the sorbent was found to be 1.53  10
4 mol/g or
1.05  10
5 mol/m2. This value is slightly lower but of the same
order of magnitude than the density of Dtc groups found for
other sorbents like mesoporous silica grafted with Dtc
(Venkatesan et al., 2003) (2.5  10
4 mol/g) and silica gel
grafted with Dtc (Venkatesan et al., 2002) (3.7  10
4 mol/g).

3.2. Hg (II) removal by Fe3O4/SiO2/NH/CS

2 particles
3. Results and discussion
3.2.1. Time profile of Hg (II) removal
3.1. Characterisation of Fe3O4/SiO2/NH/CS

2 particles Kinetic curves corresponding to the sorption of Hg (II) onto
different amounts of Fe3O4/SiO2/NH/CS
2 particles are shown in
In the present work, surface modified Fe3O4 particles were Fig. 2. The plots represent the normalized Hg (II) concentration
firstly prepared using methodologies previously described by that remained in the liquid phase vs. time in Fig. 2A and the Hg
us (Girginova et al., 2010). Fig. 1 and Table 1 summarizes (II) concentration in the MPs vs. time in Fig. 2B, both for an
relevant results obtained for the characterization of these initial concentration of 50 mg Hg (II)/L. For all the amounts of
materials and that have been collected here for monitoring Fe3O4/SiO2/NH/CS
2 particles, a decrease with time on Hg (II)
the materials properties. In brief, the powder X-ray diffraction concentration in the liquid phase was observed, even when
patterns of the materials match those of magnetite,1 whose only 0.2 mg/L were used. The time profile curves show that
identity was unequivocally proved by Mössbauer spectros- equilibrium time is attained in 24 h for the highest amount of
copy (Girginova et al., 2010). As expected, TEM analysis (Fig. 1) particles used (6 mg/L), and as expected, it increases as long as
showed single cubic particles of magnetite coated with the amount of particles decreases. This increase is particularly
amorphous silica shells. As previously reported, the cubic notorious by comparing the equilibrium time for the lowest
particles exhibited a magnetization hysteresis loop at room and the highest amount of Fe3O4/SiO2/NH/CS
2 ; for the lowest
temperature with a saturation magnetization of 62 emu/g
(Girginova et al., 2010). Finally, the surface functionalization of
the silica coated magnetite particles was monitored step by Table 1 e Assignment of infrared diagnosis bands for
step using infrared spectroscopy and the results are summa- Fe3O4 and surface modified Fe3O4 particles.
rized in Table 1. Sample Wave number (cm
1) Assignment
Due to the higher value of the stability constant for the Hg-
Fe3O4 540, 310 (s) n(FeeO)
Dtc complex is expectable that Hg (II) loading onto the func-
Fe3O4/SiO2 1070 (vs) n(SiOeSi)
tionalized particles will depend of the extent of functionali- 789 (w) n(SieOH)
zation carried out on the sorbent. Based on surface area 943 (w) n(SieOeFe)
(14.6 m2/g), sulphur content (0.98%) and taking into account Fe3O4/SiO2/NH2 786 (w) d(NeH)
that each Dtc group has two sulphur atoms, the amount of Dtc Fe3O4/SiO2/NH/CS

2 1438 (s) n(CeN)

1
(vs-very strong, s-strong, w-weak; n-stretching vibration, d-bending
Joint Committee for Powder Diffraction Studies, JCPDS vibration).
19-0629.
5778 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

amount of MPs (0.2 mg/L) the equilibrium time was achieved
only after 96 h, while for the highest amount of MPs (6 mg/L) the
equilibrium was achieved in less than 24 h (Fig. 2). This
difference suggests that equilibrium time can be drastically
affected by small variations in sorbent amount, and it is mainly
related with the large availability of the Dtc groups. The time
profile curves also reveal that during the first 12 h, the Hg (II)
content in the particles increased quickly and then slowed
down approaching equilibrium (Fig. 2B).
3.2.2. The effect of Fe3O4/SiO2/NH/CS

2 concentration on Hg
(II) removal
The concentration of sorbent is an important parameter to
obtain quantitative metal removal, since for a given initial
concentration of metal, it influences both the contact time
necessary to reach equilibrium and the sorption capacity
(Lopes et al., 2009).
Fig. 3 shows the equilibrium removal percentage and the
equilibrium amount of Hg (II) sorbed by gram of Fe3O4/SiO2/
NH/CS


2 vs. concentration of Fe3O4/SiO2/NH/CS2 . Given the
obtained results it can be concluded that the particles
concentration strongly affects Hg (II) removal and Hg (II)
uptake per gram of particles. It is noticeable, that at equilib-
rium, the removal percentage increases with the increasing of
A 1.0
6 mg/L
1 mg/L
0.8 0.5 mg/L
0.2 mg/L
0.6
Ct/C0
Fe3O4/SiO2/NH/CS
2 concentration, from 53.2  1.3% (for
0.2 mg/L of particles) to 99.8  1.5% (for 6 mg/L of particles)
while the amount of Hg (II) sorbed per gram of Fe3O4/SiO2/NH/
CS
2 particles decreases drastically from112  3 mg/g to
9.2  0.1 mg/g. Thus, for an invariable initial Hg (II) concen-
tration, the increase of the sorbent concentration provides
greater contact surface area and increases the number of
available sorption sites, promoting the removal of Hg (II). As
the number of sorption sites increases, the concentration of
Hg (II) in the liquid phase and the amount of Hg (II) per mass of
sorbent decreases and more sorption sites remain unsatu-
rated during the sorption process. Additionally, the increase
of the amount of Hg (II) per mass of MPs will only occur as long
as the maximum capacity of the MPs is not fulfilled. This fact,
suggest that under the experimental conditions tested, the
maximum capacity of the magnetic particles was not
achieved.
Another remarkable result was the achievement of
a considerably low residual concentration (0.10  0.02 mg Hg
(II)/L) with only the application of a few milligrams of sorbent
(6 mg/L). The residual Hg (II) concentration achieved is ten
times lower than the current guideline value of the European
Union for water of drinking quality ([Hg (II)]  1 mg/L) (EU 1998).
This result clearly evidence the huge potential of this
magnetic material to decontaminate waters and to achieve
totally Hg free effluents, as is required by the Water Frame-
work Directive (EU 2000).
Based on the obtained results, 6 mg of Fe3O4/SiO2/NH/CS
2
particles per litre of Hg (II) solution (50 mg/L) should be used for
a completely effectiveness of the decontamination process,
since any further increase in the particles dose will not

significantly affect the equilibrium between the ions sorbed in

the solid phase and those remaining in solution.
0.4
3.2.3. The effect of salinity on Hg (II) removal
It is well know that in the presence of chloride ions, Hg forms
0.2
chloro-complexes of Hg (II) and consequently some sorbents
become useless in saline waters (Lopes et al., 2007). The
removal of Hg (II) by Fe3O4/SiO2/NH/CS
2 particles was studied
0.0
in the presence of two different Cl
concentrations (ca. 10 and
120
B 120 120

Removal.. (%) qe (mg/L)

100
100 100

80
80 80
qt , mg/g

60
60 60

40
40 40

20
20 20

0
0 24 48 72 96 120 0 0
t, h 0.2 0.5 1.0 6.0 0.2 0.5 1.0 6.0
Fe 3O4/SiO2/NH/CS2- concentration (mg/L)
Fig. 2 e Normalized Hg (II) concentration in the liquid
phase (A) and sorbed on the particles (B) as a function of Fig. 3 e Effect of Fe3O4/SiO2/NH/CSL
2 concentration on Hg (II)
time, for different amounts of Fe3O4/SiO2/NH/CSL2 particles. removal (%) and sorption capacity (mg/g).
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4 5779

1.0 values of the amount of Hg (II) sorbed per gram of Fe3O4/SiO2/

A NH/CS
2 particles (qe values ranged from 9.08 to 9.35 mg/g, with
0% a RSD ca. 1.5%), neither the equilibrium removal percentage,
0.8 10% which was higher than 98% for all matrices (0, 10 and 100% of
100% salinity). Even the residual concentration of Hg (II) in solution
was lower than the current guideline value of the European
0.6 Union for water of drinking quality (1 mg/L), for all three
Ct /C0

systems. For 0% of salinity, i.e. for ultra-pure water the

residual concentration achieved was 0.10  0.02 mg Hg (II)/L,
0.4 for 10% of salinity it was 0.97  0.01 mg Hg (II)/L and for 100% of
salinity, i.e. for seawater the residual concentration achieved
was 0.82  0.02 mg Hg (II)/L. Conversely, the presence of high
0.2
Cl
concentrations clearly increased the equilibrium time
from 24h (0% of salinity) to ca. 96h (10 and 100% of salinity) (see
Fig. 4) and the rate which Hg (II) is removed by the particles
0.0
decreases with the increasing of the matrix’s complexity, i.e.
10 from ultra-pure water to seawater.
B
These results suggest that besides, the equilibrium time,
the matrix of the system Hg (II)/Fe3O4/SiO2/NH/CS
2 particles
8 also influences the kinetic rate of the sorption process, as
confirmed by the rate constants (k1 and k2) of the pseudo-first-
and pseudo-second-order kinetic models (see Table 2). The
6 kinetic sorption rate constants k1 and k2 decrease with the
qt , mg/g

increasing of the salinity of the matrix, which reinforces that

the higher is the ionic strength, the slower equilibrium is
4
reached. The obtained results also suggest that during the first
hours not all Hg (II) is available for the active sites of the Fe3O4/
SiO2/NH/CS
2 particles, probably due to the formation of
2
mercury chloro-complexes. However, as the equilibrium
values of the amount of Hg (II) removed per gram of func-
0 tionalized particles (qe) did not change significantly for the
0 24 48 72 96 120 different matrices, it is expectable that as the concentration of
t, h the Hg (II) available in solution decreases the decomplexation
of the chloro-complexes occurs, since they are less stable (Ks
Fig. 4 e Normalized Hg (II) concentration in the liquid
107-1015) (Nam et al., 2003) than the complexes formed
phase (A) and sorbed on the particles (B) as a function of
between mercury and the Dtc groups (Ks 1038) (Venkatesan
time, for different percentages of salinity.
et al., 2002), increasing the concentration of Hg (II) available
in solution, for complexing with the Dtc groups present at the
100% of salinity) and the results, which are shown in Fig. 4, surface of the magnetic particles. The results also allow to
were also compared with those obtained for ultra-pure water conclude that Naþ do not compete with Hg (II) for the active
(0% of salinity). The plots in Fig. 4A represent the normalized sites of the Fe3O4/SiO2/NH/CS
2 particles, even when the Na
þ

Hg (II) concentration remaining in the liquid phase vs. time,
while the Hg (II) concentration in the MPs vs. time is repre-
sented in Fig. 4B.
According with the results, the presence of high Cl

concentrations did not change significantly the equilibrium
concentration was much higher than that of Hg (II).
Furthermore, the modelling of the kinetic process by the
pseudo-first- and pseudo-second-order models allows to
conclude that whatever the kinetic equation used, the
description of the sorption kinetics was satisfactory for all

Table 2 e First- and second-order sorption rate constants obtained for the removal of Hg (II) from matrices with different
percentage of seawater, together with experimental and fitted qe, and the goodness of the of the fittings.
Kinetic model Model’s parameters Goodness of the fit Experimental values

Pseudo - 1st order k1 h
1 SE qe mg/g SE R0 2 SS Sx/y ARE % DF qe mg/g RE %

0% 0.397 0.037 9.10 0.19 0.98 1.53 0.44 4.4 8 9.22 1.3
10% 0.183 0.016 8.89 0.22 0.98 1.69 0.43 10 9 9.08 2.1
100% 0.154 0.028 8.78 0.45 0.92 6.85 0.87 18 9 9.35 6.1
Pseudo - 2nd order k2 g/mg h SE qe mg/L SE R2 SS Sx/y ARE % DF qe mg/g RE %
0% 0.0667 0.0077 9.61 0.18 0.99 0.90 0.33 3.7 8 9.22 2.6
10% 0.0259 0.0032 9.67 0.24 0.99 1.29 0.38 7.2 9 9.08 2.3
100% 0.0224 0.0042 9.59 0.37 0.97 2.71 0.55 12 9 9.35 2.2
5780 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

matrices, as confirmed by the high values of R0 2 (0.92e0.99) The Langmuir equation assumes that adsorption occurs at
and low values of Sx/y (0.33e0.87) (see Table 2). It is clear that definite localized sites on the surface, each site being able to
both kinetic models describe very well the quantity of Hg (II) bind a single molecule of the adsorbing species. The energy of
removed in the early stages of sorption, although there are adsorption is equal for all sites and there are no interaction
slightly deviations between the experimental and the fitted forces between adjacently adsorbed molecules (Cooney, 1999).
data, particularly in the inflexion zone (Fig. 5). However, both Theoretically, a saturation value is reached, beyond which no
models are able to accurately estimate the qe value of the three further sorption can take place, which is represented by
systems, which is corroborated by the low values of RE a plateau in the equilibrium isotherm and corresponds to the
(1.3e6.1%) found between the predicted and the experimental assumption of one complete monomolecular layer of coverage
qe values in all matrices. Still, comparing the predicted and
experimental qe values it is perceptible that the Langergren
model underestimates the qe values while the pseudo-second-
order model overestimates them. Although these slight
differences, for a confidence level of 95% there is no significant
difference between the goodness of the fit of the two models
(F-test), for all tested matrices.

3.3. Sorption equilibrium

An accurate mathematical description of the equilibrium data

between the concentration of the sorbate in the liquid and the
amount in the solid phase is essential for a consistent prediction
of the sorption parameters and for quantitative comparison of
the sorption capacity of different sorbents. This mathematical
function, called isotherm, is a basic requirement for designing
any sorption system (Marin et al., 2009), and is obtained for
a specific temperature and initial sorbate concentration.
Fig. 6 shows the sorption isotherm of Hg (II) onto Fe3O4/
SiO2/NH/CS
2 particles, as well as the fit to the isotherm
models described in the experimental section. The parame-
ters of the isotherm models obtained from the corresponding
fittings are presented in Table 3.
All isotherms are positive and concave to the concentra-
tion axis, and under the experimental conditions here used,
the experimental equilibrium values of the amount of Hg (II)
sorbed in the functionalized particles increases with the
increasing of Hg (II) concentration in solution, without
reaching a saturation plateau.
The Freundlich isotherm is an empirical equation, which
does not assume that the material coverage must approach
a constant value corresponding to one complete solute
monomolecular layer as Ce gets larger. This model predicts
that qe monotonously increases with increasing Ce which,
being physically impossible, means that the Freundlich
equation should fail to describe the experimental data at high
Ce values (Cooney, 1999). However, the concentration values in
real sorption processes are considered sufficiently diluted, in
order to avoid the process entering the region where the
Freundlich equation breaks down (Cooney, 1999). According
with the obtained results, the Freundlich model provides
a good description of the experimental data (ARE¼5.8%,
R0 2¼0.98), since in the range of experimental conditions used,
the equilibrium data do not achieved a plateau at a limiting
value of Ce, suggesting the existence of heterogeneous surface
conditions. Freundlich isotherm allows to calculate two
empirical constants, kF and n, which are related to adsorption 0
capacity of the sorbent and sorption intensity, respectively.
,
The magnitudes of kF (601 mg(1
1/n) L(1/n)/g) and n (2.26) indi-
cate easy separation of Hg (II) from liquid phase and favour- Fig. 5 e Sorption kinetics modelling of Hg (II) on the
able sorption (1 < n < 10). particles, for different percentages of seawater.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4 5781

120 monolayer capacity (qm) of Fe3O4/SiO2/NH/CS
2 particles esti-

Two parameter models mated by the Langmuir model is 142 mg/g, with a 95% confi-
dence interval of [115e170 mg/g], and the Langmuir constant
100
(kL), which represents the affinity between the sorbent and
sorbate is 145 L/mg.
80 Moreover, the isotherms can be classified as irreversible
when the separation factor (RL) is 0; favourable when
qe, mg/g

0 < RL¼<1, linear when RL ¼ 1 and unfavourable when RL > 1.

