Modeling bioaugmentation with nitrifiers in membrane bioreactors

Bioaugmentation with nitrifiers is a cost-effective strategy to obtain high nitrification efficiencies at relatively low solids retention times (SRTs) (Ma et al. 2009; Bartolì et al. 2011; Szoke et al. 2011; Wett et al. 2011). The strategy consists of enriching the mixed liquor from the main-stream reactor (MSR) of an activated sludge system with nitrifying biomass collected from a side-stream reactor (SSR) where the environmental conditions are favorable for nitrifiers (Leu & Stenstrom 2010; Peng et al. 2012). Several factors, not included in the state-of-the-art activated sludge models (ASMs), can influence bioaugmentation efficiency. These include, among others, predation, large temperature differences and different nitrifying biomass in the seeding and seeded reactors. Therefore, bioaugmentation efficiency has seldom been accurately predicted using the conventional International Water Association ASMs (Salem et al. 2003; Munz et al. 2012). Bioaugmentation is difficult to predict and control (Van Limbergen et al. 1998) and its success depends on the effective establishment and metabolic adaptation of the seeded biomass in the main-stream reactor (Satoh et al. 2003). A reduction of nitrifier activity was observed (Parker & Wanner 2007) when biomass did not adapt to the new environmental conditions. Ammonia oxidizing bacteria (AOB) and nitrite oxidizing bacteria (NOB), selected at low temperatures, have been used by Cui et al. (2014) for the enrichment of nitrifying biomass treating municipal wastewater at 10 °C. In that case, bioaugmentation allowed an increase in the ammonia removal efficiency (RE_N) of 43%, with an overall ammonia removal rate of 85% in the main-stream MSR; however, the temperature difference (ΔT) between the SSR and MSR was not investigated.

In full-scale wastewater treatment plants (WWTPs), the ΔT between an SSR fed with anaerobic digester supernatant (characterized by an temperature higher than 30 °C) and MSR could be higher than 20 °C during the winter period (Mannucci et al. 2014) and could affect bioaugmentation efficiency due to the high sensitivity of the nitrifying biomass to temperature changes (Coskuner & Curtis 2002).

Several studies conducted in bench scale using continuous-flow stirred-tank (Lee et al. 2011) and sequencing batch reactor (SBR) (Head & Oleszkiewicz 2004; Hwang & Oleszkiewicz 2007), in pilot scale with conventional activated sludge systems (Plaza et al. 2001) and in full-scale WWTPs showed that the temperature difference between seeding and seeded reactors influences bioaugmentation efficiency.

The Arrhenius equation is commonly used to provide an estimation of the temperature effect on nitrification. The equation showed some important limitations when applied to the prediction of the effects of the AOB bioaugmentation process (Wett et al. 2011). The exact role of temperature changes still remains to be evaluated in detail, and under more controlled conditions.

Compared to conventional activated sludge systems, a membrane bioreactor (MBR) offers more accurate control of the solids retention time (SRT) and total undifferentiated bacteria retention. This can facilitate the evaluation of ASMs' suitability to predict bioaugmentation and to estimate the effect of temperature change (Zhang et al. 2009). The use of MBRs both for the seeding and seeded reactors has not been investigated so far. The community survivability and functional stability of the seeded biomass are common problems in bioaugmentation applications and the investigation of the fate of inoculated biomass remains an open issue (Yu et al. 2012).

The objective of this work was to model the bioaugmentation process in pilot scale using two MBRs operating in different conditions in terms of the ammonia loading rate (ALR), SRT and hydraulic retention time. The existing ASMs will be evaluated based on simulation of the bioaugmentation process. The fate of the seeded biomass was analyzed by comparing shifts in the nitrifying population in mixed liquor suspended solids (MLSS) before and after bioaugmentation started using fluorescent in situ hybridization (FISH) analysis.

