Integration of methane removal in aerobic anammox-based granular sludgereactorsCelia M. Castro-Barrosa, Long T. Hoa, Mari K. H. Winklera,b and Eveline I. P. Volcke a

aDepartment of Biosystems Engineering, Ghent University, Gent, Belgium; bDepartment of Civil and Environmental Engineering,University of Washington, Seattle, WA, USA

ABSTRACTCombined partial nitritation-anaerobic ammonium oxidation (anammox) processes have beenwidely applied for nitrogen removal from anaerobic digestion reject water. However, suchstreams also contain dissolved methane that can escape to the atmosphere, hence contributingto global warming. This study investigates the possibility of integrating methane removal inaerobic anammox-based granular sludge reactors, through modelling and simulation. Methaneremoval could be established through aerobic methane-oxidizing bacteria (MOB), denitrifyinganaerobic methane-oxidizing bacteria (damoB, NO2

− + CH4 → N2 + CO2), and/or archaea (damoA,NO3

− + CH4 → NO2− + CO2). The simulation results demonstrated that the combined removal of

nitrogen and methane was feasible at low dissolved oxygen conditions. Aerobic MOB were themain responsible microorganisms for removing methane. A sensitivity analysis of key kineticparameters showed a shift in the methanotrophic populations depending on the mostfavourable parameters for each microbial group, while keeping high nitrogen and methaneremoval efficiencies. Possible methane stripping during aeration could be limited by increasingthe depth within the reactor column at which aeration was supplied. Overall, the integration ofmethane removal in aerobic anammox-based granular sludge reactors seems to be a promisingprocess option to reduce the carbon footprint from wastewater treatment.

ARTICLE HISTORYReceived 7 December 2016Accepted 19 May 2017

KEYWORDSAnammox; methaneoxidation; granular sludge;modelling; simulation

1. Introduction

Wastewater treatment negatively contributes to globalwarming by emitting the greenhouse gases carbondioxide (CO2), nitrous oxide (N2O) [1], and methane(CH4) [2]. CH4 is a strong greenhouse gas that accountsfor 34 CO2 equivalents over a 100-year horizon [3]. CH4

can be emitted after its formation in the sewagesystem or can be stripped to the atmosphere from the

reject water from the anaerobic digester. Even thoughthe energy recovery of dissolved CH4 from reject watermay not be economically attractive, its removal couldsignificantly decrease the carbon footprint of wastewatertreatment plants (WWTPs) [2].

About a decade ago, it was found that anaerobic CH4

oxidation coupled to denitrification can take place [4].Denitrifying anaerobic methane oxidation (damo) can

© 2017 Informa UK Limited, trading as Taylor & Francis Group

CONTACT Eveline I. P. Volcke eveline.volcke@ugent.be Department of Biosystems Engineering, Ghent University, Coupure links 653, 9000 Gent, BelgiumThis article was originally published with errors. This version has been corrected. Please see Erratum (http://dx.doi.org/10.1080/09593330.2017.1383671).

Supplemental data for this article can be accessed at https://doi.org/10.1080/09593330.2017.1334709

ENVIRONMENTAL TECHNOLOGY, 2018VOL. 39, NO. 13, 1615–1625https://doi.org/10.1080/09593330.2017.1334709

be carried out by denitrifying anaerobic methane-oxidiz-ing bacteria (damoB), namely Candidatus Methylomir-abilis oxyfera, which couple CH4 oxidation to nitritereduction [5,6] (Equation (1)) and by denitrifying anaero-bic methanotrophic archaea (damoA), such as Candida-tus Methanoperedens nitroreducens, that uses nitrateas an electron acceptor [7] (Equation (2)).

damoB

3CH4 + 8NO−2 + 8H+ � 3CO2 + 4N2 + 10H2O,

(1)

damoA 2CH4 + 8NO−3 � 2CO2 + 8NO−

2 + 4H2O. (2)

Apart from its dissolved CH4 content, reject water is in thefirst place characterized by high nitrogen concentrations inthe form of ammonium and by high temperatures, and istherefore suitable for the application of anaerobicammonium oxidation (anammox) technology [8], implyingimportant energy and chemical savings compared to con-ventional nitrogen removal over nitrate. Ammonium canbe conveniently removed from reject water with a com-bined partial nitritation–anammox process [9,10], whichcan be implemented in one or two stages. Partial nitrita-tion–anammox reactors have become widely applied atfull-scale for the treatment of high-strength ammoniumwastewaters [11]. A lot of research interest currently goesto their integration for mainstream treatment, followinganaerobic psychrophilic treatment or an aerobic stageoperated at a low sludge retention time [12,13].

Simultaneous ammonium and CH4 removal byanammox and damo microorganisms under anoxic con-ditions has been demonstrated feasible in lab-scalesystems [14–16] and modelling approaches [17,18]. Inpractice, this system needs a preceding step realizingnitritation (i.e. oxidation of ammonium to nitrite byammonium-oxidizing bacteria [AOB]) in order toachieve the complete removal of ammonium (byanammox) and of CH4 (by damoB, Equation (1)).

