Available online at www.sciencedirect.

com

Bioresource Technology 99 (2008) 36653675

Anaerobic digestion model no. 1-based distributed parameter model

of an anaerobic reactor: I. Model development
S.J. Mu a, Y. Zeng a, P. Wu a, S.J. Lou b, B. Tartakovsky
a

b,*

Institute of High Performance Computing, 1 Science Park Road, #01-01, The Capricorn 117528, Singapore
Biotechnology Research Institute, NRC, 6100 Royalmount Avenue, Montreal, Quebec, Canada H4P 2R2

Received 4 May 2006; received in revised form 16 July 2007; accepted 16 July 2007
Available online 18 September 2007

Abstract
This work presents a distributed parameter model of the anaerobic digestion process. The model is based on the Anaerobic digestion
model no. 1 (ADM1) and was developed to simulate anaerobic digestion process in high-rate reactors with signicant axial dispersion,
such as in upow anaerobic sludge bed (UASB) reactors. The model, which was named ADM1d, combines ADM1s kinetics of biomass
growth and substrate transformation with axial dispersion material balances. ADM1d uses a hyperbolic tangent function to describe biomass distribution within a one compartment model. A comparison of this approach with a two-compartment, sludge bed liquid above
the bed, model showed similar simulation results while the one-compartment model had less equations. A comparison of orthogonal collocation and nite dierence algorithms for numerical solution of ADM1d showed better stability of the nite dierence algorithm.
 2007 Elsevier Ltd. All rights reserved.
Keywords: Distributed parameter model; ADM1; UASB reactor; Anaerobic digestion

1. Introduction
Compared to its aerobic counterpart, anaerobic digestion of wastewater features advantages such as energy
recovery in the form of methane production, production
of low sludge volume, and a capacity for degradation of
high strength wastewater (Bernard et al., 2001). While
problems of instability are infrequent for anaerobic digestors treating primary sludge, high-rate upow anaerobic
sludge bed (UASB) reactors are highly sensitive to external
disturbances such as hydraulic or organic shock loads,
which lead to imbalances in the acidogenic and methanogenic populations. As a result, the anaerobic reactors treating high-strength wastewaters are prone to instability and
may be subject to reactor failures. These problems have
motivated a number of studies on anaerobic digestion
kinetics and the development of reliable reactor models

Corresponding author. Tel.: +1 514 496 2664; fax: +1 514 496 6265.
E-mail address: Boris.Tartakovsky@cnrc-nrc.gc.ca (B. Tartakovsky).

0960-8524/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2007.07.060

to study what-if operational scenarios for improving

process stability and enhancing process control strategies.
A comprehensive structured model, known as the Anaerobic digestion model no. 1 (ADM1), was proposed by the
International Water Association task group on anaerobic
digestion (Batstone et al., 2000a,b, 2002). ADM1 takes into
account the processes of disintegration of complex solids,
hydrolysis of particulate organic materials, substrate
degradation, and biomass growth and decay. Since it was
rst introduced in 2000, ADM1 has been widely used in
modeling of anaerobic processes (Batstone et al., 2005;
Blumensaat and Keller, 2005; Jeong et al., 2005).
While the development of ADM1 is a signicant step
towards comprehensive modeling of anaerobic digestion, it
uses the assumption of ideal mixing. However, UASB reactors are often operated at high volumetric organic loads and
low recirculation-to-inuent ratios. These operating conditions are suitable for the formation of biolm granules,
but they also lead to the existence of signicant substrate
and biomass gradients in the sludge bed, which have a
profound inuence on long-term reactor performance and

3666

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

Nomenclature

stability during shock loads (Sam-Soon et al., 1991;

Kalyuzhnyi et al., 1996; Torkian et al., 2003).
Several somewhat simplied distributed parameter models of the anaerobic digestion process have already been
proposed. In the studies by Kalyuzhnyi et al. (2006) and
Schoefs et al. (2004), relatively simple reaction kinetics
were used. Batstone et al. (2005) developed a distributed
parameter model by combining the ADM1 kinetics with
the Takacs clarier model (Takacs et al., 1991), which
approximates a UASB reactor using several layers, i.e.
reactor hydrodynamics was simplied. In contrast, this
study presents a comprehensive distributed parameter
model, which combines the biotransformation kinetics of
ADM1 with the axial dispersion transport model. The validation of this model with experimental data obtained in a
laboratory-scale UASB reactor is presented in a subsequent paper (Tartakovsky et al., 2007).
2. Modeling framework
2.1. Material balances

u
v
V, Vliq
Vgas
x
xat
xmax
xmin
xtotal
z
b
s
g

linear upow velocity, m day1

stoichiometric parameter
reactor and liquid volumes, m3
reactor gas volume, m3
suspended matter concentration, kg COD m3
maximal attainable biomass concentration,
kg COD m3
attainable concentration of microorganisms in
the sludge bed, kg COD m3
attainable concentration of microorganisms in
liquid above sludge bed, kg COD m3
total biomass concentration, kg COD m3
reactor axial position, m
positive constant in Eq. (11)
hydraulic retention time, day
dimensionless reactor axial position

Subscripts
ac
acetate
at
attainable
c
composites
in
inuent ow
max
maximal
min
minimal
out
euent ow
r
recirculation ow
s
sludge compartment

Off-gas
height
gas phase
effluent

sludge
zone

Hs

xmin

xmax
sludge

influent

Fig. 1. Schematics of an UASB reactor (left panel) and a qualitative

prole of maximal attainable biomass concentration (xat) obtained using a
hyperbolic tangent function in Eq. (11) (right panel).



