Modeling the pH effects on nitrogen removal in the anammox-enriched granular sludge

The anaerobic ammonium oxidation (anammox) process is a viable alternative to nitrogen removal by the conventional nitrification/denitrification. In practice, most of the anammox-based systems utilize mixed cultures of functional consortia predominated by anammox bacteria (Niu et al. 2016). These systems are subjected to inhibition by many factors, including substrates (ammonia and nitrite), organic matter (nontoxic and toxic organic matter), specific compounds (salts, heavy metals, phosphate, sulfide), and variations in the operating conditions (Carvajal-Arroyo et al. 2013).

The pH is considered as one of the most important operating parameters affecting stoichiometry and kinetics of biological processes. The recent study by Carvajal-Arroyo et al. (2013) has indicated sharp pH optima (7.3–7.5) in the specific anammox activities (SAAs) for both suspended and granular anammox enrichments, i.e., 17.32 ± 2.13 mmol N₂/(gVSS·d) and 6.78 ± 0.71 mmol N₂/(gVSS·d), respectively (VSS: volatile suspended solids). The impact of pH extremes has been attributed to the direct effects of pH or the pH-dependent un-ionized species, free ammonia (FA) and free nitrous acid (FNA), which are presumably toxic for the anammox and nitrification processes (Fernández et al. 2012; Daverey et al. 2015). Chen et al. (2010) proposed that low pH as well as FNA was responsible for anammox inhibition in acid solutions. Puyol et al. (2014b) proposed that high pH instead of FA was responsible for anammox inhibition in alkaline solutions. The inhibition has also been rationalized based on a disruption of the intracellular proton motive force over the anammoxosome membrane, thus directly affecting energy yield and cell survival (van der Star et al. 2010).

Hultman (1973) proposed the first equation describing pH effects on the maximum specific growth rate of microorganisms. Antoniou et al. (1990) proposed another equation describing pH effects on nitrification according to active ionized enzyme theory. The same equation was also suggested by Angelidaki et al. (1993) but in a different form (also known as Michaelis pH function). However, none of these functions have been used in the anammox-enriched granular sludge.

Serralta et al. (2004) extended the activated sludge model ASM2d with pH calculations. Goel et al. (2010) verified the pH predictions of the activated sludge model ASM2d. Hao et al. (2002, 2005) described several simulation studies on the behaviour of a partial nitrification–anammox biofilm system under different process conditions, including oxygen consumption, temperature and inflow variations. However, no mathematical model with pH calculation has been applied in the anammox-enriched system.

In this study, a comprehensive mechanistic model was combined with a pH calculation tool to optimize the kinetic parameters, including the maximum specific growth rates of ammonia oxidizing bacteria (AOB) and anammox bacteria, at different pH scenarios in a single system. Two different pH functions were used to determine the inhibitory effects of pH on the maximum specific growth rates of AOB and anammox bacteria in the anammox-enriched granular sludge.

The laboratory experiments were conducted in a 4 L completely mixed batch reactor with the anammox-enriched granular sludge at different initial pH values (6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5) and constant temperature T = 30 °C. The sludge was cultivated for more than 1 year on synthetic substrates under laboratory conditions in a sequencing batch reactor (SBR) to obtain a granulation effect. The average diameter of the developed granules was approximately 730 μm. The synthetic feed contained sodium nitrite, ammonium chloride and trace elements. The initial concentrations of total ammonia and total nitrite, respectively, were in the ranges of 31–39 g N/m³ and 38–50 g N/m³ to avoid the substrate limitation, and the theoretical NO₂-N/NH₄-N ratio of 1.3 for the anammox process was maintained during all the batch experiments. In this way, two most important kinetic parameters (maximum specific growth rate of AOB and of anammox bacteria) could be estimated in terms of changing pH. The initial pH was adjusted by the addition of dilute HCl (1 mol/L) or NaOH (2 mol/L). The initial dissolved oxygen (DO) concentrations were between 0.5 and 1 mg O₂/L and stable in the same level during the course of the experiments except when pH = 6. During the course of each batch experiment, the pH and DO concentrations were not controlled. More information about the laboratory experiments can be found elsewhere (Sobotka et al. 2016, in press; Yin et al. 2016).

Concentrations of the inorganic N forms (NH₄-N, NO₃-N, NO₂-N) were determined in the filtered mixed liquor every 30 min during the course of the batch experiments. The measurements were carried out using cuvette tests (Hach Lange GmbH, Germany) in a Xion 500 spectrophotometer (Dr Lange GmbH, Germany). The DO, pH and temperature were monitored continuously during the course of each batch experiment. The biomass composition for setting the initial conditions in modeling was determined using the metagenomic analysis (Sobotka et al. in press).

The comprehensive model Mantis2, pre-installed in GPS-X 6.4 (Hydromantis, Canada), extends ASM2d and ADM1 (Anaerobic Digestion Model No. 1) with side-stream partial nitritation/denitritation and anammox process for nitrogen removal. The addition of a special algebraic pH solver also allows estimation of the actual pH value in the anammox-based process by solving a set of algebraic equations for a charge balance along with the equilibrium equations for each ionic species.

