w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

Available at www.sciencedirect.com

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

Activated sludge model (ASM) based modelling of membrane

bioreactor (MBR) processes: A critical review with special
regard to MBR specificities

A. Fenu a,*, G. Guglielmi b,1, J. Jimenez c, M. Spèrandio d,e,f, D. Saroj g, B. Lesjean h,

C. Brepols i, C. Thoeye a, I. Nopens j
a
Aquafin NV, Department of Research and Products Development, Dijkstraat 8, 2630 Aartselaar, Belgium
b
Department of Civil and Environmental Engineering, University of Trento, Via Mesiano, 77 e 38100, Trento, Italy
c
Anjou Recherche, Veolia Eau, Chemin de la Digue, BP 76, 78 603 Maisons Laffitte Cedex, France
d
Université de Toulouse; INSA, UPS, INP; LISBP, 135 Avenue de Rangueil, F-31077 Toulouse, France
e
INRA, UMR792 Ingénierie des Systèmes Biologiques et des Procédés, F-31400 Toulouse, France
f
CNRS, UMR5504, F-31400 Toulouse, France
g
Department of Urban Water and Sanitation, UNESCO-IHE Institute for Water Education, Westvest 7, 2611 AX Delft, Netherlands
h
Kompetenzzentrum Wasser Berlin, Cicerostr. 24, 10709 Berlin, Germany
i
Erftverband, Am Erftverband 6, D 50126 Bergheim
j
BIOMATH, Department of Applied Mathematics, Biometrics and Process Control, Ghent University, Coupure Links 653, 9000 Gent, Belgium

article info abstract

Article history: Membrane bioreactors (MBRs) have been increasingly employed for municipal and indus-
Received 7 October 2009 trial wastewater treatment in the last decade. The efforts for modelling of such wastewater
Received in revised form treatment systems have always targeted either the biological processes (treatment quality
9 April 2010 target) as well as the various aspects of engineering (cost effective design and operation).
Accepted 3 June 2010 The development of Activated Sludge Models (ASM) was an important evolution in the
Available online 11 June 2010 modelling of Conventional Activated Sludge (CAS) processes and their use is now very well
established. However, although they were initially developed to describe CAS processes,
Keywords: they have simply been transferred and applied to MBR processes. Recent studies on MBR
Membrane bioreactors (MBR) biological processes have reported several crucial specificities: medium to very high sludge
Modelling retention times, high mixed liquor concentration, accumulation of soluble microbial
Activated sludge models products (SMP) rejected by the membrane filtration step, and high aeration rates for
Kinetic parameters scouring purposes. These aspects raise the question as to what extent the ASM framework
Influent fractionation is applicable to MBR processes. Several studies highlighting some of the aforementioned
Soluble microbial products (SMP) issues are scattered through the literature. Hence, through a concise and structured
Exo polymeric substances (EPS) overview of the past developments and current state-of-the-art in biological modelling of
MBR, this review explores ASMebased modelling applied to MBR processes. The work aims
to synthesize previous studies and differentiates between unmodified and modified
applications of ASM to MBR. Particular emphasis is placed on influent fractionation, bio-
kinetics, and soluble microbial products (SMPs)/exo-polymeric substances (EPS) modelling,

* Corresponding author. Tel.: þ32 3450 4511.

E-mail addresses: [email protected] (A. Fenu), [email protected] (G. Guglielmi), [email protected]
(J. Jimenez), [email protected] (M. Spèrandio), [email protected] (D. Saroj), boris.lesjean@kompetenz-
wasser.de (B. Lesjean), [email protected] (C. Brepols), [email protected] (C. Thoeye), ingmar.nopens@
ugent.be (I. Nopens).
1
Present address: E.T.C. Engineering srl, Via Praga, 7, 38121 Trento, Italy.
0043-1354/$ e see front matter ª 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2010.06.007
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4273

and suggestions are put forward as to good modelling practice with regard to MBR
modelling both for end-users and academia. A last section highlights shortcomings and
future needs for improved biological modelling of MBR processes.
ª 2010 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4273
1.1. General overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4275
2. Application of unmodified ASMs to MBR processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4276
2.1. ASMs application to MBRs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4276
2.2. Influent fractionation for unmodified ASMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4277
2.2.1. Experimental measurement of influent COD fractions in MBR studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4277
2.2.2. “Assessment of influent COD fractions with trial-and-error methods in MBR studies” . . . . . . . . . . . . . . 4278
2.2.3. Outlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4278
2.3. Process kinetics and stoichiometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4278
2.3.1. Parameter sensitivity in MBR vs CAS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4279
2.3.2. Nitrification kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4279
2.3.3. Denitrification kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4282
2.3.4. Phosphorus removal kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4282
2.3.5. Oxygen uptake rate (OUR) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4283
2.3.6. Oxygen transfer rate (a-factor) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4283
2.3.7. Sludge production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4283
2.3.8. Outlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4284
3. Application of modified ASMs to MBR processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4284
3.1. Integrating membrane rejection studies in ASM models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4284
3.1.1. EPS and SMP definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4284
3.1.2. Modelling objectives of modified ASMs to MBR processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4285
3.2. Influent fractionation for modified ASMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4285
3.3. Overview of stand-alone EPS and SMP models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4286
3.3.1. Stand-alone EPS models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4286
3.3.2. Stand-alone SMP models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4286
3.4. Overview of ASM-extensions incorporating EPS/SMP concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4288
3.5. Model identification e UAP/BAP kinetics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4289
3.5.1. Outlines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4289
4. Outlook and future perspectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4290
4.1. Influent characterization for the unmodified ASMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4290
4.2. Process kinetics for the unmodified ASMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4290
4.3. Process kinetics for the modified ASMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4290
4.4. Application of ASM models at full-scale, hydraulics and aeration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4291
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4291
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4291
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4291

1. Introduction and a reduced foot-print were seen as able to mitigate the

higher capital and operational cost. Recently, technological
The membrane activated sludge process, commonly referred advances (i.e., improved configurations) and scientific inves-
to as membrane bioreactor (MBR), combines an activated tigations have led to further significant cost reduction.
sludge process and membrane micro or ultra-filtration to In early efforts on dynamic modelling of wastewater
separate the treated water from the mixed liquor. The first treatment processes, only two state variables were considered
membrane bioreactors were developed in the late 1960s, in describing degradation of substrate and formation of biomass
“side-stream” configurations, but market penetration became by first order kinetics (McKinney, 1962). However, the later
significant only after the commercialization of submerged MBR models incorporated an increasing number of state variables
configurations (Judd, 2006), which proved to be energetically and process descriptions based on widely accepted Monod-
superior. Since the advent of MBR technology, factors such as type kinetics. The increased understanding of various
the need to comply with stringent environmental legislation complex processes further enhanced the models and
4274 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

Nomenclature OHO Ordinary Heterotrophic Organism

OUR Oxygen Uptake Rate
AOB Ammonia Oxidizing Biomass
PAO Phosphate Accumulating Organisms
AS Activated Sludge
PE People Equivalent
ASP Activated Sludge Process
PN Proteins
ASM Activated Sludge Model
PHA Polyhydroxyalkanoate
BAP Biomass Associated Products
PS Polysaccharides
bA Autotrophic decay rate
qfe Maximum Rate For Fermentation
bH Heterotrophic decay rate
qpp Rate Constant for Storage XPP
BNR Biological Nutrient Removal
rs Specific substrate utilization rate
CAS Conventional Activated Sludge
rS,BAP Production/Consumption Rate of BAP
CODEPS Chemical Oxygen Demand Content of EPS
RSF Relative Sensitivity Function
CODcell CellularChemical Oxygen Demand
SBAP BAP concentration
CODsol,in Soluble Influent Chemical Oxygen Demand
SCOD Soluble COD concentration
CODb Biodegradable Chemical Oxygen Demand
Si Inert Soluble Fraction
CODtot,IN Total influent Chemical Oxygen Demand
sMBR Submerged Membrane Bioreactor
DO Dissolved Oxygen
SMP Soluble Microbial Products
DOC Dissolved Organic Carbon
SNH Ammonia Concentration
EPS Exoecellular Polymeric Substances
SNO Nitrate Concentration
fB Fraction of Biomass that ends up as Microbial
SPO4 Orthophosphate Concentration
Products
SRT Sludge Retention Time
fBAP Fraction of BAP Produced during Cell Lysis
Ss Biodegradable Soluble Fraction
fUAP Fraction of UAP Produced during Substrate
SUAP UAP concentration
Production
TKN Total Kjeldahl Nitrogen
fp Fraction of inert COD generated in biomass lysis
UAP Utilization Associated Products
F/M Food to Micro organisms ratio
VSS Volatile Suspended Solids
HRT Hydraulic Retention Time
WWTP Waste Water Treatment Plant
sCOD SolubleChemical Oxygen Demand
Xaut Autotrophic Biomass Concentration
KBAP Half saturation coefficient for BAP
XDNPAO Fraction of denitrifiers in the PAO biomass
kh,BAP Hydrolysis Rate of BAP
Xi Inert particulate fraction
Keps EPS formation coefficient
Xhet Heterotrophic Biomass Concentration
Kmp Half saturation Coefficient For Microbial Products
XPAO PAO Biomass Concentration
KNH Half saturation Coefficient For Ammonia
Xsto Organics stored by heterotrophs
KNO Half saturation Coefficient For Nitrate
Xs Biodegradable Particulate Fraction
KOA Autotrophic Saturation Coefficient for Oxygen
XTSS Total Sludge Suspended Solid Concentration
KOH Heterotrophic Saturation Coefficient for Oxygen
Xu Non-biodegradable Organic Particulate Matter
kh,UAP Hydrolysis Rate of UAP
YH Yield Coefficient For Heterotrophic Biomass
kUAP UAP formation rate coefficient ofHeterotrophs
YHD Yield Coefficient For Heterotrophic Biomass In
kUAPa UAP formation rate coefficient ofAutotrophs
Anoxic Conditions
KSMP SMP Half saturation coefficient forHeterotrophs
YPAO Yield Coefficient (Biomass/PHA)
iTS_X” Ratio between the mg/l of Xtss and the COD of the
YSMP Yield coefficient for growth on SMP
different fractions
YMP Yield coefficient for growth on MP
LC-OCD Liquid Chromatography - Organic Carbon
Yobs Observed Yield
Detection
gUAP,H Heterotrophic UAP Formation Constant
MBR Membrane Bioreactor
gMP,A Autotrophic MP formation constant
MLSS Mixed Liquor Suspended Solids
gMP,H Heterotrophic MP formation constant
MLVSS Mixed liquor volatile suspended solids,
gUAP,A Autotrophic UAP formation constant
MP Microbial products
mH Heterotrophicmaximum growth rate
MW Molecular Weight
mA Autotrophicmaximum growth rate
NOB Nitrite Oxidizing Biomass
mSMP SMP maximum growth rate
OM Organic Matter
hNO3 Anoxic reduction factor

a steady-state aerobic model was developed (Marais and treatment systems with a variety of configurations for carbon,
Ekama, 1976) which later evolved into a well established nitrogen and even phosphorous removal. Activated Sludge
dynamic model (Dold et al., 1980). The dynamic model Modelling or ASM-modelling represented an important mile-
included key hypotheses viz. bi-substrate and death regener- stone in modelling of biological treatment processes. ASMs
ation. Such early dynamic models showed their very useful were initially developed to describe Conventional Activated
application to design, optimization and control of wastewater Sludge (CAS) processes under correspondingly typical
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4275

operating conditions. Nevertheless, they have been used since MBRs treating domestic wastewater. However, applications of
the late nineties to simulate biomass kinetics in MBRs systems models at different scales (lab, pilot and full-scale) are
as well, provided that some necessary adaptations are made considered.
to allow for the specific behaviour of these systems.
The specificities of the MBR biological process compared to 1.1. General overview
the CAS process have been extensively reported (Ng and Kim,
2007): higher sludge retention times (SRTs), higher mixed Since MBRs encompass the Activated Sludge Process (ASP) as
liquor concentration and viscosity, accumulation of microbial their fundamental process, ASMs have been applied for the
products rejected by the membrane filtration step, high biological modelling of MBRs as reported in various studies
aeration rates for scouring and good nitrification perfor- (inter alia Chaize and Huyard, 1991; Wagner and Rosenwinkel,
mance. These all influence and characterize the MBR process 2000; Wintgens et al., 2003; Delrue et al., 2008). ASMs are
behaviour. The logical question one can pose now is whether robust dynamic models which are widely used for activated
all current knowledge about ASM based modelling for CAS is sludge based wastewater treatment process understanding,
simply transferable to MBR systems, or in other words, how design, optimization and control. There are various versions
the current MBR process understanding can be merged into of ASMs, and one or another is preferred on account of various
the ASM framework. MBR literature has been very prolific factors e.g. modelling objective, desired accuracy, calibration
lately but studies and results on biological modelling are effort, ease of use and relevance of process types, etc. Various
scattered, and a systematic overview of the state-of-the-art of versions of such dynamic models viz. ASM1, ASM2, ASM2d,
all scientific work performed to date in this area is missing. ASM3 (Henze et al., 2000) have been developed during the two
More specifically, both academics and end-users of biological decades since the introduction of the first version i.e. ASM1
modelling processes are in need of a concise summary to (Henze et al., 1987), by assimilating the developments in
support their decision making process in terms of MBR process understanding pertaining to wastewater treatment
design and operations. This review paper aims to (i) provide systems.
a concise overview of the past developments and current The advent of ASM1 introduced the Gujer matrix form of
state-of-the-art in biological modelling of MBR systems and (ii) model presentation, which assimilates all the process
highlight weaknesses in the current models and propose descriptions in a condensed and elegant way. ASM1 does not
future needs. incorporate biological phosphorous modelling although the
For reasons of clarity, the review is broken down into two process had already been established prior to the advent of
major sections: unmodified versus modified ASM models for ASM1. However, most of the plant at that time did not
MBR. The former contains the literature studies where the incorporate this process and only required predictions of C
plain ASM models, as described in Henze et al. (2000), are used and N removal, aeration capacity requirements and sludge
as such in MBR applications and where only parameter esti- production. Later, biological P-removal gained importance
mations have been performed to improve the model perfor- and was therefore included in ASM2. The understanding of
mance (i.e. matching model predictions with measured denitrification in the biological phosphorous removal
experimental data). Unmodified ASM models also contain processes evolved further, resulting in ASM2d, which incor-
efforts that expand the description of certain biokinetic porated the processes pertaining to denitrifying PAOs.
processes, but without adding state variables to the model ASM2d might not have been considered very important if
state vector. The decision to include those applications here only carbon and nitrogen removal had been targeted;
was taken from the perspective that these models are more however, the model played an important role in the under-
readily applicable in practice. standing of the complexity of combined nitrogen and phos-
With modified ASM models, we target all applications phorous removal processes (Henze et al., 2000). Its practical
where the plain ASM models are extended in terms of (1) application was hampered by the large number of parame-
biokinetic process models and (2) SMP/EPS models. Here, ters in the model. Hence, in the mean time, ASM1 continued
additional state variables are introduced in the model either to be the state-of-the-art model for dynamic modelling in
for a better description of certain processes already present in wastewater engineering, despite certain defects that became
ASMs or to allow for the description of processes formerly apparent in its application, e.g. no nitrogen and alkalinity
lacking in ASMs (e.g. EPS/SMP). limitations for heterotrophic bacteria were included, no
End-users will most probably be interested in the former differentiated decay rates of nitrifiers under aerobic and
section as the models described in that section are intended anoxic conditions were considered and intracellular storage
for use in practice, whereas the latter section focuses on of PHAs was not addressed. The introduction of ASM3 (Gujer
academic work to improve process understanding and, hence, et al., 1999) aimed to correct the defects of ASM1 and pre-
is still in a more basic research phase and further away from sented a new standard for ASM based modelling. The original
practical use. ASM3 did not incorporate biological P-removal (unlike ASM2
Both sections deal with the typical aspects that are meth- or ASM2d), chemical precipitation, growth of filamentous
odologically encountered during a modelling process such as organisms or pH calculations. However, these processes can
(i) influent characterization and (ii) determination/calibration be connected as add on modules (Henze et al., 2000). One
of kinetics and stoichiometry. A final section highlights example of this is the extended ASM3 for biological P-
shortcomings and the authors’ conception of future needs for removal (Rieger et al., 2001).
improved biological modelling of MBR processes. Finally, to The process descriptions of anaerobic processes are not
limit the scope of the paper, discussions have been kept to part of ASM1/ASM3. Their application is limited to (aerobic)
4276 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

