See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/300411262

A simple model to describe the performance of

highly-loaded aerobic COD removal reactors
Article April 2016
DOI: 10.1016/j.bej.2016.04.004

READS

288

4 authors, including:
Olaf Duin

Mark Van Loosdrecht

WS Hollandse Delta

Delft University of Technology

5 PUBLICATIONS 38 CITATIONS

841 PUBLICATIONS 36,756 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate,

letting you access and read them immediately.

SEE PROFILE

Available from: Mark Van Loosdrecht

Retrieved on: 12 July 2016

Biochemical Engineering Journal 112 (2016) 94102

Contents lists available at ScienceDirect

Biochemical Engineering Journal

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

Regular article

A simple model to describe the performance of highly-loaded aerobic

COD removal reactors
Jeremy Smitshuijzen a , Julio Prez a, , Olaf Duin b , Mark C.M.van Loosdrecht a
a
b

Department of Biotechnology, Faculty of Applied Sciences, Delft University of Technology, Julianalaan 67, 2628 BC Delft, The Netherlands
Waterschap Hollandse Delta (WHD), Handelsweg 100, 2988 DC Ridderkerk, The Netherlands

a r t i c l e

i n f o

Article history:
Received 6 October 2015
Received in revised form 21 March 2016
Accepted 5 April 2016
Available online 8 April 2016
Keywords:
Adsorption
Biosorption
Dynamic simulation
AB process
A-stage
Modelling

a b s t r a c t
Highly-loaded aerobic chemical oxygen demand (COD) removal reactors (also known as A-stage) include
two main processes, COD removal by heterotrophic biomass and adsorption of COD on the activated
sludge. A simple model to describe highly-loaded aerobic COD removal reactors has been developed. A
one-year-set of measured data from a full scale wastewater treatment plant has been used to calibrate
the efciency of the adsorption and to evaluate the ability of the mathematical model to describe the
measured data in both steady state and dynamic conditions. The approach lumped the efciency of the
settler and the adsorption with a simple but powerful approach which includes the use of the measured
sludge retention time (SRT) and the settling efciency. The effects of dynamics in terms of (i) seasonality
(for water temperature, ow rate and concentration of pollutants), and (ii) daily variations in ow rate
were investigated. Results showed how during winter the low water temperatures negatively impacted
the efciency of the A-stage, producing a higher COD concentration in the efuent, which eventually could
impact the performance of the nitrogen removal in the B-stage. General guidelines for the application of
the model to similar highly loaded A-stage reactor systems were provided.
2016 Elsevier B.V. All rights reserved.

1. Introduction
Biological removal of organic material representing a given
chemical oxygen demand (COD) by activated sludge is associated
to the growth of heterotrophic bacteria. Heterotrophic bacteria
take-up the (biodegradable) COD, a fraction is mineralized to
CO2 and water, another fraction can be stored as intracellular
biopolymers for later use and the rest is used for biomass synthesis, increasing the amount of bacteria. In the 1970s a simple
conguration emerged in Europe for the treatment of wastewaters with a high fraction of industrial efuents: the so-called
Adsorption-Belebungsverfahren (AB process), a two-stage wastewater treatment plant (WWTP) [13]. This process features a very
high food-to-microorganism (F/M) ratio in the rst stage (Adsorption or A-stage) at a short hydraulic residence time (HRT, about
30 min), and a short sludge residence time (SRT, 312 h). The second
stage (Belebungs or in English Biooxidation, B-stage) is an aeration
stage with a low F/M ratio. The F/M ratio in the adsorption stage

Corresponding author.
E-mail addresses: j.o.perezcanestro@tudelft.nl, julio.perez@uab.es (J. Prez).
http://dx.doi.org/10.1016/j.bej.2016.04.004
1369-703X/ 2016 Elsevier B.V. All rights reserved.

typically ranges from 2 to 10 g BOD/g VSS/day (where BOD stands

for biological oxygen demand and VSS for volatile suspended solids)
[4]. In the second stage, biological oxidation takes place at a much
lower F/M ratio of less than 0.1 g BOD/g VSS/day [4]. The adsorption
stage is operated at a low oxygen concentration (close to zero) [4].
The A-stage was considered to be of key importance against the
effect of toxic shock loads for the nitrifying bacteria in the B-stage
activated sludge process, while simultaneously being able to collect a high fraction of COD as sludge and use it for biomethanation
[4].
The treatment conguration was developed and introduced in
the 70-ies, but got in disuse in the past decades due to efforts to
reduce total nitrogen in the treatment plants. Nowadays, there is
still a fraction of the WWTP in operation in Europe that retained
the AB process conguration. The A-stage achieves a high removal
efciency, typically removing 7080% of total COD and 8095% of
suspended solids [5]. The production of biogas out of the excess
sludge is enhanced because of the short SRT imposed; thus the
digestion characteristics of the sludge improve considerably [6].
Due to this highly efcient energy production to energy consumption ratio, the A-stage has been identied as a convenient option for
COD removal to be combined with the use of nitritation/anammox