60
The separation factor (RL) can be calculated using the
follow equation:
40
1
RL ¼
Langmuir 1 þ kL C0
20 Freundlich
The calculated RL value was 0.12, corroborating that Hg (II)
sorption on Fe3O4/SiO2/NH/CS
2 particles is favourable.
0 The Langmuir-Freundlich isotherm, also known as Sips
0.000 0.005 0.010 0.015 0.020 0.025
isotherm, is an equation with three fitting coefficients with
physical meaning (Nt, a and m), that describes the relationship
120
of the equilibrium concentration of the sorbate between the
Three parameter models
solid and liquid phase in heterogeneous systems (Umpleby
100 et al., 2001). As the name implies, this isotherm is a combina-
tion of the Langmuir and Freundlich isotherms. At low sorbate
concentrations it can be reduced to the Freundlich isotherm,
80
while at high sorbate concentrations, it predicts a monolayer
qe, mg/g

sorption capacity characteristic of the Langmuir isotherm (Ho

60 et al., 2002; Marin et al., 2009). This isotherm is capable of
modelling both homogeneous and heterogeneous binding
40 surfaces. The value of exponent m on Sips equation was 0.705,
which means that Hg (II) sorption onto the functionalized
particles is more of Langmuir type than that of Freundlich,
20 Sips
since for m¼1 the Sips equation (Eq. (9)) reduces to the Lang-
Redlich-Peterson
muir equation (Eq. (8)) in which the variable a corresponds
0 directly to binding affinity (k). The predicted value of Nt was
0.000 0.005 0.010 0.015 0.020 0.025 higher than the corresponding value (qm) of the Langmuir
Ce , mg/L model (Table 3).
Redlich-Peterson isotherm also incorporate features of
Fig. 6 e Equilibrium isotherms of Hg (II) on the particles at both Langmuir and Freundlich equations, approximating to
21 ± 1  C. Henry’s law at low concentrations (b¼0), while at higher
concentrations its behaviour approaches that of the
Freundlich isotherm (Ho et al., 2002). For b¼1 the Redlich-
Peterson isotherm reduces to the Langmuir form. In this
of the adsorbing species on the adsorbent (Cooney, 1999; study, the value of the exponent b approximates to 1 (0.860),
Kocaoba, 2007). Langmuir plot shows a higher difference suggesting like the Sips equation, that the equilibrium data
between experimental equilibrium data and the predicted can preferably be fitted by the Langmuir model rather the
ones (ARE¼14%), although no significant differences were Freundlich.
observed between the goodness of the fit of the Freundlich Among the three-parameter models, the Sips isotherm
and Langmuir models (F-test for 95% confidence level). The provides the highest value of R0 2 and the lowest values of ARE,

Table 3 e Isotherm constants of two- and three-parameter models for Hg (II) sorption on magnetic particles at 21 ± 1  C.
Isotherm Model’s parameters Goodness of the fit

Freundlich kF mg(1
1/n)L(1/n)/g SE n SE R0 2i SS Sx/y ARE % DF

601 109 2.26 0.21 0.98 148 5.43 5.8 5
Langmuir kL L/mg SE qm mg/g SE R2 SS Sx/y ARE % DF
145 29 142 11 0.98 130 5.11 14 5
Sips a L/g SE Nt mg/g SE m SE R2 SS Sx/y ARE % DF
16.9 30.7 206 101 0.705 0.211 0.98 102 4.51 10 4
Redlich-Peterson kRP L/g SE aRP (L/mg)b SE b SE R2 SS Sx/y ARE % DF
26188 13485 114 44 0.860 0.227 0.98 122 4.93 14 4
5782 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4

SS and Sx/y, suggesting that this model is more appropriate to
describe the experimental equilibrium data than the Redlich-
Peterson. Among the two-parameter models, the Langmuir is
in agreement with the experimental data and with the vari-
ables obtained from the three-parameter models and
provides higher value of R0 2 and lower values of SS and Sx/y
than the Freundlich. However, the comparison (F-test)
between the values of Sx/y obtained for all models indicates
that for 95% confidence level there are no significant differ-
ences between the goodness of the fit of the different models,
in the range of experimental conditions studied (Ce
[0e0.025 mg/L]).
3.4. Application of Fe3O4/SiO2/NH/CS

2 particles in
natural waters
Besides the application of Fe3O4/SiO2/NH/CS
2 particles to
seawater, already discussed in a former section, the feasibility
of the functionalized particles were also tested in river water.
Likewise for the seawater, the time profile curve (plot not
shown) reveals a decrease on Hg (II) concentration with time.
However, probably due to a higher complexity of the matrix
and higher levels of organic matter, the equilibrium time in
the river water was attained in 240 h against the 96 h neces-
sary to achieve equilibrium in seawater. The percentage of Hg
(II) removal obtained was ca. 97% for river water and ca. 99%
for seawater, while the residual Hg (II) concentration in the
liquid phase was, respectively, 1.20  0.07 and 0.82  0.02 mg
Hg (II)/L. Those results suggest that in the case of river water,
the amount of particles employed should be slightly higher
than 6 mg/L in order to achieve a completely effectiveness of
the decontamination process in this type of natural waters;
however, it must be highlight the high effectiveness of Fe3O4/
SiO2/NH/CS
2 particles to remove Hg (II) from water, even from
natural waters, where the high complexity of the matrix could
undermine their performance.
3.5. Comparison with other sorbents
The effectiveness of the Fe3O4/SiO2/NH/CS
2 particles by
means of maximum Hg (II) sorption capacities, is quantita-
tively compared with other sorbents, particularly with algal
biomass (Tuzun et al., 2005), Romanian clays (Hristodor et al.,
2010), ETS-4 titanosilicate (Lopes et al., 2009) and furfural
carbon (Yardim et al., 2003) (Table 4). The maximum Hg (II)
sorption capacities of these sorbents range between 122 and
246 mg/g and are of the same order of magnitude of that found
for Fe3O4/SiO2/NH/CS
2 particles.
Compared to other sorbents containing Dtc ligands, the
maximum Hg (II) sorption capacity of Fe3O4/SiO2/NH/CS
2
particles is considerable higher than that found for meso-
porous silica (MCM-41-Dtc) (Venkatesan et al., 2003) and silica
gel (Si-Dtc) (Venkatesan et al., 2002), despite the lower Dtc
surface coverage (respectively 2.5  10
4 and 3.7  10
4 mol/g
against 1.5  10
4 mol/g of the magnetic particles). Compa-
rable binding capacities were observed between Fe3O4/SiO2/
NH/CS-2 particles and other sorbents, like mesoporous silica
grafted with other sulphur ligands as thiol (Mercier and
Pinnavaia, 1998), mercaptan (Mattigod et al., 1999) and 1-
furoyl thiourea urea (Mureseanu et al., 2010) (see Table 4).
While sorbents like benzoylthiourea-modified mesoporous
silica (Antochshuk et al., 2003) and thiol functional organo-
ceramic composite (Nam et al., 2003), with higher sulphur
surface coverage (ca. 10
3 mol/g) and/or multifunctional
ligands, possess several “active” groups toward mercury ions
and exhibit considerable higher sorption capacities, on the
other hand they do not offer the possibility of magnetic
separation.

Table 4 e Residual mercury concentration (Ce) and sorption capacity (qm) of other sorbents for Hg (II).
Adsorbent qm (mg/g) Ce (mg/L) Ref.

Biosorbents Rice husk ash 6.72 n.a (Feng et al., 2004)

Bacillus sp. 7.94 20 (Green-Ruiz, 2006)
Eucalyptus bark 33.1 n.a (Ghodbane and Hamdaoui, 2008)
Seaweed biomass 84.7 n.a (Zeroual et al., 2003)
Yeast cells 93.4 n.a (Yavuz et al., 2006)
Algal biomass 122 n.a (Tuzun et al., 2005)
Clays and zeolitic Zeolitic mineral 10.1 n.a (Gebremedhin-Haile et al., 2003)
materials Clay 152 n.a (Hristodor et al., 2010)
ETS-4 titanosilicate 246 <1 (Lopes et al., 2009)
Carbons Activated carbon 25.8 n.a (Rao et al., 2009)
Carbon aerogel 34.9 w5000 (Goel et al., 2005)
Activated carbon 43.8 n.a (Ranganathan, 2003)
Furfural carbon 174 n.a (Yardim et al., 2003)
Sorbents containing Thiol functional organoceramic composite 726 <1 (Nam et al., 2003)
sulphur ligands Dithiocarbamate grafted on mesoporous silica 40.1 n.a. (Venkatesan et al., 2003)
Dithiocarbamate grafted on silica gel 61 n.a. (Venkatesan et al., 2002)
Thiol functionality grafted on mesoporous silica 110e301 n.a. (Mercier and Pinnavaia, 1998)
Mesoporous silica containing mercaptan groups 26e270 401 (Mattigod et al., 1999)
Benzoylthiourea-modified mesoporous silica 1000 n.a. (Antochshuk et al., 2003)
Mesoporous silica grafted with 1-furoyl thiourea urea 122 n.a. (Mureseanu et al., 2010)
Dithiocarbamate grafted on magnetite particles 142e206 <1 This study

n.a. not available.

 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4
It must also be highlighted that the Fe3O4/SiO2/NH/CS
2 references
particles together with thiol functional organoceramic
composite and ETS-4 titanosilicate were the only sorbents,
mentioned in Table 4, able to reduce the initial mercury
concentration to values lower than 1 mg/L (Table 4). This fact is
due not only to their sorption capacity but also because the
majority of the studies use initial Hg concentrations
extremely high and nothing realistic of the degree of
contamination found in the environment.
4. Conclusions
The sorption capacity towards Hg (II) of silica coated magne-
tite particles derivatized with dithiocarbamate groups was
studied in batch mode under different experimental condi-
tions. It was confirmed that silica coated magnetite particles
functionalized with Dtc groups are effective sorbents for Hg
(II) removal from synthetic and natural waters; however the
sorption process is strongly dependent on contact time and
particles concentration. The presence of higher concentra-
tions of Cl
and Naþ ions did not affect the amount of Hg (II)
removed per gram of sorbent at equilibrium but reduced the
rate at which Hg ions were removed from solution, as
confirmed by pseudo-first- and pseudo-second-order kinetic
models.
The Sips model provides a good description of the equi-
librium data, predicting a maximum sorption capacity of
206 mg/g at 211  C, which is quite high compared with the
sorption capacities found in the literature for other materials.
Furthermore, this study confirmed the effectiveness of silica
coated magnetite particles grafted with Dtc groups in two
distinct types of natural waters, seawater and river water.
However, the functionalized magnetite particles exhibited
a slightly higher performance in seawater than in river water.
It must be highlighted that in both ultra-pure water and
seawater, only 6 mg/L of functionalized particles was suffi-
cient to achieve a residual concentration lower than 1 mg/L,
which is the current acceptable value for drinking water
quality. Our results emphasize the advantages of these Dtc
functionalized particles, such as high affinity towards
mercury ions, selective removal, and large efficiency in high
complex matrices. Additionally, the easy separation of the
sorbent from solution due to the magnetite ferrimagnetic
properties opens new prospects in the design of high-
performance sorbents for environment remediation and
mercury ions recovery.
Acknowledgements
C.B. Lopes thanks Fundação para a Ciência e Tecnologia for
a Post-Doc grant (SFRH/BPD/45156/2008).
Appendix. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.watres.2011.08.057.
5783
Antochshuk, V., Olkhovyk, O., Jaroniec, M., Park, I.S., Ryoo, R.,
2003. Benzoylthiourea-modified mesoporous silica for
mercury(II) removal. Langmuir 19 (7), 3031e3034.
Cooney, D.O., 1999. Adsorption Design for Wastewater
Treatment. Lewis Publishers, Boca Raton (FL).
Daniel-da-Silva, A.L., Trindade, T., Goodfellow, B.J., Costa, B.F.O.,
Correia, R.N., Gil, A.M., 2007. In situ synthesis of magnetite
nanoparticles in carrageenan gels. Biomacromolecules 8 (8),
2350e2357.
EU, 1998. Council Directive 98/83/EC on the quality of water
intended for human consumption. Official Journal 330,
0032e0054.
EU, 2000. Directive 2000/60/EC of the European Parliament and of
the Council of the European Union, establishing a framework
for Community action in the field of water policy. Official
Journal of the European Communities 327/1.
EU, 2008. Directive 2008/105/EC of the European Parliament and of
the Council of the European Union, on environmental quality
standards in the field of water policy, amending and
subsequently repealing Council Directives 82/176/EEC, 83/513/
EEC, 84/156/EEC, 84/491/EEC, 86/280/EEC and amending
Directive 2000/60/EC of the European Parliament and of the
Council. Official Journal of the European Communities 348/84.
Feng, X., Fryxell, G.E., Wang, L.Q., Kim, A.Y., Liu, J., Kemner, K.M.,
1997. Functionalized monolayers on ordered mesoporous
supports. Science 276 (5314), 923e926.
Feng, Q.G., Lin, Q.Y., Gong, F.Z., Sugita, S., Shoya, M., 2004.
Adsorption of lead and mercury by rice husk ash. Journal of
Colloid and Interface Science 278 (1), 1e8.
Freundlich, H., 1906. Concerning adsorption in solutions.
Zeitschrift Fur Physikalische ChemieeStochiometrie Und
Verwandtschaftslehre 57 (4), 385e470.
Gebremedhin-Haile, T., Olguin, M.T., Solache-Rios, M., 2003.
Removal of mercury ions from mixed aqueous metal solutions
by natural and modified zeolitic minerals. Water Air and Soil
Pollution 148 (1e4), 179e200.
Ghodbane, I., Hamdaoui, O., 2008. Removal of mercury(II) from
aqueous media using eucalyptus bark: kinetic and equilibrium
studies. Journal of Hazardous Materials 160 (2e3), 301e309.
Girginova, P.I., Daniel-Da-Silva, A.L., Lopes, C.B., Figueira, P.,
Otero, M., Amaral, V.S., Pereira, E., Trindade, T., 2010. Silica
coated magnetite particles for magnetic removal of Hg2þ from
water. Journal of Colloid and Interface Science 345 (2),
234e240.
Goel, J., Kadirvelu, K., Rajagopal, C., Garg, V.K., 2005. Investigation
of adsorption of lead, mercury and nickel from aqueous
solutions onto carbon aerogel. Journal of Chemical
Technology and Biotechnology 80 (4), 469e476.
Green-Ruiz, C., 2006. Mercury(II) removal from aqueous solutions
by nonviable Bacillus sp from a tropical estuary. Bioresource
Technology 97 (15), 1907e1911.
Hasan, S.H., Srivastava, P., 2009. Batch and continuous
biosorption of Cu2þ by immobilized biomass of Arthrobacter
sp. Journal of Environmental Management 90 (11),
3313e3321.
Ho, Y.S., McKay, G., 1999. Pseudo-second order model for sorption
processes. Process Biochemistry 34 (5), 451e465.
Ho, Y.S., Porter, J.F., McKay, G., 2002. Equilibrium isotherm studies
for the sorption of divalent metal ions onto peat: copper,
nickel and lead single component systems. Water Air and Soil
Pollution 141 (1e4), 1e33.
Hristodor, C., Copcia, V., Lutic, D., Popovici, E., 2010.
Thermodynamics and kinetics of Pb(II) and Hg(II) ions removal
from aqueous solution by Romanian clays. Revista De Chimie
61 (3), 285e289.
5784 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 7 3 e5 7 8 4
Kocaoba, S., 2007. Comparison of Amberlite IR 120 and dolomite’s
performances for removal of heavy metals. Journal of
Hazardous Materials 147 (1e2), 488e496.
Lagergren, S., 1898. Zur theorie der sogenannten adsorption
gelöter stoffe. Kungliga Svenska Vetenskapsakademiens.
Handlingar 24 (4), 1e39.
Lopes, C.B., Otero, M., Coimbra, J., Pereira, E., Rocha, J., Lin, Z.,
Duarte, A., 2007. Removal of low concentration Hg2þ from
natural waters by microporous and layered titanosilicates.
Microporous and Mesoporous Materials 103 (1e3), 325e332.
Lopes, C.B., Otero, M., Lin, Z., Silva, C.M., Rocha, J., Pereira, E.,
Duarte, A.C., 2009. Removal of Hg2þ ions from aqueous
solution by ETS-4 microporous titanosilicate-kinetic and
equilibrium studies. Chemical Engineering Journal 151 (1e3),
247e254.
Marin, A.B.P., Aguilar, M.I., Meseguer, V.F., Ortuno, J.F., Saez, J.,
Llorens, M., 2009. Biosorption of chromium (III) by orange
(Citrus cinensis) waste: batch and continuous studies.
Chemical Engineering Journal 155 (1e2), 199e206.
Mattigod, S.V., Feng, X.D., Fryxell, G.E., Liu, J., Gong, M.L., 1999.
Separation of complexed mercury from aqueous wastes using
self-assembled mercaptan on mesoporous silica. Separation
Science and Technology 34 (12), 2329e2345.
Mercier, L., Pinnavaia, T.J., 1998. Heavy metal lan adsorbents
formed by the grafting of a thiol functionality to mesoporous
silica molecular sieves: Factors affecting Hg(II) uptake.
Environmental Science and Technology 32 (18), 2749e2754.
Mureseanu, M., Reiss, A., Cioatera, N., Trandafir, I., Hulea, V.,
2010. Mesoporous silica functionalized with 1-furoyl thiourea
urea for Hg(II) adsorption from aqueous media. Journal of
Hazardous Materials 182 (1e3), 197e203.
Nam, K.H., Gomez-Salazar, S., Tavlarides, L.L., 2003. Mercury(II)
adsorption from wastewaters using a thiol functional
adsorbent. Industrial and Engineering Chemistry Research 42
(9), 1955e1964.
Ranganathan, K., 2003. Adsorption of Hg(II) ions from aqueous
chloride solutions using powdered activated carbons. Carbon
41 (5), 1087e1092.
Rao, M.M., Reddy, D., Venkateswarlu, P., Seshaiah, K., 2009.
Removal of mercury from aqueous solutions using activated
carbon prepared from agricultural by-product/waste. Journal
of Environmental Management 90 (1), 634e643.
Starvin, A.M., Rao, T.P., 2004. Removal and recovery of mercury(II)
from hazardous wastes using 1-(2-thiazolylazo)-2-naphthol
functionalized activated carbon as solid phase extractant.
Journal of Hazardous Materials 113 (1e3), 75e79.
Tuzun, I., Bayramoglu, G., Yalcin, E., Basaran, G., Celik, G.,
Arica, M.Y., 2005. Equilibrium and kinetic studies on
biosorption of Hg(II), Cd(II) and Pb(II) ions onto microalgae
Chlamydomonas reinhardtii. Journal of Environmental
Management 77 (2), 85e92.
Umpleby, R.J., Baxter, S.C., Chen, Y.Z., Shah, R.N., Shimizu, K.D.,
2001. Characterization of molecularly imprinted polymers
with the Langmuir-Freundlich isotherm. Analytical Chemistry
73 (19), 4584e4591.
Venkatesan, K.A., Srinivasan, T.G., Rao, P.R.V., 2002. Removal of
complexed mercury from aqueous solutions using
dithiocarbamate grafted on silica gel. Separation Science and
Technology 37 (6), 1417e1429.
Venkatesan, K.A., Srinivasan, T.G., Rao, P.R.V., 2003. Removal of
complexed mercury by dithiocarbamate grafted on
mesoporous silica. Journal of Radioanalytical and Nuclear
Chemistry 256 (2), 213e218.
Wang, J.L., Chen, C., 2009. Biosorbents for heavy metals removal
and their future. Biotechnology Advances 27 (2), 195e226.
Yardim, M.F., Budinova, T., Ekinci, E., Petrov, N., Razvigorova, M.,
Minkova, V., 2003. Removal of mercury (II) from aqueous
solution by activated carbon obtained from furfural.
Chemosphere 52 (5), 835e841.
Yavuz, H., Denizli, A., Gungunes, H., Safarikova, M., Safarik, I.,
2006. Biosorption of mercury on magnetically modified yeast
cells. Separation and Purification Technology 52 (2), 253e260.
Zeroual, Y., Moutaouakkil, A., Dzairi, F.Z., Talbi, M., Chung, P.U.,
Lee, K., Blaghen, M., 2003. Biosorption of mercury from
aqueous solution by Ulva lactuca biomass. Bioresource
Technology 90 (3), 349e351.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Influence of operational parameters on nitrogen removal

