Including Nitrite as an Intermediate in Simultaneous Nitrification/Denitrification Membrane-Aerated Biofilm Reactor Models

Abstract

Our understanding of biofilm function has been greatly increased with the aid of one-dimensional biofilm models. However, to date, there has been no evaluation of how the inclusion or exclusion of nitrite within nitrification/denitrification biofilm models may impact simulated prediction of the biofilm community composition and effluent water quality. As such, several variations of commonly assumed nitrification and denitrification models with a common set of kinetic parameters were simulated in a one-dimensional membrane-aerated biofilm model to identify the differences in simulation output with respect to simultaneous nitrification and denitrification. Our results indicated that the inclusion of nitrite as an intermediate affected the biofilm community composition, microbial activity, and effluent water quality. Even under circumstances when the predicted water quality results were similar, significant differences persisted in the microbial activity. For example, chemical oxygen demand (COD) removal was independent of the including nitrite as an intermediate; however, the processes that contributed to COD removal (i.e., aerobic oxidation vs. anoxic denitrification) varied considerably between model assumptions, and as a result, affected differences between the predicted oxygen demand. Simulations also helped to identify inefficiencies caused by nitrite looping, a phenomenon in which NO₂⁻ cycles between oxidation within aerobic regions and reduction within anoxic regions. Overall, results indicated that if one-dimensional biofilm models are used to inform our understanding of fundamental biofilm processes, then the manner in which nitrite is modeled must be carefully considered to avoid introducing modeling artifacts into our interpretation of simulation results.

Introduction

Biofilm models are rapidly evolving to emulate the complexity of processes implicit to real biofilms. In particular, nitrification and denitrification reactions within mathematical simulations have expanded from simple models capable of predicting effluent ammoniacal-nitrogen (NH_x) and nitrate (NO₃⁻) concentrations in an activated sludge plant (Henze et al., 2000) to more complex models capable of simulating the fate and production of all major forms of nitrogen (e.g., Hiatt and Grady, 2008).

A variety of models have been used by researchers to simulate nitrogen dynamics in biofilms (Table 1). As a basis for the current work, all permutations of one-step and two-step nitrification and denitrification biopathways are considered (Table 2). Essentially, the models are defined by the inclusion or exclusion of nitrite (NO₂⁻) within nitrification and denitrification biopathways. Inclusion of NO₂⁻ increases the number of processes modeled, thereby increasing the complexity of the model. Two different methodologies for modeling denitrification have been typically adopted by past modeling efforts. The first category, parallel denitrification, assumes that both NO₃⁻ and NO₂⁻ can be reduced to nitrogen gas (N₂), but NO₃⁻ reduction does not proceed via the intermediate product NO₂⁻ before terminating as N₂ (Matsumoto et al., 2007; Downing and Nerenberg, 2008; Lackner et al., 2008; Wang et al., 2009). The second category, sequential denitrification, assumes that NO₃⁻ reduction must proceed through NO₂⁻ before being reduced to N₂ (Tiedje, 1998).

Table 1.

Assumptions Used in Previously Published Models

Nitrification pathway	Denitrification and/or Anammox pathway	References
2-Step	Anammox	Koch et al. (2000)
2-Step	Anammox	Hao et al. (2002)
1-Step	1-Step	Shanahan and Semmens (2004)
1-Step	1-Step	Lee and Park (2007)
2-Step	Anammox	Terada et al. (2007)
2-Step	2-Step parallel	Matsumoto et al. (2007)
1-Step	1-Step	Wolf et al. (2007)
2-Step	2-Step parallel	Elenter et al. (2007)
2-Step	2-Step parallel and Anammox	Lackner et al. (2008)
2-Step	2-Step parallel and Anammox	Liu and Horn (2012)
2-Step	2-Step parallel	Wang et al. (2009)

Anammo×® (Pagues BV, Balk, The Netherlands), anaerobic ammonium oxidation.

Table 2.

Permutations of Single-Step and Two-Step Nitrification and Denitrification Pathways

Model abbrv.	Description	Simplified stoichiometry	No. of processes
1N1D	One-step nitrification	NH_x→NO₃⁻	7
	One-step denitrification	NO₃⁻→N₂
1N2D Seq	One-step nitrification	NH_x→NO₃⁻	14
	Sequential denitrification	NO₃⁻→NO₂⁻→N₂
2N1D	Two-step nitrification	NH_x→NO₂⁻→NO₃⁻	13
	One-step denitrification	NO₃⁻→N₂
2N2D Par	Two-step nitrification	NH_x→NO₂⁻→NO₃⁻	18
	Parallel denitrification	NO₃⁻→N₂ and NO₂⁻→N₂
2N2D Seq	Two-step nitrification	NH_x→NO₂^-→NO₃⁻	18
	Sequential denitrification	NO₃⁻→NO₂⁻→N₂

1N1D, one-step nitrification, one-step denitrification; 1N2D Seq, one-step nitrification, two-step sequential denitrification; 2N1D, two-step nitrification, one-step denitrification; 2N2D Seq, two-step nitrification, two-step sequential denitrification; 2N2D Par, two-step nitrification, parallel denitrification.

