Bioresource Technology 88 (2003) 131–136

Performance of a glucose fed periodic anaerobic baffled reactor

under increasing organic loading conditions: 1. Experimental results
K. Stamatelatou a, V. Vavilin b, G. Lyberatos a,*

a
Department of Chemical Engineering, University of Patras, 1 Karatheodori St., 26500 Patras, Greece
b
Water Problems Institute of Russian Academy of Sciences, Gubkina str. bd. 3, 119991 Moscow, Russia
Received 6 June 2002; received in revised form 7 October 2002; accepted 23 October 2002

Abstract
The influence of the organic loading rate on the performance of an innovative reactor, the periodic anaerobic baffled reactor
(PABR) was examined. A laboratory-scale PABR of four compartments being fed with a glucose based synthetic medium performed
with high stability while the feed organic load was doubled from 12.5 to 25 and then to 50 g COD/l. Finally the feed concentration
was increased to 75 g COD/l. The successive step changes in the feed concentration lasted for 20, 15, and 7 d, respectively. The COD
removal efficiency of the PABR was satisfactory in the first two transitions (
97.5 and 96%). In the third transition (OLR ¼ 18:75
g COD/l/d) the reactor failed as the pH dropped to 4. The concentrations of butyric and valeric acids increased as the organic
loading was increased and eventually they became greater than the concentration of acetic and propionic acids.
Ó 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Periodic anaerobic baffled reactor; Dynamic response; Organic loading rate; Glucose

1. Introduction they are present in high concentrations, have motivated

The development of anaerobic bioreactor technology
resulted in significant reduction of the required hy-
draulic retention times and allowed anaerobic digestion
to become a major alternative biological method of
wastewater treatment, that offers fundamental benefits
(Lettinga, 1995). The reactor stability against imposed
organic shock loads is a very important aspect in diges-
ter design, especially for agro-industrial wastes treat-
ment, because of the variable nature of the wastes
(Nachaiyasit and Stuckey, 1997a).
The response of the anaerobic process to overloading
depends on many factors, such as the size and duration
of the overloading, the type of the wastewater, the ac-
tivity state of the biomass, etc. Nevertheless, the com-
mon symptoms that follow an organic shock are an
increase in the dissolved COD concentration due to
volatile fatty acid (VFA) accumulation, and a concom-
itant drop in the pH if the system lacks the necessary
buffering capacity. The toxic effects of the VFAs, when
*
Corresponding author. Fax: +30-2610-99-30-70.
E-mail address: lyberatos@chemeng.upatras.gr (G. Lyberatos).
substantial research from which no specific conclusions
can be drawn as to the toxicity levels of the VFAs
(Angelidaki et al., 1993). It is the interrelation between
the different groups of microorganisms that enables the
anaerobic consortium to tolerate the increased levels of
VFAs, by providing a thermodynamically feasible car-
bon flow from the complex organic compounds to
methane and carbon dioxide.
The reactor design may influence and even modulate
the dynamics of different microorganism groups. This
is the case for some of the successfully applied high-rate
systems such as the upflow anaerobic sludge blanket re-
actor (UASB) (Lettinga et al., 1980) and the anaerobic
baffled reactor (ABR) (Bachmann et al., 1982, 1985). In a
UASB type reactor, the biomass tends to form granules.
The structure of the granules protects the more sensitive
microorganism groups from adverse changes in the en-
vironmental conditions such as organic overloading.
The ABR has been also shown to be very stable when
overloaded, a feature attributed to its compartmental-
ized structure. Cases of overloading were studied thor-
oughly by Nachaiyasit and Stuckey (1997a,b), who
reported the high COD removal efficiency (90%) of an
8-l ABR, when subjected to a pulse change in the feed