efficiency and microbial communities in a full-scale activated
sludge process

Young Mo Kim a, Hyun Uk Cho b, Dae Sung Lee c, Donghee Park d,*, Jong Moon Park b,**
a
Department of Civil and Environmental Engineering, University of Massachusetts, Amherst, MA 01003, USA
b
Department of Chemical Engineering, School of Environmental Science and Engineering, Division of Advanced Nuclear Engineering,
Pohang University of Science and Technology, Pohang 790-784, Republic of Korea
c
Department of Environmental Engineering, Kyungpook National University, Daegu 702-701, Republic of Korea
d
Department of Environmental Engineering, Yonsei University, Wonju 220-710, Republic of Korea

article info abstract

Article history: To improve the efficiency of total nitrogen (TN) removal, solid retention time (SRT) and
Received 5 May 2011 internal recycling ratio controls were selected as operating parameters in a full-scale
Received in revised form activated sludge process treating high strength industrial wastewater. Increased biomass
17 August 2011 concentration via SRT control enhanced TN removal. Also, decreasing the internal recy-
Accepted 29 August 2011 cling ratio restored the nitrification process, which had been inhibited by phenol shock
Available online 3 September 2011 loading. Therefore, physiological alteration of the bacterial populations by application of
specific operational strategies may stabilize the activated sludge process. Additionally, two
Keywords: dominant ammonia oxidizing bacteria (AOB) populations, Nitrosomonas europaea and
Activated sludge Nitrosomonas nitrosa, were observed in all samples with no change in the community
Operational parameter composition of AOB. In a nitrification tank, it was observed that the Nitrobacter populations
Nitrogen removal consistently exceeded those of the Nitrospira within the nitrite oxidizing bacteria (NOB)
qPCR community. Through using quantitative real-time PCR (qPCR), nirS, the nitrite reducing
Nitrifying bacteria functional gene, was observed to predominate in the activated sludge of an anoxic tank,
Denitrifying bacteria whereas there was the least amount of the narG gene, the nitrate reducing functional gene.
ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction reactions in which nitrate or nitrite is transformed into

Activated sludge process is one of the most widely used bio-
logical treatments of wastewaters containing carbon and
nitrogen pollutants. Biological nitrogen removal has tradi-
tionally been accomplished using autotrophic nitrification
and heterotrophic denitrification. Nitrification is carried out
in two sequential steps via two distinct groups of bacteria:
ammonia oxidizing bacteria (AOB) and nitrite oxidizing
bacteria (NOB). Denitrification consists of consecutive
gaseous forms (N2 or N2O).
Although efficient and reliable in treating industrial
wastewater, activated sludge process is susceptible to distur-
bances and toxic loadings (Juliastuti et al., 2003; Mertoglu
et al., 2008; Kim et al., 2009). In particular, the activity of
nitrifying bacteria in wastewater treatment plants is sensitive
to shifts in the process’s pH and temperature, ammonia/
nitrite concentrations, oxygen concentration and the pres-
ence of toxic compounds, often leading to process failure

* Corresponding author. Tel.: þ82 33 760 2435; fax: þ82 33 760 2571.
** Corresponding author. Tel.: þ82 54 279 2275; fax: þ82 54 279 2699.
E-mail addresses: dpark@yonsei.ac.kr (D. Park), jmpark@postech.ac.kr (J.M. Park).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.063
5786 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
(Eighmy and Bishop, 1989; Kim et al., 2007). Fluctuation inflow,
organic loads, solid retention time (SRT) and lack of nutrient
matter also inhibit the nitrification process (Hallin et al., 2005).
Even though nitrifying bacteria have been fairly well studied,
nitrification is difficult to maintain stably throughout the
activated sludge process (Kim et al., 2009).
Actually, many current full-scale activated sludge processes
treating industrial wastewater have long experienced trouble
with instability, but previous reports have simply focused on
monitoring the emergence of the process failure (Kim et al.,
2007, 2008, 2009). The causes of the process instability have
not been clearly identified; thus appropriate solutions have not
yet been suggested (Kim et al., 2011a,b). Moreover, research
targeting a full-scale wastewater treatment process (WWTP)
has rarely been attempted due to either sampling problems or
difficulties in proper WWTP selection. Results from lab scale
experiments have proved difficult to extrapolate to real WWTP
conditions (Kim et al., 2007, 2009). In addition, little is known
about the microbial ecology of nitrifying and denitrifying
bacteria in full-scale activated sludge processes treating high
strength industrial wastewater containing toxic compounds
like thiocyanate, ammonia and phenol. Therefore, this study
aimed to identify the causes underlying current difficulties
achieving and maintaining the legal discharge of total nitrogen
(TN) in the final effluent of a full-scale activated sludge process.
To improve the efficiency of TN removal, SRT and internal
recycling ratio controls as main operation parameters were
attempted in an unstable process. Meanwhile, bacterial pop-
ulations responsible for biological nitrogen removal were
investigated in relation to changes in the efficiency of TN
removal of a full-scale activated sludge process treating
wastewater from a coke plant. Diversity surveys assessing the
relationship between bacterial populations and activity to the
overall processing conditions may lead to an understanding of
the basis of process instability and be of help in designing better
process monitoring while avoiding operational failure.
2. Materials and methods
2.1. Wastewater treatment plant operation and samples
A full-scale WWTP of a coke manufacturing plant in Pohang,
Korea employs a single sludge along with the recycling of
Fig. 1 e Schematic diagram of a full-scale activated sludge process treating high strength industrial wastewater.
nitrified effluent e the pre-denitrification activated sludge
process. This system is composed of an anoxic tank (900 m3),
an aerobic tank (3600 m3) and a nitrification tank (1800 m3)
with a combined volume of 6300 m3 and treats about 300 m3
of cokes wastewater per hour (Fig. 1). Nitrified effluent was
recycled from the nitrification tank to the anoxic tank at the
rate of about 600 m3 per hour. Return activated sludge
collected from the secondary clarifier was pumped back to
the anoxic tank at the rate of about 300 m3 per hour. Both the
aerobic and nitrification tanks were aerated. In the course of
this study, the concentrations of pollutants in the raw
wastewater were as follows: 2025e3150 mg/L of chemical
oxygen demand (COD), 680e1070 mg/L of biochemical
oxygen demand (BOD5), 642e916 mg/L of total organic carbon
(TOC), 268e715 mg/L of phenol, 177e236 mg-N/L of total
nitrogen (TN), 76e140 mg-N/L of ammonia, 190e297 mg-
SCN/L of thiocyanate (SCN) and 12.4e17.2 mg-CN/L of
total cyanides. The mixed-liquor suspended solids (MLSS) of
the system were controlled at about 1800 mg/L, the average
hydraulic retention time (HRT) was 0.9 day and the average
SRT was 15 days. The SRT was controlled by removal of
excess sludge, resulting in different MLSS concentrations of
the system. The temperature range of the tanks varied
between 33 and 36  C. The pH of the influent, anoxic tank and
nitrification tank was maintained at 9.0, 7.5 and 7.0, respec-
tively. The dissolved oxygen (DO) concentration in the
aerobic and nitrification tank was more than 4.0 mg/L, while
the DO level of the anoxic tank was maintained below
0.3 mg/L. Functional stability of the system was defined and
quantified by the effluent concentration of TN.
MLSS samples for this study were taken from the last
sections of the anoxic and nitrification tanks weekly for 3
months (AugusteOctober). Nitrification and denitrification
activity was measured directly on fresh samples. For the DNA
based studies, each sample of 1.0 mL was dispensed into
a 1.5 mL sterile tube and centrifuged at 13000 g for 10 min. The
supernatant was decanted and the pellet was stored at 20  C
before being used.
2.2. Process monitoring and chemical analysis
The collected samples were centrifuged at 3500 rpm for 3 min
(MF550, Hanil Sci. Ind., Korea), and the supernatants were
analyzed as follows: according to standard methods (APHA,
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5 5787

1998), chemical oxygen demand (COD), ammonia, phenol
and thiocyanate (SCN) were analyzed by the colorimetric
method with a spectrophotometer (Genesys TM-5, Spec-
tronic Inc., USA). After distillation, the cyanide (CN)
concentration was determined by the pyridine-pyrazolone
method. Nitrite and nitrate ions were measured with an
ion chromatograph (ICS-1000, Dionex Co., USA). Total orag-
nic carbon (TOC), inorganic carbon (IC), and TN were
measured with a TOC/TN analyzer (TOC-V csu, TNM1,
Shimadzu Co., Japan).
2.3. Microbial activity test
To investigate microbial activities of nitrifiers and denitrifiers
in each tank at the WWTP, nitrification and denitrification
rates were estimated weekly through batch experiments with
synthetic medium containing ammonia or nitrate. Batch
experiments for nitrification and denitrification activity were
carried out in 500 mL Erlenmeyer flasks filled with 100 mL of
test solution containing 50 mg-N/L of ammonia and nitrate
ion, respectively. Without any pretreatment each flask was
inoculated with fresh activated sludge (about 2000 mg/L),
which was sampled directly from each full-scale tank, and
then agitated on a thermostatic shaker at 200 rpm and 35  C,
maintaining the pH at 7.5. The specific nitrification and
denitrification rates were calculated applying the equation
provided by Kim et al. (2011a).
2.4. DNA extraction
The pellet was washed with 1 mL of deionized and distilled
water (DDW) and centrifuged at 16,000 g for 5 min to ensure
a maximal removal of residual medium. The supernatant was
carefully removed and the pellet resuspended in 100 mL of
DDW. All DNA in the suspension was immediately extracted
using an automated nucleic acid extractor (Magtration System
6 GC, PSS, Chiba, Japan). Purified DNA was eluted with 100 mL
of TriseHCl buffer (pH 8.0) and stored at 20  C for further
analyses.
2.5. T-RFLP analysis
T-RFLP was used to analyze the nitrifying bacteria community
in the pre-denitrification process reactor based on the known
16S rRNA genes of ammonia oxidizing bacteria (AOB) and
nitrite oxidizing bacteria (NOB), as described in the protocol of
a previous study (Siripong and Rittmann, 2007). Because of the
low concentration of DNA from the nitrifiers, it was amplified
through nested PCR, using the universal primers 11f and 1492r
(Table 1), followed by the specific amplification of nitrifier
genes (Nitrifier-specific reverse primer: Nso1225r, NIT3r,
Ntspa685r, Forward primer: Eub338f included phosphor-
amidite dye 6-FAM (Table 1)) (Siripong and Rittmann, 2007).
2 mL of template DNA was used for the universal amplification
step and 1 mL of the universal amplification product as the

Table 1 e Primers and probes used in T-RFLP and qPCR.