The degree of model complexity is often chosen based upon the bulk liquid NO₂⁻ concentrations. For example, if the NO₂⁻ concentrations in the bulk liquid are minimal, the modeler may choose to exclude NO₂⁻ as in the suite of activated sludge models (ASM) (Henze et al., 2000); however, this approach may be ill advised since NO₂⁻ could be a critical component governing the rate of biological processes within the biofilm, even though it does not appear to be of any significance given the bulk water quality data (e.g., de Beer et al., 1997).

A survey of one-dimensional biofilm models detailed the various nitrification and denitrification models that have been used in past works (Table 1). Aside from Elenter et al. (2007), every model that used two-step nitrification studied nitritation in which NO₂⁻ was the desired end product of nitrification. Even though sequential denitrification was the most used two-step mechanism for ASM (Sin et al., 2008), the authors of this manuscript are not aware of a one-dimensional biofilm model published in literature that has considered modeling denitrification in this manner. Although a thorough comparison of nitrification and denitrification reactor model results with and without nitrite as an intermediate have not been investigated previously, critical modeling aspects such as kinetics and rate expressions have been discussed elsewhere (Carrera et al., 2004; Latimer et al., 2009).

One of the primary reasons that research interest has been directed toward nitrogen removal pathways in biofilm-based technologies is the proficiency of biofilms to develop a redox-stratified condition that supports efficient nitrogen elimination pathways such as simultaneous nitrification and denitrification (SND) and deammonification (combination of partial nitritation and anaerobic ammonium oxidation). The membrane-aerated biofilm reactor (MABR), in which, oxygen-permeable membranes serve as both a means of biofilm immobilization and aeration, is one such technology that is particularly adept at supporting both SND and deammonification (Syron and Casey, 2008).

This study aims to examine the effects of commonly applied nitrification and denitrification models on one-dimensional membrane-aerated biofilm models performing SND. To achieve this goal, we modeled the biofilm community composition and water quality of a membrane-aerated biofilm using five combinations of nitrification—denitrification pathways using a common set of kinetic parameters (Table 2). In addition, we also explored the differences due to the type of waste stream treated and associated nitrogenous loading rate.

Experimental Protocols

Development of mathematical models

A multispecies MABR was modeled using the one-dimensional biofilm reactor compartment of the simulation software Aquasim 2.1g (Reichert, 1998). A diffusive link within the Aquasim software was used to link a completely mixed gas compartment with the base of the biofilm compartment such that a counter-diffusion biofilm was simulated (Debus and Wanner, 1992). The biofilm was assumed to be rigid (i.e., porosity throughout the biofilm is constant) with a constant liquid volume fraction (ɛ_l) of 0.75 (Koch et al., 2000). The simulated reactor had a confined reactor volume of 5×10⁻³ m³ and was supplied with oxygen via 250 siloxane membranes, each 300mm long with outer and inner diameters of 2 and 0.8 mm, respectively, that yielded a biofilm attachment surface area of 0.75 m². Since the microbial population is attached directly to the oxygenation source (i.e., the shell-side of the membrane), oxygen transfer efficiencies are relatively high compared to other aerated biological processes. Additionally, oxygen gradients within the biofilm can be better controlled, thereby allowing redox-stratified conditions necessary for SND and deammonification to develop throughout the biofilm's depth.

Five models based upon the inclusion or exclusion of nitrite as an intermediate within nitrification and denitrification models were used as the basis for this study (Table 2). One-step nitrification was catalyzed by a lumped microbial group, autotrophic nitrifying organisms (ANO), while two-step nitrification was characterized by ammonium oxidizing organisms (AOO) and nitrite oxidizing organisms (NOO). Ordinary heterotrophic organisms (OHO), anaerobic ammonium oxidizing organisms (AMO), and particulate undegradable organics (U) were incorporated into all model variations. Stoichiometric parameters for AMO were selected from an elemental balance conducted by Takacs et al. (2007). All kinetic and stoichiometric parameters included in the model simulations, as well as the stoichiometric matrix, are detailed in the Supplementary Data (Supplementary Tables S1–S4). Parameter notation was applied according to a recently proposed standardized framework for wastewater treatment modeling (Corominas et al., 2010). The reaction rate expressions used in this study were similar to Matsumoto et al. (2007) (Supplementary Table S5).

The one-step nitrification, one-step denitrification (1N1D) model was the only simulation that did not actively model AMO since NO₂⁻ production was not included. Heterotrophic growth occurred during aerobic and anoxic conditions with a reduction factor to account for a decreased activity when NO_x (NO₂⁻ or NO₃⁻) was utilized as the electron acceptor. All models assumed that NH_x was released during decay and assimilated during growth reactions. Endogenous decay was modeled under aerobic and anoxic conditions for heterotrophs and autotrophs (Manser et al., 2006). Oxygen flux, J_O2, followed a mechanistic oxygen transfer model derived via an oxygen mass balance around the lumen of a hollow fiber membrane (Gilmore et al., 2009): \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {\rm J}_{\rm O2} = \rm nQ_{\rm air}(\rm HS_{\rm O2,F}-S_{\rm O2,G} \left[ exp \left( - \frac {\pi dD_{O2, silicone} K_{G} l} {Q_{air} {\delta}} \right) - 1 \right] \tag {1} \end{align*} \end{document}

