wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 2

Avai lab le a t www.sc iencedi rec t .com

journa l homepage : www.e lsev ie r . com/ loca te /wat res

The nature of the carbon source rules the competitionbetween PAO and denitrifiers in systems for simultaneousbiological nitrogen and phosphorus removal

Javier Guerrero 1, Albert Guisasola*, Juan A. Baeza 2

Departament d’Enginyeria Quımica, Escola d’Enginyeria, Universitat Autonoma de Barcelona, 08193 Bellaterra (Barcelona), Spain

a r t i c l e i n f o

Article history:

Received 8 February 2011

Received in revised form

15 June 2011

Accepted 17 June 2011

Available online 30 June 2011

Keywords:

Carbon source

EBPR

Nitrate

OHO

PAO

VFA

* Corresponding author. Tel.: þ34 93 581 187E-mail addresses: franciscojavier.guerrer

(J.A. Baeza).1 Tel.: þ34 93 581 4798; fax: þ34 93 581 2012 Tel.: þ34 93 581 1587; fax: þ34 93 581 201

0043-1354/$ e see front matter ª 2011 Elsevdoi:10.1016/j.watres.2011.06.019

a b s t r a c t

The presence of nitrate in the theoretical anaerobic reactor of a municipal WWTP aiming

at simultaneous C, N and P removal usually leads to Enhanced Biological Phosphorus

Removal (EBPR) failure due to the competition between PAO and denitrifiers for organic

substrate. This problem was studied in a continuous anaerobiceanoxiceaerobic (A2/O)

pilot plant (146 L) operating with good removal performance and a PAO-enriched sludge

(72%). Nitrate presence in the initially anaerobic reactor was studied by switching the

operation of the plant to an anoxiceaerobic configuration. When the influent COD

composition was a mixture of different carbon sources (acetic acid, propionic acid and

sucrose) the system was surprisingly able to maintain EBPR, even with internal recycle

ratios up to ten times the influent flow rate and COD limiting conditions. However, the

utilisation of sucrose as sole carbon source resulted in a fast EBPR failure. Batch tests with

different nitrate concentrations (0e40 mg L�1) were performed in order to gain insight into

the competition for the carbon source in terms of P-release or denitrification rates and P-

release/C-uptake ratio. Surprisingly, no inhibitory or detrimental effect on EBPR perfor-

mance due to nitrate was observed. A model based on ASM2d but considering two step

nitrification and denitrification was developed and experimentally validated. Simulation

studies showed that anaerobic VFA availability is critical to maintain EBPR activity.

ª 2011 Elsevier Ltd. All rights reserved.

1. Introduction Removal (EBPR) and nitrification/denitrification for simulta-

Biological nutrient removal (BNR) is considered the most

economical and sustainable technology to meet the increas-

ingly stricter discharge requirements in wastewater treat-

ment. Although many wastewater treatment plants (WWTP)

have already adapted their operation to meet stringent

nutrient discharge limits, many others do not satisfy these

requirements due to failures in the BNR processes. WWTP

configurations integrating Enhanced Biological Phosphorus

9; fax: þ34 93 581 2013.o@uab.cat (J. Guerrero), al

3.3.ier Ltd. All rights reserved

neous C, N and P removal require an anaerobic reactor after

the inlet and the presence of an aerobic reactor before the

settling process to promote Polyphosphate Accumulating

Organisms (PAO) growth. Nitrification in the aerobic reactor

may result in nitrate recirculation to the anaerobic reactor

through the external recycle. This availability of nitrate in the

supposedly anaerobic reactor is one of the most reported

causes of EBPR failure in full-scaleWWTP. Some studies (Kuba

et al., 1994; Patel andNakhla, 2006) indicated that the presence

bert.guisasola@uab.cat (A. Guisasola), juanantonio.baeza@uab.cat

.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 24794

of nitrate prevented anaerobic P-release and thus EBPR

activity, which only occurred after nitrate depletion

(<1 mg L�1). Two different hypotheses have been reported in

the literature so far. On one hand, some studies (Van Niel

et al., 1998; Saito et al., 2004) linked the detrimental effect of

nitrate on EBPR to the presence of some denitrification inter-

mediates (nitrite or nitric oxide) which would have an inhib-

itory effect on PAO. On the other hand, the EBPR failure could

be due to an insufficient amount of organic carbon, COD, as

substrate for nutrient removal. The presence of nitrate would

trigger the activity of ordinary heterotrophic organisms (OHO)

which would reduce nitrate using COD as electron donor and

result in less COD available for PAO growth.

Several external carbon sources have been studied to

balance the abovementioned COD deficiency in wastewaters

(Gerber et al., 1986; Jones et al., 1987;Winter, 1989; Appeldoorm

et al., 1992; Isaacs et al., 1994; Hallin et al., 1996). Among those,

acetic acid was suggested as the most effective carbon source

for improving BNR. However, Cho and Molof (2004) reported

that acetic acid was preferentially degraded by denitrifying

bacteria over PAO, which were outcompeted for the carbon

source. For this reason, the delicate balance between organic

carbon, N and P levels has a major impact to enhance the P

removal in BNR systems. In this sense, EBPR failures from

urban wastewater with low or medium organic content have

been reported (Tasli et al., 1999). Another factor to consider in

theeffectivenessof EBPR is thenatureof the carbonsource that

plays the electron donor role. Randall et al. (1997) proved that

the presence of volatile fatty acids (VFA) was imperative to

obtain a high P-removal capacity. In addition, Pijuan et al.

(2004) and Oehmen et al. (2006) also proved that propionic

acid favoured PAO enrichment.