ASP with possible extension to include the anoxic conditions a laboratory scale MBR fed with domestic wastewater, aimed
and (partial) anaerobic reactor integration (for Bio-P processes to model effluent COD, TKN and sludge production at two
only). For a complete description of anaerobic processes, other HRT values (viz. 8h and 2h) and very high SRT (nearly 100 d).
models such as the anaerobic digestion model (ADM) are used, The MBR system was modelled with ASM1 using default
which are not the part of current review. values of parameters (Henze et al., 1987). The predicted
The assumptions underlying the ASMs related with effluent COD was reported to be slightly lower than that
influent fractions might no longer be applicable when indus- observed, and the predicted TKN was found quite close to the
trial wastewaters are mixed; various facets of influent frac- observed value. However, the major disagreement was
tionation are outlined in the Section 2.2. reported on solids concentration. The model predicted
a lower solids concentration than observed, and the solids
concentration prediction was relatively better at higher HRT.
2. Application of unmodified ASMs to MBR The probable reason was thought to be the very high SRT (i.e.
processes 100 days). These outcomes illustrate that a non-calibrated
ASM1 is able to give a reasonable estimate of effluent COD
The objective of this section is to give a concise overview of and TKN, but is insufficient for very low HRT and very high
the application of unmodified ASMs. The terminology SRT systems. Hence, this imposes care in the application of
“unmodified” should be interpreted in terms of modelling those models and in the investigation of appropriate param-
scope as well as model structure. Unmodified ASM models are eter sets valid for these systems under variable operational
to be used for modelling MBRs when objectives are similar to conditions. This sets the scene for investigating the where-
those originally stipulated for the use of ASM for CAS systems abouts of the encountered limitations.
(Henze et al., 2000): process design, effluent characterization, The application of ASM1 moved towards better under-
oxygen demand and sludge production. In these cases, plain standing of model parameters and, hence, a more systematic
ASM models are used. The section includes efforts where calibration, taking into account the nature of the MBR
plain ASMs were calibrated for MBR or where slight modifi- biology and specific operating conditions. The ASM1 appli-
cations to biokinetic processes were performed using solely cation on side-stream MBR by Jiang et al. (2005) stressed the
the state variables defined in the original ASMs. This section importance of various sensitive biokinetic parameters and
includes the models that are closest to being applicable in influent wastewater characterization. More recently, Delrue
practice and is, thus, very useful for end-users. It includes an et al. (2008) commented that despite some difficulties,
introduction and treats two important aspects: (i) influent ASM1 is suitable for modelling MBR plants if influent char-
characterization and (ii) determination/calibration of kinetics acterization and systematic calibration of aeration can be
and stoichiometry. taken care of.
The incorporation of storage phenomena (Krishna and van
2.1. ASMs application to MBRs Loosdrecht, 1999; Gujer et al., 1999) is a unique feature of ASM3
and might play a role in the case of MBRs on account of
The application of ASMs are presumably meant for ASP possibilities of low organic load conditions (Wintgens et al.,
operation in the ranges of conventional ASP operating 2003). Nevertheless, ASM1 has been shown to be sufficient
parameters, e.g. SRT range 3e15 d, HRT range of 3e5 h and where conditions are not favourable to storage phenomena
MLSS range 1.5e4 g/L for completely mixed systems (Metcalf (Delrue et al., 2008). In the aim of modelling MBRs over a large
and Eddy, 2003). A recent study on design and operating range of SRTs, Sperandio and Espinosa (2008) used ASM1 and
experience with municipal MBRs in Europe has reported the ASM3 and commented that ASM models could provide satis-
ranges of various parameters (Itokawa et al., 2008). The HRT of factory prediction of aerobic biological processes in
13 MBR plants have been reported to be in the range of submerged MBRs, although these could be improved for high
2.8e8.1 h, with most of the plant operating at HRT between 4 SRT conditions.
and 6 h. The MLSS of 11 MBR plants have been reported to be in Studies so far are not conclusive as to whether ASM1 or
the range of 7e13.5 g/L, with most of the plant operating at ASM3 is better for MBRs. It appears that the application of
MLSS higher than 10 g/L. Further, the SRT values of 7 plants ASMs, in their original forms, often needs careful calibration
have been reported to range between 15 and 40 days. For of parameters, especially for sludge production and nitrifi-
municipal MBR applications, it seems reasonable to define the cation modelling. The issue of the significance of high SRT,
SRT below 15 days as “low SRT range”, and SRT above 40 days which was a matter of further attention even in early MBR
as “high SRT range”. modelling studies (Chaize and Huyard, 1991), remains a rele-
Efforts have been made over the last 15 years towards vant point. It has been reasoned in recent research (inter alia
appropriate application of ASMs for MBRs. While early trials Massé et al., 2006; Sperandio and Espinosa, 2008) that high
(Chaize and Huyard, 1991) used the very basic form of ASM1, SRT operation of MBRs is linked with corresponding influence
using default parameter values, performing no systematic on MBR specific sludge production and autotrophic biology.
calibration or influent characterization, recent efforts have Throughout, it can generally be observed that all the recent
presented various aspects of systematic calibration of key and efforts aiming at an accurate biological modelling of MBRs
sensitive parameters along with emphasis on the influence of focus on MBR specificities (e.g. high SRT operation,
influent wastewater characterization in terms of various ASM membrane retained microbial metabolites etc.) and the cor-
based fractions (Delrue et al., 2008; Sperandio and Espinosa, responding parameter adjustment and modifications
2008). The early study of Chaize and Huyard (1991), based on required in ASMs.
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4277

2.2. Influent fractionation for unmodified ASMs

Table 1 e Comparative evaluation of the two methods for
wastewater characterization. Total COD: 579 g mL3
The monitoring practice for municipal WWTPs typically relies (adapted from Jiang et al., 2005).
on collecting influent and effluent wastewater samples to
Parameter Chemicalebiological PhysicaleChemical
assess the inlet organic and nutrient loading and the impact (g m3) (g m3)
on the receiving water body. However, when process model-
SI 33 (5.7%) 33
ling is carried out, a more detailed characterization of feed-
SS 214 (37.0%) 228
water COD is needed, which supplies the modeller with
XI 58 (10.0%) 141
information about the degree of biodegradability (readily, XS 274 (47.3%) 177
slowly and inert) and the physical aggregation state (soluble,
particulate) of the influent substrate. Generally, two different
approaches are reported in the MBR-modelling literature for
COD characterization: (i) integration of chemical/physical/ biotank for the “physical-chemical” method under steady-
biological methods and (ii) application of “trial-and-error” state conditions (13245 g m3 for the physical-chemical
procedures which aim to fit experimental observations (i.e. method versus 10020 g m3 experimentally measured).
MLSS in the biotank) with model simulation by tuning the A bench-scale MBR (0.016 m3) was modelled by Sperandio
different COD fractions in a reasonable range of values. In the and Espinosa (2008) with ASM1 and ASM3 over a wide range
following subsections inferences from these two different of sludge ages (10O110 days), with special focus on waste-
approaches are presented and discussed. water characterization and excess sludge. During the four
experimental periods, a combined physical-biological method
was used for influent wastewater fractionation: the non-
2.2.1. Experimental measurement of influent COD fractions in biodegradable soluble COD (SI) was assumed to be equal to the
MBR studies effluent COD, the inert particulate COD (XI) was measured at
Jiang et al. (2005) compared two methods for wastewater the end of long-term BOD measurement (30 days), the slowly
characterization to calibrate ASM1 in a side-stream biodegradable (XS) and readily biodegradable COD (SS) were
membrane bioreactor (sludge age w20 days, HRT e8 h). The obtained by combining respirometry (Spèrandio and Paul,
first method (also referred to as “chemical-biological”) 2000) and 0.45 mm filtration. The simulated MLVSS trend
combined the respirometric technique proposed by indicated that ASM1 was able to predict sludge production at
Vanrolleghem et al. (1999) and physical separation (filtration SRTs shorter than 50 days, while an overestimation was
on 0.45 mm fibreglass paper). First, the 0.45 mm filtration was observed at SRT ¼ 110 days. In contrast, an underestimation is
carried out to quantify soluble versus particulate COD. Once reported by the authors in the application of ASM3 at SRT
assessed, the soluble and particulate biodegradable COD (SS values of 10 days and 30 days, and a slightly better prediction
and XS respectively) were determined with a respirometric of the MLVSS concentration at 110 days. According to the
test (S0/X0 ¼ 1/200). By assuming the soluble inert fraction of authors, most of the discrepancy between the ASM1 model
influent COD (SI) to correspond to 90% of effluent COD and the experimental observations is due to the large quantity
(according to Vanrolleghem et al., 2003), the authors deter- of non-biodegradable organic particulate matter (XU) related
mined the inert particulate fraction (XI) as the difference to the death-regeneration concept, which is replaced by the
between CODtot,IN and (SS þ SI þ XS). In order to evaluate the endogenous decay concept in ASM3. In terms of surplus
effectiveness of this method, a trial-and-error procedure was sludge production (expressed as Yobs) the work points out
also used to determine XI and XS, by comparing a steady-state that, the longer the SRT, the more relevant the impact of the
COD mass balance with measured and simulated MLSS inert fraction in the influent feedwater (XI) on the sludge
concentration in the biotank. The two methods provided very composition. Therefore, the authors recommend introducing
close results (e2.7% on XS value). The second approach applied a slow hydrolysis mechanism in the standard ASM1 and ASM3
by the authors (“physical-chemical method”) used the to correctly predict sludge production at very long SRT. This
STOWA protocol for wastewater characterization (STOWA, also matches the conclusion of other groups (Rosenberger et
1996; STOWA, 1999, Roeleveld and van Loosdrecht, 2002) al., 2000; Rosenberger, 2003), who demonstrated that, for
and coupled physical separation (0.45 mm) with BOD analysis. SRTs above 80d, an hydrolysis factor could be used to better
The soluble inert COD (SII) in the influent was calculated as simulate the MBR sludge production. Witzig et al. (2002) later
90% of soluble effluent COD and soluble biodegradable COD demonstrated that there was actually no hydrolysis at high
(SS) was given by the difference (CODsol,in e SI). Then, biode- sludge age, but that the bacteria went into maintenance mode
gradable COD (CODb ¼ SS þ XS) was calculated as a function of and did not grow any more, mathematically impacting the
BOD5 and XS was calculated from CODb and SS; finally, XI was sludge yield in a similar way to hydrolysis (Drews et al., 2005).
determined as the difference between CODtot,IN, SI, SS and XS. This fact, known for pure cultures, was shown to be also
The mean results of the wastewater fractionation on 16 applicable for mixed cultures in activated sludge.
influent samples are summarized in Table 1. The most In a recent work, a small pilot-scale MBR (sludge age e36
significant difference between the two approaches occurred days) was modelled by Sarioglu et al. (2009) with a modified
in the XI fraction, which was significantly higher when ASM1 to describe the simultaneous nitrification-denitrifica-
determined with the “physical-chemical” method compared tion process. All the growth and decay processes were coupled
to the “chemical-biological” one. This discrepancy reflected with switching functions defining the diffusion limits of
a relevant overestimation of the MLSS concentration in the oxygen and substrate. A combined physical-chemical-
4278 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

biological method was used for COD characterization; the

Table 2 e Results of the COD fractionation of a large scale
biodegradable COD (CODb) was determined according to MBR working on combined sewer influent (Erftverband,
Roeleveld and van Loosdrecht (2002), while the SI and XI 2001).
fractions were measured according to Germirli et al. (1991) and
Dry weather Rain weather
Orhon et al. (1999) respectively. Once SS was calculated as the
difference between influent soluble COD and SI, the XS frac- XI, % of COD 20 40
SI, % of COD 8 15
tion was determined by mass balance. This calibration resul-
ted in a fairly good fit of MLVSS and MLSS in the biotank.
A respirometry-based approach for wastewater fraction-
ation has been reported by Guglielmi et al. (2009), who cali- calibration of influent COD fractions was reported by the
brate an extended version of ASM3 that includes the authors, due to the shorter HRT and reduced buffering
simultaneous growth of heterotrophs on both storage prod- volume.
ucts (XSTO) and readily biodegradable substrate (SS), originally
proposed by Sin et al. (2005). The model has been successfully 2.2.3. Outlines
validated in terms of sludge production under dynamic Summing up the inferences for ASM applications to
conditions on a large pilot-scale MBR run under an SRT of membrane bioreactors, the available literature highlights the
20O25 days. In detail, the total biodegradable COD was crucial role that design/operational SRT plays in influent COD
determined according to Ekama and Marais (1986), whereas characterization. Particularly, the actual operational sludge
the soluble biodegradable COD was estimated according to the age can influence the choice of a proper method for the
single-OUR method proposed by Ziglio et al. (2001), with determination of XS and XI fractions since at very long SRT
sodium acetate as standard. Afterwards, SI was calculated as a slow but significant hydrolysis of the “inert” fraction takes
the difference between filtered COD in the influent (0.45 mm) place. From the modelling point of view, this results in
and the soluble biodegradable COD, being usually equal to a general overestimation of the sludge production. This issue
90e95% of COD concentration in the permeate. Finally, XI was is strictly related to the biodegradability concept itself, which
quantified with a mass balance. depends not only on the substrate characteristics and the
substrate/biomass affinity but also on the time available for
2.2.2. “Assessment of influent COD fractions with trial-and- biological degradation. In terms of practical process design
error methods in MBR studies” and operation, this aspect has great relevance for MBR
When ASM-modelling tools are used for process optimization installations with noticeable seasonal fluctuations of the
in a full-scale installation, there is typically a general tendency influent organic loading (e.g. tourist areas), where SRT can
towards “trial and-error” procedures. increase up to 50e60 days or more. In this case, a trial-and-
In the application of ASM1 to a full-scale MBR in Guéthary, error method tuning XS and XI seems more appropriate. On
France (10,000 PE, 1600 m3 d1; SRT ¼ 30O60 days; F/ the other hand, when operating at more conventional sludge
M ¼ 0.02O0.05 kgBOD5 kgMLVSS1 d1), Delrue et al. (2008) ages (<30O40 days), respirometry-based fractionation seems
compared three different methods for soluble/particulate to be able to provide satisfactory characterization.
fractionation, namely filtration on fibreglass filter, 0.1 mm
membrane filtration and coagulation-flocculation followed by
a 0.1 mm filtration. Then, in order to divide COD into biode- 2.3. Process kinetics and stoichiometry
gradable/non-biodegradable organic matter, the method
proposed by Roeleveld and van Loosdrecht (2002) was Next to influent characterization (previous section) and
compared with a trial-and-error procedure based on MLVSS module structure (unaltered in this case), kinetic and stoi-
steady-state mass balance (fitting method). In their conclu- chiometric parameters provide degrees of freedom for
sion, the combination of fibreglass filtration and MLVSS fitting matching ASM model predictions with experimentally
was chosen as the most reliable protocol for influent COD collected data.
characterization. The standard ASM models come with default parameter
A model-based optimization of a full-scale MBR was values, obtained from wide experience at both lab-scale and
carried out at Rödingen, Germany (Erftverband, 2001), in full-scale. However, the inclusion of membranes in activated
which the influence of the weather conditions on the COD sludge systems may cause associated changes in these default
fractionation was described. The plant (3000 PE, SRTe25 days; kinetic parameters compared to CAS systems. The main
combined sewer) was modelled in ASM1 and calibrated by effects on process kinetics are possibly due to (i) specific
means of a trial-and-error procedure on the observed sludge biomass selection (high SRTs, free bacteria retention), (ii) high
production over two different periods (57 and 28 days). A biomass concentration, and (iii) high hydrodynamic
comparative evaluation of inert fractions (SI and XI) is shown constraints imposed by continuous or cyclic air scouring of
in Table 2, with an increase of both non-biodegradable the membrane element instead of quiescent conditions in the
components during rainy events. The catchment area of the settler. The crucial parameters for design and control of MBRs
MBR was small but widely ramified and the sewer had only are: the sludge suspended solids (XTSS) impacting the excess
a mild slope so that effects of run-off, the flushing out of sludge production and the oxygen transfer rate (a-factor), the
deposits and dilution by the rain water were very pronounced removed and residual nitrogen species (SNH, SNO), the residual
and produced a visible effect on the plant. Unlike what is phosphorus concentration (SPO4) and the oxygen consumption
observed in CAS modelling, the need for a more frequent rate (OUR, and SO).
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4279

