Journal of Biotechnology 126 (2006) 475–487

Start-up of the Anammox process in a membrane bioreactor

C. Trigo, J.L. Campos, J.M. Garrido

∗

, R. M´endez

Department of Chemical Engineering, Faculty of Engineering, University of Santiago de Compostela,

Avda Lope G´omez de Marzoa, E-15782 Santiago de Compostela, Galicia, Spain

Received 30 December 2005; received in revised form 20 April 2006; accepted 4 May 2006

Abstract

The start-up of an Anammox process was studied in a membrane sequencing batch reactor (MSBR) in which a submerged

hollow fibre membrane module was used to retain the biomass. The reactor was seed with Anammox biomass and fed using the
Van de Graaf medium. During a first operating stage, salt precipitation was observed and interfered with microbial activity and
caused a decrease of the nitrogen removal rate of the reactor from 100 to only 10 mg l

−1

per day. Salt precipitation was avoided

by diminishing adequately the Ca and P concentrations of the Van de Graaf medium during the last operating stage. This action
increased quickly the activity of the system, and nitrogen removal rate reached up to 710 mg l

−1

per day with almost full nitrite

removal. Sporadic flotation of the sludge was observed in the MSBR. The use of the membrane avoided biomass wash-out from
the system. Moreover, a surprising fact was that Anammox biomass did not grow in flocs in the MSBR, but in granules. This
fact showed that this kind of microorganisms have a trend to grow in aggregates. Results indicated that the use of the MSBR
could be a suitable system for nitrogen removal by using the Anammox reaction.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Anammox; Denitrification; Granule; Membrane; MSBR; Wastewater

1. Introduction

The removal of the nitrogen present both in munic-

ipal and industrial wastewaters, mainly as ammonium,
is carried out conventionally by means of the combina-
tion of two biological processes, nitrification and deni-
trification. This procedure is suitable for the treatment
of nitrogenous wastewaters rich in biodegradable car-

∗

Corresponding author. Tel.: [phone]x16778;

fax: [phone].

E-mail address:

[e-mail]

(J.M. Garrido).

bon, but it results expensive in those wastewaters with
low biodegradable matter and high nitrogen concen-
trations, such as, e.g. old landfill leachates or effluents
from the anaerobic digestion of sludge in wastewa-
ter treatment plants (WWTP). In this case the addi-
tion of an external organic matter source (methanol
or acetic acid) for the denitrification stage is neces-
sary in order to obtain the nitrogen removal. This
increases the operating costs in the wastewater treat-
ment plant due to the cost of the chemicals added
and the treatment of the additional sludge that is
generated.

0168-1656/$ – see front matter © 2006 Elsevier B.V. All rights reserved.

doi:

10.1016/j.jbiotec.2006.05.008

476

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

An alternative, to these conventional processes, is

the combination of two biological processes: the par-
tial nitrification of ammonium to nitrite by means of
nitrifying bacteria and the denitrification of nitrite to
dinitrogen gas by using ammonium as electron donor,
the Anammox process (

Mulder et al., 1995; Van de

Graaf et al., 1995

). The Anammox process is a biolog-

ical mediated reaction in which ammonia is oxidized
to nitrogen gas using nitrite as the electron acceptor
under anaerobic conditions. In Eq.

(1)

the stoichiom-

etry of this reaction proposed by

Strous et al. (1998)

is shown, in which ammonium and nitrite, in almost
equimolar ratio, react to produce dinitrogen gas and a
small quantity of nitrate.

NH

4

+

+ 1.32NO

2

−

+ 0.066HCO

3

−

+ 0.13H

+

→ 1.02N

2

+ 0.26NO

3

−

+ 0.066CH

2

O

0

.5

N

0

.15

+ 2.03H

2

O

(1)

The Anammox process should be combined with a pre-
vious partial nitrification stage, in which around 50% of
ammonia should be converted to nitrite. This could be
obtained either by manipulating the temperature and
the HRT using the Sharon process (

Hellinga et al.,

1998

), by inhibition of nitrite oxidizing bacteria by free

ammonia (

Balmelle et al., 1992

) or by manipulating the

dissolved oxygen concentration (

Garrido et al., 1997

).

The Anammox process presents advantages for the
treatment of effluents with deficiency in organic mat-
ter, compared to the nitrification–denitrification pro-
cess. This process allows the reduction of the oxygen
requirements and carbon dioxide emission to the atmo-
sphere and less production of sludge in the WWTP.
However, the practical application of the Anammox
process is still limited by its long start-up periods due
to the very low growth rates (0.072 d

−1

measured at

32

◦

C) and biomass yield generated per ammonia nitro-

gen consumed (0.088 g g

−1

) of these microorganisms

(

Jetten et al., 1997

). Moreover, an additional problem

is caused by loss of a fraction of the sludge washed out
with the effluent. For this reason, an efficient system or
operation strategy in order to avoid biomass wash-out
with the effluent is required. To achieve high biomass
retention it is very important during the start-up, which
can take months or even a year in laboratory scale reac-
tors. Even a slight loss of biomass supposes a delay in
the time required to obtain the desired loading rate.

The development of reactors using the Anammox

process is still recent. The first Anammox reactors were
biofilm reactors, e.g. fixed bed reactor, fluidised bed
reactors and gas lift reactor (

Van de Graaf et al., 1996;

Strous et al., 1997; Sliekers et al., 2003; Dapena-Mora
et al., 2004a

). However, some of these systems did not

show as the most adequate to avoid the Anammox
biomass wash-out. In order to improve the biomass
retention and the stability process, the sequencing
batch reactor (SBR) was successfully used to grow
Anammox biomass (

Strous et al., 1997, 1999; Dapena-

Mora et al., 2004a

). These reactors were operated with

an additional mechanical stirring in order to improve
the biomass retention and prevent the entrapment of
nitrogen bubbles, therefore increasing the stability of
the process. Other systems that were used with suc-
cess were a reactor containing non-woven media for
biomass immobilisation (

Furukawa et al., 2003

) and

an upflow system seed with anaerobic granular sludge
(

Imajo et al., 2004

). However, a fraction of the gener-

ated biomass is inevitably washed out with the effluent
in all these systems, especially during unstable peri-
ods due in many cases to overloads, which provoke the
biomass flotation.

