Article pubs.acs.org/est
Modeling Electron Competition among Nitrogen Oxides Reduction and N2O Accumulation in Denitrification Yuting Pan, Bing-Jie Ni, and Zhiguo Yuan* Advanced Water Management Centre (AWMC), The University of Queensland, St Lucia, Brisbane, QLD 4072, Australia S Supporting Information *
ABSTRACT: Competition for electrons among different steps of denitrification has previously been shown to occur, and to play an important role in the accumulation and emission of N2O in wastewater treatment. However, this electron competition is not recognized in the current denitrification models, limiting their ability to predict N2O accumulation during denitrification. In this work, a new denitrification model is developed for wastewater treatment processes. It describes electron competition among the four steps of denitrification, through modeling the carbon oxidation and nitrogen reduction processes separately, in contrast to the existing models that directly couple these two types of processes. Electron carriers are introduced to link carbon oxidation, which donates electrons to carriers, and nitrogen oxides reduction, which receives electrons from these carriers. The relative ability of each denitrification step to compete for electrons is modeled through the use of different affinity constants with reduced carriers. Model calibration and validation results demonstrate that the developed model is able to reasonably describe the nitrate, nitrite, and N2O reduction rates of a methanol-utilizing denitrifying culture under various carbon and nitrogen oxides supplying conditions. The model proposed, while subject to further validation, is expected to enhance our ability to predict N2O accumulation in denitrification.
■
INTRODUCTION Denitrification is one of the key processes in biological nitrogen removal from wastewater. Heterotrophic denitrification converts nitrate generated from autotrophic nitrification to nitrogen gas (N2) thus removing nitrogen from wastewater. Denitrification consists of four consecutive reduction steps, which produce nitrite (NO2−), nitric oxide (NO), and nitrous oxide (N2O) as three obligatory intermediates. Each reduction step is catalyzed by one or more specific reductase enzymes.1 The accumulation of denitrification intermediates and the subsequent emission of NO or N2O to the atmosphere are highly undesirable. N2O is a potent greenhouse gas with a 300fold stronger radiative force than carbon dioxide, and is also a primary ozone-depleting substance in the 21st century.2,3 De Haas and Hartley showed that the carbon footprint of a typical biological nitrogen removal wastewater treatment plant would increase by approximately 30% when just 1% of denitrified nitrogen is emitted as N2O.4 Therefore, N2O accumulation during denitrification should be avoided. The accumulation of a certain denitrification intermediate is thought to result from unbalanced rates of the neighboring steps of denitrification. Some environmental conditions such as dissolved oxygen (DO), pH, and sulfide have previously been reported to play important roles in intermediate accumulation.5−11 It is likely that variations of these factors exert different effects on the enzymes responsible for different reduction steps. These different effects could be related to the © 2013 American Chemical Society
distinctive structures of various enzymes, or their different locations inside of a cell, both of which could lead to different sensitivities of the enzymes to environmental conditions. Either way, the unbalanced rates would be caused by changed enzymatic activities. There is also the possibility that the four denitrification steps could exert influence on each other through electron competition, thus leading to unbalanced rates. During denitrification, competition for electrons may occur among the four reduction steps when the electron supply rate from the oxidation process could not meet the demand for electrons by the four reduction steps.12 This competition was clearly demonstrated in our previous study with the use of a methanol-utilizing denitrifying sludge.13 It was found that, with electron supply being the limiting step, the culture did not allocate electrons according to the maximum turnover rates of the different nitrogen oxide reductases. Rather, the reduction of nitrite was prioritized over the other denitrification steps, consequently leading to N2O accumulation.13 The electron competition process likely plays an important role in N2O accumulation at low COD/N ratio conditions,14,15 or when polyhydroxybutyrate (PHB), an intracellular carbon Received: Revised: Accepted: Published: 11083
May 25, 2013 August 31, 2013 September 3, 2013 September 3, 2013 dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