Simultaneous ammonium and CH4 removal in aerobicreactors, by combining partial nitritation–anammoxsystems with damo conversions in a one-stage system,so far has not been realized experimentally. Mathemat-ical models are excellent tools to explore the feasibilityof this new process, to give guidelines for directed exper-imental work and hence save time. Chen et al. [19] mod-elled simultaneous ammonium and CH4 removal in amembrane biofilm reactor to supply oxygen while avoid-ing CH4 stripping. They assumed that aerobic methaneoxidizers (MOB) would not be present in the systemsince oxygen and CH4 diffused from opposite sides inthe biofilm. Given that membrane systems are relativelycostly and prone to fouling, the present study assessedthe potential integration of CH4 conversion in aerated

partial nitritation–anammox granular sludge reactors,reducing CH4 stripping by keeping the aerationminimal. In granular sludge reactors, biomass is grownin the form of dense, fast-settling granules, resulting incompact systems, which allow a high loading rate. Gran-ular sludge reactors have been successfully applied fornitrogen removal through partial nitritation–anammoxfrom reject water [11]; they also hold potential for nitro-gen removal from the mainstream of WWTPs [20]. Theintegration of methane removal in addition to nitrogenremoval in existing or future partial nitritation–anammox granular sludge reactors would extend thereactor performance with minimizing emissions ofmethane as a greenhouse gas, and is thus very promis-ing. In this study, the potential of combined methaneand nitrogen removal in partial nitritation granularsludge reactors is assessed for the first time.

The model presented in this study is the first toinclude not only damo bacteria and archaea, but alsoMOB as a potentially important microbial group for CH4

removal, besides the bacterial populations playing arole in nitrogen removal (AOB, nitrite oxidizing bacteria[NOB], and anammox bacteria) and heterotrophs(growing on decay products), in a one-stage aerobicanammox-based granular sludge reactor. Microbial com-petition in the granules was investigated through simu-lation, determining the key parameters that govern thepresence or absence of the different populations. A sen-sitivity analysis was carried out to assess the influence ofthe microbial parameter values in this respect. Finally,the potential minimization of CH4 stripping through ade-quate reactor design was addressed as well.

2. Materials and methods

2.1. Modelling simultaneous nitrogen andmethane removal

A model was set up to describe the growth and decay ofdamoB, damoA, anammox bacteria, AOB, NOB, MOB, andheterotrophic bacteria. Nitrification and anammox reac-tions were modelled based on Volcke et al. [21], hetero-trophic transformations were modelled according toHenze et al. [22], and aerobic methane oxidation wasmodelled based on Arcangeli and Arvin [23]. ThedamoB process was modelled according to Winkleret al. [18], and the growth and decay of damoA(besides other species) were added in this study as inChen et al. [19]. The interactions between all microbialgroups involved are summarized in Table 1.

The stoichiometric matrix and kinetic expressions con-cerning the microorganisms involved in the CH4 and nitro-gen removal (i.e. including themethanotrophic populations

1616 C. M. CASTRO-BARROS ET AL.

and anammox bacteria) are presented in the supplemen-tary information (SI) in Tables S1.1 and S1.2, and the corre-sponding parameter values are listed in Table S1.3. Notethat damoB could use both nitrite and ammonium astheir nitrogen source for growth. Since nitrite is involvedin the catabolic metabolism of damoB, this substrateneeds to be present for the survival of the bacteria. There-fore, nitrite was considered the nitrogen source fordamoB in this study, as in the model of Winkler et al. [18]and in accordance with the findings of Ettwig et al. [24].The oxygen inhibition constants for damoB and damoA(KiO2_damoB/A) are not available from the literature andwere considered to be the same as that for anammox bac-teria, 0.01 g O2 m

−3 [25]. The nitrite inhibition constant(KiNO2_damoA) and CH4 half-saturation constant (KCH4_damoA)for damoA were assumed to be the same as that fordamoB, 40 g N m−3 and 0.19 g COD m−3, respectively.

2.2. Granular sludge reactor model

The abovementioned bioconversion reactions wereimplemented in a one-dimensional biofilm model inthe Aquasim simulation environment [26]. A biofilmcompartment was used, considering spherical biomassparticles (granules) with a fixed granule size of0.75 mm. The use of a fixed granule size rather thantaking into account a more detailed granule size distri-bution is sufficient to assess the overall reactor behaviour[27], as is done in this study. The reactor volume was400 m3, of which 100 m3 was occupied by particulatematerial, comprising both active biomass and inerts gen-erated during the decay, and the remaining volume(300 m3) was occupied by the bulk liquid.

The interphase (gas–liquid) transfer rate for CH4 andoxygen was included in themodel to assess CH4 strippingin the system (Section 3.5). The partial pressure of the dis-solved gases was modelled as a function of the reactorheight as in Daelman et al. [28] (see Section S6 in the SI).

2.3. Set-up of the simulation study

The reactor behaviour was simulated at a liquid flow rateof 1000 m3 d−1 and a fixed CH4 influent concentration of100 g COD m−3 (25 g CH4 m

−3) as in Winkler et al. [18]and Chen et al. [19]. The influent was assumed to notcontain any other organic carbon source, which is areasonable assumption for reject water [25] and allowsstraightforward interpretation of the simulation results.Ammonium was the sole source of nitrogen in the influ-ent. Its concentration was varied over a realistic range forreject water, from 100 [29] up to 1500 g N m−3, includ-ing typical values for reject water treated by partial nitri-tation–anammox [24]. A sufficiently long simulation timewas applied to reach steady state in terms of both bulkliquid concentrations and solid concentration profileswithin the granules.