Target Primer/probe Sequence (5’-3’) References

For T-RFLP
Bacterial 16S rDNA 11f 5’-GTTTGATCCTGGCTCAG-3’ Kane et al., 1993
1492r 5’-TACCTTGTTACGACTT-3’ Lin and Stahl, 1995
Bacterial I6S rDNA Eub338f 5’-(6-FAM)-ACTCCTACGGGAGGCAGC-3’ Amann et al., 1990
AOB 16S rDNA Nso1225r 5’-CGCCATTGTATTACGTGTGA-3’ Mobarry et al., 1996
Nitrobacter 16S rDNA NIT3r 5’-CCTGTGCTCCATGCTCCG-3’ Wagner et al., 1995
Nitrospira 16S rDNA Ntspa685r 5’-CGGGAATTCCGCGCTC-3’ Regan et al., 2002
For qPCR
Bacterial 16S rDNA 1055f 5’-ATGGCTGTCGTCAGCT-3’ Ferris et al., 1996
1392r 5’-ACGGGCGGTGTGTAC-3’ Ferris et al., 1996
16STaq 1115 5’-(6-FAM)-CAACGAGCGCAACCC-(TAMRA)-3’ Harms et al., 2003
AOB 16S rDNA CTO 189fA/Ba 5’-GGAGRAAAGCAGGGGATCG-3’ Hermansson and Lindgren, 2001
CTO 1891Ca 5’-GGAGGAAAGTAGGGGATCG-3’ Hermansson and Lindgren, 2001
RTlr 5’-CGTCCTCTCAGACCARCTACTG-3’ Hermansson and Lindgren, 2001
TMP1 5’-(6-FAM)-CAACTAGCTAATCAGRCATCRGCCGCT-(TAMRA)-3’ Hermansson and Lindgren, 2001
Nitrospira spp. 16S rDNA NSR 1113f 5’-CCTGCTTTCAGTTGCTACCG-3’ Dionisi et al., 2002
NSR 1264r 5’-GTTTGCAGCGCTTTGTACCG-3’ Dionisi et al., 2002
NSR 1143Taq 5’-(6-FAM)-AGCACTCTGAAAGGACTGCCCAGG-(TAMRA)-3’ Harms et al., 2003
Nitrobacter spp. I6S1DNA Nitro 1198f 5’-ACCCCTAGCAAATCTCAAAAAACCG-3’ Graham et al., 2007
Nitro 1423r 5’-CTTCACCCCAGTCGCTGACC-3’ Graham et al., 2007
Nitro 1374Taq 5’-(6-FAM)-AACCCGCAAGGAGGCAGCCGACC-(TAMRA)-3’ Graham et al., 2007
narG gene narG 1960m2f 5’-TAYGTSGGGCAGGARAAACTG-3’ Lòpez-Gutièrrez et al., 2004
narG 2050m2r 5’-CGTAGAAGAAGCTGGTGCTGTT-3’ Lòpez-Gutièrrez et al., 2004
nirS gene nirS If 5’-TACCACCCSGARCCGCGCGT-3’ Braker et al., 1998
nirS 3r 5’-GCCGCCGTCRTGVAGGAA-3’ Braker et al., 1998
nirK gene nirK 876 5’-ATYGGCGGVCAYGGCGA-3’ Henry et al., 2004
nirK 1040 S’-GCCTCGATCAGRTTRTGGTT-S’ Henry et al., 2004
nosZ gene nosZ 2f 5’-CGCR ACGGC AAS AAGGTSM SSGT-3’ Henry et al., 2006
nosZ 2r 5’-CAKRTGCAKSGCRTGGCAGAA-3’ Henry et al., 2006