In the next subsection, the parameter sensitivity will be moderately to very influential, depending on process opera-
primarily discussed. The following subsections will look more tional conditions. In fact, Jiang et al. (2005) studied the local
deeply at the impact of MBR on the kinetics of the different parameter sensitivity in a model calibrated for a specific case,
processes. with specific operating conditions (SRT ¼ 20 days, DO
range ¼ 3e8 mg/l). At different operating conditions, the
2.3.1. Parameter sensitivity in MBR vs CAS sensitivity of the parameters might change. That is why
Given the similar nature of the process, it would be reasonable sensitivity analysis is systematic and necessary in every
to think that the influence of most kinetic constants is calibration exercise. An alternative to this approach is the use
comparable in CAS and MBR processes. The reasons pointed of global sensitivity analysis which explores the entire
out above might, however, lead to different system behav- parameter space and yields average sensitivities (and not just
iours. During the calibration exercise, sensitivity analysis can locally around a single point in this space), as performed by
help by assessing the most influential parameters for a given Benedetti et al. (2008), with an ASM2 model for BSM2
process and biological reaction (sludge production, uptake of (Benchmark Simulation Model no.2). Global methods are,
COD, denitrification and nitrification). The results of such however, always more computationally expensive and inter-
a sensitivity analysis for ASM1 applied to an MBR are given in pretation of their results also needs scrupulous care.
Table 3. Relative sensitivity functions (RSF) were calculated by
Jiang et al. (2005) for several model variables (Y) towards all 2.3.2. Nitrification kinetics
model parameters (q), as described by the equation (1). As autotrophic bacteria are very sensitive to the environ-
mental conditions, the impact of including membranes for the
q dY
RSF ¼ (1) solid-liquid separation on the kinetics of nitrogen removal has
Y dq
received considerable attention. Ammonium consumption
RSFs rescale the sensitivity function in order to allow easy and residual concentration is sensitive to nitrification
comparison and fair ranking accounting for different orders of parameters, in decreasing order of influence: mA, bA, KNH, KOA
magnitude in parameters and variables. Ranking is done and YA (Jiang et al., 2005). In ASMs, the nitrification rate is
according to the absolute values of RSF as shown in Table 3. controlled by the concentration of active autotrophic bacteria
A finite difference approach is typically used to compute the (Xaut) stabilized in the process, which is imposed by the
derivative. conversion yield YA, the influent nitrifiable nitrogen, and
The influent wastewater characterization, with particular the decay rate bA. The nitrification rate is basically linked to
regard to XI, XS, SS and SNH fractions, appears to be extremely the product mAXaut. For this reason, the parameters bA and mA
influential on MLSS concentration and effluent quality, which are highly correlated and it is impossible to identify them
confirms the importance of the previous discussion (Section simultaneously if the active biomass is not stabilized at
2.2). The stoechiometric parameter (YH) and kinetic parame- different levels, i.e. data should be collected at different SRTs
ters (bH, bA, mmaxH and mmaxA) are very to moderately influen- in a continuous process (Sperandio and Espinosa, 2008).
tial on the MLSS concentration and effluent quality, whereas Munz et al. (2008) highlighted the difference between the
the remaining parameters do not seem to be candidates for ammonium-nitrogen concentration in CAS and MBR effluents
change when performing a calibration. (both with 50 days SRT): for MBRs the nitrification was more
Petersen et al. (2002) used the same sensitivity function stable and complete. In the same way Parco et al. (2006) found
applied to CAS (with an SRT of 8.6 days). The most influential a specific autotrophic uptake rate for MBR 1.8 times higher than
parameters were the same as for MBR but with different for conventional Biological Nutrient removal (BNR). By oper-
degrees of importance. Kinetics parameters (KOA, KNH and kh) ating CAS and MBR in the same conditions (SRT), Manser et al.
which had no influence in Jiang et al.’s (2005) model became (2005) did not observe any significant difference in the

Table 3 e Results of sensitivity analysis with the calibrated ASM1 model.

RSF Steady-state calibration

XTSS SS SNO SNH

RSF<0.25 fp, bA, ka, kh, KNH, fp, bA, ka, kh, KNH, KNO, fp, bA, ka, kh, KNH, KNO, fp, ka, kh, KNO, KOA, KOH,
(Not influential) KNO, KOA, KOH, KS, KX, KOA, KOH, KX, mmaxA, KOA, KOH, KS, KX, mmaxA, KS, KX, mmaxH, YA, YH,, SNO
mmaxA, mmaxH, YA, SNO YA, YH SS, SNH,SNO mmaxH, YA, SNO
0.25 < RSF<1 bH bH KS bH Y H bA, KNH, SNH
(Moderately influential) - (YH, XI)*- - XS*- - (YH, bA, KNH, KOA, SS)*- - (bA, KNH, SNH, KOH,
KOA, kh, SS, XS)*-
1 < RSF<2 YH, SNH mmaxH (SNH, XS)* mmaxA, SS
(Very influential) - (YH,, mmaxH, Ks, kh)* - YH, *
RSF>2 XI, XS, SS XS XS, SS, SNH XS
(Extremely influential) - mmaxA*-

RSF results with the calibrated ASM1 model (Jiang et al., 2005): MBR with SRT ¼ 20 (In standard character).
RSF results with calibrated ASM1 model (Petersen et al., 2002) CAS: with SRT 8.6 days (In bold and with *).
4280 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

maximum specific ammonium uptake rate but authors The KNH value seemed to be linked to internal transfer
showed that the ammonium uptake rate became higher in MBR resistance but also to external mass transfer resistance. KNH
compared to the CAS during transient shock loads, especially was reported to change from 0.3 to 0.6 when MLSS increases
at low temperature and relatively low DO (dissolved oxygen). from 3 to 8 g/l, pointing to an increase of viscosity (Spèrandio
The difference between nitrifiers in MBRs and CAS may et al., 2005; Sarioglu et al., 2009) and an increase of external
depend on cumulative factors such as: (i) the different micro- transfer resistance. When MLSS increased, ammonia uptake
organism selection (not demonstrated); (ii) the higher bio- rate (AUR) and mA decrease indicating a concentration effect
availability of substrates which can be due to the smaller size (Pollice et al., 2008; Ramphao et al., 2006). Parco et al. (2006)
of flocs (Manser et al., 2005); and (iii) the tendency of nitrifiers explained that the MLSS effect is related to ammonia diffu-
clusters to grow at different places in the flocs or with lower sion limitations and not to DO transfer limitation.
density (presence of Ammonia Oxidizing Bacteria aggregates Regarding KOA, ASM1 models fitted to respirometric data
on the surfaces of MBR flocs was observed only by Munz et al. resulted in a KOA range of 0.18e0.4 mgO2/L in MBR (Spérandio
2008). These findings might however also depend on et al., 2005; Jiang, 2007; Jimenez et al., 2008) and tended to be
a different mass balance, i.e. higher autotrophic biomass lower than the ASM1 default value of 0.4 mgO2/l. A parameter
concentration can be maintained by a better retention of set from the different studies on MBRs in urban wastewater
solids (for example, it is clear that solids loss in the activated treatment is reported in Table 4. Manser et al. (2005) showed
sludge can contribute to a more or less significant decrease of that values differed significantly from MBR to CAS. In detail,
nitrifiers depending on the SRT). In this sense, to better he observed lower KOA values for MBR (0.18 mg O2/L) in
understand this section, the reader must consider that lower comparison with the CAS (0.79 mg O2/l) and explained this
affinity constants or higher growth rates do not necessary variation with the lower transfer resistance induced by MBR
mean higher removal rates, since the nitrification mass smaller floc size (which became negligible for floc sizes under
balance must be taken into account entirely. 100 mm). Similar conclusions have been drawn in other studies
Several authors have found discrepancies when modelling (Jiang, 2007; Jimenez et al., 2008) as shown in Tables 4 and 5.
MBR processes with ASM1 using the initial default values of Jimenez et al., (2008) observed that KOA had a very high
Henze et al. (1987) for nitrification (Jiang, 2007; Sperandio and sensitivity to nitrification (based on RSF results) when DO
Espinosa, 2008). Sperandio and Espinosa (2008) used ASM1 and concentration was lower than 2 mg/l and average floc size was
ASM3 to calibrate an MBR working at a large range of SRT, lower than 35 mm.
through the parameters mA and bA. The authors reported that However, it is not always obvious that floc sizes are lower
ASM1 default values (0.8; 0.04d1) overestimated ammonia in MBRs than in CAS, since they depend on the hydrodynamic
removal for all the SRTs studied, whereas ASM3 (1, 0.15d1) conditions and on the type of aeration (intermittent or
gave better results but minimized the SRT influence. The data continuous). An inverse tendency was obtained by Spèrandio
obtained from this study led to a mA ¼ 0.45 d1 and et al. (2005), who reported results indicating that an sMBR can
bA ¼ 0.04 d1. Autotrophic growth rate is known to be variable have larger flocs and larger values for KNH, if aeration and
from one process to another even for activated sludge shear stress are limited. Similar results were recently found by
processes. This means that more data will be necessary to Sarioglu et al. (2008, 2009). On the other hand, floc size is also
conclude on the best set of parameters for MBR, i.e. mA should influenced by SRT (Massé, 2004). This author showed that an
be measured properly in future works on MBR. Since this SRT increase from 9 to 106 days led to a diminution of the
parameter is bacteria-specific, an analysis of microbial species average floc size in the MBR from 240 to 70 mm. Deflocculation
(AOB, NOB) by molecular techniques in MBRs would probably of biomass could be due to the increase in aeration when
help in interpreting the observed changes in mA. MLSS increases, or to the reduction of F/M ratio, leading to the
Concerning the better behaviour of MBRs reported by Munz decrease in the active fraction of the sludge (Spèrandio et al.,
et al. (2008), Parco et al. (2006), and Manser et al. (2005), an 2005).
explanation may be offered by the reduction of the half- Therefore, although there is a tendency in MBRs to find
saturation constants KNH and KOA, which directly influence a value for the half-saturation constants, KOA and KNH, lower
the residual ammonia concentration model predictions. than the ASM1 default values, these parameters depend on
Regarding the KNH, experiments on MBRs in the literature the operating conditions (SRT, MLSS concentration, viscosity,
report values from 0.15 to 1 mgN/l and tend, in some cases, to oxygen concentration, floc size distribution).
be lower than the ASM1 default value (ASM default values is 1 Concerning the autotrophic decay rate (ba), Manser et al.
mgN/l), leading to an improvement of ammonia transfer (2006) did not find significant differences between CAS and
(Delrue et al., 2008; Spérandio et al., 2005; Jiang et al., 2009; MBR concerning the ammonia oxidizing biomass (AOB), but
Jimenez et al., 2008; Erftverband, 2001, 2004). A parameter a slight difference concerning the nitrite oxidizing biomass
set from the different studies on MBRs in urban wastewater (NOB): AOB ba was found to be 0.13d1, both in CAS and MBR,
treatment is reported in Table 4. On the other hand, Manser et while the NOB ba was 0.28d1 and 0.17d1 for CAS and MBR
al. (2005) reported observed KNH values lower than ASM1 respectively.
default values but similar for CAS (0.14 mg N/l) and MBR The conversion yield YA does not seem to be influential
(0.13 mg N/l), concluding that the limitations induced by larger (see Table 3) but can become relevant when dissolved oxygen
floc density were not that significant for ammonia transfer. It sensitivity is studied under dynamic conditions (Jiang et al.,
should also be noted that KNH also varied significantly from 2005). Moreover, this parameter was measured by Jiang
one activated sludge to another (Sperandio and Espinosa, et al., (2005), yielding the value (0.25 gN/gCOD), close to the
2008). ASM default value (0.24gN/gCOD).
Table 4 e Parameter set from the different studies on MBR in urban wastewater treatment.
Model Units Experimental Default Jiang et al., Spérandio Manser Jiang Sarioglu Delrue, Jimenez Erftverband RWTH Range
methods ASM1 2005 et al., 2005 et al. (2005) 2007 et al., 2008 2008 et al., 2008 2001, 2004 2008 of values
references
ASM1 ASM1 ASM3 ASM1 ASM2d ASM1 ASM1 ASM1 ASM1 ASM1
endogenous modified
decay model

SRT d 20 10e110 20 38 30e60 15

Nitrification mmaxA d-1 [2], [5]; [7], 0.8 0.45 1 0.8 0.45e1.00

w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4
[8], [6],[10]
ba d-1 [1], [5], [7], 0.05e0.15 0.08 0.04 0.055 0.06 0.15 0.04e0.15
[8], [10]
KNH mgN [8], [10] 1 0.25e0.6 0.2 2 1 1 0.1 0.10e2.00
-NH4/l
KOA mgO2/l [4], [8], 0.4 0.18 0.2 1.25 1 0.25 0.18e1.25
[9], [10]
YA gCOD/gN [3], [8] 0.24 0.25 0.24

Denitrification %XI % COD 15 17.5 15 e

COD oxidation YH gCOD [3], [8] 0.67 0.72 0.66 0.67 0.52e0.92 0.63e0.67
Sludge prod. /gCOD
bH d-1 [2], [4], [8] 0.62 0.25 0.4 0.24 0.03e0.47 0.24e0.4
KO,NOB mgN/l [9] 0.5 0.13 2 1 0.13e2
KOH mgO2/l [4], [8] 0.2 0.05 1 0.22 0.1 0.05e1

Comment Jiang et al., 2005: YH with acetate addition overestimated, inducing an underestimation of fp
Spérandio et al., 2005: Assessment of (ma, ba) valuable in a large range of SRT
Manser et al. (2005): Accurate estimation of KOA: Correlation with floc size distribution
Jiang, 2007: Biological P-removal calibration (cf. paragraph P-removal)
Sarioglu et al., 2008: Specifically adapted for modelling simultaneous nitrification-denitrification
Jimenez et al., 2008: Correlation with floc size distribution
Erftverband, 2001, 2004: Assessment for simulation of timeline over several weeks

[1] Henze et al. (1987), [2] Ekama et Marais (1986), [3] Solfrank et Gujer (1991), [4] Kappeler et Gujer (1992), [5] Lesouef et al. (1992), [6] Nowak (1994), [7] Spanjers et al. (1995), [8] Vanrolleghem et al. (1999),
[9] Ficara et al. (2000) [10] Van Haandel (2007).