An overview of the different simulations carried outand the specific settings for each case is provided in theSI (Table S2.1). While all bioconversions (Table S1.2) wereincluded in most simulations, some simulations wererun without damoA and/or damo B (i.e. by excluding reac-tions 9–10 and/or 7–8, respectively, from Table S1.2) toassess their effect on the nitrogen removal. In order todeal with the significant differences found in the literaturefor the half-saturation constants of the methanotrophs(methane half-saturation constant for damoA anddamoB, KCH4_damo; ammonium half-saturation constantfor MOB, KNH4_MOB; and nitrite half-saturation constantfor damoB, KNO2_damoB), a sensitivity analysis was per-formed to determine their influence on the competitionamong methanotrophic communities (Table S2.2). Gas–liquid interphase transport was neglected in most simu-lations, corresponding with an ideal scenario withoutCH4 stripping, to assess the maximum CH4 conversionpotential. The oxygen concentration was then set at afixed value between 0.1 and 2.0 gO2 m

−3 (with a resol-ution of 0.1 g O2 m

−3), reflecting ideal oxygen control ata constant set point. Besides, a series of simulationswere carried out to assess the extent of CH4 strippingand its potential mitigation by adding the aeration atdifferent depths in the reactor (SI, Section S6).

3. Results and discussion

3.1. Maximum nitrogen and CH4 removalefficiencies

The maximum nitrogen and CH4 removal efficienciesachieved as a function of the influent ammonium concen-tration, as well as the corresponding dissolved oxygen(DO) concentration ranges, are summarized in Figure 1

Table 1. Overview of microbial competition for substrates (S) anddependencies through formed products (P) during simultaneousnitrogen and CH4 removal in aerobic reactors.

NH+4 NO−

2 NO−3 O2 N2 CH4

Aerobic populationsAOB S P SNOB S P SMOB S SAerobic heterotrophs S

Anaerobic populationsAnammox bacteria S S P PdamoB S P SdamoA P S SAnaerobic heterotrophs (NO−

2 ) S PAnaerobic heterotrophs (NO−

3 ) P S

ENVIRONMENTAL TECHNOLOGY 1617

(see SI, Section S3, Figures S3.1 and S3.2 for detailed simu-lation results). Nitrogen removal efficiencies between89% and 99%were obtained for influent ammonium con-centrations up to 1000 g N m−3, when applying lowoptimal oxygen concentrations (<0.5 g O2 m

−3). Atthese oxygen concentrations, NOB were outcompetedby anammox bacteria and autotrophic nitrogen removalwas established. For higher influent ammonium concen-trations, the nitrogen removal efficiency decreased sig-nificantly, down to 52.6% for an influent ammoniumconcentration of 1500 g N m−3, because the correspond-ing ammonium loading rate exceeded the biomass-specific removal capacity (under-dimensioning of thereactor). The fluctuation in the nitrogen removal effi-ciency for influent ammonium concentrations up to1500 g N m−3 (Figure 1) may be explained through itshigh sensitivity to the DO concentration. The nitrogenremoval efficiency varied around 30.7 ± 15.7% whenchanging the DO by ± 0.1 g O2 m

−3 (SI, Figure S3.2),which was the resolution for the DO concentrationapplied in this simulation study. The sensitivity of thenitrogen removal efficiency to the DO concentrationwas observed also in an aerobic anammox-based granu-lar sludge reactor without CH4 removal [16]. Note that abroader range of optimal oxygen concentrations corre-sponding to maximum nitrogen removal was obtainedin the presence of an organic substrate in the influentdue to the conversion of nitrate to nitrite by heterotrophiccommunities [30].

Almost complete (≥99%) CH4 removal was foundfor influent ammonium concentrations of 200 g N m−3

or higher (Figure 1). The corresponding optimal DOconcentration interval giving high CH4 removal effi-ciencies became broader with increasing influentammonium concentrations (Figure 1). Interestingly, this

DO concentration interval always included the DO formaximum nitrogen removal. Thus, maximum CH4

removal can be achieved under the same (DO) con-ditions that lead to maximum nitrogen removal. Onlyin the case of a low influent ammonium concentration(NH+

4 = 100 g N m−3, keeping a constant CH4 concen-tration), the maximum CH4 removal was only 71% andrequired a slightly higher DO concentration (DO =0.2 g O2 m

−3) than the one corresponding to maximumnitrogen removal (DO = 0.1 g O2 m

−3). The maximumCH4 removal efficiencies found in the present study(≥99%), without considering CH4 stripping, are higherthan those obtained in the simulation study of Chenet al. [19] (93%), where a membrane biofilm bioreactorwas used and CH4 stripping was avoided by supplyingoxygen through the membrane, despite the higher CH4

surface loading rate applied in this study (0.25 g CODm−2 d−1 in this study and 0.05–0.1 g COD m−2 d−1 inChen et al. [19]). The difference in CH4 removal efficiencyobtained in both studies could be related to the MOBactivity, which was not included in the model of Chenet al. [19].