oci
o
oci z; t
o
Di z; t

 ui z; tci z; t
oz
oz
oz
ot
ri z; t rT;i z; t

For an axially dispersed UASB reactor (Fig. 1a) the

material balance of each component ci in the liquid phase
takes the following form (Kalyuzhnyi et al., 2006):

liquid
zone
recirculation

D
H
Hs
N
OLR
pgas;H2 O
Patm
qgas
qin
qrec
qs
Q
r
rT
R
Rw
s
sgas
sion
t
T

regression coecients (Eqs. (12) and (13))

orthogonal collocation method matrices
reactor cross section area, m2
Bodenstein number
liquid phase component concentration, kg
COD m3
dispersion coecient, m2 day1
total reactor height, m
sludge bed height, m
number of internal collocation points
1
organic loading rate, kg COD m3
R day
partial pressure of water vapor, bar
total gas pressure, bar
biogas ow rate, m3 day1
inuent ow rate, m3 day1
external recirculation ow rate, m3 day1
bypass ow rate, m3 day1
total input ow, m3 day1
biotransformation rate, kg COD m3 day1
gas component transfer rate, kmol m3 d1
gas law constant, bar m3 kmol1 K1
sludge washout factor
soluble matter concentration, kg COD m3
gas concentration, kg COD m3
ion concentration, kg mole m3
time, day
operating temperature, K

bypass

a, b
A, B
Ar
Bo
c

where Di is the dispersion coecient, ui is the upow velocity, z is the axial position, t is the time, ri is the net trans-

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

formation rate of the ith component calculated as a sum of

specic kinetic rates multiplied by corresponding stoichiometric coecients, and rT,i is the transfer rate of gas components. The liquid phase components included in the
model are soluble organic matters (si), suspended particulate matters (xi), and ions (sion,i). The distributed parameter
model takes into account the same 33 liquid phase components as ADM1. A detailed description of liquid phase
components and transformation rates ri(z, t) can be found
elsewhere (Batstone et al., 2002).
The Danckwerts boundary conditions (Danckwerts,
1953) for the above liquid component material balance
equations are given in the following two equations:
Di z; t

oci
ui z; tciz0  ci;in ;
oz

oci
0;
oz

z0

zH

2
3

where ci,in is the inuent concentration of the ith component and H is the reactor height.
Because of biogas bubbles generated during the biodegradation of organic compounds in the liquid phase, gas
transfer occurs within the liquid phase while gas transfer
at the gasliquid interface at the top of the liquid can be
neglected. Material balances for the gases are therefore
position-dependent. The dispersion and convective transport of bubbles in the liquid are considered negligible in
comparison with the gas transfer rate. The following equation describes the concentration changes of gas-phase
components:
osgas;j z; t
1 o
rT;j z; t 
q z; tsgas;j z; t
ot
Ar oz gas

where j is the index of biogas components, which are H2,

CH4 and CO2. sgas denotes the concentration of the jth biogas compound, Ar is the reactor cross section, and qgas is
the volumetric ow rate of biogas, which is dened as
RT
qgas z; t
Ar
P atm  pgas;H2 O

Z H
rT;H2 z; t rT;CH4 z; t

rT;CO2 z; t dz

16
64
0
5
where Patm and pgas;H2 O denote the total gas pressure and
the partial pressure of water vapor, respectively; T is the
operating temperature; and R is the gas constant.
The boundary conditions of the gas components are
sgas,j = 0 at z = 0 and dsgas,j/dt = 0 at z = H.
The overall mass balances of the reactor shown in Fig. 1
are given below
qin qrec qs Q

qin ci;in qrec ci;r qs ci Qci;0

3667

where qin, qrec, and qs are ow rates of the inuent, recirculation, and bypass streams, respectively, Q is the total
input ow, ci,in, ci,r and ci,0 are the concentrations of the
ith component in the corresponding ow. The bypass ow
(qs) in Eq. (7) is expected to be dependent on factors such
as linear upow velocity and organic load (Singhal et al.,
1998; Garuti et al., 2004). However, in the absence of a
strict theory the model does not use any specic
dependence.
Overall, the distributed parameter model consists of 36
partial dierential equations (material balance equations
for 12 soluble matters, 12 particulate matters, 9 ions and
3 gases), 72 ordinary dierential equations (two boundary
conditions for 36 components) and 33 algebraic equations
(overall mass balances for 33 liquid components), which
will be solved by a numerical algorithm.
For a completely stirred tank reactor (CSTR), material
balances of liquid phase components are reduced to the following form (Batstone et al., 2002):
19
X
dV liq ci t
qin ci;in t  qout ci t V liq
vi;j rj
dt
j1

Also, the CSTR gas phase component balance is given by

V gas

dS gas;j
qgas sgas;j rT;j V liq
dt

where
qgas

r

RT
rT;CH4
T;H2

rT;CO2
V liq
P atm  pgas;H2 O
16
64

10

All notations of the CSTR model are the same as in Eqs.