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

(anammox stands for anaerobic ammonia oxidation), for nitrogen

removal, with the aim to achieve energy neutral or even energy
producing WWTP [79].
The high performance obtained with the A-stage is due to a
simultaneous conventional removal of COD by heterotrophic bacteria together with the adsorption of an important fraction of the
COD, including particulate solids and colloidal particles. This COD
adsorbed onto the sludge can be removed in the settler without
having been metabolized by bacteria, allowing for short HRT and
SRT. This adsorption, also named biosorption, is a process taking
place in a short period of time, in the order of minutes [10,11].
The efciency of adsorption is known to be linked to complex process that drive the sludge characteristics, like type and fraction
of extracellular polymeric substances (EPS), presence of storage
compounds and even dissolved oxygen (DO) levels in the reactor
[1012]. In fact, the DO concentration was identied as a key operating condition, since it negatively affected the adsorption due to
poor aeration intensity [10,12].
Because the A-stage has been considered as an old process,
researchers paid little attention to the modelling of this process.
Haider et al. [13] presented general guidelines for the selection
of kinetic parameters for heterotrophs in AB processes (mainly a
higher maximum specic growth rate and lower half-saturation
coefcient for COD compared to those in conventional activated
sludge models). Moreover they proposed a particular fractionation of the soluble COD, to allow for a none complete conversion
of soluble COD in the A-stage due to the use of a higher specic growth. Their modelling approach focused on describing the
removal of soluble COD but not on how to model the fraction
of COD adsorbed in the A-stage. A detailed model was presented
by Nogaj et al. [14]. In that model, getting insight in the process
was the main driving force during model development. The strategy yielded a rather complex mathematical model, in which new
state variables were added to the generally accepted activated
sludge model No. 1 (ASM1 [15]) stoichiometry matrix, including
concentrations of colloidal biodegradable organics, colloidal nonbiodegradable organics from the inuent, extracellular polymeric
substances (EPS) and particulate storage compounds. The model
would require a set of measurements that are usually not available in the regular operation of WWTP. To date, there is no simple
model available to describe the main processes in a highly loaded
aerobic COD removal reactor. Such a model would aid in the current
research and development effort devoted to improve the efciency
of A-stage systems [12,1620]. The core of the energy production
and therefore the success of an energy neutral WWTP lies on the
efciency of the A-stage [9]. Even more, the performance of the subsequent nitritation/anammox stage might depend on the capacity
of A-stage to remove COD and solids efciently [21,22].
In the WWTP of Dokhaven (Rotterdam, The Netherlands) an AB
process is operational since 1987 [23]. The use of anammox for
nitrogen removal in the mainstream is under consideration [24].
Therefore, a model able to describe the performance of the current
A-stage can be used as a tool for the simulation of the future scenario in which the A-stage is followed by nitritation/anammox for
nitrogen removal. In addition, this model may help practitioners to
better understand and apply this technology.
Here, we developed a simple mathematical model, based on the
use of the habitual measurements in a WWTP and ASM1. The goal
was the mathematical description of the COD removal in a highlyloaded aerobic COD removal reactor with a minimal complexity.
The model predictions at steady state and under dynamic conditions are compared to a measured data set covering a period of one
year. The model provides a simple description of the performance
and presents general guidelines that could be followed to model
other similar highly loaded A-stage reactor systems.

95

Fig. 1. Different state variables used in the model of the A-stage for organic compounds and their conversions. The dashed line separates the soluble and particulate
COD in the gure. Since slowly biodegradable organics (XS ) are potentially partly
soluble, the line crosses there. Particulates included: Heterotrophic bacteria (XHET ),
inert non-biodegradable organics (XI ), and a fraction of XS . Soluble organic compounds included: biodegradable organics (SS ), a fraction of XS , and soluble inert
non-biodegradable organics (SI ).

2. Model description
2.1. State variables, and general modelling approach
The model is based on ASM1 [15] but nitrite concentration
was considered as an additional state variable as it will be further
detailed in the Section 2.4. Kinetics and stoichiometry. The different
state variables used to describe the different organic compounds
concentrations were the following: biodegradable organics (SS ),
soluble inert non-biodegradable organics (SI ), slowly biodegradable organics (XS ) and particulate inert non-biodegradable organics
(XI ). Heterotrophic bacteria (XHET ) were considered the main bacterial population. A more extensive distribution of the biodegradable
organics, as well as different variables for storage polymers (like
in ASM2 [15]; or like used in Nogaj et al. [14]) were considered
unnecessary for the purpose of this model. A schematic overview
of the model showing the main processes under consideration can
be found in Fig. 1. The SS was used as substrate for heterotrophic
growth. Since the slowly biodegradable organics (XS ) can be soluble
or particulate [15], a 40% soluble and 60% particulate was assumed
(see Section 2.5 for details). A large fraction of the XS (independently
if they are soluble or particulate) and XI would be adsorbed in the Astage, for details regarding the modelling approach to adsorption
see below Section 2.2). Furthermore XS can be hydrolysed to SS
[15]. The decay of XHET produces XS and XI . Colloidal COD has not
been included as a separate variable, to avoid unnecessary complexity. Also storage COD and a difference between faster or slower
biodegradable soluble COD were not used in this model. It was
assumed that a xed percentage of XS is adsorbed in the A-stage
and later separated in the settler. Therefore the exact composition
of the particulate COD was less important.
For simplicity it was assumed that there would be no biological activity or buffering capacity in the settler. This is in line with
operation of the settlers in Dokhaven WWTP where only very

96

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

Fig. 2. Overall process conguration of the AB process. The grey rectangle indicates the process units and streams described by the mathematical model. Notation used for
the equations when implementing the model is shown (Q stands for ow-rate and X for biomass concentration).

limited amounts of sludge are present in the settlers. We have not

included nitrication in the model. Nitrication in the A-stage is
a minor process, given the high COD/N ratio in the reactor. However, nitrication could be easily added to the model description.
Chemical phosphate removal was not described with the model,
but could be added with conventional approaches.
2.2. A-stage conguration
A typical A-stage conguration has one or several (in parallel)
aerated sludge tanks with a plug ow conguration (Fig. 2). After
the aerated tanks a settler is used for sludge separation and efuent
extraction. In the Dokhaven plant the settlers are rectangular with
scrapers and an overow at the end. From those settlers the sludge
will be recirculated to the A-stage and a part will be removed as
excess sludge. This excess sludge will be pumped to the anaerobic
digester to produce biogas (Fig. 2). The inuent of the A-stage, does
not only consists of sewage water, but also of recirculation water
from the plant efuent (for obtaining some denitrication, [25])
and the stream coming from the anaerobic digester (recirculation
sidestream).