It can be noted that the granule size will affect theconversions, as demonstrated before for partial nitrita-tion–anammox reactors [21,27]. For larger granules, therange of bulk oxygen concentration corresponding toautotrophic nitrogen removal (out-competition of NOBby anammox bacteria) is broader [30], while themaximum nitrogen removal capacity (N2 production)decreases and the optimal bulk oxygen concentrationrequired to achieve this maximum nitrogen removalincreases [30]. Winkler et al. [18] studied the influenceof granule size on the combined conversion ofmethane (by damoB) and nitrogen (by anammox) innon-aerated granular sludge reactors and also foundlower removal efficiencies for larger granules. Likewise,in aerated granular sludge reactors, a lower methaneand nitrogen removal efficiency for larger granules dueto larger diffusion limitations is expected.

3.2. Contribution of functional groups to CH4 andnitrogen removal

Considering the cases with maximum nitrogen removal(Figure 1), MOB were the main contributors to CH4

removal in most cases. Over 58% of CH4 was removedby MOB, except for influent ammonium concentrationsof 100 g N m−3, with no MOB contribution on the CH4

removal; and 400 g N m−3, with only 18% CH4 removedby MOB (Figure 2). DamoB only contributed to CH4

removal for an influent ammonium concentration of400 g N m−3. For influent ammonium concentrationshigher than 800 g N m−3, the nitrogen loading exceeded

Figure 1. Maximum nitrogen and CH4 removal efficiencies andcorresponding DO concentration (range) as a function of influentammonium concentration. Simulation results obtained for anideal scenario, without CH4 stripping.

1618 C. M. CASTRO-BARROS ET AL.

the biomass removal capacity, leading to high nitriteaccumulation (SI, Figure S3.1), inhibiting damoA andleaving MOB as the only contributors for CH4 removal.Between 400 and 800 g N m−3, nitrite did not accumu-late at the (low) DO concentrations (≤0.2 g O2 m

−3) cor-responding to maximum nitrogen removal efficiency,resulting in a relatively important participation ofdamoA in CH4 removal (28–41% for influent ammoniumconcentrations in the range 600–800 g N m−3).

For an influent ammonium concentration of100 g N m−3, damoA were the only microorganismsremoving CH4 (9.1% removal efficiency). DamoB andMOB did not survive in this case due to the limiting sub-strate availability (nitrite for damoB and oxygen andammonium for MOB). DamoB only participated in theCH4 removal for the maximum nitrogen removal scen-arios for an influent ammonium concentration of 400 gNH+

4 − Nm−3 (Figure 2). In this case, AOB producedmore nitrite than required by anammox bacteria toconvert the remaining ammonium, resulting in a slightaccumulation of nitrite in the system (9.6 g N m−3, SI,Figure S3.1C), hence creating a competitive edge fordamoB (note that damoB have a lower affinity fornitrite than anammox bacteria, Table S1. 3). This con-clusion is supported by a simplified calculation basedon the ammonium consumed by AOB (SI, Table S5).

As for nitrogen removal, most of the nitrogen gas wasproduced by anammox bacteria (SI, Figure S4.1). The het-erotrophs, growing on decay products, only had a smallshare in the direct nitrogen formation, with a maximumproduction of 3.4% of the total nitrogen formed forthe scenario with influent ammonium of 400 g N m−3.Note that heterotrophic bacteria could also contributeto nitrogen formation indirectly by reducing nitrate tonitrite, which could be then further converted to

nitrogen gas by anammox bacteria [30]. Besides, thisscenario was the only one where damoB had a relativelysignificant contribution (11%) to the nitrogen pro-duction, still being eight times lower than the contri-bution by anammox bacteria (86% of the nitrogenproduced).

The damo process not only involves CH4 removal, butalso contributes to nitrogen removal, either directlythrough the conversion of nitrite to nitrogen gas bydamoB, or indirectly through the conversion by damoAof nitrate to nitrite, which could further be taken up byanammox, heterotrophic bacteria, or damoB and con-verted to nitrogen gas. The specific contribution of thedamo process to the nitrogen removal in the systemwas studied for those optimal scenarios (maximum nitro-gen and CH4 removal according to Figure 1) in whichdamoA and/or damoB were present. The simulationsfor these scenarios (influent ammonium concentrationsfrom 100 to 800 g N m−3, see Figure 3) were repeatedwithout including damoA and/or damoB in the model(Figure 3). For influent ammonium concentrations of100, 200, and 500 g N m−3, the presence of damoA didnot influence the nitrogen removal. For these scenarios,the total conversion of nitrate to nitrite (by heterotrophsand damoA) was the same whether or not the activity ofdamoA was included in the model (Figure S4.2), and thusthe global nitrogen removal was the same – the nitriteobtained by either damoA or heterotrophs was takenup by anammox bacteria, yielding nitrogen gas. At300 g NH+

4 − Nm−3, the presence of damoA led to aslightly higher nitrogen removal efficiency. The activityof damoA at 400 g NH+

4 − Nm−3 was very low (1.9%CH4 removed by damoA, Figure 2), not influencing sig-nificantly the nitrogen removal. For this specific case

Figure 2. CH4 removed by each methanotrophic functionalgroup (influent CH4 concentration = 100 g COD m−3) in termsof the influent ammonium concentration, for the scenarioswith maximum nitrogen and CH4 removal efficiency.