(1)(7).
2.2. Biodegradation kinetics
Kinetic dependencies describing the biotransformation
of organic matter and the growth of microorganisms were
adopted from ADM1. Kinetic parameters and model
inputs are those used in ADM1 benchmark simulations
(Rosen and Jeppsson, 2002). ADM1 kinetics is briey
reviewed below, while a detailed description can be found
elsewhere (Batstone et al., 2002).
It is assumed that organic particulate polymers disintegrate into inert materials, carbohydrates, proteins and fats,
which are then hydrolyzed to sugars, amino acids and long
chain fatty acids (LCFA). Sugars and amino acids are fermented to generate propionate, butyrate, valerate, acetate
and hydrogen. Propionate, butyrate and valerate are further degraded to acetate and hydrogen. LCFA can be
directly degraded to acetate and hydrogen. Methane is produced by both the degradation of acetate and the reduction
of carbon dioxide by hydrogen.
The model considers seven microbial trophic groups.
Kinetic equations take into account the processes of

3668

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

microbial growth and biomass decay. Biomass growth

rates are proportional to degradation rates of organic
matter and are described by Monod-like dependencies.
Also, the kinetic equations take into account the inhibitive
eects of pH, hydrogen, ammonium, and LCFA.
2.3. Sludge bed modelling
Biomass granulation is a distinct feature of UASB reactors. Anaerobic biolm granules, which have diameters of
0.54 mm, are composed of a consortium of anaerobic
microorganisms as well as of inert organic materials. The
settling of the granules is aected by both the upward
liquid ow and the apparent settling velocities (Kalyuzhnyi
et al., 2006) and results in the formation of a sludge bed.
The liquid above the bed only contains ne suspended biomass particles. In this study, the sludge distribution in the
reactor is modeled by a hyperbolic tangent function, which
denes a maximal attainable biomass concentration, xat(z),
at each reactor position


xmax  xmin
 1  tanhbz  H s xmin
xat z
2


11

where xmax and xmin are the total attainable concentrations

of microorganisms in the sludge bed and the liquid above
the sludge bed, respectively; Hs is the sludge bed height,
and b is a positive constant. The resulting sludge distribution is qualitatively shown in Fig. 1b. Furthermore, sludge
bed parameters (xmax, xmin and Hs) are dependent on the
upow liquid velocity (u) and the organic loading rate
(OLR), i.e. xmax = f1(u, OLR), Hs = f2(u, OLR). In the absence of detailed experimental data and an appropriate theory, the following empirical linear dependences are
proposed:
xmax a0 a1 u a2 OLR
H s b0 b1 u b2 OLR

12
13

where a0a2 and b0b2 are the regression parameters. It is

worth noting that volatile suspended solid (VSS) concentrations in the sludge bed often decrease with increasing
upow velocity and that the sludge bed expands, i.e. it
can be expected that a1 < 0 and b1 > 0. Because increasing
the OLR results in a higher biogas production, the inuence of OLR is expected to be similar to that of upow
velocity. Eqs. (12) and (13) can only be used following a
parameter estimation procedure and its applicability is
limited to a narrow range of operating parameters. Furthermore, regression constants should be periodically reestimated.
The retention of biomass granules in the sludge bed is
described by multiplying the transport term (u) in Eq. (1)
by a washout factor (Rw). The washout factor also uses a
hyperbolic tangent function

14
Rw z tanhxtotal z  xat z 1=2
P
where xtotal z i xi z is the total biomass concentration
at reactor position z, xat(z) is dened by Eq. (11), and xi(z)
is the concentration of the ith microorganism
(i = 5, . . . , 11). Eq. (14) implies that if the total biomass
concentration exceeds the maximal attainable biomass concentration, then Rw = 1, and the transport term in Eq. (1) is
the same as for all other liquid phase components. Otherwise, Rw = 0 and the transport term in Eq. (1) will be equal
to zero, i.e. the biomass is retained in the reactor. This approach uses continuous functions, which are capable of
describing biomass accumulation within the sludge bed,
biomass washout in the liquid zone, and the existence of
a layer of ne biomass particles above the bed. Biomass
concentration proles obtained using Eqs. (11)(14) were
in good agreement with the sludge proles measured in
UASB reactors (Kalyuzhnyi et al., 1996; Torkian et al.,
2003).
By incorporating Eq. (14) into Eq. (1) the mass balance
of the microorganisms can be rewritten as follows:


oxi
o
oxi z; t
o
Dx;i z; t

 ux z; txi z; t rx;i z; t
oz
oz
oz
ot
15
where ux(z, t) = ui(z, t)Rw(z) is the upow biomass velocity.
The transport term of the above equation is extended as
follows:
d
dRw z
ux z; txi z; t ui z; txi z; t
dz
dz
dui z; t
Rw zxi z; t
dz
dxi z; t
16
ui z; tRw z
dz
This approach is essentially similar to other methods, such
as biomass distribution modeling using a fraction coecient a, which is dened as the ratio of bacteria in the liquid
to that of the entire reactor (Bernard et al., 2001). The a
can also be dened as the fraction of biomass transferred
by the liquid ow (Vavilin et al., 2003).
2.4. Numerical methods
For the sake of comparison, the distributed parameter
model described above was solved by two dierent numerical methods, namely the orthogonal collocation method
(OC) (Finlayson, 1980) and the nite-dierence method
(FD) (Rice and Do, 1995).
2.4.1. Orthogonal collocation method
For the OC method, the material balance (Eq. (1)), was
rewritten as follows:
"
#

 N 2
N 2
dci;j 1 1 X
dDi
ui X
Bj;k ci;k 
Aj;k ci;k f ui ; DRw;i


si Boi k1
dt
dg H 1 k1
ri;j rT;i;j
17

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

where si is the hydraulic retention time, s = uRw/H; Bo is

the Bodenstein number, Bo = D/uH; ci,j denotes the ith
liquid component (i = 1, . . . , 33) at the jth collocation point
(j = 1, . . . , N), and N is the number of internal collocation
points. A(N+2)(N+2) and B(N+2)(N+2) are the orthogonal
collocation matrices for rst and second order derivates,
respectively. The transport term f(ui, DRw,i) described
washout of microbial species, hence for the microorgandR
isms f ui ; DRw;i uxi dzw;i and f(u, DRw,i) = 0 for all other
components.
The boundary conditions dened in Eqs. (2) and (3) are
rewritten as follows:
N 2
1 X
A1;k ci;k ci;0  ci;1 0
Boi k1
N 2
X

