COD Fractionation of Palm oil mill effluent (POME): Typical Treatment,

Biodegradability and Modeling

Salmiati1, M.R. Salim1, Z. Ujang1* and G. Olsson2**

1
Institute of Environmental and Water Resource Management (IPASA), Universiti Teknologi Malaysia, 81310 Johor,
Malaysia. (*Email: zaini@utm.my)
2
Lund University, SE-22100 Lund, Sweden (**Email: wstgustaf@gmail.com)

ABSTRACT
This study aims to establish the scientific link between typical treatment and biodegradability of different COD
fractionations of palm oil mill effluent (POME), by means of respirometric analysis and model calibration. Typical
treatment processes are determined by the treatment methods applied to the POME. POME treatments are operated
under dynamic conditions, where the microbial response can include, besides bacterial growth, several COD removal
mechanisms, and particularly the storage in form of polymers. Biodegradability-related COD fractionation was
characterized as the availability of readily biodegradable (Ss), slowly biodegradable (Xs), storage polymer (XSTO), inert
organic matter (SI and XI), and yield of heterotrophic organism (YH). 
 To verify the experimental results, modeling
studies have been conducted using the activated sludge model (ASM) models, ASM1 and ASM3. The models require
characterization of POME using COD-fractionation. The characteristics of fermented POME for anaerobic treatment
using COD-fractionation indicates the presence of up to 35% readily biodegradable organic substrate (Ss) and COD-
fractions due to availability of free fatty acids in POME. However, the percentage of Ss is higher compared to the raw
POME, which represents only 10% of the total COD. ASM3 proved to be a better framework to model all experimental
results compared to ASM1.

Keywords: Fractionation, activated sludge model (ASM), OUR, respirometry, yield

INTRODUCTION
Activated sludge modeling has gained a new perspective with the introduction of substrate storage
into activated sludge models. Substantial research has shown that the storage mechanism plays an
important role under dynamic substrate conditions experienced in treatment plants. ASM1 was
primarily developed to yield a good description of the sludge production and consumption patterns
of electron acceptors [1]. Storage phenomena have been encountered in modeling enhanced
biological nitrogen removal and carbon removal processes in ASM3 [2].

Storage of excess substrate available under feast conditions allows microorganisms capable of
substrate storage to survive on the accumulated substrate reserves when no external substrate is
present (famine conditions). These microorganisms have the benefit of a more balanced growth and
an advantage in competition [3].

The mixed culture process from activated sludge is able to store PHB. PHB production using a
mixed culture is feasible due to the role of PHA as metabolite intermediate of wastewater treatment.
Using only conventional Monod-type kinetics to predict competition (between storage of substrate
and balanced growth) is obviously not enough to accurately describe what happens in wastewater
treatment processes.

A better knowledge of substrate removal mechanism in activated sludge plants is particularly

needed for modeling purpose. Growth and other substrate removal mechanisms such as storage,
accumulation and sorption are typical phenomena of the microbial response to dynamic conditions
[4]. Experimental work carried out on the activated sludge processes would give some
understanding of the microbial behavior as well as the substrate removal performances. An
empirical model has been developed based on the study. Such a model is hopefully able to simulate
the behavior of organisms in dynamically fed systems.
MATERIALS AND METHODS
Two types of bioreactor were designed and used to meet metabolic activities in one cycle. The
anaerobic part was designed in a cylindrical shape with a working volume of 19 litres, interior
diameter 19.5 cm and height 90 cm. The sampling ports were designed along 10 cm height
intervals from the bottom. The reactor was operated at room temperature (28 ± 2oC) and at varied
sludge and hydraulic retention times (SRT/HRT). At each HRT, the reactor was operated for six
weeks to reach steady state conditions that were established when the variation in the product
concentrations were constant (effluent VFAs and COD concentrations).

