the use of the anaerobic baffled

7/30/2019 The Use of the Anaerobic Baffled

1/20

REVIEW PAPER

THE USE OF THE ANAEROBIC BAFFLED REACTOR

(ABR) FOR WASTEWATER TREATMENT: A REVIEW

WILLIAM P. BARBER*M and DAVID C. STUCKEY**M

Department of Chemical Engineering and Chemical Technology, Imperial College of Science,Technology and Medicine, Prince Consort Road, London SW7 2BY, U.K.

(First received May 1998; accepted in revised form August 1998)

AbstractA review concerning the development, applicability and possible future application of the an-aerobic baed reactor for wastewater treatment is presented. The reactor design has been developedsince the early 1980s and has several advantages over well established systems such as the upow an-aerobic sludge blanket and the anaerobic lter. These include: better resilience to hydraulic and organicshock loadings, longer biomass retention times, lower sludge yields, and the ability to partially separatebetween the various phases of anaerobic catabolism. The latter causes a shift in bacterial populationsallowing increased protection against toxic materials and higher resistance to changes in environmentalparameters such as pH and temperature. The physical structure of the anaerobic baed reactor enablesimportant modications to be made such as the insertion of an aerobic polishing stage, resulting in areactor which is capable of treating dicult wastewaters which currently require several units, ulti-mately signicantly reducing capital costs. # 1999 Elsevier Science Ltd. All rights reserved

Key wordsanaerobic baed reactor, anaerobic digestion, reactor development, performance, solidsretention, molids odelling, full-scale.

INTRODUCTION

The successful application of anaerobic technology

to the treatment of industrial wastewaters is criti-

cally dependent on the development, and use, of

high rate anaerobic bioreactors. These reactors

achieve a high reaction rate per unit reactor volume

(in terms of kg COD/m3 d) by retaining the biomass

(Solids Retention Time, SRT) in the reactor inde-

pendently of the incoming wastewater (Hydraulic

Residence Time, HRT), in contrast to Continually

Stirred Tank Reactors (CSTRs), thus reducing reac-

tor volume and ultimately allowing the application

of high volumetric loading rates, e.g. 1040 kgCOD/m3 d (Iza et al., 1991). High rate anaerobic

biological reactors may be classied into three

broad groups depending on the mechanism used to

achieve biomass detention, and these are xed lm,

suspended growth, and hybrid. There are currently

900 full-scale installations in the world today

(Habets, 1996), and they are distributed as follows:

Upow Anaerobic Sludge Blanket (UASBsus-

pended growth) 67% (Lettinga et al., 1980); CSTR

12%; Anaerobic Filter (AFxed lm) 7% (Young

and McCarty, 1969); other 14%. The highest load-

ing rates achieved during anaerobic treatment to

date are attributed to the ``Anaerobic Attached

Film Expanded Bed'' (AAFEB) reactor (120 kg

COD/m3 d, Switzenbaum and Jewell (1980)), but its

inherent complexity and high operating costs limit

its practical use on a wide scale.

Around the same time as Lettinga developed the

UASB, McCarty and co-workers at Stanford

noticed that most of the biomass present within an

anaerobic Rotating Biological Contactor (RBC,

Tait and Freidman (1980)) was actually suspended,

and when they removed the rotating discs they

developed the Anaerobic Baed Reactor (ABR,

McCarty (1981)). However, baed reactor unitshad previously been used to generate a methane

rich biogas as an energy source (Chynoweth et al.,

1980). Although not commonly found on a large

scale, the ABR has several advantages over other

well established systems, and these are summarised

in Table 1.

Probably the most signicant advantage of the

ABR is its ability to separate acidogenesis and

methanogenesis longitudinally down the reactor,

allowing the reactor to behave as a two-phase sys-

tem without the associated control problems and

high costs (Weiland and Rozzi, 1991). Two-phase

operation can increase acidogenic and methano-genic activity by a factor of up to four as acido-

genic bacteria accumulate within the rst stage

Wat. Res. Vol. 33, No. 7, pp. 15591578, 1999# 1999 Elsevier Science Ltd. All rights reserved

Printed in Great Britain0043-1354/99/$ - see front matterPII: S0043-1354(98)00371-6

*Author to whom all correspondence should be addressed.[Tel. +44-171-594-5591; Fax: +44-171-594-5604, E-mail: [email protected]].

1559


2/20

(Cohen et al., 1980, 1982), and dierent bacterial

groups can develop under more favourable con-

ditions. The advantages of two-phase operation

have been extensively documented (Pohland and

Ghosh, 1971; Ghosh et al., 1975; Cohen et al.,

1980, 1982). These benets have catalysed the devel-

opment of other staged reactor congurations such

as the ``Multiplate Anaerobic Reactor'' (El-

Mamouni et al., 1992; Guiot et al., 1995), ``Upow

Staged Sludge Bed (USSB)'' (van Lier et al., 1994,

1996) and the ``Staged Anaerobic Filter'' (Alves et

all., 1997), all of which have showed considerable

potential for wastewater treatment. Disadvantages

of the baed reactor design at pilot/full-scale

include the requirement to build shallow reactors to

maintain acceptable liquid and gas upow vel-

ocities, and problems with maintaining an even dis-

tribution of the inuent (Tilche and Vieira, 1991).

However, despite its many potential advantages

over other high rate anaerobic reactor designs, and

the ever-increasing number of publications, there

has never been any attempt to collate all this infor-

mation in a review. Hence, the objective of thispaper is to review the currently available literature

on the ABR, focusing on reactor development, hy-

drodynamics, performance, biomass characteristics

and retention, modelling, full-scale operation and a

comparison with other well established alternatives.

Finally, based on the review, a closing section will

discuss future prospects for the ABR.

REACTOR DEVELOPMENT

The ABR is a reactor design which uses a series

of baes to force a wastewater containing organic

pollutants to ow under and over (or through) thebaes as it passes from the inlet to the outlet

(McCarty and Bachmann, 1992). Bacteria within

the reactor gently rise and settle due to ow charac-

teristics and gas production, but move down the

reactor at a slow rate. The original design is shown

in Fig. 1(C), although Fig. 1(A) is more commonlyrecognised. However, in order to improve reactor

performance several modications have been made

(Fig. 1(B and DJ)). The main driving force behind

reactor design has been to enhance the solids reten-

tion capacity, but other modications have been

made in order to treat dicult wastewaters (e.g.

with a high solids content, Boopathy and Sievers

(1991)), or simply to reduce capital costs (Orozco

(1997), Fig. 1(F)). A summary of the main altera-

tions is shown in Table 2.

In 1981, Fannin et al. (1981) added vertical baf-

es to a plug-ow reactor treating high solids sea

kelp slurry (Fig. 1(C)) in order to enhance the reac-

tor's ability to maintain high populations of slowly

growing methanogens, which were being replaced

by the inuent solids. With a constant loading rate

of 1.6 kg COD/m3 d methane levels increased from

30 to over 55% with a methane yield of 0.34 m3/kg

VSS after the baes were added. In a later study,

Bachmann et al. (1983) compared the performance

of two baed reactors before and after narrowing

the downow chambers and slanting the bae

edges (Fig. 1(A) and Table 2). Although methane

production rates and reactor eciency were

improved in the modied design, a decrease in the

methane content of the biogas was also noted.

Despite the alterations, the performance of bothreactors was inferior to an anaerobic lter and

rotating biological disc operated under the same

conditions. COD removal eciencies were 82, 92

and 90% for the modied bae, anaerobic lter,

and rotating biological disc reactors respectively.

The next major change occurred with the devel-

opment of the rst of several hybrid designs (Tilche

and Yang, 1987, Fig. 1(E)). The motivation behind

the alterations was based on enhancing solids reten-

tion for high strength wastewater treatment. The

reactor was signicantly larger than those used pre-

viously, and incorporated a solids settling chamber

after the nal compartment. Solids washed out

from the baed reactor were collected in thesettling chamber and subsequently recycled to the

rst compartment. Packing was also positioned at

the liquid surface of each compartment with ran-

domly packed Pall rings in the rst two chambers,

and a deeper, structured, modular corrugated block

which had a high voidage in the third chamber.

Bioocs, which became buoyant due to a reduction

in density caused by high gas production, were

retained in the rst chamber due to the packing.

Higher loading rates were possible with this struc-

ture due to minimal solids washout during elevated

gas mixing. Each gas chamber was separated per-

mitting the measurement of gas composition andproduction from each compartment. Although ben-

ecial in this regard, the separation of the gas can

Table 1. Advantages associated with the anaerobic baed reactor

Advantage

Construction1 Simple design2 No moving parts3 No mechanical mixing4 Inexpensive to construct5 High void volume6 Reduced clogging7 Reduced sludge bed expansion8 Low capital and operating costsBiomass1 No requirement for biomass with unusual settling properties2 Low sludge generation3 High solids retention times4 Retention of biomass without xed media or a solid-settlingchamber5 No special gas or sludge separation requiredOperation1 Low HRT2 Intermittent operation possible3 Extremely stable to hydraulic shock loads4 Protection from toxic materials in inuent5 Long operation times without sludge wasting6 High stability to organic shocks

William P. Barber and David C. Stuckey1560


3/20

also enhance reactor stability by shielding syn-

trophic bacteria from the elevated levels of hydro-

gen which are found in the front compartments of

the baed reactor.