tion.29,30 More recently, Hiatt and Grady,23 Kampschreur et al.,31 and Ni et al.18 proposed four-step denitrification models with the aim to predict the accumulation of all denitrification intermediates. However, in all these multistep denitrification models, as in the single-step model, the reduction of a nitrogen oxide compound and the oxidation of organic carbon are directly coupled in a single oxidation−reduction reaction with a stoichiometric relationship obtained through electron balance. In this way, these models ignore the fact that the nitrogen oxides reduction and carbon oxidation are carried out by different enzymes with their specific kinetics, and consequently either of the two processes could limit the rate of denitrification. In addition, this coupling approach describes each denitrification step independently with its rate not being affected by the other denitrification steps that draw electrons from the same electron supply. Such a structure defect in the previous denitrification models limits their ability to model electron competition during denitrification. Further, although the reaction rates of all reduction processes are considered to be dependent on the carbon concentration through the use of a Monod kinetic term with respect to carbon (i.e., ((SS)/(KS + SS))), the possibility of electron production or transfer being a limiting step in the coupled oxidation− reduction reaction is not recognized or modeled. Indeed, the carbon oxidation rate is modeled as the sum of the carbon requirements by all denitrification steps, with the underlying assumption that electron supply will always be able to meet the predicted total electron demand. 2.2. Development of a New Denitrification Model Incorporating Electron Competition. The model developed in this work decouples the carbon oxidation and the nitrogen oxide reduction processes. Electron carriers are introduced as a new component in this model to link carbon oxidation to nitrogen oxides reduction. Mred (reduced mediator) and Mox (oxidized mediator), defined as the reduced and oxidized forms of electron carriers,32 respectively, are included in the model as two new state variables. Because the electron carrier pool (with the NADH/ NAD+ pool as an example) is relatively small compared with its turnover rate, the continued availability of Mox for carbon oxidation and Mred for nitrogen oxides reduction relies on their concomitant regeneration.33,34 In modeling, this recirculation loop could be reflected by an increase in Mred being balanced by a decrease in the Mox and vice versa (Mred ⇋ Mox + 2e− + 2H+), with the total level of Mred and Mrox being held constant, i.e., SMred + SMox= Ctot,35,36 where Ctot is the total concentration of electron carriers. Carbon Oxidation Process. Because the experimental data to be used in this work for model calibration and validation were previously obtained by use of a methanol-utilizing denitrifying culture (as detailed in Pan et al.13), methanol is used here as a model carbon source. The oxidation of methanol is modeled by the following two processes (R1 and R2), describing the catabolic and anabolic methanol oxidation reactions, respectively. C5H7O2N is used to describe the biomass formed.37 Mox would receive electrons produced by carbon oxidation and be reduced to Mred: Methanol conversion to CO2:
storage compound with a relatively low biodegradation rate, is used as an electron donor for denitrification.16 These conditions could lower the carbon oxidation rate and hence the electron supply rate. In addition, electron competition could partially explain the accumulation of intermediates thought to be induced by other environmental conditions. For example, it is known that low pH causes a stronger inhibitory effect on N2O reduction than on the other steps of denitrification,12 thus leading to N2O accumulation.10 However, Pan et al.9 observed, with an enriched methanol-utilizing denitrifying culture, that at a low pH of 6.5 N2O accumulation occurred during nitrate reduction under endogenous denitrification conditions (no methanol supply) but not under exogenous denitrification conditions (with methanol supply). This suggests that the typically observed pH effect on N2O accumulation may be partially related to electron competition. Mathematical models are widely applied to predict nitrogen removal during wastewater treatment.17 These models are gaining more attention recently for the prediction of N2O accumulation and emission during both nitrification and denitrification. Several models have been proposed for the prediction of N2O accumulation during nitrification.18−21 Ni et al.22 successfully applied one of these models to describe N2O emissions from two full-scale wastewater treatment plants. Models for predicting N2O accumulation during heterotrophic denitrification, for example the Activated Sludge Model for Nitrogen (ASMN) by Hiatt and Grady,23 have also been proposed. However, these models do not consider electron competition among different steps of denitrification, and have previously been shown to be unable to predict N2O accumulation under carbon-liming conditions.9 This study aims to develop a new denitrification model that takes electron competition among different steps of denitrification into account, and can be used as a practical tool for predicting N2O accumulation during denitrification in wastewater treatment. To this end, the complex biochemical reactions and electron transfer processes involved in biological denitrification are lumped into one oxidation and four reduction reactions that are linked through electron carriers. Electron competition among the four steps of denitrification is modeled through the use of different affinity constants with respect to reduced electron carriers for different enzymes. The model is demonstrated using experimental data previously reported for a methanol-utilizing denitrifying culture.