For these reasons, further investigation is needed

to increase biomass retention inside the reactor espe-
cially in those cases where the Anammox activity of the
inoculum is very low. This is an important challenge
in order to scale up Anammox systems from labora-
tory to industrial scale, in which the start-up could be
done using secondary sludge from WWTP that is even
bioaugmented with the Anammox biomass generated
in small scale laboratory units. In fact, there is only
one reference of an industrial scale unit in Rotterdam
WWTP (

Van Loosdrecht and Salem, 2005

).

An alternative for obtaining full biomass retention

in Anammox systems might be the use of membrane
biological reactors (MBR) for the treatment of the
wastewaters. In the last 20 years, membrane technol-
ogy has been utilized to promote biomass retention
instead of secondary clarifiers in WWTPs. MBR sys-
tems are compact reactors, which may operate with
high biomass concentrations and an absolute control of
solids and hydraulic retention times. Limitations inher-
ent to MBR processes are the cost of membranes and
operative costs due to fouling and higher energy con-
sumption compared to traditional WWTPs. Since the
MBR retains all organisms, it could ideally be suited

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

477

to slow growing cultures such as Anammox bacteria,
enhancing the start-up of the process. The use of an
Anammox MBR was referred for first time by other
authors (

Wyffels et al., 2004

). However, there is still a

limited knowledge concerning the behaviour of Anam-
mox MBRs during the start-up, the factors that would
inferred in the process, the kind of aggregates formed
by Anammox biomass, or the recovery of the capacity
of these systems after episodes of nitrite build-up. The
aim of the present work is to present the results obtained
in a membrane sequencing batch reactor (MSBR) used
to promote the growth of Anammox biomass, which
could be an alternative to other suspended or biofilm
systems.

2. Materials and methods

2.1. Reactor

In

Fig. 1

is depicted the MSBR used during the

experiments, containing a submerged ultrafiltration
hollow fibre membrane module. The system had a max-
imum working volume of 5 l. The reactor was operated

Fig. 1. Scheme of the experimental system. (1) MSBR, (2) stirrer,
(3) membrane module, (4) influent tank, (5) peristaltic pump of the
feeding media, (6) permeate pump, (7) permeate tank and (8) pro-
grammable logic controller.

at a fixed temperature of 35

◦

C by means of ther-

mostated jacket. The mixture inside the reactor was
achieved with a mechanical stirrer. Norprene tubing
and connections were used in order to avoid the dif-
fusion of oxygen. The hollow fibre membrane was
provided by Zenon, Environmental Inc. The ultrafiltra-
tion membrane module with a pore size of 0.04

␮m was

used to ensure the complete retention of the suspended
solids into the reactor. The small dimensions of the
reactor and the presence of a mechanical stirrer forced
to design a module with the membranes arranged in
a circular configuration with the extremes joined to a
tube where permeate was collected.

The MSBR was seeded at the operating day 0

with enriched anaerobic ammonium-oxidising granu-
lar sludge from a laboratory scale SBR (

Dapena-Mora

et al., 2004b

). The initial biomass concentration in

the membrane bioreactor was 0.125 g l

−1

. The specific

Anammox activity of this biomass was 0.3 g g

−1

per

day.

2.2. Feeding media and strategy of operation

The reactor was fed with three different synthetic

media (synthetic media 1, 2 and 3) during the three
main operating stages (stages 1, 2 and 3) of the research
(

Table 1

). The feeding synthetic medium 1 was the one

described by

Van de Graaf et al. (1996)

which is, at the

moment, the most used synthetic medium to operate
the Anammox process (

Strous et al., 1997; Kuai and

Table 1
Composition of the three feeding synthetic media fed to the MSBR,
in mg l

−1

Compound

Synthetic
medium 1

a

Synthetic
medium 2

Synthetic
medium 3

(NH

4

)

2

SO

4

75.3–283.4

66.4–79.7

66.4–1727

NaNO

2

83.7–315.4

73.9–88.7

73.9–1823

NaNO

3

8.5

KHCO

3

1000

KH

2

PO

4

50

10

10

CaCl

2

·2H

2

O

226

22.6

5.65

MgSO

4

·7H

2

O

58.6

FeSO

4

6.25

EDTA

6.25

Trace elements

a

1.25 ml/l

Composition of all compounds in media 2 and 3 were the same as
synthetic medium 1, except for the concentrations of (NH

4

)

2

SO

4

,

NaNO

2

, KH

2

PO

4

and CaCl

2

·2H

2

O that are indicated.

a

Described by

Van de Graaf et al. (1996)

.

478

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

Verstraete, 1998; Imajo et al., 2004; Dapena-Mora et
al., 2004a

). The other two synthetic media were based

in the

Van de Graaf et al. (1996)

medium. The con-

centration of calcium salt was diminished from 226 to
22.6 and 5.65 mg l

−1

in the synthetic medium 2 and

synthetic medium 3, respectively, and the concentra-
tion of phosphorus was set at 10.0 mg l

−1

for these two

synthetic media.

The control of the SBR was carried out with a PLC

system (CPU224, Siemens). The reactor was operated
in cycles of 6 h (

Dapena-Mora et al., 2004b

). Each cycle

comprised four stages. During the first one, the reac-
tor was continuously fed and mixed for 330 min. In the
second stage, the stirrer was put off during 9 min allow-
ing the biomass to settle. The third stage consisted of
the permeation of the supernatant liquid, which was
removed from the reactor by creating a membrane
under-pressure with a peristaltic pump, during 18 min.
Finally, in the fourth stage, part of the permeated liquid
was backwashed during 3 min to minimise the mem-
brane fouling. The operational cycle of the reactor is
detailed in

Fig. 2

.