In order to treat swine wastewater containing a

high content of small particulate material,

Boopathy and Sievers (1991) further modied the

baed reactor. The main problems associated with

the treatment of swine wastewater in a baed reac-

tor were the inability to produce a oating sludge

layer which would enhance solids retention, and,

the high velocities associated with the baes causedsignicant washout of solid material. Therefore, the

baed reactor was modied to reduce upow liquid

velocities and to accept whole waste. The rst com-

partment of a two-chamber unit was doubled in

size to 10 l and this was followed by a second com-

partment of 5 l (Fig. 1(G)). Performance character-

istics and solids retention capabilities were

compared with a three-chamber unit with equal

volume chambers. The additional chamber in the

three-compartment unit, together with physical

modications, provided a longer solids retention

time and superior performance than the reactor

with only two compartments. This was in contrast

to earlier ndings (Sievers, 1988), when no dier-ence was found in treatment eciency compared

with compartment number in unmodied reactors.

Fig. 1. Variations of the baed reactor. (A) Single gas headspace, (B) individual gas headspace, (C)vertical, (D) horizontal, (E) hybrid with settling zone, (F) open top, (G) enlarged rst compartment,(HJ) various packing arrangements: (H) up-comers, (I) down-comers, (J) entire reactor. Key:

W = Wastewater, B = Biogas, E = Euent, S = Solids, (shaded areas represent random packing).

The anaerobic baed reactor: a review 1561


4/20

The larger compartment in the two-compartment

reactor acted as a natural lter and provided su-perior solids retention for the small particles. The

reactor collected twice the amount of solid material

(20.9 g/l) than the reactor with three chambers. This

was further substantiated in the solids washout

data, which was lower in the two-compartment

reactor despite showing lower treatment eciency.

Further analysis showed that despite losing more

solids, the three-compartment reactor was more e-

cient at converting the trapped solids to methane.

REACTOR HYDRODYNAMICS

Flow patterns

The hydrodynamics and degree of mixing that

occur within a reactor of this design strongly inu-

ence the extent of contact between substrate and

bacteria, thus controlling mass transfer and poten-

tial reactor performance. In 1992, Grobicki and

Stuckey conducted a series of residence time distri-

bution studies by tracking the fate of an inert tracer

(Li+) in the euent of a number of baed reactors

(48 chambers), both with and without biomass, at

various HRTs, and incorporated the data into

``Dispersion'' and ``Tanks In Series'' models pre-

viously described by Levenspiel (1972). The models

provided a useful method to calculate the degree of

mixing and the amount of unused volume (known

as ``dead space'') within the reactor. They found

low levels of dead space (80% in a CSTR (Stuckey, 1983).

Dead space increased to 18% on the addition of

8 g VSS/l, however, no direct correlation between

hydraulic dead space and HRT could be drawn. At

low HRT, the presence of biomass had no signi-

cant eect on hydraulic dead space, which was

found to be a function of owrate and number ofbaes. This contrasted with biological dead space,

which was found to be a function of biomass con-

centration, gas production, and owrate, and which

increased with increasing owrates. At high loadingrates caused by low HRT, gas production as well as

increased owrates kept sludge beds partly uidised.

Therefore, the contradictory eects of hydraulic

and biological dead space prevented a correlation

being derived between HRT and overall dead space.

Biological dead space was established as the major

contributor to overall dead space at high HRT, but

its eect decreased at lower HRT since gas pro-

duction disrupted channelling within the biomass

bed. Severe channelling, caused by large hydraulic

shocks, was found to be benecial since most of the

biomass was not entrained in the ow, and this

resulted in low washout and a fast recovery in per-formance (Grobicki and Stuckey, 1992; Nachaiyasit

and Stuckey, 1997c). Nevertheless, investigations of

the hydrodynamics to date have not taken into

account various other factors which are probably

important, and these include biogas mixing eects,

viscosity changes due to extracellular polymer pro-

duction, and biomass particle size. In addition, no

work has been done of the rate at which solid par-

ticles/biomass move down the reactor.

Eect of euent recycle

Recycling the euent stream tends to reduce

removal eciency because the reactor approaches a

completely mixed system, and therefore the mass

transfer driving force for substrate removal

decreases despite a small increase in the loading

rate. The eect of loading rate and increasing re-

cycle ratios on performance is shown in Table 3.

Chynoweth et al. (1980) observed a positive eect

caused by recycling twenty percent of the euent,

when the methane yield increased by over 30%.

The addition of a recycle stream was also found to

alleviate the problems of low pH caused by high

levels of volatile acids at the front of the reactor,

and discourage gelatinous bacterial growth at the

reactor inlet for the treatment of a complex protein

carbohydrate wastewater (Bachmann et al., 1983).Another benet of recycle is the dilution of toxi-

cants and reduction of substrate inhibition in the

Table 2. Development of the ABR

Fig. Modication Purpose Ref.

1(C) addition of vertical baes to a plug-ow reactor

enhances solids retention to allowbetter substrate accessibility tomethanogens

Fannin et al., 1981

1(A) (i) down ow chambers narrowed (i) encourages cell retention in up owchambers

Bachmann et al., 1983

(ii) slanted edges on baes (40458) ( ii) rou tes ow t owa rds cent re of compartment encouraging mixing

1(E) (i) settling chamber (i) enhances solids retention Tilche and Yang, 1987(ii) packing positioned at top of eachchamber

(ii) prevents washout of solids

( ii i) s epar at ed ga s chamber s ( iii ) eas e and con tr ol of ga smeasurement, provides enhancedreactor stability

1( G) e nlar gemen t of r st chamber bet ter tr eat abil it y of hig h sol idswastewater

Boopathy and Sievers, 1991



5/20

inuent. (Bachmann et al ., 1985; Grobicki and

Stuckey, 1991).

From theoretical considerations, recycle should

have a negative eect on reactor hydrodynamics by

causing increased mixing (which encourages solids

loss, and disrupts microstructures of bacteria living

in symbiotic relationships (Henze and Harremoes,

1983)) and enhancing the amount of dead space

(Grobicki and Stuckey, 1992; Nachaiyasit, 1995). In

her thesis, Nachaiyasit (1995) showed that dead

space doubled to approximately 40% when the re-

cycle ratio was increased from zero to 2. The author

also reported a sudden loss of solids when the re-cycle ratio was doubled. Increasing recycle has also

been linked to an increase in the sludge volume

index using anaerobic lters (Matsushige et al .,

1990).

Mixing caused by recycle has also been found to

cause a return to single phase digestion, therefore

the benets arising from the separation of acido-

genic and methanogenic phases are partially lost.

Bachmann et al. (1985) noticed that methanogenic

activity was more uniformly distributed over the

whole reactor after recycle was used. The conse-

quences of this observation are scavenging bacteria

(such as Methanosaeta) will end up at the front ofthe reactor where harsh conditions of high substrate

concentration, high hydrogen partial pressure and

low pH will make them relatively inactive, and

poorly scavenging acid producing bacteria pushed

towards the rear of the reactor will be starved since

less substrate will be available. Nachaiyasit (1995)

discovered a fall in both gas production and

methane composition down the reactor when the re-

cycle ratio was increased.

The overall benets of recycle are unclear, and ul-

timately its use will depend on the type of waste

being treated. If pH problems are severe, the inu-

ent has high levels of toxic material, or high loading

rates are preferred then recycle will be benecial.

However, as can be seen in Table 4, the disadvan-

tages of recycle show that it should be used with

caution, and only when absolutely necessary.

REACTOR PERFORMANCE

Start-up

The overall objective of start-up is the develop-

ment of the most appropriate microbial culture for

the waste stream in question. Once the biomass has

been established, either as a granular particle or a

oc, reactor operation is quite stable. The import-

ant factors governing the start-up of anaerobic reac-

tors have been summarised in the literature

(Stronach et al., 1986; Weiland and Rozzi, 1991;

Hickey et al., 1991), and will not be discussed here.

A collection of data obtained during reactor start-up is shown in Table 5.

Initial loading rates should be low so that slow

growing micro-organisms are not overloaded, and

both gas and liquid upow velocities should be low

so that occulent and granular growth is encour-

aged. The recommended initial loading rate is ap-

Table 3. Reactor performance vs increasing recycle ratio

Recycle ratio Reactor volume (l) In uent COD (g/l) Organic loadingrate (kg/m3 d)

COD removal (%) Ref.

0 13 8 2.70 93b

Bachmann et al., 1985a

0 10 4 4.80 99 Nachaiyasit and Stuckey, 1995b

0.1 10 4 4.80 98 Nachaiyasit and Stuckey, 1995b


0.5 13 8 2.86 88c




2.2 13 8 3.85 81c



3 13 8 3.42 91 Bachmann et al., 1985a

5 13 8 5.76 77 Bachmann et al., 1985a

6 13 8 6.83 75 Bachmann et al., 1985a

9.6 13 8 11.01 68 Bachmann et al., 1985a

11.7 13 8 16.92 55 Bachmann et al., 1985a

13.8 13 8 17.62 60 Bachmann et al., 1985a

aRecycle ratios calculated from data supplied based on R = 0 for retention time of 71 h, organic loading rates converted from hydraulic

loading rates supplied.

b

Loading rates calculated from recycle ratio data.

c

Nutrient limited conditions.