2. MODEL DEVELOPMENT 2.1. Existing Denitrification Models. The Activated Sludge Models (ASM) Nos. 1, 2, and 3,17 published by the International Water Association, are widely accepted by both the scientific community and practitioners for the modeling of wastewater treatment processes. However, these models were developed with the aim to describe the removal of organic carbon and nitrogen, and for the removal of phosphorus in the case of ASM2, in activated sludge wastewater treatment. Denitrification was modeled as a single-step process, with nitrate being directly reduced to N2. To predict denitrification intermediates accumulation, denitrification needs to be modeled as a multiple-step process. Two-step denitrification models, with nitrite as a denitrification intermediate, have been proposed to predict the accumulation of nitrite.24−28 Three-step denitrification models, with nitrite and N2O as denitrification intermediates, have also been proposed with the purpose of predicting N2O accumula-
CH3OH + 3Mox + H 2O → CO2 + 3Mred
(R1)
Methanol assimilation to biomass: 11084
dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
Table 1. Process Matrix for the Developed Model SS
SMox
SMred
1-R1
−1
−3
3
2-R2
−1
−1
1
1
−1
⎞ ⎛ ⎞ ⎛ S SMred ⎟⎟ ⎟ × ⎜⎜ R3 = rNO3, max × X × ⎜ HB NO3 ⎝ KNO3 + SNO3 ⎠ ⎝ KMred ,1 + SMred ⎠
1/2
−1/2
⎞ ⎞ ⎛ ⎛ SMred S ⎟⎟ ⎟ × ⎜⎜ R 4 = rNO2, max × X × ⎜ HB NO2 ⎝ KNO2 + SNO2 ⎠ ⎝ KMred ,2 + SMred ⎠
1/2
1/2
−1/2
⎞ ⎞ ⎛ ⎛ SMred S ⎟⎟ ⎟ × ⎜⎜ R 5 = rNO , max × X × ⎜ HB NO K + S K + S ⎝ NO ⎝ Mred ,3 Mred ⎠ NO ⎠
−1
1
−1
⎞ ⎞ ⎛ ⎛ SMred S ⎟⎟ ⎟ × ⎜⎜ R 6 = rN 2O , max × X × ⎜ HB N 2O ⎝ KN 2O + SN 2O ⎠ ⎝ KMred ,4 + SMred ⎠
process
SNO3
−1
3-R3
SNO2
SN2O
1
−1
4-R4
SNO
1
−1
5-R5
6-R6
X
kinetic rate expressions
⎛ SS ⎞ ⎛ ⎞ SMox R1 = rCOD , max × X × (1 − Y ) × ⎜ ⎟×⎜ ⎟ ⎝ KS + SS ⎠ ⎝ KMox + SMox ⎠ 1/5
7
⎛ SS ⎞ ⎛ ⎞ SMox R 2 = rCOD , max × X × Y × ⎜ ⎟×⎜ ⎟ + + K S K S ⎝ Mox ⎝ S S⎠ Mox ⎠
SMred + SMox = Ctot
CH3OH + Mox + →
expressed on an electron basis.38 Thus, a single yield coefficient is used in the model, in line with all previous multistep denitrification models.18,23 It should be noted that Mred and Mox are two lumped parameters used in the model. In reality, the oxidation of methanol to carbon dioxide and to biomass (biomass is less reduced than methanol) will deliver electrons to several electron carriers, including pyrroloquinoline quinone (PQQ), nicotinamide adeninedinucleotide (NAD),36 and hydrogen.39 Many other electron carriers, including various cytochromes,40 are also subsequently involved in the transfer of the electrons to nitrogen oxides as the terminal electron acceptors. The use of such lumped parameters reduces the complexity of the model, making the implementation, application, and comprehension of the model easier.