The operational strategy of the reactor consisted of

increasing the nitrogen loading rate (NLR) applied to
the reactor by means of increasing the ammonium and
nitrite concentrations in the influent media, once the
nitrite concentration was close to 0 in the effluent. The
pH of the feeding medium was adjusted around 8.0 by
means of H

2

SO

4

(1 M) addition. The HRT was fixed at

1 day.

2.3. Analytical methods

Nitrate, nitrite and ammonium concentrations were

determined spectrophotometrically and biomass con-
centrations were determined as volatile suspended
solids (VSS), according to Standard Methods (

APHA,

1985

). The sludge volumetric index (SVI) was deter-

mined according to

Ramalho (1991)

. The elemental

analysis of the surface of biomass aggregates was per-

formed using a scanning electron microscope (SEM;
Leica 440). Elemental analysis of C, H, N and S of a
biomass sample was done by using Fisons EA-1108
elemental analyser and O content by using Carlo Erba
1108 elemental analyser.

The content of Anammox bacteria in the biomass

samples from the reactors was followed by fluores-
cence in situ hybridisation (FISH) (

Amann, 1995

).

This analysis was performed with a set of fluorescent-
labelled 16S rRNA-targeted probes according to the
procedure described by

Amann (1995)

. Probes used

for FISH and the formamide concentrations used dur-
ing hybridisation were the mixture EUB 338I, EUBII
and EUBIII for all the eubacteria (

Daims et al., 1999

),

the probe PLA 46 for the Planctomycetales and the
probe Amx 820 for the Candidatus “Brocardia anam-
moxidans” and Candidatus “Kuenenia stuttgartiensis”
(

Strous et al., 1998

) labelled with fluos and Cy3 flu-

orochromes. For analysis of the slides, an epifluores-
cence microscope (Axioskop 2 plus, Zeiss) in combina-
tion with a digital camera (Coolsnap, Roper Scientific
Photometrics) were used.

Batch experiments to determine the specific anam-

mox activity (SAA) were performed according to the
methodology described elsewhere (

Dapena-Mora et

al., 2004a

), based on the measurement along time of

the overpressure generated in closed vials by the nitro-
gen gas produced.

3. Results

The MSBR was operated during 375 days. From the

obtained results (

Fig. 3

), the experimental period can

be divided in three different experimental stages, which
were coincident with the periods in which the three
different synthetic media were fed: stage 1, from oper-
ating day 0 till day 80, in which the original medium
developed by Van de Graaf was fed; stage 2, from
operating day 80 till day 183, in which was used a

Fig. 2. Operational cycle strategy of the MSBR. The time length of each operating phase is indicated in the grey box (min).

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

479

Fig. 3. Evolution of the ammonium concentration in the influent (

䊉)

and the effluent (

) and the nitrite concentration in the influent ()

and the effluent (

) along the operational period (mg N/l).

modified Van de Graaf medium with a lower concen-
tration of calcium and phosphorus; and stage 3, from
operating day 183 to the end of the experiments in
which another modified Van de Graaf medium with
a lower concentration of calcium than the two others
was fed. Results showed that the nitrogen removal rate
(NRR) of the MSBR diminished during the period 1
and did not vary in period 2. NRR is expressed here
as the loss of nitrogen from ammonia and nitrite due
to the Anammox reaction and is indicated in terms
of mass per unit of reactor volume and unit of time.
During stage 3 NRR increased with time till a max-
imum value of 710 mg l

−1

per day. Throughout the

operation, both the ammonium and nitrite concentra-
tions were gradually increased from 10 to 390 mg l

−1

,

as the capacity of nitrogen removal of the system
increased.

3.1. Stage 1 (from day 0 till day 80)

During the first operating days, after the inoculation

the system showed almost full nitrite removal an even
near increase in the nitrogen removal rate (NRR) up to
100 mg l

−1

per day. However, after operating day 10,

NRR started to decrease (

Fig. 4

). The NRR obtained

on the day 18 was around 17 mg N l

−1

per day. These

facts coincided with an important increase in the non-
volatile suspended solids (NVSS) concentration from
50 to 150 mg l

−1

and an increase of the percentage of

NVSS in the biomass from 26 to 49% (

Fig. 5

), as well

as with the breakage of the granules (

Fig. 6

). Due to

this last fact, an excess of agitation was considered to be

Fig. 4. Total nitrogen (

), ammonium (䊉) and nitrite () removal

rates in the reactor during experimental stage 1 (A) and experimental
stage 3 (B).

the reason of the loss of activity of the system, and the
stirring rate was reduced from 75 to 45 rpm on the oper-
ating day 29. However, the activity still diminished, in
spite of diminishing the stirring speed in order to reduce
shear stress on the biomass. Another cause that was
considered as possibly responsible for the low activity
of the biomass was the presence of inhibitory concen-
trations of oxygen in the reactor. In order to prevent
the presence of oxygen in the reactor, the reactor was
hermetically closed and argon gas was fluxed to avoid
any accidental air entrance in the system. No improve-
ment of the Anammox activity was observed, but even
the nitrogen removal rate decreased until it reached a
minimum value of 5 mg l

−1

per day, at around day 70.

The seed was composed by small red colour gran-

ules, but biomass colour gradually changed to light
chestnut colour with time (

Fig. 6

) that occurred simul-

taneously with the observed NVSS accumulation. Cal-
cium phosphate precipitation was considered as a fea-
sible reason of the observed increase in the NVSS
concentration, and a cause of loss of biomass activity
in the reactor during this period.

480

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

Fig. 5. Evolution of the total suspended solids (

䊉) and NVSS ( ) in the MSBR (primary y-axis) and NVSS percentage () referred to TSS

during the experiment (secondary y-axis). Arrow points out the increase in the NVSS that took place along stage 1.