Table 4. Advantages and disadvantages of euent recycle

Advantages Disadvantages

1 Front pH increased 1 Overall eciency reduced2 Reduction of inuent toxicity and substrateinhibition

2 Increased solids loss

3 Higher loading rates possible 3 Increased hydraulic dead space4 Better substrate/biomass contact 4 Disruption of bacterial communities and bioocs

5 Encourages one-phase digestion



6/20

proximately 1.2 kg COD/m3 d (Henze and

Harremoes, 1983), however, successful start-up of a

pilot scale ABR has been achieved at signicantly

higher primary loading rates (Table 5, Boopathy

and Tilche (1991)). Although Nachaiyasit (1995)

originally noted adequate performance with an in-

itial loading rate of 13 kg COD/m3 d, an accumu-

lation of intermediate products caused reactor

souring and eventual failure after two weeks of oper-

ation. A possible way to prevent failure by overload-

ing was employed in 1980 by Chynoweth and co-

workers. In order to stimulate the growth of methano-

genic archea, pulses of methane precursors (acetate

and/or an acetate/formate mixture) were added

directly before raising loading rates, and these wereeective in minimising the shock caused by a sudden

increase in organic loading. Alternative methods to

prevent failure include the adjustment of pH in the

rst compartment (Grobicki, 1989). A recent study

(Barber and Stuckey, 1997) has shown that maintain-

ing an initially long detention time (80 h) which is

reduced in a stepwise fashion during which time sub-

strate concentration is kept constant, provides greater

reactor stability and superior performance than a

reactor started-up with a constant and low detention

time coupled to a stepwise increase in substrate con-

centration. These ndings were linked to better solids

accumulation, promotion of methanogenic popu-

lations, and faster recovery to hydraulic shocks in thereactor started at the longer retention time.

Treatment applications

This section reviews the performance of the

baed reactor while treating a variety of waste-

waters, in particular, low and high strength, low

temperature, high inuent solids and sulphate con-

taining waste. Tables 6 and 7 and Fig. 2 summarise

the available literature.

Low strength treatment. Various authors have

treated low strength wastewaters eectively in the

ABR, as shown in Table 8. Dilute wastewaters

inherently provide a low mass transfer driving forcebetween biomass and substrate, and subsequently

biomass activities will be greatly reduced according

to Monod kinetics. As a result, treatment of low

strength wastewaters has been found to encourage

the dominance of scavenging bacteria such as

Methanosaeta in the ABR (Polprasert et al., 1992).

Hassouna and Stuckey (1998), have shown that no

substantial change occurred in the population of

acid producing bacteria down the length of a reac-

tor treating dilute milk waste, indicating the lack of

signicant population selection at low COD concen-

trations.

It appears that biomass retention is enhanced sig-

nicantly due to lower gas production rates

suggesting that low hydraulic retention times

(6 4 2 h) are feasible during low strength treat-

ment. Orozco (1988) noted decreasing overall gas

production with increasing HRTs, and this implied

possible biomass starvation in later compartments

at longer retention times. Another important conse-

quence of low retention times when treating dilute

wastewaters is an increase in hydraulic turbulence,

which can lower apparent Ks values (Kato et al.,

1997) thus enhancing treatment eciency.

Witthauer and Stuckey (1982) observed irregular

COD removal in baed reactors run at low loading

rates and long retention times when treating dilute

synthetic greywater. These problems were associated

with low sludge blankets (inoculum contained less

than 3 g VSS/l) caused after long periods of biomass

settling. Channels were formed within the low blan-

kets and this resulted in low gas productivity in most

of the sludge blanket except for around the channels.

Hence, biogas mixing was greatly reduced and this

resulted in minimal biomass/substrate mass transport.

In contrast, anaerobic lters, operated under the same

conditions, outperformed the baed reactors, even

after their suspended biomass was ushed out in a

hydraulic shock experiment. The authors rec-

ommended that when treating dilute wastewater,

baed reactors should be started-up with higher bio-

mass concentrations (than used in their study) inorder to obtain a suciently high sludge blanket (and

better gas mixing)in as short a time as possible.

Table 5. Start-up data for the ABR

LRa

initialTime

binitial

LRLR

increasedTime

increased LRLR

nalStart-uptimec (d)

InitialVSS (g/l)

Ref.

1 (ramp increase) 4 57 NGd

Boopathy and Sievers, 19912 (ramp increase) 20 >60 NG Bachmann et al., 19830.4 NG 0.53 NG 0.8 >60 NG Yang and Moengangongo, 19874.33 40 10.26 22 12.25 62 4.01 Boopathy and Tilche, 19911.2 7 2.4 10 4.8 77 8.77 Grobicki, 19890.97 NG NG NG 12.25 78 4.01 Boopathy and Tilche, 19922.2 90 2.6 135 3.5 90 NG Boopathy et al., 1988

13.04 failed 18 Nachaiyasit, 19954.35 failed 18 Nachaiyasit, 19951.2 NG 2.4 NG 4.8 >95 18 Nachaiyasit, 1995NG (ramp increase) 20 >100 NG Fox and Venkatasubbiah, 1996

1.2 53 2.4 24 4.8 128 18 Barber and Stuckey, 1997

aLR = loading rate in kg COD/m3 d.bThe amount of time spent at each loading rate (d).cStart-up time quoted is the time required forreactor to reach steady state.

dNG = data not given.



7/20

Table6.Performancedataonbaedreactorsystemsa

Substra

te

Volume(l)

Chambers

Biomass(gVSS/l)

InletCOD(mgCOD/l)

Loadingrate(kg/m

3/d)

CODremoval(%)

HRT(h)

Temperature(8C)

Ref.

Undilutedseakelp

9.8

5

600036,000

0.42.4

360

35

Chynowethet

al.,1980

Diluted

seakelp

10

4

1.6

35

Fanninet

al.,1981,1982

10

4

67,20089,600

5.66.4

288336

35

10

4

80,000

1.6

1200

35

Carboh

ydrateprotein

6.3

5

71007600

220

7982

35

Bachmannet

al.,1983

Syntheticgreywater

8

6

480

0.10.4

6384

4884

2533

WitthauerandStuckey,1982

Carboh

ydrateprotein

6.3

5

8000

2.536

5593

4.871

35

Bachmannet

al.,1985

Diluted

swinewastewater

20

70

H140

37

Xinget

al.,1991

Carboh

ydrateprotein

10

8

18

4000

1.24.8

98,93

20,80

35

NachaiyasitandStuckey,1995

Pharmaceuticalwastewater

10

5

20,000

20

3668

24

35

FoxandVenkatasubbiah,1996

Phenolic

5

2025

22003192

1.672.5

8394

H24

21

Holtet

al.,1997

Glucose

6

5

100010,000

220

7299

12

35

Baeet

al.,1997

Carboh

ydrateprotein

10

8

18

10004000

1.24.8

98

2080

35

BarberandStu

ckey,1998

Domesticsewage/industrialwaste

394,000

8

315c

0.85

70

10.3

15

Orozco,1997

Carboh

ydrateprotein

10

8

18

4000

1.24.8

7583,9397,962

0,20,20

15,25,35

NachaiyasitandS

tuckey,1997a

Carboh

ydrateprotein

10

8

18

4000

4.89.6

9098

20

35

NachaiyasitandS

tuckey,1997b

Carboh

ydrateprotein

10

8

18

4000

4.818

5298

120

35

NachaiyasitandS

tuckey,1997c

aContainscalculatedresults,eitherfromgraphsorfrom

supplieddata.

bAlsowithshockloadingof96kg/m3

d.cBOD5

value.



8/20

High strength treatment. Whereas low retention

times are possible and even necessary for dilute

wastewaters, the opposite applies when treated con-

centrated waste. This is mainly due to the high gas

mixing caused by improved mass transfer between

the biomass and substrate. This will result in high

biomass wastage, and has led to modications in

the reactor design in order to enhance solids reten-

tion (see Section 2). A brief summary of the litera-

ture available on high strength treatment is shown

in Table 9.

When Boopathy and Tilche (1991) changed the

inuent to a 150 l hybrid reactor from 115 g

COD/l molasses alcohol stillage with a loading

rate of 12.25 kg/m3 d to raw alcohol molasses

(990 g COD/l, OLR = 28 kg COD/m3 d) they

noticed an increase in overall gas production of

over 65% within 3 weeks, a drop in COD

removal of 20%, a fall in the methane compo-

sition of the biogas by 20% for one week which

then recovered (implying initial overloading of

methanogens), and an approximate increase in

volatile suspended solids of 50% within 3 weeks.

Higher levels of gas production increased sludge

bed expansion, but the improved settling ability

of the biomass may have reduced the eects of

solids loss caused by the gas (Boopathy and

Tilche, 1991). This observation was partially con-

rmed in an earlier study (Boopathy et al., 1988)

where no increase in solids loss or decrease in

performance were noted when loading rates were

increased from 2.6 to 3.5 kg COD/m3 d. However,

minimal solids were lost to the euent at equally

low loading rates in the work by Boopathy and

Tilche (1991), but levels increased to 17 g VSS/l

at higher loading rates. (The reactor contained

approximately 1.25 g VSS/l of reactor in a 150 l

volume.) According to kinetic considerations, high

substrate concentrations will encourage both fast

growing bacteria, and organisms with high Ksvalues, and methane production will be derived

mainly from acetate decarboxylation by

Table 7. Potential methane yields from baed reactors

Wastewater OLR (kg/m3

d) Methane yield (m3

/kg VSS/d) Ref.

Swine manure 48 0.761.28 Boopathy and Sievers, 1991Swine manure 1.8 0.27 Yang and Moengangongo, 1987Carbohydrate/protein 4.8 0.11 Nachaiyasit and Stuckey, 1995Carbohydrate/protein 4.8 0.22 Grobicki, 1989Sea kelp 2.4 0.35 Chynoweth et al., 1980Molasses 20 1.25 Boopathy and Tilche, 1991Phenol 1.672.5 0.260.34 Holt et al., 1997Slaughterhouse 1.824.73 0.130.18 Polprasert et al., 1992

Fig. 2. Performance eciency against various loading rates.