To be able to impose in the simulations a given sludge residence

time (SRT), the excess sludge ow (Qw ) was chosen as manipulated
variable. The set of variables related to the calculation of the Qw to
be applied were presented in Fig. 2.
From the process diagram in Fig. 2, the following equation was
deduced:
SRT =

XA VA
QW Xrw + (QA QW Qr ) Xout

where XA is the biomass concentration in the reactor (i.e. the sum of

Xhet + Xs + Xi ), VA is the volume of the reactor, Xrw and Xout are the
concentrations of particulate COD in each one of the streams leaving
the settler, and QA and Qr are the ow rates after the reactor and that
recirculated, respectively. The concentrations of particulate COD
could be expressed as:
X rw =

QA XA fsettler
,
Qr + Qw

Instead of having a kinetic process to describe adsorption, and

because adsorption is known to happen in the time scale of minutes [10,11], a simple approach was chosen to represent the effects
of adsorption in the mathematical model. A xed percentage of XS
coming out of the A-stage is removed with a given fraction (fsettler )
from the ow towards the B-stage. The fraction fsettler has been
assumed constant for the whole year. This fraction accounts for (i)
the efciency of the adsorption, and (ii) the efciency of the settler, and can be estimated easily from the plant data. It is in fact
a simple and powerful approach, although it has the disadvantage
to lump in one parameter two different processes (adsorption and
settling). The advantage is that it allows to calculate the effects of
adsorption without adding a set of state variables to describe more
in detail -and with a more mechanistic approach- the adsorption
process. This simplicity was crucial because the driving force for the
development of the model was the later use in simulations in which
the A-stage was coupled to a CANON reactor under fully dynamic
conditions, including daily and seasonal variations of ow rate and
changes in temperature, to test the feasibility of the implementation of CANON for mainstream treatment [26]. A more complex
approach could result in a mechanistically more correct description
at the expense of having to estimate more unknown parameters.

X out =

QA XA 1 fsettler
Q A Qw Qr

(2)

where the coefcient fsettler was dened as the fraction of XS

removed in the settler (due to both adsorption to the sludge ocs
and settling efciency), hence
fsettler = 1

2.3. SRT and description of adsorption

(1)

Xout (Qin,A Qw )
QA XA

(3)

Eq. (4) can be obtained by substitution of Eq. (2) in Eq. (1), yielding the excess sludge ow (Qw) as a function of the SRT (SRT), the
fraction of solids removed (fsettler ), the volume of the A-stage (VA ),
the ow-rate into the A-stage (Qin ) and the recirculation ow (Qr ).

Q w = Q r/

f settler
(f settler 1) +

VA
SRT (Q r +Q in,A )

(4)

No specic features of the model are unique for the description

of highly loaded COD removal reactors. However, the parameter
fsettler is mainly used to describe the adsorption process, although
it includes also the efciency of the settling.
2.4. Kinetics and stoichiometry
Heterotrophic growth description is based on ASM1 [15], but
three different processes were used: aerobic growth, and two possible anoxic growth reactions using either nitrite or nitrate as
electron acceptors [27]. For denitrication a sequential mechanism
was used, in which nitrate is rst converted to nitrite and then
to nitrogen gas. Hydrolysis of the slowly biodegradable organics
(XS ) and decay of heterotrophic bacteria was modelled based on
the ASM1 approach [15]. All kinetic and stoichiometric parameters

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

97

Fig. 3. Schematics of the modelling procedure, indicating what are the inputs required by the model (with solid grey arrows pointing the model A-stage box) and the (main)
model outputs, shown by the arrow with (green) pattern. ASM1 = activated sludge model No. 1, *modied by including nitrite as state variable. fsettler = parameter lumping
adsorption and efciency of the settling, # calculated through Eq. (3) from measured data of the wastewater treatment plant (WWTP). (For interpretation of the references to
colour in this gure legend, the reader is referred to the web version of this article).