a A mixture of CTO 189fA/B and CTO 189fC at the weight ratio of 2:1 was used as the forward primer.
5788 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
template for the nitrifier-specific amplification. Finally, the
PCR products were purified and 16S rRNA gene amplicons
were digested with MspI restriction endonuclease (Siripong
and Rittmann, 2007). Digested PCR products were run
through an ABI 3130XL Genetic Analyzer (Applied Biosystems,
Foster City, USA) at the SolGent Company (Korea). The peak
results were analyzed using the Peak Scanner software v 1.0
(https://products.appliedbiosystems.com; Applied Bio-
systems, Foster City, USA). Details of the PCR conditions,
product purification and restriction digestion are provided
elsewhere (Siripong and Rittmann, 2007).
2.6. qPCR analysis
To investigate the changes in the nitrifying and denitrifying
bacteria populations occurring during the process perfor-
mance, all qPCR assays were performed using a 7300 Real
Time PCR system (Applied Biosystems, Foster City, USA).
To determine the amount of the nitrifying bacteria, four
independent qPCR assays were conducted by quantifying
total bacterial 16S rDNA, ammonia oxidizing bacterial 16S
rDNA, Nitrospira spp. 16S rDNA, and Nitrobacter spp. 16S rDNA
(Table 1). Each capillary tube was separately loaded with 2 mL
of template DNA (at 14e26 ng/mL), followed by 4.0 pmol of the
forward and reverse primers (1 mL), together with 2.0 pmol of
the TaqMan probe (0.5 mL) corresponding to each primer and
probe set, 12.5 mL of TaqMan Universal PCR Master Mix (No
4304437 Applied Biosystems, New Jersey, USA), and PCR-grade
sterile water for a final volume of 25 mL. The amount of total
bacterial 16S rDNA was amplified using primer (1055f and
1392r) (Ferris et al., 1996). The TaqMan probe (16STaq1115)
was modified by the 1114f primer (Harms et al., 2003). The PCR
program was 2 min at 50  C, 10 min at 95  C; 45 cycles of 30 s at
95  C, 60 s at 50  C, and 40 s at 72  C. To determine the amount
of AOB 16S rDNA genes, two forward primers (CTO 189A/B
and CTO 189C), one reverse (RT1r), and the TaqMan probe
(TMP1) were used as described previously by Hermansson and
Lindgren (2001). The PCR program for AOB 16S rDNA quanti-
fication included 2 min at 50  C, 10 min at 95  C; 40 cycles of
30 s at 95  C, 60 s at 60  C. The Nitrospira spp. 16S rDNA
primers (NSR 1113f/NSR 1264r) (Dionisi et al., 2002) and the
TaqMan probe (NSR 1143Taq) (Harms et al., 2003) were tested.
PCR amplification consisted of 2 min at 50  C, 10 min at 95  C;
50 cycles of 30 s at 95  C, 60 s at 60  C. Lastly, the amount of
Nitrobacter spp. from Graham et al. (2007) was amplified using
primer (Nitro 1198f/Nitro 1423r) and TaqMan probe (Nitro
1374Taq). The program used for amplification was 2 min at
50  C, 10 min at 95  C; 50 cycles of 20 s at 94  C, 60 s at 58  C,
and 40 s at 72  C.
Meanwhile, the denitrifying functional genes were quan-
tified with SYBR Premix Ex Tag (Takara, Japan). Amplifica-
tion reactions were performed in a total volume of 25 mL
containing 2 mL of template DNA (at 17.5e22.5 ng/mL), 4.0 pmol
of the forward and reverse primers (1 mL), together with 0.5 mL
(1X) of the ROX reference dye (50X), 12.5 mL of SYBR Premix,
and PCR-grade sterile water. The qPCR program for 16S rDNA
amplification using primer 1055f and 1392r was 30 s at 95  C;
30 cycles of 15 s at 95  C, 20 s at 55  C, and 31 s at 72  C. Primers
designed by López-Gutiérrez et al. (2004) were used to deter-
mine the amount of narG gene. The PCR program for narG gene
quantification included 30 s at 95  C; 35 cycles of 15 s at 95  C,
30 s at 58  C, and 31 s at 72  C. The nirS gene PCR amplification
using primers (nirS 1f and nirS 3r) (Braker et al., 1998) con-
sisted of 30 s at 95  C; 30 cycles of 15 s at 95  C, 20 s at 60  C, and
31 s at 72  C. The PCR condition for nirK gene included 30 s at
95  C; 30 cycles of 15 s at 95  C, 30 s at 58  C, and 31 s at 72  C.
Lastly, Primers (nosZ 2f and nosZ 2r) designed by Henry et al.
(2006) were used to determine the amount of nosZ gene. The
program used for amplification was 30 s at 95  C; 30 cycles of
15 s at 95  C, 30 s at 60  C, and 31 s at 72  C.
All experiments were performed in duplicate per sample
and all PCR runs included control reactions without the
template. The specificity of each PCR assay was confirmed
using both melting curve analysis and agarose gel electro-
phoresis. Gene copy numbers were calculated by comparing
threshold cycles obtained in each PCR run with those of
known standard DNA concentrations. Standards were
prepared using serially diluted plasmid DNA with 103e108
gene copies/mL. Standard curves for the 16S rDNA, AOB,
Nitrospira spp., Nitrobacter spp., narG, nirS, nirK, and nosZ assays
were generated by plotting the threshold cycle values versus
log10 of the gene copy numbers. The amplification efficiency
(E ) was estimated using the slope of the standard curve
through the following formula: E ¼ (101/slope)  1. The effi-
ciency of PCR amplification for each gene was between 90%
and 100%.
2.7. Cloning and sequencing
Prior to cloning, the amplified unlabeled AOB 16S rRNA genes
fragments were purified using the PCR purification kit (Sol-
Gent, Korea). Purified PCR products were ligated into pGEM-T
Easy cloning vectors (Promega, USA) and transformed into
competent Escherichia coli One-Shot Mach 1-T1 (Invitrogen,
USA), as described in the manufacturer’s protocol. Trans-
formants were selected by ampicillin resistance and blue-
ewhite screening was performed to identify clones with
inserts. Seventy-four white colonies were selected and culti-
vated. Primers T7 and SP6 were used to perform colony PCR
and to verify that the insert size was correct. Following PCR
confirmation of insert size, the amplified inserts were run on
2% (wt/vol) agarose gels. The samples containing inserts of the
estimated size were used for subsequent sequencing. The 16S
rRNA gene inserts were sequenced through an ABI 3130XL
Genetic Analyzer (Applied Biosystems, Foster City, USA) at the
SolGent Company (Korea). Database homology searches for
these sequences were performed using the BLAST program in
the National Center for Biotechnology Information (NCBI)
database.
3. Results and discussion
3.1. Functional performance of the full-scale activated
sludge process
The process performance of the full-scale wastewater treat-
ment system during the study is presented in Fig. 2. Although
the pre-denitrification activated sludge process is simple,
various microbial reactions occur sequentially under both
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
Fig. 2 e Variation of influent- and effluent-concentrations of (a) COD; (b) TOC; (c) phenol; (d) SCNL; (e) ammonia; (f) TN in the
full-scale process during the study period (solid circles: influent, open circles: effluent, solid triangles: removal efficiency).
anoxic and aerobic conditions. During the study,
2025e3150 mg/L of COD was fed into the anoxic tank where
75e80% of it was subsequentially removed. Residual COD
flowed into the aerobic tank and then further removed,
resulting in a COD removal efficiency of 80e88%. The removal
pattern of TOC was similar to that of COD. There was no
further degradation of any residual COD or TOC in the nitri-
fication tank. This implies that the remaining residual organic
carbon was non-biodegradable. Phenol flowed into the full-
scale process in the range of 268e715 mg/L and increased to
be 75% higher than normal concentrations of about 400 mg/L
in the ninth week (Fig. 2c). Most phenol was first degraded in
the anoxic tank, with remaining phenol being almost
completely degraded in the aerobic tank. Transformation of
organic carbons such as phenol to inorganic carbon took place
in the anoxic tank due to denitrification and fermentation
reactions by various heterotorphes (Kim et al., 2009). The
inorganic carbon was consumed by autotrophic nitrifiers and
thiocyanate-degrading bacteria under aerobic conditions.
Almost all SCN, which varied from 190 to 297 mg/L in the
influent, was removed in the aerobic tank (Fig. 2d). Various
autotrophic bacteria in activated sludge are known to degrade
SCN under aerobic conditions ðSCN þ 2O2 þ 2H2 O/
NHþ 2
4 þ SO4 þ CO2 Þ (Lee et al., 2008). Total cyanides in the
range of 12.4e17.2 mg/L flowed into the anoxic tank and
1.7e3.8 mg/L of it flowed into the aerobic tank (data not
shown). Additional degradation of cyanides did not take place
under aerobic conditions. The incomplete removal of total
cyanides was due to the existence of ferric cyanide ðFeðCNÞ3
6 Þ,
which is known to undergo very slow biodegradation (Kim
et al., 2008).
The main role of the nitrification tank was to convert the
ammonium ion into nitrite and/or nitrate. The average NHþ 4 N
5789
concentration range in the influent was about 76e90 mg-N/L,
except for shock loading of ammonia (Fig. 2e). Some amounts of
ammonia were newly generated due to cell lysis and degrada-
tion of SCN in the anoxic and aerobic tanks, respectively.
During stable operation, the final effluent concentration of
NHþ 4  N in the nitrification tank remained less than 10 mg-N/L,
while the NO 2  N concentration was in the range of 17e20 mg-
N/L. Detected NO 3  N concentration values were less than
3.0 mg-N/L. Denitrification was promoted by recycling the
nitrite/nitrate formed via nitrification back to the anoxic tank.
Complete denitrification was consistently achieved in the
anoxic tank until 11 weeks regardless of ammonia and phenol
shock loading. Excluding this shock loading of ammonia, the
average TN in the influent was 190 mg-N/L (Fig. 2f). The effluent
TN concentration remained below 40 mg-N/L, corresponding to
an average TN removal efficiency of 85% (note that regulations
stipulate a TN concentration less than 60 mg-N/L for discharge
into surface water in South Korea). Also, the effluent TN
concentration of less than 40 mg-N/L indicated that the system
was functionally stable throughout the study.
3.2. Influence of operational parameters on nitrogen
removal performance of the full-scale activated sludge
process
Nitrification of the full-scale activated sludge process treating
high strength pollutants targeted in our study has been
unstable during the past several summers due to an abnormal
influx of pollutants such as phenol, SCN, ammonia and
cyanide. Therefore, responsive system operations have been
proposed to accomplish discharge level regulations under
various situations.
5790 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
Fig. 3 shows the effect of operational parameters on the
effluent nitrogen concentration of the full-scale activated
sludge process. In the first week, although TN concentration
remained below the legal discharge level (i.e., 60 mg-N/L),
NO 2  N concentration, a by-product of nitrification, was only
7 mg-N/L (Fig. 3e). This indicated inadequate nitrification
performance. The nitrification tank was operated at an MLSS
concentration of about 1800 mg/L at 15 days of SRT (Fig. 3c),
which was lower than the 2500e3000 mg/L MLSS level of
general activated sludge processes treating cokes wastewater
(Manekar et al., 2011). To increase the nitrifying biomass
having a slow growth rate, the SRT was lengthened from 15
days to 20 days by reducing excess sludge removal. This
resulted in an increase in MLSS concentration. For 2 weeks,
the MLSS concentration in the nitrification tank increased
Fig. 3 e Variation of operational parameters: (a) SRT and
internal recycling ratio; (b) MLSS concentration and MLVSS/
MLSS ratio in the nitrification tank; (c) MLSS concentration
and MLVSS/MLSS ratio in the anoxic tank; (def) effluent-
concentrations of ammonia, nitrite nitrate, TN in the
nitrification tank during the study period.
from 1800 mg/L to 2000 mg/L (Fig. 3c). In the final effluent of
the nitrification tank, the ammonia concentration was below
20 mg-N/L while the NO 2  N concentration gradually
increased to 18 mg-N/L (Fig. 3d and e). TN concentration was
consistently maintained under the legal discharge level of
60 mg-N/L. It is believed that this rather long SRT may lead to
an increase in the slow growing nitrifying bacteria and
support their dominance, resulting in better nitrification
performance (Teck et al., 2009).
At the beginning of the fourth week, however, an abrupt
change in the effluent nitrogen concentration of the nitrifi-
cation tank occurred (Fig. 3f). The ammonia concentration
sharply increased to over 40 mg-N/L and the TN concentration
exceeded its legal discharge level of 60 mg-N/L. Initially, it was
doubted that a decrease in nitrification activity had occurred
as a result of inhibition by cyanide or phenol. However,
analysis results of the influent and effluent revealed that the
shock loading of ammonia was to blame. In the fourth week,
the ammonia concentration in the influent sharply increased
to 140 mg-N/L, a level more than twice normal loading (data
not shown). Meanwhile, the NO 2  N concentration in the
effluent did not decrease, implying that the nitrification
performance had not been inhibited. For high volumetric
ammonia removal, the process was controlled as nearly
infinite SRT without removal of excess sludge, resulting in an
accumulation of MLSS in the system for 1 week. It is known
that high volumetric loading can be achieved by maintaining
a high MLSS (Rittmann and McCarty, 2001). At the end of 4
weeks, the MLSS concentration in the system significantly
increased to 2800 mg/L (Fig. 3d). This led to a gradual decrease
in both the ammonia and TN concentrations in the final
effluent. In addition, as soon as the usual nitrogen concen-
tration in the influent flowed into the system after the fifth
week, the effluent nitrogen removal efficiency improved more
than previously. The ammonia concentration was controlled
at less than 10 mg-N/L and the TN concentration was main-
tained much lower than the legal level. The NO 2 N
concentration of about 17 mg-N/L was produced by nitrifica-
tion. Complete denitrification was consistently achieved in
the anoxic tank. As a result, the increased MLSS concentration
in the system by long SRT ensured stable nitrification of the
full-scale activated sludge process during ammonia shock
loading. These results implied that the selection of an
adequate biomass concentration by SRT control can be vital in
achieving the desired efficiency of the process.
However, full-scale process performance took a sudden
turn for the worse in the ninth week (Fig. 3f). Ammonia
concentration increased to 87 mg-N/L while NO 2 N
concentration decreased to 3.5 mg-N/L in the effluent of the
nitrification tank (Fig. 3d and e). The decrease of nitrite
concentration reflected incomplete nitrification in the nitrifi-
cation tank. Since nitrification was significantly inhibited, TN
concentration abruptly increased from 30 to 120 mg-N/L. The
analysis of variations in the pollutants’ concentrations in the
influent indicated that shock loading of phenol was one of the
causes for the nitrification failure (Fig. 2c). As the influent
concentration of phenol increased from 300e400 mg/L to
715 mg/L, more phenol flowed into the aerobic tank. The
inhibitory effect of phenol on nitrification is well known.
Previous research has shown that periods of nitrification
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
failure coincide with increased phenol concentration in the
wastewater (Figuerola and Erijman, 2010).
To prevent deterioration of the nitrification performance
from phenol inhibition, the internal recycling ratio was
controlled from 2.0 (600 m3/h) to 0.5 (150 m3/h) and long SRT
was consistently maintained at 20 days. By reducing the
internal recycling ratio, phenol removal performance was
enhanced in the anoxic tank, resulting in inflow of less phenol
into the nitrification tank. In addition, the decreased recycling
ratio increased the retention time of both the biomass and
effluent in the nitrification tank. The increased retention time,
in turn, provided both adequate nitrification reaction time and
contact stabilization, allowing nitrifiers to rebound from any
prior inhibition. In addition, it is known that high internal
recycling values have a negative effect on the maximum
specific growth rate of nitrifiers (Jimenez et al., 2011). Mean-
while, the long SRT prevented the nitrifying biomass from
washing out of the system. As a result, as soon as the phenol
concentration in the influent decreased to below 500 mg/L
after the tenth week (Fig. 2c), the TN removal efficiency
improved and a discharge level less than 50 mg-N/L could be
achieved (Fig. 3f). The effluent ammonia concentration in the
tenth week quickly decreased from 87 mg-N/L to 25 mg-N/L
and in the eleventh week was lower than 10 mg-N/L achieving
the same functional stability period as between the fifth and
eighth week (Fig. 3d). NO 2  N production also recovered from
phenol inhibition and NO 3 N concentration slightly
increased to 4.1 mg-N/L (Fig. 3e). This meant that nitrification
performance had totally recovered. In addition, the increased
retention time may be helpful in allowing NOB to generate
nitrate ion from nitrite substrate.
At the end of the twelfth week, although the effluent
ammonia and TN concentrations were stably maintained,
NO 2  N production began to decrease from 18 to 9 mg-N/L
(Fig. 3e). MLSS concentration climbed to 3400e3600 mg/L in
the system in spite of consistent SRT. These results indicated
that an increase of organic matter may cause proliferation of
heterotrophic microorganisms in the nitrification tank,
increasing uptake ammonia for their growth. Contrary to the
increase in MLSS concentration, MLVSS/MLSS ratio in the
nitrification tank quickly fell to 0.65, resulting in a decrease in
nitrification efficiency (Fig. 3b and d). In the anoxic tank, the
MLVSS/MLSS ratio sharply decreased to 0.7 with NO 2 N
concentration of 6.0 mg/L, resulting in a decrease in denitri-
fication efficiency (data not shown). This decreased VSS
concentration may result from an imbalance between feed
and biomass, since influent concentrations like organic and
nitrogen matter gradually decreased after the ninth week,
corresponding to an increase of MLSS concentration by
consistently long SRT. Consequently, these results in the
twelfth week implied that nitrogen removal performance of
the full-scale process in the future could be vulnerable to
environmental factors.
3.3. Microbial activity
Batch experiments to observe the variations of nitrification
and denitrification activities in the activated sludge process
were carried out. The specific nitrification and denitrification
rate in each batch test was analyzed through variations of
5791
ammonia, nitrite and nitrate concentrations. As illustrated in
Fig. 4, nitrifying bacteria activity was affected by process
conditions. Until the third week, increased SRT did not influ-
ence the specific nitrification rate, maintaining at about
5.0 mg-N/g-VSS$h. In the fourth week, however, the specific
nitrification rate decreased to 3.7 mg-N/g-VSS$h due to
a sudden increase in the MLSS concentration. Between the
fifth and eighth week, the specific nitrification rate gradually
increased from 4.3 to 6.3 mg-N/g-VSS$h, irrespective of a small
increase in MLSS concentration. In this period, nitrogen
removal efficiency in the full-scale system achieved almost
90%. In the ninth week, nitrification performance of the full-
scale process sharply decreased; on the other hand, the
batch test revealed only slightly decreased nitrification
activity, decreasing from 6.3 to 5.9 mg-N/g-VSS$h. This indi-
cated that there was discordance between the potential rate
(batch test) and the actual rate (full-scale performance)
regarding nitrification performance. One may conclude that,
although nitrifying bacteria possess good potential activity,
difficult to identify environmental factors influence their
actual performance in the full-scale process, leading to
decreased nitrification.
During the study, the specific denitrification rate pattern
was similar to variations in nitrification activity. The specific
denitrification rate remained in the range of 2.8e4.8 mg-N/g-
VSS h. Like the variation of the nitrification rate, the specific
denitrification rate decreased in the fourth week, due to
a considerable increase in the MLSS concentration, then
gradually increased to 4.8 mg-N/g-VSS h. Even in the ninth
week when phenol shock loading incidentally occurred, the
specific denitrification rate was not affected and remained
consistent. Contrary to the nitrification performance, deni-
trification in the full-scale process had a stable rate - similar to
its batch tests. This implied that denitrification may be less
affected than nitrification by environmental factors. Mean-
while, the MLVSS/MLSS ratio in both the anoxic and nitrifi-
cation tanks gradually decreased. As shown in Fig. 4, however,
both the nitrification and denitrification rates in the batch
tests were consistent, regardless of the nitrogen removal
performance of the full-scale process. As a result, the decrease
of MLVSS did not lead to any loss of microorganism activity.
Fig. 4 e Variation of denitrification and nitrification
activities by batch test during the study period.
5792 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
This may result from a consistent percentage of active
bacterial populations among total microorganisms.
3.4. T-RFLP analysis of AOB and NOB populations
We identified the nitrifying bacterial communities present in
an activated sludge process using T-RFLP designed for the
identification of AOB and NOB with terminal fragment (TF)
lengths (Regan et al., 2002). Fig. 5 shows electropherograms of
AOB, Nitrobacter-specific NOB, and Nitrospira-specific NOB
present in the nitrification tank of the full-scale process,
respectively. As shown in Fig. 5a, AOB-targeted T-RFLP
allowed us to differentiate between AOB groups. All samples
showed a peak at 164 bp, a signature of Nitrosomonas europaea/
eutropha and Nitrosomonas marina lineage (Table 2). Because
the influents originate from industrial wastewater, marine
AOB species need not be considered. Besides the major peak at
164 bp, we detected another peak at 273 bp, representing the
potential presence of N. europaea/eutropha, Nitrosomonas oligo-
tropha, Nitrosomonas cryotolerans, or Nitrosomonas communis
lineage (Table 2).
To better understand the AOB community present in the
process, AOB 16S rRNA gene based cloning and sequencing
was performed using the AOB-target primer (Nso1225r and
Eub338f) without fluorescent dye (Table 1). Sixty-eight of total
74 AOB clones from the reactor were closely associated with
N. europaea in the N. europaea/eutropha lineage and Nitro-
somonas nitrosa in the N. communis lineage, but the AOB clone
related to the Nitrosospira lineage was not detected. As a result,
based on the 16S rRNA gene sequences, microorganisms cor-
responding to the peaks at 164 bp and 273 bp could be
Fig. 5 e T-RFLP profiles of (a) AOB, (b) Nitrobacter, (c) Nitrospira in the nitrification tank during the study period.
identified as N. europaea and N. nitrosa, respectively (Fig. 5a).
Thus, the high peak at 164 bp implies the dominance of the N.
europaea within AOB in this wastewater treatment system,
irrespective of variations in the full-scale process perfor-
mance; N. nitrosa had a minor presence. N. europaea has been
widely observed in WWTPs (Lydmark et al., 2007; Siripong and
Rittmann, 2007), while N. nitrosa has previously been detected
on occasion in activated sludge treating industrial wastewater
(Layton et al., 2005). Despite variations in environmental
conditions such as MLSS concentration, internal recycling
ratio and influent characteristics for the AOB, the total selec-
tive pressure in the full-scale process has been insufficient to
induce a population shift (Hallin et al., 2005; Kim et al., 2011b).
This finding also indicated that low diversity of AOB pop-
ulations may be conducive to nitrification failure.
Based on Nitrobacter-specific T-RFLP, Fig. 5b shows
a prominent peak at 137 bp, characteristic of the Nitrobacter
species. Meanwhile, in the fifth week, the peak at 214 bp was
dominant along with the 137 bp peak, but disappeared after
the ninth week. This Nitrobacter sp. corresponding to 214 bp
peak may be affected by phenol. We also found TF sizes at 92,
162, 245 and 273 in the samples. These unexpected peaks
could be the result of incomplete digestion, uncharacterized
Nitrobacter species or imperfectly matched primer (Siripong
and Rittmann, 2007). The results of Nitrospira-specific T-RFLP
showed four dominant peaks at 135, 192, 272 and 334 bp
(Fig. 5c). The peak at 135, 192 and 272 corresponds to several
Nitrospira clones in the database. The 334 TF belongs to one of
the Nitrospira moscoviensis strains (Siripong and Rittmann,
2007). The Nitrospira species corresponding to the peak at
272 bp was a consistently dominant population, while
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
Table 2 e Expected TF sizes and their corresponding AOB
and NOB groups based on T-RFLP of 16S rDNA gene
(Siripong and Rittmann, 2007).
TF size (bp) Nitrifying bacteria group
164e166, 276 Nitrosomonas europaea/eutropha lineage
276 Nitrosomonas oligotropha lineage
276 Nitrosomonas cryotolerans lineage
166 Nitrosomonas marina lineage
276 Nitrosomonas communis lineage
105e107 Nitrosospira lineage
141, 196 Nitrobacter species
133, 194, 265e267, 277,333 Nitrospira species
N. moscoviensis sp. corresponding to the peak at 334 bp was
found to be predominant when nitrification performance of
the full-scale process began to stabilize in the fifth week.
Despite production of small amounts of nitrate by NOB in the
nitrification tank, harsh environmental conditions for NOB
may stimulate diversity of the NOB species.
3.5. qPCR analysis of nitrifying and denitrifying bacteria
populations
Fig. 6a shows the changes in the 16S rRNA gene copies for the
total bacteria, AOB, Nitrobacter, and Nitrospira, quantified using
qPCR assays in the nitrification tank of the full-scale process.
In all samples the total bacterial population in the nitrification
tank ranged from 4.8  1012 to 3.8  1013 copies/L. These values
are the same order of magnitude as those obtained from
activated sludge samples of WWTPs (Limpiyakorn et al., 2005).
As the MLSS concentration increased as a result of SRT
control, the total bacterial population increased to
3.8  1013 copies/L in the seventh week, indicating an eight-
fold increase compared to that in the first week.
The concentration of AOB identified using the AOB 16S
rDNA assay gradually increased and the variation was not
large, compared to that of the NOB population. In the tenth
week, an approximately 9-fold increase was observed in the
Fig. 6 e Changes in copies per liter of (a) the total bacteria, AOB, Nitrobacter, and Nitrospira, (b) the total bacteria, narG, nirS,
nirK, nosZ in the full-scale activated sludge process during the study period.
5793
number of AOB 16S copies/L over the first week. This result
indicated that newly controlled SRT and the internal recycling
ratio may ensure continued AOB population. In addition, such
a gradual increase in the AOB population may affect the
increase in nitrification activity observed in the batch test. In
the ninth week when the nitrification performance of the full-
scale process was severely inhibited, there was no observed
decrease in the AOB number. However, any change in the AOB
number generally coincided with the variation of NO 2 N
concentration produced in the nitrification tank. Meanwhile,
the percentages of the AOB within the total bacteria varied
from 1.07 to 3.29% in the nitrification tank. This result is
similar to the values for activated sludge samples obtained
from systems treating industrial wastewater (0.01e9.3%)
(Layton et al., 2005). But we could not identify any correlation
with the nitrification activity.
We observed coexisting Nitrospira and Nitrobacter genera
for NOB. The Nitrospira and Nitrobacter populations in the
initial operating condition were similar, at 2.0  109 copies/L
and 2.8  109 copies/L, respectively. However, a shift to
Nitrobacter sp. in the NOB community was observed
throughout the study. The 16S rRNA gene concentration of the
Nitrobacter increased to a range of 2.8  109 to 4.8  1010 copies/
L, and the percentages of the Nitrobacter population within the
total bacteria also sharply increased from 0.03 to 0.16% in the
nitrification tank, as the MLSS concentration increased and
more nitrite was produced. Finally, the Nitrobacter populations
in the nitrifying system were consistently higher than the
Nitrospira populations throughout the study. On the other
hand, the number of Nitrospira gradually decreased from
2.0  109 to 1.1  109 copies/L, until the third week. When high
concentration of ammonia in the influent flowed into the
activated sludge process, a 5-fold increase was observed in the
number of Nitrospira 16S copies/L along with an increase in the
number of Nitrobacter sp. However, in the ninth week when
high concentrations of phenol flowed into the system, a sharp
decrease of Nitrospira population was observed. The
percentage of the Nitrospira population among all bacteria
shrank to 0.01%. Previous research has reported that Nitrospira
5794 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
were far more sensitive to toxic compounds than Nitrobacter
(Blackburne et al., 2007; Kim et al., 2011a). Meanwhile, despite
coexisting Nitrospira and Nitrobacter genera for NOB (Fig. 6a)
and generation of NO 3  N as the final product through
nitrification activity in the batch test (data not shown), low
concentrations of NO 3  N were produced in the nitrification
tank of the full-scale process (Fig. 2). This may reflect the short
retention time necessary for NOB to react with nitrite ions in
the full-scale process.
Meanwhile, the abundance of narG, nirS, nirK and nosZ genes
of denitrifying bacteria was investigated during the study in
the activated sludge process. The narG, nirS, nirK and nosZ
target molecules were less abundant than the 16S rDNA gene
copies for the total bacteria: total bacteria ranged from
9.9  1012 to 3.9  1013 copies/L; narG ranged from 2.0  109 to
9.2  109 copies/L; nirS ranged from 1.3  1012 to
1.7  1013 copies/L; nirK ranged from 1.1  1010 to
1.0  1011 copies/L; nosZ ranged from 2.1  1011 to
2.4  1012 copies/L (Fig. 6b). In the activated sludge process
treating industrial wastewater, gene copy numbers per liter of
the nirS gene exceeded those of the narG and nosZ genes for all
sampling points. This trend implies that there is a greater
abundance of genes for the nitrite reducing genes than for the
nitrate and nitrous oxide reducing genes. Meanwhile, for nirS,
the copy numbers of genes detected were much higher than
those for nirK at all sampling points. It is known that more
taxonomically diverse nirK denitrifiers are more sensitive to
environmental changes than the nirS denitrifiers; however, the
latter are more abundant (Yoshie et al., 2004). To evaluate the
ratio of denitrifiers relative to total bacteria, the percentages of
denitrification genes in proportion to 16S rDNA were calcu-
lated, resulting in proportions of approximately 0.02%, 26.9%,
0.30%, and 4.77% for narG, nirS, nirK, and nosZ genes, respec-
tively. The maximum amount of nirS relative to 16S rDNA was
43%, confirming the high proportion of denitrifiers to total
bacteria in this activated sludge process. On the other hand,
the smallest amount of the narG gene, the nitrate reducing
functional gene, in the anoxic tank for all samplings may result
from low concentrations of NO 3  N in the nitrified effluent.
Lastly, in the twelfth week, both nitrifying and denitrifying
populations decreased along with a decrease in the MLVSS/
MLSS ratio, resulting in a slight decrease in both nitrification
and denitrification efficiencies of the full-scale process.
However, in the batch tests both the nitrification and denitrifi-
cation rates remained steady. This may result from little loss of
active bacterial populations among total microorganisms. Also,
in actual environment conditions, there may be more factors
inhibiting the activity of microorganisms in the full-scale
process than in the batch test. Consequently, it was very diffi-
cult to identify any relationships between the bacterial pop-
ulations, their activity and the process performance in the full-
scale process. However, to prevent failure of the process
performance, it is important to monitor any sudden decrease in
the populations of important bacteria such as nitrifiers.
4. Conclusions
In a full-scale activated sludge process, SRT and internal
recycling ratio controls were selected as the main operating
parameters to improve TN removal efficiency. Increased
biomass concentration via SRT control enhanced TN removal.
In addition, decreasing the internal recycling ratio restored
nitrification activity which had been inhibited by phenol
shock loading. These results indicate that application of
specific operational strategies can change the physiological
state of the activated sludge process’s bacterial populations
(Hallin et al., 2005). Therefore, proper operational strategies
should be tailored to accommodate the possibility of erratic
changes in the composition of the influent via consistent
monitoring of its components to achieve and maintain
a stable process.
Acknowledgments
The authors thank David Nielsen for assistance during this
work. This research was supported by WCU (World Class
University) program through the National Research Founda-
tion of Korea funded by the Ministry of Education, Science and
Technology (R31-30005) and the Advanced Biomass R&D
Center (ABC) of Korea Grant funded by the Ministry of
Education, Science and Technology (ABC-2010-0029800). Also,
this work was financially supported by the second phase of
the Brain Korea 21 Program in 2011 and Korea Ministry of
Environment (MOE) as ’Human resource development Project
for Energy from Waste & Recycling’.
references
Amann, R.I., Binder, R.J., Olson, S., 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, 1919e1925.
APHA, 1998. Standard Methods for the Examination of Water and
Wastewater. APHA, AWWA, WPCF, twentieth ed. American
Public Health Association, Washington, DC, USA.
Blackburne, R., Vadivelu, V.M., Yuan, Z.G., Keller, J., 2007.
Kinetic characterisation of an enriched Nitrospira culture
with comparison to Nitrobacter. Water Research 41,
3033e3042.
Braker, G., Fesefeldt, A., Witzel, K.P., 1998. Development of PCR
primer systems for amplification of nitrite reductase genes
(nirK and nirS ) to detect denitrifying bacteria in environmental
samples. Applied and Environmental Microbiology 64,
3769e3775.
Dionisi, H.M., Layton, A.C., Harms, G., Gregory, I.R., Robinson, K.G.
, Sayler, G.S., 2002. Quantification of Nitrosomonas
oligotropha-like ammonia-oxidizing bacteria and Nitrospira
spp. from full-scale wastewater treatment plants by
competitive PCR. Applied and Environmental Microbiology 68,
245e253.
Eighmy, T.T., Bishop, P.L., 1989. Distribution and role of bacterial
nitrifying populations in nitrogen removal in aquatic
treatment systems. Water Research 23, 947e955.
Ferris, M.J., Muyzer, G., Ward, D.M., 1996. Denaturing gradient gel
electrophoresis profiles of 16S rRNA-defined populations
inhabiting a hot spring microbial mat community. Applied
and Environmental Microbiology 62, 340e346.
Figuerola, E.L., Erijman, L., 2010. Diversity of nitrifying bacteria in
a full-scale petroleum refinery wastewater treatment plant
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 8 5 e5 7 9 5
experiencing unstable nitrification. Journal of Hazardous
Matererials 181, 281e288.
Graham, D.W., Knapp, C.W., Van Vleck, E.S., Bloor, K., Lane, T.,
Graham, C.E., 2007. Experimental demonstration of chaotic
instability in biological nitrification. ISME Journal 1, 385e394.
Hallin, S., Lydmark, P., Kokalj, S., Hermansson, M., Sörensson, F.,
Jarvis, A., Lindgren, P.E., 2005. Community survey of
ammonia-oxidizing bacteria in full-scale activated sludge
processes with different solids retention time. Journal of
Applied Microbiology 99, 629e640.
Harms, G., Layton, A.C., Dionisi, H.M., Gregory, I.R., Garrett, V.M.,
Hawkins, S.A., Robinson, K.G., Sayler, G.S., 2003. Real-time PCR
quantification of nitrifying bacteria in a municipal wastewater
treatment plant. Environmental Science and Technology 37,
343e351.
Henry, S., Baudouin, E., López-Gutiérrez, J.C., Martin-Laurent, F.,
Brauman, A., Philippot, L., 2004. Quantification of denitrifying
bacteria in soils by nirK gene targeted real-time PCR. Journal of
Microbiological Methods 59, 327e335.
Henry, S., Bru, D., Stres, B., Hallet, S., Philippot, L., 2006.
Quantitative detection of the nosZ gene, encoding nitrous
oxide reductase, and comparison of the abundances of 16S
rRNA, narG, nirK, and nosZ genes in soils. Applied and
Environmental Microbiology 72, 5181e5189.
Hermansson, A., Lindgren, P.E., 2001. Quantification of ammonia-
oxidizing bacteria in arable soil by real time PCR. Applied and
Environmental Microbiology 67, 972e976.
Jimenez, J., Melcer, H., Parker, D., Bratby, J., 2011. The effect of
degree of recycle on the nitrifier growth rate. Water
Environment Research 83, 26e35.
Juliastuti, S.R., Baeyens, J., Creemers, C., 2003. Inhibition of
nitrification by heavy metals and organic compounds: the ISO
9509 test. Environmental Engineering Science 20, 70e90.
Kane, M.D., Poulsen, L.K., Stahl, D.A., 1993. Monitoring the
enrichment and isolation of sulfate-reducing bacteria by using
oligonucleotide hybridization probes designed from
environmentally derived 16s ribosomal RNA sequences.
Applied and Environmental Microbiology 59, 682e686.
Kim, Y.M., Park, D., Lee, D.S., Park, J.M., 2007. Instability of
biological nitrogen removal in a cokes wastewater treatment
facility during summer. Journal of Hazardous Matererials 141,
27e32.
Kim, Y.M., Park, D., Lee, D.S., Park, J.M., 2008. Inhibitory effects of
toxic compounds on nitrification process for cokes
wastewater treatment. Journal of Hazardous Matererials 152,
915e921.
Kim, Y.M., Park, D., Lee, D.S., Jung, K.A., Park, J.M., 2009. Sudden
failure of biological nitrogen and carbon removal in the full-
scale pre-denitrification process treating cokes wastewater.
Bioresource Technology 100, 4340e4347.
Kim, Y.M., Lee, D.S., Park, C., Park, D., Park, J.M., 2011a. Effects of
free cyanide on microbial communities and biological carbon
and nitrogen removal performance in the industrial activated
sludge process. Water Research 45, 1267e1279.
Kim, Y.M., Cho, H.U., Lee, D.S., Park, C., Park, D., Park, J.M., 2011b.
Response of nitrifying bacterial communities to the increased
thiocyanate concentration in pre-denitrification process.
Bioresource Technology 102, 913e922.
5795
Layton, A.C., Dionisi, H., Kuo, H.W., Robinson, K.G., Garrett, V.M.,
Meyers, A., Sayler, G.S., 2005. Emergence of competitive
dominant ammonia-oxidizing bacterial populations in a full-
scale industrial wastewater treatment plant. Applied and
Environmental Microbiology 71, 1105e1108.
Lee, C., Kim, J., Do, H., Hwang, S., 2008. Monitoring thiocyanate-
degrading microbial community in relation to changes in
process performance in mixed culture systems near washout.
Water Research 42, 1254e1262.
Limpiyakorn, T., Shiohara, Y., Kurisu, F., Yagi, O., 2005.
Communities of ammonia-oxidizing bacteria in activated
sludge of various sewage treatment plants in Tokyo. FEMS
Microbiology Ecology 54, 205e217.
Lin, C., Stahl, D.A., 1995. Comparative analyses reveal a highly
conserved endoglucanase in the cellulolytic genus
Fibrobacter. Journal of Bacteriology 177, 2543e2549.
López-Gutiérrez, J.C., Henry, S., Hallet, S., Martin-Laurent, F.,
Catrou, G., Philippot, L., 2004. Quantification of a novel
group of nitrate-reducing bacteria in the environment by
real-time PCR. Journal of Microbiological Methods 57,
399e407.
Lydmark, P., Almstrand, R., Samuelsson, K., Mattsson, A.,
Sörensson, F., Lindgren, P.E., Hermansson, M., 2007. Effects of
environmental condition on the nitrifying population
dynamics in a pilot wastewater treatment plant.
Environmental Microbiology 9, 2220e2233.
Manekar, P., Biswas, R., Karthik, M., Nandy, T., 2011. Novel two
stage bio-oxidation and chlorination process for high strength
hazardous coal carbonization effluent. Journal of Hazardous
Matererials. doi:10.1016/j.jhazmat.2011.02.006.
Mertoglu, B., Semerci, N., Guler, N., Calli, B., Cecen, B., Saatci, A.M.
, 2008. Monitoring of population shift in an enrich nitrifying
system under gradually increased cadmium loading. Journal
of Hazardous Matererials 160, 495e501.
Mobarry, B.K., Wagner, M., Urbain, V., Rittmann, B.E., Stahl, D.A.,
1996. Phylogenetic probes for analyzing abundance and
spatial organization of nitrifying bacteria. Applied and
Environmental Microbiology 62, 2156e2162.
Regan, J.M., Harrington, G.W., Noguera, D.R., 2002. Ammonia- and
nitrite-oxidizing bacterial communities in a pilot-scale
chloraminated drinking water distribution system. Applied
and Environmental Microbiology 68, 73e81.
Rittmann, B.E., McCarty, P.L., 2001. Environmental Biotechnology:
Principles and Applicatons. McGraw-Hill Science.
Siripong, S., Rittmann, B.E., 2007. Diversity study of nitrifying
bacteria in full-scale municipal wastewater treatment plants.
Water Research 41 (5), 1110e1120.
Teck, H.C., Loong, K.S., Sun, D.D., Leckie, J.O., 2009. Influence of
a prolonged solid retention time environment on nitrification/
denitrification and sludge production in a submerged
memebrane bioreactor. Desalination 245, 28e43.
Wagner, M., Rath, G., Amann, R., Koops, H.P., Schleifer, K.H., 1995.
In situ identification of ammonia-oxidizing bacteria.
Systematic and Applied Microbiology 18, 251e264.
Yoshie, S., Noda, N., Tsuneda, S., Hirata, A., Inamori, Y., 2004.
Salinity decreases nitrite reductase gene diversity in
denitrifying bacteria of wastewater treatment systems.
Applied and Environmental Microbiology 70, 3152e3157.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Reduced microbial attachment by D-amino acid-inhibited