4281
4282 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

Table 5 e Biokinetic parameters and size distribution found in MBR modelling literature.
Mean floc size (mm) K_OA gO2/m3 K_OH gO2/m3 K_O,NOB gO2/m3

ASM1 0.40 0.20 0.50

Manser et al. (2005) 35 0.18 0.05 0.13
CAS (Manser et al., 2005) 307 0.79 0.16 0.47
Jimenez et al. (2008) 35 0.25 0.10 e
Jiang et al. (2009) 30e50 0.20 0.20 0.17
Sarioglu et al. (2008) SNdN N.d. 1.25 1.00 e
Sarioglu et al. (2009) SNdN N.d High viscosity 2.00 1.75
(SVI ¼ 800e1000 ml/g)

N.d.: not determined.

SNdN: simultaneous nitrification and denitrification.

In order to obtain a clear independent estimation of nitri- an influence on oxygen transfer and consequently on oxygen
fication parameters, tailored batch tests are suggested. transfer in the denitrification zone. In fact, some MBR
Substrate excess or substrate absence conditions would be configurations present sludge recirculation from the aerated
necessary to respectively determine mA and bA, (Henze, 1992; membrane tank to the denitrification zone and, hence, the
Vanrolleghem et al., 1999; Lesouef et al., 1992) while batch level of dissolved oxygen has an important effect on denitri-
tests with substrate addition at various oxygen concentra- fication potential (Sarioglu et al., 2008). Dissolved oxygen level
tions would be necessary to estimate the oxygen half-satu- can inhibit denitrification in anoxic growth reaction model-
ration constant (Kappeler and Gujer, 1992; Ficara et al., 2000). ling, through the parameter KOH. Manser et al. (2005) found in
YA and KNH could be estimated with ammonia addition in MBRs, a KOH value of 0.05 mgO2/l, lower than the 0.16 mgO2/l
batch tests, as in methods described by Vanrolleghem et al. for a CAS process working at the same conditions. However,
(1999), Van Haandel and Van Der Lubbev (2007). this parameter also depends on system hydrodynamics and
configuration.
2.3.3. Denitrification kinetics
Nitrate concentration in the bioreactors and in the effluent 2.3.4. Phosphorus removal kinetics
depends on both nitrification and denitrification. It is depen- From an extensive study on an MBR UCT process, Parco et al.
dent on the nitrified nitrogen, which is linked to the amount of (2007) concluded that kinetic parameters for biological
nitrogen entering the system and indirectly to the amount of P-removal are comparable in MBR and CAS. The anaerobic
nitrogen assimilated into the sludge. But more significantly, P-release and acetate consumption rates, and the anoxic and
removal of nitrate depends on the amount of easily and slowly aerobic P-uptake rates were very close to the range of values in
biodegradable substrate used for denitrification (Xs, Ss), and the literature for conventional BNR systems with mixed
also the heterotrophic yield YH (Jiang et al., 2005). cultures. Moreover, the rates obtained for different concen-
Obviously residual nitrate also depends on denitrification trations of volatile suspended solids (VSS) indicated no effect
kinetic parameters. However, the higher the SRT, the more of the sludge concentration on these rates. Acetate
sensitive the denitrification rate will be to total biodegradable consumption rates were zero order with respect to acetate
COD (Xs þ Ss) and to the endogenous respiration. Specific concentration in agreement with the studies in the literature.
denitrification rates are extremely dependent on SRT via mass Aerobic and anoxic P-uptake rates indicate a relatively low
organic load (denitrification rates measured in MBRs are often fraction of denitrifiers in the PAO biomass (XDNPAO/XPAO was
very low at high SRT) and consequentially design guidelines around 15 to 36%).
(Pinnekamp, 2006) recommend a ratio VAE/VAX ¼ 50/50%, Jiang et al. (2008) used ASM2d to predict phosphorous
whereas it could be at least 75/25% in CAS. removal. With default parameters, the model overestimated
Denitrification rates were measured by Parco et al. (2007) nitrate concentration while underestimating phosphorous
for an MBR in UCT configuration (20 days SRT and a mass concentration. The authors calibrated the model simulta-
load of 0.14 gCOD/gMLSS.d). Conventional denitrification rates neously reducing SA production in the anaerobic compart-
of 0.25 mgNO3/mgSS.d (similar to CAS) were obtained. The ment and the aerobic/anoxic phosphorous uptake rate
authors concluded that kinetic parameters for denitrification (qfe ¼ 1d1, qpp ¼ 1.1 d1 and hNO3,PAO ¼ 0.4), by trial and error.
could be applied directly to MBR BNR systems. From these However, it should be noted that, after including the SMP
results, the conclusion is that the reduction factor for anoxic concept, these parameters could be restored to default ASM2d
growth (hg), and the anoxic heterotrophic yield (YHD) are values.
probably not different in MBR compared to CAS. In contrast to Recently, MBR research at full-scale (Silva et al., 2009)
the nitrification process, denitrification is apparently less showed that unexpected high biological phosphorous
modified by the membrane configuration. However, half- removal was obtained in MBRs not designed for EBPR. In the
saturation constants (KNO, KOH) which control the effect of low full-scale MBR located in Schilde (Belgium), significant bio-
concentrations of nitrate or oxygen on the kinetics have not logical P removal was reported though the process scheme did
been specifically determined yet for MBR. As for the nitrifica- not contain anaerobic compartments, effluent nitrate
tion, Manser et al. (2005) showed that floc size distribution had concentration was generally higher than 3 mg/l, (Bixio et al.,
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4283

2006; Fenu et al., in press). Silva et al., (2009) explained this the floc were monitored but not identified as variables having
phenomenon by the high floc compactness and density of EPS a significant influence on a-factor within the range of
inducing anaerobic micro-niches. The higher sludge viscosi- parameters studied.
ties in MBR systems (when compared to CAS systems), might
also impede mixing and hydrodynamic homogeneity, creating 2.3.7. Sludge production
artificial local anaerobic zones. These assumptions have not Total sludge suspended solid concentration and excess sludge
been fully demonstrated and need further investigation. production are clearly of great importance in WWTPs. A
realistic prediction of the concentration of active components
2.3.5. Oxygen uptake rate (OUR) (XHET, XAUT, XPAO) is in fact crucial for dynamic simulations.
The dynamics of the oxygen uptake rate (OUR) are generally MBRs systems are run at average to very high SRTs. In
considered sensitive to kinetic parameters. For example, these operating conditions, suspended solids are mainly
a highly loaded period can result in Xs accumulation, whereas composed of inert particulate matter, originating from both
during low loaded periods hydrolysis of these slowly biode- influent and biomass decay. High SRTs increase the amount of
gradable compounds controls the OUR dynamics (Ekama and energy spent on maintenance rather than on growth, and
Marais, 1986). However, in MBR processes this conclusion is no average biomass concentration notoriously increases, due to
longer valid during very low-loaded operation (high SRT). In the reduced quantity of biomass wasted. In aerobic zones,
that condition, the actual oxygen uptake rate (instantaneous specific OUR will decrease as a result of inert compounds
oxygen demand) varies mainly with the flux of biodegradable accumulation and reduction in the active fraction of biomass
substrate (Ss, Xs and SNH) entering the system. The stoichio- (Tan et al., 2008).
metric parameters (YH, YA) are then more important than the Moreover, in MBR systems operated at very high SRTs,
kinetics, as no Xs, Ss or SNH accumulation is normally biodegradation of COD fractions, considered inert in ASMs as
observed in that condition. described in the previous sections, induces an overestimation
of sludge production. This is definitely a major drawback of
2.3.6. Oxygen transfer rate (a-factor) ASM models for modelling MBR in a large range of SRTs
The a-factor is defined as the ratio of the volumetric transfer (Sperandio and Espinosa, 2008).Changes in the parameters YH,
coefficient under process conditions (i.e. with mixed liquor) to bH and fp can severely modify the predicted XTSS (Jiang et al.,
the clean water transfer coefficient. This is therefore 2005; Sperandio and Espinosa, 2008). The experimental esti-
a normalized parameter depending primarily on the mixed mation of these parameters has been attempted by several
liquor characteristics. Together with the hydrodynamic research groups and the main outcomes are reported below.
conditions (such as aeration types and powers), it determines Jiang et al., (2005) measured the conversion yield of hetero-
the standard aeration efficiency (SAE) of aeration systems trophs (YH) in MBR processes, through respirometric
(Krause and Cornel, 2007). In these respects, the energy effi- measurement with acetate. The obtained YH (0.72 gCOD/gCOD
ciency of activated systems can be described as inversely at 23  C) was higher than the default value in ASM1 (0.67
proportional to the a-factor, and the modelling of the biolog- gCOD/gCOD at 20  C), but this higher value is not specific to
ical aeration demand in MBRs should account for the effect of MBRs and it is probably due to the storage phenomenon with
specific process conditions on the a-factor. acetate which is easily converted to PHA (Majone et al., 1999;
The a-factor is mainly influenced by the viscosity of the van Loosdrecht and Heijnen, 2002). YH values obtained with
activated sludge which is associated with the MLSS concen- a single carbon source should be considered with precaution
tration (Krause et al., 2003; Germain et al., 2007). As the as they are substrate-specific. A similar value can be obtained
viscosity is unknown in many cases, Krause (2005) gives an with acetate as a pure substrate in the activated sludge
approximate function of the a-factor versus the MLSS process. The mean value of 0.63e0.67 gCOD/gCOD is then
concentration of a ¼ e0.056MLSS for the design of municipal more valuable for domestic wastewater considering the large
MBRs. As a result, a then equals 0.5 for an MLSS concentration variety of carbon source (carbohydrates, proteins, alcohols,
of 12 g/L. carboxylic acids.).
Actually, a bandwidth of values can be found. However, Heterotrophic decay rate (bh) seems to be close to the ASM
the lower a-factor in MBR applications constitutes a signifi- default value. In a first study, Jiang et al. (2005) measured bH
cant difference compared with the CAS processes: while for with respirometric tests (Vanrolleghem et al., 1999) and found
an MLSS concentration lower than 5 g/L the a-factor may be a value of 0.25 d1, lower than the default value (0.4 d1). The
in the range 0.7e1, it drops to a value of about 0.5 for 10 authors explained this lower value by the decrease of preda-
gMLSS/L and may go down as low as 0.2 for 20 gMLSS/L. tion in the MBR. However, in a subsequent study using ASM2,
Despite this, other factors also contribute to this behaviour. Jiang (2007) used the default value for bH in ASM2 (0.4 d1) with
Results from Germain et al. (2007) suggested that bound a good fit of the mixed liquor COD. Sarioglu et al. (2008) esti-
carbohydrates and soluble COD also had a secondary impact mated bH ¼ 0.24 d1 (endogenous respiration concept), similar
on the a-factor: a -factor increased with increasing bound to the value in ASM3.
carbohydrates (facilitating the formation of large flocs), and Modifications of YH, bH, and fp parameters need to be
decreased with increasing soluble COD (probably due to the carefully examined as they play a major role in other processes.
presence of surfactants), although with specific impact about For example an increase of YH would reduce the electron
half of that generated by the MLSS concentration. Note that consumption, i.e. oxygen consumption or nitrate removal. For
the parameters of bound proteins, bound COD, soluble this reason, it is considered preferable to fit XI, which is logi-
carbohydrate, soluble protein, and mass median diameter of cally wastewater specific, rather than other parameters.
4284 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

Table 4 summarizes the parameter set found by the biopolymers is in fact a specificity of MBR systems compared
authors who have calibrated MBR systems in urban waste- with conventional ASP.
water treatment. It includes some comments and references The above mentioned fractions accumulate in the mixed
to methods. liquor (inter alia, Drews et al., 2007; Malamis and Andreadakis,
2009), and are susceptible of biodegradation. Shin and Kang
2.3.8. Outlines (2003) report an initiation of the reduction of microbial prod-
Nitrification parameters seem to be the most affected by the ucts at SRTs longer than 10e20 days. This reduction is
differences between CAS and MBR. However, they depend on commonly agreed to be more significant for polysaccharides
both hydrodynamic and operational conditions. It is relatively than for proteins (inter alia, Al-Halbouni et al., 2008). With
risky to choose mA, bA, YA, and KNH and KOA by combining increasing SRTs, a reduction in the molecular weight distri-
values obtained from different studies, as the coherence of bution of the mixed liquor supernatant particles has also been
each set of parameters is necessary. It is thus recommended reported (Shin and Kang, 2003; Ng et al., 2006). In contrast, the
to give special attention to these parameters, measuring them accumulation of microbial products did not appear to trigger
with proper methods if necessary. any deviation from the expected biomass metabolic activity
Denitrification rates seem to be similar to CAS ones, and (Shin and Kang, 2003).
consequently parameters are not different. An exception is Current ASM models fail to account for many of these
the oxygen half-saturation constant KOH (used as inhibition specific MBR features. They neither distinguish between
parameter of oxygen for denitrifiers). This parameter could be protein and polysaccharide fractions, nor account for shifts in
affected, depending on the oxygen level in the denitrification the molecular weight distribution. But it is important to
zone and the small floc size which could ease the oxygen mass realize that, in systems with low organic loads (such as MBRs),
transfer. Moreover, a detailed characterization of the anoxic the retained molecules may have a significant impact on the
hydrolysis rate and the endogenous respiration rate is still metabolic path, allowing further use of carbon based metab-
needed, controlling both the slow and endogenous denitrifi- olites (Furumai and Rittmann, 1992). These shortcomings in
cation rate. the current ASM models could be overcome by expanding the
Concerning kinetic parameters for biological phosphorus models with EPS/SMP concepts.
removal, data and literature are still insufficient on this topic.
Phosphorus removal performance in MBRs seems to be 3.1.1. EPS and SMP definition
slightly better than in CAS in some cases, but data are still EPS are a complex mixture of Proteins (PN), acid, poly-
needed to demonstrate whether these differences are saccharides (PS), lipids, DNA and humic acids. They surround
explained by kinetic parameters of PAO species or different cells, create a matrix for microbial flocs and films (Liao et al.,
operating conditions (local anaerobic zone in biological tank, 2001) and allow micro-organisms to live continuously at high
absence of anaerobic zone of settlers,.). cell densities in stable mixed population communities
(Wisniewski, 1996; Laspidou and Rittmann, 2002a,b). They are
further differentiated into bound EPS, the fraction bound to
3. Application of modified ASMs to MBR the sludge flocs, and soluble EPS, the fraction able to move
processes freely between sludge flocs and the surrounding liquor
(Rosenberger and Kraume, 2002). As it is difficult to differen-
The objective of this section is to give a concise overview of tiate between SMP and soluble EPS, the latter are commonly
the application of modified ASMs. The terminology “modified” denominated as SMP. It must also be inferred that EPS analysis
should be interpreted in terms of modelling scope as well as relies on its extraction from the sludge flocs. So far, the
model structure. This section discusses the following aspect: scientific community has not agreed on a standard procedure,
(i) The impact of the phase separation mechanism on the and a comparison of data set generated from different
ASMs (ii) The influent fractionation for modified ASMs (iii) An extraction methods becomes complex (Judd, 2006; Liu and
overview of the stand-alone EPS and SMP models (iv) An Fang, 2002).
overview of ASM extensions with EPS/SMP concepts (v) The SMPs are defined as soluble cellular components or debris
over-parametrization of modified ASMs. that are released during cell lysis, diffuse through the cell
membrane, are lost during synthesis, or are excreted for
3.1. Integrating membrane rejection studies in ASM some purpose. Substrate utilization, biomass decay, and EPS
models hydrolysis are believed to be the major processes contrib-
uting to SMP formation. With respect to “SMP formation
The phase separation mechanism is the main technological through substrate utilization”, if intermediate products of
distinction between CAS and MBR systems. The mechanism is metabolic processes are included in the SMP definition,
essentially a sieving effect performed by membranes with substances that do not have a microbial origin but come
a nominal pore size of typically 0.02 up to 0.2 mm: all particles directly from the substrate tend to be wrongly included
whose size is larger than the membrane pore size are retained, (Noguera et al., 1994). The problem has been taken into
whereas the smaller dissolved fractions are not. Hence, flocs, serious consideration by a number of researchers, but within
bacteria, biopolymers such as polysaccharides and proteins, the scopes of this paper, it is sufficient to refer to SMPs as
and organic colloids are in great extent retained. Humic and “any soluble material that leaves the effluent from a biolog-
low molecular weight substances can instead pass the ical system that was not present in the influent”, (Barker and
membrane (Drews et al., 2006). The retention of the Stuckey, 1999).
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4285

