ARTICLE

Kinetic Model of a Granular Sludge SBR:

Influences on Nutrient Removal
M.K. de Kreuk,1 C. Picioreanu,1 M. Hosseini,1 J.B. Xavier,2 M.C.M. van Loosdrecht1
1
Department of Biotechnology, Delft University of Technology, Julianalaan 67, 2628BC Delft,
The Netherlands; telephone: 0031-15-2781551; fax: 0031-15-2782355;
e-mail: [email protected]
2
Biomathematics group, Instituto de Technologia Quı́mica e Biológica, Universidade Nova
Lisboa, R. Qta Grande 6, 2780 Oeiras, Portugal
Received 23 May 2006; accepted 25 August 2006
Published online 18 December 2006 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/bit.21196

Introduction
ABSTRACT: A mathematical model was developed that can Aerobic granular sludge technology is a new and promising
be used to describe an aerobic granular sludge reactor, fed
with a defined influent, capable of simultaneously removing
development in the field of aerobic wastewater treatment.
COD, nitrogen and phosphate in one sequencing batch This technology is based on sequencing batch reactors
reactor (SBR). The model described the experimental data (SBRs), with a cycle configuration chosen such that a strict
from this complex system sufficiently. The effect of process selection occurs for fast settling aerobic granules and that
parameters on the nutrient removal rates could therefore be these granules are oxygen limited instead of substrate
reliably evaluated. The influence of oxygen concentration,
temperature, granule diameter, sludge loading rate, and
limited. This leads to the growth of stable and dense granules
cycle configuration were analyzed. Oxygen penetration under anaerobic/aerobic conditions (De Kreuk et al., 2005a;
depth in combination with the position of the autotrophic Dulekgurgen et al., 2003; Li et al., 2005). Compared to
biomass played a crucial role in the conversion rates of the conventional activated sludge systems, a system based on
different components and thus on overall nutrient removal aerobic granular sludge has several advantages, namely
efficiencies. The ratio between aerobic and anoxic volume in
the granule strongly determines the N-removal efficiency as
compactness, lower operational and construction costs and
it was shown by model simulations with varying oxygen lower energy requirement. Many laboratory studies (among
concentration, temperature, and granule size. The optimum others Beun et al., 1999; Liu and Tay, 2004; Morgenroth
granule diameter for maximum N- and P-removal in the et al., 1997; Tay et al., 2002), a feasibility study (De Bruin
standard case operating conditions (DO 2 mg L
1, 208C) et al., 2004) and a pilot study (De Bruin et al., 2005) showed
was found between 1.2 and 1.4 mm and the optimum COD
loading rate was 1.9 kg COD m
3 day
1. When all ammonia
the potential of aerobic granular sludge SBRs, and the next
is oxidized, oxygen diffuses to the core of the granule step is to apply this technology in practice.
inhibiting the denitrification process. In order to optimize Mathematical modeling has proven very useful to study
the process, anoxic phases can be implemented in the complex processes, such as the aerobic granular sludge
SBR-cycle configuration, leading to a more efficient overall systems (Beun et al., 2001; Lübken et al., 2005). Biological
N-removal. Phosphate removal efficiency mainly depends
on the sludge age; if the SRT exceeds 30 days not enough
processes in the granules are determined by concentration
biomass is removed from the system to keep effluent phos- gradients of oxygen and diverse substrates. The concentra-
phate concentrations low. tion profiles are the result of many factors, for example,
Biotechnol. Bioeng. 2007;97: 801–815. diffusion coefficients, conversions rates, granule size,
ß 2006 Wiley Periodicals, Inc. biomass spatial distribution, and density. All of these
KEYWORDS: aerobic granular sludge; modeling; nutrient factors tightly influence each other, thus the effect of
removal; SBR; simultaneous nitrification/denitrification separate factors cannot be studied experimentally. More-
(SND); P-removal over, due to the long sludge age in granule systems (usually
up to 70 days, De Kreuk and Van Loosdrecht, 2004), lengthy
experiments should be carried out before we can speak of
a steady state approaching system. Therefore, a good

Correspondence to: M.K. de Kreuk

Contract grant sponsors: STOW; STWA; FCT/MCTES, Portugal
Contract grant numbers: TNW99.262; DPC5577; SFRH/BPD/11485/2002

ß 2006 Wiley Periodicals, Inc. Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007 801
computational model for the granular sludge process the well-established AQUASIM simulation software for
provides significant insight in the most important factors environmental applications (Reichert, 1998). A combina-
that affect the nutrient removal rates and in the distribution tion of completely mixed reactor and biofilm reactor
of different microbial populations in the granules. Further, compartments (based on Wanner and Gujer, 1986, provided
the model could also be used for process optimization and by AQUASIM, was used to simulate the mass transport and
for the scale-up and design of a full-scale reactor for the new conversion processes occurring in the bulk liquid and in the
aerobic granular sludge technology (or NeredaTM). granules of the SBR system. The biological conversion
The computational model used in this study is based on processes are described using stoichiometric and kinetic
previously developed models. The SBR and granular sludge parameters from the models published by Hao et al. (2001,
descriptions are principally as in the model of Beun et al. 2002b)) and Meijer (2004).
(2001), which described heterotrophic organisms storing The model was used to simulate the conversion processes
acetate (without phosphate accumulation) in a feast-famine that occur in a laboratory scale reactor and compare
regime in combination with autotrophic organisms for them with those obtained experimentally under the same
nitrogen removal. Because the design of NeredaTM aims at conditions. Furthermore, it was used to predict the nutrient
phosphate removal via the selection of slow growing removal efficiencies under different circumstances. Special
phosphate accumulating organisms (PAO), the existing attention has been paid to the distribution of the different
aerobic granular sludge model was extended with conver- microbial species in the granules, as resulting from the
sion processes involving PAO as described in the models of model simulations. This distribution, in combination with
continuous activated sludge systems by Hao et al. (2001) the oxygen penetration depth, gives insight in the overall
and Meijer (2004). The results from the present model reactor performance and its sensitivities.
simulations are compared here to different laboratory scale
experiments. This model will be used here for evaluation of Conversion Processes
the overall reactor performance only. A detailed modeling The model accounts for biomass distribution among seven
study of the population dynamics within aerobic granular particulate components (Fig. 1). Three active biomass types
sludge is described in Xavier et al. (Submitted). include the phosphate-accumulating organisms (XPAO)
and the nitrifiers (ammonia-oxidizers, XNH and nitrite-
Materials and Methods oxidizers, XNO). One inactive biomass component (XI) was
defined. Special pools are defined for internally stored
Laboratory Scale Set-Up biomass: polyphosphate (XPP), poly-hydroxybutyrate
Experiments were performed in 3 L SBRs (internal diameter (XPHB), and glycogen (XGLY). Seven dissolved components
6.25 cm) operated as bubble column (SBBC) or airlift are considered relevant for the biological conversions
reactor (SBAR). Air was introduced via a fine bubble aerator and process stoichiometry: oxygen ðSO2 Þ, acetate (SAc),
at the bottom of the reactors (4 L/min). Dissolved oxygen ammonium ðSNH4 Þ, nitrite ðSNO2 Þ, nitrate ðSNO3 Þ, nitrogen
(DO) was measured as percentage of oxygen saturation ðSN2 Þ, and phosphate ðSPO4 Þ.
and could be controlled via adding N2 or air to the It was assumed that acetate would be consumed by PAO
recirculation flow of the off gas through the reactor. pH was and converted to a storage polymer (polyhydroxybutyrate,
maintained at 7.0  0.2 by dosing 1 M NaOH or 1 M HCl.
Hydraulic retention time was 5.6 h and the substrate
load, measured in chemical oxygen demand (COD), was
1.6 kgCOD m
3day
1.
The reactors were operated in successive cycles of 3 h:
60 min of anaerobic feeding from the bottom of the reactor;
112 min aeration; 3 min settling; 5 min effluent discharge.
The composition of the influent media were (A) Na acetate
63 mM, MgSO47H2O 3.6 mM, KCl 4.7 mM and
(B) NH4Cl 35.4 mM, K2HPO4 4.2 mM, KH2PO4 2.1 mM,
and 10 ml/L trace element solution. From each medium (A and
B) 0.15 L per cycle were dosed together with 1.3 L tap water.
Exact methods for reactor operation, trace element
solution, and measurement of the experimental data used
for model calibration and comparison are described in De
Kreuk et al. (2005a,b).
Figure 1. Schematic representation of the model set-up in AQUASIM. Each
compartment contains 7 soluble components (SO2, SAc, SNH4, SNO2, SNO3, SPO4, SN2) and 7
Model Description particulate components (XPAO, XNH, XNO, XI, XPP, XPHB, XGLY) involved in 21 metabolic
processes. All soluble and particle components are found in effluent and recirculation
The mathematical model describing the performance of the links.
laboratory-scale granular sludge SBR was implemented in