AN 2;k ci;k 0

18
19

k1

The resulting set of dierential-algebraic equations (DAE)

consisted of 36 N ordinary dierential equations (ODEs)
and 36 2 algebraic equations. The ODEs represented the
materials balances of 36 compounds at the N internal collocation points, and the algebraic equations described the
boundary conditions. Since large dierences in the magnitudes of the bioreaction kinetic rates existed, the model system was highly sti (Pind et al., 2002). Therefore, a sti
equation solver ode23t of MATLAB? 7.01 (MathWorks
Inc., Natick, Massachusetts, USA) was used to solve Eqs.
(17)(19).
2.4.2. Finite dierence method
The nite dierence method, in contrast to the orthogonal collocation method, evaluates the derivative at a discrete point by using the information about discrete
variables close to that point rather than from all discrete
variables (Rice and Do, 1995). In this work, the central
nite dierence approximation with an error of order h2
was used for the internal nite dierence points. The entry
and exit boundary conditions were approximated by the
forward and the backward approximations, respectively.
By specifying N internal nite dierence points, i.e.
N + 1 equally spaced intervals, this method resulted in
36 N ordinary dierential equations (ODEs) and 36 2
algebraic equations. Similarly to the collocation method,
the model equations were solved using ode23t equation
solver.
3. Results and discussion
3.1. Model parameters and initial values
This study used model parameters and operating conditions employed in ADM1 benchmark simulations (Rosen
and Jeppsson, 2002) for comparison of the distributed
parameter and ADM1 (CSTR) models. The values of stoichiometric, biochemical, and physiochemical parameters of

3669

Table 1
Operating conditions in ADM1 benchmark simulations
Parameter
3

Vliq (m )
Vgas (m3)
qin (m3 d1)
Frecycle (m3 d1)
ba, dimensionless
a

Value
1400
100
70
280a
0.25a

Parameter

Value
2

Ar, reactor area (m )

T, operating temperature (K)
1
OLR (kg m3
R d )
Hs, sludge height (m)

93.33a
308.15
60
10a

Assumed for ADM1d simulations.

ADM1 used in the simulations are given in the Appendix

and the operating conditions are given in Table 1. The
ADM1d parameters used to calculate biomass prole in
Eqs. (11) and (14) were set at xmax = 15 kg COD m3,
xmin = 0.15 kg COD m3, and b = 0.2. The values of xmax
and xmin were chosen to approximate sludge concentration
in benchmark simulations. For the sake of simplicity, a
zero bypass ow was assumed in ADM1d simulations.
In benchmark simulations composite materials comprised 91% of total CODs, and acetate comprises 81% of
VFAs. Consequently, ADM1d and ADM1 were compared
by analyzing the concentrations of composites (xxc), acetate (sac), and methane. Meanwhile, there was no one prevailing microorganism in the consortium, hence the total
microorganism concentration was used.
At the startup of each simulation, steady-state outputs
of the ADM1 benchmark simulations were used as initial
values for ADM1. For ADM1d simulations the model
was rst integrated for a period of 200 days with an
OLR of 3 kg COD m3 d1, which is the same as in benchmark simulations, and the resulting steady state outputs
were used as initial values for all subsequent simulations.
In order to compare the performance of the two models,
OLR was varied from 3 to 4 kg COD m3 d1. The ratio
of each component concentration to the total inuent
COD concentration remained equal to the original ratio
in the benchmark simulation.
3.2. Comparison of numerical methods
The eciencies of orthogonal collocation and nite difference numerical methods for ADM1d solution were compared using the benchmark model parameters described in
the previous section and integrating the model for a period
of 20 days. Also, a volumetric (per reactor volume) OLR of
1
3 kg COD m3
and a recirculation rate of 280 m3 d1
R d
were used. Model outputs obtained at the end of each integration period were compared.
For the OC method, three runs were carried out by setting the number of internal points to 4, 8, and 16, respectively. Four runs were carried out using the nite
dierence (FD) method with 10, 20, 30, and 50 internal
points. The resulting spatial distributions of key state variables at the end of the integration period are shown in
Fig. 2. The simulations were performed with MATLAB
7.0.1 (Mathworks, USA) on a Pentium 4 PC with a

16

12

OC-04

OC-08

OC-16

FD-10

FD-20

FD-50

Acetate, kg COD m-3

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

Composites, kg COD m-3

3670

0.2

0.1

10

15

10

15

d
7.2

pH

Biomass, kg COD m-3

0.3

6.8
0

10

15

Position, m

10

15

Position, m

Fig. 2. Spatial distribution of composite materials (a), acetate (b), total biomass (c), and pH (d) calculated using OC (solid gray line) and FD (solid black
line) algorithms. OC calculations were carried out with 4, 8, and 16 internal points. FD calculations were carried out with 10, 20, and 50 internal points.