The aerobic part was fabricated in two double-jacketed laboratory-scale reactors with six litres
effective volume. The operation of the aerobic reactor was performed as a sequencing batch reactor
(SBR) system under feast-famine regime which was conducted on samples of fermented POME
taken from an anaerobic reactor. The cycle of operation system depends on the substrate
concentrations. The detailed features of the aerobic system are described in Table 1. The dissolved
oxygen (DO) concentration was measured as percentage of the saturation concentration (100% =
9.1 mgL-1). DO concentration and pH were measured continuously. In order to control the oxygen
concentration properly, the gas concentration was controlled using a gas flow meter.

The oxygen uptake rate (OUR) vessel was designed for a volume of 25 mL to ensure good
monitoring of ‘endogenous respiration’ of microorganisms. This respirometer equipped with a DO
probe and a magnetic stirring bar was connected to the reactor and placed on a stirrer equipment.
The biomass was pumped directly from the aerobic reactor to the vessel and after three minutes the
recirculation pump was switched off. The decrease in DO concentration during one minute was then
measured and electronically recorded using the DAPS software (ISTEK, Korea). In the feast phase,
the biological activity was high and the DO concentration decreased rapidly. In order to have more
measurements in a relatively short feast phase, the DO concentration was measured and recorded
for every 30 seconds. The DO values were plotted against time. A straight line of “best fit” was
calculated and its slope was determined. The slope represents the OUR of the biomass for a certain
time during one cycle. Once the OUR is determined, the oxygen transfer coefficient (K ) in the
La
aerated reactor can be calculated from the following formula:

Where: OUR = oxygen uptake rate [mg O /L/min]

2
K = oxygen transfer coefficient [L/min]
La
o
C = oxygen saturation concentration in water at 20 C [mg O /L]
s 2
C = dissolved oxygen concentration in the reactor [mg O /L]
2

Table 1. Operating phase with POME as substrate

Experiment (s) Operating time (min)
Aerobic Aerobic Aerobic Draw/discharge
mineral feeding reactor
feeding
Growth 0 - 60 0 - 60 60 -330 330 - 340
DO (pretreated POME) No fill - 0 - 600 -
Mic ae (pretreated POME) No fill - 0 - 500 -
RESULTS AND DISCUSSION
OUR measurements were conducted to characterize the COD-fractionation. The COD-value covers
a number of organic materials of varying biological qualities. This helps to determine the
availability of readily biodegradable (Ss), slowly biodegradable (Xs), storage polymer (XSTO), inert
organic matter (SI and XI), and yield of the heterotrophic organisms (YH). The calculation models
are as follows [5,6].

Table 2 and Figure 1 show the comparison of COD-fractionation in a typical treatment of POME. It
indicates that raw POME in fed-batch gives a higher value for all COD-fractionations as compared
to other typical POME treatments. In addition, the diluted POME was prepared by collecting 65%
of supernatant from raw POME to be filled-up with 6 L, whereas the fermented POME was carried
out from the anaerobic reactor that hydrolyzes long chain fatty acids in POME. From the trend
shown in Figure 2, the higher value of the total COD was not influenced by the COD-fractionation
(especially Ss), but it is according to the typical treatment in POME. It clearly shows that the
fermented POME, Ss gives similar value for COD-fractionation as compared to diluted POME.

Table 2. Comparison concentration of COD fractionation

Influent COD COD fractionation (g/L)
total (g/L) Ss Xs XSTO XH SI + XI
a
Raw POME 50.4 9.6 22.9 - - 17.9
b
Diluted POME 18.9 3.6 9.2 - - 6.1
c
Fermented POME 10.1 3.6 3.8 - - 2.8
Note: (a) in fed-batch reactor (Zie Wei et al, 2007; Md Din 2007), (b) collected from initial discharge, (c) in fed batch
reactor collected from anaerobic reactor; XSTO is additional cell internal storage of PHA excluding XH in ASM3.