were taken from the literature. The microbial kinetics and stoichiometry used in the model are given in Tables A1A3 (Appendix
A). The growth yield for heterotrophic bacteria (YHET ) was set to
0.5 gCOD/gCOD [6]. This value is rather low when compared to the
standard values assumed in ASM1 (0.67 gCOD/gCOD [15]). However, in the highly loaded A-stage a yield of 0.5 gCOD/gCOD could
be used since the extent of storage is expected to be limited since
a famine period does not exist. In contrast to ASM3, storage is not
modelled explicitly in ASM1 [15], leading to a somewhat high yield
coefcient when compared to normal heterotrophic growth.
2.5. Dokhaven: data collection
The WWTP in Dokhaven (Rotterdam, The Netherlands) was used
as source of full scale data for model evaluation. The inuent of the
A-stage is composed of three streams: sewage, recirculation from
the B-stage and the recirculation from the excess sludge treatment
line. A detailed characterization of the sewage, the recirculation
from the B-stage and the overall inuent of the A-stage is presented
in Table 1.
Besides the model assumptions regarding conversion of COD
and adsorption, it is very important to clarify how the standard
measurements in the WWTP (solids and COD concentrations) are
converted into the different state variables for organic compounds.
For simulation purposes all the organic state variables are shown
in gCOD/m3 . The full scale data provided information about total
COD in gCOD/m3 and TSS (Total Suspended Solids) in gTSS/m3 ,
including the ash content. Potassium dichromate (K2 Cr2 O7 ) has
been used for all COD measurements. Since the ash content from
the TSS was on average 20%, the TSS was multiplied with 0.8 to
get the amount of VSS (Volatile Suspended Solids) in gVSS/m3 .
The VSS was divided by 0.75 to get gCOD/m3 from gVSS/m3 [15].
From the inuent VSS, 1/4 was assumed to be XHET,in . Furthermore
the assumption has been made that 3/8 of the inuent VSS would
consist of inert particulates (XI,in ) and the other 3/8 of the inuent VSS is XS,in . The XS has however been assumed to consist of
both particulate COD (60%) and soluble COD (40%). The concentration of NH4 + was only measured in the outow of the A-stage,

therefore the concentration of NH4 + in the inuent has been set

such that the concentration in the outow of the A-stage in the
model is consistent with the measured data. The NOX concentration (sum of NO2 + NO3 concentrations) in the A-stage inuent
consists mainly of NOX produced in the sidestream and efuent
recycling. The fraction coming from the sidestream treatment was
assumed to be NO2 , due to partial nitritation in the sidestream,
although the measured data from Dokhaven never showed a distinction between NO2 and NO3 (Table 1). And the fraction coming
from the efuent recycling was assumed to be NO3 (Table 1). The
values used for the state variables are consistent with the typical
ranges of inuent concentrations as shown in the ASM2 [15].
Since the assumption has been made that the total amount of
COD in the efuent consists of soluble inerts (SI ) (Table 1), the recirculation stream does also have the same concentration of SI . There
is however a small fraction of particulate COD that is also recirculated and measured as efuent VSS, which was assumed to be inert
(Table 1). Furthermore the assumption has been made that all the
NOX present in the efuent consist of NO3 , due to the nitrication
in the B-stage (Table 1).
The set of parameters used as inputs for the A-stage model
is shown in Table 2, these are the average values obtained from
the historical data les from Dokhaven WWTP. The measured data
were used to calculate the SRT. The SRTA is set at 0.3 days in the
summer (>17 C) and to 0.33 d (<17 C) in the winter.
The parameter fsettler was calculated through Eq. (3) from measured data of the WWTP. The value of the parameter fsettler was
rather constant during the year, with slight average difference
between winter (0.983) and summer (0.985). As a rst trial, no
change was applied in the model, and just a whole-year-average
was used (0.984 with standard deviation of the mean of lower than
104 ) (Table 2).
The modelling procedure is schematically depicted in Fig. 3,
where also the inputs (detailed in Table 2) and outputs, key characteristics of the model, and the use of the fsettler parameter to lump
adsorption and settling efciency were highlighted.

98

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

Table 1
Inuent concentrations based on the averages of the measured plant data. The calculations required to link the measured data to the concentrations of the state variables
used in the model are also indicated.
Sewage + Sidestream

Recirculation

Variable

Value(gN/m3 )

Calculation

Value(gN/m3 )

SNH4+
SNO2SNO3-

22
0.7
0
(gCOD/m3 )
120
48
72
33
122
395

Assumeda
1 Inuent NOX

0
0
12
(gCOD/m3 )
0
0
8
33
0
41

XS
XHET
XI
SI
SS
Total COD
a

(3/8 Inuent VSS/0.75)/0.6

1/4 Inuent VSS/0.75
3/8 Inuent VSS/0.75
1 Efuent COD
Inuent COD SI XS XHET XI

(from B-stage)
Calculation

1 Efuent NOx

1 Efuent VSS/0.75
1 Efuent COD

Inow A-stage
Value(gN/m3 )

Calculation

15
0.5
4.5
(gCOD/m3 )
75
31
46
33
77
262

Qin SNH4,in /(Qin + Qrecirc )

Qin SNO2,in /(Qin + Qrecirc )
Qrecirc SNO3,recirc /(Qin + Qrecirc )
Qin XS,in /(Qin + Qrecirc )
Qin XHET,in /(Qin + Qrecirc )
(Qin XS,in + Qrecirc XS,recirc )/(Qin + Qrecirc )
(Qin + Qrecirc ) SI,in /(Qin + Qrecirc )
Qin SS,in /(Qin + Qrecirc )

Assumed to t the measured ammonium concentration in the efuent.

Table 2
Parameters used for the simulations.
Parameter
3

Volume A-stage (VA , m )

Solids Retention Time A-stage (SRTA , d)
Particulate separation A-stage (fsettler,A )
Dissolved oxygen (DOA, gO2 /m3 )
Temperature (T, C)
Inuent (Qin, m3 /d)
Recirculation ow (Qrecirc, m3 /d)
Recirculation ow A-stage (Qr, m3 /d)
a
b
c

Steady state simulations

Dynamic simulations

4790
0.30
0.984
0.5b
17
119,363
71,458
72,000

4790
0.30/0.33a
0.985/0.983a
0.5b
variablec
variablec
variablec
variablec

Summer/winter.
Controlled at this value in the 4th tank, see details in Section 2.6.
From the measured data set, see more details in Section 2.8.