Figure 3. Total nitrogen removal efficiency at different influentammonium concentrations obtained when considering bothdamoB and damoA (□), only damoA (◊), only damoB (Δ), andno damo processes (○) in the model. Bulk oxygen concentration0.1 g O2 m

−3 for 100–300 g NH+4 − Nm−3 and 0.2 g O2 m

−3 for400–800 g NH+

4 − Nm−3, corresponding to optimal scenarios(maximum nitrogen and CH4 removal according to Figure 1).

ENVIRONMENTAL TECHNOLOGY 1619

(400 g NH+4 − Nm−3), the presence of damoB decreased

the nitrogen removal in the system, from 98% (no damoprocesses considered) to 95% (damoB + damoA includedor only damoB, since the activity of damoA was negli-gible) (Figure 3). At relatively high influent ammoniumconcentrations (600–800 g NH+

4 − Nm−3), damoA werepresent and increased the nitrogen removal efficiency(Figure 3) through their synergetic interaction withanammox bacteria: nitrate produced by anammox bac-teria was converted by damoA to nitrite, which wasthen again available for anammox bacteria, leading tohigher nitrogen removal efficiencies. The positive effectof damoA on anammox conversion is comparable tothe influence of heterotrophic bacteria, which may alsoimprove the nitrogen formation by anammox throughthe conversion of nitrate to nitrite [30]. In this study,the contribution of damoA to the conversion of nitrateto nitrite is higher than that of heterotrophs, since theinfluent does not contain organic matter, so thegrowth of heterotrophs can only take place on decayproducts released.

3.3. Microbial distribution

The microbial distribution in the granular sludge systemfor the scenarios with maximum nitrogen removal effi-ciencies (Figure 1) is displayed in Figure 4 (fraction ofeach population) and Figure 5 (relative distributioninside the granules). AOB and anammox bacteria domi-nated the granules, while the fraction of methanotrophs(damoA, damoB, and/or MOB) was low (Figure 4), corre-sponding to the relatively high ammonium load com-pared to the CH4 load. NOB were not present in thesystem at the low DO concentrations required formaximum nitrogen and CH4 removal. AOB andaerobic MOB governed the aerobic outer part of thegranules (Figure 5). Anammox bacteria and damoA

shared the space inside the granule, and for the specificcase where damoB were present (influent ammoniumconcentration 400 g NH+

4 − Nm−3, Figure 5(A)), thesebacteria were located in the inner part of the anoxiczone. The relative location of anammox bacteria anddamoB is in accordance with the findings of Winkleret al. [18], for an anoxic (non-aerated) granular sludgesystem.

The presence of the individual methanotrophicgroups (damoA, damoB, and MOB) for a range of influ-ent ammonium and bulk oxygen concentrations is sum-marized in the coexistence graph (Figure 6). The regionsbetween the thick black lines correspond to the scen-arios with maximum nitrogen removal. DamoA ordamoA + damoB (regions of A and A + B in Figure 6)were favoured at relatively low ammonium concen-trations where the conversion of ammonium was high(see cases in Figure S3.1 at DO >0.1 g O2 m

−3). Inthese cases, where ammonium was not in excess, thelow affinity of MOB for ammonium (nitrogen sourcefor their growth) impeded their growth and damoAand damoB took advantage over MOB, also giventheir higher affinity for methane (KCH4_damo= 0.19 andKCH4_MOB= 0.26 g COD m−3, Table S1. 3). However,when sufficient ammonium was available as nitrogensource for biomass growth, MOB were competitivebecause of their high growth rate (μmax_MOB = 3.5 d−1,μmax_damoA = 0.036 d−1, and μmax_damoB = 0.050 d−1,Table S1.3). DamoA coexisted with MOB in all cases,except when there was a high accumulation of nitrite,causing their inhibition (KiNO2_damoA = 40 gNH+

4 − Nm−3). DamoA were inhibited and only MOBsurvived at high ammonium concentrations and DO≠0.2 g O2 m

−3 (no accumulation of nitrite at 0.2 g O2 m−3,

see Figure S3.1).Apart from the methanotrophic competition, damoB

compete also with anammox bacteria for nitrite. In thisrespect, damoB only survived when the nitrite concen-tration in the system was not limiting. DamoB are theleast favoured anaerobic methane oxidizing population,and never appeared alone, but always with damoA ordamoA +MOB (Figure 6). The presence of damoBbesides damoA and MOB typically took place whenammonium was not in excess and at relatively highDO concentrations. For an influent ammonium of400 g N m−3, damoB were also present at a relativelylow DO of 0.2 g O2 m

−3, and even were the main con-tributors to CH4 removal (80% CH4 removed by damoB,Figure 2). MOB still survived at a low DO concentrationof 0.1 g O2 m

−3 due to their higher or equal affinity foroxygen than the other aerobic populations (KO2_MOB=0.2; KO2_AOB= 0.3; and KO2_HA= 0.2 g O2 m

−3) and theirhigh growth rate.

Figure 4. Fraction of biomass and particulate inerts as a functionof the influent ammonium concentrations, for the maximumnitrogen removal scenarios (see Figure 1).

1620 C. M. CASTRO-BARROS ET AL.

3.4. Sensitivity analysis for methanotrophic half-saturation constants

A sensitivity analysis was performed to assess the influ-ence of the methane half-saturation constant fordamoA and damoB (KCH4_damo), the ammonium half-sat-uration constant for MOB (KNH4_MOB), and the nitritehalf-saturation constant for damoB (KNO2_damoB) on thepresence of the methanotrophs and on the reactor per-formance in terms of nitrogen and CH4 removal efficien-cies (Table S2.2). The analysis was performed for thescenario with maximum nitrogen and CH4 removal inwhich the three methanotrophic groups (damoA,damoB, and MOB) were present, that is, for influentammonium = 400 g N m−3 and DO = 0.2 g O2 m

−3 (seeFigures 1 and 6).