A completely mixed tank layout was set up in GPS-X 6.4 with a working volume of 4 L, using a DO controller and pH setup. The measured process variables, including NH₄-N, NO₃-N, NO₂-N, DO, pH, temperature and mixed liquor volatile suspended solids as well as the constant microbial community composition (anammox, nitrifiers, heterotrophs) were imported for the simulation of the nitrogen removal processes at the different pH scenarios (6, 6.5, 7, 7.5, 8, 8.5, 9, 9.5, 10, 10.5).

For model calibration, a special GPS-X utility called ‘Optimizer’ was used. Two kinetic parameters, the maximum specific growth rates of AOB (μ_max,AOB) and anammox bacteria (μ_max,Anam), were optimized based on the Nelder–Mead simplex method with the maximum likelihood objective functions, using the measured NH₄-N and NO₂-N concentrations as the target variables, while NO₃-N was not used as a target variable. The remaining kinetic and stoichiometric parameters (55 parameters) (Table S1, available with the online version of this paper) were adopted without any change from the Mantis2 model (Hydromantis, Canada). The 95% liner-approximation confidence limits for the parameter estimates were calculated from the variance–covariance matrix. Furthermore, correlation matrices (μ_max,AOB and μ_max,Anam) at the different pH scenarios were developed to evaluate the degree of correlation between the adjusted parameters.

The adjusted model parameters (μ_max,AOB, μ_max,Anam) were from the optimization. The peak value μ_max,AOB at the different pH scenarios (from 6 to 10.5) was 0.81 d⁻¹ at pH = 7.5 (T = 30 °C), which was much smaller than the Mantis2 value of 1.80 d⁻¹ (T = 30 °C). The peak value of μ_max,Anam at the different pH scenarios (from 6 to 10.5) was 0.366 d⁻¹ at pH = 7.5 (T = 30 °C), which was much higher than the Mantis2 value of 0.037 d⁻¹ (T = 30 °C). Both parameters (μ_max,AOB and μ_max,Anam) decreased greatly (90%) when pH deviated from the estimated pH optimum (pH = 7.5) to pH extremes (pH = 6 or pH = 9.5).

Both Hultman pH function and Michaelis pH function gave a superior description of the measured data (R² close to 1).The pH optima for AOB and anammox bacteria were 7.4 and 7.6, respectively. The maximum specific growth rate of AOB at the pH optimum was 0.81–0.85 d⁻¹ (at T = 30 °C). The maximum specific growth rate of anammox bacteria at the pH optimum was 0.36–0.38 d⁻¹ (at T = 30 °C), which could be attributed to the fast-growing anammox-enriched cultures (0.33 d⁻¹ at 30 °C, 0.39 d⁻¹ at 37 °C, 0.59 d⁻¹ at 35 °C) (Isaka et al. 2006; Bae et al. 2010; Lotti et al. 2015) rather than the slow-growing anammox cultures (0.02 d⁻¹ at 30 °C, 0.04–0.077 d⁻¹ at 32–33 °C, 0.13–0.19 d⁻¹ at 37 °C) (van de Graaf et al. 1996; van der Star et al. 2007; Hao et al. 2009).

Subsequently, the SAAs were calculated based on the NH₄-N and NO₂-N measurements. Predicted SAAs were obtained through model simulations after applying the optimized maximum specific growth rates for each examined pH. The Michaelis pH function was also used to fit the measured SAAs (Table 3).

The estimated optimal pH values for AOB and anammox bacteria were 7.4 and 7.6, respectively. The maximum specific growth rates of AOB and anammox bacteria at the pH optima were 0.81–0.85 d⁻¹ and 0.36–0.38 d⁻¹ (at T = 30 °C), respectively. The measured SAAs, predicted SAAs by Mantis2 and fitted SAAs by the Michaelis pH function at the pH optima were 0.895, 0.858 and 0.831 gN/(gVSS·d), respectively. The pH effects on nitrogen removal in the anammox-enriched granular sludge could be described by the two examined pH functions, i.e., the Michaelis pH function and the Hultman pH function. The impact of the pH extremes on the anammox-enriched granular sludge resulted from the inhibition of FA (at high pH) and FNA (at low pH) and the inhibition of the pH extremes on the intracellular proton transfer for cell survival. Therefore, the existing mechanistic models need extensions with respect to the pH inhibition functions on the maximum specific growth rates of both AOB and anammox bacteria.

This study was financially supported by the National Science Centre (Poland) under project no. UMO-2011/01/B/ST8/07289. During this study, Xi Lu and Zhixuan Yin were visiting researchers at Gdansk University of Technology within the framework of the CARBALA Project (CARbon BALAncing for nutrient control in wastewater treatment), People Maria Curie Actions (FP7-PEOPLE-2011-IRSES).