where n is the number of membranes; Q_air is the volumetric airflow rate; H is the nondimensional Henry's law constant; S_O2,F and S_O2,G are the concentrations of oxygen at the biofilm base and within the membrane lumen, respectively; d, l, and δ are the inner diameter, length, and thickness of the membranes, respectively; D_O2,silicone is the diffusion coefficient of oxygen in silicone, and K_G is the membrane/air partition coefficient. The biofilm's area, A, was modeled as if grown on a cylindrical base such as a hollow fiber membrane: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {\rm A} = \pi \rm nl (d + 2 \delta + 2z) \tag{2} \end{align*} \end{document}

where z is the distance perpendicular from the substratum. The biomass detachment velocity, u_de, was calculated in the same manner as Lackner et al. (2008): \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}\rm { u_ { de } = u_F \left( \frac { L_F } { L_ { F , max } } \right)^2 \tag { 3 } } \end{align*} \end{document}

where u_F is the biofilm growth velocity, L_F is the biofilm thickness, and L_F,max is the maximum biofilm thickness.

Model simulations

The biofilm was discretized into 20 grid points, and each simulation spanned 100 days with time steps of 0.01 days. Within the 100 days of simulation time, steady state was reached with respect to a biomass concentration, soluble species concentration, and biofilm thickness. The treatment of three different wastewaters was simulated to determine the impact of the various nitrification and denitrification mechanisms on a range of waste types. The simulated wastewaters were a primary effluent characteristic of a domestic wastewater treatment plant, a septic tank effluent, and a centrate sidestream from an anaerobically digested sludge. Waste stream characteristics and modeled parameters are detailed in Table 3. Simulated biofilm thickness and intramembrane partial pressure were chosen based on an iterative process to maximize total nitrogen (TN) removal via SND for each waste stream whenever simulated using the two-step nitrification, two-step sequential denitrification (2N2D Seq) model type.

Table 3.

Waste Stream Characteristics

Waste type	Refs.	COD (g-COD/m³)	NH₄⁺ (g-N/m³)	HCO₃⁻ (mol/m³)	L_max (μm)	pO₂ (kPa)
Domestic	Tchobanoglous et al. (2003)	400	40	6.0^a	300	40
STE	Lowe et al. (2007)	400	58	10.3	300	28.5
Centrate sidestream	Arnold et al. (2000)	1300	600	40^b	500	28.5

Sufficient alkalinity for complete nitrification.

Assumed as 50% of the stoichiometric alkalinity requirement for nitrification.

COD, chemical oxygen demand; STE, septic tank effluent.

To more clearly identify the inherent differences that could be propagated by including nitrite as an intermediate within nitrification and denitrification biofilm models, three separate studies with varying levels of complexity were examined. Study 1 only examined the centrate sidestream at a nitrogen-loading rate of 1.25 g-N/day (1.67 g-N/m²-day). Since only one waste type and loading rate were simulated, it was possible to evaluate the differences between the nitrification and denitrification models types for a broad range of outputs, including substrate removal rates/efficiencies, substrate profiles inside of a simulated biofilm, biomass profiles inside of a simulated biofilm, and process contribution toward the overall substrate flux. Study 2 determined whether waste type/strength impacted the variances between model types as noted in Study 1, but only evaluated chemical oxygen demand (COD) and TN removal as well as O₂ flux. Finally, Study 3 extended the previous two studies to include the differences between steady-state models due to changes in mass loading rates of each wastewater.

Process contribution toward substrate removal rates

Contributions of each process toward the total flux, J, of soluble species S_i were quantified by using a trapezoidal approximation: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {J_{S_{i , j}}} =& \sum\nolimits_{k = 1}^{N - 1} [ 0.5(j_{X_i , z_k} + j_{X_i , z_{k + 1}}) (z_{k + 1} - z_k) \\& \cdot V{S_{i , j}} \cdot 0.5 ( A_{z_k} + A_{z_{k + 1}}) ] \tag {4} \end{align*} \end{document}

where \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$J_{S_{i , j}}$$ \end{document} is the contribution of process j to the flux of compound S_i, \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$j_{X_{i , z_k}}$$ \end{document} is the growth rate of species X_i at a biofilm thickness of z_k, k is the biofilm node number, N is the total number of nodes, ν is the stoichiometric coefficient for species S_i during process j, and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$A_{z_k}$$ \end{document} is the surface area at a biofilm thickness of z_k per Equation (2).