Decreasing the methane affinity of damo species (i.e.increasing KCH4_damo) led to their out-competition byMOB, which was complete for KCH4_damo ≥ 1 g COD m−3

(Figure 7(A)). The KCH4_damo for damoB and damoA inthe present study is 30 times lower (higher affinity)than that in Chen et al. [19] (KCH4_damo = 0.19 vs.5.888 g COD m−3, respectively). While Chen et al. [19]did not include MOB in the model to describe their

system, our results indicate that MOB could haveplayed an important role.

Lower values of the ammonium half-saturation con-stant for MOB (KNH4_MOB) than the one used in thisstudy (2 g COD m−3) led to the dominance of MOBamong methanotrophs (Figure 7(B)). Even thoughammonium only serves as a nitrogen source for thebiomass growth of MOB, the change in KNH4_MOB playsan important role, especially at low ammonium concen-trations, which leads to their growth limitation.

Decreasing the nitrite half-saturation constant fordamoB (KNO2_damoB) (hence increasing the competitive-ness for nitrite) resulted in increasing damoB in thesystem, but barely affected the anammox populationabundance (Figure 8(A)), because of the high ammoniumload compared to the CH4 load (41.7 g N m−3 h−1 and10.4 g COD m−3 h−1, respectively) and the relativelyhigh yield of anammox bacteria (0.170 g COD [g N]−1)compared to damoB (0.0835 g COD [g COD]−1). The pres-ence of MOB and damoA increased with increasingKNO2_damoB, while the overall CH4 removal efficiencywas not affected (Figure 8(B)). In this study, anammoxbacteria are assumed to have a higher nitrite affinity

Figure 5. Microbial distribution inside the granules for various influent ammonium concentrations. Axis x represents the radius of thegranule, from 0 (the innermost part of the granule) to 0.75 mm (the outer part of the granule).

ENVIRONMENTAL TECHNOLOGY 1621

than damoB (KNO2_damoB = 0.6; KNO2_AN = 0.005 g N m−3).Chen et al. [19] assumed the opposite, favouring damoBwhen competing with anammox bacteria for nitrite(KNO2_damoB = 0.01; KNO2_AN = 0.05 g N m−3). This studydemonstrates that the presence of damoB is influencedby KNO2_damoB (Figure 8(B)). Higher values of KNO2_damoB

as in the study of Chen et al. [19] imply a more advan-tageous situation for anammox bacteria relative todamoB, which could affect the CH4 removal.

3.5. Effect of reactor operation and design on CH4

stripping

The effect of aeration intensity (kLa) on the fate of CH4 issummarized in Figure S6. In case of surface aeration (i.e.aeration depth = 0), high nitrogen removal was achieved

at the expense of high CH4 stripping (maximum nitrogenremoval achieved = 98% at kLaO2 = 500 d−1 with 56.5% ofCH4 stripped, Figure S6.A).

CH4 stripping could be reduced through the partialrecirculation of the off-gas, as already applied in certainfull-scale aerobic anammox-based granular sludge reac-tors as a means of aeration control [31]. Recirculationof part of the off-gas would increase the CH4 partialpressure and thus lead to less CH4 stripping. However,CH4 in the non-recirculated part of the off-gas wouldstill end up in the atmosphere. A second option to mini-mize CH4 stripping, possibly combined with recirculation,would be to use subsurface aeration rather than surfaceaeration [28] and to locate the (CH4 containing) influentsupply line inside the reactor. Installing the aerationsupply deeper in the tank results in an increased gas

Figure 6. Coexistence of methanotrophic populations as a function of influent ammonium and DO concentrations for regions withsignificant CH4 removal efficiency (10–50%, see Figure S3.2). The section between the thick black lines corresponds to optimal nitrogenremoval (see Figure S3.2). The total methanotrophic population made up 1–10% of the total particulate fraction, except for regionsmarked with *, where the methanotrophs represented less than 1%. Methanotrophic populations were included when their presencewas≥ 0.1% of the methanotrophic community. M = MOB; A = damoA; B = damoB.

Figure 7. Influence of (A) the methane half-saturation constant for damo organisms (KCH4_damo) and (B) the ammonium half-saturationconstant for MOB (KNH4_MOB) on the total fraction of methanotrophic community among all particulate components, relative fraction ofmethanotrophic microorganisms within the total methanotrophs, and nitrogen and CH4 removal efficiencies. Influent NH+

4 =400 g N m−3; DO = 0.2 g O2 m

−3. The box indicates the case with standard parameter values applied in this study.