1 NH3 5
3 1 C5H 7O2 N + Mred + H 2O 5 5
(R2)
Nitrogen Reduction Processes. Four sequential nitrogen oxides reduction steps are described in R3−R6. Mred donates two electrons to a nitrogen oxide (nitrate, nitrite, NO, and N2O) and is reoxidized to Mox: NO3− reduction to NO2−, mediated by nitrate reductases: NO3− + Mred → NO2− + Mox + H 2O
NO2−
(R3)
reduction NO, mediated by nitrite reductases:
NO2− +
1 1 Mred + H+ → NO + Mox + H 2O 2 2
(R4)
NO reduction to N2O, mediated by NO reductases: 1 1 1 1 NO + Mred → N2O + Mox + H 2O 2 2 2 2
3. MODEL CALIBRATION AND VALIDATION 3.1. Experimental Data for Model Evaluation. Experimental data previously reported in Pan et al.13 are used for the model calibration and validation. Pan et al.13 used a methanoland nitrate-fed denitrifying culture, developed in an 8-L labscale sequencing batch reactor, to study the electron competition during denitrification. Extensive batch and fedbatch tests (35 in total) were performed in a 300-mL sealed reactor under anaerobic conditions. Methanol and various nitrogen oxides were supplied to the mixed liquor in each test. Two methods were employed to supply methanol to the batch/ fed-batch test reactor: (1) pulse feeding at the beginning of the test to a non-rate-limiting concentration (6.25 mmol/L); (2) controlled slow and continuous feeding throughout the test, with four different loading rates (6.3, 3.9, 1.1, 0.7 mmol/(gVSS × h)). Thus, there were five methanol-supplying schemes in total. Within each methanol-supplying scheme, seven different nitrogen oxides addition methods were applied (all added at the beginning of the test to non-rate-limiting concentrations of 0.7−2.5 mmol/L), namely the addition of a single nitrogen oxide: (1) NO3−, (2) NO2−, (3) N2O; or the simultaneous addition of two or three nitrogen oxides: (4) NO3− and NO2−, (5) NO3− and N2O, (6) NO2− and N2O, and (7) NO3−, NO2−, and N2O. The detailed experimental setup, experimental
(R5)
N2O reduction to N2, mediated by N2O reductases: N2O + Mred → N2 + Mox + H 2O
(R6)
The above reactions are summarized in Table 1, which also contains the kinetic expression proposed for each of the reactions. Model components and model parameters are defined and explained in Tables S1 and S2, respectively, in the Supporting Information (SI). By decoupling the carbon oxidation (R1 and R2) and nitrogen oxides reduction processes (R3−R6) through electron carriers, the carbon oxidation and nitrogen reduction processes are modeled separately. The possibility of the carbon oxidation or electron transfer being a limiting step in denitrification is thus considered in the model. The electron competition process among the four denitrification steps can be modeled through assigning different values to the affinity constants of the enzymes responsible for Processes R3, R4, R5, and R6 with respect to Mred, which are provided by Processes R1 and R2. Although different reduction steps are associated with different free energy changes, the maximum growth yields achieved with different nitrogen oxides are almost identical if 11085
dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
Table 2. Parameter Values Used for the Denitrification Model variable
values
rCOD,max rNO3,max rNO2,max rNO,max rN2O,max KS KHB NO3 KHB NO2 KHB NO
8.46 3.99 5.27 50 20 0.1 1.79 × 10−3 4.13 × 10−3 1.07 × 10−5
unit mmol/(gVSS mmol/(gVSS mmol/(gVSS mmol/(gVSS mmol/(gVSS mmol/L mmol/L mmol/L mmol/L
source × × × × ×
h) h) h) h) h)
variable
values
unit
KHB N2O KMox KMred,1 KMred,2 KMred,3 KMred,4 Ctot Y
2.5 × 10−2 1 × 10−2 × Ctot 4.58 × 10−3 3.93 × 10−4 1 × 10−3 × Ctot 3.23 × 10−3 1 × 10−2 0.5