0960-8524/03/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 0 - 8 5 2 4 ( 0 2 ) 0 0 2 7 6 - 6
132 K. Stamatelatou et al. / Bioresource Technology 88 (2003) 131–136
concentration (corresponding to a change in the organic
loading rate from 4.8 to 18 g COD/l/d).
Fig. 1. (a) Front view of a four compartment PABR: Each compartment
(defined by solid lines) consists of a downflow and an upflow section.
The two sections are separated with a baffle (dotted line). The feed enters
compartment (A) through a port attached to the downflow section,
passes below the baffle, comes up in the upflow section, and then enters
the downflow section of the next compartment (B) through external
tubes (shown in (b)). The same flow pattern is repeated in all compart-
ments up to the last one (D). The direction of the flow inside the reactor
is depicted with dotted arrows and outside with solid arrows, while the
influent and the effluent with thick arrows. (b) Top view of a four com-
partment PABR: The compartments are defined by solid lines and the
baffles are depicted with dotted lines as in (a). 1; 2; . . . ; 12 are electronic
valves positioned on tubes outside the reactor. The valves change the
role of each compartment (influent, effluent or intermediate) over time at
a pre-selected switching period. In this paper, the valves switched every
1/2 d, so that the switching period (a full cycle of switching) was 2 d. Note
that the external tubes connect the upflow section () of the compart-
ment with the downflow section () of the subsequent compartment.
The flow direction is counterclockwise (solid arrow), but the switching of
the valves (role of each compartment) is clockwise (thick arrows), i.e.
after switching, the influent enters compartment D (instead of A) and the
effluent comes out from compartment C (instead of D) and so on.
Skiadas and Lyberatos (1998) developed a modified
version of the ABR, the periodic ABR (PABR). This
reactor is depicted in Fig. 1(a). It consists of four com-
partments (each consisting of a downflow and an upflow
part) arranged in a circular manner so that, for one
fourth of a given period of time (typically 2 d), each
compartment serves as the feed compartment, for two
fourths as an intermediate compartment and for one
fourth, it serves as the effluent compartment. A system
of 12 computer-controlled valves is used to change the
role of each compartment periodically (Fig. 1(b)). The
flow pattern from the influent to the effluent changes,
thus, with time ðtÞ, within a given switching period, T ,
as follows: A–B–C–D for 0 < t < T =4, D–A–B–C for
T =4 < t < T =2, C–D–A–B for T =2 < t < 3T =4 and B–
C–D–A for 3T =4 < t < T .
The frequency of compartment role switching (de-
fined as the inverse of the period) is an operational pa-
rameter that can be manipulated to optimize the PABR
performance. Skiadas et al. (2000) studied the effect of
switching frequency on PABR performance for various
dilution rates and concluded that, operation of the
PABR at high switching frequencies leads to higher
performance for high dilution rates, while the reverse is
the case for low dilution rates.
On an experimental basis, the behavior of a 15 l
PABR fed with glucose was studied by Skiadas and
Lyberatos (1998) at different switching periods (1, 1.5
and 2 d), but a constant organic loading rate (4.55
g COD/l/d). The mean COD removal was high in all
cases (>94%) as the organic loading rate was rather low.
The objective of this work was to assess the ability of
the PABR to handle high organic loading rates result-
ing from sudden changes in the feed concentration.
2. Methods
The reactor that was used was made of stainless steel
and consisted of a four compartment PABR with a total
useful volume of 15 l (Skiadas and Lyberatos, 1998).
The reactor was kept at 35 °C by being immersed in a
constant temperature bath. The reactor was completely
interfaced with a PC that regulated the valve switching.
The PABR was fed with a glucose based synthe-
tic medium, which also contained casein hydrolysate
(GibcoBril) and yeast extract (Merck), as sources of
amino acids and vitamins, in a proportion of 0.367 g/g C
of glucose as well as 5 g/l NaHCO3 , 10 ml/l of a solution
of (NH4 )2 HPO4 (0.721 g/l) and 10 ml/l of a solution with
trace metals (Table 1).
The dissolved COD concentration was measured as
described in standard methods (APHA, 1989). The
acetic, propionic, butyric and valeric acids concentra-
tions were measured by a gas chromatograph equipped
 K. Stamatelatou et al. / Bioresource Technology 88 (2003) 131–136 133