Results

Hypothesized differences caused by including nitrite as an intermediate

As a starting point for this study, some of the hypothesized differences in model output as caused by nuances in each model's nitrification and denitrification biopathways are highlighted below. Obviously, one-step nitrification models (1N1D and one-step nitrification, two-step sequential denitrification [1N2D Seq]) would not accurately capture the flux of NO₂⁻. As such, it should be noted that this article did not attempt to compare the accuracy of the five nitrification and denitrification model types; rather, this study was an attempt to compare differences in their simulation output under a consistent set of kinetic parameters. Even though AMO species were included in the modeling framework, it would be expected that the AMO activity would be negligible since the simulations were designed to optimize the degree of SND under relatively high oxygen fluxes when compared to MABR modeling studies specifically optimized for AMO growth (Lackner et al., 2008). The 1N2D Seq model type had questionable utility since it was the only model type that had not been modeled in literature. It was included in this exercise to provide exhaustive coverage of all permutations attainable via one-step and two-step nitrification and denitrification models. Although not used in practice, the 1N2D Seq model type should provide a unique simulation in which, no aerobic NO₂⁻ sinks exist; thereby, all of the effluent NO₂⁻ would be attributable to incomplete denitrification. Another unique situation was expected for the 2N2D Par (two-step nitrification, parallel denitrification) model type in which, the modeler could quantify the contribution of short-cut denitrification (i.e., the mass of NO₂⁻ reduced directly to N₂ via denitrifying OHO before being oxidized to NO₃⁻ by NOO). Since short-cut denitrification was modeled using a specific biopathway for the 2N2D Par model type, it was hypothesized that this model would report the highest TN removal when compared to the other model types, especially for centrate sidestream due to its carbon-limited nature and the inherently efficient utilization of COD that is characteristic of short-cut denitrification (Ahn, 2006). Finally, the 2N2D Seq model type could produce a special scenario called nitrite looping: a phenomenon in which, NO₂⁻ that was reduced by denitrifying OHOs was reoxidized to NO₃⁻ by NOOs rather than being further reduced to N₂ via the denitrification pathway (Latimer et al., 2009; Landes et al., 2011). Since NO₂⁻ reduction efficiency could be diminished as a result of nitrite looping, lower TN fluxes were expected for the 2N2D Seq model type.

Study 1—identifying variations in model output

All of the results for Study 1 were based upon the simulated capacity of a MABR to treat centrate at a nitrogen loading of 1.25 g-N/day (1.67 g-N/m²-day). Oxygen flux was adjusted to optimize SND for the 2N2D Seq model. Percent COD (∼100%) and NH_x (56.2–61.9%) removal were similar for all models types, while TN removal efficiencies displayed some variance between model types (Table 4). As hypothesized, the 2N2D Par model exhibited the highest TN removal efficiency (32.2%), whereas the 2N2D Seq model produced the lowest percent TN removals (20.8%). Effluent NH_x concentrations varied as much as 34 g-N/m³ between model types. The only model type that resulted in substantial bulk liquid NO₂⁻ concentrations (86 g-N/m³) was the 1N2D Seq model type in which, an unexpectedly large fraction of incomplete denitrification was simulated. Although this result seems like an outlier in the current study, incomplete denitrification may be a reasonable occurrence within MABRs characterized by thin, oxygen penetrated biofilms since oxygen completely inhibits the nitrite reductase, while partially inhibiting the nitrate reductase (Phillips et al., 2002).

Table 4.

Steady-State Water Quality Performance as a Function of Model Type and Waste Type at a J_N=1.25 g-N/day (1.67 g-N/m²-day)

Waste type	Model type	COD removal (%)	NH_x removal (%)	TN removal (%)	Effluent NH_x (g-N/m³)	Effluent NO₂ (g-N/m³)	Effluent NO₃ (g-N/m³)
Centrate sidestream	1N1D	99.8	60.5	29.4	237.3	0.0	186.2
	1N2D Seq	100.0	60.2	28.8	239.0	86.0	102.3
	2N1D	99.7	61.2	30.9	232.8	0.1	182.0
	2N2D Seq	100.0	56.2	20.8	262.9	0.7	211.5
	2N2D Par	99.8	61.9	32.2	228.8	0.0	178.1
Domestic	1N1D	93.7	80.8	45.6	7.7	0.0	14.1
	1N2D Seq	94.5	89.0	54.7	4.4	8.2	5.5
	2N1D	93.7	82.7	47.9	6.9	0.2	13.8
	2N2D Seq	94.4	88.4	51.2	4.6	0.9	14.0
	2N2D Par	93.8	85.6	53.3	5.8	0.1	12.9
STE	1N1D	95.7	92.5	65.1	4.3	0.0	15.8
	1N2D Seq	96.6	93.6	75.2	3.7	9.1	1.5
	2N1D	95.7	92.8	67.3	4.2	0.2	14.6
	2N2D Seq	96.6	93.4	64.6	3.8	1.3	15.3
	2N2D Par	95.9	93.1	72.3	4.0	0.0	12.0

The differences in model output were further investigated by plotting the steady-state substrate concentration profiles in a biofilm (Fig. 1). Concentration profiles within the biofilm were relatively flat for NH_x and NO_x (i.e., the concentration at z=0 was equal to the concentration at z=L_max) indicating that the diffusional flux was equivalent to the flux of microbial conversion. Nonetheless, unique nitrogen profiles were generated by the sequential denitrification model types. The 1N2D Seq model produced elevated NO₂⁻ concentrations, while the 2N2D Seq model resulted in the highest NH_x (Fig. 1).

FIG. 1.

Steady-state substrate profiles at a nitrogen loading of 1.25 g-N/day (1.67 g-N/m²-day) for treatment of centrate sidestream waste.