2.8 GHz CPU. The required CPU times for the OC method
were 27, 270, 367 and 1631 s for 4, 8, 12 and 16 internal
points, respectively. The required CPU times for the FD
algorithm were 112, 19, 448, and 1596 s for 10, 20, 30,
and 50 internal points, respectively.
Overall, the FD algorithm had better stability and it
required somewhat shorter CPU time. Fig. 2 shows that
for all four state variables the proles obtained using the
FD algorithm converged when the number of nite points
reached 20. Increasing N from 20 to 30 and then to 50 did
not show signicant dierences between the proles. Overall, N = 20 can be considered as an acceptable compromise
between numerical accuracy and computational time.
For the OC method, obvious dierences between N = 4
and N = 8 were observed for all state variables except xxc
(Fig. 2), indicating that N = 4 is insucient. Although
the proles obtained at N = 8 almost converged to those
obtained by the FD method, unstable results were obtained
for N = 16 (Fig. 2bd).
The numerical instability of the orthogonal collocation
method with a large number of collocation points can be
explained by the stiness and nonlinearity of the distributed parameter model. Rice and Do (1995) pointed out
that though the orthogonal collocation method was superior to the nite dierence method in some problems, the
latter algorithm was easily applicable to many problems
with complex geometries. In the collocation method, the
collocation points were non-uniformly distributed. Consequently, OC was less ecient in calculating fast changes
in ion and gas balances and in modeling biomass distribution, which exhibits a sharp concentration drop at the
interface of the sludge bed and liquid zones. Overall, the

nite dierence method outperformed the orthogonal collocation method in solving the distributed parameter model
of the UASB reactor. In the subsequent simulations, a 20point FD algorithm, which oered satisfactory numerical
accuracy within an acceptable computational time was
used.
3.3. Comparison of the one- and two-compartment models
An alternative to modeling sludge distribution using a
hyperbolic tangent function (Eq. (14)) within a single compartment model was to use a two-compartment model,
which distinguishes between the sludge bed and the liquid
zones of the reactor. To compare the two approaches, a
two compartment model was developed in which the sludge
bed and liquid zones were described by distributed parameter and ideal mixing submodels, respectively. For the sake
of simplicity, the sludge bed compartment was still
described by Eqs. (1)(7) and (11)(16), but the value of
xmin was increased to 3 kg COD m3 as compared to
0.15 kg COD m3 for a single compartment model. The
CSTR material balances of the bulk liquid above the sludge
compartment were those of ADM1 (Eqs. (8)(10)).
Both single and two-compartment models were solved
using the FD algorithm with N = 20. A period of 100 days
of reactor operation was simulated. In the simulations, the
OLR was maintained at 3 kg COD m3 d1 for the rst 20
days, then the OLR was changed to 4 kg COD m3 d1
and maintained at this level to the end of the integration
period. Fig. 3ac shows the resulting spatial distribution
of composite materials, acetate, and pH at 20 d. It can be
seen that the resulting proles are similar. As expected,

10

0
0

10

Acetate, kg COD m-3

0.3

15

0.2

0.1

0
0

10

7.3

Composites, kg COD m-3

15

3671

0
0.4

Acetate, kg COD m-3

Composites, kg COD m-3

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

25

50

75

e
0.3
0.2
0.1
0
0

15

25

50

75

7.2

100

100

pH

pH

7.2
7.1

7.1

7
0

10

15

Position, m

25

50

75

100

Time, day

Fig. 3. Comparison of spatial (ac) and temporal (df) proles of key state variables calculated using one-compartment (solid black line) and twocompartment (solid gray line) distributed parameter models. Spatial proles correspond to t = 20 day. Temporal proles show euent concentration of
composite materials (d), acetate (e), and pH in euent (f).

the outputs of the two-compartment model exhibited discontinuity at the interface of the sludge bed and the overlying liquid, while the single compartment model
provided continuous outputs. A comparison of time proles of euent concentrations also showed qualitatively
similar results (Fig. 3df), although the two-compartment
model predicted lower concentrations of composite materials and acetate (Fig. 3d and e). Overall, owing to a smaller
number of model equations and continuous concentration
proles, the single compartment model was a better choice
for reactor modeling.
3.4. ADM1 versus ADM1d
An important feature of ADM1d was its ability to simulate spatial distribution of substrates and microorganisms
in the reactor, as shown in Fig. 2 where all state variables
exhibited distinct gradients. Also, the model was capable
of at least qualitative modeling of the sludge bed, as shown
in Fig. 2c. A more detailed analysis of spatial proles of
state variables in Fig. 2 showed that the concentration of
composite materials monotonously decreased with increasing reactor height (Fig. 2a), while there was a peak of acetate at around 2 m from the reactor inlet (Fig. 2b).
Notably, the acetate peak coincided with a drop in pH

(Fig. 2d). These proles were consistent with the existence

of acidogenic and methanogenic zones in an UASB reactor. Because of limited mixing and a relatively high residence time, the disintegration and hydrolysis of
particulate substrates as well as the production of degradation intermediates, such as volatile fatty acids, mainly proceeded in the bottom part of the sludge bed. The
production of volatile fatty acids was followed by the methanization process, which was more intensive in the upper
part of the sludge bed. The existence of an acidogenic zone
in an UASB reactor has been observed experimentally
(Sam-Soon et al., 1991). Notably, the stratication of metabolic activities decreased with increasing mixing, in particular due to increasing external recirculation.
In order to compare the responses of ADM1 and
ADM1d to variations in organic loading rate (OLR) and
the inuent-to-recirculation ratio, a simulation of UASB
reactor operation at four dierent combinations of OLRs
and external recirculation rates was carried out. In the simulation, each set of operating conditions was maintained
for a period of 100 days to obtain steady state outputs,
i.e. a total simulation time of 400 days was used. At the
startup of simulation, steady state values of the benchmark
ADM1 simulation and the ADM1d simulation shown in
Fig. 3 were used as initial values. The OLR was set at