The experimental setup consisted of two MBRs as shown in Figure 1. The first pilot-scale MBR (MBR1 – the seeding reactor) consisted of a pre-denitrification tank (30 L), a nitrification tank (30 L) and a filtration tank (10 L) equipped with three flat membranes (DF-10 Kubota, Osaka, Japan). After the startup phase, MBR1 was fed with synthetic high-strength ammonia influent simulating anaerobic digester supernatant (650 ± 18 mg N-NH₄⁺ L⁻¹; 250 ± 50 mg COD_tot L⁻¹) and operated for more than 600 days with an SRT of 20 d and an ALR of 560 mg N-NH₄⁺ L⁻¹ d⁻¹. The second pilot-scale MBR (MBR2 – the seeded reactor) consisted of a pre-denitrification tank (130 L), a nitrification tank (280 L) and a filtration tank (30 L) equipped with a hollow fiber filtration module (ZW10 GE-Zenon Environmental, Trevose, Pennsylvania, USA). MBR2 was fed with real domestic wastewater (32.3 ± 5 mg N-NH₄⁺ L⁻¹; 286 ± 75 mg COD_tot L⁻¹) continuously collected from the sewer at the Cuoiodepur WWTP (San Romano – San Miniato, Pisa, Italy). The ALR in MBR2 was close to 47.1 ± 7.8 mg N-NH₄⁺ L⁻¹ d⁻¹, while the dissolved oxygen (DO) concentration and pH were the same in both MBRs: pH = 7.5 ± 0.5, DO = 4 ± 0.5 mg L⁻¹. In MBR1, temperature was controlled and maintained at 20 ± 0.5 °C, while in MBR2, temperature varied from 24 to 6 °C during the experiment, depending on the ambient temperature. MBR2 was operated under steady-state conditions for more than 400 days with an SRT of 2.5 d and without any external seeding before bioaugmentation started, with a constant flow of 2.5 L d⁻¹ of nitrifying sludge from the filtration tank of MBR1 (Figure 1). The bioaugmentation phase started after 440 days and lasted more than 150 d.

A series of kinetic batch tests was performed in order to estimate the kinetic parameters of the nitrifying biomass in both MBRs. Each test was performed with 1 L of MLSS collected from the nitrification tanks of both MBRs and maintained in mixed and aerated conditions until reaching an endogenous respiration phase. A fixed amount of NH₄Cl was dosed in order to obtain an initial ammonia concentration in the range 18–22 mg N-NH₄⁺ L⁻¹. Such concentrations were neither limiting nor inhibitory, based on previous observations and literature data (Munz et al. 2011). The pH was controlled at 8 ± 0.1 and the DO concentration was maintained higher than 4 mg L⁻¹ through fine bubble aeration. Temperature was maintained at 20 ± 0.2 °C. Samples were collected every 10 minutes and the N-NH₄⁺ concentration was determined through colorimetric analysis using cuvette kits (HACH-LANGE, Berlin, Germany) and a spectrophotometer (XION 500, HACH-LANGE, Berlin, Germany).

To evaluate the effect of continuous nitrifying biomass seeding in MBR2, kinetic tests were replicated both in the bioaugmented and non-bioaugmented periods. Each test was done in triplicate.

Molecular analysis was performed by FISH in order to monitor the microbial communities before (day 430) and after (day 480) the bioaugmentation started in both MBRs. In situ hybridization was performed according to the conventional protocol (Manz et al. 1992) utilizing the eubacterial universal probe EUB_338214I (Amann et al. 1990), and probe b-AO233 (Stephen et al. 1998) specific for betaproteobacterial AOB, Nsm_156 (Mobarry et al. 1996) specific for the genus Nitrosomonas, NIT3 (Wagner et al. 1996) specific for the genus Nitrobacter, NTSPA714 (Loy et al. 2002) specific for the phylum Nitrospira and probe Nsv443 (Mobarry et al. 1996) specific for the genus Nitrosospira.

An ASM with a two-step nitrification–denitrification, termed ASMN (Hiatt & Grady 2008), was used to describe the processes (autotrophic and heterotrophic biomass) in both pilot plants. Batch tests were performed after steady-state conditions were achieved. The active AOB biomass concentration was evaluated by performing the ammonia mass balance between the influent and effluent, assuming the yield coefficient, Y_AOB, and decay coefficient, b_AOB, of 0.18 mg chemical oxygen demand (COD) (mg N-NH₄⁺)⁻¹ (Kaelin et al. 2009) and 0.17 d⁻¹ (Jubany et al. 2008), respectively. The ASMN model that separately represents AOB and NOB populations was used to calibrate the maximum specific growth rate for AOB (μ_max,AOB) and the half-saturation constant for ammonia (K_NH) based on the results of batch tests.