9/20

Methanosarcina sp. and hydrogen scavenging

methanogens (such as Methanobrevibacter and

Methanobacterium). Subsequently Methanosarcina

sp. was observed as the dominant bacterial species

in bioocs formed during high strength treatment

(Boopathy and Tilche, 1991). (See Section 5.2.)

Low temperature treatment. At low/ambient tem-

peratures van Lier et al. (1996), found signicant

advantages with respect to reactor performance for

staged reactors when compared with completely

mixed systems. From Table 6 it can be seen that

the vast majority of work done so far on the baed

reactor has been conducted in the mesophilic tem-

perature range. However, the baed reactor has

been run as low as 138C (Orozco, 1988), although

the most extensive study at low temperatures in the

baed reactor was carried out by Nachaiyasit andStuckey (1997a, Table 10).

Generally, biochemical reactions double in rela-

tive activity for every 108C increase in temperature

in accordance with the van `t Ho rule over a

restricted temperature range. In spite of this,

Nachaiyasit (1995), found no signicant reduction

in overall COD removal eciency when the tem-

perature of an ABR was dropped from 35 to 258C,

with steady state reached after only two weeks.

However, lower catabolic rates caused by elevated

Ks values (according to Arrhenius kinetics) at the

front of the reactor caused a shift in acid pro-

duction towards the rear, although overall removal

was unaected. An increase in VFA production

caused a simultaneous reduction in pH and an in-

itial increase in gas phase hydrogen that quickly

returned to below background levels. The deeper

penetration of the VFAs down the reactor should

potentially improve the growth yields of the metha-

nogens in the latter compartments. The results

showed that slower growing organisms exhibited a

greater sensitivity to a fall in temperature compared

to bacteria with faster growth kinetics, and this is

in accordance with literature ndings (Cayless et

al., 1989; Kotsyurbenko et al., 1993; Borja et al.,

1994; Speece, 1996). Similar high treatment ecien-

cies at ambient temperature have also been noted

for a medium strength phenolic wastewater (Holt et

al., 1997).

Nachaiyasit and Stuckey (1997a) further reduced

the temperature to 158C, and a fall in overall e-

ciency of 20% was noted after one month. Changes

in performance down the reactor occurred over a

long period of time in contrast to CSTRs. This is

advantageous since the slow response would inher-ently provide more protection to shocks than in

other reactor systems. However, despite the fact

that the reactors were kept for long periods of time

at reduced temperatures (12 weeks) their perform-

ance did not improve despite the increased inter-

mediate acid concentrations, which according to

Monod kinetics should encourage more biomass

growth to compensate for the increased substrate

levels. This may be due to the fact that Ks increases

substantially as temperature falls, (Lawrence and

McCarty, 1969) leaving low levels of VFAs that

cannot be degraded.

This study also found that the fraction of VFAs

in the euent in terms of COD had reduced signi-

cantly. VFAs contributed to approximately a third

of the COD at 158C, and two thirds at 258C, indi-

cating that the production of refractory material

(termed as Soluble Microbial Products (SMPs),

Table 8. Selected low strength performance data

Wastewater HRT (h) COD (mg/l) COD removal (%) OLR (kg/m3/d) Gas produced (v/v/d) Ref.

inuent euent

Greywater 84 438 109 75 0.13 0.025 Witthauer and Stuckey, 1982Greywater 48 492 143 71 0.25 0.05 Witthauer and Stuckey, 1982Greywater

a84 445 72 84 0.13 0.025 Witthauer and Stuckey, 1982

Sucroseb

6.8 473 74 74 1.67 0.49 Orozco, 1988Sucrose

b8 473 66 86 1.42 0.43 Orozco, 1988

Sucroseb

11 441 33 93 0.96 0.31 Orozco, 1988Slaughterhouse 26.4 730 80 89 0.67 0.72 Polprasert et al., 1992Slaughterhouse 7.2 550 110 80 1.82 0.33 Polprasert et al., 1992Slaughterhouse 2.5 510 130 75 4.73 0.43 Polprasert et al., 1992

aTemperature at 258C.

bTemperatures lower than 168C. All other work shown in table performed in a mesophilic temperature range

Table 9. Selected high strength treatment data

Wastewater Raw molasses Molasses alcohol stillage Swine waste Whisky distillery

Inuent COD (g/l) 990 115.8 58.5 51HRT (h) 850 138636 360 360Reactor volume (l) 150 150 15 6.3Temperature (8C) 37 37 35 30OLR (kg/m

3d) 28 4.320 4 2.23.46

COD removal (%) 50 7088 6269 >90Biogas production (v/v/d) >5 >2.3 2.93.2 1.23.6Ref. Boopathy and Tilche, 1991 Boopathy and Tilche, 1992 Boopathy and Sievers, 1991 Boopathy et al., 1988



10/20

Rittmann et al. (1987)) increased substantially at

lower temperatures. In conclusion, the work found

that a combination of decreased catabolic rates,

increased Ks, and higher levels of refractory ma-

terial caused inferior performance at 158C, but that

a drop in temperature from 35 to 258C had negli-

gible eects on overall reactor performance despitepredictions from the van `t Ho rule. This has been

observed before in biolm/oc based reactors where

mass transfer limited biomass activity (Hickey et

al., 1987). However, Nachaiyasit's work did not

consider the eects of nutrient (especially iron)

bioavailability, which may be reduced at lower tem-

peratures (Speece, 1996), nor did it investigate the

signicance of temperature on ionisation equilibria

which inevitably controls the potential toxicity of

materials, some of which may be tolerated at higher

temperatures (Sawyer et al., 1994).

High solids treatment. In early work, Chyno-

weth's group in Illinois (1980, 1981) used baed

reactors to generate methane from sea kelp as an

alternative energy source. Although the COD of the

kelp was not quoted, the feed contained 15% total

solids, which were ground and chopped. Practical

problems associated with feeding solids were over-

come by applying the substrate by syringe. During

a particular run, signicant solids build-up was

observed in the rst compartment after 2 weeks of

operation. The solids build-up reduced micro-

organism contact with the substrate therefore mini-

mising hydrolysis and subsequent bioconversion.

Performance was signicantly improved after manu-

ally agitating the reactor for a short time period.

Solid material was also found to physically displacebiomass within the reactor indicating that modi-

cations to the ABR would be required for high

solids treatment.

In 1991, Boopathy and Sievers modied the

baed reactor (see Section 2) to treat high strength

swine waste (see Table 6) containing 51.7 g/l total

solids. When a loading rate of 4 kg COD/m3 d with

a retention time of 15 d was applied, removal rates

for COD (70 and 80%), and total solids (60 and

74%) were achieved for two- and three-compart-

ment reactors respectively. Solids retention times

were experimentally determined to be over 20 d in

both reactors. The study found that the majority ofthe protein fraction of the solids was retained

within the reactor, compared with a lower retention

of cellulose/hemicellulose, and a virtual loss of all

lipid material, although the authors oered no ex-

planation to the cause. Previous work in the same

laboratory had shown protein to be dicult to

degrade but a great potential source of methane,

hence its detention proved to be signicant in reac-

tor performance.Sulphate treatment. Fox and Venkatasubbiah

(1996), investigated the eects of sulphate reduction

in the ABR by treating a sulphate containing phar-

maceutical wastewater up to a nal strength of 20 g

COD/l with a COD:SO4 ratio of 8:1. At steady

state, 50% COD removal and 95% sulphate re-

duction was possible with a detention time of 1 day.

Reactor proles showed that sulphate was almost

completely reduced to sulphide within the rst

chamber, and a concomitant increase in sulphide

levels down the reactor indicated that sulphate

redirected electron equivalents to hydrogen sulphide

in preference to methane.

After altering the COD:SO4 ratio by adding glu-

cose, isopropanol and sulphate, the authors noted a

fall in potential sulphate reduction from >95% at

COD:SO4=150:1 to


11/20

BIOMASS CHARACTERISTICS AND RETENTIONCAPABILITIES

Bacterial populations

With the unique construction of the ABR various

proles of microbial communities may develop

within each compartment. The microbial ecology

within each reactor chamber will depend on the

type and amount of substrate present, as well as

external parameters such as pH and temperature. In

the acidication zone of the ABR (front compart-

ment(s) of reactor) fast growing bacteria capable of

growth at high substrate levels and reduced pH will

dominate. A shift to slower growing scavenging

bacteria that grow better at higher pH will occur

towards the end of the reactor.

Various techniques have been applied to describe

the population dynamics within the ABR, and the

results are summarised in Table 11. By far the most

common observation involved the shift in popu-

lation of the two acetoclastic methanogens

Methanosarcina sp. and Methanosaeta sp. At high

acetate concentrations Methanosarcina outgrows

Methanosaeta due to faster growth kinetics (dou-

bling time 1.5 d compared with 4 d for

Methanosaeta), however, at low concentrations

Methanosaeta is dominant due to its scavengingcapability (Ks=30 mg/l compared with 400 mg/l for

Methanosarcina (Gujer and Zehnder, 1983)).

