A Kinetic Model for Anaerobic Digestion of

Biological Sludge

Spyros G. Pavlostathis and James M . Gossett

School of Civil and Environmental Engineering, Cornell University, lthaca,
New York 14853
Accepted for publication November 25, 1985

The principal objective of this study was the development

and evaluation of a comprehensive kinetic model capable
of predicting digester performance when fed biological
sludge. Preliminary conversion mechanisms such as cell
death, lysis, and hydrolysis responsible for rendering viable biological sludge organisms to available substrate
were studied in depth. The results of this study indicate
that hydrolysis of the dead, particulate biomass-primarily consisting of protein-is the slowest step, and
therefore kinetically controls the overall process of anaerobic digestion of biological sludge. A kinetic model
was developed which could accurately describe digester
performance and predict effluent quality.

the possible preliminary conversion mechanisms may

be hydrolysis and not cell death or lysis.
The purpose of this article is to examine directly the
preliminary conversion step in detail, considering
mechanisms such as cell death, lysis and hydrolysis,
so as to evaluate their relative importance in the anaerobic digestion of biological sludges. This is done in
the context of a conceptual model for biological sludge
digestion, which is here presented and evaluated.

MODEL DEVELOPMENT
INTRODUCTION

Conceptual Model of Biological Sludge Digestion

Previous studies,* dealing with the anaerobic digestion of activated sludge (AS) have shown that kinetic
models which assume that methanogenesis is the sole
rate-limiting step are inapplicable in the case of biological sludges. Anaerobic degradation of AS requires
that the potentially-degradable portion of viable AS
organisms must first be converted to available substrate by preliminary conversion mechanisms such as
cell death, lysis, and/or hydrolysis. It was shown2 that
for solids retention times (@IG) values of practical interest, acidogenesis and methanogenesis are not process-controlling steps in digesters receiving biological
sludges. The predominant degradable constituent in
the effluents of such digesters is particulate protein,
indicating that hydrolysis and/or other preliminary
conversion mechanism(s) is (are) Limiting from the point
of view of substrate availability.
Which of the postulated, preliminary-conversion
mechanisms is most kinetically limiting can only be
inferred from these previous investigations. Their direct study was not undertaken. Examination of continuous-flow anaerobic digestion data2 from digesters
which were fed autoclaved sludge showed that a significant preliminary barrier remains, even after autoclaving. Since autoclaving induces cell death and lysis,
these data suggest that the most kinetically limiting of

A conceptual representation of the processes involved in the anaerobic digestion of AS is shown in

Figure 1. Gossett and Belser demonstrated that the
biodegradable fraction of AS consists almost exclusively of the biodegradable portion of viable, activated
sludge organisms. The biodegradable portion of viable
AS cells may be thought of as comprising two substrate
pools: one particulate and one soluble. Upon death,
the cell membrane ruptures (i.e. lysis occurs), and part
of the intracellular, soluble, degradable COD (i.e. soluble BOD) (presumably the lowest molecular weight
components) is immediately released. Then, the remaining intracellular soluble BOD pool of a dead cell
changes with time because of soluble BOD production
via intraceliular hydrolysis (IH) and loss of soluble
BOD due to diffusion. Much of the IH may be due to
remaining, native intracellular enzymes, uncontrolled
upon cell death. Contributions to IH from exoenzymes
produced by the anaerobic microflora may also be supposed. The dead cell particulate BOD is additionally
subject to extracellular hydrolysis (EH), presumed to
be primarily induced by the active digester microorganisms. The extracellular soluble BOD pool is subject
to decrease because of uptake by the digester microflora, especially by the acid-forming bacteria. From
that point on the model follows the classical two-phase

Biotechnology and Bioengineering, Vol. XXVIII, Pp. 1519-1530 (1986)

0 1986 John Wiley & Sons, Inc.

CCC 0006-3592/86/101519-12$04.00

EXTRACELLULAR
ACIW~ESIS

M*

. co,

cell membrane, although ruptured, still provides a barrier. Whether or not the damaged membrane provides
a significant barrier to release depends on the relative
size of the membrane perforations compared with the
size of the material leaking out. Evidence is presented
in a later section which suggests that diffusion from
within a damaged cell membrane is not a significant,
kinetic limitation with respect to substrate release.
If diffusion limitations are ignored, then the intraand extracellular hydrolysis rates are additive and may
be collectively referred to as hydrolysis. Accordingly, the conceptual model presented in Figure 1 can
be simplified as is shown in Figure 2.
Equations

Figure 1. Conceptual model for anaerobic digestion of biological

solids (I.H. is intracellular hydrolysis; E.H. is extracellular hydrolysis; kd is the death rate coefficient; kh, and kh, are the intra- and
extracellular hydrolysis rate coefficients; k , is the soluble BOD diffusion rate coefficient; and y is the cell soluble BOD immediately
released, fraction of total BOD).

anaerobic digestion scheme: acidogenesis and methanogenesis.

Validation of the above-outlined model requires the
direct estimation of all constants and parameters involved. Due to the complexity of the system, difficulties are encountered in developing analytical techniques to accomplish this without simplification. For
example, note that the conceptual model equates death
and lysis. From the point of view of substrate availability, cell lysis is more important than death. However, lysis is difficult to assay, compared with the quantification of death. Certainly, all cells which lyse are
dead. If death results from predation by anaerobic protozoa, lysis may be said to be pretty much concurrent with death. But for other, nonlytic causes of death,
how soon does lysis follow death?
Attempts were made in the present study to measure
cell lysis rate, but they were not successful, as is discussed in a later section. On the other hand, it is believed that once the cell is dead, the permeability barrier provided by the plasma membrane is rapidly
destroyed, and most of the soluble cell contents (presumably the lowest molecular weight components) leak
out.3 Thus, from the point of view of substrate availability, we assume here that there is practically no lag
between death and lysis. Hence, we refer to both as a
singular death/lysis mechanism.
Upon cell death/lysis a good portion of the intracellular soluble BOD is believed to be released almost
immediately. The remaining intracellular soluble BODoriginally present in the viable cell and also produced
via intracellular hydrolysis of particulate BOD following cell death-must diffuse out of the dead cell if the
1520