For all model types, O₂ penetrated to a depth of ∼270 μm (54% of L_max) before the O₂ concentrations dropped below the O₂ half-saturation constant for nitrifiers (K_O2=0.5 g-O₂/m³). The depth of simulated O₂ diffusion agreed relatively well with an experimental study that treated a carbon-limited waste (total organic carbon:N=1.1) using an MABR in which, O₂ diffused 18–43% of the total biofilm depth (Terada et al., 2003). Finally, the HCO₃⁻ concentrations throughout the biofilm were well below the HCO₃⁻ half-saturation constant for ANO/AOO (K_HCO3=1.78 mol/m³) (Fig. 1), thereby limiting the NH_x removal efficiencies for all model simulations to ∼60% (Table 4). It should be noted that the estimated NH_x removal efficiencies would only be ∼50% if not for the denitrification-produced alkalinity.

Although water quality was relatively similar for all models, there were significant differences in the predicted biofilm community composition. The sequential biofilm models (1N2D Seq and 2N2D Seq) were characterized by sharper transitions in the OHO volume fraction indicating that heterotrophic bacteria occupied a smaller fraction of the biofilm depth (Fig. 2). Since less competition between nitrifying and heterotrophic bacteria would exist for the sequential model types due to the lower OHO populations in the aerated portions of the biofilm, it was assumed that the sequential model types would simulate higher rates of the nitrifying activity. On the contrary, the sequential model types produced the lowest percent NH_x removals when compared to the other model types (Table 4). Although the NH_x removal efficiencies for the 1N2D Seq model type did not deviate significantly from the remaining models, the 2N2D Seq simulations produced unique NH_x results. Interestingly, the primary cause for the contrast between 2N2D Seq model simulations and the remaining models was explained by the competition for oxygen exerted by NOO as shown by the distinctively large NOO volume fraction profile (Fig. 2). The maximum NOO volume fraction within a given biofilm slice for the 2N2D Seq model was 7.8%, whereas the other two model types that considered NOO (two-step nitrification, one-step denitrification [2N1D] and 2N2D Par) displayed maximum values of 3.5%. Similarly, as determined by integrating the curves in Fig. 2, the cumulative NOO biomass concentration for the 2N2D Seq model was 0.76 g/m², almost two-fold greater than the 2N1D and 2N2D Par model types that both reported NOO concentrations of 0.41 g/m².

FIG. 2.

Steady-state biomass profiles at a nitrogen loading of 1.25 g-N/day (1.67 g-N/m²-day) for treatment of centrate sidestream waste.

The final aspect of Study 1 sought to investigate the disparities in process contribution toward a given substrate's flux as caused by including nitrite as an intermediate. Equation (4) was applied to each process that contributed to the removal and/or production of COD, TN, and O₂. In this manner, a detailed itemization of each reaction rate's contribution to the net flux of COD, TN, and O₂ was elucidated as shown in Fig. 3A, D, and G. Aerobic oxidation of COD was marginal for most model types with contributions toward the net COD flux ranging from 12% for the 1N1D model to 1.6% for the 2N2D Seq model. As cited for the slightly higher TN removal rates achieved in the 2N2D Par model, the contribution of short-cut denitrification was quantified to be 1.9% (Fig. 3E). For the sequential model types, the flux of COD via reduction of NO₃^- should account for 40% of the total anoxic COD removal rate based upon stoichiometry; however, this value was simulated as 50% and 64% for the 1N2D Seq and 2N2D Seq models, respectively (Fig. 3A). In addition to the high bulk liquid NO₂⁻ concentrations, the divergence from stoichiometric expectations further confirmed that incomplete denitrification occurred in the 1N2D Seq model type. On the other hand, the 2N2D Seq simulation departed from stoichiometric expectations due to a reason other than incomplete denitrification. Instead, nitrite looping propagated the inefficiencies noted for COD removal.

FIG. 3.

Process contributions toward overall COD removal rate (A–C), TN removal rate (D–F), and oxygen consumption rate (G–I). All simulations were conducted at a nitrogen loading of 1.25 g-N/day (1.67 g-N/m²-day) for the treatment of the centrate sidestream waste.

Within Study 1, five of the simulated results corroborated that nitrite looping occurred in the 2N2D Seq model: (1) TN removal rates were lower than other model types (Table 4) since more NO₂⁻ was being oxidized rather than reduced following the first step of denitrification, (2) NOO populations were higher than other model types (Fig. 2) despite lower NH_x removal (Table 4), (3) the production of NO₂⁻ by AOO was 0.68 g-N/day (0.91 g-N/m²-day), while the consumption of NO₂⁻ by NOO was 1.10 g-N/day (1.46 g-N/m²-day), (4) the consumption of O₂ by NOO was 10% higher than stoichiometrically predicted (Fig. 3G), and (5) the amount of COD removed during denitrification with NO₃⁻ as the electron acceptor was 24% higher than stoichiometrically predicted (Fig. 3A).