Table 1 30 8
Stock solution of the trace metals used in the feed synthetic medium of 25 A 6
the PABR 20 #1 #2 #3 42
Compound Concentration (mg/l) 15
10
CaCl2 2H2 O 22,500
5
NH4 Cl 35,900
0 8
MgCl2 6H2 O 16,200 25 B 6
KCl 117,000 20 #1 #2 #3 42

COD concentration, g/l

MnCl2 4H2 O 1800 15
CoCl2 6H2 O 2700 10
H3 BO3 513 5

pH
CuCl2 2H2 O 243 0 8
Na2 MoO4 2H2 O 230 25 C 6
ZnCl2 189 20 #1 #2 #3 4
2
NiCl2 6H2 O 200 15
H2 WO4 10 10
FeSO4 700 5
0 8
25 D 6
20 4
Table 2 #1 #2 #3 2
Operating conditions of the PABR during the experiment (HRT ¼ 4 d, 15
Switching period ¼ 2 d) 10
5
Phase COD concentration Organic loading rate Duration 0
in the feed (g COD/l) (g COD/l/d) (d) -2 0 20 24 28 32 36 40 44
Initially 12.5 3.125 Time, d
#1 25 6.25 20
#2 50 12.5 15 
Fig. 2. COD concentration (j), and pH ( ) during the transitions of
#3 75 18.75 7 the PABR under a step change in the organic loading (#1) 12:5 ! 25
g COD/l, (#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l (A, B, C and D
denote the different compartments, Fig. 1).
with a flame ionization detector. The pH was deter-
mined by a portable pH-meter.
The experiment was conducted at a constant hy- 100
draulic retention time (4 d) and switching period (2 d) 95
and consisted of three phases. In each phase, the feed 90
concentration was increased instantly to a new value 85
(Table 2) and was kept constant until the PABR reached 80
75
a steady periodic state (SPS). The term ‘‘SPS’’ is used
70
here to qualify the response of the PABR when it re- #1 #2 #3
65
mained the same from period to period. Samples were 60
taken every 12 h (one fourth of the period, 2 d) at which 55
the role of the compartments (first after the feeding 50
point, etc.) changed. -4 0 4 8 12 16 20 24 28 32 36 40 44
Time, d

Fig. 3. COD removal efficiency during the transitions of the PABR

3. Results and discussion
The PABR was subjected to three successive step
changes in the feed concentration. The results are shown
in Figs. 2–7. In the first transition the feed concentration
was increased from 12.5 to 25 g COD/l. Within a period
of 2 d, the response of each PABR compartment oscil-
lated between a higher COD value, observed when the
compartment was the feeding compartment, and a lower
COD value which occurred when the compartments was
the effluent compartment (Fig. 2). The amplitude of the
oscillations increased immediately after the disturbance
was imposed and reached a stable periodic state after the
eighth day. The experiment was continued until it was
under a step change in the organic loading (#1) 12:5 ! 25 g COD/l,
(#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l.
clear that a SPS was reached. The mean COD removal
was estimated to be
97.5% which was a slight decrease
from the initial value of 98.2%. The pH varied at normal
levels between 6.8 and 7.2.
In the second transition, the feed concentration was
further increased from 25 to 50 g COD/l. The PABR
showed a periodic behavior as in the first experiment,
but, as expected, the dissolved COD concentration val-
ues oscillated more intensely than before (Fig. 2). Nev-
ertheless, the effluent COD concentration remained low.
134 K. Stamatelatou et al. / Bioresource Technology 88 (2003) 131–136

3000 2500
2500 A 2000 A
2000 #1 #2 #3 #1 #2 #3
1500
1500
1000 1000
500 500
Acetic acid concentration, mg/l

0 0
2500 B 2000 B
2000 #1 #2 #3 1500 #1 #2 #3
1500
1000 1000
500 500
0 0
2500 C 2000 C
2000 #1 #2 #3 #1 #2 #3
1500
1500
1000 1000
500 500
0 0
2500 D 2000 D
2000 #1 #2 #3 1500 #1 #2 #3
1500
1000 1000
500 500
0 0
-2 0 20 24 28 32 36 40 44 -2 0 20 24 28 32 36 40 44
Time, d Time, d
Fig. 4. Acetic acid concentration during the transitions of the PABR Fig. 6. Butyric acid concentration during the transitions of the PABR
under a step change in the organic loading (#1) 12:5 ! 25 g COD/l, under a step change in the organic loading (#1) 12:5 ! 25 g COD/l,
(#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l (A, B, C and D denote (#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l (A, B, C and D denote
the different compartments, Fig. 1). the different compartments, Fig. 1).