The following assumptions and premises were considered in the derivation of model equations: 1) all
variables are expressed on a COD unit basis; in fact,
all except microorganism concentrations are expressed
as degradable COD; 2) the inherent, net biodegradable
fraction (fd) of viable AS organisms and anaerobic microflora is the same; 3) we do not differentiate between
cell death and lysis, but we assume that lysis occurs
soon after death; 4) upon death/lysis, all soluble intracellular material is released as soluble substrate,
and it is taken as a constant fraction of the total cell
degradable COD (i.e., diffusion of soluble material out
of the damaged cell is rapid-this assumption is justified later); and 5 ) facultative AS bacteria which may
ultimately survive are not considered as part of the
acid-phase, active biomass.
The death/lysis rate is considered to be first-order:

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986

dXfs
dt

r d = p -

-kdXfs

Dum/LysIs

k,

oRocyys

kh

SUBSTRATE
REQUIRING
HYDROLYSIS

where X t S is the viable AS biomass concentration in

the digester (g/L) and kd is the death/lysis rate coefficient (day-').
The hydrolysis rate is assumed to be a first-order
reaction with respect to the concentration of degradable particulate COD:

Mass Balance on Degradable Portion of Active

Acidogenic Biomass (fdX$)

erAX20

X 3
YAkASAX$

+ vfd[K:' c

SA

bAXt]

0 (8)

Mass Balance on Products

where F is the degradable, particulate (nonviable) COD

concentration (g/L) and kh is the hydrolysis rate coefficient (day-'). In terms of the complex conceptual
model (Fig. 1)-and given our assumption that there
is no diffusional limitation for transport of solubilized
material out of the damaged cell-then this kh is really
the sum of intracellular and extracellular hydrolysis
rate coefficients.
The microbial growth rate and substrate utilization
rate for any particular group of digester bacteria are
based on the Monod equation4:

dX
dt

rx = - -

dS
r"=-dt

Q(Po - PI +

YkSX
K, + S

bX

(4)

+S

where X is the active microorganism concentration (g/L);

S is the substrate concentration (g/L); Y is the microbial
growth yield coefficient (g biomass produced/g substrate utilized); k is the maximum specific substrate
utilization rate (day- '); K , is the half-velocity coefficient (g/L); and b is the microorganism decay coefficient (day - I ) .
For the case of a completely mixed, continuous-flow
reactor without recycle, at steady-state, the following
mass balances can be derived (definitions of terms
follow):

Mass Balance on Degradable Portion of Viable AS

Microorganisms (fdXts)
efd

(X:',"

Xts)

VfdkdX:'

VkASAX2 - VYAkASAX2
Kf + SA
Kf + SA

(3)

kSX
-~
K,

We may define (as did Eastman and Ferguson5) a

quantity, P , which is the concentration of volatile acids
formed by acidogenic bacteria, plus any influent volatile acids. Parameter P differs from the actual, observed concentration of volatile acids remaining, due
to their consumption by methanogenic bacteria. Thus,
P is the sum of observed volatile acids, plus methane
formed, plus methanogenic biomass formed, all in COD
units. The mass balance on P then becomes:

(5)

where Q is the hydraulic flow rate (L/day); V is the

reactor volume (L); X;p,S is the influent, viable AS biomass COD concentration (g/L); Fo is the influent degradable particulate (nonviable) COD concentration
(g/L); y is the fraction of total degradable COD in
X,"' released as "soluble" COD upon deathilysis; SS
is the influent soluble substrate COD to the reactor
(g/L); X i o is the influent active acidogenic microorganism concentration (g COD/L); Po is the influent product
concentration (g COD/L); and fd is the net biodegradable fraction of active biomass. (Note that superscript
AS refers to activated sludge and superscript A refers
to the acid phase).
The last mass balance was derived based on the
following reasoning: COD is conserved in an anaerobic
digestion system and the utilized substrate COD is either
incorporated into cellular material or converted to
products for microbial energy. Therefore, the difference between the substrate utilized and the net cellular
COD produced represents the amount of product
COD.^,^

Mass Balance on Nonviable Substrate Requiring

Hydrolysis (F)

Q(F0

F)

VkhF

V(I - 7 )fdkdx,"'

(6)

(7)

By adding the five steady-state equations and dividing by Q the following expression is obtained:

Mass Balance on Soluble Substrate for Acid

Phase (9')

Q(SS - S A ) + VkhF

VYfdk&fiS
-

kASAXi
Kf + SA

The hydraulic retention time ( 6 ) (equal to solids retention time for systems without biomass recycle) ap-

PAVLOSTATHIS AND GOSSETT: ANAEROBIC DIGESTION OF SLUDGE

1521

pearing in the equation is equal to V/Q. From eq. (8),

we obtain:

which substituted into Equation (10) gives:

fAX$rjj

XCS) + (Fa - F ) + ( S t - SA) + xto

X:[1 + (1 -fd)Pf9]
+ (Po - P ) = 0 (IOa)

The influent, viable AS biomass concentration is given

by the following equation:

X$rjj

D COD,,
fd

where D is the ultimate biodegradability of AS (fraction) and CODi, is the influent COD concentration (g/L).
An explicit solution for each variable can be found by
algebraic manipulation of the five steady-state equations:

substrate COD/g biomass COD day-); Xf is the total

methanogenic biomass (g COD/L); (Sg)effecis the effective influent volatile acid COD to the methane phase
(g/L); and the rest of the symbols are as defined in
previous sections.
The connection between the acid phase and methane
phase of the present model is made via the product
formed in the former phase, which is essentially the
influent to the last phase, i.e.,
= P. Since
the product equation includes a correction for acidogenic biomass produced in the system, the only correction of the processed COD for the digester performance equation [i.e., percent of COD destroyed
(%COD,,,,)] is the methanogenic biomass:

The four basic equations of the model [Eqs. (I3),

(14), ( I S ) , and (16)] can be linearized in order to de-

termine the kinetic constants involved. However, the

biggest analytical difficulty faced in the present study,
which was the inability to differentiate (and measure)
AS biomass from the acidogenic and methanogenic
biomass produced in the digester, made such a determination impossible. To overcome these problems, a
number of batch experiments were conducted, to independently provide important constants.