1622 C. M. CASTRO-BARROS ET AL.

solubility (CH4 and O2), implying less CH4 stripping, but iscounteracted by a stronger saturation (in CH4) ordepletion (in O2) of the rising gas bubbles (higher gasretention time). Figure 9 summarizes the effect of thedepth of aeration supply on the amount of CH4 stripped,consumed, and remaining unconverted in the effluent,for the cases with the highest nitrogen removal effi-ciency (details on the underlying simulations are givenin Figure S6). CH4 stripping was reduced from 56% to23% (i.e. by 60%) while keeping the same nitrogenremoval efficiency (98%) when increasing the aerationsupply depth from 0 (i.e. in the case of surface aeration)to 8 m (Figure 9). A further increase (to 12 m) only slightly

reduced the CH4 stripping and increased the CH4 conver-sion (from 77% to 80% CH4 converted for an aerationdepth of 8 and 12 m, respectively), which may not bewarranted by the higher energy costs required tosupply oxygen at this depth and by the slightly increasedDO concentration needed to achieve the correspondingoptimal nitrogen removal efficiency.

4. Conclusions

The possible integration of CH4 removal in an aerobicanammox-based granular sludge reactor was assessedthrough a simulation study.

. A model was set up including not only damo bacteriaand archaea, but also aerobic MOB as a potentiallyimportant microbial group for CH4 removal, besidesthe bacterial populations playing a role in nitrogenremoval.

. Simultaneous nitrogen and CH4 removal wasdemonstrated feasible for low DO concentrations(DO <0.5 g O2 m

−3), achieving up to 99% removal effi-ciencies for both substrates in an ideal case withoutCH4 stripping.

. While nitrogen removal efficiency was found to bevery sensitive to the DO concentration, maximumCH4 removal was obtained in a broader oxygen con-centration interval. This DO concentration intervalalways included the DO for maximum nitrogenremoval; thus, maximum CH4 removal was achievedunder the same DO conditions as that for maximumnitrogen removal.

. The presence of damo archaea improved the nitrogenremoval efficiency by converting the nitrate formedby anammox bacteria.

Figure 8. Effect of the nitrite half-saturation constant for damoB (KNO2_damoB) on the relative abundance of populations competing withdamoB for substrate: (A) competition with anammox bacteria and anaerobic heterotrophs (HNO2) for nitrate and (B) competition withdamoA and MOB for CH4. Influent NH+

4 = 400 g N m−3; DO = 0.2 g O2 m−3. The box indicates the case with standard parameter values

applied in this study.

Figure 9. Influence of the depth of aeration supply (liquidcolumn height) on dissolved CH4 in the effluent, CH4 consumed,and CH4 stripped, as well as corresponding CH4 and nitrogenremoval. Results obtained for maximum nitrogen removal scen-arios for each depth (kLa O2 = 500, 450, 400, and 375 for aerationdepth = 0, 4, 8, and 12 m, respectively), influent NH+

4 =700 g N m−3, Qinf = 1000 m3 d−1.

ENVIRONMENTAL TECHNOLOGY 1623

. High CH4 removal efficiencies were maintained atdifferent influent ammonium and DO concentrationsand when varying kinetic parameters. While a shiftcould be noted in the methane-oxidizing population,aerobic methane oxidizing bacteria were the mainresponsible microorganisms for CH4 removal in allcases.

. CH4 stripping during aeration could be limited byincreasing the aeration supply depth within thereactor (23% CH4 stripped at 8 m depth), whilekeeping the nitrogen removal efficiency high.

Acknowledgements

This publication reflects only the authors’ views and the Euro-pean Union is not liable for any use that may be made of theinformation contained therein. Luis Corbalá-Robles is acknowl-edged for his help in simulating biofilm systems in Aquasim.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

The research leading to these results has received funding fromthe People Program (Marie Curie Actions) of the EuropeanUnion’s Seventh Framework Programme FP7/2007–2013,through the REA agreement 289193 – Project SANITAS; GhentUniversity Special Research Fund (BOF) 2015 Finalizing doctoralscholarship [grant number 01DI4415]; and EU Marie Curie Intra-European fellowship: [grant number PIEF-GA-2012-329328 –anMOgran].

ORCID

Eveline I. P. Volcke http://orcid.org/0000-0002-7664-7033

References

[1] Kampschreur MJ, Temmink H, Kleerebezem R, et al.Nitrous oxide emission during wastewater treatment.Water Res. 2009;43(17):4093–4103.

[2] Daelman MRJ, van Voorthuizen EM, van Dongen UGJM,et al. Methane emission during municipal wastewatertreatment. Water Res. 2012;46(11):3657–3670.

[3] IPCC. Climate change 2013: The physical science basis.Working group I contribution to the fifth assessmentreport of the intergovernmental panel on climate change.Cambridge (UK): Cambridge University Press; 2013.

[4] Raghoebarsing AA, Pol A, van de Pas-Schoonen KT, et al. Amicrobial consortium couples anaerobic methane oxi-dation to denitrification. Nature. 2006;440(7086):918–921.

[5] Ettwig KF, Butler MK, Le Paslier D, et al. Nitrite-drivenanaerobic methane oxidation by oxygenic bacteria.Nature. 2010;464(7288):543–548.

[6] Luesken FA, Sanchez J, van Alen TA, et al. Simultaneousnitrite-dependent anaerobic methane and ammonium

oxidation processes. Appl Environ Microb. 2011;77(19):6802–6807.

[7] Haroon MF, Hu SH, Shi Y, et al. Anaerobic oxidation ofmethane coupled to nitrate reduction in a novel archaeallineage. Nature. 2013;500(7464):567–570.