Figure 1. Trends of the COD fraction in typical treatment in POME

In this study, the S was recorded at high value, about 35% of the total COD compared to the raw
S
and diluted POME, both S values were about 18 -20% of the total COD as shown in Table 3. This
S
is in agreement with other studies [7] and it can be concluded that S can be increased up to 31% of
S
the total COD if the substrate contains a mixture of sludge and bulking agent, as also indicated in
this study. For example, inert particulate represents a large part of the total COD but is much less
biodegradable than the sludge, according to the considered process time scale. Therefore, as
observed in this study, X was always higher than S which indicates that a third of the influent
S S,
(fermented, fed-batch and raw POME) contained high amounts of slowly biodegradable matter,
even after pre-treatment. This may be the residual solids or separate oil in the POME which is still
difficult to digest. The total inert organic matters (S plus X ) were detected at a range 20 – 30% of
I I
the total COD.

The correct assessment of the S is important because this fraction is conceived as the rate limiting
S
substrate component for heterotrophic growth (X ). It is also related to the oxygen uptake rate
H
(OUR) measurement as observed in this study. The biodegradable fraction of the present study (S
S
and X ) is recorded at more than 50% of the total COD. Thus, this POME is considered as a rich-
S
substrate for PHA production. This study utilized the Activated Sludge Model 3 (ASM3), which
includes cell internal storage compounds (X ), which requires the biomass to be modeled with
STO
cell internal structure. Therefore, X is provided to compare the degree of fraction inside the
STO
heterotrophic organisms. The availability fraction of X
STO in the medium indicates that the PHA
production occurred intracellularly in the biomass component. Table 4 shows the storage of PHA in
the inclusion body of the bacteria.

Table 3. Percentages of COD-fractionation in POME

Influent COD COD fractionation (%)
total (%) Ss Xs XSTO XH SI + XI

Raw POME a 100 20 45 - - 35

Diluted POME b 100 19 48 - - 33
Fermented POME c 100 35 38 - - 27
Note: (a) in fed-batch reactor (Zie Wei et al, 2007; Md Din 2007), (b) collected from initial discharge, (c) in fed batch
reactor collected from anaerobic reactor; XSTO is additional cell internal storage of PHA excluding XH in ASM3.

Table 4. COD-fractionation after the treatment of the fermented POME

Experiment COD COD fractionation (g/L)
total (g/L) Ss Xs XSTO XH SI + XI

Fermented POME a 10.14 3.55 3.82 - - 2.77

50% substrate 6.13 0.32 2.03 1.92 0.32 1.54
25% substrate 2.79 0.14 0.92 0.81 0.21 0.71
(a) in fed batch reactor collected from Anaerobic reactor; (b) feeding 3L in the Aerobic reactor; (c) feeding 1.5L in
the Aerobic reactor

Table 4 depicts the wastewater biological fractionation using a respirometric analysis after the
treatment. Samples were prepared from fermented POME. The goal of this experiment was to
identify the biological fractions in the fed-batch systems. The COD removal obtained from this
system was above 50% from fermented POME. The removal of COD could be considered
acceptable since the study was conducted for only 4-6 hours.

In this study, fermented POME has been characterized to include a readily biodegradable COD
fraction (S ), which contains a considerable amount of VFA. VFA can be stored as PHA by
S
bacteria under dynamic conditions and recent studies have shown that the substrate storage
phenomena play an important role on the biochemical transformation involved in the biological
treatment of POME. In addition, fermented POME also has particulates which are of the slowly
hydrolysable COD fraction (X ). However, some slowly biodegradable matter (X ) may actually be
S S
soluble in SS [6,8]. The amended model suggests that the readily biodegradable COD fraction (S )
S
is utilized and converted to storage polymer (XSTO). It is assumed that the SS fraction is first
converted to the storage polymers and growth on these polymer starts after the total depletion of all
readily biodegradable COD in the bulk solution [9].