3.2. Stage 2 (from day 80 till day 183)

In order to avoid the NVSS accumulation observed

in the previous stage, phosphorus and calcium concen-
trations in the feeding medium were diminished on
the operating day 80. The concentration of calcium
salt was lowered 10 times (from 226 to 22.6 mg l

−1

)

and the phosphorus salt one, 5 times (from 50 to
10 mg l

−1

). The activity remained very low, 5 mg l

−1

per day, through this stage and did not vary during
almost 100 experimental days. The diminution of the
concentration of calcium and phosphorus in the influent
did not increase the capacity of the reactor. In order to
determine whether there was still active biomass inside
the reactor, a FISH assay was carried out. It could be
appreciated that although in slow proportion, there was
still active Anammox biomass. This indicated that the
residual anaerobic ammonia oxidation capacity of the

system was a consequence of the corresponding bio-
logical mediated reaction.

During this stage, an increase in biomass concentra-

tion, in terms of volatile suspended solids concentration
(VSS), was observed from 0.2 to 0.4 g l

−1

. Despite

the diminution of a 90 and 80% for Ca and P con-
centrations, respectively, the NVSS concentration in
the reactor increased achieving around 0.4 g l

−1

. The

fraction of NVSS in the biomass did not vary and
remained around 50% during the whole experimental
stage. Taking these data into account it was verified that
the precipitation was still present, which indicated that
the selected phosphorus and calcium concentrations in
medium 2 were not the adequate ones in order to avoid
salt precipitation in the system. For this reason, an ele-
mental analysis of the surface and the microscopically
observation of the surface of the biomass was carried
out by means of SEM (

Fig. 7

A). Mass percentage of

Fig. 6. Evolution of the appearance of the Anammox biomass during the first operating days of stage 1.

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

481

Fig. 7. Elemental analysis carried out with the SEM system, indi-
cating the percentage composition by mass of the most abundant
elements, of the surface of biomass samples during stage 2 at the
operating day 170 (A) and during stage 3 at the operating day 300
(B), respectively.

17.3% calcium and 7.8% phosphorus were detected in
the biomass surface using the SEM. The molar rela-
tionship between Ca and P was 1.71, which is close to
that of 1.5 of calcium phosphate, this indicated that the
precipitates were formed by calcium salts, especially
by the calcium phosphate salt.

3.3. Stage 3 (from day 183 on)

At the beginning of this period, an additional reduc-

tion in the concentration of calcium salts in the medium
was assayed. The concentration of calcium salt was
reduced 75%, from 22.6 to 5.65 mg l

−1

in order to avoid

or at least to reduce the precipitation observed in the
two previous stages. Few days after reducing the con-

centration of calcium in the feeding medium both the
activity and the nitrogen uptake of the system increased
quickly. During the first 45 operating days of this stage,
the nitrogen capacity of the system increased 5 times,
from 4.8 to 25.6 mg l

−1

per day (

Fig. 4

). This fact was

observed again during the next 45 days, and NRR of
the system increased till 142.7 mg l

−1

per day. A max-

imum NRR of 710 mg l

−1

per day was obtained in this

system, 185 days after the second reduction in the cal-
cium concentration. The doubling time of Anammox
biomass in this system was estimated in 18 days. These
values were calculated as the experimental time needed
to double the NRR in the reactor during this stage.

Nitrite accumulation took place in the effluent

between the operating days 290 and 350. Concen-
trations around 10 and 20 mg l

−1

in terms of nitrite

nitrogen were detected. This fact caused a partial inhi-
bition of the biomass. As a result, during this period
the nitrogen-loading rate was not increased in order to
avoid accumulations of higher nitrite concentrations.
Nitrite accumulation also was the cause of the spo-
radic sludge flotation events that were observed in the
reactor. In this respect, the use of a membrane avoided
the wash-out of the biomass with the effluent from the
system and the consequent loss of capacity of the sys-
tem that was referred for SBR and gas lift systems
during nitrite accumulation periods (

Dapena-Mora et

al., 2004a,b

). Nitrite accumulation was sorted out at

the operating day 355, when the feeding was stopped
during an operating cycle. This strategy make feasible
to diminish the concentration of both ammonium and
nitrite in the reactor. The diminution of the concentra-
tion of the inhibitory nitrite made feasible to recover
the activity and the NRR of the reactor. In fact nitrite
concentration after operating day 350 was almost fully
removed and its concentration was below 1 mg l

−1

.

It was possible to avoid the precipitation of cal-

cium phosphate, due to the second reduction of the
concentration of calcium. This fact was demonstrated
not only by determining the concentration of NVSS
(

Fig. 5

) but also analysing chemically the surface of

the biomass.

Fig. 7

B shows clearly the reduction of P

and Ca percentage in the biomass surface with regard to
those observed during period 2. Moreover, the concen-
tration of NVSS kept approximately constant around
0.2 g l

−1

indicating that no additional salt precipitation

took place. Additionally, biomass colour varied from
light chestnut to red colour and a mass percentage of

482

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

12.7% nitrogen was detected in the surface (

Fig. 7

B). It

was observed an increase in the concentration of VSS
from 0.27 to 0.96 g l

−1

. These facts cause a gradual

decrease of the percentage of NVSS in the biomass
from 53 till 18%.

Elemental analysis of whole biomass samples was

realised indicating that the molecular formula and
biomass yield (in terms of g of biomass per g of
ammonia) were CH

2.04

O

0.49

N

0.16

and 0.070 g g

−1

,

respectively. These results were very similar to those
obtained by

Strous et al. (1998)

of CH

2

O

0.5

N

0.15

and

0.088 g g

−1

. The stoichiometry of the reaction was also

determined, indicating that 1.22 mol of nitrite were
consumed and 0.22 mol of nitrate produced by mol of
ammonia consumed, which are lower than the values
of 1.32 and 0.26, respectively, indicated in Eq.