3.1.2. Modelling objectives of modified ASMs to MBR In this work, efforts at ASM model extensions with SMP
processes and EPS concepts in the literature are reviewed along with
From a end-user point of view, expanding the models with their relevance for MBRs.
EPS/SMP concepts is suggested only if the following modelling
objectives are pursued: 3.2. Influent fractionation for modified ASMs

3.1.2.1. Linking biology with fouling. The extension of ASM When introducing the SMP concentration as a state variable in
models with soluble microbial product (SMP) and Exo Poly- the model, one of the key issues for COD fractionation is the
meric Substances (EPS) concepts is undoubtedly crucial when determination of inert components (Si, Xi) with proper
predicting membrane fouling. SMP/EPS have been proved to methods, considering the fact that new “inert” substances can
accumulate in MBR systems as a consequence of a high be produced in the process as by-products of the microbial
membrane rejection and low biodegradability (Shin and Kang, metabolism.
2003; Drews et al., 2007; Liang et al., 2007). Their influence on In this sense, an early modification of ASM1 was proposed
fouling, or their use as indirect indicators of fouling propensity by Lu et al. (2001) by modelling a bench-scale MBR fed with
through biomass deflocculation (De la Torre et al., 2009), has synthetic sewage. Wastewater was characterized by means of
been evaluated and acknowledged by numerous researchers batch tests aimed at quantifying the inert soluble fraction
(inter alia Rosenberger et al., 2005, 2006; Jarusutthirak and Amy, according to the method proposed by Henze et al. (1987).
2007; Le-Clech et al., 2006; Zhang et al., 2006; Wu et al., 2007). Activated sludge from the bioreactor (MLSSe13,000 g m3) was
Based on these considerations, the SMP/EPS would be washed by ultrafiltered effluent and, after washing, it
a necessary input for a filtration model and, hence, needs to be was diluted to about 2000 O 3000 g m3. The filtered feedwater
incorporated in the biological model in order to explain or was afterwards fed to the sludge with an initial COD concen-
predict the filtration performance of the system. tration of 200 g m3. The steady value of soluble COD
concentration reached after approximately 4e5 h was
3.1.2.2. Soluble COD prediction. The supernatant of settled assumed to represent the inert soluble COD in the influent
mixed liquor is mainly composed of living cells and microbial (SII). This concentration was slightly reduced in order to take
products. So far it is certain that, in municipal applications, the presence of SMP into account. Similarly, the XI fraction
effluent soluble organic matter of activated sludge water from was determined by the same method using substrate pre-
conventional activated sludge systems is mainly SMP treated by sonication and by subtracting the SI contribution.
(Namkung and Rittmann, 1986; Jarusutthirak and Amy, 2007) Finally, SS and XS were given by the difference between
and the same can be said for MBR processes (Lu et al., 2002; Le- soluble COD and SI and particulate COD and XI respectively.
Clech et al., 2006; Rosenberger et al., 2006; Aquino and More recently, an extension of ASM1 to the SMP concept
Stuckey, 2008). SMP and EPS are microbial products, not has been proposed by Di Bella et al. (2008) in an integrated
active cells, and represent energy that is not invested in cell model for physical-biological wastewater organic removal in
growth (Laspidou and Rittmann, 2002a,b). Ignoring SMP and MBRs. The model couples the SMP-extended ASM1 firstly
EPS formation could lead to a general overestimation of true proposed by Lee et al. (2002) with a physical model accounting
cellular growth rates and this would severely under predict for organics removal in both the membrane and the cake-
the COD effluent (Jiang et al., 2008). This is commonly artifi- layer. Here, the COD fractionation comes from a multi-step
cially corrected by an overestimation of Si in the conventional process including the trial-and-error on MLSS and effluent
ASM1 model. Finally, it should be pointed out that the COD COD concentrations and an automatic calibration with 10,000
effluent predictions are not a real concern in municipal MBR Monte Carlo simulations intended to define the values for the
systems since values are generally low and stable. most sensitive parameters of the whole model (including the
physical sub-model). Interestingly, in this work, a relevant
 Model high SRT processes fraction of influent COD is attributed to the heterotrophic
active biomass in the wastewater (Xhet), which is usually
Furumai and Rittmann (1992), investigated the interactions assumed to be negligible in the above mentioned research.
between nitrifiers and heterotrophs. They reported that, in In the successful application of a modified ASM3 including
a CAS at high SRT, SMP formed by nitrifiers promoted SMP to a pilot-scale MBR treating pre-settled municipal
heterotrophic growth and reduced the minimum substrate wastewater working at almost 50 days, Oliveira-Esquerre et al.
concentration necessary for heterotrophs, allowing hetero- (2005) used OUR profiles to quantify the SS and XS fractions in
trophic growth at very low influent organic concentration. the influent feed. In detail, they measured the inert soluble
MBR processes generally run with average to high SRTs and fraction as COD concentration in the permeate after 12 h’
the above considerations could be relevant at low F/M ratio (in aeration, thus degrading the SMP contribution.
MBRs, easily as low as 0.01 g COD/g MLSS.d). To sum up, the implementation of additional processes for
Bound EPS has been experimentally shown to accumulate specific MBR modelling purposes in the ASM framework is not
with decreasing SRTs (Ng and Hermanowicz, 2005; Massé reflected in a different COD fractionation. As for unmodified
et al., 2006), with a corresponding trend on SMP activated sludge models, a general trend towards the trial-
(Rosenberger et al., 2006; Al-Halbouni et al., 2008). Beyond the and-error fractionation is observed when longer sludge ages
prediction on membrane filtration, a proper modelling of this are operated, whereas the respirometry-based approach is
fraction is beneficial for sludge production prediction. generally preferred at more conventional values of SRT.
4286 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

3.3. Overview of stand-alone EPS and SMP models yield a ratio of bound EPS to VSS (bEPS/VSS) close to the
experimental value of PN-like-EPS to biomass. Keps resulted in
Some models have been developed as stand-alone descrip- 0.03 mgCODEPS/mgCODcell/d, which is in accordance with
tions of the concepts of production and degradation of EPS Robinson et al. (1984).
and SMP. Others have focused on integrating the latter
concepts into the ASM type of models. This distinction is used 3.3.2. Stand-alone SMP models
in the overview given below. SMP models have been developed since the late eighties.
Namkung and Rittmann (1986) put forward the following SMP
3.3.1. Stand-alone EPS models subdividision: 1. UAP, i.e. SMP that are associated with
An early model to characterize microbial products formation substrate metabolism and biomass growth and are produced
was proposed by Luedeking and Piret (1959) for the fermen- at a rate proportional to the rate of substrate utilization. 2.
tation of lactic acid (Eq (1) in Table 6). In this equation, the first BAP, i.e. SMP that are associated with biomass decay and are
term accounts for EPS formation associated with a first order produced at a rate proportional to the concentration of
growth (with K1 as the fraction of substrate electron shunted biomass. This method of subdivision has been widely
to EPS formation) and the second term represents EPS accepted.
formation associated with a non-growth term. The main There seems to be a general consensus on UAP formation
objections to this simple model were (i) the inconsistent and degradation mechanisms. UAP formation results from
production rate values proved by Luong and Muchaldani substrate utilization and is proportional to the rate of
(1988); (ii) the fact that a mechanism of EPS dissolution was substrate utilization and biomass concentration (Luedeking
not included although theoretical evidence supported this and Piret, 1959; Namkung and Rittmann, 1986; Lu et al., 2001;
hypothesis (Laspidou and Rittmann, 2002a,b). Laspidou and Rittmann, 2002a,b). Contrary to Laspidou and
Laspidou and Rittmann, 2002a,b studied the EPS mass Rittmann’s approach, Lu et al. (2001) differentiated the active
balance in a continuous flow reactor (Eq (2) in Table 6). In this biomass into Xhet and Xaut eqs. (3e6) in Table 7) and UAP
equation, the first term is the product of the substrate utili- cannot be consumed but only produced by Xaut (Fig. 1).
zation rate rs, the active biomass Xa and the part of substrate Recently, a modelling study conducted with aid of LC-OCD
electrons shunted to EPS formation, Keps. The second term and batch tests characterization (Jiang et al., 2008) further
quantifies the rate of EPS loss due to hydrolysis, using a first- hypothesized two types of UAPs: the UAPs produced during
order relationship with respect to EPS. The hydrolyzed EPS storage formation of readily biodegradable COD would have
would then become soluble EPS or SMPs. a lower MW and would be biodegradable; the UAPs produced
The Laspidou and Rittmann model was criticized on 2 during the utilization of storage products would instead have
major points: a higher MW and would be more refractory. However the
study was limited to the modelling of the first proposed UAP
(i) The formation of bound EPS is said to be growth-associ- type.
ated, and produced in direct proportion to substrate Unlike for the UAP case, there is no consensus on BAP
utilization. During transient conditions of organic shock production and degradation mechanisms. With regard to the
loads this would theoretically lead to very high EPS BAP production, Laspidou and Rittmann (2002a,b) suggested
concentrations. Aquino and Stuckey (2008) did not that BAP were produced solely by EPS hydrolysis (Eq.(7) in
observe this, and proposed to model the formation of EPS Table 7). This hypothesis was shown to be weak. (i) Hydro-
as a non-growth-associated product (eq. (3) in Table 6). lyzed (soluble) EPS and BAP revealed different physicochem-
This entails that the EPS formation rate can only be high ical properties (Ramesh et al., 2006). Aquino and Stuckey
at high concentrations of biomass. (ii) The EPS hydrolysis/ (2008) demonstrated that both soluble EPS and cell lysis
dissolution rate was 0.17 d1 in Laspidou and Rittmann’s products were the sources of BAP (Eq.(8) in Table 7). (ii)
works, but was recently reduced to 0.02e0.03 d1 by Jang Regarding the kinetics, Jang et al. (2006) found that the BAP/
et al. (2006), and Aquino and Stuckey, 2008. In this respect, UAP kinetics of the Laspidou and Rittmann model could not be
a valuable hypothesis is that the high concentration of applied in their specific tests. When employing a maximum
“hydrolysis end products” would reduce the hydrolysis utilization rate for UAP/BAP, they found that the UAP/BAP
rate. (Jang et al., 2006). utilization exceeded the formation. It was therefore assumed
that UAP/BAP accumulation inhibited the degradation rate.
It is interesting to note that, regarding EPS formation In the ASM models experiences, EPS are generally not
kinetics, Aquino and Stuckey (2008) proposed to calculate the considered and BAP are not produced from EPS hydrolysis but
EPS formation rate Keps, so that the steady-state model would from cell lysis processes (Jiang et al., 2008) or additionally from

Table 6 e EPS model rates in literature.

Leudeking and Laspidou and Rittmann, 2002a,b (Eq. (2)) Aquino and Stuckey, 2008 (Eq. (3))
Piret., 1959 (Eq. (1))

reps ¼ K1 m X þ K2 X rEPS ¼ Keps rs Xa  Khyd EPS rEPS ¼ Keps Xa  Khyd EPS

 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4287

Table 7 e Summary of main formation and degradation rates for UAP and BAP.
UAP formation rate
Laspidou and Rittmann, 2002a,b (Eq. (4)) Lu et al., 2001 (Eq. (5))

S rUAP ¼ mhet Xhet or rUAP ¼ maut Xaut 

rUAP ¼ K1 rquap Xbm
Ks þ S

UAP degradation rate

Laspidou and Rittmann, 2002a,b (Eq. (6)) Lu et al., 2001 (Eq. (7))

UAP SSMP
rUAP ¼ rquap Xbm rUAP ¼ mSMP Xhet 
Kuap þ UAP KSMP þ SSMP

BAP formation rate

Laspidou and Rittmann, 2002a,b (Eq. (8)) Aquino and Stuckey, 2008 (Eq. (9))

rBAP ¼ Khyd EPS rBAP ¼ K2 X þ Khyd EPS

BAP degradation rate
Laspidou and Rittmann, 2002a,b (Eq. (10)) Jiang et al., 2008 (Eq. (11))

BAP rbap ¼ Kh;bap Sbap XH

rbap ¼ qbap Xbm
Kbap þ BAP

(*) Switching functions are not reported since their introduction depends on the process where UAP is formed or degraded.