2500 2500
2000 A #1 #2 #3 2000 A
#1 #2 #3
1500 1500
1000 1000
500 500
Propionic acid concentration, mg/l

0 0
Valeric acid concentration, mg/l

2000 B 2000 B
#1 #2 #3 #1 #2 #3
1500 1500
1000 1000
500 500
0 0
2000 C #1 #2 #3 2000 C
#1 #2 #3
1500 1500
1000 1000
500 500
0 0
2000 D #1 #2 #3 2000 D
1500 1500 #1 #2 #3
1000 1000
500 500
0 0
-2 0 20 24 28 32 36 40 44 -2 0 20 24 28 32 36 40 44
Time, d Time, d
Fig. 5. Propionic acid concentration during the transitions of the Fig. 7. Valeric acid concentration during the transitions of the PABR
PABR under a step change in the organic loading (#1) 12:5 ! 25 under a step change in the organic loading (#1) 12:5 ! 25 g COD/l,
g COD/l, (#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l (A, B, C and D (#2) 25 ! 50 g COD/l, (#3) 50 ! 75 g COD/l (A, B, C and D denote
denote the different compartments, Fig. 1). the different compartments, Fig. 1).
 K. Stamatelatou et al. / Bioresource Technology 88 (2003) 131–136
The resulting COD removal was almost 96% which was
quite satisfactory. The pH dropped to lower values, but
still continued to vary within the normal range of 6.5–
7.2. In the third transition, the organic loading increase
was chosen to be smaller than in the previous experi-
ments, because the possibility of attaining a new SPS,
while increasing the influent COD was judged to be
rather small. Under the organic loading of 75 g COD/l,
the dissolved COD concentration in the effluent com-
partment could not be reduced to low levels. This re-
sulted in continuously increasing values of COD in all
compartments. The pH dropped below 6 and eventually
reached a value of 4 in three of the four compartments
(A, C and D), which indicated an irreversible state of the
digester going sour. As a consequence, the PABR per-
formance was dramatically low (
70%) when the ex-
periment terminated (Fig. 3) and obviously it tended to
decrease more.
Since feeding was shared equally in all four com-
partments of the PABR, each compartment bore the
stress of the feed organic loading only during one fourth
of the switching period (set at 2 d). During the rest three
fourths of the switching period, the compartment was
allowed to alleviate the stress until it was time to become
the influent compartment again. The exposure of the
microorganisms to a high organic loading rate for a
prolonged time could enhance the inhibitory effect of the
acidic fermentation products on them and limit the rate
of their recovery. In this perspective the organic loading
rate of 18.75 g COD/l/d (transition #3) for the given
switching period of 2 d was too high. Shortening the
switching period could probably improve the PABR
performance, since it would result in smaller but more
frequent feeding intervals. This is supported by simula-
tion results obtained in previous works (Stamatelatou,
1999; Skiadas et al., 2000).
It should also be noted that the alkalinity of the feed,
provided in the form of NaHCO3 , was kept unchanged
in all three transitions and equal to 5 g/l which corre-
sponded to an alkalinity concentration of almost 3000
mg CaCO3 /l. Since no significant production of alka-
linity was expected during the anaerobic digestion of the
specific synthetic medium fed to the reactor, the buf-
fering capacity of the reactor proved to be insufficient in
the case of the third transition. As a result, the fluctu-
ations in the pH were intense and stopped the PABR
from attaining a stable operation.
The gradual increase in dissolved COD throughout
the experiments was due to the accumulation of the
(total) VFAs, which showed a similar periodic response
as well. However, it is interesting to examine how the
concentrations of the individual VFAs were built up.
Before implementing any change in the feed concentra-
tion, acetate and propionate were the only VFAs in
significant concentrations, with the latter being in higher
concentrations. During the first transition, both acids
135
increased but the propionate concentration was still
sensibly higher than acetate (Figs. 4 and 5). During the