802 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
PHB) during the anaerobic period. Converting glycogen and Stoichiometric and kinetic parameter values are given in
degrading poly-P generate energy needed for the active Appendix 2 and Appendix 3, respectively. A few model
uptake of acetate. During the aerobic period, the stored PHB parameters were altered with respect to the original models,
is used for growth, maintenance and restoring the glycogen as can be read from the reference column of these
and poly-P pools. The formation of glycogen and the Appendices.
formation of poly-P are always coupled to the consumption Since only one kinetic model describes all the different
of PHB and therefore, these aerobic or anoxic processes processes and the experimental results in both the aerobic
cannot be read separately. In the aerobic period, also granules and in the reactor, the following parameters needed
nitrification takes place, with ammonia converted via nitrite to be adjusted. The reduction factor for process rates under
into nitrate. All anoxic processes can take place on nitrite anoxic conditions (h) was chosen 0.3, since this described
and nitrate as electron acceptors, depending on their best the experimentally measured values for nitrate
availability. Also endogenous respiration for autotrophic production. This low reduction value is needed, because
organisms is included. When no PHB, glycogen nor acetate in this model denitrification can occur on both nitrite and
is available, PAO will decay. For simplicity and to maintain nitrate. In other models, denitrification often only takes
the granule structure, the decayed biomass was considered place on nitrate. Allowing denitrification with nitrate and
to be turned into inert material (Beun et al., 2001). To limit nitrite simultaneously and not adding an inhibition factor
the calculation time for the simulation, no other hetero- for the preference of one of the two electron acceptors would
trophic organisms (XH) were included in the model. This is double the denitrification rates, producing unrealistic
justified by the fact that a very small amount of the incoming results if no reduction factor (h) was used.
acetate is left at the start of the aerobic period, which could Part of the phosphate removal in aerobic granular sludge
be consumed by aerobic heterotrophs and by PAO (Fig. 2). is removed by (biologically induced) precipitation (De
The lack of acetate during the aerobic period will limit Kreuk et al., 2005a). For simplicity, precipitation was not
the growth and will result in the total disappearance of included in the model. As an alternative, the maximum
max
heterotrophic organisms other than the ones capable of fraction of poly-phosphate ðfPP Þ was increased from
anaerobically storing substrate (as shown by Xavier et al., 0.35 (Wentzel et al., 1989) to 0.65. Consequently, the
Submitted). Therefore, omitting the heterotrophs that do maximum fraction of stored PHB had to be adjusted in
not store substrate is not expected to greatly influence the order to generate enough PHB for the increased poly-P
long-term (‘‘steady-state’’) simulation results. uptake.
The stoichiometry and kinetics of biological conversions Another model adjustment concerns the expressions for
are derived from the Delft bio-P model for activated sludge the reaction rates. In order to prevent the occurrence of
(Meijer, 2004; Murnleitner et al., 1997; van Veldhuizen et al., negative values for solute concentrations, switch functions
1999) and presented with the Activated Sludge Model were used (Appendices 1 and 3). These continuous S-shaped
(ASM) notation (Henze et al., 1999) in Appendix 1. functions shut down a process consuming a reactant whose
concentration is nearly zero and, in this way, make the
simulations numerically more stable.

Reactor Description
Liquid circulation pattern. In the laboratory scale SBR,
influent is fed from the bottom of the reactor with a flow
pattern close to plug flow through the settled bed of granules
(De Kreuk and Van Loosdrecht, 2004). At the end of each
cycle, the effluent is discharged from a port at half of the
reactor height. The AQUASIM software does not allow the
volume of the bulk liquid in the biofilm compartment to
vary in time. In order to circumvent this and simulate the fill
and discharge process, two linked compartments have to be
defined. A completely mixed liquid compartment with
variable volume is connected to the biofilm compartment. A
high fluid circulation rate (Qexchange) between the two
compartments ensures the same bulk liquid concentrations
in both compartments (Fig. 1).
We chose to represent the flow pattern in the reactor as
Figure 2. Measured concentrations during a cycle measurement (points; PO4-P completely mixed because the simulation in AQUASIM of
(*), Acetate (^), NH4-N (^), NO3-N (~), NO2-N (*)) and simulated results with the different liquid circulation regimes in different phases of the
present model (lines), in the standard case (DO 2 mgL
1, 208C, granule diameter
1.1 mm). SBR cycle would imply enormous difficulties. It is expected
that the liquid circulation in the feeding phase affects the

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 803

Biotechnology and Bioengineering. DOI 10.1002/bit
diffusion rate of acetate into the granule. Consequently, it (Fig. 2). This simulation case was performed under the same
also determines the stored substrate in the anoxic interior of operational conditions as applied during this standard
the granules available for denitrification during the aerobic operation of the SBR. The DO was controlled at 2 mg  L
1
period (simultaneous nitrification/denitrification, SND). during the aerobic phase, the temperature was 208C and
With a plug-flow feeding pattern, the substrate concentra- influent concentrations were as given in Table 4. The initial
tion in the lowest part of the settled granule bed is equal to concentrations of the biomass in the granule were:
the influent substrate concentration. In the model, however, PAO (XPAO) 50,000 gCOD  m
3; ammonia oxidizers
the influent is directly mixed with the remaining bulk liquid (XNH) and nitrite oxidizers (XNO) 5,000 gCOD  m
3 each;
in the reactor, leading to a decreased concentration in the inerts (XI) 1 gCOD  m
3; glycogen concentration (XGLY)
bulk liquid surrounding the granules. To compensate for 22,500 gCOD  m
3; PHB concentration (XPHB) 36,000
this lower substrate concentration, the diffusion coefficient gCOD  m
3; poly-P concentration (XPP) 29,250 gP  m
3
of acetate in the AQUASIM model during feeding was (initial concentration of the storage polymers was set to
increased arbitrarily. 90% of the maximum fraction). These initial values were
Cyclic SBR operation. The model biofilm compartment based on the overall density of aerobic granular sludge
contains a biofilm volume, which accounts for all the (80 gTSS  L
1) and a sufficient amount of storage polymers
granules and bulk liquid volume (Fig. 1). The volume of this for the conversions.
compartment was fixed at 1.5 L. The completely mixed The operation of the SBR under different conditions
compartment contained the remainder of the liquid volume (such as changed DO concentrations, the influence of
in the system, with a maximum volume of 1.6 L. The granule diameter and of sludge load) was simulated for
influent (1.5 L) entered the reactor in the biofilm 365 days, which corresponds to about 8 h of computing time
compartment during the first 60 min of the cycle. In the on a Pentium 4, 2.4 GHz processor. After 200 days of
model, aeration of the bulk liquid was switched on during simulation, the effluent concentrations were compared at a
the next 112 min (3 h cycle) or 232 min (4 h cycle). The last 5 days interval. The fluctuations of effluent concentrations
8 min of the cycle effluent is discharged from the mixed over these intervals were less than 1%. Inside the simulated
compartment. Other parameters needed for the AQUASIM granule, in the outer layer from radius 0.22 to 0.55 mm
model are given in Appendix 4. (87.5% of the volume of the granule), the biomass
Number and size of granules. The number of granules, in concentration did not change more than 1% per 5 days
combination with the granule diameter, can affect the after 350 days of simulation. In the inner zone of the granule
outcome of the simulation, because this influences the (radius 0–0.22 mm) the biomass concentrations of the
overall liquid/granule interfacial area. The granules grown in autotrophic organisms changed with less than 4% per
the laboratory had the following characteristics: granule size 5 days. These results showed that, in the simulations, after
distribution between 0.36 and 3.7 mm (average diameter 365 days the reactor effluent and the granule composition
1.1 mm, standard deviation 0.45); aspect ratio 0.72; shape could be considered sufficiently close to a steady state. In the
factor 0.74; dry weight in reactor 18 gVSS  L
1; ash content experiments at laboratory scale, measurements are mostly
37%; SVI8 15 mL  g
1 (De Kreuk et al., 2005b). The performed between 4 months and 1.5 years after an
laboratory reactor contains thus a wide variation in granule operation change, which is in agreement with the simulated
sizes, which also change in time. However, the use of a 365 days of operation.
system with a granule size distribution was avoided in this
model, since it would significantly increase the complexity of
Results
the numerical computations and it is not expected to
contribute to a better understanding of the system.
Standard Case: Evaluation of the Model With
Therefore, in the simulations the diameter chosen was
Experimental Data
1.1 mm, which is the most representative for aerobic
granules in our reactor. The number of granules was set to The standard operation of the granule reactor was simulated
600,000, resulting in a comparable amount of biomass as first. The results of this simulation, in steady state reached
present in the laboratory scale reactor. The sensitivity of the after 365 days, are shown in Figure 2, together with the
size and the number of granules was evaluated in several concentrations measured in the laboratory scale reactor
simulations. A radius-dependent biofilm area was set in during a cycle with the same operational conditions.
AQUASIM to correctly represent the spherical symmetry of Conversion rates measured in the experiments were
the granules. compared to the ammonium and phosphate consumption
and to the net nitrite and nitrate production rates obtained
from the model. Maximum conversion rates observed in the
Simulation Strategy
experiments were determined by taking the largest slope of
Since experimental data measured during one cycle are at least three data points in the cycle. This was compared to
different in time, depending on granule size, morphology the average conversion rate over the same time interval in
and reactor operation, a representative cycle measurement the simulation. The conversion rates resulting from the
was chosen to compare to a standard simulation case model in this standard case did not differ with more than