(1)

. The

obtained production of nitrate per mol of ammonia was
similar to that referred to by

Wyffels et al. (2004)

of

0.20 in an Anammox MBR system. However, these
authors indicated a stoichiometry of only 1.05 mol of
nitrite consumed per mol of ammonia consumed.

FISH assays indicated that most of the active

biomass that grew in the MSBR was composed by
Anammox microorganisms. With regard to the activ-
ity of the sludge, the activity assays indicated a gradual
increase of SAA till values between 0.35 and 0.45 g g

−1

per day, that are similar to those obtained by other
authors in an SBR system (

Dapena-Mora et al., 2004a

).

During the period in which nitrite accumulation took
place, between operating days 290 and 350, the activ-
ity suffered a decrease till a value of 0.2 g g

−1

per day,

that was a consequence of the partial inhibition of the
biomass by nitrite. Once the nitrite accumulation dis-
appeared, SAA value was recovered.

One surprising fact in the MBR was the aspect and

settling properties of the sludge. The value of the SVI
gradually decreased during 73 operating days of stage
3 from around 125 till 60 ml g

−1

, which indicates an

increase of the biomass density. Moreover, biomass
growth did not occur as suspended biomass but in gran-
ules with an irregular cauliflower appearance (

Fig. 8

A

and B). The former fact was similar to those results
observed by

Dapena-Mora et al. (2004b)

in a pre-

vious research done with a SBR Anammox system.
They observed that SVI changed from 108 to 63 ml g

−1

after around 80 operating days. However, these
authors indicated that the biomass grew as flocculent
sludge.

Fig. 8. Photograph of granules of biomass (A) and microphotograph
of a granule obtained in the SEM (B), during the operating stage 3.

3.4. Membrane fouling

An additional objective of this research was to study

the effect of the biological process on the behaviour
of the membrane. In this sense to study the behaviour
of the membrane permeability during the operation
and its fouling was very important. The monitoring
of the membrane fouling was made by following the
evolution of the transmembrane pressure. The foul-
ing of the membrane module was very low during the
whole experimental period. Membrane fouling control
involved only one method that was backwashing with
permeate after the end of the permeation cycle. Due to
the characteristics of the reactor, neither coarse bubble
gasification at the bottom of the membranes nor the

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

483

utilization of chemicals were applied. Moreover, the
external chemical washing of the membrane was nec-
essary only after approximately 3 months of operation.
This probably was a consequence of the characteristics
of the solids retained in the reactor, a mixture of NVSS
and biomass that tended to grow in granules with low
fouling capacity. Transmembrane pressure increased
progressively during every permeation period, proba-
bly as result of the deposition of an outer layer of solids
on the membrane. Backwashing had a positive effect
to remove solids deposition, since the value of trans-
membrane pressure at the beginning of each cycle was
recovered, this being around 32 kPa. Once the maxi-
mum transmembrane pressure attained a value of 60
or 70 kPa at the end of the permeation period, the
membrane was replaced by another and was externally
cleaned with commercial sodium hypochlorite, diluted
to a concentration of 250 mg l

−1

.

4. Discussion

4.1. Application of membrane bioreactors to the
operation of the anammox process

The operation of the MSBR was unsuccessful dur-

ing the first two stages, but successful during the
third. The maximum NRR reached during stage 3 was
approximately of 700 mg l

−1

per day. The average total

nitrogen removal efficiency was of 73.6% in stage 3.
Nevertheless, this was a consequence of the use of an
operation strategy, in which ammonia nitrite ratio was
higher than the ratio suggested by stoichiometry. The
kinetic and stoichiometric parameters of the Anammox
process, found in the membrane system are similar
to those found in other reactors. The stoichiometric
parameters found, nitrate produced to ammonia con-
sumed and nitrite consumed to ammonia consumed
were 0.22 and 1.22 mol mol

−1

, respectively, similar

to those found by

Strous et al., 1999

. FISH analysis

showed the presence of Anammox microorganisms in
the sludge. These facts demonstrate that the process
was carried out by Anammox microorganisms. On the
other hand, the use of a membrane did not shorten the
period of time that is necessary to operate at high NLR.
Moreover, the doubling time observed in the MSBR
was 18 days which was similar to 19 days obtained by

Dapena-Mora et al. (2004b)

in a SBR, but higher than

the 11 days referred by

Strous et al. (1999)

.

The influent concentrations and the NRR treated in

this study are comparable to those previously obtained
by

Dapena-Mora et al. (2004b)

and by

Van Dongen

et al. (2001)

of 600 and 700 mg l

−1

per day, respec-

tively, in SBR systems. Moreover, NRR was in between
the values of 650 and 1100 mg l

−1

per day referred by

Wyffels et al. (2004)

in an Anammox MBR system

fed with pre-filtered reject water from the dewater-
ing of digested sludge in a wastewater treatment plant.
Values of NRR higher than those of the MSBR were
referred by

Dapena-Mora et al. (2004a)

and by

Strous

et al. (1997)

of 2000 and 1800 mg l

−1

per day, respec-

tively, in fluidised bed systems. Futhermore,

Fux et al.

(2002)

obtained NRR of 1800 mg l

−1

per day in an

SBR. However, it is important to indicate that these sys-
tems underwent many problems of stability in periods
of overload, provoked both clogging of dinitrogen gas
and sludge flotation. These problems of stability could
be prevented by means of the use of membrane biore-
actors, since the wash-out would be avoided owing to
the presence of the membranes. Consequently, lower
times to recover the activity of the system would be
necessary. Furthermore, it is important to point out
that the operation of these systems was carried out
with high concentrations of biomass, in many cases
inoculated from other systems, which made feasible
to achieve high NRR since the reactor’s start-up. In the
fluidised bed reactor, concentrations of biomass of 10 g
VSS/l would be necessary to reach NRR of 1500 mg l

−1

per day. Nevertheless, in the MSBR it was possible
to remove 700 mg l

−1

per day with less than 2 g l

−1

biomass.