Model Equations

The model equations are as follows:

PROGRAM

-X:[1

+ (1

- fd)bAO]

(17)

In the last phase, i.e. methanogenesis, we make use

of the Monod model for substrate and biomass concentrations (Note that superscript B refers to methanogenesis). Ignoring X:a,

where SB is the effluent soluble COD (i.e., volatile acid

COD) (g/L); K! is the composite, half-velocity coefficient ( = Kyetic + Kyopionic fo r this case, g COD/L);
bB is the decay coefficient for methanogens (day- I);
YB is the yield coefficient for methanogens (g biomass
COD/g substrate COD utilized); kB is the maximum
specific substrate utilization rate for methanogens (g
1522

OF STUDY

Skeptics might note that there are sufficient kinetic

parameters in the complex model depicted in Figure
2, such that one could-through convenient selection of parameter values-fit almost any performance
data! Legitimate evaluation of the model requires the
independent measurement of parameters employed in
it. However, it is not necessary to do this for all parameters-only the important ones. For example, it
will later be shown through a sort of sensitivity analysis
that the important, limiting steps are those relating to
preliminary conversion of substrate to available, soluble form, and the later conversion of acetate and propionate to methane. Acidogenesis is relatively rapid.
This is to be expected, based upon previous
Thus, it makes little sense to do anything more than
to expropriate kinetic parameters for acidogenesis from
literature values. [The one exception was the yield
coefficient for the acidogenic bacteria ( YA),which was
measured, though with only modest s ~ c c e s s . ~ ]
Hence, many phases of study described in subsequent sections were undertaken with the purpose of
independently measuring sludge and kinetic parameters of importance to the proposed model. Some examples: the deathhysis rates of viable AS organisms
in anaerobic environments were measured in continuous flow systems, using an oxygen uptake technique;

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986

hydrolysis rates were inferred from solubilization studies employing batch-fed digesters which received differing amounts of active, anaerobic seed. Measurements of other important model parameter values (e.g.,
ultimate biodegradability of autoclaved and intact AS;
half-velocity coefficients for acetate and propionate
utilization) were presented previously.2In a final analysis phase of study, the various independently measured parameter values were employed with the simplified version of the model depicted in Figure 2, allowing
comparison of observed, continuous-flow digestion
performances with model predictions. For this comparison, digester performance data presented elsewhere were employed.*
MATERIALS AND METHODS

Reactor configurations and modes of operation for

the laboratory generation of activated sludge and its
subsequent anaerobic digestion were described previously.2 The AS was produced in a 150-L fill-and-draw
reactor fed with a soluble synthetic waste. The resulting sludge volatile suspended solids were thus entirely of biological origin. The digesters were constructed from 2-L Pyrex bottles fitted with holes and
rubber stoppers to allow for feeding, withdrawal of
digested sludge, mixing of the digester contents (mechanically), sampling of the mixed liquor and measurement of the gases. The digesters were fed on a
once-per-hour basis via a timer, a series of solenoid
valves and two metering pumps with variable-speed
drives. Digested sludge was withdrawn once o r twice
a day via a sampling port.
Batch Digestion Experiments

A variation of the serum bottle technique outlined

earlier2was used to investigate the hydrolysis (actually
solubilization) of AS in anaerobic environments. In this
case, seed and media were transferred anaerobically
to the serum bottles-already
containing the substrate-separately in order to permit the use of various
amounts of seed. Autoclaved AS solids were employed
as substrate. Autoclaving was chosen as a cell disruption technique.2 It was thought that autoclaved AS
solids would represent cell particulate matter left after
cell death and lysis and requiring further hydrolysis
(see the Hydrolysis Rate Coefficient section). All bottles received the same amount of defined media (20
mL) and the same amount of substrate (15 mL from a
stock suspension of ca. 20 g COD/L). The amounts of
seed used were 0, 20, and 40 mL of digester mixed
liquor receiving intact AS as feed. These samples are
later referred to as Bo,B , , and B2, respectively. Twelve
replicates were prepared; two were opened immediately after filling with media and/or seed (for assay of
initial conditions and substrate measurements), and the

rest were opened one-by-one each time that a measurement was effected. The incubation time was one
month and the room temperature was 35
1C. The
bottles were manually shaken twice per day.
On a predetermined schedule, the gas productions
were measured in all replicates, and then one bottle
from each group was opened and the following analyses were performed on its contents: pH, soluble COD,
and volatile acids. The gas composition was also measured. The particulate substrate COD was accurately
measured only at the beginning of the experiment. This
is the difference between total and soluble COD of the
composite sample (i.e., substrate + seed
media)
after subtracting seed and media blanks.
The autoclaved AS solids fraction was prepared
using the following technique: settled AS was autoclaved (121C, 30 min), cooled, then centrifuged (104g
for 1 h), and the centrate was wasted. The pellet was
resuspended in distilled water and centrifuged again
(lo4g for 30 min). The last step was repeated one more
time. Finally, the centrate was wasted and the pellet
was suspended in distilled water, yielding the solid
sample.

Cell Death/Lysis Rate(s)

An oxygen uptake rate (OUR) technique was used
for estimation of the death rate constant (kd) of AS
microorganisms fed to continuous flow digesters operated at different @IG. The decrease in OUR-measured in preaerated (30 min) digester mixed liquor under substrate saturation conditions (ca. 300 mg
glucose/L)-with incubation time (i.e. digester retention time) was assumed to represent microbial death.
The use of this OUR technique for kd estimation is
certainly questionable: The OUR is more a measure
of biological activity, and activity and viability are not
necessarily synonyms. It is conceivable that the OUR
of a biological community may decrease without an
equal decrease in viability (i.e., percent of viable organisms of the total). For example, in the case under
discussion, facultative bacteria originally in the AS,
now growing in an anaerobic environment, may not
exert high OUR immediately upon reintroduction to
the aerobic environment, although their viability may
not have been decreased. Another concern about using
the OUR technique with anaerobic media is the strictly
chemical oxygen consumption that reduced substances
may exert under aerobic conditions. In order to assay
the significance of this source of oxygen consumption,
ethanol (95%) was added to an aliquot of digester mixed
liquor, left to stand for 15 min, then aerated for 15 min
prior to measurement of OUR. The OUR was practically zero over 25 min of measurement. This led us to
conclude that, although there is a chemical oxygen
consumption exerted by various reduced substances
in the digester mixed liquor, this rate is almost negli-