hydrolysis of particulate biodegradable organic matter (Lu et al., 2008; Zarragoitia-González et al., 2008; Silva et al.,
et al., 2001). 1998; Lu et al., 2001, 2002). Drews et al. (2007) studied the
The BAP degradation has been considered by many impact of ambient conditions on the rejection of MBRs.
researchers (Lu et al., 2001; Laspidou and Rittmann, 2002a,b; According to the latter authors, the rejection of SMP compo-
Aquino and Stuckey, 2008; Oliveira-Esquerre et al., 2006) as nents appears to decrease at higher temperatures and higher
a direct degradation and only Jiang et al., (2008) assumed nitrification activity. The modelled rejection factors should
hydrolysis (Eq. (12) at Table 7). Jiang reported that 63% of the thus take experimental observations in considerations.
BAP had a molecular weight larger than 20 kDa, which makes The predominance of UAP over BAP or vice versa needs to be
it unlikely that such large molecules pass the cell membranes discussed for two main reasons (i) One of the two fractions
directly. They therefore concluded that BAP were hydrolyzed could have more impact in terms of soluble COD, and an SMP
yielding fermentable COD. Overall, Jiang et al. (2008) proposed fraction could be neglected leading to a simplification of the
3 steps for each biomass type to determine the complete BAP model. The exclusion of a fraction would be of great benefit in
dynamics, (formation, hydrolysis, degradation). terms of modelling since it would reduce the number of
The SMP equilibrium concentration results from produc- variables. In fact ASM models in the early nineties used to
tion/degradation mechanisms but also depends on SMP consider only BAP fractions. However, the process conditions
retention by the membrane. The latter aspect is given little may determine the predominant fraction case by case. A rule
consideration in MBR modelling. Authors have mainly of thumb is that the BAP fraction tends to predominate at high
measured or fitted it as a steady value, independently of the SRTs (Furumai and Rittmann, 1992) or in steady-state condi-
process specificities. The percentage of SMP permeating through tions (Aquino and Stuckey, 2008), while UAP predominate in
the membrane ranges from 0 to 100% (Jang et al., 2006; Jiang the SMP production when the rate of substrate degradation is
high (Aquino and Stuckey, 2008; Laspidou and Rittmann,
2002a,b); (ii) One of the two fractions could have more
impact in terms of fouling predictions. Zhang et al., (2007)
operating an MBR fed with external carbon source proved
that the concentration of large molecule OM was greater in
BAP than that in UAP, (over 18% of molecules with more than
100 K), being the main cause of the increasing resistance.
Conversely, Rosenberger et al. (2006) observed that the SMP
fraction produced at different sludge retention times in MBR
units operated with municipal wastewater led to a specific
long-term fouling rate (per SMP mass unit) higher with 15d
SRT than with 8d SRT, even though much more SMP was
produced at 8d SRT than at 15d SRT. This demonstrates that
the SMP fraction produced by the biological system differs not
only in quantity but also in quality depending on the envi-
Fig. 1 e Alpha-value in dependence of MLSS (Krause and ronmental conditions, with crossed impact of both criteria on
Cornel, 2007). the fouling propensity of the mixed liquor.
4288 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

3.4. Overview of ASM-extensions incorporating EPS/ were calibrated by fitting but the SMP concentrations were not
SMP concepts shown. Sensitivity tests report that YSMP, and gUAP,A have
a strong influence on the majority of output variables while
The SMP concept was incorporated into activated sludge bBAP,h and mSMP mainly affect NH4 and NO3. The sensitivity
model No. 1 (ASM1) in the early 90s (Orhon et al., 1989; Artan tests confirm that, once SMP are embedded in the ASM model,
et al., 1990). First, a very simple SMP model including only their intermediate role in decay and substrate metabolism is
BAP was developed (Orhon et al., 1989). So-called SP (equiva- very relevant. Identification of the kinetic parameters intro-
lent to BAP) is produced proportional to the hydrolysis of duced is thus crucial for the prediction of the effluent quality.
particulate COD (Xs) and they are assumed non-biodegrad- Ahn et al. (2006) tried to include EPS in the ASM-SMP
able. The model was further developed to include UAP (Artan models, though their use in ASM works has generally been
et al., 1990). However, this model combines the concepts and avoided. EPS were modelled in eq.(1) in Table 6, thus excluding
degradation kinetics of UAP and BAP resulting in strong an explicit EPS loss term, and BAP were produced uniquely by
parameter correlations. Moreover, the model lacks experi- EPS hydrolysis. For the modelling process, 5 processes were
mental support. added: hydrolysis of EPS, EPS formation by heterotrophs, EPS
Lu and coworkers have incorporated a very complex SMP formation by autotrophs, UAP formation by heterotrophs, and
model into ASM1 (Lu et al., 2001) and ASM3 (Lu et al., 2002) in UAP formation by autotrophs (Fig. 2). In order to describe the
MBR studies (Fig. 2). However, the COD of their SMP model is newly added processes, 8 new parameters were introduced
not balanced, i.e. the loss of substrate COD is not equal to the (mSMP, kUAP, kUAPa, kBAP, KSMP, kEPS, kEPSa, and KXEPS), yielding an
sum of formed UAP COD, formed biomass COD and consumed over parameterized model. In this model, the effect of SRT was
oxygen. In addition, 8 SMP-related parameters are introduced, not observed in SMP production but rather in EPS production,
but the experimental results available for model calibration and the experimental results showed good agreement with
are limited to steady state soluble COD (SCOD) measurements the simulation results. EPS concentration was sensitive to
(Jiang et al., 2008). The kinetic parameters are estimated by Keps, Kbap and SRT, while Kuap sensitivity was low. However,
calibration but strong parameter correlation may hamper kinetic parameters are not reported, SMP behaviour not
correct determination. Thus, the fitting does not convincingly described, and the model lacks an appropriate calibration
demonstrate the validity of the model structure and param- (Jiang et al., 2008).
eter values. Oliveira-Esquerre et al., (2006) introduced SMPs in ASM3.
Other authors have also adopted the Lu model. Zarragoitia- UAP and BAP were lumped into a general term MP. Overall, the
González et al. (2008), adopted the Lu ASM1 model (with the model comprised 5 new SMP kinetic parameters (gMP,H, gMP,A,
exception of nitrification processes) and linked it with kMP, fb, Ymp) whose values were adopted from Lu et al. (2001),
a membrane fouling model where SMPs were used to estimate and 2 new processes. The work focused on understanding the
the bound EPS concentration in the sludge supernatant peculiarities of an SMP model in ASM3 when compared to the
according to the equation SUAPþSBAP/0.8XTSS. Kinetic param- ASM1. In the ASM1 model it is assumed that slowly biode-
eters of Lu’s model were partly adopted. This work did not gradable substrates are hydrolyzed before their use for
include SMP modelling results. growth. In the ASM3 model it is assumed that all organic
Di Bella’s (2008) work was also based on Lu’s model. The substrates are directly converted into stored material and that
work was important since it focused on describing the fate of stored compounds are subsequently used as a carbon and
COD through the pilot MBR by distinguishing the COD removal energy source for growth purposes. Consequently, as the
contribution of (i) the physical membrane and (ii) the cake specific rate of hydrolysis of MPs in ASM1 is considerably
layer according to its depth. COD effluent was fairly well lower than the specific rate of storage in ASM3, it becomes
predicted. However, the model parameters were not in line a rate-limiting factor in the uptake of MPs. This explains why
with other experiments (Table 8). SMP kinetic parameters ASM3 gave a markedly low MP concentration (0.75 gCOD m3),

Fig. 2 e Concepts of the Lu et al. model (2001) (left) and the Ahn et al. model (2006) (right).
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4289

Table 8 e SMP/EPS Kinetic parameters in recent literature works.

EPS/SMP modelling

Affinity constants Formation Degradation EPS

KUAP KBAP SMP UAP BAP KBAP KUAP fSMP KEPS KHYD
1
[gCODSMP/l] *** ** [d ] [mgCODproduct/mgCODcell/d] [d1]

Furumai and Rittmann, 1992 0.5 0.2 0.1* 1

Laspidou and Rittmann, 2002a,b 100 85 0.05 0.17 0.07 1.27 0.18**** 0.17
Jang et al., 2006 100 85 0.05 0.02 0.07 0.4 0.18**** 0.02
Aquino and Stuckey, 2008 500 500 0.2 0.0034 0.03 1.2 0.03 0.03

ASM related works

Affinity constants Formation Degradation rates

KSMP mSMP KBAP fSMP fUAP fBAP KSMP Kh,BAP Kh,UAP

[gCODSMP/l] [d1] [d1] [] [] [] [d1] [d1] [d1]

Lu et al., 2001 30 0.7 0.4 0.38

Lu et al., 2002 60 2.5 0.01 0.3
Lee et al., 2002 30 0.7 0.4 0.1
Oliveira-Esquerre et al., 2006 30 0.4
Zarragoitia-González et al., 2008 30 0.7 0.22 0.38
Di Bella et al., 2008 133 8.3 0.33 0.82
Jiang et al., 2008 0.0963 0.0215 7.4E-7 0.0102

(*) in [mgCODproduct/mgCODcell/d]; (**) in [mgCODuap/mgCODsub]; (***) in [mgCODcell/mgCODsmp]; (****) in [mgCODeps/mgCODsub].

while a value of 80 gCOD m3 was obtained with ASM1. Both monitoring of a batch system in famine conditions. UAP
models would, however, give similar values of MP concen- parameters were calculated by PN and PS concentration
tration if no storage was considered in ASM3 and MPs were monitoring of a batch system in external substrate excess
directly used for bacterial growth, as assumed in ASM1. conditions. The degradation rate equations were chosen in
Jiang et al. (2008) introduced the SMP concept in an ASM2d order to minimize the number of unknown parameters. The
model. This work differed considerably from the previous determination of the unknown parameters was reported by
ASM-SMP models. BAP and UAP were degraded by hydrolysis the authors as complex, especially in the case of UAP,
steps, creating 3 new processes and imposing variations in 13 because BAP and acetate could interfere in the UAP
other processes. SMP kinetic parameters, production and measurements. The main criticism with regard to this batch
hydrolysis rate of SMPs (subdivided by UAPs and BAPs) were approach could reside in: (i) The BAP batch test is run up to
investigated through specially designed experiments. The SCOD values of 200 mg COD/l. This value is very high when
experiments were associated with an LC-OCD characteriza- compared to WWTP supernatant COD; (ii) UAP tests are
tion, offering interesting outcomes. Jiang et al.’s (2008) UAP substrate specific; (iii) Both BAP and UAP tests are dilution
and BAP model peculiarities are in this review discussed in the dependent as shown by Laspidou and Rittmann; (iv) The
stand-alone models section. washing of biomass prior experiments is arguable because
the initial soluble COD of the supernatant would conse-
3.5. Model identification e UAP/BAP kinetics quently be very low. But it is questionable whether the rate of
SMP production/degradation are concentration dependent.
Two problems have been identified with SMP/EPS models. The This issue deserves further attention.
kinetic parameters are not easily determinable experimen-
tally and the models are usually over parameterized. Saroj 3.5.1. Outlines
et al., (2008) infer that incorporation of EPS/SMP in ASM The use of an ASM expansion with the EPS/SMP concept is
would tend to worsen the model identification process, which encouraged only if the following modelling objectives are to be
is crucial in any ASM calibration exercise. pursued: (i) linking biology with fouling, (ii) soluble COD
As can be seen in Table 8, in order to overcome these predictions (iii) Model high SRT processes.
problems several recent models made use of the Lu et al. The ASM extension with EPS/SMPs concept creates diffi-
model identification values, but this is not a good practise if culty in identifying the newly proposed parameters. Modellers
SMP dynamics are related to specific processes or influent have to implement strategies to reduce these parameters
composition. The lack of validation campaigns confirming according to experiences reported in the literature and the
the theoretical models is striking. Few researchers have process specificities. Parameter reduction strategies have
made efforts in the experimental determination of the new been adopted by (i) coupling UAP/BAP into SMPs (Oliveira-
kinetic parameters. Jiang et al. (2008) calibrated their model Esquerre et al., 2006); (ii) excluding EPS modelling (Lu et al.,
once BAP and UAP kinetic parameters had been experimen- Oliveira-Esquerre et al., Lee et al. Jiang et al.); (iii) modelling
tally derived. BAP parameters were calculated by sCOD the EPS (as sole AS foulants) with equations external to the
4290 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

ASM model (Saroj et al., 2008); (iv) excluding a water fraction Inorganic Suspended Solids (ISS) in the ASM framework. This
UAP/BAP when negligible (Orhon et al., 1989). model assumes ISS in the sludge to be due to the accumula-
It would always be preferable to determine the kinetics tion of influent mineral solids and the uptake of inorganic
parameters by specific batch tests, as suggested by Jiang et al., dissolved solids (IDS) by ordinary heterotrophic organism
(2008), despite the limitation of this approach as discussed (OHO) and phosphate accumulation organisms (PAO). In this
above. case, a suitable calibration of parameters ”fOHO” and “fPAO”
(inorganic solids content for OHO and PAO respectively) could
“mask” the gradual solubilization of influent inorganics at
4. Outlook and future perspectives long sludge ages. In addition, a solubilization mechanism
should be mathematically introduced in the ASM matrix.
In this final section, the authors would like to point out what
they feel as the main shortcomings of the current state-of- 4.2. Process kinetics for the unmodified ASMs
the-art of and future challenges in biological modelling of
MBRs. The section is structured similarly to the previous Concerning the biokinetic parameters of unmodified ASM1,
sections. most of the biological processes have been widely investi-
gated. Little, on the other hand, is known about the possible
4.1. Influent characterization for the unmodified ASMs specificities of biokinetics for phosphorus removal in MBR.
The available literature does not show significant differences
Notwithstanding the fact that a large number of ASM appli- between MBR and CAS calibration parameters. However,
cations have been reported in MBRs, some key factors still recent microbial studies have shown the presence of PAOs in
have to be further investigated and considered in wastewater several full-scale MBR not specifically designed for biological
characterization. One of these relates to the assessment of the phosphorous removal. Considering that a volume percentage
active heterotrophic biomass Xhet in the influent wastewater of the anoxic reactor is anaerobic is not an applicable solution
that, although usually neglected in CAS modelling, needs to be in all cases (since nitrate supernatant concentration may not
better addressed when modelling membrane bioreactors. In allow it) and predicting phosphate effluent concentrations in
fact, from a theoretical point of view, longer sludge ages lead these conditions remains rather problematic. This aspect
to decreases in the percentage of active biomass in the MLVSS. needs further investigation.
Therefore, the higher the SRT, the less negligible the new
biomass entering the plant via the influent becomes. This 4.3. Process kinetics for the modified ASMs
contribution is influenced by the characteristics of the sewage
system (separate/combined), the presence and typology of EPS and SMP concentrations are sensitive to biomass
pre-treatment units, the residence time in the sewage pipe- production, hydrolysis and degradation rate. Few detailed
lines and the possible presence of toxic compounds due to experimental data are available for calibrating the rates of
industrial discharges. Possibly, a combination of well-known each process independently. Models including the EPS/SMP
respirometric methods for Xhet determination (Kappeler and concept show that EPS constitute a large amount of the
Gujer, 1992) and advanced biomass characterization tools organic reserve in a bioreactor and their hydrolysis can impact
such as flow-cytometry (Ziglio et al., 2002; Foladori et al., 2007) the SMP release in the sludge water very significantly. Slow
can be adopted at experimental level to actually understand hydrolysis of microbial products (in aerobic, anoxic or anaer-
the need for further calibration efforts. obic conditions) is a poorly understood process which defi-
Another open issue in the application of ASMs to long-SRT nitely needs more attention.
membrane bioreactors is the fate of inorganic compounds in The SMP equilibrium concentration results from produc-
the influent (mineral suspended solids), whose concentration tion/degradation mechanisms but also depends on SMP
clearly depends on the upstream operation units (presence/ retention by the membrane. The latter aspect is not suffi-
absence of sand removal and/or primary settling tank, fine ciently considered in the works reviewed and authors have
screening meshes, run off, infiltration into the sewers). Some mainly measured or fitted it as a steady value, independent of
experimental evidence reports solubilization of inorganic the process variations. The percentage of SMP permeating
solids entering the system: Laera et al. (2005) showed the loss through the membrane ranges from 0 to 100% (Jang et al.,
of a significant amount of inorganic particulate matter in 2006; Jiang et al., 2008; Zarragoitia-González et al., 2008; Lu
a bench-scale MBR operated with no sludge wasting, and they et al., 2001, 2002). Quantification of this transport depending
suggested that hydrolysis and solubilization produce mole- on membrane type, process conditions and fouling charac-
cules with a smaller size than the membrane pores, which teristics needs further investigation.
leave the reactor with the effluent permeate. In the above The subdivision among UAP and BAPs has been widely
mentioned work, Sperandio and Espinosa (2008) found that up acknowledged and modelling efforts for their production and
to 50% of influent mineral solids were converted into soluble consumption are under way. However, recent observations
inorganics at a sludge age of 110 days. Therefore, in the tend to show how the SMP/bound EPS ratio (i.e. the state of
specific case of ASM3 application to MBRs, this would mean at flocculation of the biomass) also depends on process distur-
very long SRT an adjustment of the absolute values of bances such as shock loads, shear stress, temperature, pH and
composition parameters for MLSS (iTS_XI, iTS_XS, iTS_XBM, iTS_X- oxygen shocks (inter alia Drews et al., 2007). An important
STO). An alternative option could be given by the introduction question is whether the current models are valid for inclusion
of the model proposed by Ekama and Wentzel (2004) for of these process disturbance effects. Most probably they are
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4291