Granule formation has been referred for both aero-

bic (

Beun et al., 1999

) and anaerobic systems (

Hulshoff

Pol et al., 2004

). Granule formation in aerobic sys-

tems depends in different parameters as e.g. loading
rate, shear stress, oxygen concentration. However the
most important selection mechanism to promote gran-
ulation is hydraulics, especially when settling is used to
promote the separation of the biomass and the treated
water. Microorganisms in granules have higher settling
rates than those growing in dispersed form or in flocs
than are more sensitive to be washed out with the efflu-
ent. For the case of the anaerobic systems as UASB
reactors, granulation has been also based on the selec-
tive wash-out of dispersed sludge or the retention of
biomass in, resulting in an increased growth of retained
heavier sludge agglomerates.

484

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

On the other hand,

Cicek et al. (1999)

has shown

that in systems with submerged membranes, biomass
tended to grow in suspension instead of in granules.
These researchers justify this because of the better
access to the substrate by the biomass and the lesser
problems of diffusion than those of the granules. In this
respect, formation of granular biomass in the MSBR
is an unexpected result of this research. Anammox
granule formation was previously referred by other
authors in an upflow sludge blanket system (

Imajo et

al., 2004

) and gas lift system (

Dapena-Mora et al.,

2004a

), in which the selective wash-out of dispersed

biomass favoured the growth of granules. However,
from our results, it seems that other mechanisms, not
the hydraulic, would be responsible for granule forma-
tion for Anammox biomass.

Fern´andez et al. (2006)

,

operating a SBR with flocculent Anammox biomass,
observed a change in the aspect of the sludge from
flocculent to granular when they added 5 g l

−1

NaCl.

The floc formation indicated by other authors could be
influenced by an interference of the granulation for-
mation by other processes as solids precipitation, or
by the same definition of what a granule is and what
a floc is. In this respect, granule formation by Anam-
mox biomass could be a consequence of an intrinsic
tendency of these microorganisms to grow into aggre-
gates (biofilms or granules).

The growth of biomass as granules might be very

positive for promoting the use of membrane Anam-
mox reactors. This grow could be a reason of the low
membrane fouling observed in our system. Granular
biomass could have better conditions of filterability
than flocculent biomass. The low SVI of 60 ml g

−1

VSS implies that it could be able to retain biomass
concentrations of up to 16 g l

−1

, which implies that the

maximum NRR that could be achieved in this system
would be limited to 5000 or 6000 mg l

−1

per day.

4.2. Precipitation of calcium phosphate salts

As was above-mentioned, the system lost its activity

owing to the precipitation of calcium phosphate salts in
the mixed liquor, when the synthetic medium described
by

Van de Graaf et al. (1996)

was used.

Rosenberger et

al. (2000)

indicated that the utilization of membranes

for the treatment of wastewater with high cellular reten-
tion times can lead to the accumulation of non-volatile
material into the reactor, partly by the accumulation of

Table 2
Ca to P ratio (g g

−1

), pH value measured in the reactor and pH that

may cause phosphate precipitation throughout the three different
operating stages

Stage

Ca/P

Reactor pH

Precipitation pH

Fact

1

5.45

7.8–8.7

<8.0

Salt precipitation
favoured

2

2.71

8.0–8.7

From 8.5 to 9.0

Salt precipitation
favoured

3

0.68

7.8–8.4

9.0

No precipitation

inert material, either because the feeding medium con-
tains these inorganic compounds or these precipitate
in the reactor. The effect of accumulation of inorganic
material in a biological reactor with membranes may be
very different. Moreover, it is not too clear the conse-
quences that to the microorganisms will have the total
retention of this inert material in the system. This can
be especially outstanding in systems with low grow-
ing microorganisms, such as Anammox biomass, since
the inorganic material may be, in some way, an obsta-
cle to the correct development of the microorganisms
(

Wagner and Rosenwinkel, 2000

).

The reasons of calcium and phosphorus salts precip-

itation during the first two operating stages can be stated
from the work of

Song et al. (2002)

. These authors indi-

cated that calcium phosphate precipitation relies on the
pH value and the Ca to P ratio of the water. In fact the
optimum pH at which precipitation occurs depended
on the Ca to P ratio of the water as suggested below:

• optimum pH value for precipitation higher than 9 if

Ca/P ratio is 1.67;

• optimum pH value for precipitation higher than 8.5

if Ca/P ratio is 3.33;

• optimum pH value for precipitation higher than 8 if

Ca/P is 5.00;

• optimum pH value for precipitation higher than 7.5

if Ca/P is 6.67.

The evolution of the pH value throughout the oper-

ation of the reactor is shown in

Table 2

. It can be

concluded that during the first two stages precipitation
occurred due to the pH and Ca to P ratio used, as the
conditions were optimum for precipitation. However,
during the operating stage 3 the operating conditions,
in which Ca to P ratio was 0.68 and pH was between
7.8 and 8.4, did not favour precipitation. These facts are
according to the evolution of NVSS and the results of

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

485

the elementary analysis of the surface of the biomass,
during the three experimental stages.

The

Van de Graaf et al. (1996)

medium that caused

the formation of precipitates in the MSBR, was used
successfully during the operation of other Anammox
reactors fed with synthetic media (

Strous et al., 1997;

Kuai and Verstraete, 1998; Imajo et al., 2004; Dapena-
Mora et al., 2004a,b

).

Dapena-Mora et al. (2004b)

referred NRR up to 600 mg l

−1

per day in an Anam-

mox SBR system operating with an inorganic solids
content between 15 and 25% TSS. Moreover, we have
also detected the presence of a high fraction of cal-
cium phosphate salts in the Anammox SBRs fed with
the Van der Graaf medium in our laboratory (data not
published). In the present study inorganic solids con-
tent was higher, up to 50% TSS. A possible reason of
the higher accumulation of NVSS in the MBR system
could be the presence of the membrane. This bar-
rier may retain the inorganic precipitation nuclei more
efficiently than in conventional SBR and other bio-
logical systems, causing the accumulation of a higher
inorganic solids percentage in the system that cov-
ered the biomass surface. Salt precipitation interfered
with microbial activity and caused a decrease of the
NRR of the reactor from 100 to only 10 mg l

−1

per

day observed during the operating stages 1, and the
low NRR of around 5 mg l

−1

per day obtained during

stage 2.