PAVLOSTATHIS AND GOSSElT: ANAEROBIC DIGESTION OF SLUDGE

1523

gible over the time that the OUR of a sample is measured (usually 20 min).
Other direct viability measurement techniques (e.g.
plate counts) were judged inappropriate primarily because of the inability to completely disperse an activated sludge floc without killing the bacteria. Plate
counts are appropriate for use where viable numbers
range over orders of magnitude. However, their accuracy is too limited to be of use in our studies. Thus,
despite its limitations, the OUR technique was selected
for monitoring relative changes in AS viability.
Initially, we intended to distinguish death from lysis,
and to separately determine death and lysis rates for
AS organisms in anaerobic environments. As a mechanism for converting membrane-enclosed, potential
substrate into available substrate, cell lysis is perhaps
more important than cell death. Attempts were made
to directly monitor cell lysis via ultraviolet (UV) absorbance (at 260 nm) of soluble material produced during batch anaerobic incubation of unseeded, intact AS
samples. Unfortunately, the absorbance at 260 nm
(corresponding to release of intracellular nucleic acid
m a t e ~ - i a l ~ .was
' ~ ) masked by absorbance at 275 nm,
probably due to amino acid production via protein hydrolysis.
The fast turnover rate of soluble materials
released upon cell lysis also limits the utility of this
technique.
Unsuccessful attempts were also made to distinguish
death from lysis using microscopic differentiation of
three categories of bacterial cells: viable, dead-intact,
and dead-damaged. Fluorescein diacetate (FDA) and
ethidium bromide (EB) stains were combined into a
single a ~ s a y . ' ~Results
.'~
were negative, in that viable
cells did not fluoresce when stained with FDA, and all
cells, regardless of viability, stained positively with
EB. The failure of the staining procedure might be due
to the nature of the particular microorganisms present
in our system. At the time, the dominant organism form
appeared to be that of a well-dispersed tetrad resembling Micrococcus spp.
Because of the inability to separately measure death
and cell lysis rates, and because a good body of evidence3
supports the notion of simultaneous cell death and lysis,
the lysis rate is equated to the cell death rate in our
model. Therefore, all limitations applied to our measurement of the death rate constant-estimated by use
of the OUR technique-apply to its use as a combined
deathhysis rate coefficient as well. For all practical
purposes, this rate coefficient is hereafter called the
death/lysis rate coefficient.

",''

Analytical Methods

The following parameters were assayed in accordance with procedures outlined in Standard MethodsIs:
pH (Sec. 423); chemical oxygen demand (COD) (di1524

chromate reflux method, Sec. 508A, with "soluble"

COD measured on samples centrifuged at lo4 g for 30
min, followed by filtration through 0.45-pm membrane
filters); dissolved oxygen (membrane electrode method,
Sec. 421F); and gas composition [gas chromatographic
method, Sec. 511B, with a silica-gel column (50C) in
series with a molecular sieve column (25"C)I. Gas production measurements and volatile acids were assayed
as described elsewhere.'

EVALUATION OF MODEL PARAMETERS

AS Organism Decay Coefficient (PSI

The average decay coefficient applicable to the aerobic, activated sludge organisms (bAS) was measured
using an OUR technique' under starvation conditions
(i.e., the AS mixed liquor was aerated over 17 days
without any exogenous substrate addition). A plot of
the logarithmic (base e) OUR vs. time resulted in a
straight line with a slope, bAS = 0.12 per day (R2 =
0.997). The bASvalue is well within the range of decay
coefficients found for mixed-culture aerobic systems:
0.10-0.20 per day.4

AS Viable Cell Degradable Fraction (f,)

Gossett and Belser' expropriated the following equation from Christensen and McCartyt6as an expression
for the AS biodegradability ( D ) , when the AS is generated from a soluble feed:

This equation can be solved for fd:

fd

D(l + bASeS)
1
(DbASeS)

The applicable microorganism decay coefficient (bAS)

was measured and found to be ca. 0.12 per day (see
the previous section). The activated sludge reactor solids retention time
was 10 days, and the ultimate
AS biodegradability (D)ranged from 51-7 to 61 S%.'
Substituting these values into eq. (22), the calculated
fd range is 0.70-0.78. Christensen and McCarty16 suggested a value of 0.8, but did not actually measure it
themselves. Gossett and Belser' found an fd value of
0.68 for a similar AS. A range offd from 0.65 to 0.80
is found in the literature (cited by Gossett and Belser').
The fd values estimated in the present study are well
within this range. Henceforth, we employed our average value of fd = 0.73 which corresponds to the
observed, average, ultimate AS biodegradability of
55.8%, based on gas data.2

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986

(eS)

Death/Lysis Rate Coefficient (&J

The kd was estimated by use of an OUR technique.

The limitations of this technique were discussed in a
previous section. For the case of a completely mixed,
continuous-flow reactor without recycle, the surviving
fraction of viable AS organisms-assuming a firstorder death rate [eq. (])]-is given by eq. (23):

XtS
- -

x;;

+k

(23)

d p

Assuming that there is a direct proportionality between

X t S and OUR, eq. (23) becomes equivalent to:

where (OUR), and OUR are the oxygen uptake rates

of the digester influent and mixed liquor, respectively
(mg O2 L I min-]).
This technique was applied to the mixed liquors from
the continuous-flow digesters receiving intact AS and
the results are shown in Figure 3. The kd value was
found to be equal to 2 per day by use of eq. (23a).
Cell Soluble BOD Pool and Diffusional
Considerations

The conceptual model (Fig. 1) accounts for intracellular soluble BOD that is immediately released upon
cell lysis, followed by continued release of additional
intracellular solubilized material under possible diffusional limitations. However, in the simplified model
(Fig. 2), it has been assumed that there is no significant,
diffusional limitation. Two questions need to be addressed: 1 ) how much of the intracellular soluble BOD
is immediately released upon cell lysis and 2) is there
any diffusional limitation that controls the rate of release of remaining soluble BOD or soluble BOD formed
via intracellular hydrolysis?
Experimentally, it seems almost impossible to quan-