AI-2 and EPS production

Huijuan Xu a, Yu Liu a,b,*

a
Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering,
Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore
b
Advanced Environmental Biotechnology Centre, Nanyang Environment & Water Research Institute,
Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore

article info abstract

Article history: This study investigated the effect of D-tyrosine on microbial attachment to hydrophilic
Received 21 March 2011 glass and hydrophobic polypropylene surfaces. Results showed that D-tyrosine did not
Received in revised form influence microbial growth, ATP and substrate utilization, but significantly inhibited the
22 August 2011 synthesis of autoinducer-2 (AI-2), eDNA and extracellular polysaccharides and proteins,
Accepted 29 August 2011 and subsequently reduced microbial attachment onto glass and polypropylene surfaces
Available online 3 September 2011 was observed. It was shown that D-amino acid would be a non-toxic agent for control of
microbial attachment.
Keywords: ª 2011 Elsevier Ltd. All rights reserved.
Microbial attachment
D-amino acid
Autoinducer-2
ATP
eDNA
EPS

1. Introduction those D-amino acids in stationary phase, which may regulate

the chemistry of the cell wall through slow production of
It had been considered that D-amino acids are excluded from peptidoglycan that is crucial for cell wall (Lam et al., 2009). It
living systems except for D-amino acids in the cell wall of has been reported that many bacteria would produce various
microorganisms. D-amino acids have been discovered in many D-amino acids just before biofilm disassembly and the release

physiological processes. The best described of D-amino acids
may be their involvement in the formation of the peptido-
glycan. Both the thick cell wall of Gram-positive bacteria and
much thinner cell wall of Gram-negative bacteria consist of
peptidoglycan which contain D-amino acids. Besides compo-
nents of bacterial cell wall, D-amino acid have been known to
regulate bacterial germination and to be incorporated into
peptides (Wood et al., 2011). The bacteria begin to synthesize
of rapid diffused small molecule D-amino acid could be
a signal to coordinate the whole population action to different
environment (Kolodkin-Gal et al., 2010).
During biofilm formation, microorganisms need first attach
onto a solid surface, followed by secretion of extracellular
polymeric substances (EPS) (Flemming and Wingender, 2010),
whereas other factors have also been reported to be essentially
involved in biofilm development. For example, both Gram-

* Corresponding author. Division of Environmental and Water Resources Engineering, School of Civil and Environmental Engineering,
Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore. Tel.: þ65 67 905 254; fax: þ65 67 910 676.
E-mail address: cyliu@ntu.edu.sg (Y. Liu).
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.061
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4
negative and Gram-positive bacteria could communicate
through small signaling molecules to coordinate population
behavior (Bassler, 1999). Autoinducer-2 (AI-2) is a species-
nonspecific signal molecule found in both Gram-negative and
Gram-positive bacteria, and can influence a mixed-species
biofilm formation between Porphyromonas gingivalis and Strep-
tococcus gordonii, and it was found that production of AI-2 by
either species was sufficient for inter-species communication
and biofilm formation (McNab et al., 2003). ATP provides a pool
of energy for most energy-consuming microbial activities, such
as signaling molecules secretion, EPS synthesis, motility and
flagella etc. In addition, the role of EPS in the formation of
biofilms has been well documented (Flemming and Wingender,
2010).
Although the role of D-amino acids in pure culture biofilm
dispersal has been reported (Kolodkin-Gal et al., 2010), little is
currently known about the effects of exogenous D-amino acids
on the formation of mixed-culture biofilm and synthesis of
cellular ATP, AI-2, eDNA and EPS, which are all essential for
biofilm development on a solid surface. Furthermore, the
present study focused more on the interaction between
D-amino acid and AI-2-mdeiated cellular communication. For
this purpose, a typical D-amino acid, D-tyrosine was used due
to its potent activity (Kolodkin-Gal et al., 2010). Therefore, this
study aimed to investigate how D-tyrosine could affect
attachment of mixed-culture microorganisms onto hydro-
phobic PP and hydrophilic glass surfaces through determining
changes in surface charge, ATP, AI-2, eDNA and EPS. It is ex-
pected that this study can offer an alternative approach for
biological control of microbial attachment on various solid
surfaces including membrane.
2. Materials and methods
2.1. Carriers for microbial attachment
Glass slides with the dimension of 24  50 mm (CEP, SPD
Scientific, Singapore) and 24  50 mm polypropylene (PP)
coupons (Kinary, Singapore) were used as biocarriers in
microbial attachment experiments. The PP coupons were
cleaned with detergent and rinsed thoroughly with distilled
water, whereas the glass slides were cleaned by being soaked
in 10% nitric acid for 24 h, and were then thoroughly rinsed
with distilled water and dried. The hydrophobicity of the
carrier surface was characterized by contact angle that was
measured using a contact angle goniometer (dataphysics OCA
20, Filderstadt, Germany). Eight measurements were made on
triplicate samples. The average water drop contact angle for
clean glass slide was 16.9  0.5 and 99.3  2.2 for PP coupons,
i.e. the PP coupons are highly hydrophobic, and hydrophilic
for glass slides.
2.2. Microbial attachment assay
Activated sludge microorganisms were taken from a local
wastewater treatment plant and acclimated with a synthetic
substrate for one month. The synthetic substrate consisted of
690 mg l1 of sodium acetate and 240 mg l1 ethanol as carbon
source, 200 mg l1 NH4Cl, 60 mg l1 K2HPO4, 15 mg l1
5797
CaCl2·2H2O, 12.5 mg l1 MgSO4·7H2O and 20 mg l1 FeSO4·7H2O
(Liu et al., 2003). Experiments were designed to investigate the
effect of D-tyrosine on microbial growth and attachment
potentials. Thus, microorganisms with and without exposure
to D-tyrosine for different times were used in 1-h static
microbial attachment experiments conducted under the same
conditions, as detailed below: (i) two series of batch experi-
ments were conducted: one served as control free of D-tyrosine,
while the other was added with 6 mg l1 of D-tyrosine (Sig-
maeAldrich, St. Louis, MO, USA); (ii) suspended biomass
cultivated with and without exposure to D-tyrosine was
collected at different exposure times of 1e4 h for 1-h microbial
attachment assay and determination of surface charge, cellular
ATP, AI-2, eDNA and EPS. The static microbial attachment was
conducted in Petri dishes mounted with one PP coupon and
one glass slide on the bottom. Suspended microorganisms
harvested from the batch reactors at different exposure times
were resuspended in 30 ml of 10 mM phosphate buffered saline
(PBS) solution with 100 mg dry biomass l1 and were made
contact with carriers for 1 h in Petri dishes. After attachment,
carriers were gently rinsed three times with distilled water to
remove loosely attached microorganisms. Fixed biomass was
quantified in terms of TOC by a TOC analyzer (ASI-V, TOC-
Vcsh, Shimadzu, Japan).
2.3. Surface charge
The colloid titration method was used to determine surface
charge of suspended microorganisms with and without
exposure to D-tyrosine (Wilen et al., 2003). Polybrene (Sigma-
eAldrich) was used as positive colloidal reagent and polyvinyl
sulfate potassium salt (PVSK) (SigmaeAldrich) as negative
reagent. For titrating negatively charged suspended microor-
ganisms, 5 ml of 0.001 N polybrene was added to the sample.
The excess polybrene was back titrated with 0.0005 N PVSK
using 100 ml of 0.1% toluidine blue (SigmaeAldrich) as the end-
point indicator. Titration was terminated when the color
changed from blue to pink, indicating that electrical neutrality
was reached. Equal volumes of polybrene in distilled water
were used as blanks. The surface charge expressed as mill
equivalents per gram of dry biomass can be determined from
the equation given below.