In general, the readily biodegradable COD (S ) has been reduced from 3.55 g O /L (fermented
S 2
POME) to 0.14 - 0.32 g O /L for the variable feeding. Meanwhile, the slowly biodegradable COD
2
(X ) is between 0.92 to 2.03 g O /L. Since the S is still present in the medium after the
S 2 S
accumulation stage, a modification of the fed-batch system must be considered to accelerate the
biodegradation process. In addition, a significant reduction of slowly biodegradable COD
concentration (X ) at approximately 40% of the initial concentration indicates that the hydrolysis
S
products (VFAs) will be degraded in a shorter period.

Respirometry has been extensively used for the experimental assessment of activated sludge
models. In this context, an OUR profile obtained from the reactor was a useful tool to differentiate
ASM1 and ASM3 from its modified version including direct growth on primary substrate
competing with storage. These OUR profiles were used to investigate model-fit performance and
identifiability for ASM1 and ASM3 and show the trend of the substrate volume if the same
substrate (VFAs) were added. The differences observed among the OUR profiles were probably
linked to the operational conditions of the treatment (i.e. SRT or alternating feed and famine
conditions).

The fittings of the models (ASM1 and ASM3) are depicted in Figure 2. The parameter estimation
errors obtained are quite small, in part because the method used is known to give too optimistic
results due to autocorrelation in the OUR data [10]. The substrate utilization and the OUR response
of the system under feast conditions were equally well simulated with both models. However, the
ASM1 could not cope with the decrease of the OUR level in the famine phase. In other words, a
first glance at Figure 2 shows that ASM1 is not able to describe the tail observed, where the storage
effect is emphasized. Many respirograms can be found in the literature with this tail, and much of
the criticism that ASM1 has received is that this tail is not predicted when the feed solely contains
readily biodegradable substrate. However, in the ASM3 model the prediction is much better and the
tail of the model is a smooth line for the famine phase. The result from this study was similar to
those observed by [8,9].

ASM1 with only one growth process can neither predict the amount of sludge produced, nor
simulate lower electron acceptor utilization rate in the famine phase. The proposed model with the
addition of the growth process on primary substrate resulted in two different specific growth rates,
where the specific growth rate in the famine phase was 23% of the feast phase as shown in Figure 2.
ASM3 generates a lower growth rate throughout the experiment which gives a lower biomass yield.
The OUR response of the system and the results of model simulation performed for ASM1 and
ASM3 with the kinetic and stoichiometric coefficients are given in Table 5 and Figure 2. Hence, the
more significant the storage effect is, the higher is the performance of ASM3 compared to ASM1.
This improvement is even observed in this experiment, where less storage can be appreciated.
However, once a good fitting is obtained, an analysis on the mechanistic interpretation of the
parameter estimation results is required.
Figure 2. OUR data and the model simulation results for the proposed model ASM1
and ASM3

Table 5. Kinetic and stoichiometric coefficients obtained for the aerobic reactor system fed with
fermented POME
Model coefficient ASM1 ASM3
µH (/day) 0.69 1.88
KS (mg COD/l) 0.92 1.79
YSTO (gCOD/gCOD) - 0.56
YH (g cellCOD/gCOD) 0.42 0.89
bH (/day) - 0.2
XH (mg COD/l) - 1400

According to ASM1, the direct growth on external substrate is the cause of the first shoulder,
whereas the ASM3 model links the first consumption to the storage of substrate into internal
polymer. These processes have different default yield values: 0.69 for the growth yield in ASM1
and 1.88 for the storage yield in ASM3, because less energy is required to store external substrate
than to produce new cells.

High values of the ASM3 growth yield, YH,STO, were also observed in the literature when fitting
ASM3 to experimental data [11,12]. These high values contradict the conceptual basis of ASM3
since the predicted growth yield values no longer have a mechanistic meaning. On the other hand,
ASM1 was not able to predict the tail often observed in the OUR obtained from the experiments.