Salt precipitation might limit the use of MSBR sys-

tems for some wastewaters. Nevertheless, it could be
feasible to perform a pre-treatment stage of the water
to avoid the possible precipitation in the MSBR. Some
authors also recommended for membrane systems
operational strategies based on low purges of biomass
in order to avoid accumulation of inorganic material
(

Rosenberger et al., 2000; Wagner and Rosenwinkel,

2000

). In this sense, a feasible solution for Anammox

MSBR treatment could be to promote simultaneous
precipitation of salts in the previous partial nitrification
stage to nitrite that is carried out before the Anammox
reactor.

5. Conclusions

The operation of the MSBR was unstable during the

first two stages due to salts precipitation. Salts precipi-
tation on the biomass surface interferes with microbial

activity and caused a decrease of the nitrogen removal
rate (NRR) of the reactor from 100 to only 10 mg l

−1

per day during operating stage 1. Experimental results
have shown that the precipitation was a result of the use
of the

Van de Graaf et al. medium (1996)

in the MSBR.

The membrane acted as a barrier that may retain the
inorganic precipitation nuclei more efficiently than in
conventional SBR and other biological systems, caus-
ing the accumulation of around 50% NVSS in the
biomass. Modification of the Ca and P concentration
of this medium were necessary to avoid precipitation.
The nitrogen loading rate could be increased during
stage 3, by avoiding salt precipitation in the system,
and nitrogen removal rate was up to 710 mg l

−1

per

day. NVSS percentage in the biomass diminished dur-
ing period 3 till 18%. On the other hand, the use of a
membrane did not shorten the period of time that is nec-
essary to obtain a system operating at high NLR. In this
respect, this could be explained because the doubling
time observed in the MSBR of 18 days was similar to
that observed by other authors as SBRs and biofilm
systems subjected to biomass losses with the effluent.
Moreover, the stoichiometric parameters of the reac-
tion found in the membrane system were similar to
those found in the bibliography.

The MSBR could be a suitable system for nitro-

gen removal using Anammox biomass. Either biomass
wash-out or contact with air were avoided by the
use of the membrane. The system maintained a good
activity even during periods in which a little amount
of nitrite accumulated and sporadic sludge buoyancy
was detected. Nitrite accumulation was removed by
stopping the feeding during an operating cycle. In
fact nitrite concentration after that action was below
1 mg l

−1

.

The behaviour of the aspect and settling properties

of the sludge were good. The value of the SVI decreased
during stage 3 from around 125 to 60 ml g

−1

VSS, indi-

cating an increase of the biomass density. Moreover,
biomass growth did not occur as suspended biomass but
in granules with an irregular cauliflower appearance.
This could be a result of the tendency of Anammox
microorganisms to grow into biofilms or granules.

The growth of biomass as granules can be very posi-

tive membrane reactors. This growth could be a reason
of the low membrane fouling observed in our system.
Granular biomass could have better conditions of fil-
terability than flocculent biomass.

486

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

Acknowledgements

To the Spanish Ministry of Science (CTQ2005-

04935) and the Xunta de Galicia through the GRAFAN
project (PGIDIT04TAM26500PR). Authors also want
to thank the Zenon, Environmental Inc. for the kind
supply of the membrane fibres.

References

Amann, R.I., 1995. In situ identification of microorganisms by

whole-cell hybridization with r RNA-targeted nucleic acid
probes. In: Akkerman, A.D.L., van Elsas, J.D., de Brujin, F.J.
(Eds.), Molecular Microbial Ecology Manual. Kluwer Academic
Publisher, Dordrecht, The Netherlands, pp. 1–15.

APHA-AWWA-WPCF, 1985. Standard Methods for Examination of

Water and Wastewater, 16th ed. APHA-AWWA-WPCF, Wash-
ington.

Balmelle, B., Nguyen, K.M., Capdeville, B., Cornier, J.C., Deguin,

A., 1992. Study of factors controlling nitrite build-up in biologi-
cal processes for water nitrification. Water Sci. Technol. 26 (5/6),
1017–1025.

Beun, J.J., Hendriksi, A., van Loosdrecht, M.C.M., Morgenroth,

E., Wilderer, P.A., Heijnen, J.J., 1999. Aerobic granulation in
a sequencing batch reactor. Water Res. 33 (10), 2283–2290.

Cicek, N., Franco, J.P., Suidan, M.T., Urbain, V., Manem, J., 1999.

Characterization and comparison of a membrane bioreactor and a
conventional activated-sludge system in the treatment of wastew-
ater containing high-molecular-weight compounds. Water Envi-
ron. Res. 71, 64–70.

Daims, H., Bruhl, A., Amann, R., Schleifer, K.H., Wagner, M., 1999.

Probe EUB338 is insufficient for the detection of all bacteria:
development and evaluation of a more comprehensive probe set.
Syst. Appl. Microbiol. 22 (3), 434–444.

Dapena-Mora, A., Campos, J.L., Mosquera-Corral, A., Jetten,

M.S.M., M´endez, R., 2004a. Stability of the Anammox process
in a gas-lift reactor and a SBR. J. Biotechnol. 110, 159–170.

Dapena-Mora, A., Arrojo, B., Campos, J.L., Mosquera-Corral, A.,

M´endez, R., 2004b. Improvement of the settling properties of
Anammox sludge in an SBR. J. Chem. Technol. Biotechnol. 79
(12), 1412–1420.