1000ðA  BÞN
Charge meq g1 SS ¼ (1)
XV
Where A is the volume of PVSK added to the sample (ml), B is
the volume of PVSK added to the blank (ml), N is the normality
of PVSK solution used (0.0005 N), V is the volume of the sample
(ml), X is the biomass concentration of the sample (g L1).
2.4. Determination of cellular ATP
The cellular ATP were extracted from freshly collected bio-
samples according to the trichloroacetic acid (TCA) method
(Chen and Leung, 2000) with some modifications. Five milli-
liter of the bacterial suspension was added into 5 ml of 5% TCA
solution, and was homogenized with an ultrasonic homoge-
nizer (Sonics & Materials, Newton, CT, USA) for 3 min.
Aliquots of 0.5 ml homogenized suspension were diluted by
ten times with Tris-Acetate-EDTA (TAE) buffer (Bio-Rad,
5798 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4
Singapore) to adjust the pH of about 7.75. The mixture was
filtered through 0.2 mm syringe filters and the collected filtrate
was stored at 20  C for further use. ATP concentration was
determined according to firefly luciferin-luciferase biolumi-
nescence method with the FLAA Adenosine 50 -triphosphate
(ATP) Bioluminescent Assay Kit (SigmaeAldrich, St. Louis, MO,
USA) as recommended in the instruction. The intensity of
luminescence was measured by a TD-20/20 Luminometer
(Turner Designs, Sunnyvale, CA, USA).
2.5. Autoinducer-2 measurement
For determination of AI-2, 10 ml suspended microorganisms
were collected and resuspended in fresh autoinducer bioassay
(AB) medium (Surette and Bassler, 1998). The resuspended
sample was well mixed and filtered through 0.2 mm syringe
filter. The filtrate was collected and stored at 20  C. The cell-
free culture supernatant was thawed before determination of
AI-2 concentration. The amount of AI-2 was measured by Vibrio
harveyi BB170 (ATCC BAA 1117) bioluminescence reporter assay
(Rickard et al., 2008). The reporter strain V. harveyi BB170
(ATCC, Manassas, VA, USA) was cultured in fresh AB medium
for 13e16 h with shaking at 30  C and then diluted 1:5000 with
fresh AB medium. One hundred and eighty microliter of the
diluted cells was added to the well of 96-well plate containing
20 ml cell-free supernatant to be tested for AI-2 activity. The 96-
well plate was incubated in a rotary shaker at 30  C. The
intensity of luminescence was measured hourly using
a microplate reader. The fold induction was converted to the
molar concentration of AI-2 by comparing the fold induction
and 4,5-dihydroxy-2,3-pentanedione (DPD) concentration, as
DPD (Omm Scientific, Dallas, USA) were used as the calibration
standard. Each filtrated sample was assayed six times in
parallel and the mean values reported.
2.6. Response of AI-2 reporter strain to D-tyrosine
In order to further investigate the effect of D-tyrosine on AI-2
repression, an AI-2 bioluminescence assay in presence and
absence of D-tyrosine was conducted. In this assay, five
thousand times diluted reporter strain V. harveyi BB170 as
described before was grown in fresh AB medium supple-
mented with 0.1e0.7 mM of DPD in the wells of a 96-well plate.
Two series of experiments were conducted: for control, the
wells were free of D-tyrosine; other wells were added with
6 mg l1 D-tyrosine. The 96-well plate was shaken in a rotary
shaker at 30  C. The light intensity was assayed over time
using a microplate reader until get the maximum fold induc-
tion of bioluminescence.
2.7. Extraction and quantification of eDNA
Extracellular DNA was extracted from suspended microorgan-
isms sample according to (Steinberger and Holden, 2005) with
modification. Five milliliter of bacterial suspension was
collected and resuspended in 0.9% NaCl solution. The resus-
pended sample was well mixed with a homogenizer (Sonics &
Materials, CT, USA). Treated cell solution was filtered through
0.2 mm syring filter and the collected filtrate was stored at 20  C
for determining eDNA concentration. The concentration of
DNA was measured by using PicoGreen dsDNA Quantification
Kit (Molecular Probes, Invitrogen, Eugene, OR, USA) following
the protocol provided by the kit. The calf thymus DNA was used
as the standard. The fluorescence intensity was recorded by
a microplate reader (BioTek, sygnergy 2, VT, USA).
2.8. Determination of extracellular polysaccharides and
proteins
Extracellular polysaccharides (PS) and proteins (PN) were
extracted from biosample by modified cold aqueous tech-
nique method (Jia et al., 1996). Ten milliliter of suspended
microorganisms was washed twice with distilled water and
centrifuged at 3500 rpm for 10 min. The settled biomass was
recovered and resuspended in 10 ml of 8.5% NaCl and 0.22%
formaldehyde solution. The mixture was homogenized for
2 min using an ultrasonic homogenizer (Sonics & Materials,
CT, USA) in an ice-water bath, and then was centrifuged at
10,000 rpm for 30 min to remove solid residues. The super-
natant was harvested for PS, PN measurement and high
performance size exclusion chromatography analysis. PS was
determined by the phenol-sulfuric acid method (Dubois et al.,
1956), whereas PN was analyzed by the modified Lowry
method (Lowry et al., 1951). Glucose and bovine serum
albumin (SigmaeAldrich) were used as the standards for PS
and PN, respectively.
High performance size exclusion chromatography (HPSEC)
analysis was carried out with Series 200 HPLC system (Perkin
Elmer, Waltham, MA, USA) equipped with a Series 200LC
quaternary pump, Series 200 autosampler, a Perkin Elmer 600
interface and a UV/Vis detector (785A). A 300  7.8 mm size
exclusion chromatography column BioSep SEC S2000 (Phe-
nomenex, Torrance, CA) was used. The mobile phase con-
sisted of 9.0 mM NaCl and 0.9 mM Na2HPO4 at pH 7.0 (Comte
et al., 2007). Extracellular polymeric substances were extrac-
ted from suspended microorganisms as mentioned above and
all samples were filtered through 0.20 mm filters prior to
injection. All measurements were conducted at 25  C, mobile
phase flow 1.0 ml min1, the sample injection 100 ml. The
detection was carried out with a UV detector at 280 nm.
2.9. Staining and visualization
In order to visualize microbial attachment, the adherent
bacteria on glass slides and PP coupons surfaces were stained
with LIVE/DEAD BacLight Bacterial Viability kits (Molecular
Probes, Eugene, OR, USA), which consisted of two nucleic acid
dyes staining on both live and dead cells: SYTO 9 and propi-
dium iodide (PI). SYTO 9 is a green-fluorescent dye which stains
both live and dead bacteria with intact and damaged cell
membranes while the red-fluorescing PI only stains dead
bacteria with damaged cell membranes. The excitation/emis-
sion maxima for these dyes are about 480/500 nm for SYTO 9
stain and 490/635 nm for PI. With an appropriate mixture of
both dyes, viable bacteria with intact cell membranes are
stained green, whereas bacteria with damaged cell membranes
fluoresce red. The color assigned to the live and dead cells
follows from the color at which the stained cells fluoresce
under laser excitation. The sample staining procedure was
carried out following the instructions in the manual. First, two
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4
hundred microliters of the mixed solution (1000 times diluted
SYTO 9 and PI from the stock solution) was added to each
attachment sample on glass slides and PP coupons. The stained
sample was then incubated in the dark at room temperature
for 15 min. After that, the sample was gently rinsed two times
with DI water to remove unbound dyes. Finally, the sample was
covered with cover slip and viewed using an Olympus Fluoview
FV300 confocal laser scanning microscopy (CLSM) (Olympus
Optical, Tokyo, Japan) with a 100X objective.
2.10. Statistical analysis
All tests were performed in triplicate otherwise stated. Results
were expressed as mean value  absolute deviation. Student t-
tests were employed for analyzing the significance of results
at the level of P < 0.05.
3. Results
3.1. Microbial attachment on glass and PP
Fig. 1a shows that attachment of microorganisms exposed to
6 mg l1 D-tyrosine was reduced significantly on glass slides
compared to the control free of D-tyrosine (Student’s t-test,
P < 0.05). After 2-h culture, attachment of microorganisms
without exposure to D-tyrosine was 10.8 mg TOC cm2 on the
glass surface, while attachment of microorganisms with
exposure to D-tyrosine was 8.5 mg TOC cm2, indicating 22%
reduction in microbial attachment caused by D-tyrosine.
Similar phenomenon was also observed in microbial attach-
ment on PP surface (Fig. 1b). These results indeed are sup-
ported by the microscopic observations (Fig. 2). It should be
noted that mixed-culture microorganisms with and without
exposure to D-tyrosine were collected at different culture
(exposure) times of 1e4 h, and used for 1-h microbial attach-
ment assays, as shown in Fig. 1. According to substrate
availability over 4-h culture, 1 h- and 2-h old microorganisms
were basically in earlier exponential and post exponential
growth phases, while 3 h- and 4 h-old microorganisms already
entered into earlier stationary and post-stationary growth
phases, respectively. It is reasonable to consider that micro-
organisms at different growth states would have different
attachment abilities as observed in the control assays (Fig. 1).
This view is strongly supported by the findings of Fletcher
Fig. 1 e Attachment of microorganisms with (-) and without (,) treatment by D-tyrosine on glass slides (a); on PP coupons
(b). Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.
5799
(1977), showing that the number of attached cells harvested
from exponential growth phase was greatest, followed by
those from stationary and decay phases. In addition, Fig. 3a
shows that addition of D-tyrosine to the culture media had no
negative effect on the TOC removal efficiency. It was shown in
Fig. 3b that the suspended biomass concentration increased
from 450 mg l1 to 630 mg l1 in the cultures supplemented
with and without D-tyrosine. These suggest that D-tyrosine is
not inhibitory to substrate utilization and microbial growth at
the concentration studied. Fig. 4 shows that suspended
microorganisms with and without exposure to D-tyrosine both
carried negative surface charge, but microorganisms exposed
to D-tyrosine carried more negative surface charge compared
to that of control free of D-tyrosine.
3.2. Cellular ATP and AI-2 contents of suspended
microorganisms
Fig. 5a showed the possible effect of D-tyrosine on energy
metabolism of suspended microorganisms. It can be seen in
Fig. 5a that the cellular ATP content did not change signifi-
cantly in the presence of D-tyrosine compared to that of
control. D-tyrosine thus does not appear to inhibit the ATP
synthesis. AI-2 as inter-species signaling molecules coordi-
nates the formation of biofilm by various species (Rickard
et al., 2008). To investigate the effect of D-tyrosine on cellular
communication, Fig. 5b showed the respective AI-2 content of
suspended microorganisms with and without D-tyrosine
addition. After 1 h culture, the AI-2 content of suspended
microorganisms without exposure to D-tyrosine was about
0.27 nmol mg1 in the control, while it decreased to about
0.19 nmol mg1 for suspended microorganisms with exposure
to D-tyrosine, i.e. D-tyrosine could suppress the synthesis or
secretion of AI-2. The response of reporter strain V. harveyi
BB170 to D-tyrosine was further studied, and results were
presented in Fig. 6, showing fold induction of luminescence in
presence and absence of D-tyrosine, and obviously lumines-
cence was suppressed significantly when the culture was
supplemented with D-tyrosine (Student’s t-test, P < 0.05). For
example, for 0.4 mM DPD, the fold induction of luminescence
was 23.6 in the control free of D-tyrosine, whereas it decreased
to 12.4 in the media with addition of D-tyrosine, indicating 48%
reduction as compared to that of control. These imply that D-
tyrosine at the concentration studied had an inhibitory effect
on AI-2 expression.
5800 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Fig. 2 e CLSM images of attachment of microorganisms without D-tyrosine treatment on glass slides (A) and PP coupons (C)
at the exposure time of 1 h; with D-tyrosine treatment on glass slides (B) and PP coupons (D) at the exposure time of 1 h.

3.3. EPS production of suspended microorganisms
EPS are composed of a variety of organic substances, in which
polysaccharides and proteins are two major components, and
play an important role in microbial attachment onto a solid
surface (Flemming and Wingender, 2010). Fig. 7 shows the
respective contents of extracellular polysaccharide (PS) and
protein (PN) in microorganisms with and without exposure to
D-tyrosine. As compared to the control, a 31% reduction in PN
and 17% decrease in PS were observed in microorganisms
after 1 h exposure to D-tyrosine, leading to a lowered PN/PS
ratio. eDNA is a unique component of the organic substances
in the matrix of suspended microorganisms. Fig. 8 shows
eDNA content of suspended microorganisms with and
without exposure to D-tyrosine. After 1 h exposure to D-tyro-
sine, eDNA was reduced to 0.006 mg g1 biomass, i.e. a 68%
reduction compared to the control. These results suggest that
D-tyrosine would significantly inhibit eDNA secretion.

Fig. 3 e Profiles of TOC removal efficiency (a) and microbial growth (b) in the cultures supplemented with (C) and without (B)
D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4 5801

Fig. 4 e Surface charge of suspended microorganisms with Fig. 6 e Response of AI-2 reporter strain in the cultures
(C) and without (B) exposure to D-tyrosine. Each point supplemented with (C) and without (B) D-tyrosine. Each
represents the mean of triplicate measurements and error point represents the mean of six measurements in parallel
bar is absolute deviation from the mean. and error bar is absolute deviation from the mean.