mmol/L mmol/(gVSS)
a a a b b
17 17 17 17
source 8c b a a b a b
18, 23, 48
a c
Parameter values estimated in this work through model calibration. bParameter values assumed based on literature information, detailed in text. Parameter values determined previously for the same culture.
rNO,max and a low value of KMred,3 are used in this model (Table 2) to ensure there is no accumulation of NO. Ctot and KMox. As Mred and Mox are lumped parameters representing reduced and oxidized forms of electron carriers, respectively, the parameters directly related to Mred and Mox, including Ctot, KMox, are all lumped parameters that are not directly measurable. Ctot was assumed to be 0.01 mmol/(gVSS) based on literature-reported values for intracellular NAD + NADH concentrations.35,41 KMox was assumed to be 1% of Ctot to ensure that the reduction of Mox would not be the rate limiting step during carbon oxidation. With the above assumptions, six parameters, namely rNO3,max, rNO2,max, rCOD,max, KMred,1, KMred,2, and KMred,4, are calibrated using experimental data. Experimental data from the 7 batch tests with methanol pulse feeding and another 7 fed-batch tests with a methanol loading rate of 1.1 mmol/(gVSS × h)13 were used to calibrate the model. Parameter values were estimated by minimizing the sum of squares of the deviations between the measured data (NO3−, NO2−, and N2O) and the model predictions for all the 14 tests used for model calibration. Parameter estimation were performed with AQUASIM for aquatic systems.42 The model was then run with these best-fit parameter values under conditions for the remaining 21 fedbatch tests to validate the model and the obtained parameter values through assessing the fit between model predictions and experimental data. The values of the estimated parameters are listed in Table 2. The parameter correlation matrix given in Table S3 (SI) shows that the correlations among most of the estimated parameters are low.43−45 This indicates these parameters in general have good identifiability. Parameters combinations of KMred,1 and rNO3,max, KMred,2 and KMred,4, rCOD,max and rNO2,max, rCOD,max and rNO3,max, and rNO2,max and rNO3,max though have correlation coefficients greater than 0.8. Therefore, their joint 95% confidence regions were investigated to further evaluate their identifiability. As shown in Figure S1 (SI), the 95% confidence regions for the all the pairs are small, with mean values lying at the center. The 95% confidence intervals are also small. These indicate that the estimated values for these parameters have a high-level of certainty. The estimated rNO3,max, rNO2,max, and rCOD,max values are generally comparable to previously reported values.1,22 The estimated values for KMred,1, KMred,2 and KMred,4 represent the affinity of the corresponding denitrification step to Mred, with lower values indicating a higher affinity and thus a higher ability to compete for electrons. The estimated KMred,2has a value that is about ten times smaller than KMred,1 and KMred,4, indicating