Experimental evidence that explicitly corroborated the occurrence of nitrite looping was not found in literature; however, the following studies have examined the topic. While quantifying the community makeup of three aerobic granular systems using fluorescent in situ hybridization and quantitative polymerase chain reaction, Winkler et al. (2012) reported that (1) the NOO/AOO ratio was unexpectedly high and (2) the nitrite oxidizing capacity was three times greater than the ammonium oxidizing capacity. The researchers hypothesized that nitrite looping could be responsible for these findings. In another instance, an activated sludge modeling study found that a calibrated full-plant simulator predicted lower effluent TN concentrations whenever a 1N1D modeling framework was simulated as opposed to a 2N2D Seq modeling framework (Latimer et al., 2009). The differences in the predicted TN concentration were attributed to nitrite looping, and its effects on the 2N2D Seq simulation results were minimized by changing half-saturation affinity constants to enhance the NO₂⁻ uptake ability of facultative OHO versus NOO. In this manner, the 2N2D Seq model was better calibrated to match experimental results for an activated sludge system, and the occurrence of nitrite looping within the simulation results was explained through the lens of kinetics. On the other hand, a previous modeling study focused on biofilms rather than suspended biomass, suggested that nitrite looping within biofilm systems may be plausible so long as denitratation-produced NO₂⁻ interacted with NOO before being either reduced to N₂ or diffusing out of the biofilm (Landes et al., 2011). Although the simulation results were attributed to diffusional transport of denitratation-produced NO₂⁻ from anoxic to aerobic regions, it was postulated that this interaction could also occur due to close spatial proximity of OHOs and NOOs or advective transport mechanisms. For the current study, nitrite looping was mathematically possible as a result of denitratation-produced NO₂⁻ diffusing from anoxic to aerobic regions.

Study 2—impact of waste type on model output

The general purpose of Study 2 was to determine whether the variations in model output, as demonstrated through Study 1, were evident whenever other waste types were simulated (Table 4 and Fig. 3). For all waste types investigated, COD removal did not vary significantly as nitrite was included. Unlike the centrate sidestream simulations, the highest NH_x and TN removal efficiencies were reported for the 1N2D Seq model. The 2N1D and 2N2D Par models exhibited similar O₂ consumption and COD removal patterns; however, as a result of short-cut denitrification, the 2N2D Par model reported higher TN removals. In general, a consistent theme regardless of the waste applied was that the models with the highest NH_x removal also reported the highest TN removal.

For the 2N2D Seq simulations, a wide variety of nitrite looping scenarios occurred. The amount of NO₂⁻ oxidized by NOO was 1.6-, 1.4-, and 1.8-fold greater than the amount of NO₂⁻ produced by AOO whenever the centrate sidestream, domestic wastewater, and septic tank effluent waste were simulated, respectively. Despite the variation in NO₂⁻ consumption by NOO, the amount of oxygen consumed by NOO was consistently higher for the 2N2D Seq model type regardless of the wastestream that was simulated (Fig. 3G–I). The contribution of NH_x assimilation toward TN removal decreased as aerobic COD consumption by OHO decreased (Fig. 3A–F). The sequential model types always reported the least contribution of aerobic COD oxidation; thereby, more COD was consumed via anoxic denitrification processes (Fig. 3A–C). Complementary to increased COD consumption via denitrification should be increased TN removal as was the case for the 1N2D Seq model type; however, nitrite looping prevented the efficient use of COD under anoxic conditions for the 2N2D Seq simulations (Fig. 3D–F).

Study 3—impact of loading rate on model output

Study 3 aided in determining whether or not implications from Studies 1 and 2 could be applied as the nitrogen mass loading rate was varied. To provide the general trends that accompanied changes in mass loading rates, only the centrate sidestream and the domestic wastewater were presented (Figs. 4 and 5). These two waste types were presented since they were most affected by nitrogen mass loading rates.

FIG. 4.

Process contributions toward overall COD removal rate (A–E), TN removal rate (F–J), and oxygen consumption rate (K–O) for simulations treating a centrate sidestream.

FIG. 5.

Process contributions toward overall COD removal rate (A–E), TN removal rate (F–J), and oxygen consumption rate (K–O) for simulations treating a domestic wastewater.

Centrate sidestream simulations

In general, COD and TN removals did not vary for the centrate sidestream, except for the lowest loading rate in which, OHO began to compete with nitrifying bacteria for O₂ (Fig. 4). Although the intramembrane air pressure was held constant for all nitrogen loading rates, the flux of O₂ that diffused into the biofilm reactor compartment increased as the microbial activity increased at higher loading rates (Fig. 4K–O). For reference, the maximum J_O2 for the centrate sidestream simulations was 12.4 g-O₂/day (16.5 g/m²-day), which indicated that the O₂ transfer efficiencies increased from 24% to 44% as the nitrogen loading increased from 1.0 to 2.0 g-N/day (1.3–2.6 g-N/m²-day). For all model types simulated with the centrate waste, bacterial assimilation accounted for ∼4% of TN removal regardless of the loading rate. Similarly, decay processes contributed for ∼4% of TN removal at all loading rates.