804 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
Table I. Effect of dissolved oxygen concentration on the overall N- and P-removal rates in the SBR during short- and long-term changes (standard case in
italic).

Overall net conversion rates

(mg-N  L
1  h
1, mg-P  L
1  h
1)

NHþ
4 consumption NO

2 production NO

3 production PO3

4 consumption

DO concentration 1 mg/L (one cycle change) 7.2 (8.5) 0 (0) 1.5 (0) 98 (144)
DO concentration 2 mg/L (long-term operation) 15.9 (18.0) 1.1 (1.1) 10.0 (14.4) 142 (156)
DO concentration 4 mg/L (long-term operation and start for 23.1 (15.1) 0 (0.2) 16.4 (2.4) 173 (199)
the one-cycle changes)
DO concentration 10 mg/L (one cycle change) 28.8 (19.4) 1.0 (n.a.) 26.0 (20.2) 197 (267)
DO concentration 10 mg/L (long-term operation) 31.9 (15.4) 1.2 (1.7) 26.1 (22.5) 193 (202)

The short-term changes are simulated with the biomass distribution as developed after 365 days at DO 4 mg/L. Rates are calculated from the slopes of the
simulated and experimentally determined concentration profiles in the reactor during a cycle. Experimental data are shown between parentheses (De Kreuk
et al., 2005a).

30% from the values found in the experiments (Table I, first was lacking the total N-removal efficiency decreased. At
row). Given all the uncertain parameters (e.g., exact granule 2 mgO2  L
1 the simulated total N-removal efficiency
size distribution and granule surface area), it was concluded was 60% (61% in the experimental data-set), while at
that the model described the experimental data satisfacto- 10 mgO2  L
1 it was only 33% (10% in the experimental
rily. Default parameters were thereby considered to be good data-set). The ammonia consumption rate at higher oxygen
enough not to calibrate the model any further and to start concentrations is particularly overestimated in the model
performing other simulations in different operational simulations, especially for the long-term oxygen change.
conditions. For this reason, the parameter values for this These higher N-removal rates and efficiencies obtained from
standard case are used to obtain all other results, unless the model simulations compared to the experiments can be
mentioned otherwise. explained by the lower biomass concentration present
in the reactor during the experiments at 10 mg O2  L
1
Short- and Long-Term DO Changes (16 gTSS  L
1 vs. 28 gTSS  L
1 during the experiment at
2 mgO2  L
1, later TSS value was used for model
To study the effects of short- and long-term DO changes, the calibration).
DO was varied for different series of model simulations and The phosphate removal at higher oxygen concentrations
the results were compared with data from laboratory is faster than at low oxygen concentrations. This is due to the
experiments (De Kreuk et al., 2005a). For long-term effects, significantly faster storage of Poly-P in aerobic conditions
the development of granular sludge was simulated during than in anoxic conditions. At higher oxygen concentrations
365 days of reactor operation at DO concentrations of 2, 4, and thus at an increased oxygen penetration depth in the
and 10 mg  L
1. The results of the 4 mg  L
1 simulation granules, more PAO will be able to store their phosphate
were then used to simulate short-term effects of a DO shift aerobically, resulting in an increased P-removal rate. At all
during one cycle to 1 mgO2  L
1 and to 10 mgO2  L
1, simulated oxygen concentrations here, the storage was high
respectively. Due to the relatively short-time interval during enough to have complete phosphate removal.
these DO shifts, the biomass distribution within the granule
did not significantly change.
The Influence of Temperature
Conversion rates obtained experimentally and by
simulation, are given in Table I. The conversion rates for Variations in process temperature (20, 15, and 88C) were
the short-term experiments show the same trend. At an simulated according to experiments performed in a
increased DO most rates are enhanced. Ammonium is laboratory scale reactor. Starting with the steady state data
consumed faster, while the net nitrate production rate of the standard case simulation at 208C (2 mg O2  L
1),
increased even more, due to lower denitrification rates in the temperature was lowered to 158C and the process was
presence of high oxygen concentration. simulated for 50 more days. Subsequently, the temperature
The simulated biomass distribution in a granule grown at was lowered again to 88C followed by an additional
2 and 10 mg O2  L
1 and the oxygen penetration depths simulation of 65 days. Both simulated periods were equal to
during the cycle are compared in Figure 3a–f. The increased the duration of the long-term temperature change experi-
oxygen penetration depth during the total aeration period ments (De Kreuk et al., 2005b). Short-term temperature
(DO 10 mg/L) reduced the competition for oxygen between shift experiments were also simulated by changing the
autotrophs and PAO in the outer layer and created the temperature to 15 and 88C during two cycles each. Especially
possibility for autotrophic organisms to accumulate in the the simulated biomass distribution in the long-term low
inner zones of the granule. This increased concentration of temperature experiments was of interest, since experiments
autotrophs enhanced the nitrification rates at higher oxygen suggested an enrichment of autotrophic organisms in deeper
concentrations (Table I), however, because an anoxic zone layers of the granules at low temperatures. Simulations

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 805

Biotechnology and Bioengineering. DOI 10.1002/bit
Figure 3. Distribution of biomass concentration in the granules (a, d, g), the absolute quantity of biomass at different positions inside the granules (b, e, h) and the oxygen
profile inside the granules at different moments in a cycle, cycle times (minutes) between brackets (c, f, i). Oxygen concentration in the bulk liquid was set at 2 mgL
1 (a, b, c)
and10 mgL
1 (d, e, f) with a temperature of 208C during 365 days simulation. The temperature set to 88C with an oxygen concentration maintained at 2.6 mgL
1 (DO 20% saturation)
during 80 days of simulation (g, h, i).

indicated that this accumulation of autotrophs was mainly ammonia consumption rate in the long-term simulation at
due to a higher oxygen penetration depth, caused by 88C compared to the short-term simulated shift. These rates
decreased activity in the outer layers of the granules. are similar to the ones found in the experiments (Table II).
The results of the simulations were largely according to Consistently, the N- and P-removal rates are significantly
those expected from the experiments (Table II). Figure 3g–h lower after short temperature shifts compared to the long-
shows an increased amount of ammonia oxidizing bacteria term shifts, both in experiments and in simulations.
in the granules simulated at 88C. This results in a higher However, it can be seen that the rates of 20 and 158C

Table II. Effect of temperature on the overall N- and P-removal rates in the SBR, during short- and long-term changes (standard case in italic).