2.6. DO prole along the A-stage

The DO concentration in the A-stage in the WWTP at Dokhaven
is measured at one point (i.e. ca. 2/3 of the reactor length) and
controlled at 0.5 mgO2 /L. The main ow is in practice treated in
8 parallel lines, with a very high length to width ratio [23], which
results in a plug ow hydrodynamics. We have chosen to describe
this hydrodynamics with 5 tanks in series (with equal volume). It
is not crucial to include the plug ow hydrodynamics for the correct description of the overall COD removal but it provides a more
realistic description of A-stage reactor. In the model, a control loop
was used to maintain a DO set point of 0.5 mgO2 /L in the 4th reactor
(i.e. ca. 2/3 of the reactor length, since the A-stage was described
with 5 tanks in series). The oxygen transfer coefcient applied in
that tank was maintained constant for the rest of tanks, whereas DO
concentration was not controlled as such in tanks 13 and 5. This
strategy created a gradient of DO concentrations along the A-stage,
mimicking more closely the conditions found in the real installation (Fig. 4). Unfortunately due to the underground construction of
the plant it was not possible to measure the DO along the length of
the A-stage.
2.7. Dynamics
To describe better the disturbances of the WWTP during the
simulations, variations in the following variables were taking into
account based on the measured data from Dokhaven (see also
Table 2):
Temperature
Seasonal ow rate
Daily ow rate (a typical dry weather daily pattern from
Dokhaven has been chosen, see an example in Fig. 5)

Fig. 4. Simulated DO concentration prole along the A-stage.

Inuent concentrations (average from the inuent of Dokhaven)

The inuent loadings (kgCOD/day and kgN/day) are constant for
every day and coupled to the daily uctuation of Qin , meaning that
at 4:00 a.m. the load is lower than at 16:00 (since the ow rate is
lower at 4:00 a.m.). The seasonal ow rate on the other hand does
not have an impact on the load. Only the concentration in which
this load is carried to the WWTP depends on the seasonal ow rate.
So at high seasonal ow rates the concentration is lower and at low
seasonal ow rate the concentrations are higher (but the load is the
same). This was done to be able to mimic the dilution of pollutants
due to rainy periods.
Note how when we use temperature as a variable we refer in
this study to the water temperature.

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

99

Fig. 5. Daily uctuation of ow rates used in the simulations. Qin,A is the sum of Qin and Qrecirc.

2.8. Simulation strategy

The model was implemented in the software package AQUASIM
v2.1., and the process diagram used was depicted in Fig. 2 (grey
area). For each one of the simulations, a rst steady-state was
achieved i.e. without any dynamic variations (see Table 1 for
details). This was done for a period of 1000 days, and constant values of substrate and biomass concentrations were checked to verify
steady state was indeed achieved.
After steady state, a period of one year was dynamically simulated based on the variations in temperature, ow rate and
composition measured at Dokhaven WWTP.
3. Results and discussion
3.1. Steady state simulations
The simulation results obtained for the set of parameters stated
in Table 1 are presented in Fig. 6. For an easy comparison, the
average values from the measured data set are also included. The
model outputs for total COD concentration in the efuent underestimated the mean value from the measured data set with less
than 10% (70 and 76 gCOD/m3 for model and mean measured values, respectively). The sum of the Xs in the excess sludge and that
in the outow A-stage is 60% of total inow Xs (Fig. 6). Following
these model outputs: (i) 40% of the Xs was hydrolysed and (ii) ca.
50% of the Xs was adsorbed in the A-stage. The major part of the
biodegradable organics (those present in the inow plus those coming from hydrolysis of XS ) were removed according to the steady
state simulation results (only 3 gCOD-biodegradable/m3 remained
in the efuent, Fig. 6). The actual solids concentration in the reactor
is 12% lower than the model prediction: 1535 and 1230 gCOD/m3 ,
for model and measured mean values, respectively. The measured
SRT value was imposed in the simulations through Eq. (3), so in
practice the model is using the measured SRT as an input. A better
description of the total solids concentration would require a calibration of the inuent fractions or yield. However the efforts in
this phase were focused on the calibration of the efciency of the
adsorption, and its link with the imposed SRT, and the description
of the COD removal efciency.

Regarding nitrogen conversions, from the measured data

8 gN/m3 (difference in nitrogen Kjeldahl concentration between
inow and outow, Fig. 6) are overall consumed whereas in the
model the nitrogen consumption is 5 gN/m3 (difference in the sum
of SNH4+ , SNO2 , SNO3 , between inow and outow, Fig. 6). Ammonium concentration of the raw wastewater was not measured
(Table 1) and therefore an ammonium concentration of 15 g/m3
assumed. A better description of the nitrogen species in the A-stage
may require measured data of nitrogen species along the A-stage
and probably including nitriers in the model.
For the steady state simulations, a correct description of the COD
removal in the A-stage is also obtained when 100% of the XS is
assumed to be particulate as regularly considered in ASM models
(see results in the Supporting information, Table S1). However, for
a better description of the dynamic simulations, 40% of the slowly
biodegradable organics (XS ) is considered as soluble for the initial
fractionation of the COD. See the next section for more information.
3.2. Dynamic simulations
The modelling results in terms of COD removal under dynamic
variations of temperature, ow rate (seasonal and daily variations,
see Fig. 7B and Fig. 5, respectively) and inuent concentrations were
assessed for a period of one year. The simulation results are presented in Fig. 7A. The model is able to describe the main trends
found in the experimental data set for COD time course concentrations, with average relative deviation of less than 10% (for individual
points highest deviation was below 30%).
The goal of the dynamic simulation was to be able to describe
mathematically the drop in COD removed by the A-stage during
winter. The important drop in the COD removal during December
2012 and January 2013 could be described with the model in the
dynamic simulation (see Fig. 7A). The drop in the performance is
linked to the decrease of heterotrophic activity due to the low temperatures (see the coupling with the temperature, Fig. 7), rather
than to a less efcient adsorption (particulate separation in the Astage is rather constant along the year, see Table 2). This assessment
was important since a temporary reduction in COD removal in the
A-stage would mean an increase in the COD/N ratio of the wastewater to be further treated in the B-stage for N removal. In case