Regarding the electron acceptor, a fraction of PAO called

denitrifying PAO (DPAO), can uptake effectively phosphorus

linked to denitrification under anoxic conditions using poly-

hydroxyalkanoates (PHA) previously accumulated under

anaerobic conditions. Hence, DPAO can be useful for

achieving simultaneous phosphorus and nitrogen removal

from wastewaters with carbon shortage (Kerrn-Jespersen and

Henze, 1993; Kuba et al., 1996). Some authors recently found

that PAO can be divided into two types with different deni-

trifying capabilities (Carvalho et al., 2007; He et al., 2007;

Flowers et al., 2009). One clade (named IA) was able to

couple nitrate reductionwith phosphorus uptake, but another

(named IIA) could only use nitrite in addition to oxygen.

- -

Fig. 1 e A2/O and MLE pilo

The overall objective of this work was to study the role

played by the nature of the carbon source in the intricate

competition between PAO and OHO for the organic substrate

under anoxic conditions. An anoxic/aerobic modified

Ludzack-Ettinger (MLE) continuous pilot plant (146 L) for

simultaneous biological organic matter, N and P removal was

operated with different internal recycle ratios to study the

detrimental effect of nitrate presence in the anaerobic reactor.

Different organic matter concentrations and compositions

were also used at different steps to induce EBPR failure.

Finally, a mathematical model to describe the behaviour of

the system was developed and validated. Different scenarios

were simulated to obtain a better understanding on the role of

the carbon source on EBPR feasibility under anoxic and aerobic

conditions.

2. Material and methods

2.1. Pilot-plant description

Thepilotplant (146L) consistedof three continuousstirred tank

reactors and one settler (Fig. 1). The plantwas initially operated

with the classical anaerobiceanoxiceaerobic (A2/O) configura-

tion for simultaneous C, N and P removal. The first reactor (R1,

28L) was anaerobic so that PAO were selected against other

OHO. Nitrate entering to the second reactor (R2, 28L) withQRINT

was denitrified by either OHO or DPAO. The third reactor (R3,

90 L) worked under aerated conditions and complete organic

matterandP removal tookplace togetherwithnitrification.The

settler (50 L) produced an effluent stream and a biomass

enriched streamwhichwas returned to the systemthrough the

external recycle (QREXT). Mixed liquor was withdrawn daily

from the three reactors in order to keep a desired sludge

retention time (SRT), around 15 � 2 days. The influent (QIN)

flow-rate was 140 L d�1 while QREXT was maintained around

125 L d�1 during all the experiments. This configuration was

maintained during 4 months under steady-state conditions.

However, during most of this work, the plant was working

withMLEconfiguration (anoxiceaerobic)whenQRINTwasmoved

to R1 (Fig. 1), which did not behave anaerobic anymore. This

configuration, typical of systems designed for only biological C

and N removal, was chosen for gaining insight into the effect of

nitrate entering to the anaerobic phase on the P removal. In fact,

t plant configurations.

Table 2 e Pilot plant conditions for each experimentalstep.

Experiment Influentcomposition,

mg L�1

(COD:N:P)

QRINT

flow-rate,L d�1

QRINT/QIN

Plantconfiguration

Step 0 400:40:10 420 3 A2/O

Step I 400:40:10 420 3 MLE

Step II 400:40:10 840 6 MLE

Step III 400:40:10 1400 10 MLE

Step IV 200:40:10 420 3 MLE

Step Va 400:40:10 420 3 MLE

a Sucrose was used as sole carbon source.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 2 4795

in most of the cases, nitrate was completely depleted in R1 as

COD resulting in an anoxiceanaerobiceaerobic configuration.

The three reactors were monitored on-line with DO (Desin

DO2-WW), pH (Desin EPH-M11) and temperature (Pt-100)

probes connected to signal converters (Desin TM-3659). R3was

alsoequippedwithammoniumandnitrateprobes (HachLange

scNH4DandscNO3D)andanon-linephosphateanalyser (Hach

Lange PHOSPHAX)with a samplefiltration system (Hach Lange

FILTRAX). On-line data weremeasuredwith a data acquisition

card (Advantech PCI-1711) connected to a PC, with Lab-

Windows CVI 2009 software for process monitoring and

control. The data acquisition card had several analog and

digital outputs for actuation over the pumps, stirrers and

valves. The pH was controlled at 7.25 � 0.05 using an on-off

controller with sodium carbonate (1 M) dosage. Dissolved

oxygen (DO) in R3was controlled at 1.75� 0.25mgDO L�1 with

an on/off controller. Synthetic influent was prepared from

a concentrated feed (Table 1) that was diluted (20:1) with tap

water. The concentration of organicmatter in the influentwas

different throughout the study (Table 2). The micronutrients

composition was adapted from Smolders et al. (1994). Sludge

frommunicipal WWTP fromManresa (Barcelona) was used to

inoculate the pilot plant. PAO content was analysed by FISH

quantification resulting in less than 2% of the total biomass.

2.2. Batch experiments

Off-line batch experiments were performed in a magnetically

stirred vessel (2 L). This system could be operated either under

anaerobic/anoxic or aerated conditions by sparging nitrogen

or oxygen gas, respectively. This gas was supplied through

a microdiffuser which ensured good transfer from gas to

liquid phase. The gas flow was controlled with a mass flow-

meter (Bronckhorst HiTec 825) to ensure a constant flow.

Table 1 e Concentrated feed composition.

Composition g L�1

Macronutrients

Sodium acetate (C2H3O2Na)a 2.20/4.39

Sodium propionate (C3H5NaO2)a 1.38/2.77

Sucrose (C12H22O11)a 0.94/1.87

Ammonium chloride (NH4Cl)a 3.06

Dipotassium phosphate (K2HPO4)a 0.74

Potassium phosphate (KH2PO4)a 0.29

Micronutrients

Magnesium sulphate (MgSO4$7H2O) 0.88

Calcium chloride (CaCl2$2H2O) 1.40

Potassium chloride (KCl) 0.38

Ferric chloride (FeCl3$6H2O) 1.50b

Potassium iodide (KI) 0.18b

Boric acid (H3BO3) 0.15b

Cobalt chloride (CoCl2$6H2O) 0.15b

Manganese chloride (MnCl2$4H2O) 0.12b

Zinc sulphate (ZnSO4$7H2O) 0.12b

Sodium molybdate (Na2MoO4$2H2O) 0.06b

Copper sulphate (CuSO4$5H2O) 0.03b

EDTA (C10H16N2O8) 10.00b

a Main components: 4/8 g COD L�1 (37.5% acetate, 37.5% propio-

nate and 25% sucrose), 0.8 g N L�1 y 0.2 g P L�1.

b Trace solution: 1 mL introduced per L of influent.