PREDtCTlON SHOWN FOR:

OUR,-^ rnpi-min-
k,-2

day-

1w

5 2j\
0

0 0

SRT,Doys

Figure 3. Estimation of deathilysis rate coefficient (kd) for intact

AS fed to continuous-flow digesters by use of oxygen uptake rate
technique (bars indicate 95% confidence intervals).

tify both the intracellular soluble BOD and its rate of

release under natural conditions, i.e., under conditions prevailing in a digester. In search of an estimation, we followed an indirect approach: it was assumed that autoclaved cells resemble dead, damaged
cells that got that way naturally, and that the quantity of soluble matter eventually released from autoclaved cells held in a sterile environment is a good
measure of the quantity of soluble matter in an intact
cell that would be capable, eventually, of permeating
a damaged membrane without requiring further hydrolysis. These are, admittedly, rather tenuous assumptions. However, microscopic examination of autoclaved cells shows them to be essentially identical
to viable cells, with intact cell walls. Heat may, of
course, damage the membrane far more than normal
death/lysis, but the literature suggests that upon death,
the life of a microbial membrane is relatively short.3
Heat may also alter the relative proportions of particulate and soluble materials within the cell, as
well as the cellular chemical composition. But finding
no other path available, the above assumptions were
made of necessity.
Intact AS was placed in glass dilution bottles closed
with heat-resistant screw-caps and autoclaved (121C,
30 min). Following sterilization, the bottles were equilibrated to 35 t 1C temperature by immersing them
in a water bath. Finally, the bottles were placed in a
controlled temperature room (35 ? 1C). At time intervals, one bottle was opened and its soluble fraction
was separated by centrifugationlfiltration.COD measurements and UV scans were taken of the autoclaved
sludge soluble fractions. During ca. 5 days following
sterilization, the UV scan showed no change, and neither did the soluble COD. For example, immediately
after sterilization, the soluble COD was 1180 mg/L,
with a coefficient of variation (i.e. lOOS/2?l equal to
1%; the total intact (i.e., before autoclaving) AS COD
was 4230 mg/L with 40 mg/L as soluble COD. This
represents ca. 28% solubilization. Subsequent soluble
COD measurements showed an average of 1200 mg/L
with a coefficient of variation equal to 5%.
From the above results, along with knowledge of the
intrinsic biodegradabilities of soluble and solid fractions (90.7 and 70. I%, respectively*), we conclude that
the intracellular soluble BOD is equal to 30% of the
total cell BOD, and that it is all immediately released
upon cell lysis. Thus, the diffusion of soluble intracellular BOD was neglected, resulting in the simplified
model (Fig. 2).
Hydrolysis Rate Coefficient (khl

The hydrolysis rate coefficient was estimated from

batch studies in which autoclaved AS solids were inoculated with differing amounts of an anaerobic culture. Autoclaved AS solids were employed as substrate

PAVLOSTATHIS AND G O S S E T : ANAEROBIC DIGESTION OF SLUDGE

1525

because elimination of the cell membrane barrier via

autoclaving allows kh values to be easily inferred from
observations of solubilization rate (after appropriate
corrections detailed below). With use of intact AS as
substrate, solubilization rate is influenced by both lysis
and hydrolysis kinetics, necessitating the numerical
solution of three simultaneous, differential equations
in order to estimate kh.8
Use of batch reactors fed autoclaved AS solids requires acceptance of at least two assumptions: that the
hydrolytic environment of a batch system can approximate that of a continuous-flow reactor, provided a
relatively large volume fraction of inoculum is employed, and that the hydrolysis rate in an anaerobic
system fed autoclaved AS solids is representative of
the rate in a comparable system fed intact AS. This
latter assumption really encompasses two others: that
autoclaved solids resemble nonviable solids which result from more natural death/lysis processes, and
that the thermal destruction of AS enzymes via the
autoclaving process-with loss of one possibly-significant source of hydrolytic enzymes in AS fed digesters-does not adversely affect the estimation of kh.
The proposed digestion model (Fig. 2) assumes a
simplistic, first-order dependence of hydrolysis rate
upon nonviable, particulate BOD concentration [eq.
(2)]. No explicit dependence on digester microorganism concentration is assumed. If production of hydrolytic exoenzymes by anaerobic microorganisms-andor
enzymatic activity following production-are
regulated processes, then resulting enzymatic activity might
indeed be relatively independent of anaerobic microorganism concentration, influenced solely by perceived
need (i.e., substrate and/or product concentrations). If
hydrolytic activity is thus regulated, it can be argued
that autoclaved AS is a better choice of substrate than
intact AS for use in batch experiments.
Some batch experiments were performed using intact AS. Estimates of kh obtained using numerical solution techniques appear unrealistically high.8 It is
thought that the pulse of AS enzymes released when
intact AS is batch-fed to an anaerobic environment
may elevate hydrolytic activity to artificial levels. In
a continuous-flow system, on the other hand, the relatively slow steady influx of these AS enzymes may
merely decrease the need of native anaerobic bacteria
to produce exoenzymes in order to keep pace with the
pressures of outflow and natural enzyme destruction/consumption. In essence, the natural regulatory
mechanism may be capable of incorporating the intact
AS contribution to enzymatic activity in continuousflow systems, but is likely overwhelmed in batch-fed
systems.
Despite obvious concerns, but considering all the
above arguments, we decided to employ autoclaved
AS solids as substrate in batch hydrolysis assays. The
simplistic first-order form of the hydrolysis model was
1526

tested by utilization of differing amounts of anaerobic

seed in the studies.
A summary of the results from the batch studies with
autoclaved AS solids is given in Table 1 . The data were
corrected for methane production, microbial growth,
and decay during batch anaerobic digestion, since these
affect the particulate and soluble COD pools. In other
words, new biomass produced at the expense of soluble COD had to be estimated and properly added to
the solubilized COD pool. The correction was based
on a COD conservation equation and the anaerobic
digestion process was viewed as a two-phase process
(i.e. acidogenesis and methanogenesis).2 Detailed data
are given only for one sample ( B , in Table 11). An
apparent difference was observed in rate and extent of
solubilization between the two samples incubated with
different amounts of seed (Bl and B2), but only after
ca. 4-5 days incubation. And, curiously, the B , systems then exhibited greater hydrolytic activity than
did B2 systems. The samples incubated without any
anaerobic seed (B,) showed solubilization almost from
the beginning of the incubation period and after ca. 7.5
days incubation, methane was detected in the gaseous
head space of these bottles, indicating active methanogenesis. Although these bottles were not deliberately seeded with anaerobic inoculum, contamination
is assumed since preparation of substrate solids, media
transfers, and other bottle preparations were effected
in a nonsterile environment. However, given the large
volumes of inoculum used in seeded systems, it is hard
to imagine how contamination could have significantly
influenced results for B 1and B2 samples.