100

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

Fig. 6. Summary of the main model outputs for the steady state simulation and direct comparison to measured data. # No measured data available for this stream; *Calculated
as XHET +0.6XS +XI ; **Calculated as XHET +XS +XI . Blank cells for measured data indicate data not available.

of a higher COD concentration reaching the B-stage, heterotrophs

could proliferate on the external surface of the granular sludge. A
higher DO set point should be applied to assure the adequate oxygen concentration in the inner layers, where ammonium is oxidized
to nitrite by ammonia oxidizing bacteria. Otherwise, if the DO set
point is kept constant at the same value after the increase in COD,
this will impact the performance of the N removal, simply because a
fraction of the oxygen will be used for aerobic COD removal by heterotrophs and not for ammonium oxidation into nitrite. In turn, this
type of disturbance might trigger nitrite-oxidizing bacteria development as pointed out in previous studies [21,22]. Proliferation of
nitrite-oxidizing bacteria is one of the identied challenges in the
implementation of autotrophic N removal in the main water line of
the WWTP [28].

3.3. Applicability of the model & further research requirements

The model here presented can be easily applied for the mathematical description of existing installations in which the particulate
separation coefcient (fsettler ) can be identied from the performance of the wastewater treatment plant (e.g. historical data les
or current values in a time frame of months or preferably one year).
COD fractionation when applying the model to other WWTPs could
be done either by following the approach here detailed or with
general guidelines available in literature [29]. A good fractionation
between soluble (S) and particulate (X) COD is essential. In the short
SRT typical for A-stage processes (here 8 h) biological conversion of
colloidal and particular COD is marginal, making the distribution of
COD over the fractions XS , XHET and XI not sensitive to the model
outcome. Further research would be required to characterize this
coefcient in different plants that may present different A-stage
performances to generalize the ability of the model for the description of a broader spectrum of highly loaded aerobic COD-removal
systems (A-stages). The effects of the SRT and the temperature
on the parameter fsettler were measured for winter and summer
(Table 2). In the narrow range of measured SRT (0.30 and 0.33 d
for summer and winter, respectively) the effect on fsettler is limited,
comprising a reduction of 0.9850.983 from summer and winter,

respectively, being the efciency of the adsorption higher at lower

SRT during the summer. An independent effect of temperature and
SRT on fsettler is not directly available from the experimental data.
Further research will be required to conrm if a longer SRT could
be the major driving force, rather than a direct temperature effect.
A wider (and possibly mechanistically based) correlation between
SRT and fsettler would be desirable to get a more general model able
either to describe different WWTPs or to use of the model for process optimization in a wider range of operation conditions. This will
require an in depth study of the effects of DO level, the potential
impact of the specic wastewater treatment among other related
factors.
4. Conclusions
It has been possible to develop a simple model to simultaneously
describe growth and adsorption of COD. The agreement between
the measured data and the simulations is remarkably good for
both steady-state conditions and dynamic simulations mimicking
the seasonality of one whole year of operation of the A-stage in
Dokhaven. The following conclusions can be drawn:
A reduced set of state variables considered in the model produced
a very appealing approach for the simplest-possible modelling of
highly loaded aerobic COD removal reactor or A-stage.
One of the key points for the calibration of the model is to have
specic data available to calculate the lumped parameter (fsettler ),
which accounts for both adsorption and settling efciencies.
Especially during winter conditions, the A-stage is not capable of completely removing SS . This may impact the potential
implementation of a partial nitritation/anammox for autotrophic
nitrogen removal in the B-stage.
Acknowledgements
JP work was supported a Marie Curie Intra European Fellowship
(GreenN2, PIEF-GA-2012-326705). This research was funded by the
SIAM Gravitation Grant 024.002.002, the Netherlands Organization

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

101

Fig. 7. A. Comparison between COD-removal from measured data (grey dashed line) and simulation data (green solid line) from one whole year. Including changing
temperature and ow-rate. The gaps in some weeks during July and August 2012 correspond to missing data points from the measured data set. B. Temperature during a
year (solid red line) and the inow (Qin,A) in the A-stage (dashed blue line). (For interpretation of the references to colour in this gure legend, the reader is referred to the
web version of this article.)

for Scientic Research. The authors acknowledge the cooperation of the plant operators at Sluisjesdijk WWTP (Rotterdam, The
Netherlands).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.bej.2016.04.004.