The pH (WTW Sentix 81) and DO (WTW CellOx 325) probes

were connected to a multiparametric terminal (WTW INOLAB

3) which was in turn connected via RS232 to a PC allowing for

data monitoring and storage. This software also manipulated

a high precision microdispenser (Crison Multiburette 2S) for

pH control with acid/base addition. More detailed information

about this equipment can be found in Guisasola et al. (2007).

The batch experiments aimed at studying the competition

for influent COD between OHO and PAO under anoxic condi-

tions and the inhibitory effect of nitrate on EBPR. The procedure

followed was : i) the vessel was filled with biomass (around

2000 mg TSS L�1) from the pilot plant when it was operating

under A2/O conditions, whichwas left under aerobic conditions

for 12 h to ensure PHA depletion and thus, to obtain the

maximum P-release after carbon source addition; ii) a pulse of

acetic acid (350mgCODL�1) at different nitrate concentration (0

or 40 mg NeNO3� L�1) was added under nitrogen-sparging

conditions and the major components (COD, P, NeNO2�,

NeNO3�)weremonitored.AfterCODdepletion,a secondpulseof

nitratewas added (20mg L�1) in the experimentswith nitrate to

monitor theanoxicP-removal. Finally, the systemwasswitched

to aerobic conditions after 10 h tomonitor aerobic P removal.

2.3. Analytical methods

Ammonium was analysed by means of a continuous flow

analyzer based on a potentiometric determination of

ammonia (Baeza et al., 1999). Nitrate and nitritewere analysed

with ionic chromatography (DIONEX ICS-2000). Phosphate

was measured by a phosphate analyser (PHOSPHAX sc) based

on the Vanadomolybdate yellow method, where a two-beam

photometer with LEDS measured the phosphate specific

yellow colour. Organic matter, mixed liquor total suspended

solids (TSS) and mixed liquor volatile suspended solids (VSS)

were analysed according to APHA (1995).

Fluorescence in situ hybridisation (FISH) technique

(Amman, 1995) coupled with confocal microscopy was used to

quantify the biomass distribution as in Jubany et al. (2009).

Hybridizations were performed using at the same time a Cy3-

labelled specific probe and Cy5-labelled EUBMIX for most

bacteria (Dains et al., 1999). Ammonium oxidising bacteria

(AOB) detection was performed with Nso190 specific probe

while NIT3 probe was used for nitrite oxidising bacteria (NOB)

hybridization (Jubany et al., 2009). PAOMIX probe was used to

quantify most of PAO bacteria (Crocetti et al., 2000) and the

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 24796

methodology proposed by Flowers et al. (2009) was used to

hybridize PAO clade IA and clade IIA. GAOMIX, DF1MIX and

DF2MIX probes were used to quantify glycogen accumulating

organisms (GAO) according to the methodology described in

Guisasola et al. (2009).

2.4. Model description

The model used (see Supplementary information) is an

extension of the well-known Activated Sludge Model 2d

(ASM2d) proposed by IWA that describes the different

processes occurring in a system for simultaneous biological

organic matter and nutrient removal (Henze et al., 2000). The

major extensionwas the inclusion of nitrite as a state variable.

Nitrite is a key intermediate to understand thebehaviour of the

different PAO fractions since some of them can use either

nitrate and/or nitrite as electron acceptor. Moreover, nitrite

may have an inhibitory effect on EBPR as some authors repor-

ted (VanNiel et al., 1998; Saito et al., 2004). The anoxic P-uptake

with nitrite as electron acceptor could be also considered

according to the different denitrifying capabilities of DPAO

biomass (He et al., 2007; Flowers et al., 2009). Hence, nitrifica-

tion was modelled as a two-step process, including AOB and

NOB. Denitrification was also described in two steps (nitrate to

nitrite and nitrite to nitrogen gas) to understand the COD fate

under anoxic conditions and the possible substrate competi-

tion with PAO. The extended model included 21 compounds,

which were divided into soluble or particulate, and 28

processes. The process kinetics, stoichiometry and parameter

valuesmatrix can be found in the Supplementary information.

The model was integrated with Matlab� using the ode15s

function, a variable order method recommended for stiff

systems. The parameter estimation of the new processes

considered was carried out by using NeldereMead Simplex

search method ( fminsearch Matlab function). The settler was

modelled using the model of Takacs et al. (1991). The starting

point for each simulation was the steady state of the system

under A2/O during 100 days in order to obtain the initial

conditions of the plant (Step 0).

3. Results

3.1. Pilot-plant operation

Tables 2 and 3 describe, respectively, the different plant

configurations and its corresponding steady-state effluent

composition. Step 0 corresponds to the starting point of this

study, the A2/O configuration (see Fig. 1). In the following steps

(Step IeIII), the plant configuration was moved to a MLE

Table 3 e Steady-state effluent composition and PAO percenta

Experiment COD (mg COD L�1) NeNH4þ (mg N L�1) N

Step 0 22.4 � 0.1 0.21 � 0.07

Step I 23.2 � 0.8 0.08 � 0.01

Step II 22.8 � 1.4 <0.05

Step III 18.4 � 0.7 0.32 � 0.15

configuration for a better assessment of the detrimental

effects of NOX (nitrate or denitrification intermediates) in the

EBPR performance. The value ofQRINT was gradually increased

among these periods. Different FISH analyses were performed

at the end of each step in order to quantify PAO sludge

content.