results, and calculation of the electron consumption rates by each step of denitrification can be found in Pan et al.13 On the basis of their experimental results, Pan et al.13 revealed that electron competition occurs under not only carbon-limiting, but also carbon-abundant, conditions. The electron distribution among the nitrogen oxide reductases is affected by the carbon loading rate, with a lower fraction of electrons distributed to the N2O reductase with a reduced carbon loading rate. N2O accumulation occurs when the electron flux going to nitrite reduction is higher than that going to N2O reduction. 3.2. Parameter Estimation and Model Validation. The new denitrification model includes 17 stoichiometric and kinetic parameters in total, as summarized in Table 2. Six of HB HB HB these parameters (i.e., KS, KHB NO3, KNO2, KNO, KN2O, and Y) are well established by previous studies. Thus, literature values were adopted for these parameters.17,22 Preliminary data analysis reveals that five of the parameters (i.e., rNO,max, rN2O,max, KMox, KMred,3, Ctot) are not identifiable from the data set available or easily measurable through additional experiments. For these parameters, we took the approach of assuming their values, as explained below, and then analyzing the sensitivity of the other calibrated parameters and the fit between model predictions and experimental data to the variation of these values (Section 4.2). The remaining six parameters, which are unique to the proposed model (i.e., rNO3,max, rNO2,max, rCOD,max, KMred,1, KMred,2, and KMred,4), are then calibrated. Maximum Nos-Mediated Reaction Rate (rN2O,max). The maximum Nos-mediated reaction rate (rN2O,max) is assumed to be 1.5 times of the experimentally determined maximum N2O reduction rate when both carbon and nitrate were in excess. This is because previously we revealed that the experimentally measured maximum Nos-mediated reaction rate was limited by the electron supply rate rather than by the Nos activity.13 Hence, the true maximum activity of the N2O reductase (i.e., rN2O,max) should be higher than the experimentally determined maximum N2O reduction rate. A factor of 1.5 was chosen arbitrarily, and the impact of this was further analyzed through sensitivity analysis. NO Reduction Related Parameters (rNO,max and KMred,3). The maximum Nor-mediated reaction rate (rNO,max) was also beyond the ability of measurement since NO was not added in any tests given its toxicity to bacteria.1 However, NO accumulation was not detected in any of the tests, suggesting that NO reduction was faster than other steps. Indeed, other studies suggest that NO reduction is usually prioritized by bacteria to avoid its toxicity,1,12 thus a relatively high value of 11086
dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
that nitrite reduction has a higher ability to compete for electrons under electron-limited conditions. The fit between experimentally measured and modelsimulated nitrate, nitrite, and N2O profiles resulting from model calibration (Figure 1, and Figure S2 in SI) and model
Figure 2. Experimental and simulated reduction rates of NO3−, NO2− or N2O in tests with methanol pulse feeding, with 7 nitrogen oxides addition schemes that include: (1) NO3−, (2) NO2−, (3) N2O, (4) NO3− and NO2−, (5) NO3− and N2O, (6) NO2− and N2O, (7) NO3−, NO2−, and N2O. The reduction rate calculation method is detailed in Pan et al.13 This figure corresponds to Figure 2a in Pan et al.13
electron competition (as detailed in Pan et al.13). Pan et al.13 further observed that the fractions of electrons distributed to N2O reductase decreased with the reduced carbon loading rate when NO2− (Figure 3c) or NO3− and NO2− (Figure 3d) were present simultaneously with N2O. The model reproduced these trends very well. Further, N2O accumulation was observed with the lowest carbon loading rate of 0.7 mmol/(gVSS × h), which was also correctly predicted by the model (SI Figure S5). These results indicate that the proposed denitrification model is able to describe the electron competition among different steps of denitrification.