Overall net conversion rates (mg-N  L
1  h
1, mg-P  L
1  h
1)

NHþ
4 consumption NO

2 production NO

3 production PO3

4 consumption

Temperature 208C (long-term operation and start for 15.9 (18.0) 1.1 (1.1) 10.0 (14.4) 142 (156)
short-term changes)
Temperature 158C (one cycle change) 13.6 (11.7) 1.0 (1.2) 9.6 (8.8) 115 (134)
Temperature 158C (long-term operation) 15.7 (16.8) 1.1 (2.1) 10.4 (13.6) 135 (151)
Temperature 88C (one cycle change) 9.8 (8.2) 0.8 (0.8) 7.4 (6.5) 84 (94)
Temperature 88C (long-term operation) 10.6 (13.6) 0.9 (0.7) 9.5 (11.9) 103 (109)

The short-term changes are simulated with the biomass distribution as developed at 208C. Rates are calculated from the slopes of the simulated and
experimentally determined concentration profiles in the reactor during a cycle. Experimental data are shown between parentheses (partly from De Kreuk et al.,
2005b).

806 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
long-term simulations do not differ largely. This shows that
the granules are adapting to a small temperature change if
the time is long enough. The PAO population increases
slightly at lower temperatures, leading to a faster phosphate
consumption rate during the long-term experiment at 88C
than during the short-term experiment.
Similar to the results of simulations with different oxygen
concentrations, enrichment of PAO and ammonia oxidizers
can be explained by the higher oxygen penetration depth
during the first period of the cycle. Increased oxygen
penetration at lower temperatures is due to decreased
activity of the organisms in the outer layers (Fig. 3i).

Influence of Granule Diameter on Nutrient

Removal Efficiency
In the laboratory scale experiments, it was found that the
average granule diameter fluctuated in time between 0.4 and
2 mm. This granule diameter is not controllable in current
systems, but experimental data strongly suggests that it is
correlated with nutrient removal (De Kreuk et al., 2005a and
Fig. 4 here). In order to evaluate how sensitive this model is
for granule size, spherical particles with diameters in the
range from 0.72 to 1.61 mm were used in the simulations. In
order to keep the total volume of the granules the same in all
simulations (0.43 L), the number of granules (NG) was
varied as well (NG between 197,000 and 2,190,000 particles).
Reactor operation for every diameter was simulated for
365 days, each time starting with initial conditions.
The N-removal followed in general the trends as found in
the experiments. However, the effect of granule diameter on
N-removal in the simulations was smaller than measured in
practice (Fig. 4a). Reading these figures, it should be noted
that, due to the cycles in the SBR, a relatively small difference
in conversion rate could result in much higher or lower Figure 4. N-removal (a) and P-removal efficiency (b) at different average
effluent concentrations and thus in a significant difference in granule diameters: measured data (*, De Kreuk et al., 2005a) and data from several
simulations (&).
removal efficiency. For example, at the simulated diameter
of 1.6 mm, an increased ammonia consumption rate
from the present 12–14 mg N  L
1  h
1 would lead to an
increased N-removal efficiency from 71% to 83%.
At small granule diameters, there is a large surface area to
Influence of Sludge Load on Nutrient Removal
biomass volume ratio. Therefore, oxygen penetrates
relatively deep into the granules. This leads to sufficient In practice wastewater treatment plants are confronted with
aerobic volume for a fast oxidation of ammonia. However, night and day fluctuations, dry weather and storm water
as soon as all ammonia and phosphate are removed from the influent, all leading to fluctuations in the sludge loading of
system, oxygen will diffuse through the whole granule and the aerobic granules. Also industrial wastewaters are often
denitrification will be inhibited. Therefore, at the same more concentrated than sewage. The model was further used
oxygen concentration overall N-removal efficiency will be to study the optimum COD loading for nutrient removal.
lower at smaller diameters. When the diameter is larger than Sludge loading was varied by changing the amount of
1.4 mm, both N- and P-removal efficiencies start decreasing. granules in the system (NG from 375,000 to 1,200,000),
At this granule size, the surface area becomes limiting for leading to a variation in sludge loading from 17.7 to 5.5 kg

1
oxygen transport and thus for the conversion processes. This COD  m
3 granules  d or loading rate of 0.8 to 2.5 kg

1
leads to a decreased ammonia oxidation rate and phosphate COD  m
3 reactor  d . Every sludge-loading rate was applied
uptake rate and to ammonia and phosphate in the effluent. for 365 days in simulations starting with the initial
Simulations and experiments reveal that the optimum conditions described in the simulation strategy section.
diameter for nutrient removal in the standard case operating From the simulations (Fig. 5), it can be observed
conditions will be between 1.2 and 1.4 mm. that the N-removal efficiency slightly increases

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 807

Biotechnology and Bioengineering. DOI 10.1002/bit
 Cycle Optimization; the Effect of an Extra
Denitrification Period
Given the fact that ammonia and phosphate are both fully
converted before the end of the SBR cycle and that nitrate
accumulates in the standard operation (Fig. 2), an extra
anoxic phase after the aeration period could increase the
total N-removal efficiency. In order to study this effect of a
denitrification period at the end of the cycle, one simulation
was performed with a 90 min aeration period, followed by
an anoxic period of 22 min, instead of the normal aeration
period of 172 min. Especially the denitrification capacity
with the remaining stored PHB was studied.
Since all phosphate is already consumed after the first
hour of aeration, the anoxic period did not negatively
influence the overall phosphate removal efficiency. It did,
however, increase the N-removal efficiency to 80%, by
significantly reducing nitrate concentration in the effluent
Figure 5. Influence of COD loading (in COD per m3 reactor volume per day) on (Fig. 6). This simulation showed that during the anoxic
the N-removal efficiency (^) and P-removal efficiency (*).
phase ammonia slightly increased due to biomass decay
processes. Of the total incoming ammonium, 17% was
discharged in the effluent as nitrate and 3% as ammonium.
with an increasing load, with an optimum of
During the simulation of optimized SBR cycle, only very
1.9 kgCOD  m
3  d
1, corresponding to a sludge loading

1 small changes were found in the pools of storage products,
rate of 13 kgCOD  m
3 granules  d . At higher sludge loading so the extra denitrification is not much affecting the PHB,
rates, the N-removal efficiency quickly deteriorates. At very
glycogen, poly-P or the PAO concentrations in the granules.
low sludge loading rates, ammonia is quickly oxidized to
Therefore, it can be concluded that an extra denitrification
nitrate. As soon as all ammonia is converted, the oxygen
period after the aerobic phase would be a logical step in the
uptake rate decreases, resulting in a fully aerobic granule and
process optimization.
thus in limited denitrification during the rest of the cycle.
The higher the sludge loading, the longer it takes to consume
all ammonia and thus the anoxic zones in the granule exist
Discussion
during a longer part of the cycle. This results in more
N-removal via denitrification at high sludge loading In general, the mathematical model developed here
rates. However, at sludge loading rates higher than described the experimental data well. However, an absolute
1.9 kg  m
3  d
1, the amount of ammonium oxidizers is prediction of the process with this model is difficult.
too low relative to the incoming ammonium, resulting in Scenario analysis clearly showed that a small decrease in a
ammonium in the effluent. This leads to a fast decrease in
the N-removal efficiency and accumulation of ammonium
in the system.
Phosphate removal efficiency shows the opposite trend
of the ammonium removal. At loading rates lower than
1.3kgCOD  m
3  d
1 (corresponding to 9 kg  m
3
1
granules  d ),
the phosphate removal efficiency decreases. This can be
explained by the fact that a decreased COD loading rate
reduces the concentration of PHB accumulated in the granules
(in the anaerobic phase). In turn, less PHB causes slower
biomass growth rates and in consequence an increasing sludge
age. For example, the solid retention time (SRT) or