Tilche and Yang (1987) and Yang et al. (1988)

compared the performance and bacterial popu-

lations of an anaerobic lter and a Hybridised

Baed Reactor (HABR) at pilot scale treating mol-

asses wastewater with maximum loading rates of

10.5 and 5.5 kg COD/m3 d for the anaerobic lter

and HABR respectively. The major ndings of the

study were: a large concentration of Methanosarcina

at the front of the baed reactor which shifted to

Methanosaeta towards the rear, compared with a

domination of Methanosaeta in the lter reactor,

and, hydrogen scavenging Methanobacterium were

observed at the front of the baed reactor using

epiuorescence microscopy.Explanations were oered to describe the lack of

Methanosarcina in the lter reactor. Firstly, the

acetate loading in the rst chamber of the HABR

was 1000 mg/l which might be close to the apparent

Ks value for Methanosarcina (data not given) and

therefore may have favoured its growth. In con-

trast, acetate levels were 10 times lower in the lter

reactor and therefore Methanosaeta had a kinetic

advantage and dominated in the reactor. Secondly,

lower supercial gas production rates in the baed

reactor (5 m/d in the rst compartment of the

HABR compared with 9 m/d in the lter) resulted

in lower gas turbulence, and therefore fewer wash-outs of bioocs compared with the anaerobic lter.

Hydrogen levels were also measured, and the high-

Table 11. Bacterial observations in the ABR

No. Observations Technique Ref.

1 Methanosarcina predominant at frontof reactor with Methanosaeta foundtowards rear

SEM, TEM, LLM Boopathy and Tilche, 1991, 1992;Tilche and Yang, 1987; Garuti et al.,1992; Yang et al., 1988

2 act iv e met han og enic fr act io n wit hi nbiomass highest at front of reactor andlowest in last chamber

ATA Bachmann et al., 1985; Orozco, 1988

3 bacteria resembling Propionibacterium,Syntrophobacter andMethanobrevibacter found in closeproximity within granules

TEM Grobicki, 1989

Methanosaeta and colonies ofSyntrophomonas also observed

4 large numbers of Methanobacterium atfront of ABR along withMethanosarcina covered granules;subsequent chambers consisted ofMethanosaeta coated ocs

EP Tilche and Yang, 1987

5 vi rt uall y al l biomass act iv ity (>8 5%)occurred in the bottom third of eachcompartment where biomass wasconcentrated; highest activity (92%)found in bottom of rst chamber

ATPA Xing et al., 1991

6 mainly Methanosaeta observed withsome cocci; no Methanosarcinaobserved

SEM Polprasert et al., 1992

7 irregular granules with gas ventscovered by single rod shaped bacteria;no predominant species observed

SEM Holt et al., 1997

8 bacteria resemblingMethanobrevibacter, Methanococcus,and Desulfovibrio found

ATPA, SEM, EP B oopa thy and Til che, 1 99 2

9 wide va rie ty of bact eri a ob serv ed atfront of reactor

SEM, TEM Boopathy and Tilche, 1991; Barberand Stuckey, 1997

Abbreviations: ATA = anaerobic toxicity assay, ATPA = ATP analysis, EP = (phase contrast) epi uorescence microscopy,LLM = light level microscopy, SEM = scanning electron microscopy, TEM = transmission electron microscopy.



12/20

est concentrations (4 104 atm) were noted in the

rst chamber of the baed reactor, and this may

explain the presence of Methanobacterium. The

results were subsequently supported by Polprasertet al. (1992) where acetate concentrations as low as

20 mg/l enabled the domination of Methanosaeta-

like bacteria throughout a four-compartment reac-

tor.

Biomass activity

Tilche and Yang (1987) and Yang et al. (1988)

also discovered that 70% of all methane produced

in the HABR came from the rst compartment,

despite having only 10% of the VSS present within

the reactor, and these ndings supported previous

work (Bachmann et al ., 1985; Orozco, 1988).

Bachmann used a procedure based on the

Anaerobic Toxicity Assay (ATA, Owen et al .

(1979)) and discovered that the active fraction of

acetate utilising methanogens as a percentage of the

total VSS varied from 5.7 to 1.8%, with the largest

values obtained at the front of the reactor and the

lowest at the rear. In a study involving an 11-com-

partment open top baed reactor treating 500 mg/l

sucrose at low temperature (13168C), Orozco

(1988) quoted activities of 1.43 g COD-CH4/m3 in

the rst seven chambers and 0.72 in chambers 7 to

11.

Xing et al . (1991), and Boopathy and Tilche

(1992) used ATP analysis to determine the relative

position of the most active bacteria. Samples weretaken from the top, middle and bottom of all three

chambers from a reactor with a working volume of

150 l treating molasses wastewater at a loading rate

of 20 kg COD/m3 d. The results showed that at

least 85% of the activity came from the bottom of

each compartment with the highest activity (92%)

measured at the base of the rst compartment.

However, the opposite trend was found in a study

treating slaughterhouse wastewater (Polprasert et

al., 1992). The reasons for this may lie in the con-

centration of intermediates, especially acetate, at

the front of the reactor. In studies where methane

activity was higher in the front compartments

(Bachmann et al., 1985; Tilche and Yang, 1987; and

Yang et al., 1988), acetate concentrations were rela-

tively high and therefore provided the best environ-

mental conditions for the growth of Methanosarcina

which can grow quickly and eciently even at pH

values as low as 6 (Speece, 1996). Another source

of methane would be from hydrogen scavenging

bacteria such as Methanobacterium (Tilche and

Yang, 1987) and Methanobrevibacter, which would

be stimulated by the higher hydrogen concen-

trations; the net eect would be a high methano-

genic activity. In contrast, with dilute wastewaters,

where acetate levels are low in the front compart-

ment (as in the work by Polprasertet al), the likelyscenario is that Methanosaeta would dominate.

However, this species grows at a far slower rate

compared to Methanosarcina and is also far more

sensitive to environmental conditions such as a

reduced pH. This would encourage the growth of

acid producing bacteria that would inevitably leadto a reduction in methane potential.

Hassouna and Stuckey (1998) showed a shift in

the activity of acid producing bacteria down the

length of an eight-compartment baed reactor.

Using the method of Owen et al. (1979), aliquots

were removed from each compartment of ABRs

treating a range of substrate concentrations. In the

foremost compartments a glucose spike was readily

converted to volatile acids within a few hours and

this contrasted with the results from subsequent

compartments which showed virtually no degra-

dation of the spike.

Granulation (and oc sizes)

Although granulation is not necessary in the

ABR for optimum performance, unlike suspended

systems such as the UASB, various reports have

noted the appearance of granules in the reactor.

Boopathy and Tilche (1991) started up HABRs (the

inoculum contained 4.01 g VSS/l) with a low initial

loading rate (0.97 kg COD/kg VSS d) and liquid

upow velocities below 0.46 m/h, in order to encou-

rage the growth of occulent and granular biomass.

Subsequently, stable granules of 0.5 mm appeared

after one month in all chambers of the reactor and

they were reported to be growing although no data

was given; microscope studies subsequently showedthat the granules were comprised primarily of acet-

oclastic methanogens. Similarly, Tilche and Yang

(1987) found Methanosarcina coated ocs held

together by brous bacteria resembling

Methanosaeta. The ocs, which were formed after

one month, were small with diameters less than

1.5 mm and were weak. Under the same loading

conditions the authors also found densely packed

granules typical of a UASB (d< 3 mm) formed

after 3 months in an anaerobic lter.

Boopathy and Tilche (1992) noticed similar par-

ticles of both types described above, which grew

from 0.5 mm after one month to 3.5 mm after three

months in a hybrid reactor. Granules, which weremade from Methanosarcina clusters, were of low

density and full of gas cavities and therefore lifted

to the surface of the reactor due to high gas and

liquid velocities during high loading. The particle

size appeared to be partially dependent on substrate

type. There was little dierence in particle size

throughout the reactor when molasses alcohol stil-

lage wastewater was treated. However, two weeks

after the substrate was altered to raw molasses with

a ten-fold increase in inlet COD a prole emerged

which showed a steady decrease in particle size

down the reactor. In addition, the sludge weight

increased from


13/20

granule size from 5.4 mm in the rst chamber down

to 1.5 mm in the last chamber of a reactor treating

dilute carbohydrate waste. However, on a labora-

tory scale, (Barber and Stuckey, 1997) oc sizeseemed to grow to a maximum near the centre of

an eight-compartment reactor and then reduce

towards the rear. Typical oc sizes were 100, 230

and 175 mm in the front, middle and rear compart-

ments respectively. These authors postulated that

the oc size was a function of both gas production

and COD concentration, with the largest particles

growing when COD concentrations were suciently

high to support growth, and gas production low

enough to avoid oc breakage.

Solids retention capability

By using a chromic oxide sesqui tracer in a highsolids swine wastewater (51 g/l), Boopathy and

Sievers (1991) managed to measure the solids reten-

tion time for two hybrid reactors running at a

hydraulic retention time of 15 d. A three-compart-

ment reactor resulted in a solids retention time of

25 d compared with 22 d for a two-compartment

unit. The two-compartment reactor had a larger in-

itial chamber, and this provided a natural ltering

action that enabled it to lose fewer solids to the

euent. Despite this, the three-compartment reactor

was found to be more ecient at converting the

trapped material into methane on the basis of cellu-

lose, lipid and protein measurements.

In a comparative study, Orozco (1988) calculated

the minimum solids retention time required to

achieve certain removal eciencies in baed and

UASB reactors under the same loading conditions,

and concluded that the solid residence time in the

UASB would have to be approximately 40% higher

than the ABR in order to achieve the same removal

rate. By assuming a series of perfectly mixed reac-

tors, Grobicki and Stuckey (1991), calculated the

solids retention times, biomass yield, and washout

of biomass under several experimental conditions.