WEST-for-optimization (DHI, Hørsholm, Denmark) was used as software for modeling. The parameter calibration was carried out by using the simplex method to minimize the mean error between observed and predicted ammonia concentration. During calibration, μ_max,AOB and K_NH were simultaneously varied in the range 0.5–1.2 d⁻¹ and 0.2–1.2 mg L⁻¹, respectively, with a step size of 0.01. The calibrated μ_max,AOB and K_NH were used to validate the ASMN based on the effluent ammonia observed in MBR1 during the whole experiment and effluent ammonia observed in MBR2 during the non-bioaugmented period.

The RE_N obtained in MBR2 during the continuous AOB seeding period (ON) was effectively increased when compared with RE_N in the non-bioaugmented (OFF) periods. However, the bioaugmentation efficiency, expressed as the increase of RE_N without and with continuous nitrifiers seeding, was not constant during the experiment, as it depended on the temperature difference between the seeding and seeded reactor. In order to explain the influence of temperature on the bioaugmentation performance, the whole MBR2 temperature range was divided into three intervals (12–15, 15–17 and 17–20 °C). The increase of RE_N due to continuous bioaugmentation was calculated for each temperature interval as the difference between RE_N with and without continuous seeding.

Temperatures higher than 17 °C did not negatively affect nitrification in MBR2, in fact a high RE_N (RE_N > 90%) was observed in periods without continuous bioaugmentation at these temperatures. An average increase in the efficiency of 4 ± 0.4% was obtained due to bioaugmentation. Within the temperature range 12–15 °C, the RE_N without bioaugmentation dropped to 75%. A low temperature in MBR2 and high ΔT between MBR1 and MBR2 lowered the effect of bioaugmentation and RE_N increased only by 1 ± 0.2% due to bioaugmentation. The highest impact (19 ± 3%) was obtained with the MBR2 temperature in the range 15–17 °C.

The AOB concentration inside MBRs during the kinetic tests was adopted as the ASMN input for the initial AOB concentration inside batch reactors. The parameters μ_max,AOB and K_NH were calibrated based on the ammonia bulk liquid concentrations obtained in the conventional batch kinetic tests and assuming the AOB decay coefficient (b_H,AOB, of 0.17 d⁻¹).

An example of the observed and predicted ammonia concentrations for the MBR2 nitrifying biomass is shown in Figure 2.

The average values of μ_max,AOB and K_NH calibrated based on the kinetic batch tests are reported in Table 1.

As reported in Table 1, the main kinetic parameters of AOB in MBR2 were not influenced by continuous seeding of nitrifiers with different kinetics. Therefore, a unique nitrifying biomass in MBR2 also in the presence of bioaugmentation was considered for further simulations (μ_{max, AOB} = 0.84 and K_N = 0.4 mg L⁻¹). Although the AOB kinetics in MBR2 did not change in the presence of bioaugmentation, nitrifier populations were influenced: Nitrosomonas spp. concentration (Table 2) in MBR2 doubled after seeding started, in accordance with the results of previous works (Yu et al. 2012).

The ASMN model was validated based on the MBR1 effluent ammonia concentrations (Figure 3) and MBR2 effluent ammonia concentrations (Figure 4) observed during the non-bioaugmented period.

In Figure 4, predicted effluent ammonia concentrations are shown for the periods with and without bioaugmentation. During the whole non-bioaugmented period (until day 440), the ASMN described accurately the effluent ammonia concentrations while changing some operational conditions, such as the SRT and ALR. For example, from day 365 to day 382, the SRT dropped during operation from 2.5 to 2.0 d.

Due to dilution with rain water, the influent ammonia concentration decreased by 40% from day 530 to day 539, while from day 556 to day 566, ALR was maintained at 22.0 ± 2.7 mg N-NH₄⁺ L⁻¹ d⁻¹, half of the average value maintained during the experimentation.