Fig. 9 shows the HPSEC spectra of the EPS extracted from
microorganisms with and without exposure to D-tyrosine. In
general, retention time in HPSEC spectrum reflects molecular
weight of a target chemical, i.e. the peak of a chemical with
higher molecular weight appears quicker. For EPS extracted
from microorganisms exposed to D-tyrosine, the peaks
showed a significant decrease in area compared with those of
the control, indicating a lowered EPS production that is
consistent with the results in Fig. 7. Two peaks were observed
for the EPS extracted from microorganisms without exposure
to D-tyrosine at the retention time of 10e13 min, while only
one peak appeared for the EPS from microorganisms exposed
to D-tyrosine. This implies that EPS with higher molecular
weight was reduced due to exposure to D-tyrosine. These
suggest that D-tyrosine would not only inhibit the EPS
production, but also can alter the EPS composition.
4. Discussion
Figs. 1 and 2 show that microbial attachments onto hydro-
phobic PP and hydrophilic glass surfaces were inhibited by
D-tyrosine. So far, little information is available for the role of D-
tyrosine in control of mixed-culture biofilm development.
Kolodkin-Gal et al. (2010) reported that D-tyrosine-triggered
release of amyloid fibers from cell surface would suppress
development of a pure culture biofilm. It has been known that
the long amyloid fiber facilitates the anchoring of cells to
various surfaces, which is essential for microbial attachment
and biofilm formation. In addition, the PP coupons used in this
study have a contact angle of 99.3  2.2 , while 16.9  0.5 for
glass slides. As observed in Figs. 1 and 2, more microorganisms
attached to hydrophobic PP than to hydrophilic glass slide. In
fact, it has been well documented that higher hydrophobicity
of a solid surface would favor microbial attachment (Liu et al.,
2004); on the contrary, coating surfaces with non-charged
hydrophilic polymers resulted in reduced cell adsorption on
a variety of surfaces (Park et al., 1998).
Fig. 3 showed that D-tyrosine did not appear to affect the
biomass growth and substrate removal efficiency. Such
observation is consistent with the results obtained from pure
culture experiments (Kolodkin-Gal et al., 2010). It had been
reported that D-amino acids at a concentration higher than
20 mg l1 would inhibit bacterial growth (Teeri and Josselyn,

Fig. 5 e Cellular ATP content (a) and AI-2 concentration (b) in the cultures supplemented with (-) and without (,)
D-tyrosine. Each point represents the mean of triplicate measurements and error bar is absolute deviation from the mean.
5802 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4

Fig. 7 e PS (a) and PN (b) contents of suspended microorganisms with (-) and without (,) exposure to D-tyrosine. Each point
represents the mean of triplicate measurements and error bar is absolute deviation from the mean.

1953). Peptidoglycan is an important mesh-like polymer
component of cell wall. In the peptidoglycan polymer, there are
two unique amino acids at the terminal of a peptide side chain
of peptidoglycan: D-alanine as opposed to its isomer L-alanine.
D-tyrosine can replace D-alanine in the peptide side chain of cell
wall (Lam et al., 2009), and further alter the cell wall-building
protein so that the peptidoglycan production would be
slowed down, i.e. D-amino acid negatively regulate the amount
of peptidoglycan production. In the presence of D-methionine,
peptidoglycan synthesis could be severely inhibited, whereas
biomass continued to grow (Caparros et al., 1992). The amount
of peptidoglycan per cell decreased significantly due to the
increased biomass and decreased peptidoglycan synthesis.
However, it had been reported that Escherichia coli would be able
to grow properly with 60% decrease of the normal peptido-
glycan content (Prats and De Pedro, 1989). The unchanged
cellular ATP synthesis in microorganisms with and without
exposure to D-tyrosine (Fig. 5a) strongly supports this.
It appears from Fig. 5b that AI-2 content of suspended
microorganisms decreased due to D-tyrosine in the culture. To
exclude other factors affecting AI-2 quorum sensing, the
response of reporter strain V. harveyi BB170 to D-tyrosine
showed that the inhibition of AI-2 regulated bioluminescence
was only observed in presence of D-tyrosine (Fig. 6), which
further confirmed that AI-2 expression was suppressed due to
the presence of D-tyrosine. In study of D-amino acids regulated
cell wall remodeling, Lam et al. (2009) found that exogenous D-
methionine produced by Vibrio cholera were incorporated into
E. coli at the same position in the peptide even though E. coli
bacterium did not produce or release D-amino acids. Thus,
these rapid diffused small D-amino acids molecules could
regulate cells releasing them and the neighboring cells of
different species. In study of biofilm inhibition by D-amino
acids, Kolodkin-Gal et al. (2010) hypothesized that D-amino
acid may play an important role of chemical signal, but
opposite to quorum sensing signal molecule, to mediate inter-
species communication for facilitating cell dispersion from
biofilm. In addition, AI-2 is known to be a cellular communi-
cation signal molecule both for Gram-negative and Gram-
positive bacteria and has a positive effect on biofilm forma-
tion (Federle and Bassler, 2003). Due to the opposite effects of
these two signal molecules on biofilm formation, D-amino acid

Fig. 8 e eDNA of suspended microorganisms with (C) and

without (B) exposure to D-tyrosine. Each point represents Fig. 9 e HPSEC chromatograms of EPS extracted from
the mean of triplicate measurements and error bar is suspended microorganisms with (---) and without (d)
absolute deviation from the mean. exposure to D-tyrosine at the exposure time of 1 h.
 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4
may co-coordinate AI-2 regulated quorum sensing as shown
in Fig. 6. However, how these two signals may co-regulate
each other needs further investigation. Furthermore, it is
speculated that D-amino acids may play a coordinating role
through regulating the gene expression to control biofilm
community structure or induce biofilm dispersion.
Fig. 7 shows that the PS and PN contents of suspended
microorganisms tended to decrease with exposure to D-tyro-
sine. Tsuruoka et al. (1984) also observed that D-amino acid
caused reduction of lipoprotein in study of D-amino acid
incorporation into peptidoglycan. It has been reported that an
incorporation of D-tyrosine into the cellular proteins of Bacillus
subtilis (Champney and Jensen, 1970) and E. coli (Miyamoto
et al., 2010). As the D-isomer has a similar shape and size to
the L-isomer molecule, the D-analog incorporated into proteins
in the place of the natural amino acid would modify the
structure of the proteins and the enzymic activity (Richmond,
1962). These would eventually lead to the reduced production
of PS and PN. As can be seen in Fig. 9, high molecular-weight
EPS at the retention time of 10e13 min disappeared in micro-
organisms exposed to D-tyrosine. In fact, in study of the effect
D-amino acid on structure and synthesis of peptidoglycan,
Caparros et al. (1992) also found a direct inhibition of
D-methionine on the production of high molecular-weight
proteins. In addition, EPS have been believed to play an
important role in microbial attachment. The reduced produc-
tion of PS and PN would result in inhibited microbial attach-
ment (Fig. 1). Oliveira et al. (1994) reported that extracellular
polysaccharides could promote a preconditioning of the
surface, making attachment more favorable, whereas Flint
et al. (1997) found that treatment the cells with trypsin or
sodium dodecyl sulfate to remove cell surface proteins resul-
ted in a 100-fold reduction in the attachment of Thermophilic
streptococci onto stainless steel.
It had been shown that eDNA would play an important role
in initial microbial adhesion to hydrophobic and hydrophilic
surfaces (Das et al., 2010). Many studies have shown that eDNA
is an important component of extracellular network that
mediates cellecell and cellesurface interactions (Bockelmann
et al., 2006; Das et al., 2010). As can be seen in Fig. 8, the pres-
ence of D-tyrosine in the culture media caused reduction of
DNA in the extracellular network, as the result, less attach-
ment was observed both on glass and PP surface (Fig. 1). These
suggest that eDNA may facilitate microbial attachment onto
both hydrophilic glass and hydrophobic PP surfaces. Further
study is needed to elucidate how D-amino acid would regulate
eDNA production. In study of the role of eDNA in Listeria mon-
ocytogenes attachment, Harmsen et al. (2010) observed that
peptidoglycan, specifically N-acetylglucosamine, together with
eDNA could induce adhesion. Inhibited production of pepti-
doglycan and subsequently eDNA by D-tyrosine would also be
responsible for reduced microbial attachment (Fig. 1). It
appears from Fig. 4 that microorganisms carried more negative
surface charge in presence of D-tyrosine. Since EPS often have
charged functional groups, the higher negative charge density
would be associated with changes in the composition and
quantity of EPS induced by D-tyrosine. According to DLVO
theory, increased negative charge would lead to strong elec-
trostatic repulsion between cell and approaching surface (Zita
and Hermansson, 1994). Hence, the extent of attachment
5803
would be less (Fig. 1) as a result of increased surface charge or
a greater level of electrostatic repulsion.
5. Conclusions
This study showed that D-tyrosine as a typical D-amino acid
could inhibit microbial attachment on both hydrophilic glass
and hydrophobic PP surfaces, while no inhibitory effect on
microbial growth, ATP synthesis and substrate utilization was
observed. It was further found that the synthesis of AI-2,
eDNA and EPS were all reduced in the presence of D-tyrosine
in the culture media. These in turn provide a plausible
explanation for the D-tyrosine-triggered reduction in micro-
bial attachment and demonstrate a mean for biological
control of microbial attachment on a solid surface.
references
Bassler, B.L., 1999. How bacteria talk to each other: regulation of
gene expression by quorum sensing. Current Opinion in
Microbiology 2 (6), 582e587.
Bockelmann, U., Janke, A., Kuhn, R., Neu, T.R., Wecke, J.,
Lawrence, J.R., Szewzyk, U., 2006. Bacterial extracellular DNA
forming a defined network-like structure. FEMS Microbiology
Letters 262 (1), 31e38.
Caparros, M., Pisabarro, A.G., Depedro, M.A., 1992. Effect of D-
amino acids on structure and synthesis of peptidoglycan in
Escherichia-coli. Journal of Bacteriology 174 (17), 5549e5559.
Champney, W.S., Jensen, R.A., 1970. Molecular events in the
growth inhibition of Bacillus subtilis by D-tyrosine. Journal of
Bacteriology 104 (1), 107e116.
Chen, G.H., Leung, D.H.W., 2000. Utilization of oxygen in
a sanitary gravity sewer. Water Research 34 (15), 3813e3821.
Comte, S., Guibaud, G., Baudu, M., 2007. Effect of extraction
method on EPS from activated sludge: an HPSEC investigation.
Journal of Hazardous Materials 140 (1e2), 129e137.
Das, T., Sharma, P.K., Busscher, H.J., van der Mei, H.C., Krom, B.P.,
2010. Role of extracellular DNA in initial bacterial adhesion
and surface aggregation. Applied and Environmental
Microbiology 76 (10), 3405e3408.
Dubois, M., Gilles, K.A., Hamilton, J.K., Rebers, P.A., Smith, F.,
1956. Colorimetric method for determination of sugars and
related substances. Analytical Chemistry 28 (3), 350e356.
Federle, M.J., Bassler, B.L., 2003. Interspecies communication in
bacteria. Journal of Clinical Investigation 112 (9), 1291e1299.
Flemming, H.C., Wingender, J., 2010. The biofilm matrix. Nature
Reviews Microbiology 8 (9), 623e633.
Fletcher, M., 1977. Effects of culture concentration and age, time,
and temperature on bacterial attachment to polystyrene.
Canadian Journal of Microbiology 23 (1), 1e6.
Flint, S.H., Brooks, J.D., Bremer, P.J., 1997. The influence of cell
surface properties of Thermophilic streptococci on attachment to
stainless steel. Journal of Applied Microbiology 83 (4), 508e517.
Harmsen, M., Lappann, M., Knochel, S., Molin, S., 2010. Role of
extracellular DNA during biofilm formation by Listeria
monocytogenes. Applied and Environmental Microbiology 76 (7),
2271e2279.
Jia, X.S., Furumai, H., Fang, H.H.P., 1996. Yields of biomass and
extracellular polymers in four anaerobic sludges.
Environmental Technology 17 (3), 283e291.
Kolodkin-Gal, I., Romero, D., Cao, S.G., Clardy, J., Kolter, R.,
Losick, R., 2010. D-amino acids trigger biofilm disassembly.
Science 328 (5978), 627e629.
5804 w a t e r r e s e a r c h 4 5 ( 2 0 1 1 ) 5 7 9 6 e5 8 0 4
Lam, H., Oh, D.-C., Cava, F., Takacs, C.N., Clardy, J., de Pedro, M.A.,
Waldor, M.K., 2009. D-amino acids govern stationary phase
cell wall remodeling in bacteria. Science 325 (5947), 1552e1555.
Liu, Y., Yang, S.F., Li, Y., Xu, H., Qin, L., Tay, J.H., 2004. The influence
of cell and substratum surface hydrophobicities on microbial
attachment. Journal of Biotechnology 110 (3), 251e256.
Liu, Y., Yang, S.F., Tay, J.H., 2003. Elemental compositions and
characteristics of aerobic granules cultivated at different
substrate N/C ratios. Applied Microbiology and Biotechnology
61 (5-6), 556e561.
Lowry, O.H., Rosebrough, N.J., Farr, A.L., Randall, R.J., 1951.
Protein measurement with the folin phenol reagent. Journal of
Biological Chemistry 193 (1), 265e275.
McNab, R., Ford, S.K., El-Sabaeny, A., Barbieri, B., Cook, G.S.,
Lamont, R.J., 2003. LuxS-based signaling in Streptococcus
gordonii: autoinducer 2 controls carbohydrate metabolism and
biofilm formation with Porphyromonas gingivalis. Journal of
Bacteriology 185 (1), 274e284.
Miyamoto, T., Sekine, M., Ogawa, T., Hidaka, M., Homma, H.,
Masaki, H., 2010. Detection of D-amino acids in purified
proteins synthesized in Escherichia coli. Amino Acids 38 (5),
1377e1385.
Oliveira, R., Melo, L., Oliveira, A., Salgueiro, R., 1994.
Polysaccharide production and biofilm formation by
Pseudomonas fluorescens: effects of pH and surface material.
Colloids and Surfaces B-Biointerfaces 2 (1e3), 41e46.
Park, K.D., Kim, Y.S., Han, D.K., Kim, Y.H., Lee, E.H.B., Suh, H.,
Choi, K.S., 1998. Bacterial adhesion on PEG modified
polyurethane surfaces. Biomaterials 19 (7e9), 851e859.
Prats, R., De Pedro, M.A., 1989. Normal growth and division of
Escherichia coli with a reduced amount of murein. Journal of
Bacteriology 171 (7), 3740e3745.
Richmond, M.H., 1962. The effect of amino acid analogues on
growth and protein synthesis in microorganisms.
Bacteriological Reviews 26 (4), 398e420.
Rickard, A.H., Campagna, S.R., Kolenbrander, P.E., 2008.
Autoinducer-2 is produced in saliva-fed flow conditions
relevant to natural oral biofilms. Journal of Applied
Microbiology 105 (6), 2096e2103.
Steinberger, R.E., Holden, P.A., 2005. Extracellular DNA in single-
and multiple-species unsaturated biofilms. Applied and
Environmental Microbiology 71 (9), 5404e5410.
Surette, M.G., Bassler, B.L., 1998. Quorum sensing in Escherichia
coli and Salmonella typhimurium. Proceedings of the National
Academy of Sciences of the United States of America 95 (12),
7046e7050.
Teeri, A.E., Josselyn, D., 1953. Effect of excess amino acids on
growth of certain Lactobacilli. Journal of Bacteriology 66 (1),
72e73.
Tsuruoka, T., Tamura, A., Miyata, A., Takei, T., Iwamatsu, K.,
Inouye, S., Matsuhashi, M., 1984. Penicillin-insensitive
incorporation of D-amino acids into cell-wall peptidoglycan
influences the amount of bound lipoprotein in Escherichia-coli.
Journal of Bacteriology 160 (3), 889e894.
Wilen, B.M., Jin, B., Lant, P., 2003. The influence of key chemical
constituents in activated sludge on surface and flocculating
properties. Water Research 37 (9), 2127e2139.
Wood, T.K., Hong, S.H., Ma, Q., 2011. Engineering biofilm
formation and dispersal. Trends in Biotechnology 29 (2),
87e94.
Zita, A., Hermansson, M., 1994. Effects of ionic strength on
bacterial adhesion and stability of flocs in a wastewater
activated sludge system. Applied and Environmental
Microbiology 60 (9), 3041e3048.
 water research 45 (2011) 5805

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Erratum

Erratum to “Bioassays as a tool for evaluating advanced

oxidation processes in water and wastewater treatment”
[Water Research 45 (2011) 4311e4340]

Luigi Rizzo
Department of Civil Engineering, University of Salerno, via Ponte don Melillo 1, 84084 Fisciano (SA), Italy

On page 4319, left column, the sentence starting in line 11 from below should read: “According to the results available in scientific
literature, AOPs were found to decrease and increase toxicity”. The subsequent sentence, “A decreased toxicity was . ., 2010).”,
should be removed.

DOI of original article: 10.1016/j.watres.2011.05.035.

E-mail address: l.rizzo@unisa.it.
0043-1354/$ e see front matter ª 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.08.013