1
sludge age at sludge loading rates 11 and 9 kg  m
3 granules  d
are, respectively, 20 and 38 days. Since phosphate can
only be removed with the biomass in the effluent, a long
SRT results in an insufficient phosphate removal from the
system. As the poly-P concentrations in the biomass increase
to high values, the rates of poly-P storage will become too
low to remove all phosphate from the influent, resulting Figure 6. Simulated effect of an extra denitrification period at the end of the SBR
cycle on the concentration profiles of N compounds (dotted lines: the standard cycle
finally in accumulation of phosphate in the reactor and in the operation, solid lines: optimised cycle).
effluent.

808 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
conversion rate can easily lead to an increased concentration granular sludge has its optimum at 600–700 m2/m3 reactor
of that component in the effluent. Because the model (corresponding to a granule diameter between 1.2 and 1.4
is based on a SBR system, this component will mm). Higher granule diameters lead to a permanent
accumulate during the next cycles, leading to effluent data anaerobic core inside the granules and not enough
incomparable to experimental results. However, the trends autotrophic biomass in the outer aerobic layers resulting
of the conversion rates are comparable to practice, so the in ammonia in the effluent. The duration of an anaerobic
model can be used to obtain insight in the process rates and core existence within an SBR cycle also plays a crucial role in
in the granule structure. the N-removal at different biomass loading rates. The longer
The main reason for the model to be less predictive is the the anaerobic zone exists, the longer denitrification takes
complexity of the granular sludge itself. First of all, the place and the higher the N-removal efficiency. As in
granules in the model are of a certain diameter and are activated sludge systems with biological phosphate removal,
assumed to be perfect spheres. In reality, we are facing a wide the phosphate removal efficiency in aerobic granules is
range of granule sizes and shapes in a reactor (De Kreuk depending on the sludge age because poly-P leaves the
and Van Loosdrecht, 2004; Etterer and Wilderer, 2001; system incorporated in the biomass. By extrapolation, we
Toh et al., 2003). Less regularly shaped granules will lead to estimate that when the SRT exceeds 30 days the amount of
an increased surface to volume ratio, this influencing phosphate removed from the reactor with the biomass is less
accordingly the oxygen transport into the granules. As can than the incoming phosphate and phosphate will be found
be seen from the simulations, mass transfer (especially the in the effluent.
oxygen penetration in the granule) is decisive for the overall Optimization of nutrient removal efficiencies involves the
performance of the (simulated) system. It is possible to cycle optimization. As it becomes clear from the simula-
include several biofilm compartments with different granule tions, the oxygen penetration depth plays a major role. The
diameters in AQUASIM, in order to simulate a granule size oxygen concentration in the bulk liquid could be stepwise
distribution, but this will make the model implementation controlled via the oxygen uptake rate. If the oxygen uptake
and the interpretation of the data rather complex and will rate decreases, it means that oxygen penetration depth
drastically increase the needed calculation time. Diffusive increases. Lowering the oxygen concentration in the bulk
mass transport has never been a direct problem in ASMs, when the oxygen uptake rate decreases is expected to lead to
because in these models biomass is assumed in suspension in a better overall N-removal efficiency. A more simple
the bulk liquid. Consequently, nitrification and denitrifica- solution is the incorporation of denitrification steps in
tion occur in different compartments, rather than simulta- the cycle as shown by simulations. The model could be
neously in the same reactor due to oxygen limitation in useful to determine at which moment in the cycle the anoxic
aerobic granules. This increases the ability of ASMs to period would be most effective. Simulations suggest that at
predict effluent concentrations and overall system perfor- the end of the cycle the PHB pool needed for denitrification
mance compared with the aerobic granular sludge systems. is still sufficient, increasing the possibilities of timing this
When an aerobic granular sludge model must describe the denitrification period.
behavior of this system in practice, like the ASM models,
more system variables are needed due to the high complexity
of the wastewater composition. In practice, the COD in the Conclusions
wastewater will partly consist of suspended COD. This needs
to be hydrolyzed first, before organisms can use it for The presented model describes the performance of a
growth. Incorporating this substrate in the model implies laboratory scale SBR, fed with defined influent, capable of
the incorporation of heterotrophic biomass as well, since simultaneously removing COD, nitrogen and phosphate in
this relatively slow hydrolysis process will result in cell- one reactor. The model describes the experimental data
external substrate during the aerobic period and thus to from this complex system sufficiently well, as the simulated
heterotrophic growth (McSwain et al., 2004; Schwarzenbeck trends in nutrient removal were similar to those obtained
et al., 2005). However, the exact rates and methods of with laboratory experiments. Simulation results underline
conversion of suspended solids in aerobic granules are yet the importance of oxygen penetration depth into the
unknown, so more laboratory research will be needed first. granules, and thus the ratio of anoxic and aerobic biomass,
The granular sludge model was useful to obtain more for overall nutrient removal. The model can be used for
insight in the biomass distribution within the granules. process understanding and thus for optimization of the
The increased concentration of autotrophic organisms in nutrient removal efficiency. It could also form a starting
the outer layers of the granules and their shift to deeper point for models to design full-scale treatment plants.
layers when oxygen is available correspond to the hypothesis
made from analyzing the experimental results. The trends in This research was funded by the Dutch Foundation for Wat. Res.
nutrient removal at varying granule diameter were also (STOWA, TNW99.262) and STW (DPC5577) within the framework
of the ‘‘aerobic granule reactors’’ project. The project is in close co-
similar to the trends we found in practice. Simulation operation with DHV-Water in Leusden, The Netherlands. J. B. Xavier
indicated that in the standard operating conditions (DO was financially supported by the FCT/MCTES, Portugal, through the
concentration 2 mg/L) the surface to volume ratio of aerobic Grant SFRH/BPD/11485/2002.

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 809

Biotechnology and Bioengineering. DOI 10.1002/bit
 810
Appendix 1: Stoichiometry and Rates for the Bioconversion Processes Considered in the Model

SO2 SAc SNH4 SNO3 SNO2 SN2 SPO4 XPAO XPP XPHB XGLY XNH XNO XI

Process Component gO2/m3 gCOD/m3 gN/m3 gN/m3 gN/m3 gN/m3 gP/m3 gCOD/m3 gP/m3 gCOD/m3 gCOD/m3 gCOD/m3 gCOD/m3 gCOD/m3 Rate

Phosphate accumulating
organisms (PAO)
Anaerobic
1 Storage of acetate
1 YPO4 YPO4
YPHB AN

ðYPHB
1Þ rSA

2 Maintenance 1
1 AN
rM
Aerobic
3 Consumption of PHB
(1
1/YPHBO)
iNBM/YPHBO
iPBM/YPHBO 1/YPHBO
1 O
rPHB
4 Storage of poly-P
1/YPPO iNBM/YPPO
(1
iPBM/YPPO)
1/YPPO 1 O
rPP
5 Glycogen formation 1
1/YGLYO iNBM/YGLYO iPBM/YGLYO
1/YGLYO 1 O
rGLY
6 Maintenance
1 iNBM iPBM
1 O
rM
Anoxic (NO3)
7 Consumption of PHB
1 NO

Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

iNBM =YPHBNO3
ð1
1=YPHBNO3 Þ=2:86
ð1
1=YPHBNO3 Þ=2:86 iPBM =YPHBNO3 1=YPHBNO3 rPHB3
8 Storage of poly-P iNBM =YPPNO3
1=ðYPPNO3  2:86Þ 1=ð2:86  YPPNO3 Þ
ð1
iPBM =YPPNO3 Þ
1=YPPNO3 1 NO3
rPP
9 Glycogen formation iNBM =YGLYNO3 ð1
1=YGLYNO3 Þ=2:86
ð1
1=YGLYNO3 Þ=2:86 iPBM =YGLYNO3 1=YGLYNO3 1 NO
rGLY3
10 Maintenance iNBM
1/2.86 1/2.86 iPBM
1 NO3
rM
Anoxic (NO2)
11 Consumption of PHB
iNBM =YPHBNO2
ð1
1=YPHBNO2 Þ=1:71 ð1
1=YPHBNO2 Þ=1:71
iPBM =YPHBNO2 1=YPHBNO2
1 NO
rPHB2
12 Storage of poly-P iNBM =YPPNO2
1=ðYPPNO2  1:71Þ 1=ðYPPNO2  1:71Þ
ð1
iPBM =YPPNO2 Þ
1=YPPNO2 1 NO2
rPP
13 Glycogen formation iNBM =YGLYNO2 ð1
1=YGLYNO2 Þ=1:71
ð1
1=YGLYNO2 Þ=1:71 iPBM =YGLYNO2 1=YGLYNO2 1 NO
rGLY2
14 Maintenance iNBM
1/1.71 1/1.71 iPBM
1 NO2
rM
Decay
15 Decay 1
fXI iNBM
iNXIfXI iPBM
iPXIfXI
1 fXI rD

Autotrophic organisms
Nitrification (XNH)
16 Growth
(3.43/YNH
1)
1/YNH
iNBM 1/(YNH) 1 O
rG;NH
17 Aerobic end. resp.
(1
fXI) iNBM
iNXIfXI
1 O
fXI rER;NH
18 Anoxic end. resp. iNBM
iNXIfXI
(
fXI)/2.86 (1
fXI)/2.86
1 NO 3
fXI rER;NH
Nitrification (XNO)
19 Growth
(1.14/YNO
1) 1/YNO
1/YNO
iNBM 1 O
rG;NO
20 Aerobic end. resp.
(1
fXI) iNBM
iNXIfXI
1 O
fXI rER;NO
21 Anoxic end. resp. iNBM
iNXIfXI
(1-fXI)/2.86 (1
fXI)/2.86
1 NO 3
fXI rER;NO

(Continued)

DOI 10.1002/bit
Appendix 1: (Continued )
Process Process rate
Phosphate accumulating organisms
Anaerobic
1 Storage of acetate AN
rSA fPHB;max
fPHB SAc
qs;max    XPAO
ðfPHB;max
fPHB Þ þ KPHBP SAc þ KAcP

2 Maintenance AN
rM KO2 ;P KNO3 ;P KNO2 ;P
mAn     XPAO
SO2 þ KO2 ;P SNO3 ;P þ KNO3 ;P SNO2 þ KNO2 ;P
Aerobic
3 Consumption of PHB O fPHB SPO4 SO2
rPHB
zO2  zNH4  kPHB     XPAO
fPHB þ Kf PHB;P SPO4 þ KPO4 SO2 þ KO2 ;P
4 Storage of poly-P O
rPP XPAO fPP;max
fPP SPO4 SO2
zO2   kPP     XPAO
XPP ðfPP;max
fPP Þ þ KPP;P SPO4 þ KPO4 ;P SO2 þ KO2 ;P
5 Glycogen formation O
rGLY XPHB fGLY;max
fPP SO2
zO2   kGLY    XPAO
XGLY ðfGLY;max
fGLY Þ þ KGLY;P SO2 þ KO2 ;P
6 Maintenance O
rM SO2
mO2   XPAO
SO2 þ KO2 ;P
Anoxic (NO3)
7 Consumption of PHB NO2 fPHB SPO4 SNO3 KO2 ;P
rPHB zNO3  zNH4  kPHB  hP;NO3      XPAO
fPHB þ Kf PHB;P SPO4 þ KPO4 SNO3 þ KNO3 ;P SO2 þ KO2 ;P
8 Storage of poly-P NO2
rPP XPAO fPP;max
fPP SPO4 SNO3 KO2 ;P
zNO3   kPP  hP;NO3      XPAO
XPP ðfPP;max
fPP Þ þ KPP;P SPO4 þ KPO4 ;P SNO3 þ KNO3 ;P SO2 þ KO2 ;P
9 Glycogen formation NO2
rGLY XPHB fGLY;max
fGLY SNO3 KO2 ;P
zNO3   kGLY  hP;NO3     XPAO
XGLY ðfGLY;max
fGLY Þ þ KGLY;P SNO3 þ KNO3 ;P SO2 þ KO2 ;P
10 Maintenance NO2
rM SNO3 KO2 ;P
zNO3  mNO3    XPAO
SNO3 þ KNO3 ;P SO2 þ KO2 ;P
Anoxic (NO2)
11 Consumption of PHB NO3 fPHB SPO4 SNO2 KO2 ;P
rPHB zNO2  zNH4  kPHB  hP;NO2      XPAO
fPHB þ Kf PHB;P SPO4 þ KPO4 SNO2 þ KNO2 ;P SO2 þ KO2 ;P
12 Storage of poly-P NO3
rPP XPAO fPP;max
fPP SPO4 SNO2 KO2 ;P
zNO2   kPP  hP;NO2      XPAO
XPP ðfPP;max
fPP Þ þ KPP;P SPO4 þ KPO4 ;P SNO2 þ KNO2 ;P SO2 þ KO2 ;P
13 Glycogen formation NO3
rGLY XPHB fGLY;max
fGLY SNO2 KO2 ;P
zNO2   kGLY  hP;NO2     XPAO
XGLY ðfGLY;max
fGLY Þ þ KGLY;P SNO2 þ KNO2 ;P SO2 þ KO2 ;P
14 Maintenance NO3
rM SNO2 KO2 ;P
zNO2  mNO2    XPAO
SNO2 þ KNO2 ;P SO2 þ KO2 ;P
Decay
15 Decay rD KPHB;P KGLY;P KAc;P
bPAO     XPAO
fPHB þ Kf PHB;P fGLY þ KGLY;P SAc þ KAc;P
Autotrophic organisms
Nitrification (XNH)
16 Growth O
rG;NH SNH4 SO2
zO2  zNH4  mNH    XNH
SNH4 þ KNH4 ;NH SO2 þ KO2 ;NH
17 Aerobic end. resp. O SO2
rER;NH zO2  bNH  SO  XNH
2
þKO2 ;NH
18 Anoxic end. resp. NO3
rER;NH SNO3 KO2 ;NH
zNO3  bNH  hNH;NO3    XNH
SNO3 þ KNO3 ;NH SO2 þ KO2 ;NH
Nitrification (XNO)
19 Growth O
rG;NO SNO2 SO2
zO2  zNO2  mNO    XNO
SNO2 þ KNO2 ;NO SO2 þ KO2 ;NO
20 Aerobic end. resp. O
rER;NO SO2
zO2  bNO   XNO
SO2 þ KO2 ;NO
21 Anoxic end. resp. NO3
rER;NO SNO3 KO2 ;NO
zNO3  bNO  hNO;NO3    XNH
SNO3 þ KNO3 ;NO SO2 þ KO2 ;NO

Note: the functions zi ¼ 1=ð1 þ e
200Si þ4 Þþ10
12 switch the reaction off for negative substrate concentrations Si.