Solids retention times varied from 7 to over 700 d

(5 < 80 h) and large deviations in the results were

attributed to varying degrees of granulation.Although a strong correlation was found to exist

between the solids retention time and HRT, the

authors suggested that caution should be exercised

when using the calculated gures due to the

assumptions of perfectly mixed behaviour. Solids

retention times of 300 d were reported by Garuti etal. (1992) using a 350 l reactor with a 15 h retention

time and this gure is far higher than those calcu-

lated by Grobicki and Stuckey (1991) under similar

conditions. These authors also calculated from the-

ory and a mass balance, that the observed yields

were very low (approximately 0.03 kg VSS/kg

COD), which implies constant biomass concen-

tration proles over time, but these ndings are in

contrast to other researchers (Boopathy and Tilche,

1991; Xing et al, 1991).

Boopathy et al. (1988) discovered that increasing

the loading rate from 2.2 to 3.5 kg COD/m3 d made

no signicant dierence to the amount of solids lost

to the euent, with a maximum of 0.5 g/l occurring

during start-up. These results were further sup-

ported in a hybrid reactor (Boopathy and Tilche,

1991) where virtually negligible euent VSS was

found with loading rates between 6 and 12.5 kg

COD/m3 d. However, a linear increase up to 17 g

VSS/l at high loading rates (28 kg COD/m3 d) was

observed. A similar correlation was also found to

exist between the Sludge Volume Index (SVI) and

the total solids lost from a pilot scale reactor

(Garuti et al., 1992). Finally, in a recent study,

Barber and Stuckey (1997) found that twice as

many solids were lost during start-up by a reactor

running at a low HRT of 20 h compared with onewhich was run on the same feed at long retention

times (80 4 40 4 20 h), and this was linked to in-

ferior COD removal since biomass accumulated fas-

ter in the reactor run at longer retention times.

MODELLING

Bachmann et al. (1983) found similar treatment

behaviour under identical conditions in an ABR,

anaerobic lter and a rotating biological disc reac-

tor. In order to predict reactor performance, an

attempt was made to develop a unied model for

the xed lm reactors and also for the ABR. Theauthors considered the sludge particles found within

the sludge bed of the ABR to be uidised spheres

Table 12. Model equations for ABR systems

No. Substrate model equations Ref.

1 dS/dt = aCSq

+QS0QS, S= S0(a/Q)CSq

Bachmann et al., 19832 Df(d

2Sf/dz2) = ( kSfXf)/(Ks+Sf) Bachmann et al., 1985

3a Sn=S0/[(1 + k1W1/Q)(1 + k2W2/Q) F F F(1 + knWn/Q)] Xing et al., 19913b Sn=[S0(1 + R)

n 1]/[(1 + R + k1W1/Q)(1 + R + k2W2/Q) F F F(1 + R + knWn/Q) (1 + R)n 1R] Xing et al., 1991

4 Df[(d2Sf/dr

2) + (2/r)(dSf/dr)] = (kXfSf)/(Ks+Sf) Nachaiyasit, 1995

Nomenclature: a = surface area per unit reactor volume (L1), C= variable-order reaction coecient, Df=molecular diusivity in bio-lm (L

2t1), k = maximum specic rate of substrate utilisation (MsMxt1

), Ks=half-velocity constant (ML3

), Q = specic ow rate(T

1), q = variable-order reaction order, r = radius of a three-dimensional spherical particle (L), R = recycle ratio, S= substrate concen-

tration (ML3), S0=inuent concentration (ML3), Sf=substrate concentration in biolm (ML

3), Sn=euent substrate concentration(ML

3), W= mass of sludge = volume/[Xf] (M), Xf=bacterial density (ML

3), z = distance normal to biolm surface (L). Numericalsubscripts refer to compartment number.



14/20

with a surface area through which the solute must

diuse for bacterial consumption. Therefore, they

used a combination of a xed lm model

(Williamson and McCarty, 1976) along with a vari-able order model (Rittmann and McCarty, 1978)

which incorporated the concepts of liquid-layer

mass transfer, Monod kinetics, and molecular diu-

sion to accurately describe the process (Table 12).

Two dierent approaches were employed; the rst

was based on the concept of a rate limiting sub-

strate (assumed to be acetate and propionate) dif-

fusing into a ``deep'' xed bacterial lm.

Application of the model was made possible by esti-

mating the specic surface area in each of the reac-

tor chambers from a data set, and then applying

the results to simulate behaviour at dierent load-

ings. Although initial predictions were good, the

model underestimated the level of COD removal at

higher loading rates. The reasons for the discrepan-

cies were given to be an unrealistic assumption of a

constant diusion layer depth which would decrease

at higher loading rates due to increased gas pro-

duction, thereby improving substrate/biomass con-

tact and ultimately reactor performance.

The second evaluation was made by assuming a

series of completely mixed dispersed growth reac-

tors using Monod kinetics. Here values of the active

micro-organism concentration were determined

within each compartment for one loading, and the

data applied to the same loading rates as with the

``xed lm'' model. The results of the second modeltermed ``the dispersed growth model'', did not give

a realistic interpretation of the data since diusional

limitations were not considered.

Further work using the xed lm model was car-

ried out by Bachmann et all. (1985) on baed reac-

tors with an inuent substrate concentration of 8 g

COD/l. The model predicted the following beha-

viour: a decrease in treatment eciency with (a)

decreasing inuent substrate concentration at con-

stant loading rates, (b) an increase in organic load-

ing at constant inuent substrate concentration, and

(c) an increase in recycle ratio at constant HRT

since the reactor approaches completely mixed

behaviour. Reactor eciency improved with redu-cing substrate concentration at constant HRT.

Some of these ndings were mirrored in the work

of Xing and Tilche (1992) on the modelling of a

hybridised form of the baed reactor which had a

working volume of 150 l, and treated 20 kg COD/

m3 d of molasses wastewater. The model focused on

the ndings of ATP testing which concluded that

virtually all of the active biomass was held within

the base of each compartment, so the biomass

weight and not concentration was used in the

model. The main assumptions of the model were:

all substrate consumption occurred within a granu-

lar sludge bed, and, the sludge bed was perfectlymixed due to gas evolution. The following predic-

tions were made from the model: at constant or-

ganic loading the treatment eciency increased with

increasing inuent substrate concentration; as HRT

was reduced the performance of the reactor

decreased; performance deteriorated with increasingloading (1116 kg COD/m3 d) with a constant

sludge weight; an improvement in COD removal

eciency was observed with increasing sludge

weight until a certain concentration was reached,

above which reactor performance becomes indepen-

dent of biomass concentration; and nally an

increase in recycle ratio coincided with a subsequent

decrease in COD removal.

Bachmann et al. (1983, 1985) assumed that the

oc diameter was very large relative to the active

biolm depth. However, this seems to be an unwar-

ranted assumption since in anaerobic biolms the

electron donor and acceptor are often the same or-

ganic, and since the three main microbial groups

are symbiotic, oc particles may have active cores

even with 3 mm diameters. In addition, many lm

supports are not perfectly at, but can be con-

sidered suciently at if the biologically active

thickness of the biolm is less than about 1% of

the radius of curvature (i.e. the radius of a sphere

plus a diusion layer, Rittmann and McCarty

(1978)). At high loading rates this is not the case in

the ABR, since sludge particles within the reactor

act as uidised spheres with a surface area through

which the substrate must diuse for consumption

(Bachmann et al., 1985). These facts imply that a

spherical model would provide a better t than asimple planar one.

Nachaiyasit (1995) derived a spherical model

using Monod kinetics combined with molecular dif-

fusion of the substrate into the biomass aggregates

based on the assumptions that: (i) substrate concen-

tration could be described by a single parameter,

COD, (ii) biomass concentration can be adequately

described by a single parameter, VSS, (iii) the bio-

mass composition is constant during balanced

growth and (iv) the biological reactions of import-

ance occur at constant temperature and pH. The

calculation of important model parameters such as

diusion layer thickness, and liquid phase mass

transfer coecient followed the techniques pro-posed by Bachmann et al. (1985). In general, the

model predicted better COD removal than was ex-

perimentally measured, and was most accurate for

high loading rates (8 and 15 g COD/l at 20 h) than

at short retention times (10, 5 h HRT with feed

concentration of 4 g COD/l), but showed large devi-

ations for the rst couple of compartments for

some of the simulations. As with previous models

certain trends appeared with the results, namely a

decrease in removal eciency with increasing re-

cycle ratios, decreasing HRTs (with xed substrate

concentration), and increasing substrate concen-

trations (with xed HRT). However, the sphericalmodel did provide a closer t than the earlier pla-

nar xed lm equations. Based on the ndings of a



15/20

sensitivity analysis that showed oc surface area

and owrate had the greatest inuence on model

predictions, the model was modied by making the

surface area a tting parameter. Nachaiyasit thencompared the results obtained from dierent models

based on the same assumptions and with the same

experimental data, and found the closest t with the

spherical model. It was concluded that while the

predictive capacity of the spherical model was not

always good, it was useful as a tool for understand-

ing the interaction between the various system par-

ameters, and therefore could be used as a basis for

the development of better predictive models.