Figure 1. Fits between experimental and simulated NO3−, NO2−, and N2O profiles achieved in model calibration: tests with methanol pulse feeding to a nonrate limiting concentration as an example. Nitrogen oxides addition scheme: (a) NO3−, (b) NO2−, (c) N2O, (d) NO3− and NO2−, (e) NO3− and N2O, (f) NO2− and N2O, and (g) NO3−, NO2−, and N2O. No accumulation of denitrification intermediates were observed in these tests. This figure corresponds to Figure 1 in Pan et al.13
4. DISCUSSION 4.1. Modeling Electron Competition and N2O Accumulation during Denitrification. The key difference between the model developed in this paper and the previous denitrification models is that the proposed model links carbon oxidation and nitrogen reduction processes through a pool of electron carriers, while the previous models directly couple the two processes. The proposed model recognizes the fact that the carbon oxidation process and the nitrogen oxides reduction processes are carried out by different components and enzymes, and should be modeled with different kinetics.12 The calibration and validation results confirmed the ability of the model to predict electron competition during denitrification. The coupled denitrification models would not be able to predict the electron competition process. Indeed, unsatisfactory results were obtained when we applied the four-step denitrification model by Hiatt and Grady22 to describe the experimental results. In the case that both the carbon source and the nitrogen oxides concentrations were in excess, the predicted nitrate, nitrite, and N2O reduction proceeded at their respective highest rates, independent of the presence or absence of other electron acceptors (SI Figure S6). Such predictions are contradictory with the experimental data, which showed that the nitrogen oxides have effects on each other’s reduction rate. This is not surprising given the fact that the electron transfer process as a rate-limiting step is not reflected in the model. Indeed, in previous multistep denitrification models, the maximum reduction rate of each denitrification
validation (Figures S3−S5 in SI) show that the model predictions match the measured data reasonably well in all 35 tests. Furthermore, the model predictions show no systematic deviations from experimental data. These results indicate that the proposed model is able to describe the four-step denitrification process under a wide range of carbon-loading and electron acceptor availability conditions. Pan et al.13 reported that, when multiple nitrogen oxides were present, electron competition leads to reduced reduction rates of all nitrogen oxides in comparison to the rates measured when a single nitrogen oxide was present. This was also well predicted by the model, as the predicted nitrogen oxide reduction rates, when electron donor is in excess, are very close to the experimentally determined results (Figure 2). Importantly, the highest nitrate, nitrite, or N2O reduction rates are always achieved only when the respective nitrogen oxide is present as the single electron acceptor. Pan et al.13 further observed that the electron distribution was affected by the intensity of electron competition. Figure 3 shows that the model-predicted electron distribution pattern reproduced the experimentally determined results, when two or more nitrogen oxide species were present. When the methanol addition scheme changed from pulse feeding to continuous loading at rates of 6.3, 3.9, 1.1, 0.7 mmol/(gVSS × h) (methanol schemes 1−5 in Figure 3), the reduced carbon supply rate affected the electron supply rate and thus intensified 11087
dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
Figure 3. Experimental and simulated electron distribution with two or three nitrogen oxides added in each test. Methanol addition schemes 1, 2, 3, 4, and 5 stand for methanol pulse feeding and methanol slow feeding with loading rates of 6.3, 3.9, 1.1, 0.7 mmol/(gVSS × h)), respectively. (a) NO3− and NO2− were added; (b) NO3− and N2O were added; (c) NO2− and N2O were added; (d) NO3−, NO2−, and N2O were added. This figure corresponds to Figure 5 in Pan et al.13
introduced to the model proposed in this work. These include (1) rCOD,max, the maximum carbon oxidation rate, which is ignored in ASMN; (2) KS, the affinity constant of carbon oxidation with respect to the carbon source concentration, which replaces the four parameters (KS1, KS2, KS3, and KS4) used in ASMN; (3) four parameters KMred,1, KMred,2, KMred,3, and KMred,4, which are the key kinetic parameters governing electron distribution in this model as discussed in Sections 4.1; and 4) C tot and K Mox , to describe the total electron carrier concentration and the affinity constant of carbon oxidation with respect to oxidized form of electron carrier, respectively. With the assumptions made in Section 3.1 and the parameter values taken directly from literature as listed in Table 2, the remaining parameters are identifiable, as evidenced by the correlation matrix and the subsequent analysis of the 95% confidence regions and confidence intervals (SI Figure S1). The impact of the assumed parameter values (made in Section 3.1) and the parameters chosen from literature on the estimates of other parameters and on the fit between the experimentally measured and model predicted NO3−, NO2−, and N2O profiles were analyzed through sensitivity analysis, with the findings summarized below. NO Reduction Related Parameters. Sensitivity analysis suggests that the estimated values of other parameters and fits between model predictions and experimental data are not sensitive to rNO,max and KMred,3, as long as the chosen values do not lead to NO accumulation. As NO accumulation during denitrification has never been reported,1 it is reasonable to choose an rNO,max value that is far higher (e.g., 10 times) than the largest value of rNO3,max, rNO2,max, and rN2O,max, together with a KMred,3 value that is far lower than the lowest value (e.g., one tenth) of KMred,1, KMred,2, and KMred,4.