Fig. 2 shows the experimental profiles of the main

compounds of the influent and the effluent during the steps

0 to III. High N and P removal (around 80% and 98% respec-

tively, as Table 3 shows) was achieved with the conventional

A2/O configuration despite a little amount of NOX entering

with the external recycle (8 mg NeNOX� L�1 where nitrate was

the predominant compound). Under anaerobic conditions

(R1), 13.6% of the total COD inlet was consumed for denitrifi-

cation of the recycled nitrate, while 53% was taken up by PAO

resulting in P-release. These percentages were calculated

assuming default growth yields: the recycle of 1 mg NeNO3� to

the anaerobic phase would consume 7.6 mg COD, whereas

1 mg PePO4�3 released would consume 2.5 mg COD as VFA

(Henze et al., 2000). FISH quantification performedduring A2/O

step clearly indicated the development of an enriched PAO

sludge (72% of PAO) comparing with PAO content in the

start-up of the plant (see Section 2.1.). Therefore, the existing

PAO were able to coexist with denitrifying OHO in the A2/O

configuration despite the nitrate inlet. This observation con-

trasted to the common textbook knowledge that a strict

anaerobic phase is mandatory to achieve EBPR and that NOX

presence in the anaerobic reactor can be detrimental to EBPR

success (Simpkins and Mclaren, 1978; Van Niel et al., 1998;

Henze et al., 2008).

When the plant configuration was changed from A2/O to

MLE (Step I, Table 2), N and P removal efficiencies slightly

decreased to 74% and 97%, respectively (Table 3). The increase

ofQRINT during step II resulted in a decrease of the effluent NOX

(more than 40%) as more NOX was brought to R1 to be deni-

trified. However, the subsequent increase of the internal

recycle (Step III) did not result in a important decrease in the

NOX (less than 15%) effluent content because the COD

concentration became limiting under anoxic conditions. The

measured effluent COD (Table 3) could be related to inert

organic components. Surprisingly, the net-P removal efficiency

was never affected during the abovementioned MLE operation

(i.e. P removalwas never lower than 85%despite the increase in

the phosphorus concentration at the end of the step III) sug-

gesting that COD was preferentially consumed for EBPR. This

fact was corroboratedwith the FISH quantification obtained for

steps IeIII (Table 3), where PAO population did not show an

important shift during the experiments. Then, a COD-limited

influent (200 mg COD L�1) was proposed in step IV to gain

more insight into the substrate competition between PAO and

ge in the sludge at the end of each experimental step.

eNOX� (mg N L�1) PePO4

�3 (mg P L�1) PAO biomass (%)

7.89 � 0.25 0.21 � 0.01 72 � 9

9.48 � 0.85 0.26 � 0.04 77 � 5

5.46 � 0.46 0.23 � 0.05 68 � 5

4.68 � 0.80 1.09 � 0.67 71 � 5

Time (days)0 5 10 15 20 25 30 35 40 45 50 55 60 65

P-P

O4-3

, N-N

H4+

and

N-

NO

X- (

mg

· L-1

)

0

10

20

30

40

50

60

70

CO

D (

mg

· L

-1)

0

100

200

300

400

500

600MLEQRint:Inf

10:1

MLEQRint:Inf

6:1

MLEQRint:Inf

3:1

A2O

Step IIIStep II Step I Step 0

Fig. 2 e Influent and effluent concentrations during the experimental steps 0eIII. (;) COD inlet, (7) COD outlet, (-)

ammonium inlet, (,) ammonium outlet, (C) phosphorus inlet, (B) phosphorus outlet and (>) NOX outlet.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 2 4797

OHO (Fig. 3). Again, EBPR was not significantly affected by the

COD decrease (P effluent concentration was always lower than

1.5 mg P L�1). On the contrary, the denitrification process was

limited by the carbon source and NOX effluent concentration

increased from 7 mg NeNOX L�1 to 15 mg NeNOX L�1 (nitrate

was around 97% of the total NOX). This observation was again

surprising and in clear disagreement with textbook knowledge

that OHO should outcompete PAO in an anoxic COD-limited

scenario.

Finally, the PAO capacity to outcompete OHO related to the

nature of the carbon source was studied in step V (Fig. 3). For

this purpose, sucrose was used as sole carbon source. The P-

removal capacity was progressively lost after only 4 days

resulting in a high P effluent concentration. Most of the

sucrose (70%) was oxidised by denitrifying OHO since NOX

effluent concentration showed a slightly decreasing trend.

3.2. Batch experiments and model calibration

A batch experiment where a pulse of acetic acid and nitrate

was simultaneously added to our sludge (see Section 2.2) is

shown in Fig. 4. P-release occurrence was not avoided with

high nitrate concentrations when acetic acid was the carbon

source. Moreover, batch tests conducted with and without

Time (days)0 1 2 3 4 5 6 7

P-P

O4-3

, N-N

H4+

and

N-

NO

X- (

mg

· L-1

)

0

4

8

12

16

20Step IV

Fig. 3 e Effluent composition and model predictions with a low

source (Step V). (,) ammonium, (>) NOX and (C) phosphorus. D

dashed line to ammonium and solid line to NOX.

nitrate in the anaerobic phase (Table 4) showed that nitrate

presence did not significantly reduce the P-release rate. These

results seem to be in disagreement with some literature

(Chuang et al., 1996; Artan et al., 1998) which reported that

nitrate presence reduces the P-release rate.