In general, more reliable parameter values would be obtained if the ASM3 model could describe
that part of the VFAs which was used directly for growth. In this case, the model would predict less
PHA production and the predicted tail would be lower and consequently closer to the experimental
data. Moreover, a decrease on the values of bH and YH, STO would be necessary to describe the tail.
The reduction of the tail (i.e. the reduction of the oxygen consumption due to the storage process) as
a function of a percentage of the substrate used directly for growth is as depicted in Figure 2. This
observation of simultaneous growth and storage on the external substrate has already been
developed in some metabolic model such as that of van Aalast-van Leeuwen et al. [13] and recently
in the work of van Loosdrecht and Heijnen [14] and Karahan et al. [9,15]. Apart from a more
reliable description of reality, considering the growth on external substrate can help overcome
another described limitation of ASM3.

The proposed procedure calculates YSTO in accordance with ASM3, with the assumption that
available readily biodegradable substrate is entirely converted into internal storage products. There
is experimental evidence however for partial storage allowing for simultaneous growth both
competing for the same pool of external substrate [4,16,17]. It could then be argued that limiting
external substrate at low F/M ratios would allow simultaneous microbial growth, as in
conventionally operated activated sludge systems, whereas a sudden increase of external substrate
with high F/M ratios after a famine period would highlight internal storage as the dominant
mechanism. Since growth and storage are commonly defined by significantly different yield
coefficients, the resulting weighted average yield value reflected by overall OUR measurements
would exhibit a changing pattern parallel to adopted F/M values. The experimental results in this
study provided a reliable indication that this was not the case for acetate utilization, which was best
interpreted with total storage.

Two runs were conducted on VFAs (different value of substrate) with initial F/M ratios of 0.025
and 0.035 gCOD respectively. As indicated in Table 6, the yield YSTO is 0.82 gCOD/(gCOD) for
0.035, a value significantly higher than for the 0.025 gCOD. This value agrees well with a YSTO of
0.9 mgCOD/(mgCOD) assumed by Karahan et al. [18] and with the experimental results of Dircks
et al. [17] for smaller volume, with consideration that the formation of glycogen from small
substrate requires less energy as compared to PHB accumulation from a higher volume of substrate.
It is stated that the maximum yield is 30% greater.

Table 6. Experimental assessment of YSTO for different volume of substrate

Set No. Reactor Composition Biomass Substrate F/M Ratio YSTO
Volume (substrate) concentration concentration (COD/COD)
(L) (mgCOD/L) (mgCOD/L)
Set 1 6.0 25% 2500 1300 0.52 0.56
Set 2 6.0 50% 3200 2900 0.90 0.82

CONCLUSION
The characteristics of fermented POME for anaerobic treatment using COD-fractionation indicates
the presence of up to 35% readily biodegradable organic substrate (Ss) COD-fractions due to
availability of free fatty acids in POME. However, the percentage of Ss is higher compared to the
raw POME, which represents only 10% of the total COD. ASM3 proved to be a better framework
than ASM1 to model all experimental results. However, it has to be taken into account that seven
parameters are estimated in this model in contrast with ASM1, where only five parameters are
estimated. In experiments with storage capability bacterial cells, ASM1 was not able to predict the
tail observed due to the internal polymer consumption, while ASM3 could describe this tail
accurately.
ACKNOWLEDGMENTS
The authors are pleased to acknowledge Ministry of Science, Technology and Innovation
(ScienceFund-79004) for funding this research and Bukit Besar Palm Oil Mill, Kulai for providing
samples of POME. Special thanks to Alia Damayanti and Mohd Azlan Ahmad for their assistance.