3672

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

3 kg COD m3 d1 for both models, which is the same as

in the benchmark simulations. The recirculation rate was
set at 70 m3 d1, which corresponded to an inuent-torecirculation ratio of 1 (phase #1, 0100 days). Then the
recirculation rate was increased to 280 m3 d1, while the
OLR was unchanged (phase #2, 100200 days). In phase
#3 (200300 days), the recirculation rate remained the
same as in phase #2, but the OLR was increased to
4 kg COD m3 d1. Finally, in phase #4 (300400 days),
the recirculation rate was increased from 280 m3 d1 to
560 m3 d1 resulting in a recirculation to inuent ratio of
8, while the OLR remained at 4 kg COD m3 d1. Because
of a high recirculation rate, this phase corresponded to
near ideal mixing and similar outputs of ADM1 and
ADM1d were anticipated.
A comparison of euent concentrations predicted by
ADM1 and ADM1d is shown in Fig. 4. In phase #1 the
upow velocity was low. As the result, ADM1d outputs

0.3

Acetate, kg COD m-3

Composites, kg m-3

for euent composite materials and acetate were lower

than those of ADM1 (Fig. 4a and b). Consequently,
ADM1d predicted slightly higher euent pH and methane
ow rate (Fig. 4c and d). When the recirculation rate was
increased from 70 up to 280 m3 d1 (phase #2), ADM1
outputs did not change because the ideal mixing model
was unable to reect changes in hydraulic mixing. On the
other hand, ADM1d outputs changed in response to
increasing recirculation rate. In particular, increased mixing resulted in higher euent concentrations of composite
materials and acetate as shown in Fig. 4a and b. This trend
was in agreement with results obtained in an UASB reactor
(Sam-Soon et al., 1991), an anaerobic lter (Yu et al., 2000)
and an anaerobic packed-bed bioreactor (Mshandete et al.,
2004).
The responses of both models to increased organic load
in phase #3 were similar. As expected, the euent concentration of composite materials increased, while a decrease

6
4
2

0.2

0.1

0
0
1400

200

400

100

200

300

400

200

300

400

d
7.3

1200

pH

CH4 flow, m3d-1

1000

800

7.1

6.9
0

100

200

300

400

100

Time, day

Time, day

7000

6000
5000
4000
3000
0

100

200

Time, day

300

400

Trophic groups, kg COD m-3

Total biomass, kg COD

Fig. 4. Dynamic simulations of an UASB reactor with external recirculation using ADM1 (solid gray lines) and ADM1d (solid black line) models. The
panels show euent concentration of composites (a) and acetate (b), as well as methane production (c) and euent pH (d).

acidogens
syntrophs
methanogens

3
2
1
0
0

10

15

Position, m

Fig. 5. Time proles of total biomass amount computed by ADM1 (solid gray line) and ADM1d (solid black line) models (a) and spatial distribution of
main trophic groups computed by ADM1d for t = 20 day (b).

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

of pH was predicted (Fig. 4a and d). At the same time, the

acetate concentration decreased after the transition period,
which can be attributed to biomass accumulation in the
reactor (Fig. 5a). An increase of the recirculation rate in
phase #4 resulted in a near ideal mixing and caused
ADM1d outputs to almost overlap with those of ADM1,
in particular for pH and methane outputs (Fig. 4c and
d). As in phase #2, the recirculation rate increase resulted
in higher predicted euent concentrations of composite
materials and acetate for ADM1d, while ADM1 outputs
remained the same.
The simulated concentrations of composites and acetate
in the reactor euent (Fig. 4a and b) of ADM1d were less
than those predicted by ADM1. This can be explained by
the fact that in ADM1 biodegradation rates were the same
at all reactor positions, while in ADM1d the rates were
position dependent. The Monod-like kinetic equations,
which were used to describe biotransformations in
ADM1d, resulted in higher degradation rates predicted at
the bottom of the sludge bed, where substrate concentrations were the highest. Consequently, an overall volumetric
rate of biodegradation was higher in ADM1d than in
ADM1 thus resulting in lower euent concentrations of
composite materials and acetate.
ADM1d provided a qualitatively correct description of
the external recirculation rate inuence on total biomass
concentration in the reactor. Each increase in the recirculation rate was followed by a decrease in total biomass quantity, while the consequence of increasing organic load was
an accumulation of biomass (Fig. 5a). An examination of
the distribution of microbial species in the reactor at day
20 (phase 1) showed similar trends for all trophic groups.
The ratio of fermentative microorganisms was slightly
higher at the bottom of the sludge bed, while the density
of methanogenic microorganisms was predicted to increase
close to the top of the sludge bed (Fig. 5b). These trends,
however, were not well pronounced.
ADM1d simulations suggested that an increase of the
rate of external recirculation above a certain value
decreased the removal eciency. At the same time, the
biogas ow rate was predicted to increase slightly (results
not shown) with increasing external recirculation because
of increased gas transfer due to higher liquid phase concentration of carbon dioxide and methane. This trend
agreed with the experimental results of Romli et al.
(1994) in a two phase upow anaerobic reactor and in
an upow packed-bed anaerobic reactor (Mshandete
et al., 2004). Romli et al. (1994) explained that the
increase in biogas production was due to increased stripping of dissolved CO2 from the liquid phase because of
improved mixing. Overall, the ideal mixing (ADM1)
model was unable to respond to changes in the type of
mixing such as changes of recirculation rate or upow
velocity. In contrast, ADM1d was able to reect the eect
of both changes in OLR and recirculation rate and

3673

showed axial concentration gradients. In addition,

ADM1d simulations suggested that an optimum recirculation rate existed and this parameter can be used in process
optimization and control.