not and model extensions are required in order to predict SMP summarizing paragraphs at the end of each section are
dynamics with regard to process operation and process intended to guide end-users in the current availability and
disturbances. applicability of specific models and techniques, their limita-
Finally, as sufficiently noted in the text, the determination tions and possible pitfalls. This should assist in modelling
of kinetic parameters of the SMP/EPS models is a major municipal laboratory, pilot and full-scale MBRs, either for
concern. Over-parametrization or lack of identifiability performance optimization or for decision support.
appears as a typical problem of ASM model extensions. Is it The experience reported in this review proves that, when
possible to mathematically overcome the problem or do we modelling purposes do not differ from effluent characterization,
need to systematically characterize each process oxygen demand and sludge production, ASMs are very relevant
experimentally? to MBR applications. However, particular care needs to be taken
since the specific conditions present in MBRs are reflected in
4.4. Application of ASM models at full-scale, hydraulics some important discrepancies when compared to ASM default
and aeration parameter values. In terms of biokinetic studies of the different
processes, the authors have preferred not to propose a general
Most modelling exercises and conclusions are based on lab- set of values, since, as research results show, the values are very
scale or pilot-scale experiments. However, in order to take the dependent on the operating conditions. Instead, researchers’
step to usage and full exploitation of the benefits of biokinetic work has been brought together in concise tables that can,
models for plant design and operation, more full-scale studies indicate common directions for each process.
are required. More attention should be paid to full-scale model In cases of ASMs developed for specific purposes related to
applications in order to confirm the applicability of the current MBR operations, and with particular care to fouling prediction
findings. This problem is not new, however, as it is a known modelling, the knowledge on SMP and EPS modelling has been
issue in CAS modelling as well. In principle, the models critically reviewed on the basis of the literature, including the
described in chapter 2 (unmodified ASM) should be suitable most recent developments. Progress in tackling the over-
for full-scale use, bearing in mind some points requiring parametrization of the extended models has been highlighted,
attention as stipulated. as this paper intends to serve as an incentive to bring scientific
Full-scale plant aeration as well as hydraulic character- efforts into practice.
ization are key issues in modelling MBR. However, they have Finally, the outlook and future perspectives have been
not received much attention to date (especially hydraulics). systematically highlighted and proposed for each section.
The contribution of coarse bubble aeration (membrane The paper provides a guide for different end-users of
scouring) to the overall oxygen supply for oxic processes mathematical models for MBR, be they people active in
needs to be better characterized. In the case of sMBR systems, process design and operations, or academics who want to
there is a need to couple valid kinetic ASM models with more learn about the current state-of-the-art or detect current
accurate hydraulic models (e.g. CFD). Intensive hydraulic shortcomings in MBR modelling.
characterization of aerobic/anoxic basins is still insufficient.
When membranes are directly submerged in the bioreactor, it
becomes difficult to correctly predict the influence of
membrane aeration on the overall performance. Fine bubble Acknowledgements
aeration (for biological needs) is generally regulated by on/off
control, for minimising energy consumption and promoting This review paper is an outcome of the “Liaison Group on MBR
denitrification through increased anoxic reactor volume. But Modelling” that was established within the coalition of Euro-
membrane coarse bubbles aeration rate is more or less pean projects, MBR-Network (http://www.mbr-network.eu/).
constant and contributes to the oxygen supply, influencing The authors would like to thank the European Commission for
nitrification in a complex hydraulic system with heteroge- its financial support through the EUROMBRA project (Contract
neous oxygen concentration. A better balance between the No. 018480, FP 6thdGlobal Change and Ecosystems), the
different sub-models (biokinetic model, aeration model, AMEDEUS project (Contract No. 018328, FP 6thdGlobal
hydraulic model) should, therefore, be investigated. Change and Ecosystems), and the MBR-TRAIN project (Marie
Curie Host Fellowship for Early Stage Research Training sup-
ported by the European Commission under the 6th Frame-
5. Conclusions work Programme with Contract No. MEST-CT-2005-021050).

A concise overview of the most recent literature on ASM-

based MBR modelling is given in this review work. Clarity was references
sought by categorizing models as unmodified or modified
ASMs, underlining the strong relationship with the modeller’s
target: modelling to improve process performance or model- Ahn, Y.T., Choi, Y.K., Jeong, H.S., Shin, S.R., 2006. Modeling of
extracellular polymeric substances and soluble microbial
ling for further understanding of the process.
products production in a submerged MBR at various SRTs.
The paper has aimed to extract relevant information that
Water Sci. Technol. 53 (7), 209e216.
seems commonly agreed on within the scientific community Al-Halbouni, D., Traber, J., Lyko, S., Wintgens, T., Melin, T.,
in this particular area, and also to highlight contradictory Tacke, D., Janot, A., Dott, W., Hollender, J., 2008. Correlation of
observations and conclusions present in scientific papers. The EPS content in activated sludge at different sludge retention
4292 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

times with membrane fouling phenomena. Water Res. 42 Fenu A., Roels J., Wambecq T., De Gussem K., Thoeye C., De
(6e7), 1475e1488. Gueldre G., Van De Steene B. Energy audit of a full scale MBR
Aquino, S.F., Stuckey, D.C., 2008. Integrated model of the production system. Desalination, in press.
of soluble microbial products (SMP) and extracellular polymeric Ficara, E., Rocco, A., Rozzi, A., 2000. Determination of nitrification
substances (EPS) in anaerobic chemostats during transient kinetics by the ANITA-DOstat biosensor. Water Sci. Technol.
conditions. Biochem. Eng. J. 38, 138e146. 41 (12), 121e128.
Artan, N., Orhon, D., Baykal, B.B., 1990. Implications of the task Foladori, P., Bruni, L., Andreottola, G., Ziglio, G., 2007. Effects of
group model I. the effect of the initial substrate concentration. sonication on bacteria viability in wastewater treatment
Water Res. 24 (10), 1259e1268. plants evaluated by flow-cytometry: faecal indicators,
Barker, D.J., Stuckey, D.C., 1999. A review of soluble microbial wastewater and activated sludge. Water Res. 41 (1), 235e243.
products (SMP) in wastewater treatment. Water Res. 33 (14), Furumai, H., Rittmann, B.E., 1992. Advanced modelling of mixed
3062e3082. population of heterotrophs and nitrifiers. Water Sci. Technol.
Benedetti, L., Batstone, D.J., Debaets, B., Nopens, I., 26 (3e4), 493e502.
Vanrolleghem, P., 2008. Global Sensitivity Analysis of the Germain, E., Nelles, F., Drews, A., Pearce, P., Kraume, M., Reid, E.,
Benchmark Simulation Model No. 2. IEMSs Fourth Biennial Judd, S.J., Stephenson, T., 2007. Biomass effects on oxygen
Meeting: International Congress on Environmental Modelling transfer in membrane bioreactors. Water Res. 41 (5),
and Software. Spain, Barcelona. 7e10 July. 1038e1044.
Bixio, D., De Wilde, W., Lesjean, B., 2006. Model-based Evaluation Germirli, F., Orhon, D., Artan, N., 1991. Assessment of the initial
of Flow Repartition Control Strategies for the CAS/MBR Dual 1 inert soluble COD in industrial wastewater. Water Sci.
Concept, Including Determination of the Sludge Technol. 23, 1077e1086.
Characteristics Amedeus Project, D52 (task 91). Guglielmi G., Avesani D., Brepols C., Foxon K., Brouckaert C.,
Chaize, S., Huyard, A., 1991. Membrane bioreactors on domestic Buckley C., 2009. Sludge production in aerobic and anaerobic
wastewater treatment sludge production and modeling membrane bioreactors, Proc. of WISA Membrane technology
approach. Water Sci. Technol. 23, 1591e1600. Conference. 13e15 May, Stellenbosch, South Africa.
De la Torre T., Iversen V., Meng F., Stüber J., Drews A., Lesjean B., Gujer, W., Henze, M., Mino, T., Loosdrecht, M., 1999. Activated
Kraume M, 2009. Searching for a universal fouling indicator for sludge model No.3. Water Sci. Technol. 39 (1), 183e193.
MBR. 5th IWA Specialised Membrane Technology Conference Henze, M., Grady, C.P.L., Gujer, W., Marais, G.v.R., Matsuo, T.,
for Water & Wastewater Treatment, 1e3 September 2009. 1987. Activated Sludge Model No.1 IAWPRC Scientific and
Beijing, China. Technical Report No.1, London (GB).
Delrue F., Racault Y., Choubert J.M., Spèrandio M., (2008), Henze, M., 1992. Characterization of wastewater for modelling of
Modelling a full scale Membrane Bioreactor using Activated activated sludge systems. Water Sci. Technol. 18 (6), 91e114.
Sludge Model n 1: challenges and solutions. IWA Regional Henze, M., Gujer, W., Mino, T., van Loosdrecht, M., 2000. Activated
Conference “Membrane Technologies in Water and Waste Sludge Models ASM1, ASM2, ASM2d and ASM3 IWA Scientific
Water Treatment” (2e4 June 2008 Moscow, Russia). and Technical Report No.9, London (GB).
Di Bella, G., Mannina, G., Viviani, G., 2008. An integrated model for Itokawa, H., Thiemig, C., Pinnekamp, J., 2008. Design and
physical-biological wastewater organic removal in a sMBR: operating experiences of municipal MBRs in Europe. Water
model development and parameter estimation. J. Membr. Sci. Sci. Technol. 58 (12), 2319e2327.
322, 1e12. Jang, N., Ren, X., Cho, J., Kim, I.S., 2006. Steady-state modeling of
Dold, P.L., Ekama, G.A., Marais, G.v.R., 1980. A general model for bio-fouling potentials with respect to the biological kinetics in
the activated sludge process. Prog. Water Technol. 12, 47e77. the sMBR. J. Memb. Sci. 284, 352e360.
Drews, A., Evenblij, H., Rosenberger, S., 2005. Potential and Jarusutthirak, C., Amy, G., 2007. Understanding soluble microbial
drawbacks of microbiology-membrane interaction in products (SMP) as a component of effluent organic matter
membrane bioreactors. Environ. Prog. 24 (4), 426e433. (EfOM). Water Res. 41 (12), 2787e2793.
Drews, A., Vocks, M., Iversen, V., Lesjean, B., Kraume, M., 2006. Jiang, T., Liu, X., Kennedy, M.D., Schippers, J.C., Vanrolleghem, P.
Influence of unsteady membrane bioreactor operation on EPS A., 2005. Calibrating a side-stream membrane bioreactor using
formation and filtration resistance. Desalination 192 (1e3), Activated Sludge Model No. 1. Water Sci. Technol. 52 (10e11),
1e9. 359e367.
Drews, A., Mante, J., Iversen, V., Vocks, M., Lesjean, B., Kraume, M. Jiang T., 2007. Characterization and modelling of soluble
, 2007. Impact of ambient conditions on SMP elimination and microbial products in membrane bioreactors. PhD thesis,
rejection in MBR. Water Res. 41 (17), 3850e3858. Ghent University, Belgium pp. 241.
Erftverband, 2001. Weitergehende Optimierung einer Jiang, T., Myngheer, S., De Pauw, D.J.W., Spanjers, H., Nopens, I.,
Belebungsanlage mit Membranfiltration e Zwischenbericht. Kennedy, M.D., Amy, G., Vanrolleghem, P.A., 2008. Modelling
(Advanced optimisation of an activated sludge plant with the production and degradation of soluble microbial products
membrane filtration e mid-term-report), Report to the (SMP) in membrane bioreactors (MBR). Water Res. 42 (20),
Ministry of Environment North e Rhine Westpahlia, Germany, 4955e4964.
Erftverband Bergheim, pp. 73e93. Jiang, T., Sin, G., Spanjers, H., Nopens, I., Kennedy, M., van der
Erftverband, 2004. Optimisation of an Activated Sludge Plant with Meer, W., Futselaar, H., Amy, G., Vanrolleghem, P., 2009.
Membrane Filtration e, Report to the Ministry of Environment Comparison of the modeling approach between membrane
North e Rhine Westpahlia, Germany, Erftverband Bergheim, bioreactor and conventional activated sludge processes.
pp. 2e16. Optimierung einer Belebungsanlage mit Water Environ. Res. 81 (4), 432e440.
Membranfiltration e Band 3, vol. 3. Jimenez J., Grelier P., Meinhold J., Tazi-Pain A., 2008. Biological
Ekama, G.A., Wentzel, M.C., 2004. A predictive model for the modelling of MBR and impact of primary sedimentation.
reactor inorganic suspended solids concentration in activated MIDW-EDS conference, Toulouse, 20e22 October 2008.
sludge systems. Water Res. 38 (19), 4093e4106. Judd, S., 2006. The MBR Book: Principles and Applications of
Ekama, G.A., Marais, G.R., 1986. Procedures for determining Membrane Bioreactors for Water and Wastewater Treatment.
influent COD fractions and the maximum specific growth rate EN, July 2006. Elsevier, ISBN 978-1-85617-481-7.
of heterotrophs in activated sludge systems. Water Sci. Kappeler, J., Gujer, W., 1992. Estimation of kinetic parameters of
Technol. 18 (6), 63e89. heterotrophic biomass under aerobic conditions and
 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4 4293