Appendix A.
See Table A1, Table A2 and Table A3.
Table A1
Stoichimetric matrix Aij (Adapted from Mozumder et al. [21]).
Aij

i component

SNH4
[gCOD m3 ] [gN m3 ]
SS

SNO2

[gN m3 ]

SNO3

[gN m3 ]

SO2

[gO2 m3 ]

SN2

[gN m3 ]

XI
XS
[gCOD m3 ] [gCOD m3 ] [gCOD m3 ]
XHET

j process
Growth
1. aerobic growth XHET

1
YHET

2. anoxic (NO2 ) growth of XHET

1
YHET NO2

3. anoxic (NO3 ) growth of XHET

1
YHET NO3

Hydrolysis
4. hydrolysis of XS
Decay

5. decay of XHET

1 i
NSS
YHET
1
iNSS
iNXB +
YHET,NO3

iNXB +

iNXB +

1
1YHET,NO
2

1.71YHET,NO
2
1YHET,NO
3
1.14YHET,NO
3

1
iNSS
YHET,NO3

iNXB fI iNXI

1 fI

1 YHET,NO
3
1.14 YHET,NO
3

1
YHET

1
1YHET,NO

2
1
1.71YHET,NO
2
1

1
iNSS

1 fI

fI

102

J. Smitshuijzen et al. / Biochemical Engineering Journal 112 (2016) 94102

Table A2
Kinetic rate expressions.
j.

Process

Process rate (d1 )

Growth of XHET.

max,HET

2
3

Growth of XHET,NO2
Growth of XHET,NO3

Reference

SO

max,HET NO2
max,HET NO3

Hydrolysis of XS

qXS,HYD

Decay of XHET

bHET XHET

SS

KS,HET +SS

O2,HET

O2,HET

+SO2

O2,HET

 +SO2

XS /XHET
KX +XS /XHET

2
KO ,HET +SO
2
2

O2,HET

Value

Stoichiometric parameters
0.5
YHET
YHET,NO2
0.53
YHET,NO3
0.53
iNXB
0.07
iNXI
0.07
iNSS
0.03
fI
0.08
kinetics at T = 30
max,HET
12 d1
qXS,HYD
4.44 d1
KNO2 ,HET
0.3
KNO3 ,HET
0.3
KS,HET
4
0.2
KO2 ,HET
KNH4 ,HET
0.02
KX
0.03
bHET
0.6
NO2 = NO3 = NOX
0.8

Reference

g COD g1 COD
g COD g1 COD
g COD g1 COD
g N g1 COD
g N g1 COD
g N g1 COD
g COD g1 COD

[6]
[30]
[30]
[21]
[21]
[15]
[15]

g N m3
g N m3
g COD m3
g O2 m3
g N m3
g XS g XHET 1
d1

[15]a
[15]a
[31]
[31]
[15] (ASM2)
[15]
[21]
[15]
Assumed H
max /20 [21]
[15]

a
Conversion of ASM1-values given by Henze et al. [15] using temperature relationship proposed by these authors (ASM2).

References
[1] B. Bhnke, Das adsorptions-belebungsverfahren, korrespondenz abwasser,
Jahrg 24 (1977) 277.
[2] B. Bohnke, B. Diering, Process for purication of efuent. Patent assignee:
Botho Bohnke (Aachen, DE), US Patent, 1988.
[3] A.I. Versprille, B. Zuurveen, Th. Stein, the AB process: a novel two stage
wastewater treatment system, Water Sci. Technol. 17 (1984) 235246.
[4] B. Boehnke, R. Schulze-Rettmer, S.W. Zuckut, Cost-effective reduction of
high-strength wastewater by adsorption-based activated sludge technology,
Water Eng. Manage. 145 (1998) 3134.
[5] W. Zhao, Y.P. Ting, J.P. Chen, C.H. Xing, S.Q. Shi, Advanced primary treatment
of waste water using a bio-occulation-absorption sedimentation process,
Acta Biotechnol. 20 (2000) 5364.
[6] M.C.M. van Loosdrecht, T. Kuba, H.M. van Veldhuizen, F.A. Brandse, J.J.
Heijnen, Environmental impacts of nutrient removal processes: a case study,
J. Environ. Eng. 123 (1997) 3340.
[7] M.S.M. Jetten, S.J. Horn, M.C.M. van Loosdrecht, Towards a more sustainable
municipal wastewater treatment system, Water Sci. Technol. 38 (1997)
171180.
[8] H. Siegrist, D. Salzgeber, J. Eugster, A. Joss, Anammox brings WWTP closer to
energy autarky due to increased biogas production and reduced aeration
energy for N-removal, Water Sci. Technol. 57 (2008) 383388.
[9] B. Kartal, J.G. Kuenen, M.C.M. van Loosdrecht, Sewage treatment with
anammox, Science 328 (2010) 702703.
[10] R. Pujol, J.P. Canler, Biosorption and dynamics of bacterial populations in
activated sludge, Water Res. 26 (1992) 209212.
[11] A. Guellil, F. Thomas, J.C. Block, J.L. Bersillon, P. Ginestet, Transfer of organic
matter between wastewater and activated sludge ocs, Water Res. 35 (2001)
143150.
[12] J. Jimenez, M. Miller, C. Bott, S. Murthy, H. De Clippeleir, B. Wett, High-rate
activated sludge system for carbon managementevaluation of crucial
process mechanisms and design parameters, Water Res. 87 (2015) 476482.
[13] S. Haider, K. Svardal, P.A. Vanrolleghem, H. Kroiss, The effect of low sludge age
on wastewater fractionation (SS ,SI ), Water Sci Technol. 47 (2003) 203209.