4. Conclusions
High rate UASB-type reactors exhibit signicant gradients of composite materials, volatile fatty acids, pH and
volatile suspended solids over the height of the reactor,
which can be only described by a distributed parameter
model. In this study, Anaerobic Digestion Model No. 1
was used as a basis for the development of a comprehensive distributed parameter model, named ADM1d.
ADM1d used a hyperbolic tangent function to describe
biomass distribution within a one compartment model.
A comparison of this approach with a two-compartment
model, which consisted of a sludge bed and a liquid above
the bed compartments, showed similar simulation results
while the one-compartment model had less equations. It
should be noted, that the hyperbolic tangent model of
the sludge bed does not take into account physical processes of granule settling and washout, but provides a
nonlinear regression model of the experimentally measured sludge distribution. This regression model should
be substituted with more complex models (e.g. Kalyuzhnyi
et al., 2006) if some insight on granular sludge dynamics is
desired.
A comparison of orthogonal collocation and nite-difference algorithms for numerical solution of the distributed
parameter model showed that both algorithms can be used,
however FD algorithm showed better stability. A comparison of ADM1 and ADM1d outputs showed that ADM1d
was better suited for modeling anaerobic reactors with limited mixing and high organic load. ADM1d was able to
simulate the impacts of both organic loading and external
recirculation on process performance while ADM1, which
uses an assumption of ideal mixing, only simulated the
impact of organic loading. The capacity of ADM1d for
predicting spatial distribution of state variables can be very
useful in developing a systematic approach for the design
of UASB reactors.

Acknowledgements
This work was funded by NRC-A*STAR Collaborative
Research Program for BRI-NRC, Canada, and by Singapore-NRC Joint Research Programme (A*STAR Grant)
for IHPC, Singapore. The authors are thankful to D. Batstone, U. Jeppsson, and C. Rosen for providing the Matlab
code of ADM1 and to M.F. Manuel and D. Lyew for suggestions on improving the manuscript. This is NRC paper
#49082.

3674

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

Appendix. ADM1 parameters

Parameter
A. Stoichiometric
fsI,xc
fxI,xc
fch,xc
fpr,xc
fli,xc
Nxc
NI
Naa
Cxc
CsI
Cch
Cpr
Cli
CxI
Csu
Caa
ffa,li
Cfa
fh2,su
fbu,su

Value
parameters
0.1
0.2
0.2
0.2
0.3
0.0376/14
0.06/14
0.007
0.03
0.03
0.0313
0.03
0.022
0.03
0.0313
0.03
0.95
0.0217
0.19
0.13

B. Biochemical parameters
kdis
0.5
khyd,ch
10
khyd,pr
10
khyd,li
10
KS,IN
1E4
km,su
30
KS,su
0.5
pHUL,aa
5.5
pHLL,aa
4
km,aa
50
KS,aa
0.3
km,fa
6
KS,fa
0.4
KI,h2,fa
5E6
km,c4
20
KS,c4
0.2
KI,h2,c4
1E5
km,pro
13
C. Physiochemical parameters
R
0.08314
Kw
2.08E14
Ka,va
1.38E5
Ka,bu
1.50E5
Ka,pro
1.32E5
Ka,ac
1.74E5
Ka,co2
4.94E7
Ka,IN
1.11E9
kA,Bva
1E+8
kA,Bbu
1E+8
a

M = kmol m3.

Unit

Parameter

Value

Unit

kmol N (kg COD)1

kmol N (kg COD)1
kmol N (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1
kmol C (kg COD)1

kmol C (kg COD)1

fpro,su
fac,su
Nbac
Cbu
Cpro
Cac
Cbac
Ysu
fh2,aa
fva,aa
fbu,aa
fpro,aa
fac,aa
Cva
Yaa
Yfa
Yc4
Ypro
Cch4
Yac
Yh2

0.27
0.41
0.08/14
0.025
0.0268
0.0313
0.0313
0.1
0.06
0.23
0.26
0.05
0.4
0.024
0.08
0.06
0.06
0.04
0.0156
0.05
0.06

kmol C
kmol C
kmol C
kmol C
kmol C

kmol C

kmol C

d1
d1
d1
d1
M
d1
kg COD m3

KS,pro
KI,h2,pro
km,ac
KS,ac
KI,nh3
pHUL,ac
pHLL,ac
km,h2
KS,h2
pHUL,h2
pHLL,h2
kdec,Xsu
kdec,Xaa
kdec,Xfa
kdec,Xc4
kdec,Xpro
kdec,Xac

0.1
3.5E6
8
0.15
0.0018
7
6
35
7E6
6
5
0.02
0.02
0.02
0.02
0.02
0.02

kg COD m3
kg COD m3
d1
kg COD m3
Ma

kA,Bpro
kA,Bac
kA,Bco2
kA,BIN
pgas,h2o
Patm
kLa
KH,co2
KH,ch4
KH,h2

1E+8
1E+8
1E+8
1E+8
0.0557
1.013
200
0.0271
0.00116
7.38

d1
kg COD m3
d1
kg COD m3
kg COD m3
d1
kg COD m3
kg COD m3
d1
bar M1 K1
M
M
M
M
M
M
M
M1 d1
M1 d1