An extended version of ASM2d, detailed in section 2.4 and

in the Supplementary information, was used for a better

understanding of the causes of the experimentally observed

EBPR non-deterioration. The proposed model with default

ASM2d parameters predicted EBPR failure when the QRINT

increased, in contrast to the experimental data. Fig. 4 was

used formodel calibration purposes and the biomass diversity

was fixed according to FISH quantification results: 72% of PAO,

4.0% of AOB, 6.7% of NOB and 0.5% of GAO (not included in the

model). The rest (18%) were considered OHO. The parameters

obtained after the model calibration process are presented in

Table 5.

3.3. Pilot plant simulations

The utilisation of the model with the calibrated parameters

allowed a proper description of the experimental results when

increasing the QRINT during Steps IeIII (Fig. 5). No EBPR failure

was predicted in coincidence with the experimental results,

Time (days)

0 1 2 3 4 5 6 70

4

8

12

16

20Step V

COD inlet (Step IV) and results with sucrose as sole carbon

otted line belongs to the model prediction for phosphorus,

Time (h)0 2 4 6 8 10 12 14

0

20

40

60

80

100

120

HA

c (m

g C

OD

·L-1

)

0

100

200

300

400Nitrate pulse: 20 ppm

Aer

obic

con

ditio

ns

Ano

xic

cond

ition

s

P-P

O4-3

and

N-

NO

X- (

mg

· L-1

)

Fig. 4 e Experimental batch test (2L) for model calibration

purposes (;) COD, (>) NOX and (C) phosphorus. Dotted

line belongs to the phosphorus behaviour described by the

model, solids line to NOX and dashed line to COD.

Table 5 e Parameters obtained in themodel calibration ofthe batch experiment with acetate and nitrate.

Parameters ASM2d value(20 �C)

Calibratedvalue

Units

qPHA 3.00 5.00 mg XPHA mg

XPAO�1 d�1

qPP 1.50 0.60 mg XPP mg

XPAO�1 d�1

qPAO 1.00 0.56 d�1

hNO3 ; PAO 0.60 0.07 e

hNO2 ; PAOa e 0.90 e

hNO3 ; OHO 0.80 0.90 e

hNO2 ; OHOa e 0.90 e

a These parameters do not appear in ASM2d model (Henze et al.,

2000).

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 24798

even when a low COD content influent was used (step IV,

Fig. 3). Moreover, a simulation based study was performed in

order to investigate the competition between PAO and OHO

for the carbon source. Two different issues were analysed: i)

the effect of COD influent concentration (with the same

sucrose/VFA ratio shown in Table 1) on the competition of

PAO and OHO ii) how the EBPR could be affected by the nature

of the carbon source (i.e. different VFA and sucrose ratioswere

simulated) under COD limiting conditions (200 mg L�1). Each

scenario was simulated during 7 days to mimic the experi-

mental conditions of steps IV and V. As the model predicted

(Fig. 6), the EBPR failure was lower with the calibrated model

as the capacity of PAO to outcompete OHO for the carbon

source was strengthened with the estimated parameters.

4. Discussion

4.1. Batch experiments

The simultaneous presence of nitrate and organic matter did

not prevent net-P removal, i.e. P-release and subsequent P-

uptake. However, lower P/C ratios (Table 4) were obtained

when nitrate was present. It could be argued that simulta-

neous P-release and anoxic-P-uptake was occurring, but this

hypothesis should lead to a certain decrease in the P-release

Table 4 e Major transformations obtained in the batchstudies with acetic acid.

Initial NeNO3�

concentration

0 mg L�1 40 mg L�1

P-Release Rate (g PePO4�3 g TSS�1 d�1) 1.27 1.22

Nitrate uptake rate (g NeNO�3 g TSS�1 d�1) e 0.20

COD uptake rate (g COD g TSS�1 d�1) 2.35 3.01

P-release/C-uptake (P mmol/C mmol) 0.43 0.35

rate, which was not detected when comparing experiments

with and without nitrate. Another hypothesis to explain the

lower P/C ratios obtainedwith nitrate presence and acetic acid

as sole carbon source is the simultaneous oxidation of COD by

OHO with nitrate as electron acceptor.

Denitrification activity, i.e. nitrate reduction, was also

observed without COD presence (Fig. 4) which could be

attributed to the presence of DPAO. The denitrification capa-

bilities of the different existing PAO is discussed in Oehmen

et al. (2010) according to the classification proposed by

Flowers et al. (2009), which specified that every PAO pop-

ulation (clade IA and clade IIA) could denitrify from nitrite, but

only some of them (clade IA) could do this process from

nitrate. These observations are in agreement with our results.

We detected that part of our PAO could denitrify from nitrate

since FISH quantification resulted in 22.3% of clade IA and

77.6% of clade IIA from the total PAO population quantified

with PAOMIX probe.

4.2. Feasibility of P removal in MLE system

EBPR deterioration was not observed after switching its

configuration toMLE despite the increasedNOX inlet and thus,

high P removal was obtained. These results seem to challenge

the widely accepted idea that denitrifying OHO outcompete

PAO when competing for the electron donor.

The carbon source used in this work is a combination of

propionic acid, acetic acid and sucrose. Different results can

be found in the literature depending on the carbon source

used. Some authors (Kuba et al., 1994; Patel and Nakhla, 2006)

indicated that P-release should only occur when denitrifica-

tion is completed (NeNO3� < 1mg L�1) withmost of the carbon

sources except with acetic acid. With the latter, some reports

have found that simultaneous nitrate reduction and P-release

is observed (Gerber et al., 1986; Chuang et al., 1996; Artan et al.,

1998). In any case, OHO is usually considered to have a pref-

erence for acetic acid in the denitrification process (Cho and

Molof, 2004; Elefsiniotis et al., 2004). With respect to pro-

pionic acid, it is considered as an optimum carbon source for

PAO growth. Some authors observed that propionate may be

relatively easily sequestered and metabolized by PAOs

compared to other microorganisms (Pijuan et al., 2004;