Domestic wastewater simulations

NH_x and TN removal efficiencies declined at the highest loading rate whenever the model type did not include NO₂⁻ as an intermediate (1N1D, 2N1D, and 2N2D Par), while overall COD removal efficiencies did not greatly vary as a function of the model type (Fig. 5). Upon closer inspection, it was revealed that the contribution of oxic and anoxic processes toward COD removal varied between the model types (Fig. 5A–E). For all nitrogen loading rates, the sequential denitrification model types (1N2D Seq and 2N2D Seq) consistently had a larger anoxic process contribution toward COD removal than model types that omitted sequential denitrification (1N1D, 2N1D, and 2N2D Par). This was most evident at the highest loading rate in which, the oxygen consumption trend increased for aerobic heterotrophs, while it decreased for nitrifying organisms whenever nonsequential denitrification models were simulated (Fig. 5K–O). As a consequence of a less nitrification activity and less anoxic COD consumption, nonsequential denitrification models also reported decreased TN removal efficiencies at the highest liquid loading rate (Fig. 5F–J). Considering that the maximum J_O2 for domestic wastewater simulations was 16.4 g-O₂/day (21.9 g/m²-day), the simulated O₂ transfer efficiencies ranged from 26% to 45% for nonsequential denitrification models and from 26% to 49% for the sequential denitrification model types, which was similar to the values obtained with the centrate sidestream. Bacterial assimilation accounted for ∼14% of TN removal regardless of the loading rate, and for the lowest nitrogen loading rate, the vast majority of the total TN removal was attributable to bacterial assimilation. In contrast, decay processes contributed marginally to TN removal (<2.5%) at all loading rates.

Conclusions

The purpose of this study was to identify the inherent differences that could be propagated by including nitrite as an intermediate within nitrification and denitrification biofilm models. The overarching conclusion reached as a result of this study was that mathematical simulation results varied only slightly with respect to water quality (with the exception of the 2N2D Seq results) and more profoundly with biofilm speciation based upon the inclusion or exclusion of nitrite. Interestingly, the greatest disparities in model results occurred for seemingly identical model types, specifically models that applied parallel denitrification (2N2D Par) versus sequential denitrification (2N2D Seq).

Magnitude of dissimilarity between model types was shown to be a function of waste type and mass loading rate. Global implications could not be applied with respect to the impact of including nitrite as an intermediate; nonetheless, some general statements could be claimed based upon the conditions simulated in this article. COD removal rates did not vary between model types even though the processes that contributed to the COD flux did vary. The sequential denitrification model types (1N2D Seq and 2N2D Seq) had the lowest aerobic COD consumption contributions, the 1N2D Seq model type consistently had elevated NO₂⁻ concentrations when compared to the remaining models, and the rate of TN removal was proportional to the rate of NH_x removal. This study also introduced a method to itemize each processes contribution toward the flux of a given soluble species. Since biofilm models are often used to educate our understanding of microbial interactions, inclusion or exclusion of nitrite as an intermediate in biofilms models will play a significant role in the interpretation of how varying microbial processes interact. Finally, the impact of nitrite looping on the 2N2D Seq model type was highlighted. The current study demonstrated that nitrite looping was mathematically possible as a result of denitratation-produced NO₂⁻ diffusion from anoxic to aerobic regions whereby it was oxidized to NO₃⁻.

Footnotes

Acknowledgment

The authors would like to thank NASA and the Texas Space Grant Consortium for the generous support of this research.

Author Disclosure Statement

All authors declare that no competing financial interests exist.

Nomenclature

References

Ahn

2006. Sustainable nitrogen elemination biotechnologies: A review. Process Biochem., 41:1709.

Arnold

, Bohm

, Wilderer

P.A.

2000. Application of activated sludge and biofilm sequencing batch reactor technology to treat reject water from sludge dewatering systems: A comparison. Water Sci. Technol., 41:115.

Carrera

, Jubany

, Carvallo

, Chamy

, Lafuente

2004. Kinetic models for nitrification inhibition by ammonium and nitrite in a suspended and an immobilized biomass systems. Process Biochem., 39:1159.

Corominas

, Rieger

, Takacs

, Ekama

, Hauduc

, Vanrolleghem

, Oehmen

, Gernaey

, van Loosdrecht

, Comeau

2010. New framework for standardized notation in wastewater treatment modelling. Water Sci. Technol., 61:841.

de Beer

, Schramm

, Santegoeds

C.M.

, Kuhl

1997. A nitrite microsensor for profiling environmental biofilms. Appl. Environ. Microbiol., 63:973.

Debus

, Wanner

1992. Degradation of xylene by a biofilm growing on a gas-permeable membrane. Water Sci. Technol., 26:607.

Downing

, Nerenberg

2008. Effect of bulk liquid BOD concentration on activity and microbial community structure of a nitrifying, membrane-aerated biofilm. Appl. Environ. Microbiol., 81:153.

Elenter

, Milferstedt

, Zhang

, Hausner

, Morgenroth

2007. Influence of detachment on substrate removal and microbial ecology in a heterotrophic/autotrophic biofilm. Water Res., 41:4657.

Gilmore

, Little

, Smets

, Love

2009. Oxygen transfer model for a flow-through hollow-fiber membrane biofilm reactor. J. Environ. Eng., 135:806.

10.

Hao

, Heijnen

, van Loosdrecht

2002. Sensitivity analysis of a biofilm model describing a one-stage completely autotrophic nitrogen removal (CANON) process. Biotechnol. Bioeng., 77:266.

11.

Henze

, Gujer

, Mino

, van Loosdrecht

M.C.M.

2000. Activated Sludge Models ASM1, ASM2, ASM2d and ASM3. London: IWA.

12.