Table I. Influent particulate COD remaining during batch anaerobic

digestion of autoclaved sludge solid samples (mg/L).

0
0.5
1
1.5
2
2.5
3.5
4.5
5.5
7.5
8.5
12.5
13.5
18.5
22.5
24.5
30.5

2745
2650
2585
2440
2330
2320
2230
2155
2040

3265
3055

2350

2480

1830
1785

2020
1660

2220
2040

1395
1330
-

1370

3265
2990

2805

2735

2620
2540
2490

2650
2550
2530

1350
1205

1890
1790

Values corrected for inoculum contribution.

Subscripts indicate relative amounts of anaerobic seed used:
Bo = 0%; B , = 22.2%; and B, = 44.4% (v/v).
a

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986

Table 11. Measured and calculated parameters for autoclaved sludge solids sample B , during batch anaerobic digestion (all parameters
in mg CODIL; seed-blank values subtracted).
Incubation
time
(days)
0
0.5
1.5
2.5
3.5
4.5
7.5
13.5
18.5
24.5
30.5
a

Measured
soluble COD

Cumulative
methane COD

Sum of soluble
and cumulative
methane COD

Acetic

propionic

sum

Xi?

xb

Sum

Corrected
soluble COD

50
140
290
265
140
35
20
<5
10
<5

0
35
85
255
440
580
715
1020
1335
1635
1740

50
175
375
520
580
615
735
1020
1345
1635
1750

15
80
160
140
50
10
10
< 5b
<5
<5
<5

15
40
110
130
95
10
5

30
120
270
270
145
20
15
-

0
30
80
115
125
125
135
165
190
205
230

0
5
5
15
25
35
40
60
75
80
80

0
35
85
130
150
160
175
225
265
285
310

50
210
460
650
730
775
910
1245
1610
1920
2060

10

Volatile acids

<5 b
<5
1 5
<5

Biomass formed

Xi? and XB, are total acidogenic and methanogenic biomasses formed respectively (see ref. 2 for details).
The lower detection limit of individual volatile acids was ca. 5 mg/L.
Sum of measured soluble COD, cumulative methane COD and total biomass formed.

A second batch experiment was performed with autoclaved AS solids, with results virtually identical to
the first experiment. The divergence in activities of B1
and B2 systems after the first 4-5 days of incubation
remains unexplained.
Solubilization during the first few days of incubation
was found to fit the assumed model of eq. (2) very well.
Naturally, we decided to base kh estimates on these
initial data, rationalizing that errors in all those corrections required to adjust observed solubilization in
order to estimate true solubilization may be the explanation for anomalous results obtained as incubation
proceeded. The corrections are less important near the
beginning of incubation.
The B l systems showed ultimate COD solubilizations averaging 63.1% at 30 days incubation. This
compares reasonably well with ultimate digestibilities
of 69.5% earlier reported for autoclaved AS solids incubated for 85 days.2 Note that the data of Table I1
indicate that solubilization equates with degradability,
since practically all soluble COD is converted to
methane.
With 63.1% as an estimate of the total, ultimately
solubilizable fraction of initially fed COD, the remaining solubilizable praticulate BOD (F)was estimated
for both B , and B2 systems using cumulative solubilization vs. time data. Upon integration, eq. (2) gives:

B 1 via eq. (24). A kh value of 0.16 day- was obtained.

Likewise, the B2 sample data gave kh = 0.14 day-
(R2 = 0.92). For all practical purposes, an average kh
value of 0.15 day- I is proposed.
The observation that kh appears relatively independent of anaerobic microorganism concentration supports its omission from the hydrolysis model. The
goodness of fit between eq. (2) and the initially obtained
solubilization data supports the specific form (i.e. firstorder assumption) of the model.
Questions remain concerning the extrapolation of kh
values measured in batch systems to application in
continuous-flow systems. Care was taken to ensure
that abiotic parameters (e.g., pH, alkalinity, redox potential, etc.) in the serum bottle assays were close, if
not identical, to values observed in the continuousflow digesters also employed in this research. Biotic
parameters, though, are difficult to quantify or control.
Applicability of kh values so-obtained is evaluated in
the next section.

AUTOCUMD SLUDGE SOLIDS (81

(24)
where Fo is the maximum solubilizable particulate COD
(g/L). Thus, the hydrolysis rate coefficient (kh)can be
estimated-in the case of autoclaved sludge solidsfrom the slope value of the straight line obtained by
plotting In (FIFO)vs. time.
Figure 4 shows how the kh was estimated for sample

kh -0.16 day

-1.5
0.0

0.5

1.0

1.5

2.0

2.5

3.0

TIME.Days

Figure 4. Hydrolysis rate coefficient (kh)evaluation for autoclaved

sludge solids (batch experiment, sample B , ) .