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 811

Biotechnology and Bioengineering. DOI 10.1002/bit
Appendix 2: Stoichiometric Coefficients for the Bioconversion Processes Shown in Appendix 1

Symbol Definition Value Unit Reference

Composition factors
(XPAO,XNH, XNO)
iNBM Nitrogen content of biomass 0.07 gN  gCOD
1
BM (1)
iNXI Nitrogen content of inert particulate COD 0.02 gN  gCOD
1
XI (1)
iPBM Phosphorus content of biomass 0.02 gP  gCOD
1
BM (1)
iPXI Phosphorus content of inert particulate COD 0.01 gP  gCOD
1
XI (1)
Phosphate accumulating
organisms (XPAO)
fXI Fraction of inert COD generated in decay 0.1 gCODXI  gCOD
1
XPAO (1)
YPO4 Anaerobic yield for phosphate release 0.5 gP  gCOD
1
SAc (2)
YPHB Anaerobic yield for PHB formation 1.5 gCODPHB  gCOD
1
SAc (2)
YPHBO Aerobic yield for PHB degradation 1.39 gCODPHB  gCOD
1
XPAO (3)
YPPO Aerobic yield for poly-P formation 4.42 gP  gCOD
1
XPAO (3)

YGLYO Aerobic yield for glycogen formation 1.11 gCODGLY  gCOD
1

XPAO (3)
YPHBNO3 Anoxic yield for PHB degradation on nitrate 1.7 gCODPHB  gCOD
1
XPAO (3)
YPPNO3 Anoxic yield for poly-P formation on nitrate 3.02 gP  gCOD
1
XPAO (3)
YGLYNO3 Anoxic yield for glycogen formation on nitrate 1.18 gCODGLY  gCOD
1
XPAO (3)
YPHBNO2 Anoxic yield for PHB degradation on nitrite 1.7 gCODPHB  gCOD
1
XPAO (3)
YPPNO2 Anoxic yield for poly-P formation on nitrite 3.02 gP  gCOD
1
XPAO (3)

YGLYNO2 Anoxic yield for glycogen formation on nitrite 1.18 gCODGLY  gCOD
1
XPAO (3)
Autotrophic organisms: ammonium
oxidizers (XNH) and
nitrite oxidizers (XNO)
fXI Fraction of inert COD generated in 0.1 gCODXI  gCOD
1
XNH;NO (1)
endogenous respiration
YNH Yield for growth of ammonium oxidizers 0.15 gCODXNH  gN
1 (4)

1
YNO Yield for growth of nitrite oxidizers 0.041 gCODXNO  gN (4)

(1) Henze et al. (1999); (2) Smolders et al. (1994); (3) Murnleitner et al. (1997) (4) Wiesmann (1994).

812 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
Appendix 3: Kinetic Parameters for the Bioconversions Shown in Appendix 1

Symbol Definition Value Unit Ref.

Phosphate accumulating organisms (XPAO)
bPAO Rate constant for lysis and decay 0:4  euPAO ðT
20Þ 1/d (1)
fGLY,max Maximum ratio of stored glycogen and biomass 0.5 — (2)
fPHB,max Maximum ratio of stored PHB and biomass 0.8 — —
fPP,max Maximum ratio of stored Poly-P and biomass 0.65 — —

1
1
kGLY Glycogen formation rate 0:93  euPAO ðT
20Þ gCOD2GLY  gCODPHA  gCODXPAO  d
1 (2)
uPAO ðT
20Þ
1
1
kPHB PHB degradation rate 5:51  e gCOD  gCOD  d (2)
kPP Poly-P formation rate 0:45  euPAO ðT
20Þ gP2  gCOD
1  d
1 (3)
KAc,P Half-saturation coefficient for acetate 4 gCOD  m
3 (4)
KGLY,P Half-saturation coefficient for glycogen 0.01 gCOD  m
3 (2)
KNO2 ;P Half-saturation coefficient for nitrite 1 gN  m
3 —
KNO3 ;P Half-saturation coefficient for nitrate 1 gN  m
3 —
KO2 ;P Half-saturation coefficient for oxygen 0.2 gO2  m
3 (1)
KPHB;P Half-saturation coefficient for PHB 0.01 gCOD  m
3 (1)
KPO4 Half-saturation coefficient for phosphate for growth 0.001 gP  m
3 —
KPO4 ;P Half-saturation coefficient for phosphate for 1 gP  m
3 (2)
Poly-P formation
KPP,P Half-saturation coefficient for Poly-P 0.01 gP  m
3 (4)

1
mAn Anaerobic maintenance 0:05  e uPAO ðT
20Þ gP  gCOD  d
1 (3)
mNO2 Anoxic maintenance on nitrite 0:11  euPAO ðT
20Þ gCOD  gCOD
1  d
1 (3)
mNO3 Anoxic maintenance on nitrate 0:11  euPAO ðT
20Þ gCOD  gCOD
1  d
1 (3)
mO2 Aerobic maintenance on oxygen 0:06  euPAO ðT
20Þ gO2  gCOD
1  d
1 (3)
qs,max Maximum acetate consumption rate 8  euPAO ðT
20Þ gCOD  gCOD
1  d
1 (2)
zNH4 Switching function for ammonium 1=ð1 þ e
200  S NH4 þ4 Þ þ 10
12 — —
zNO2 Switching function for nitrite
200  S NO2 þ4
12 — —
1=ð1 þ e Þ þ 10
zNO3 Switching function for nitrate
200  S NO3 þ4
12 — —
1=ð1 þ e Þ þ 10
zO2 Switching function for oxygen
200  S O2 þ4
12 — —
1=ð1 þ e Þ þ 10
hP;NO2 Reduction factor under anoxic conditions (NO2) 0.3 — —
hP;NO3 Reduction factor under anoxic conditions (NO3) 0.3 — —
uPAO Temperature coefficient 0.063 — (8)
Autotrophic organisms: ammonium
oxidizers (XNH) and nitrite oxidizers (XNO)
bNH Lysis and decay rate coefficient (XNH) 0:1  euNH ðT
20Þ d
1 (6)
bNO Lysis and decay rate coefficient (XNO) u
0:06  e NO ðT
20Þ d
1 (6)
KNH4 ;NH Half-saturation coefficient for ammonium (XNH) 2.4 gN  m
3 (6)
KNO2 ;NO Half-saturation coefficient for nitrite (XNO) 0.238 gN  m
3 (5)
KNO3 ;NH Half-saturation coefficient for nitrate 1.0 gN  m
3 —
inhibition of decay (XNH)
KNO3 ;NO Half-saturation coefficient for nitrate inhibition 1.0 gN  m
3 —
of decay (XNO)
KO2 ;NH Half-saturation coefficient for oxygen (XNH) 0.3 gO2  m
3 (6)
KO2 ;NO Half-saturation coefficient for oxygen (XNO) 0.1 gO2  m
3 (5)
hNH;NO3 Reduction factor under anoxic conditions (XNH) 0.3 — —
hNO;NO3 Reduction factor under anoxic conditions (XNO) 0.5 — —
mNH,max Maximum specific growth rate (XNH) 0:4  euNH ðT
20Þ d
1 Half the value (6)
mNO,max Maximum specific growth rate (XNO) 1:1  e uNO ðT
20Þ d
1 (6)
uNH Temperature coefficient (XNH) 0.094 — (7)
uXO Temperature coefficient (XNO) 0.061 — (7)

(1) Henze et al. (1999); (2) Meijer (2004); (3) Murnleitner et al. (1997); (4) Henze et al. (1999); (5) Wijffels and Tramper (1995); (6) Wiesmann (1994);
(7) Hao et al. (2002a); (8) Henze et al. (2000).