It seems that a combination of theoretical con-

siderations and experimental ndings can be used

together in order to generate models with a more

realistic t. Since the accuracy of any model

depends critically on the wastewater and substrate

used, kinetic data should be experimentally deter-

mined for each compartment once the reactor is at

steady state (Bachmann et al., 1985) by using simple

bioassays. Such an approach may have enabled

Nachaiyasit's spherical model to give a more realis-

tic t at the front of the reactor. All modelling so

far has used acetoclastic methane production as the

rate-limiting step. However, it is evident from

Section 5, that the structure of the reactor will

cause a shift in the population dynamics of the two

species (Methanosarcina and Methanosaeta) respon-

sible for acetate consumption. Since both archea

dier widely in kinetic ability, acetate loadings andpH will have an eect on reactor performance in

each compartment. For reactors treating medium to

high strength wastes acetate consumption at the

front of the reactor will be higher than for a low

strength waste. This will result in most of the COD

being removed in the front of the reactor. For low

strength wastes, acetate loadings will be low and

this will encourage growth of Methanosaeta with lit-

tle COD removed in the acidication zone.

Channelling has been shown to be an important

phenomenon in the ABR (see Section 3) and will

aect the accuracy of any model. In order to take

channelling into account it is necessary to calculate

the number N of ideally mixed reactors in seriesusing tracer studies (see Section 3.1). The results of

these experiments could then be input as the num-

ber of real compartments into a reactor in series

model. Also, correlations are available which show

the eect of hydrodynamic dispersion on the sub-

strate diusion coecient (Bear, 1972).

Furthermore, by calculating the minimum solids

retention time (Orozco, 1988), it should be possible

to determine the correct compartment size for a

given treatment eciency. On the basis of the litera-

ture it seems that for most cases only 24 compart-

ments are necessary for adequate COD removal.

However, reactors with more compartments will befar more resistant to hydraulic and organic shocks,

since they will protect against the shift in acid pro-

duction towards the rear. Therefore a compromise

will exist between optimal (required) compartment

number, ``safe'' compartment number, and also

upow liquid velocity.Despite the less than perfect predictive capabili-

ties of the models described above, there is an

urgent need to generate models for larger scale reac-

tors. Boopathy and Tilche (1991) pointed out that

at larger scale a greater evolution of gas per com-

partment cross sectional area can be expected, and

this would cause an increase in mixing which would

subsequently improve mass transfer rates leading to

greater eciencies, but perhaps increased solids

loss.

Finally, it is also necessary to model reactor

behaviour when hydrolysis is the rate-limiting step,

as is the case with high solid inuents and lipid con-

taining wastewaters, since by assuming rst-order

rate kinetics it is possible to calculate the minimum

solids retention time to achieve a given eciency

(Pavlostathis and Giraldo-Gomez, 1991). With

wastewaters containing a large amount of particu-

late material, it seems likely that COD removal will

be low at the front end, and that the VFA prole

formed will be shifted down the reactor, unless the

reactor is modied, (Boopathy and Sievers, 1991) or

extra compartments added, a drop of eciency may

result.

FULL-SCALE EXPERIENCE

The performance data of a full-scale plant, treat-

ing domestic waste from a small town in Columbia

(Tenjo, population


16/20

reactors experienced several practical problems

during early operation. Hydraulic shocks increased

solids washout, and poor screening of solid material

caused the plastic packing media to oat with gas

production. These problems were overcome by

using a by-pass pipe during the rainy season and

improved screening facilities. The reactors per-

formed well with approximately 70% COD re-

duction and 80% removal of suspended solids over

a two-month period. Varying the volumetric load

between 0.4 and 2 kg/m3 d had no eect on removal

eciency. However, the author concluded that a

polishing lagoon was necessary to achieve discharge

quality euent. Work is currently being undertakento provide a wastewater treatment plant for a larger

town.

Although a detailed economic study was not pre-

sented, construction costs for the baed reactor

were 20% less than those for UASB reactors in

Columbia running at ambient temperature, and ve

times less than a conventional activated sludge

plant for a small town.

FUTURE PROSPECTS FOR THE ABR

The ABR shows promise for industrial waste-

water treatment since it can withstand severehydraulic and organic shock loads, intermittent

feeding, temperature changes, and tolerate certain

toxic materials due to its inherent two-phase beha-

viour. Despite comparable performance with other

well established technologies (Table 14), its future

use will depend on exploiting its structure in order

to treat wastewaters which cannot be readily trea-

ted. Outlined below is a list of possible processes

that are feasible in the ABR.

In situ aerobic polishing

Unpublished work in this laboratory has shown

that an aerobic polishing step can be inserted within

an ABR with no detrimental eect on reactor per-

formance. This is due to the fact that ``aerophobic''

methanogens can remain active even when oxygen

is present, and whilst inside immobilised aggregates

methanogenic archea are well shielded from oxygen

by layers of facultative bacteria (Lettinga et al.,

1997). Also, processes which inherently require both

anaerobic and aerobic treatment (or detoxication)

can be dealt with within a single reactor unit, such

as black hemp liquors, wood extractives, coal pro-

cessing industry, petrochemical, and textile dye

wastewaters (Lettinga, 1995) thus signicantly redu-

cing capital costs.

Total nitrogen removal. Work is currently being

undertaken to treat ammonia containing waste-

waters with an anaerobic/aerobic baed reactor fortotal nitrogen removal. Ammonia present in the

wastewater passes through the anaerobic compart-

Table 14. Treatment eciencies for various reactor congurations

Feedstock Reactortype

Reactorvolume (l)

Inlet COD(g/l)

Loadingrate (kg/m3 d)

CODremoval (%)

Ref.

Carbohydrate ABR 6.3 7.1 1 79 Bachmann et al., 1983ABR 6 110 220 7299 Bae et al., 1997ABR 75 0.440.47 0.961.66 8493 Orozco, 1988

UASB 4.8 110 220


17/20

ments largely unmetabolised, and is then oxidised

to form nitrates and nitrites at the rear of the reac-

tor. These can then be recycled to the anaerobic

section where they act as alternative electron accep-

tors and are reduced to nitrogen.

Complete sulphur removal. Sulphate is reduced at

higher redox potentials than that at which methano-

genesis begins (Henze and Harremoes, 1983), and

will therefore be converted to hydrogen sulphide at

the front of a baed reactor at the expense of

methanogenesis (Fox and Venkatasubbiah, 1996).

Micro-aerobic polishing could be achieved within

an aerobic stage to produce elemental sulphur,

which could be recovered eliminating the need of a

separate trickling lter unit.

Enhancement of two-phase properties (better pH and

temperature control)

The optimum pH for a two-phase system has

been widely quoted to be approximately 5 (Ghosh

et al., 1975; Aivasidis et al., 1988; Speece, 1996).

This implies that less buering would be required in

a baed reactor since the pH is routinely above 6

in the rst compartment. Alternatively, buering

and/or nutrients could be added separately in latter

compartments to provide optimal conditions forscavenging methanogens.

CONCLUSIONS AND RECOMMENDATIONS

Laboratory, pilot and full-scale work has shown

that the ABR is capable of treating a variety of

wastewaters of varying strength (0.45 < 1000 g/l),

over a large range of loading rates (0.4 < 28 kg/

m3 d), and with high solids concentrations with sat-

isfactory results (Table 6). Long biomass retention

times are possible without granulation and solids/

liquid separation devices, and a selective pressure

exists which enhances the development of appropri-ate bacterial populations in various parts of the

reactor. This reactor conguration confers consider-

able resistance to toxic materials, shields syntrophic

bacteria from elevated hydrogen levels, and results

in high removal eciencies even at low hydraulic

retention times (26 h). The physical structure of

the reactor allows various modications to be

made, such as an in situ aerobic polishing stage,

resulting in providing the capability to treat waste-

waters that currently require at least two separate

units, therefore substantially reducing capital costs.

However, in order to enhance the commercial po-

tential of the ABR, more work still remains to be

done in the following areas: modelling the fate of

SMPs, solids, intermediate products, and COD

removal; nutrient requirements; treatment of toxic

wastewaters (e.g. polychlorinated aliphatics, nitrated

organics, xenobiotics, haloaromatics, surfactants)

which have been treated with success anaerobically;

and an improved understanding of the factors con-

trolling bacterial ecology. Finally, Table 15 shows a

list of recommendations based on this review of the

literature.

AcknowledgementThe authors would like to thankProfessor Chynoweth for his generous help with providing

material for this review and the BBSRC for nancial sup-port.

REFERENCES

Aivasidis A., Bastin K. H. and Wandrey C. (1988)Optimisation of selection stress in a chemostat.Anaerobic Digestion, IAWPRC, 3546.

Alves M. M., Pereira M. A., Mota M., Novais J. M. andColleran E. (1997) Staged and non staged anaerobic l-ters: microbial selection, hydrodynamic aspects and per-formance. Proceedings of the 8th InternationalConference on Anaerobic Digestion, Vol. 2, Sendai,Japan, pp. 5663.

Bachmann A., Beard V. L. and McCarty P. L. (1983)Comparison of Fixed Film Reactors with a Modied

Sludge Blanket Reactor, Fixed Film BiologicalProcesses for Wastewater Treatment, ed. Y. C. Wu andE. D. Smith. Noyes Data, NJ.