The above sudden changes in the influent ammonia loads did not affect the effluent quality, and ammonia concentrations were always lower than 1 mg L⁻¹. The ASMN model output (Figure 4) with (black line) and without (grey line) nitrifier seeding underlines a higher stability of the system due to the presence of bioaugmentation during the last period.

Since all operational conditions (except for the MBR2 temperature) were kept constant during the bioaugmentation period, at least from day 440 to day 530, ΔT between MBR1 and MBR2 and its effect on nitrifying seeded biomass were identified as the main reason causing the high discrepancy between experimental data and model predictions from day 505 to day 525.

Three ΔT intervals were defined to evaluate the ASMN capability to describe the experimental RE_N in MBR2 during bioaugmentation. As long as ΔT was lower than 10 °C, the ASMN model was able to describe MBR2 effluent quality in the presence of bioaugmentation, and it depicted clearly the positive relationship between the improvement of RE_N due to bioaugmentation with increasing ΔT.

Discrepancies of up to 29% between the predicted and the observed RE_N were observed with ΔT higher than 10 °C (Figure 5).

ASMN does not include any adaptation mechanisms related to sudden temperature changes and, as a consequence, does not describe bacterial activity reduction due to ΔT higher than 10 °C. The observed phenomena can be explained when assuming that the investigated biomass was subjected to an adaptation phase and a subsequent lag phase dependent on temperature shock. It was observed that, after exposure to physical stress such as sudden temperature decrease, a lag phase occurred; during this phase, cell growth was limited and specific proteins were produced (cold shock proteins) (Beales et al. 2004; Yamanaka & Inouye 2001).

In order to include the effect of temperature shock, the ASMN was modified. The effect of temperature shock on the seeded AOB was evaluated by considering two AOB populations with different sets of kinetic parameters. Different temperature sensitivities for AOB grown in MBR2 (AOB_MBR2) and for seeded AOB (AOB_MBR1) were considered. A simplified linear model was used to describe θ_{AOB_MBR1} variations as a function of temperature shock and to predict effluent ammonia concentrations in the bioaugmented period from day 440 to day 550. Three sets of RE_N experimental data obtained with different ΔT were used to calibrate the Arrhenius temperature correction factor of the seeded AOB (θ_{AOB_MBR1}): 6, 9 and 12.5 °C (Table 3). The temperature correction factor of AOB in MBR2 (θ_{AOB_MBR1}) was maintained constant (equal to 1.07) for all simulations.

This increase in θ_{AOB_MBR1} with increasing ΔT between the seeded and seeding reactor allowed better representation of the results (Figure 6) and an error (between observed and predicted RE_N) lower than 5% was obtained for all the tested ΔT. Even though not based on mechanistic grounds, the use of a simple linear relationship between θ_{AOB_MBR1} and ΔT is a potentially useful solution for practical purposes in conditions similar to those investigated.

The study confirmed the feasibility of modeling bioaugmentation effects using a single nitrifying biomass in both seeded and seeding reactors only with a low temperature difference (ΔT < 10 °C) between side-stream and main-stream systems. ΔT had a very significant impact on the effect of bioaugmentation in nitrification and would have to be taken into consideration in order to improve mathematical modeling of bioaugmentation. The use of a double-AOB model also improved the capability of ASM to predict bioaugmentation in the presence of high ΔT.

The biological mechanisms responsible for the decrease in bacterial growth due to sudden temperature changes should be lumped together in an increase in temperature sensitivity (θ_{AOB_MBR1}) of the seeded biomass inoculated into the main-stream reactor. As reported by Oleszkiewicz & Berquist (1988) and Guo et al. (2010), the temperature correction factor depended on the studied temperature range. In this work, the temperature in the MBR2 reactor varied in the range 7–20 °C, which was the typical temperature of municipal wastewater. Further studies are needed to evaluate the effect of the same ΔT in different temperature ranges.

The authors acknowledge the European Union for supporting this work through the Carbala Marie Curie Irses program (Carbala project 295176). This paper was presented at WWTmod2014 and the fruitful discussions are kindly acknowledged.