PAVLOSTATHIS AND GOSSElT: ANAEROBIC DIGESTION OF SLUDGE

1527

Table 111. Summary of model parameters with typical values used in this study
(continuous-flow digestion).
Values for
Parameter
Influent COD concentration [COD,, (g/L)J
Ultimate digestibility [D(fraction)]
Net biodegradable fraction of active biomass Ud)
Cell soluble degradable COD immediately
released [ y (fraction of total degradable COD)]
AS cell death rate coefficient [kd (day-')]
Hydrolysis rate coefficient [ k , (day- ')I
Decay coefficient for acid formers [bA(day-')]
Yield coefficient for acid formers [ Y A (g COD
biomass/g COD utilized)]
Maximum specific substrate utilization rate for
acid formers [kA (g COD utilized/g biomass
COD day-')]
Half-velocity coefficient for acidogenesis
[K;' (g C O D W I
Decay coefficient for methanogens [bB(day-')]
Yield coefficient for methanogens [ YB (g COD
biomass/g COD utilized)]
Maximum specific substrate utilization rate for
methanogens [kB(g COD utilized/g biomass
COD day-')]
Half-velocity coefficient for methanogenesis
[K? (g COD/L)I

MODEL PREDICTIONS OF DIGESTER

PERFORMANCE

A summary of the model parameter values is given

in Table I11 for both intact and autoclaved sludges. The
values for parameters such asfd, y, kh, and kd were
measured directly in this study and are discussed in
previous sections of this article. Values for CODinand
D were presented previously.2 Values for bA and YA
were chosen based on results discussed elsewhere.*
The values of kA and K$ were chosen to minimize the
nonvolatile acid effluent degradable soluble COD based
on continuous-flow digestion data which showed that
this component is neglible.2 The biokinetic coefficients
for methanogens were chosen from literature values"
(for bB and YE)o r by fitting of acetate and propionate
data in this study' (for kB and K:).
Based on the above-presented parameter values,
digester performance predictions were obtained. For
autoclaved AS (Fig. 5), eqs. (13)-(20) were employed,
with X$$ = 0; Po = 0.26 g COD/L (i.e., influent measured volatile acid COD); S$ = y D COD,, - Po; and
Fo = ( 1 - y)D CODin. For intact AS (Fig. 6), eqs.
(12)-(20) were employed, with S$ = Fo = Po = 0. The
fit of the mode1 is quite good.
An additional performance data point was obtained
for digesters receiving intact AS. At the completion of
the previously reported continuous-flow digestion
studies,2 the feed rate was increased to the digester
1528

Intact
AS

Autoclaved
sludge

14.8
0.558
0.73

14.95

0.30
2.0
0. I5
0.10
0.20

0.722
0.73
0.30
-

0.15

0.10
0.20

8.0

8.0

0.045

0.045

0.015
0.057

0.015

6.2

6.2

0.045

0.125

0.057

operated at the lowest @ P I G , decreasing retention time

to three days. Operation continued for 10 days, after
which its level of performance was noted. This data
point is included in Figure 6.

DISCUSSION

The model adequately predicts digester performance

for intact sludge using the hydrolysis rate coefficient
estimated by batch digestion of autoclaved sludge sol-

50
AUTDCUMD SLUDGE

c3

-30

z
P
+
P
+

0 20
3

t-L11

8
10
n

MODEL PREOlCnONS
0-0.722

k "-0.1 5 doy

-'

Figure 5. Continuous-flow digester performance, model predictions for autoclaved sludge (bars indicate 95% confidence intervals
on data).

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986

1 0 1 2

SRT.Days

Figure 6. Continuous-flow digester performance, model predictions for intact AS (bars indicate 95% confidence intervals on data).

ids, together with the ultimate digestibility data and

other parameter values presented in Table 111.
A sensitivity analysis was performed in order to
quantify the effect of each parameter on the digester
performance prediction. The results are shown in Table IV. The base-line values are the same values
reported in Table 111. From these results, it is obvious
that the ultimate digestibility has a major impact on
digester performance predictions. Parameters such as
fd, y, kd. kh, bA, YA, and YB have less impact on the
model prediction. The rest of the parameters have almost negligible effect on the model predictions. The
results of the sensitivity analysis support our experimental approach, which was to concentrate on independent measurement of important parameters, such
as D and kh, while being satisfied with literature values
for many others.
However, if erroneous interpretations are to be
avoided, these results should be used with caution. All
that is shown in Table IV is the percent change of the
Table IV. Sensitivity analysis of model parameters (Intact AS;
GI@
= 8 days).

Parameter

Base-line
parameter
value

Percent change in predicted

digester performance with
10% increase in parameter
value (%)

14.8
0.558
0.73
0.30
2.0
0.15

0.1
10.1
0.8
2.0
0.5
2.4

0.10
0.20

0.4
-1.5

8.0
0.045
0.015
0.057
6.2

0
0

0.045

For parameter units, see Table 111

0
- 0.5

0. I
-0.1

model prediction (at @IG = 8 days) for a 10% change

in each parameter. Two points are noteworthy: all 14
parameters are interrelated, and therefore the percent
change in model prediction around a base-line for
a 10% change of one parameter depends on the level
(value) of the other thirteen parameters, and all parameters are not expected to have equal uncertainties.
Based on the kd and kh values found in this study
for intact AS, and on the results of the sensitivity analysis (Table IV), we can conclude that hydrolysis is a
more kinetically limiting mechanism than death/lysis,
with respect to conversion of viable AS organisms into
available substrate. Of course, there is uncertainty associated with the kd value (2 day-) reported in this
study due to the technique used to measure it. What
should not be overlooked here is that no matter the
exact value of kd, these data suggest that it is a large
number compared with the kh value. Digester performance becomes a very insensitive function of kd.
Part of the influent viable AS biomass may adjust to
the anaerobic environment (e.g., facultative AS bacteria) and become active acidogenic biomass. In this
case, the acidogenic biomass is underestimated by the
present model, and kd measurements might be in error.
Presently, the above point cannot be either proved or
disproved due to lack of accurate techniques for differentiation and estimation of acidogenic, methanogenic, and viable AS bacterial population densities.
However, previous anaerobic digestion studiess20 have
shown that fastidious, nonmethanogenic, obligate anaerobic bacteria are present in numbers one or two
orders of magnitude greater than facultative bacteria.
On the other hand, most of the principal genera of AS
bacteria recorded in taxonomic studies are found to be
strict aerobes.21Thus, the importance of AS facultative
bacteria as active anaerobic digester acidogenic bacteria seems to be very minimal.
The superiority of the present model vis-a-vis models which consider methanogenesis rate-limiting,
lies in the analytical prediction of the digester effluent characteristics. As was discussed previously,2 the
ORourke model-formulated for raw primary sludge
digestion and based largely on lipids degradation-ignores effluent particulate BOD of other waste components (e.g. protein) and assumes complete hydrolysis of the influent degradable COD for @IG higher
than ca. 7 days. It works well for digestion of largely
lipid waste. In the case of AS digestion, ca. 99% of
the total effluent COD is particulate, showing that hydrolysis is not complete, even for 0PIGhigher than 25
days. Particulate protein comprises between 80 and
90% of the total effluent degradable COD.
The present model represented by eqs. (131420) can
easily be adapted to situations where the digester influent contains significant contributions of nonbiological, degradable sources, such as in the digestion of
combined municipal primary and waste activated