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 813

Biotechnology and Bioengineering. DOI 10.1002/bit
Appendix 4: Other Parameters Used in the Simulations, With Values According to the Experimental Conditions
Symbol Definition Value Unit
D Diffusion coefficient in the granule 0.0002/water fraction m2  dS1
DAc Diffusion coefficient of acetate in the granule in the feeding phase 5D m2  dS1
NG Number of granulesa 600,000 —
rmax Maximum radius of the granulesa 0.0055 m
SAc,in Influent concentration acetate 396 gCOD  mS3
SNH4 ;in Influent concentration ammonia 50 gN  mS3
SPO4 ;in Influent concentration phosphate 20 gP  mS3
tcycle Total cycle timea 0.125 days
Vreactor Total reactor volume 0.003 m3
Vbiofilm Volume of the biofilm compartmentc 0.0015 m3
rX Density of the biomass 350,000 gCOD  m
3
rPHB Density of the storage polymers (PHB, glycogen and poly-P)b 1  108 gCOD  m
3
a
This was the situation in the basic simulation of 365 days. In other simulations these data can vary according to the description in the text.
b
The density of the storage polymers was chosen extremely high, since the volume taken up by the polymers is already included in the volume of biomass
(Beun et al., 2001).
c
The biofilm compartment in AQUASIM contains bulk liquid, biofilm water and biofilm biomass fractions.

References Henze M, Gujer W, Mino T, Matsuo T, Wentzel MC, Marais GvR, Van
Loosdrecht MCM. 1999. Activated Sludge Model No.2d, ASM2D.
Beun JJ, Hendriks A, van Loosdrecht MCM, Morgenroth M, Wilderer PA, Water Sci Technol 39(1):165–182.
Heijnen JJ. 1999. Aerobic granulation in a sequencing batch reactor. Henze M, Gujer W, Mino T, van Loosdrecht M. 2000. Activated sludge
Water Res 33(10):2283–2290. models; ASM1, ASM2, ASM2d and A SM3. London: IWA Publishing.
Beun JJ, Van Loosdrecht MCM, Heijnen JJ. 2001. N-removal in a granular 130 p.
sludge sequencing batch airlift reactor. Biotechnol Bioeng 75(1): Li XM, Yang GJ, Yang Q, Zeng GM, Liao DX, Hu MF, Wu YM. 2005.
82–92. Simultaneous phosphorus and nitrogen removal by aerobic granular
De Bruin LMM, De Kreuk MK, van der Roest HFR, Van Loosdrecht sludge in single SBR system. In: Bathe S, De Kreuk MK, Mc Swain BS,
MCM, Uijterlinde C. 2004. Aerobic granular sludge technology, alter- Schwarzenbeck N, editors. Aerobic granular sludge. London, UK:
native for activated sludge technology? Water Sci Technol 49(11–12): IWA. p 71–78.
1–9. Liu Y, Tay J-H. 2004. State of the art of biogranulation technology for
De Bruin LMM, Van der Roest HF, de Kreuk MK, Van Loosdrecht MCM. wastewater treatment. Biotechnol Adv 22(7):533–563.
2005. Promising results Pilot research aerobic granular sludge technol- Lübken M, Schwarzenbeck N, Wichern M, Wilderer PA. 2005. Modelling
ogy at WWTP Ede. In: Bathe S, de Kreuk MK, McSwain BS, Schwar- nutrient removal of an aerobic granular sludge lab-scale SBR using A
zenbeck N, editors. Aerobic Granular Sludge. London, UK: IWA SM3. In: Bathe S, De Kreuk MK, Mc Swain BS, Schwarzenbeck N,
Publishing. p 135–142. editors. Aerobic granular sludge. London, UK: IWA. p 103–110.
De Kreuk MK, Van Loosdrecht MCM. 2004. Selection of slow growing McSwain BS, Irvine RL, Wilderer PA. 2004. The effect of intermittent
organisms as a means for improving aerobic granular sludge stability. feeding on aerobic granule structure. Water Sci Technol 49(11):19–25.
Water Sci Technol 49(11–12):9–19. Meijer SCF. 2004. Theoretical and practical aspects of modelling activated
De Kreuk MK, Heijnen JJ, van Loosdrecht MCM. 2005a. Simultaneous sludge processes [PhD.]. Delft: Technical University. 204 p
COD, nitrogen and phosphate removal by aerobic granular sludge. Morgenroth E, Sherden T, van Loosdrecht MCM, Heijnen JJ, Wilderer PA.
Biotechnol Bioeng 90(6):761–769. 1997. Aerobic granular sludge in a sequencing batch reactor. Water Res
De Kreuk MK, Pronk M, Van Loosdrecht MCM. 2005b. Formation of 31(12):3191–3194.
aerobic granules and conversion processes in an aerobic granular sludge Murnleitner E, Kuba T, van Loosdrecht MCM, Heijnen JJ. 1997.
reactor at moderate and low temperatures. Water Res 39(18):4476– An integrated metabolic model for the aerobic and denitrifying bio-
4484. logical phosphorus removal. Biotechnol Bioeng 54(5):434–450.
Dulekgurgen E, Yesiladali K, Ovez S, Tamerler C, Artan N, Orhon D. 2003. Reichert P. 1998. AQUASIM 2.0—Computer program for the Identifica-
Conventional morphological and functional evaluation of the micro- tion and Simulation of Aquatic Systems. Version 2.0. Dubendorf:
bial populations in a sequencing batch reactor performing EBPR. EAWAG.
J Environ Sci Health Part a-Toxic/Hazardous Substances & Environ Schwarzenbeck N, Borges JM, Wilderer PA. 2005. Treatment of dairy
Eng 38(8):1499–1515. effluents in an aerobic granular sludge sequencing batch reactor. Appl
Etterer T, Wilderer PA. 2001. Generation and properties of aerobic granular Microbiol Biotechnol 66(6):711–718.
sludge. Water Sci Technol 43(3):19–26. Smolders GJF, van der Meij J, Van Loosdrecht MCM, Heijnen JJ. 1994.
Hao X, van Loosdrecht MCM, Meijer SCF, Heijnen JJ, Qian Y. 2001. Model- Model of the anaerobic metabolism of the biological phosphorus
based evaluation of denitrifying P removal in a two-sludge system. removal process; stoichiometry and pH influence. Biotechnol Bioeng
J Environ Eng 127(2):112–118. 43:461–470.
Hao X, Heijnen JJ, Van Loosdrecht MCM. 2002aModel-based Tay JH, Liu QS, Liu Y. 2002. Aerobic granulation in sequential sludge
evaluation of temperature and inflow variations on a partial nitrifica- blanket reactor. Water Sci Technol 46(4–5):13–18.
tion ANAMMOX biofilm process. Water Research 36(19):4839– Toh SK, Tay JH, Moy BYP, Ivanov V, Tay STL. 2003. Size-effect on the
4849. physical characteristics of the aerobic granule in a SBR. Appl Microbiol
Hao XD, Heijnen JJ, van Loosdrecht MCM. 2002b. Sensitivity analysis Biotechnol 60(6):687–695.
of a biofilm model describing a one-stage completely autotrophic van Veldhuizen HM, van Loosdrecht MCM, Heijnen JJ. 1999. Modelling
nitrogen removal (CANON) process. Biotechnol Bioeng 77(3):266– biological phosphorus and nitrogen removal in a full scale activated
277. sludge process. Water Res 33(16):3459–3468.

814 Biotechnology and Bioengineering, Vol. 97, No. 4, July 1, 2007

DOI 10.1002/bit
Wanner O, Gujer W. 1986. A multispecies biofilm model. Biotechnol Wijffels RH, Tramper J. 1995. Nitrification by immobilized cells. Enzyme
Bioeng 28(3):314–328. and Microbial Technology 17(6):482–492.
Wentzel MC, Ekama GA, Loewenthal RE, Dold PL, Marais GR. Xavier JB, de Kreuk MK, Picioreanu C, Van Loosdrecht MCM. Submitted.
1989. Enhanced polyphosphate organism cultures in activated Microbial populations dynamics and nutrient conversions in aerobic
sludge systems. Part II: Experimental behavior. Water SA 15(2):71–88. granular sludge described by multi-scale modelling. Environ Sci
Wiesmann U. 1994. Biological nitrogen removal from wastewater. Adv Technol.
Biochem Eng/Biotechnol 51:113–154.

de Kreuk et al.: Kinetic Model of a Granular Sludge SBR 815

Biotechnology and Bioengineering. DOI 10.1002/bit