subsequent transitions, acetate approached and slightly
overcame propionate, which was maintained at the same
levels. In the third transition, when the PABR showed
that it was becoming unstable, acetate and propionate
seemed to be steadily varying below 2000 mg/l each,
with no signs of built up.
The production of butyrate and valerate increased
during the first transition (Figs. 6 and 7), but it was
greater during the second transition, during which these
acids approached the levels of acetate and propionate.
Finally, during the third transition, butyrate and valer-
ate concentrations dramatically increased and prevailed
over the other measured acids, showing a tendency for
further accumulation (above 4000 mg/l each). Since the
time of pH dropping below 6 and the time of rapid
butyrate and valerate accumulation coincided, it seems
that the two facts were correlated.
A reasonable explanation may be that the anaerobic
microorganisms producing acidic fermentation prod-
ucts, under stressing conditions, prefer forming acids of
lower dissociation constant (such as butyrate and val-
erate) over acetate and propionate, as an attempt to
prevent further acidification. When organisms with a
mixed-acid fermentation pattern are present, a decrease
in their environmental pH often results in a shift from
acid to alcohol production. This possibility was, how-
ever, not investigated here. The shifts in fermentation
carbon and electron flow pathways following a decrease
in the pH are often the result of their effort to maintain
their intracellular pH. For this purpose, anaerobes
also use other mechanisms to adjust their cytoplasmic
conditions, including cellular morphology, membrane
structure and protein synthesis (Lowe et al., 1993).
The accumulation of butyrate and valerate was also
attributed to a potentially high concentration of hy-
drogen or formate, which renders the utilization of the
higher acids by the microorganisms thermodynamically
impossible. In such a case, the high concentrations of
butyrate and valerate are probably the results of process
imbalance.
In previous studies (Ahring et al., 1995; Hill and
Bolte, 1989), it was noted that butyrate, valerate and
their isomers were highly sensitive to disturbances and it
was proposed they be used as process indicators so as to
provide an early warning of an impeding process im-
balance. The sensitivity of these parameters was verified
in the present experiments as well, since they responded
immediately after every increase in the organic loading.
Especially during the second transition, the abrupt ac-
cumulation of these acids could be a sign that the system
was subjected to a severe stress. Nevertheless the PABR
operated in stability and showed an excellent COD re-
moval efficiency (96%). However, this organic loading
rate seemed to have brought the PABR close to its
136 K. Stamatelatou et al. / Bioresource Technology 88 (2003) 131–136
limits, under the particular strategy followed to stress
the reactor.
The PABR might be able to handle higher organic
loading rates, if the OLR were increased more gradually.
In this case, the biomass could have the time to adjust
better to the new conditions and the reactor would
perform more efficiently. When, however, there is a
necessity to change the organic loading rate abruptly
within a limited time interval, as is often the case in
practice, the PABR could be operated successfully as in
the case presented here, as long as the increase was not
excessive. The feasible increase in the OLR, while
maintaining stable operation, might be further in-
creased, by allowing pH control through alkalinity and/
or optimal selection of the switching period.
The long duration of PABR experiments before a
stable periodic state is reached makes this assessment
particularly difficult and time-consuming. Appropriate
mathematical models that can adequately describe the
PABR reactor could prove very useful in the effort to
assess the limiting OLR for a given waste type. Such a