PAVLOSTATHIS AND G O S S E T : ANAEROBIC DIGESTION

OF SLUDGE

1529

sludges. (Of course, the accuracy of the model in such

cases has not been demonstrated.) In the simplest case
where the applicable hydrolysis coefficient for the nonbiological BOD roughly equals that of the lysed, biological solids, all that is necessary is the addition of a
finite Fo value to eq. (14). If the hydrolysis coefficients
differ for biological and nonbiological solids, separate
mathematical treatment of the two sludges is required
down to the point of soluble COD production. From
that point onward, the presented model may be used,
regardless of the soluble COD source. Additionally,
however, it should be noted that the feed composition
may influence the half-velocity coefficient for methanogenesis (K:); if significant lipid material is present,
use of a greatly increased value of K: may be in order,
to properly reflect the expected high effluent contributions from long-chain fatty acids.
In prediction of dynamic behavior the complex model
presented in this article is perhaps necessary. However, for many practical applications, it may be unwarranted. If desired, the model may be simplified,
based on results presented here. Model predictions
suggest that death/lysis and acidogenesis mechanisms
are sufficiently rapid that they may be neglected, at
least for retention times of practical interest. Doing so
results in essentially the model proposed by Gossett
and Belser, * but where the preliminary conversion step
of concern has been properly acknowledged as a hydrolysis step, rather than a death/lysis step.
CONCLUSIONS

Based on the results of this study, the following conclusions can be drawn. In the conversion of viable,
biological solids to available substrate for the anaerobic microflora, mechanisms such as death, lysis and
hydrolysis play an important role. Results suggest that
hydrolysis of degradable, particulate, dead AS biomass
is much slower than death or lysis. Therefore, hydrolysis appears to be the rate-controlling step in the
case of anaerobic digestion of biological solids. As a
result, ca. 99% of the degradable, effluent COD is particulate (mostly protein).
The death rate coefficient (kd)-estimated by use of
an oxygen uptake rate technique-of viable AS biomass under anaerobic conditions (2 per day) is more
than 16 times higher than the decay coefficient (bAS)
under aerobic starvation conditions (0.12 per day).
An anaerobic digestion model for biological sludge
was developed and evaluated. The four major steps in
this model are: viable cell death/lysis; hydrolysis of
particulate dead biomass; acidogenesis; and methan-

1530

ogenesis. A sensitivity analysis of the parameters involved in this model showed that the most important
parameter is the ultimate digestibility of the sludge,
followed by other parameters such as those relating to
hydrolysis and cell-soluble fraction. Overall, the model
proved to be adequate for predicting the performance
of a laboratory-scale digester receiving a biological solids feed. An extension of the model for use in cases
of combined primary and waste-activated sludge appears feasible.
This research was supported by the U.S. Environmental Protection Agency through Grant No. R 809500-01-0. This article
has not been subjected to the Agencys required peer and
administrative review and, therefore, does not necessarily
reflect the views of the Agency and no official endorsement
should be inferred.

References
I . J. M. Gossett and R. L. Belser, J . Environ. Eng. Div. ASCE,
108, 1101 (1982).
2. S. G. Pavlostathis, A Kinetic Model for Anaerobic Digestion
of Waste Activated Sludge, Ph.D. thesis, Cornell University,
Ithaca, NY, 1985.
3. T. D. Brock, Biology of Microorganisms, 3rd. ed. (PrenticeHall, Englewood Cliffs, NJ, 1979).
4. A. W. Lawrence and P. L. McCarty, J . Sanit. Eng. Div. ASCE,
96, 757 (1970).
5. J. A. Eastman and J. F. Ferguson, J . Water Pollut. Control
Fed., 53, 352 (1981).
6 . P. L. McCarty, Int. J. Air Water Pollut., 9, 621 (1965).
7. S. Ghosh, Biotechnol. Bioeng. Symp.. 11, 301 (1981).
8. S. G. Pavlostathis and J. M. Gossett, unpublished.
9. J. R. Postgate and J. R. Hunter, J . Gen. Microbiol., 29, 233
( 1962).
10. R. A. Laddaga and R. A. MacLeod, Can. J . Microbiol., 28,414
(1982).
11. K. M. Sorrells, R. A. Cowman, and H. E. Swaisgood, J . Bacteriol., 112, 474 (1972).
12. K. Ohmiya and Y. Sato, Agric. Biolog. Chem., 42, 7 (1978).
13. J. L. Jarnagin and D. W. Luchsinger, Stain Technol., 55, 253
( 1980).
14. J. T. Kvach and J. R. Veras, Int. J. Leprosy, 50, 183 (1982).
15. American Public Health Association, Standard Methods for the
Examination of Water and Wastewater, 15th ed., (American
Public Health Association, Washington, DC, 1980).
16. D. R. Christensen and P. L. McCarty, J. Water Pollut. Control
Fed., 147, 2652 (1975).
17. J. T. ORourke, Kinetics of Anaerobic Treatment at Reduced
Temperatures, Ph.D. thesis, Stanford University, Stanford,
CA, 1968.
18. D. F. Toerien, M. L. Siebert, and W. H . J. Hattingh, Water
Res., 1, 497 (1967).
19. P. G. Thiel, D. F. Toerien, W . H. J. Hattingh, J. P. Kotze, and
M. L. Siebert, Water Res., 2, 391 (1968).
20. R. A. Mah and C. Sussman, Appl. Microbiol., 16, 358 (1968).
21. E. B. Pike, Aerobic Bacteria, in Ecological Aspects of UsedWater Treatment C. R. Curds and H. A. Hawkes, Eds. (Academic, London, 1975), Vol. 1.

BIOTECHNOLOGY AND BIOENGINEERING, VOL. 28, OCTOBER 1986