Hiatt

, Grady

2008. An updated process model for carbon oxidation, nitrification, and denitrification. Water Environ. Res., 80:2145.

13.

Koch

, Egli

, Van der Meer

, Siegrist

2000. Mathematical modeling of autotrophic denitrification in a nitrifying biofilm of a rotating biological contactor. Water Sci. Technol., 41:191.

14.

Lackner

, Terada

, Smets

B.F.

2008. Heterotrophic activity compromises autotrophic nitrogen removal in membrane-aerated biofilms: Results of a modeling study. Water Res., 42:1102.

15.

Latimer

R.J.

, Pitt

, Dold

, Takacs

, Lynch

T.J.

2009. Kinetic parameters for modeling two-step nitrification and denitrification: A case study. Water Sci. Technol., 59:631.

16.

Landes

, Jackson

W.A.

, Morse

2011. Nitrite looping: plausible phenomenon or mathematical nuance? Proceedings of the Water Environment Federation, Los Angeles, CA. WEFTEC, 2011Session 51–603219.

17.

Lee

, Park

2007. One-dimensional mixed-culture biofilm model considering different space occupancies of particulate components. Water Res., 41:4317.

18.

Liu

S.T.

, Horn

2012. Effects of biofilm geometry on deammonification biofilm performance: A simulation study. Bioresour. Technol., 116:252.

19.

Lowe

K.S.

, Rothe

N.K.

, Tomaras

J.M.B.

, DeJong

, Tucholke

M.B.

, Drewes

, McCray

J.E.

, Munakata-Marr

2007. Influent Constituent Characteristics of the Modern Waste Stream from Single Sources: Literature Review04-DEC-1aWater Environment Research Foundation: Alexandria, VA.

20.

Phillips

, Laanbroek

H.J.

, Verstraete

2002. Origin, causes and effects of increased nitrite concentrations in aquatic environments. Rev. Environ. Sci. Biotechnol., 1:115.

21.

Manser

, Gujer

, Siegrist

2006. Decay processes of nitrifying bacteria in biological wastewater treatment systems. Water Res., 40:2416.

22.

Matsumoto

, Terada

, Tsuneda

2007. Modeling of membrane-aerated biofilm: Effects of C/N ratio, biofilm thickness and surface loading of oxygen on feasibility of simultaneous nitrification and denitrification. Biochem. Eng. J., 37:98.

23.

Reichert

1998AQUASIM 2.0—User Manual. CH-8600 DubendorfSwitzerland: Swiss Federal Institute for Environmental Science and Technology (EAWAG), 214.

24.

Shanahan

, Semmens

2004. Multipopulation model of membrane-aerated biofilms. Environ. Sci. Technol., 38:3176.

25.

Sin

, Kaelin

, Kampschreur

, Takacs

, Wett

, Gernaey

, Rieger

, Siegrist

, van Loosdrecht

2008. Modelling nitrite in wastewater treatment systems: A discussion of different modelling concepts. Water Sci. Technol., 58:1155.

26.

Syron

, Casey

2008. Membrane-aerated biofilms for high rate biotreatment: Performance appraisal, engineering principles, scaleup, and development requirements. Environ. Sci. Technol., 42:1833.

27.

Takacs

, Vanrolleghem

, Wett

, Murthy

2007. Elemental balance based methodology to establish reaction stoichiometry in environmental modeling. Water Sci. Technol., 56:37.

28.

Terada

, Hibiya

, Nagai

, Tsuneda

, Hirata

2003. Nitrogen removal characteristics and biofilm analysis of a membrane-aerated biofilm reactor applicable to high-strength nitrogenous wastewater treatment. J. Biosci. Bioeng., 95:170.

29.

Terada

, Lackner

, Tsuneda

, Smets

2007. Redox-stratification controlled biofilm (ReSCoBi) for completely autotrophic nitrogen removal: The effect of co- versus counter-diffusion on reactor performance. Biotechnol. Bioeng., 97:40.

30.

Tchobanoglous

, Burton

F.L.

, Stensel

H.D.

2003. Wastewater Engineering: Treatment and Reuse. New York: McGraw-Hill, 1819.

31.

Tiedje

J.M.

1988. Ecology of denitrification and dissimilatory nitrate reduction to ammonia. Zehnder

A.J.B.

Biology of Anaerobic Microorganisms. New York: John Wiley & Sons, 179.

32.

Wang

, Terada

, Lackner

, Smets

, Henze

, Xia

, Zhao

2009. Nitritation performance and biofilm development of co- and counter-diffusion biofilm reactors: Modeling and experimental comparison. Water Res., 43:2699.

33.

Winkler

M.K.H.

, Bassin

J.P.

, Kleerebezem

, Sorokin

D.Y.

, van Loosdrecht

M.C.M.

2012. Unravelling the reasons for disproportionate AOB and NOB in aerobic granular sludge. Appl. Microbiol. Biotechnol., 94:1657.

34.

Wolf

, Picioreanu

, van Loosdrecht

2007. Kinetic modeling of phototrophic biofilms: The PHOBIA model. Biotechnol. Bioeng., 97:1064.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.13 MB

0.04 MB

0.03 MB