Fern´andez, I., V´azquez-Pad´ın, J.R., Mosquera-Corral, A., Campos,

J.L., M´endez, R., 2006. Biofilm and granular systems to improve
Anammox biomass retention. 7th IWA Speciality Conference on
Small Water and Wastewater Systems, March 7–10. Mexico D.F.,
Mexico.

Furukawa, K., Rouse, J.D., Yoshida, N., Hatanaka, H., 2003. Mass

cultivation of anaerobic ammonium-oxidizing sludge using a
novel nonwoven biomass carrier. J. Chem. Eng. Jpn. 36 (10),
1163–1169.

Fux, C., Boehler, M., Huber, P., Brunner, I., Siegrist, H.R., 2002. Bio-

logical treatment of ammonium-rich wastewater by partial nitrita-
tion and subsequent anaerobic ammonium oxidation (Anammox)
in a pilot plant. J. Biotechnol. 99, 295–306.

Garrido, J.M., van Benthum, W.A.J., van Loosdrecht, M.C.M., Hei-

jnen, J.J., 1997. Influence of dissolved oxygen concentration
on nitrite accumulation in a biofilm airlift suspension reactor.
Biotechnol. Bioeng. 53, 168–178.

Hellinga, C., Schellen, A.A.J.C., Mulder, J.W., Van Loosdrecht,

M.C.M., Heijnen, J.J., 1998. The SHARON process: an inno-
vative method for nitrogen removal from ammonium-rich waste
water. Water Sci. Technol. 37, 135–142.

Hulshoff Pol, L.W., de Castro Lopes, S.I., Lettinga, G., Lens, P.N.L.,

2004. Anaerobic sludge granulation. Water Res. 38, 1376–
1389.

Imajo, U., Tokutomi, T., Furukawa, K., 2004. Granulation of Anam-

mox microorganisms in up-flow reactors. Water Sci. Technol. 49
(5/6), 155–163.

Jetten, M.S.M., Logemann, S., Muyzer, G., Robertson, L.A., De

Vries, S., van Loosdrecht, M.C.M., Kuenen, J.G., 1997. Novel
principles in the microbial conversion of nitrogen compounds.
Antonie van Leeuwenhoek 71 (1/2), 75–93.

Kuai, L., Verstraete, W., 1998. Ammonium removal by the oxygen-

limited autotrophic nitrification–denitrification system. Appl.
Environ. Microbiol. 64, 4500–4506.

Mulder, A., van de Graaf, A.A., Robertson, L.A., Kuenen, J.G.,

1995. Anaerobic ammonium oxidation discovered in a denitri-
fying fluidized bed reactor. FEMS Microbiol. Ecol. 16, 177–
184.

Ramalho, R.S., 1991. Introduction to Wastewater Treatment Pro-

cesses, 2nd ed. Academic Press, London.

Rosenberger, S., Witzig, R., Manz, W., Szewzyk, U., Kraume, M.,

2000. Operation of different membrane bioreactors: experimental
results and physiological state of the microorganisms. Water Sci.
Technol. 41 (10/11), 269–277.

Sliekers, A.O., Third, K., Abma, W., Kuenen, J.G., Jetten, M.S.M.,

2003. CANON and Anammox in a gas-lift reactor. FEMS Micro-
biol. Lett. 218, 339–344.

Song, Y., Hahn, H.H., Hoffmann, E., 2002. Effects of solu-

tion conditions on the precipitation of phosphate for recov-
ery. A thermodynamic evaluation. Chemosphere 48 (10), 1029–
1034.

Strous, M., van Gerven, E., Kuenen, J.G., Jetten, M., 1997. Ammo-

nium removal from concentrated waste streams with the anaero-
bic ammonium oxidation (Anammox) process in different reactor
configurations. Water Res. 31, 1955–1962.

Strous, M., Heijnen, J.J., Kuenen, J.G., Jetten, M.S.M., 1998. The

sequencing batch reactor as a powerful tool for the study of slowly
growing anaerobic ammonium-oxidizing microorganisms. Appl.
Microbiol. Biotechnol. 50, 589–596.

Strous, M., Kuenen, J.G., Jetten, M.S.M., 1999. Key physiological

parameters of anaerobic ammonium oxidation. Appl. Microbiol.
Biotechnol. 65, 3248–3250.

Van de Graaf, A.A., Mulder, P., de Bruijn, P., Jetten, M.S.M.,

Robertson, L.A., Kuenen, J.G., 1995. Anaerobic oxidation of
ammonium is a biologically mediated process. Appl. Environ.
Microbiol. 61, 1246–1251.

Van de Graaf, A.A., de Bruijn, P., Robertson, L.A., Jetten,

M.S.M., Kuenen, J.G., 1996. Autotrophic growth of anaerobic
ammonium-oxidizing microorganisms in a fluidized bed reactor.
Appl. Environ. Microbiol. 142, 2187–2196.

C. Trigo et al. / Journal of Biotechnology 126 (2006) 475–487

487

Van Dongen, U., Jetten, M.S.M., Van Loosdrecht, M.C.M., 2001.

The SHARON-Anammox process for treatment of ammonium
rich wastewater. Water Sci. Technol. 44, 153–160.

Van Loosdrecht, M.C.M., Salem, S., 2005. Biological treatment of

sludge digester Liquids. In: Proceedings of the IWA Specialized
Conference on Nutrient Management in Wastewater Treatment
Processes and Recycle Streams, Krakow, Poland, September
19–21, 2005, pp. 13–22.

Wagner, J., Rosenwinkel, K.H., 2000. Sludge production in mem-

brane bioreactors under different conditions. Water Sci. Technol.
41 (10/11), 251–258.

Wyffels, S., Boeckx, P., Pynaert, K., Zhang, D., Van Cleemput, O.,

Chen, G., Verstraete, W., 2004. Nitrogen removal from sludge
reject water by a two-stage oxygen-limited autotrophic nitri-
fication denitrification process. Water Sci. Technol. 49 (5/6),
57–64.