REFERENCES
[1] Henze, M., Grady Jr., C.P.L., Gujer, W., Marais, G.v.R. and Matsuo, T. (1988). Activated
Sludge Model No. 1, IAWPRC Scientific and Technical Report No. 1. IAWPRC, London.
[2] Gujer, W., Henze, M., Mino, T. and van Loosdrecht, M.C.M. (2000). Activated Sludge Model
No.3. In: Activated Sludge Models ASM1, ASM2, ASM2D and ASM3, Henze, M., Gujer, W., Mino,
T., van Loosdrecht, M. (eds.) IWA Scientific and Technical Report No.9. IWA London. ISBN: 1
900222 24 8.
[3] van Loosdrecht, M. C. M., Pot M. A. and Heijnen, J. J. (1997). Importance of bacterial storage
polymers in bioprocess. Water Science Technology. 35, 41-47.
[4] Majone, M., Dircks, K. and Beun, J. J. (1999). Aerobic storage under dynamic conditions in
activated sludge processes: The state of the art. Water Sci. Technol. 39(1), 61-73.
[5] International Water Association (IWA). (2000). Activated Sludge Models ASM1, ASM2, ASM2d
and ASM3. IWA Task Group on Mathematical Modeling for Design and Operation of Biological
Wastewater Treatment, IWA Publishing, London.
[6] Petersen, B., Gernaey, K., Henze, M. and Vanrolleghem, P.A. (2000). Calibration of Activated
Sludge Models: A Critical Review of Experimental Designs. In: Biotechnology for the
Environment. Eds. Agathos, S. and Reineke, W., Focus on Biotechnology, Vol. 3. Kluwer
Academic Publishers, Dordrecht, The Netherlands. 80.
[7] Tremier, A., Guarda, A., Massioni, C., Paul, E. And Martel, J.L. (2005). Respirometric method
for characterizing the organic composition and biodegradation kinetics and the temperature
influence on the biodegradation kinetics, for a mixture of sludge and bulking agent to be co-
composted. Bioresource Tehnology. 96, 169-280.
[8] Canudas, A. G. (2005). Modelling Biological Organic Matter and Nutrient Removal Processes
from Wastewater Using Respirometric and Titimetric Techniques. PhD Dissertation. Universitat
Autonoma de Barcelona, Bellaterra.
[9] Karahan, G.O., Dogruela, S., Dulekgurgen, E. and Orhon, D. (2008). COD fractionation of
tannery wastewaters: Particle size distribution, biodegradability and modeling. Water Research. 42,
1083 – 1092
[10] Dochain, D. and Vanrolleghem, P.A. (2001). Dynamical Modelling and Estimation in
Wastewater Treatment Processes, IWA Publishing, London.
[11] Beccari, M., D. Dionisi, Giuliani, A., Majone, M. and Ramadori, R. (2002). Effect of different
carbon sources on aerobic storage by activated sludge. Water Sci. Technol. 45(6), 157-168.
[12] Holenda, B. (2007). Development of Modeling, Control and Optimization Tools for Activated
Sludge Process. PhD. Thesis. University of Pannonia.
[13] van Aalast-van Leeuwen, M. A., Pot, M. A., Van Loosdrecht, M. C. M. and Heijnen, J. J.
(1997). Kinetic modeling of poly (β-hydroxybutyrate): Production and consumption by Paraccus
Pantotrophus under dynamic substrate supply. Biotechnol. Bioeng. 55, 773-782.
[14] van Loosdrecht, M. C. M. and Heijnen, J. J. (2002). Modeling of activated sludge processes
with structured biomass. Water Sci. Technol. 45, 13-23.
[15] Karahan, O., van Loosdrecht, M.C.M. and Orhon, D. (2007). Modeling the utilization of starch
by activated sludge for simultaneous substrate storage and microbial growth. Biotechnology
Bioengineering. 94 (1), 43–53.
[16] Beun, J.J., Paletta, F., van Loosdrecht, M.C.M. and Heijnen, J.J. (2000) Stoichiometry and
kinetics of polyhydroxybutyrate metabolism in aerobic, slow growing, activated sludge cultures.
Biotechnology Bioengineering. 67, 379-389.
[17] Dircks, K., Beun, J.J., van Loosdrecht, M.C.M., Heijnen, J.J. and Henze, M. (2001) Glyogen
metabolism in aerobic mixed culture. Biotechnology and Bioengineering. 73, 85-94.
[18] Karahan, G. O., Artan, N., Orhon, D., Henze, M. and van Loosdrecht, M.C.M. (2001).
Respirometric assessment of storage yield for different substrates. Water Science Technology.
46(1–2), 345–352.