characterization of wastewater for activated sludge Massé A., 2004. Bioréacteurs à membranes immergées pour le
modelling. Water Sci. Technol. 25 (6), 125e139. traitement des eaux résiduaires urbaines: spécificités physico-
Krishna, C., Van Loosdrecht, M.C.M., 1999. Substrate flux into chimiques du milieu biologique et colmatage. Ph.D., INSA
storage and growth in relation to activated sludge modelling. Toulouse, no. 759, 15, décembre 2004.
Water Res. 33 (14), 3149e3161. Massé, A., Spérandio, M., Cabassud, C., 2006. Comparison of
Krause, S., 2005. Untersuchungen zum Energiebedarf von sludge characteristics and performance of a submerged
Membranbelebungsanlagen. Schriftenreihe WAR 166, membrane bioreactor and an activated sludge process at high
Darmstadt, Techn. Universität, Diss., 2005. In: Research into solids retention time. Water Res. 40 (12), 2405e2415.
the Energy Consumption of Membrane Bioreactor Plants, ISBN Metcalf, Eddy, 2003. Wastewater Engineering: Treatment and
3-932518-62-4. Reuse, fourth ed.. Mc Graw Hill.
Krause, S., Cornel, P., Wagner, M., 2003. Comparison of different Munz, G., De Angelis, D., Goria, R., Moric, G., Casarci, M.,
oxygen transfer testing procedures in full scale membrane Lubello, C., 2008. Process efficiency and microbial monitoring
bioreactors. Water Sci. Technol. 47 (12), 169e176. in MBR and conventional activated sludge process treatment
Krause S., Cornel P., 2007. Membrane Bioreactor’s Energy Demand of tannery wastewater. Bioresour. Technol. 99 (18), 8559e8564.
in Wastewater Treatment. 4th IWA International Membrane Namkung, E., Rittmann, B.E., 1986. Soluble microbial products
Conference, 15the17th May 2007, Harrogate, UK (SMP) formation kinetics by biofilms. Water Res. 20 (6),
Laera, G., Pollice, A., Saturno, D., Giordano, C., Lopez, A., 2005. 795e806.
Zero net growth in a membrane bioreactor with complete Ng, H.Y., Hermanowicz, S.W., 2005. Specific resistance to filtration of
sludge retention. Water Res. 39 (20) 5241e4249. biomass from membrane bioreactor reactor and activated sludge:
Laspidou, C.S., Rittmann, B.E., 2002a. Non-steady state modeling effects of exo-cellular polymeric substances and dispersed
of EPS, SMPs, and active and inert biomass. Water Res. 36 (8), micro-organisms. Water Environ. Res. 77 (2), 187e192.
1983e1992. Ng, H.Y., Tan, T.W., Ong, S.L., 2006. Membrane fouling of
Laspidou, C.S., Rittmann, B.E., 2002b. A unified theory for EPS, submerged membrane bioreactors: impact of the mean cell
SMPs, and active and inert biomass. Water Res. 36 (11), residence time and the contributing factors. Environ. Sci.
2711e2720. Tecnol. 40, 2706e2713.
Le-Clech, P., Chen, V., Fane, T.A.G., 2006. Fouling in membrane Ng, N.L.A., Kim, A.S., 2007. A mini-review of modelling studies on
bioreactors used in wastewater treatment. J. Membr. Sci. 284 MBR treatment for municipal wastewaters. Desalination 212
(1e2), 17e53. (1e3), 261e281.
Lee, Y., Cho, J., Seo, Y., Lee, J.W., Ahn K.-, H., 2002. Modelling of Noguera, D.R., Araki, N., Rittmann, B.E., 1994. Soluble microbial
submerged membrane bioreactor process for wastewater products (SMP) in anaerobic chemostats. Biotechnol. Bioeng.
treatment. Desalination 146, 451e457. 44, 1040e1047.
Lesouef, A., Payraudeau, M., Rogalla, F., Kleiber, B., 1992. Oliveira-Esquerre, K.P., Narita, H., Yamato, N., Funamizu, N.,
Optimizing nitrogen removal reactor configurations by onsite Watanabe, Y., 2005. Modelling of a Conventional MBR for
calibration of the IAWPRC Activated sludge model. Water Sci. Wastewater Treatment. 21st Center of Excellence Program
Technol. 25 (6), 105e123. Publication. Department of Urban and Environmental
Liang, S., Liu, C., Song, L., 2007. Soluble microbial products in Engineering, Hokkaido University, Japan.
membrane bioreactor operation: behaviors, characteristics, Oliveira-Esquerre, K.P., Narita, H., Yamato, N., Funamizu, N.,
and fouling potential. Water Res. 41 (1), 95e101. Watanabe, Y., 2006. Incorporation of the concept of microbial
Liu, H., Fang, H.H.P., 2002. Extraction of extracellular polymeric products formation into ASM3 and the modelling of
substances (EPS) of sludges. J. Biotechnol. 95, 249e256. a membrane bioreactor for a wastewater treatment. Braz. J.
Lu, S.G., Imai, T., Ukita, M., Sekine, M., Higuchi, T., Fukagawa, M., Chem. Eng. 23 (4), 461e471.
2001. A model for membrane bioreactor process based on the Orhon, D., Artan, N., Cimcit, Y., 1989. The concept of soluble
concept of formation and degradation of soluble microbial residual product formation in the modelling of activated
products. Water Res. 35 (8), 2038e2048. sludge. Water Sci. Technol. 21, 339e350.
Lu, S.G., Imai, T., Ukita, M., Sekine, M., Higuchi, T., 2002. Modeling Orhon, D., Karahan, Ö, Sözen, S., 1999. The effect of residual
prediction of membrane bioreactor process with the concept of microbial products on the assessment of the particulate inert
soluble microbial product. Water Sci. Technol. 46 (11e12), 63e70. COD in wastewaters. Water Res. 33, 3191e3203.
Luedeking, R., Piret, E.C., 1959. A kinetic study of lactic acid Parco, V., Wentzel, M., Ekama, G., 2006. Kinetics of nitrogen
fermentation batch process at controlled pH. J. Biochem. removal in a MBR nutrient removal activated sludge system.
Microbiol. Technol. Eng. 1 (4), 393e412. Desalination 199, 89e91.
Luong, J.H., Muchaldani, A., 1988. Kinetic of bio polymers Parco, V., du Toit, G., Wentzel, M., Ekama, G., 2007. Biological nutrient
synthesis: a revisit. Enzym. Microb. Technol. 10, 326e329. removal in membrane bioreactors: denitrification and
McKinney, R.E., 1962. Mathematics of complete mixing activated phosphorus removal kinetics. Water Sci. Technol. 56 (6), 125e134.
sludge. J. Sanitary Eng. ASCE 88 (3), 87e113. Petersen, B., Gernaey, K., Henze, M., Vanrolleghem, P., 2002.
Majone, M., Dircks, K., Beun, J.J., 1999. Aerobic storage under Evaluation of an ASM1 procedure on a municipal-industrial
dynamic conditions in activated sludge processes. The state of wastewater treatment plant. J. Hydroinf. 04 (1), 38.
the art. Water Sci. Technol. 39 (1), 61e73. Pinnekamp, J., 2006. Municipal Water and Waste Management:
Malamis, S., Andreadakis, A., 2009. Fractionation of proteins and Membrane Technology for Waste Water Treatment. In:
carbohydrates of EPS in a MBR. Bioresour. Technol. 100 (13), Friedrich, H. (Ed.). FIW Verlag, Aachen, ISBN 3-939377-01-5.
3350e3357. Pollice, A., Laera, G., Saturno, D., Giordano, C., 2008. Effects of
Manser, R., Gujer, W., Siegrist, H., 2005. Consequences of mass sludge retention time on the performance of a membrane
transfer effects on the kinetics of nitrifiers. Water Res. 39 (19), bioreactor treating municipal sewage. J. Membr. Sci. 317,
4633e4642. 65e70.
Manser, R., Gujer, W., Siegrist, H., 2006. Decay processes of Ramesh, A., Duu-Jong, L., Hong, S.G., 2006. Soluble microbial
nitrifying bacteria in biological wastewater treatment products (SMP) and soluble extracellular polymeric substances
systems. Water Res. 40 (12), 2416e2426. (EPS) from wastewater sludge. Environ. Biotechnol. 73, 219e225.
Marais, G.V.R., Ekama, G.A., 1976. The activated sludge process. Ramphao, M.C., Wentzel, M.C., Ekama, G.A., Alexander, W.V.,
Part 1 e steady-state behaviour. Water SA 2, 163e199. 2006. A comparison of BNR activated sludge systems with
4294 w a t e r r e s e a r c h 4 4 ( 2 0 1 0 ) 4 2 7 2 e4 2 9 4

membrane and settling tank solid-liquid separation. Water Sperandio, M., Espinosa, M.C., 2008. Modelling an aerobic
Sci. Technol. 53 (12), 295e303. submerged membrane bioreactor with ASM models on a large
Rieger, L., Koch, G., Kuehni, M., Gujer, W., Siegrist, H., 2001. The range of sludge retention time. Desalination 231 (1e3), 82e90.
EAWAG bio-P module for activated sludge model no.3. Water Spérandio, M., Paul, E., 2000. Estimation of wastewater biodegradable
Res. 35 (16), 3887e3903. COD fractions by combining respirometric experiments in
Robinson, J.A., Trulear, M.G., Characklis, W.G., 1984. Cellular various S0/X0 ratios. Water Res. 34 (4), 1233e1246.
reproduction and extracellular polymer formation by STOWA, 1996. Methods for Wastewater Characterization,
Pseudomonas aeruginosa in continuous culture. Biotechnol. Inventory and Guidelines. STOWA-report 97-23, Hageman
Bioeng. 26, 1409e1417. Fulfilment, Zwijndrecht, The Netherlands. (in Dutch).
Roeleveld, P.J., van Loosdrecht, M.C.M., 2002. Experience with STOWA, 1999. Wastewater Characterization of Raw and Pre-
guidelines for wastewater characterization in The treated Wastewater, the Influence of Primary Sedimentation
Netherlands. Water Sci. Technol. 45 (6), 77e87. and Pre-Precipitation. STOWA-report 99-13, Hageman
Rosenberger, S., Witzig, R., Manz, W., Szewzyk, U., Kraume, M., Fulfilment, Zwijndrecht, The Netherlands (in Dutch).
2000. Operation of different membrane bioreactors: Tan, T.W., Ng Yong, H., Leong Ong, S., 2008. Effect of mean cell
experimental results and physiological state of the micro- residence time on the performance and microbial activity of
organisms. Water Sci. Technol. 41 (10e11), 269e277. pre-denitrification submerged membrane bioreactors.
Rosenberger, S., Kraume, M., 2002. Filterability of activated sludge Chemosphere 70, 387e396.
in membrane bioreactors. Desalination 146, 373e379. Van Haandel, A.C., Van Der Lubbev, J.G.M., 2007. Handbook
Rosenberger S., (2003), Charakterisierung von belebtem Schlamm Biological Waste Water Treatment e Design and Optimization
in Membranbelebungsanlagen. Dissertation, of Activated Sludge Systems. Quist Publishing, Leidschendam,
Forschrittsberichte VDI-Verlag, Reihe 3, Nr. 769 ISBN 3-18- The Netherlands, ISBN 978-90-77983-22-5. EN.
376903-4, 160 S. Vanrolleghem, P.A., Spanjers, H., Petersen, B., Ginestet, P.,
Rosenberger, S., Evenblij, H., Poele te, S., Wintgens, T., Laabs, C., Takacs, I., 1999. Estimating (combinations of) activated sludge
2005. The importance of liquid phase analyses to understand model n 1 parameters and components by respirometry.
fouling in membrane assisted activated sludge processesdsix Water Sci. Technol. 39 (1), 195e214.
case studies of different European research groups. J. Membr. Vanrolleghem, P.A., Insel, G., Petersen, B., Sin, G., De Pauw, D.,
Sci. 263 (1e2), 113e126. Nopens, I., Dovermann, H., Weijers, S., Gernaey, K., 2003. A
Rosenberger, S., Laabs, C., Lesjean, B., Gnirss, R., Amy, G., Jekel, M. comprehensive model calibration procedure for activated
, Schrotter, J.-C., 2006. Impact of colloidal and soluble organic sludge models. Proc.. of WEFTEC 03, October 11-15 2003, Los
material on membrane performance in membrane bioreactors Angeles, CA, USA.
for municipal wastewater treatment. Water Res. 40 (4), Van Loosdrecht, M.C.M., Heijnen, J.J., 2002. Modelling of activated
710e720. sludge processes with structured biomass. Water Sci. Technol.
Sarioglu, M., Insel, G., Artan, N., Orhon, D., 2008. Modelling of 45 (6), 13e23.
long-term simultaneous nitrification and denitrification Wagner, J., Rosenwinkel, K.H., 2000. Sludge production in
(SNDN) performance of a pilot scale membrane bioreactor. membrane bioreactors under different conditions. Water Sci.
Water Sci. Technol. 57 (11), 1825e1833. Technol. 41 (10e11), 251e258.
Sarioglu, M., Insel, G., Artan, N., Orhon, D., 2009. Model evaluation Wintgens, T., Rosen, J., Melin, T., Brepols, C., Drensla, K.,
of simultaneous nitrification and denitrification in Engelhardt, N., 2003. Modelling of a membrane bioreactor
a membrane bioreactor operated without an anoxic reactor. J. system for municipal wastewater treatment. J. Membr. Sci.
Membr. Sci. 337, 17e27. 216 (1e2), 55e65.
Saroj, D.P., Guglielmi, G., Chiarani, D., Andreottola, G., 2008. Wisniewski C., (1996), Etude du comportement de cultures mixtes
Modeling and simulation of membrane bioreactors by en bioreacteur à membrane. Cinétiques réactionnelles et
incorporating simultaneous storage and growth concept: filtrabilité. PhD-Thesis. Univeristy Montpellier II.
a special attention to fouling while modelling the biological Witzig, R., Manz, W., Rosenberger, S., Krüger, U., Kraume, M.,
process. Desalination 221, 475e482. Szewzyk, U., 2002. Microbiological aspects of a bioreactor with
Shin, H.S., Kang, S.T., 2003. Characteristics and fates of soluble submerged membranes for aerobic treatment of municipal
microbial products in ceramic membrane bioreactor at wastewater. Water Res. 36 (2), 394e402.
various sludge retention times. Water Res. 37 (1), 121e127. Wu, Z., Wang, Z., Zhou, Z., Yu, G., Gu, G., 2007. Sludge rheological
Silva, D.G.V., de, Urbain, V., Abeysinghe, D.H., Rittmann, B.E., and physiological characteristics in a pilot-scale submerged
1998. Advanced analysis of membrane-bioreactor membrane bioreactor. Desalination 212, (1e3), 152e164.
performance with aerobic-anoxic cycling. Water Sci. Technol. Zarragoitia-González, A., Schetrite, S., Alliet, M., Jáuregui-Haza, U.
38 (4e5), 505e512. , Albasi, C., 2008. Modeling of SMBRs: conceptual study about
Silva A.F., Carvalho G., Lousada Ferreira M., Nieuwenhuijzen van link between activated sludge biokinetics, aeration and
A., Guglielmi G., Crespo J.G., Reis M.A.M., Crespo M.T.B., 2009. fouling process. J. Membr. Sci. 325, 612e625.
Microbial population structure of pilot and full scale Zhang, J., Chua, H.C., Zhou, J., Fane, A.G., 2006. Factors affecting
membrane bioreactors. Book of Proceedings of final MBR the membrane performance in submerged membrane
Network workshop, 31 Marche1 April, Berlin (Germany). bioreactors. J. Membr. Sci. 284, (1e2), 54e66.
Sin, G., Guisasola, A., De Pauw, D.J.W., Baeza, J.A., Carrera, J., Zhang, H.F., Sun, B.S., Zhao, X.H., Qi, G.S., 2007. Generation of
Vanrolleghem, P.A., 2005. A new approach for modelling soluble microbial products in membrane bioreactor. China
simultaneous storage and growth processes for activated Environ. Sci. 27 (4), 539e542.
sludge systems under aerobic conditions. Biotechnol. Bioeng. Ziglio, G., Andreottola, G., Foladori, P., Ragazzi, M., 2001.
92 (5), 600e613. Experimental validation of a single-OUR method for wastewater
Spèrandio, M., Massé, A., Espinosa, M.C., Cabassud, C., 2005. RBCOD characterisation. Water Sci. Technol. 43 (11), 119e126.
Characterization of sludge structure and activity in Ziglio, G., Andreottola, G., Barbesti, S., Boschetti, G., Bruni, L.,
submerged membrane bioreactor. Water Sci. Technol. 52 Foladori, P., Villa, R., 2002. Assessment of activated sludge
(10e11), 401e408. viability with flow cytometry. Water Res. 36 (2), 460e468.