Table 15. Recommendations based on literature ndings

Recommendations

Start-up low initial loading rates will encourage granule/oc growthpulses of methane precursors (e.g. acetate) have been successfully used to encourage methanogenicgrowth and dampen the eects of increases in loading ratestart-up with long retention times reduces solids loss due to low liquid upow velocities and,promotes higher methanogen populations in every compartment

Recycle recycle is benecial with respect to diluting toxicants in feed stream, increasing front pH andreducing production of foam and SMPs, but has several disadvantages outlined in Table 4

Low strength wastewater low retention time enables better mass transport due to improved hydraulic mixing and reducesbiomass starvation in latter compartmentsmethane production will originate from scavenging bacteria (Methanosaeta)

High strength wastewater long retention times reduce solids washout caused by high gas production, otherwise the reactormay be modied (by adding packing) to decrease biomass lossmethane production will be mainly due to Methanosarcina, and hydrogen scavenging methanogens

High solids wastewater a larger front compartment has proved to be eective in treating wastewater with a high solidscontent

Temperature reducing temperature to 258C from 358C has no eect on easily degradable waste, furtherdecreases in temperature are detrimental on reactor performance, this may be due to potentialtoxicity, nutrient bioavailability and slower kinetic ratesreactors started-up and kept at lower temperatures perform consistently well



18/20

Bachmann A., Beard V. L. and McCarty P. L. (1985)Performance characteristics of the anaerobic baedreactor. Wat. Res. 19(1), 99106.

Bae J.-H., Song K.-B. and Cho K.-M. (1997) Comparison

of operational characteristics of UASB and ABR: or-ganic removal eciency and the variations of PH2 andPCO. Proceedings of the 8th International Conferenceon Anaerobic Digestion, Vol. 1, Sendai, Japan, pp. 164171.

Barber W. P. and Stuckey D. C. (1997) Start-up strategiesfor anaerobic baed reactors treating a syntheticsucrose feed. Proceedings of the 8th InternationalConference on Anaerobic Digestion, Vol. 2, Sendai,Japan, pp. 3239.

Barber W. P. and Stuckey D. C. (1998) Inuence of start-up strategies on the performance of an anaerobic baedreactor. Environ. Technol. 19, 489501.

Bear J. (1972) Dynamics of Fluids in Porous Media.Elsevier, New York.

Boopathy R. and Sievers D. M. (1991) Performance of a

modied anaerobic baed reactor to treat swine waste.Trans. ASAE 34(6), 25732578.Boopathy R. and Tilche A. (1991) Anaerobic-digestion of

high-strength molasses waste-water using a hybrid an-aerobic baed reactor. Wat. Res. 25(7), 785790.

Boopathy R. and Tilche A. (1992) Pelletization of biomassin a hybrid anaerobic baed reactor (HABR) treatingacidied waste-water. Bioresource Technol. 40(2), 101107.

Boopathy R., Larsen V. F. and Senior E. (1988)Performance of anaerobic baed reactor (ABR) intreating distillery waste-water from a Scotch Whiskyfactory. Biomass 16(2), 133143.

Borja R., Banks C. J. and Wang Z. (1994) Stability andperformance of an anaerobic down ow lter treatingslaughterhouse wastewater under transient changes inprocess parameters. Biotechnol. Appl. Biochem. 20, 371

383.Cayless S. M., da Motta Marques D. M. L. and Lester J.

N. (1989) The eect of transient loading, pH and tem-perature shocks on anaerobic lters and uidised bedsEnviron. Technol. Lett. 10(11) 951968.

Chang Y. J., Nishio N. and Nagai S. (1995)Characteristics of granular methanogenic sludge growthon phenol synthetic medium and methanogenic fermen-tation of phenolic wastewater in a UASB reactor. J.Ferm. Bioeng. 79(4), 348353.

Chynoweth D. P., Srivastra V. J. and Conrad J. R. (1980)Research study to determine the feasibility of producingmethane gas from sea kelp. Annual Report for GeneralElectric Company, IGT Project 30502, Institute of GasTechnology, IIT Centre, 3424 S. State Street, Chicago,IL 60616.

Cintoli R., Disabatino B., Galeotti L. and Bruno G.(1995) Ammonium uptake by zeolite and treatment inUASB reactor of piggery wastewater. Wat. Sci. Technol.32(12), 7381.

Cohen A., Breure A. M., van Andel J. G. and vanDeursen A. (1980) Inuence of phase separation on theanaerobic digestion of glucose, I. Maximum COD-turn-over-rate during continuous operation. Wat. Res. 14,14391448.

Cohen A., Breure A. M., van Andel J. G. and vanDeursen A. (1982) Inuence of phase separation on theanaerobic digestion of glucose, II. Stability and kineticresponses to shock loadings. Wat. Res. 16, 449455.

El-Mamouni R., Rouleau D., Mayer R., Guiot S. R. andSamson R. (1992) Comparison of the novel multiplateanaerobic reactor with the upow anaerobic sludgeblanket reactor. 46th Purdue Industrial Waste

Conference Proceedings. Lewis, Chelsea, MI 48118.Fannin K. F., Srivastra V. J., Conrad J. R. and

Chynoweth D. P. (1981) Marine biomass program: an-

aerobic digester system development. Annual Report forGeneral Electric Company, IGT Project 65044, Instituteof Gas Technology, IIT Centre, 3424 S. State Street,Chicago, IL 60616.

Fannin K. F., Srivastra V. J., Mensinger J., Conrad J. R.and Chynoweth D. P. (1982) Marine biomass program:anaerobic digester process development. Annual Reportfor General Electric Company, IGT Project 65044 and30547, Institute of Gas Technology, IIT Centre, 3424 S.State Street, Chicago, IL 60616.

Fox P. Venkatasubbiah V. (1996) Coupled anaerobic/aerobic treatment of high-sulphate wastewater with sul-phate reduction and biological sulphide oxidation. Wat.Sci. Technol. 34(56), 359366.

Garuti G., Dohanyos M. and Tilche A. (1992) Anaerobicaerobic combined process for the treatment of sewagewith nutrient removal: the Ananox1 process. Wat. Sci.Technol. 25(7), 383394.

Ghosh S., Conrad J. R. and Klass D. L. (1975) Anaerobicacidogenesis of sewage sludge. J. WPCF 47, 3045.

Grobicki A. M. W. (1989) Hydrodynamic characteristicsand performance of the anaerobic baed reactor. Ph.D.dissertation, Department of Chemical Engineering,Imperial College, London, U.K.

Grobicki A. M. W. and Stuckey D. C. (1989) The role offormate in the anaerobic baed reactor. Wat. Res.23(12), 15991602.

Grobicki A. M. W. and Stuckey D. C. (1991) Performanceof the anaerobic baed reactor under steady state andshock loading conditions. Biotechnol. Bioeng. 37, 344355.

Grobicki A. M. W. and Stuckey D. C. (1992)Hydrodynamic characteristics of the anaerobic baedreactor. Wat. Res. 26, 371378.

Guiot S. R., Sa B., Frigon J. C., Mercier P., MulliganC., Tremblay R. and Samson R. (1995) Performances ofa full-Scale novel multiplate anaerobic reactor treating

cheese whey euent. Biotechnol. Bioeng. 45, 398495.Gujer W. and Zehnder A. J. B. (1983) Conversion pro-

cesses in anaerobic digestion. Wat. Sci. Tech. 15, 127167.

Habets L. (1996) Overview of industrial anaerobic wastewater treatment. Industrial Anaerobic WasteWaterTreatment Conference, 18th September, SCI, London.

Hassouna S. and Stuckey D. C. (1998), in preparation.Henze M. and Harremoes P. (1983) Anaerobic treatment

of wastewater in xed lm reactors: a literature review.Wat. Sci. Technol. 15(8/9), 1101.

Hickey R. F., Wu W.-M., Viega M. C. and Jones R.(1991) Start-up, operation, monitoring and control ofhigh-rate anaerobic treatment systems. Wat. Sci.Technol. 24, 207255.

Hickey R. F., Vanderwielen J. and Switzenbaum M. S.

(1987) Eects of organic toxicants on methane pro-duction and hydrogen gas levels during the anaerobicdigestion of waste activated sludge. Wat. Res. 21, 14171427.

Hilton M. G. and Archer D. B. (1988) Anaerobic diges-tion of a sulphate-rich molasses wastewater: inhibitionof hydrogen sulphide production. Biotechnol. Bioeng.31, 885888.

Holt C. J., Matthew R. G. S. and Terzis E. (1997) A com-parative study using the anaerobic baed reactor totreat a phenolic wastewater. Proceedings of the 8thInternational Conference on Anaerobic Digestion, Vol.2, Sendai, Japan, pp. 4047.

Iza J., Colleran E., Paris J. M. and Wu W.-M. (1991)International workshop on anaerobic treatment technol-ogy for municipal and industrial wastewaters: summarypaper. Wat. Sci. Technol. 24(8), 116.

Kanekar P., Sarnaik S. and Kelkar A. (1996) Microbialtechnology for management of phenol bearing dyestuwastewater. Wat. Sci. Technol. 33(8), 4751.



19/20

Kato M. T., Field J. A. and Lettinga G. (1997) The an-aerobic treatment of low strength wastewaters.Proceedings of the 8th International Conference onAnaerobic Digestion, Vol. 1, Sendai, Japan, pp. 356

363.Kotsyurbenko O. R., Nozhevhikova A. N. and Zavarzin

G. A. (1993) Methanogenic degradation of organic mat-ter by anaerobic bacteria at low temperature.Chemosphere 27, 17451761.

Lawrence A. W. and McCarty P. L. (1969) Kinetics ofmethane fermentation in anaerobic treatment. J. WPCF41, R1R17.

Lettinga G. (1995) Anaerobic digestion and wastewatertreatment systems. Antonie van Leeuwenhoek 67, 328.

Lettinga G., Field J., van Lier J., Zeeman G. andHulsho Pol L. W. (1997) Advanced anaerobic waste-water treatment in the near future. Wat. Sci. Technol.35(10), 512.

Lettinga G., Hobma S. W., Hulsho Pol L. W. and deZeeuw

the use of the anaerobic baffled

Documents