Oehmen et al., 2006). Finally, sucrose is a complex carbon

Reactor 1

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

P-P

O4-3

, N-N

H4+ a

nd N

- N

OX

-

(m

g · L

-1)

0

10

20

30

40

50

60

Reactor 2

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

P-P

O4-3

, N-N

H4+

and

N-

NO

X-

(mg

· L-1

)

0

10

20

30

40

50

60

Reactor 3

Time (days)0 5 10 15 20 25 30 35 40 45 50 55 60 65 70

P-P

O4-3

, N-N

H4+

and

N-

NO

X-

(m

g · L

-1)

0

10

20

30

40

50

60

MLEQRint:Inf

10:1

MLEQRint:Inf

6:1

MLEQRint:Inf

3:1

A2O

Step IIIStep II Step I Step 0

MLEQRint:Inf

10:1

MLEQRint:Inf

6:1

MLEQRint:Inf

3:1

A2O

Step IIIStep II Step I Step 0

MLEQRint:Inf

10:1

MLEQRint:Inf

6:1

MLEQRint:Inf

3:1

A2O

Step IIIStep II Step I

Step 0

Fig. 5 e Model validation. Pilot plant behaviour and model predictions when increasing QRINT (Steps 0eIII). (,) ammonium,

(>) NOX and (C) phosphorus. Dotted line belongs to the phosphorus model prediction, dashed line to ammonium and solid

line to NOX.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 2 4799

source of the synthetic wastewater. Under anaerobic condi-

tions, fermentative bacteria (FB) should be consuming sucrose

rather than PAO or OHO and produce VFA, while in presence

of nitrate it could be an alternative carbon source for deni-

trifying bacteria.

Our experimental results indicate that, under nitrate

limited conditions (i.e. with less nitrate entering the anaerobic

zone than the amount required to oxidise all the influent COD)

COD (mg·L-1)

100 150 200 250 300 350 400

N-N

OX

- and

P-PO

4-3 (

mg·

L-1

)

0

5

10

15

20

25

30

35

A

Fig. 6 e Simulation and experimental results to study the effec

source (B) in the EBPR process. (>) belong to NOX and (B) to pho

ASM2d, black symbols the calibrated model and grey symbols

most of the COD is consumed linked to the EBPR process (Step

IeIII). PAO population remained more or less constant during

the different steps as FISH quantification shown (Table 3). This

was observed under continuous nitrate entrance for more

than 60 days. Batch experiments showed that P-release rate

did not differ when nitrate was added (Table 4). Under COD-

limited conditions (i.e. in step IV the amount of nitrate

entering the anaerobic reactor was enough to oxidise all the

VFA content of the total organic carbon source

0% 25% 50% 75% 100%

N-N

OX

- and

P-PO

4-3 (

mg·

L-1

)

0

5

10

15

20

25

30

35B

t of influent COD content (A) and the nature of the carbon

sphorus. White symbols represent the default values of

the experimental values.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 24800

influent COD) PAO were able to uptake carbon source rapidly

and a high P removal was observed (Fig. 3). NOX content

increased since OHO could not denitrify completely the

influent NOX due to substrate limitations (i.e. the complete

denitrification of NOX present in the anaerobic phase would

have required a COD inlet of 390 mg L�1, 195% higher than the

experimental COD influent concentration). These results do

not agree with the preferential utilization of acetic acid by

OHO reported by Cho andMolof (2004). This capacity of PAO to

outcompete OHO for the carbon source is intrinsically linked

to the nature of the organic matter and to the population

distribution in the sludge (i.e. the high amount of PAO present

in the system). Step V (Fig. 3) showed that the utilization of

a complex carbon source (i.e. sucrose) could favour the OHO

denitrification process against EBPR. Phosphorus effluent

concentration increased from 0.57mg L�1 to 6.63 mg L�1 while

NOX outlet presented a decreasing trend. Under those condi-

tions, PAO biomass needed the VFA produced by FB in sucrose

fermentation. When, sucrose and NOX coexisted most of the

sucrose was oxidised by denitrifying OHO (around 70% of the

COD influent content). This fact could explain the EBPR failure

when sucrose was the sole carbon source in contrast to the

situation when VFA were added.

4.3. Model validation and simulation based study

The experimental results obtained could not be reliably

described using the extended AMS2d model with default

parameters. For that reason, a calibration process was per-

formed (Table 5). The higher maximum rate of PHA storage

(qPHA) obtained after the calibration process could suggest that

the PAO population could be more effectively consuming VFA

thanwith the standard ASM2d values (Henze et al., 2000). This

increase is necessary to describe that PAO would be more

favoured than OHO in terms of VFA competition. Conse-

quently, P-release capacity was almost not affected by NOX

presence. The reduction factor for denitrification from nitrate

for PAO (sNO3, PAO) indicates a low capacity to denitrify from

nitrate to nitrite. However, it was enough to obtain the deni-

trification rates registered in the experimental data. This fact

was in agreement with FISH quantification results when PAO

clade IA and clade IIA were quantified (22.3% and 77.6%

respectively of the total PAO bacteria). Finally, the results of

themodel calibration also indicated a lower P-uptake rate and

lower PAO growth rate than the standard values of ASM2d,

which did not affect the EBPR process in our case.

Fig. 6 shows the results obtained in the model based

study, where several scenarios with the default and cali-

brated parameters (Table 5) were simulated. The non-cali-

brated model predicted the total EBPR failure (Fig. 6A) when

COD content was below 200 mg L�1 (i.e. P effluent concen-

tration was the same as in the influent). In contrast, only

a partial EBPR failure was observed with the calibrated

model, even under strong COD limiting conditions

(100 mg L�1). It should be noted that the denitrification

process was more limited by the influent COD reduction with

the calibratedmodel resulting in an effluent with high nitrate

content. The PAO capacity to outcompete OHO for the carbon

source would explain this fact. Steady-state experimental

points were also included in Fig. 6. As can be observed, model

predictions properly described the experimental phosphorus

values. However, less NOX effluent content was experimen-

tally obtained in contrast to model predictions, suggesting

that denitrification process was more efficient at practice.