model is developed in a companion paper (Stamatelatou
et al., in press).
4. Conclusions
The PABR, fed with an easily biodegradable sub-
strate (glucose), exhibited high stability under conditions
of high organic loading (12.5 g COD/l/d). This renders
the new suggested digester design very promising re-
garding the treatment of organically strong wastewaters.
The PABR, although subjected to a 100% increase in
the organic loading rate twice, it still operated with high
performance (>96%). A further increase in the organic
loading rate by 50% though, resulted in the reactor be-
coming sour. The PABR is expected to perform better
under the same conditions of organic loading if addi-
tional alkalinity is provided to the system or a more
gradual increase in the organic loading is followed, or if
the switching period is optimised.
The levels of butyrate and valerate increased sub-
stantially during the second transition and even more in
the last transition, while acetate and propionate did not
follow. As a result, it is verified that these VFAs could
serve as an indicator of reactor stressing, even at an
early stage, when the other indicators (such as COD
concentration, pH, etc.) are still within a normal range
of values.
Acknowledgement
This work was partly supported by NATO Collabo-
rative Linkage Grant under reference EST.CLG975008.
References
Angelidaki, I., Ellegaard, L., Ahring, B.K., 1993. A mathematical
model for dynamic simulation of anaerobic digestion of complex
substrates: focusing on ammonia inhibition. Biotechnol. Bioeng.
42, 159–166.
APHA (American Public Health Association, American Water Works
Association, Water Pollution Control Federation). Standard
Methods for the Examination of Water and Wastewater, 17th
ed., Washington, 1989.
Ahring, B.K., Sandberg, M., Angelidaki, I., 1995. Volatile fatty acids
as indicators of process imbalance in anaerobic digestors. Appl.
Microbiol. Biotechnol. 43, 559–565.
Bachmann, A., Beard, V.L., McCarty, P.L., 1982. Comparison of
fixed-film reactors with a modified sludge blanket reactor. Pro-
ceedings of the First International Conference on Fixed-Film
Biological Processes, II, pp. 1192–1211.
Bachmann, A., Beard, V.L., McCarty, P.L., 1985. Performance
characteristics of the anaerobic baffled reactor. Water Res. 19,
99–106.
Hill, D.T., Bolte, J.P., 1989. Digester stress as related to iso-butyric
and iso-valeric acids. Biol. Wastes 28, 33–37.
Lettinga, G., Hobma, S.W., Klapwijk, A., Van Velsen, A.F.M., De
Zeeuw, W.J., 1980. Use of the upflow sludge blanket (UAS) reactor
concept for biological wastewater treatment. Biotechnol. Bioeng.
22, 699–734.
Lettinga, G., 1995. Anaerobic digestion and wastewater treatment
systems. Antonie van Leeuwenhoek 67, 3–28.
Lowe, S.E., Jain, M.K., Zeikus, J.G., 1993. Biology, ecology and
biotechnological applications of anaerobic bacteria adapted to
environmental stresses in temperature, pH, salinity or substrates.
Microbiol. Rev. 57, 451–509.
Nachaiyasit, S., Stuckey, D.C., 1997a. The effect of shock loads on the
performance of an anaerobic baffled reactor (ABR), 1. Step
changes in feed concentrations at constant retention time. Water
Res. 31, 2737–2747.
Nachaiyasit, S., Stuckey, D.C., 1997b. The effect of shock loads on the
performance of an anaerobic baffled reactor (ABR), 2. Step and
transient hydraulic shocks at constant feed strength. Water Res. 31,
2747–2755.
Skiadas, I.V., Lyberatos, G., 1998. The periodic anaerobic baffled
reactor. Water Sci. Technol. 38 (8–9), 401–408.
Skiadas, I.V., Gavala, H.N., Lyberatos, G., 2000. Modelling of the
periodic anaerobic baffled reactor (PABR) based on the retaining
factor concept. Water Res. 34, 3725–3736.
Stamatelatou, K., 1999. Optimization of anaerobic digestion systems.
Thesis University of Patras, Greece (in Greek).
Stamatelatou, K., Lokshina, L., Vavilin, V., Lyberatos, G. Perfor-
mance of a glucose fed periodic anaerobic baffled reactor (PABR)
under increasing organic loading conditions: 2. Model prediction.
Bioresource Technology, in press.