NH4,HET

SNO3
KNO3,HET +SNO3

2
KO ,HET +SO
2
2

Unit

SNH4
+SNH4

SNO2
KNO2,HET +SNO2

SO

Table A3
Parameters.
Parameter

+ NOX

XHET
SS

KS,HET +SS
SS

KS,HET +SS

KO ,HET
2
KO ,HET +SO
2

[21]

SNO2
SNO2 +SNO3

SNO3
SNO2 +SNO3

SNO

SNH4
+SNH4

XHET

[21]

SNH4
+STAN

XHET

[21]

NH4,HET

3
KNO ,HET +SNO
3
3

NH4,HET

XHET

[15]
[15]

[14] T.M. Nogaj, A.A. Randall, J.A. Jimenez, I. Takacs, C.B. Bott, S. Miller,
Mathematical modeling of carbon removal in the high-rate activated sludge
system: model presentation and application, in: Proceedings of the 86th
Annual Water Environment Federation Technical Exhibition and Conference
[CDROM], Chicago, Illinois, Oct. 59; Water Environment Federation:
Alexandria, Virginia, 2013.
[15] M. Henze, W. Gujer, T. Mino, M.C. van Loosdrecht, Activated Sludge Models
Asm1, Asm2, Asm2d and Asm3 Scientic and Technical Report; No. 9), IWA
Publishing, London, 2000.
[16] H.Q. Ge, D.J. Batstone, J. Keller, High-rate activated sludge process combined
with biological phosphorus removal as a sustainable approach for wastewater
treatment, in: IWA Specialist ConferenceGlobal Challenges: sustainable
Wastewater Treatment and Resource Recovery, Kathmandu, Nepal, 2014.
[17] S. Hu, D. Batstone, D. Solley, C. Hertle, J. Keller, Sustainable wastewater
treatment and resources recovery through optimal technology integration, in:
IWA Specialist ConferenceGlobal Challenges: Sustainable Wastewater
Treatment and Resource Recovery, Kathmandu, Nepal, 2014.
[18] M.W. Miller, P. Regmi, J. Jimenez, S. Murthy, B. Wett, C.B. Bott, Optimizing
absorption-style high rate activated sludge for BNR and energy recovery, in:
IWA Specialist ConferenceGlobal Challenges: Sustainable Wastewater
Treatment and Resource Recovery, Kathmandu, Nepal, 2014.
[19] A. Rahman, T. Wadhawan, E. Khan, R. Riffat, I. Takcs, H. De Clippeleir, B. Wett,
J.A. Jimenez, A. Al-Omari, S. Murthy, Characterizing and quantifying
occulated and adsorbed chemical oxygen demand fractions in high-rate
processes, in: IWA Specialist Conferenceglobal Challenges: Sustainable
Wastewater Treatment and Resource Recovery, Kathmandu, Nepal, 2014.
[20] B. Wett, M. Hell, J. Jimenez, I. Takacs, C. Bott, S. Murthy, Operational and
structural A-stage improvements for high-rate carbon removal, in: IWA
Specialist ConferenceGlobal Challenges: Sustainable Wastewater Treatment
and Resource Recovery, Kathmandu, Nepal, 2014.
[21] Md. S.I. Mozumder, C. Picioreanu, M.C.M. van Loosdrecht, E.I.P. Volcke, Effect
of heterotrophic growth on autotrophic nitrogen removal in a granular sludge
reactor, Environ. Technol. 35 (2014) 10271103.
[22] S. Lackner, A. Terada, B.F. Smets, Heterotrophic activity compromises
autotrophic nitrogen removal in membrane aerated biolms: results of a
modeling study, Water Res. 42 (2008) 11021112.
[23] H.A. Meijer, Rotterdam-Dokhaven sewage treatment plant; a large sewage
treatment plant in the midst of a developing residential quarter, Water Sci.
Technol. 20 (1988) 267274.
[24] T. Lotti, R. Kleerebezem, Z. Hu, B. Kartal, M.K. de Kreuk, C. van Erp Taalman
Kip, J. Kruit, T.L.G. Hendrickx, M.C.M. van Loosdrecht, Pilot-scale evaluation of
anammox-based mainstream nitrogen removal from municipal wastewater,
Environ. Technol. 36 (2014) 11671177.
[25] H. Kroiss, F. Klager, S. Winkler, G. Wandl, K. Svardal, Nutrient removal process
development and full implementation at the 4 million p.e. main treatment
plant of Vienna, Austria, Water Sci. Technol. 50 (2004) 1926.
[26] J. Smitshuijzen, Modelling the AB-process to Implement Cold CANON. Master
thesis, TU Delft, 2015.
[27] G. Sin, D. Kaelin, M.J. Kampschreur, I. Takcs, B. Wett, K.V. Gernaey, L. Rieger,
H. Siegrist, M.C.M. van Loosdrecht, Modelling nitrite in wastewater treatment
systems: a discussion of different modelling concepts, Water Sci. Technol. 58
(2008) 11551170.
[28] M.K.H. Winkler, R. Kleerebezem, J.G. Kuenen, J. Yang, M.C.M. van Loosdrecht,
Segregation of biomass in cyclic anaerobic/aerobic granular sludge allows the
enrichment of anaerobic ammonium oxidizing bacteria at low temperatures,
Environ. Sci. Technol. 45 (2011) 73307337.
[29] P.J. Roeleveld, M.C.M. van Loosdrecht, Experience with guidelines for
wastewater characterisation in The Netherlands, Water Sci. Technol. 45
(2006) 7787.
[30] A. Muller, M.C. Wentzel, R.E. Loewenthal, G.A. Ekama, Heterotroph anoxic
yield in anoxic aerobic activated sludge systems treating municipal
wastewater, Water Res. 37 (2003) 24352441.
[31] E. Alpkvist, C. Picioreanu, M.C.M. van Loosdrecht, A. Heyden,
Three-dimensional biolm model with individual cells and continuum EPS
matrix, Biotechnol. Bioeng. 94 (2006) 962979.