These divergent results could be explained if one takes into

account that the model was calibrated when the pilot plant

was operated under A2/O conditions. FISH quantification

results showed an increase on PAO clade IA population after

step III (from 22.3% to 35.6%) and thus, an increase in the

denitrification capacity would explain the experimental

results. Regarding the results where the nature of the carbon

source was analysed (Fig. 6B), the predictions of both models

were quite different. The P removal was almost negligible

and denitrification was not deteriorated with the different

influent tested with the default parameters, because it is

assumed that the carbon source is preferentially used for

denitrification rather than for EBPR. With the calibrated

model the simulations results showed that P and N removal

presented an inverse behaviour. EBPR capacity was highly

affected by the VFA influent content and thus, P removal

decreased as the VFA influent content also decreased. On the

contrary, denitrification was favoured when the influent was

enriched in a complex carbon source (sucrose). These results

may be very helpful in view the designing new systems

for simultaneous biological C, N and P removal when the

influent wastewater composition is known. Also, it could be

used to explain the reasons why some EBPR failures are

observed with nitrate presence and how to solve them.

5. Conclusions

This work shows how the nature of the carbon source rules

the competition between PAO and denitrifying OHO in

systems for simultaneous biological nitrogen and phosphorus

removal. After switching the operation of an A2/O pilot plant

to MLE operation, no inhibitory effect on EBPR due to NOX

presence in the anaerobic phase was observed. When the

carbon source presented a high VFA content, PAO could

outcompete OHO even under anoxiceaerobic configuration.

Heterotrophic denitrification activity was more affected than

EBPR when working with low influent COD.

EBPR failedwhen amore complex compound (sucrose) was

used as a sole carbon source. In that case, NOX presence had

an inhibitory effect in EBPR, not to inhibit the P-release

process itself but to prevent the fermentation process for the

VFA production.

A model was developed and experimentally validated

which explained the EBPR feasibility with nitrate presence

under anoxiceaerobic conditions. The model calibration

allowed a better understanding of the experimental results in

terms of kinetic parameters of PAO and OHO. The simulation

of different scenarios evidenced again that the PAO pop-

ulation could be more effective than OHO consuming VFA

even under anoxic conditions. Although no EBPR failure was

observed even under COD limiting conditions (100 mg L�1),

VFA presence was demonstrated as the key point to trigger

EBPR activity. The calibrated model was validated as a helpful

tool to set the limits to avoid EBPR failures linked to nitrate

presence.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 2 4801

Acknowledgments

Javier Guerrero is grateful for the grant received from the

Spanish government. This work was supported by the Spanish

Ministerio de Ciencia y Tecnologıa (CTM2010-20384). The

authors are members of the GENOCOV research group (Grup de

RecercaConsolidatde laGeneralitatdeCatalunya, 2009SGR815).

Appendix. Supplementary information

Supplementary data associated with this article can be found,

in the online version, at doi:10.1016/j.watres.2011.06.019.

r e f e r e n c e s

Amman, R.I., 1995. In situ identification of microorganisms bywhole cell hybridization with RNA-targeted nucleic acidprobes. In: Ackkermans, A.D.L., van Elsas, J.D., de Bruijn, F.J.(Eds.), Molecular Microbial Ecology Manual. Kluwer AcademicPublications, Dordrecht, Holland.

APHA, 1995. Standard Methods for the Examination of Water andWastewater, nineteenth ed.. American Publishers HealthAssociation, Washington, DC, USA.

Appeldoorm, K.J., Kortstee, J.J., Zehnder, A.J.B., 1992. Biologicalphosphate removal by activated sludge under definedconditions. Water Res. 26, 453e460.

Artan, N., Tasli, R., Orgur, N., Orhon, D., 1998. The fate ofphosphate under anoxic conditions in biological nutrientremoval activated sludge systems. Biotechnol. Lett. 20 (11),1085e1090.

Baeza, J., Gabriel, D., Lafuente, J., 1999. An expert supervisorysystem for pilot WWTP. Environ. Modell. Softw 14, 383e390.

Carvalho, G., Lemos, P.C., Oehmen, A., Reis, A.M., 2007.Denitrifying phosphorus removal: linking the processperformance with the microbial community structure. WaterRes. 41, 4383e4396.

Cho, E., Molof, A.H., 2004. Effect of sequentially combiningmethanol and acetic acid on the performance of biologicalnitrogen and phosphorus removal. J. Environ. Manage. 73,183e187.

Chuang, S.H., Ouyang, C.F., Wang, Y.B., 1996. Kinetic competitionbetween phosphorus release and denitrification on sludgeunder anoxic conditions. Water Res. 30 (12), 2961e2968.

Crocetti, G.R., Hugenholtz, P., Bond, P.L., Schuler, A., Keller, J.,Jenkins, D., Blackall, L.L., 2000. Identification ofpolyphosphate-accumulating organisms and design of 16SrRNA-directed probes for their detection and quantitation.Appl. Environ. Microbiol. 66 (3), 1175e1182.

Dains, H., Bruhl, A., Amann, R., Schleider, K.H., Wagner, M., 1999.The domain-specific probe EUB338 is insufficient for thedetection of all bacteria: development and evaluation ofa more comprehensive probe set. Syst. Appl. Microbiol. 22 (2),434e444.

Elefsiniotis, P., Wareham, D.G., Smith, M.O., 2004. Use of volatilefatty acids from an acid-phase digester for denitrification. J.Biotecnol. 114 (3), 289e297.

Flowers, J., He, S., Yilmaz, S., Noguera, D., McMahon, K.D., 2009.Denitrification capabilities of two biological phosphorusremoval sludges dominated by different ‘CandidatusAccumulibacter’ clades. Environ. Microbiol. Rep. 1 (6),583e588.