[8] Strous M, Heijnen JJ, Kuenen JG, et al. The sequencingbatch reactor as a powerful tool for the study of slowlygrowing anaerobic ammonium-oxidizing microorgan-isms. Appl Microbiol Biot. 1998;50(5):589–596.

[9] Fux C, Boehler M, Huber P, et al. Biological treatment ofammonium-rich wastewater by partial nitritation and sub-sequent anaerobic ammonium oxidation (anammox) in apilot plant. J Biotechnol. 2002;99(3):295–306.

[10] van der Star WRL, Abma WR, Blommers D, et al. Startup ofreactors for anoxic ammonium oxidation: experiencesfrom the first full-scale anammox reactor in rotterdam.Water Res. 2007;41(18):4149–4163.

[11] Lackner S, Gilbert EM, Vlaeminck SE, et al. Full-scale partialnitritation/anammox experiences – an application survey.Water Res. 2014;55:292–303.

[12] Morales N, Val del Río A, Vázquez-Padín JR, et al.Integration of the Anammox process to the rejectionwater and main stream lines of WWTPs. Chemosphere.2015;140:99–105.

[13] Cao Y, van Loosdrecht MCM, Daigger GT. Mainstreampartial nitritation–anammox in municipal wastewatertreatment: status, bottlenecks, and further studies. ApplMicrobiol Biotechnol. 2017;101:1365–1383.

[14] Hu SH, Zeng RJ, Haroon MF, et al. A laboratory investi-gation of interactions between denitrifying anaerobicmethane oxidation (DAMO) and anammox processes inanoxic environments. Sci Rep. 2015;5:320.

[15] Shi Y, Hu SH, Lou JQ, et al. Nitrogen removal from waste-water by coupling anammox and methane-dependentdenitrification in a membrane biofilm reactor. EnvironSci Technol. 2013;47(20):11577–11583.

[16] Xie GJ, Cai C, Hu S, et al. Complete nitrogen removal fromsynthetic anaerobic sludge digestion liquor through inte-grating anammox and denitrifying anaerobic methaneoxidation in a membrane biofilm reactor. Environ SciTechnol. 2016;51(2):819–827.

[17] Chen XM, Guo JH, Shi Y, et al. Modeling of simultaneousanaerobic methane and ammonium oxidation in a mem-brane biofilm reactor. Environ Sci Technol. 2014;48(16):9540–9547.

[18] Winkler MK, Ettwig KF, Vannecke TP, et al. Modellingsimultaneous anaerobic methane and ammoniumremoval in a granular sludge reactor. Water Res. 2015;73:323–331.

[19] Chen XM, Guo JH, Xie GJ, et al. A new approach to simul-taneous ammonium and dissolved methane removalfrom anaerobic digestion liquor: a model-based investi-gation of feasibility. Water Res. 2015;85:295–303.

[20] Lotti T, Kleerebezem R, Hu Z, et al. Pilot-scale evaluation ofanammox-based mainstream nitrogen removal frommunicipal wastewater. Environ Technol. 2015;36(9):1167–1177.

[21] Volcke EIP, Picioreanu C, De Baets B, et al. Effect of granulesize on autotrophic nitrogen removal in a granular sludgereactor. Environ Technol. 2010;31(11):1271–1280.

[22] Henze M, Gujer W, Mino T, et al. Activated sludge modelsASM1, ASM2, ASM2d and ASM3. London: IWA; 2000.

1624 C. M. CASTRO-BARROS ET AL.

[23] Arcangeli JP, Arvin E. Modelling the growth of a methano-trophic biofilm: estimation of parameters and variability.Biodegradation. 1999;10(3):177–191.

[24] Ettwig KF, Shima S, van de Pas-Schoonen KT, et al.Denitrifying bacteria anaerobically oxidize methane inthe absence of archaea. Environ Microbiol. 2008;10(11):3164–3173.

[25] Hao XD, Heijnen JJ, van Loosdrecht MCM. Sensitivityanalysis of a biofilm model describing a one-stage com-pletely autotrophic nitrogen removal (CANON) process.Biotechnol Bioeng. 2002;77(3):266–277.

[26] Reichert P. Aquasim – a tool for simulation and data-analy-sis of aquatic systems. Water Sci Technol. 1994;30(2):21–30.

[27] Volcke EIP, Picioreanu C, De Baets B, et al. Thegranule size distribution in an anammox-based granu-lar sludge reactor affects the conversion - implications

for modeling. Biotechnol Bioeng. 2012;109(7):1629–1636.

[28] Daelman MRJ, Van Eynde T, van Loosdrecht MCM, et al.Effect of process design and operating parameters onaerobic methane oxidation in municipal WWTPs. WaterRes. 2014;66:308–319.

[29] Henze M, van Loosdrecht MCM, Ekama GA, et al. Biologicalwastewater treatment: principles, modelling and design.London: IWA; 2008.

[30] Mozumder MSI, Picioreanu C, van Loosdrecht MCM, et al.Effect of heterotrophic growth on autotrophic nitrogenremoval in a granular sludge reactor. Environ Technol.2014;35(8):1027–1037.

[31] Castro-Barros CM, Daelman MRJ, Mampaey KE, et al. Effectof aeration regime on N2O emission from partial nitrita-tion-anammox in a full-scale granular sludge reactor.Water Res. 2015;68:793–803.

ENVIRONMENTAL TECHNOLOGY 1625