(kg COD)1
(kg COD)1
(kg COD)1
(kg COD)1
(kg COD)1

(kg COD)1

(kg COD)1

d1
kg COD m3
d1
d1
d1
d1
d1
d1

M1 d1
M1 d1
M1 d1
M1 d1
bar
bar
M bar1
M bar1
M bar1
M bar1

S.J. Mu et al. / Bioresource Technology 99 (2008) 36653675

References
Batstone, D.J., Keller, J., Newell, R.B., Newland, M., 2000a. Modelling
anaerobic degradation of complex wastewater. I: model development.
Bioresour. Technol. 75, 6774.
Batstone, D.J., Keller, J., Newell, R.B., Newland, M., 2000b. Modelling
anaerobic degradation of complex wastewater. II: parameter estimation and validation using slaughterhouse euent. Bioresour. Technol.
75, 7585.
Batstone, D.J., Keller, J., Angelidaki, I., Kalyuzhnyi, S.V., Pavlostathis,
S.G., Rozzi, A., Sanders, W.T.M., Siegrist, H., Vavilin, V., 2002.
Anaerobic Digestion Model No. 1 (ADM1). IWA Publishing, London,
UK.
Batstone, D.J., Gernaey, K.V., Steyer, J.-P., Schmidt, J.E., 2005. A
particle model for UASB reactors using the Takacs Clarier Model.
In: Proceedings of the First International Workshop on the IWA
Anaerobic Digestion Model No. 1, Copenhagen, Denmark, pp. 129
136.
Bernard, O., Hadj-Sadok, Z., Dochain, D., Genovesi, A., Steyer, J.-P.,
2001. Dynamical model development and parameter identication for
an anaerobic wastewater treatment process. Biotechnol. Bioeng. 75,
424438.
Blumensaat, F., Keller, J., 2005. Modelling of two-stage anaerobic
digestion using the IWA anaerobic digestion model no. 1 (ADM1).
Water Res. 39, 171183.
Danckwerts, P.V., 1953. Continuous ow systems. Chem. Eng. Sci. 2, 1
13.
Finlayson, B.A., 1980. Nonlinear Analysis in Chemical Engineering.
McGraw-Hill, New York.
Garuti, G., Leo, G., Pirozzim, F., 2004. Experiments and modeling on
biomass transport inside upow sludge blanket reactors intermittently
fed. Water S.A. 30, 97106.
Jeong, H.-S., Suh, C.-W., Lim, J.-L., Shin, H.-S., 2005. Analysis and
application of ADM1 for anaerobic methane production. Bioprocess
Biosyst. Eng. 27, 8189.
Kalyuzhnyi, S.V., Sklyar, V.I., Davlyatshina, M.A., Parshina, S.N.,
Simankova, M.V., Kostrikina, N.A., Nozhevnikova, A.N., 1996.
Organic removal and microbiological features of UASB-reactor under
various organic loading rates. Bioresour. Technol. 55, 4754.
Kalyuzhnyi, S.V., Fedorovich, V.V., Lens, P., 2006. Dispersed plug ow
model for upow anaerobic sludge bed reactors with focus on granular
sludge dynamics. J. Ind. Microbiol. Biotechnol. 22, 221237.

3675

Mshandete, A., Murto, M., Kivaisi, A.K., Rubindamayugi, M.S.T.,

Mattiasson, B., 2004. Inuence of recirculation ow rate on the
performance of anaerobic packed-bed bioreactors treating potatowaste leachate. Environ. Technol. 25, 929936.
Pind, P.F., Angelidaki, I., Ahring, B.K., Stamatelatou, K., Lyberatos, G.,
2002. Monitoring and control of anaerobic reactors. Adv. Biochem.
Eng./Biotechnol. 82, 135182.
Rice, R.G., Do, D., 1995. Applied Mathematics and Modeling for
Chemical Engineers. John Wiley & Sons.
Romli, M., Greeneld, P.F., Lee, P.L., 1994. Eect of recycle on a twophase high-rate anaerobic wastewater treatment system. Water Res.
28, 475482.
Rosen, C., Jeppsson, U., 2002. Anaerobic COST benchmark model
description. Department of Industrial Electrical Engineering and
Automation, Lund University, Lund, Sweden.
Sam-Soon, P., Loewenthal, R.E., Wentzel, M.C., Moosbrugger, R.E.,
1991. Eects of a recycle in upow anaerobic sludge bed (UASB)
systems. Water S.A. 17, 3745.
Schoefs, O., Dochain, D., Fibrianto, H., Steyer, J.P., 2004. Modeling and
identication of a partial dierential equation model for an anaerobic
wastewater treatment process. In: Proceedings of the 10th World
Congress on Anaerobic Digestion, Montreal, Canada, vol. 1, pp. 343
347.
Singhal, A., Gomes, J., Praveen, V.V., Ramachandran, K.B., 1998. Axial
dispersion model for upow anaerobic sludge blanket reactors.
Biotechnol. Prog. 14, 645648.
Takacs, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of the
clarication-thickening process. Water Res. 25, 12631271.
Tartakovsky, B., Mu, S.J., Zeng, Y., Lou, S.J., Guiot, S.R., Wu, P., 2007.
Anaerobic digestion model no 1-based distributed parameter model of
an anaerobic reactor: II. model validation. Bioresour. Technol. 99,
36763684.
Torkian, A., Eqbali, A., Hashemian, S.J., 2003. The eect of organic
loading rate on the performance of UASB reactor treating slaughterhouse euent. Resour. Conserv. Rec. 40, 111.
Vavilin, V.V., Rytov, S.V., Lokshina, S.Y., Pavlostathis, S.G., Barlaz,
M.A., 2003. Distributed model of solid waste anaerobic digestion:
eects of leachate recirculation and pH adjustment. Biotechnol.
Bioeng. 81, 6673.
Yu, H., Wilson, F., Tay, J.H., 2000. Prediction of the eect of
recirculation on the euent quality of anaerobic lters by empirical
models. Water Environ. Res. 72, 217224.