Gerber, A., Mostert, E.S., Winter, C.T., Villiers, R.H., 1986. The effectof acetate and other short-chain carbon compounds on thekinetics of biological nutrient removal. Water S.A. 12 (1), 7e12.

Guisasola, A., Vargas, M., Marcelino, M., Lafuente, J., Casas, C.,Baeza, J.A., 2007. On-line monitoring of the enhancedbiological phosphorus removal processes using respirometryand titrimetry. Biochem. Eng. J. 35, 371e379.

Guisasola, A., Qurie, M., Vargas, M.M., Casas, C., Baeza, J.A., 2009.Failure of an enriched nitrite-DPAO population to use nitrateas an electron acceptor. Process Biochem. 44, 689e695.

Hallin, S., Rothman, M., Pell, M., 1996. Adaptation of denitrifyingbacteria to acetate and methanol in activated sludge. WaterRes. 30, 1445e1450.

He, S., Gall, D.L., McMahon, K.D., 2007. “CandidatusAccumulibacter” population structure in enhanced biologicalphosphorus removal sludge as revealed by polyphosphatekinase genes. Appl. Environ. Microbiol. 73 (18), 5865e5874.

Henze, M., Gujer, W., Mino, T., van Loosdrecht, M., 2000. ActivatedSludge Models ASM1, ASM2, ASM2d, ASM3. IWA Publishing,London.

Henze, M., van Loosdrecht, M., Ekama, G., Brdjanovic, D., 2008.Biological Wastewater Treatment. IWA Publishing, London,ISBN 1843391880, pp. 162e169.

Isaacs, S.H., Henze, M., Søeberg, H., Kummel, M., 1994. Externalcarbon source addition as a means to control an activatedsludge nutrient removal process. Water Res. 28, 511e520.

Jones, P.H., Tadwalkar, A.D., Hsu, C.L., 1987. Enhanced uptake ofphosphorus by activated sludge: effect of substrate addition.Water Res. 21, 301e308.

Jubany, I., Lafuente, J., Carrera, J., Baeza, J.A., 2009. Automatedthresholding method (ATM) for biomass fractiondetermination using FISH and confocal microscopy. J. Chem.Technol. Biotechnol. 84 (8), 1140e1145.

Kerrn-Jespersen, J.P., Henze, M., 1993. Biological phosphorusuptake under anoxic and aerobic conditions. Water Res. 27,617e624.

Kuba, T., Wachtmeister, A., van Loosdrecht, M.C.M., Heijnen, J.J.,1994. Effect of nitrate on phosphorus release in biologicalphosphorus removal systems. Water Sci. Technol. 30 (6),263e269.

Kuba, T., van Loosdrecht, M.C.M., Heijnen, J.J., 1996. Phosphorusand nitrogen removal with minimal COD requirement byintegration of denitrifying dephosphatation and nitrificationin a two-sludge system. Water Res. 30 (7), 1702e1710.

Oehmen, A., Saunder, A.M., Vives, M.T., Yuan, Z., Keller, J., 2006.Competition between polyphosphate and glycogenaccumulating organisms in enhanced biological phosphorusremoval systems with acetate and propionate as carbonsource. J. Biotechnol. 123, 22e32.

Oehmen, A., Lopez-Vazquez, C.M., Carvalho, G., Reis, M.A.M., vanLoosdrecht, M.C.M., 2010. Modelling the population dynamicand metabolic diversity of organisms relevant in anaerobic/anoxic/aerobic enhanced biological phosphorus removalprocesses. Water Res. 44, 4473e4486.

Patel, J., Nakhla, G., 2006. Interaction of denitrification and Premoval in anoxic P removal systems. Desalination 201, 82e99.

Pijuan, M., Saunders, A.M., Guisasola, A., Baeza, J.A., Casas, C.,Blackall, L.L., 2004. Enhanced biological phosphorus removalin a sequencing batch reactor using propionate as the solecarbon source. Biotechnol. Bioeng. 85 (1), 56e67.

Randall, A.A., Benefield, L.D., Hill, W.E., Nicol, J.P., Boman, G.K.,Jing, S.R., 1997. The effect of volatile fatty acids on enhancedbiological phosphorus removal and population structure inanaerobic/aerobic sequencing batch reactors. Water Sci.Technol. 35 (1), 153e160.

Saito, T., Brdjanovic, D., van Loosdrecht, M.C.M., 2004. Effect ofnitrite on phosphate uptake by phosphate accumulatingorganisms. Water Res. 38, 3760e3768.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 4 7 9 3e4 8 0 24802

Simpkins, M.J., Mclaren, A.R., 1978. Consistent biologicalphosphate and nitrate removal in an activated sludge plant.Progr. Water Tech. 10 (5e6), 433e442.

Smolders, G.J.F., van der Meij, J., van Loosdrecht, M.C.M.,Heijnen, J.J., 1994. Model of the anaerobic metabolism of thebiological phosphorus removal process: stoichiometry and pHinfluence. Biotechnol. Bioeng. 42 (1994), 461e470.

Takacs, I., Patry, G.G., Nolasco, D., 1991. A dynamic model of theclarification thickening process. Water Res. 25 (10), 1263e1271.

Tasli, R., Orhon, D., Artan, N., 1999. The effect of substratecomposition on the nutrient removal potential of sequencingbatch reactors. Water S.A. 25 (3), 337e344.

Van Niel, E.W.J., Appeldoorn, K.J., Zehnder, A.J.B., Kortstee, G.J.J.,1998. Inhibition of anaerobic phosphate release by nitric oxidein activated sludge. Appl. Environ. Microbiol. 64, 2925e2930.

Winter, C.T., 1989. The role of acetate in denitrification andbiological phosphate removal in modified Bardenpho systems.Water Sci. Technol. 21, 375e385.