step is actually a lumped term of the maximum nitrogen compound reduction rate and the maximum carbon oxidation rate.18,23 In contrast, in the proposed model, Processes 1 and 2 model the electron supply processes while Processes 3, 4, 5, and 6 describe the electron consumption processes. The two types of processes are linked through the electron carriers. The relative electron consumption rates of the four denitrification steps determine the electron distribution. While all parameters involved in the nitrogen reduction steps (R3, R4, R5 and R6) would affect the electron distribution, the affinity constants with respect to electrons (KMred,1,KMred,2, KMred,3, and KMred,4), introduced into the denitrification model for the first time in this study, are the key kinetic parameters governing electron distribution. When the electron supply is rate limiting, a low SMred concentration would occur, which would subsequently limit all the four reduction processes. The different values for the four affinity constants with respect to electrons (KMred,1, KMred,2, KMred,3, and KMred,4) determine the competitiveness of different reduction steps for electrons when SMred is low. As discussed above, nitrite reductase seems to have a higher ability to compete for electrons in comparison to nitrate and N2O reductases. These estimated parameter values enable the model to reproduce the experimental observation that the fraction of electrons distributed to nitrite reductase increased as the intensity of electron competition increased (Figure 3). For example, the fraction of electrons distributed to nitrite reductase increased from 13% to 29%, and from 13% to 30%, respectively, as the intensity of electron competition increased with the decrease in carbon loading rate (Figure 3c and d). 4.2. Complexity and Identifiability of the Proposed Model. In comparison to ASMN,22 several new parameters are 11088
dx.doi.org/10.1021/es402348n | Environ. Sci. Technol. 2013, 47, 11083−11091
Environmental Science & Technology
Article
tion is also identified as an important source of N2O.19,46,47 Modeling of the latter processes to predict N2O production by nitrifiers is attracting considerable attention in the past few years, with several models proposed already.18,20−22 The integration of the model proposed in this work with a nitrifier-N2O model can serve as a powerful tool to estimate the overall N2O emission factor and to explore the effect of operational conditions on N2O dynamics in wastewater treatment systems. The model developed in this paper has so far been calibrated and validated with experimental data obtained from a mixed culture in a laboratory system receiving methanol as the sole electron donor. Calibration and validation using other cultures receiving different carbon sources including sludge from fullscale wastewater treatment systems are still needed for the model to be developed into a useful tool for practical applications.
Ctot. Sensitivity analysis shows that if Ctot is increased by 100%, the estimated values of KMred,1, KMred,2, and KMred,4 would also increase by around 100%. In other words, the ratios between these parameters and Ctot do not change. Examination of the kinetic expressions indicates that relative ratios between KMred,1, KMred,2, KMred,3, and KMred,4 rather than their absolute values are important for the reaction rate. Indeed, the estimated values of other parameters (rNO3,max,rNO2,max,rCOD,max) almost remain unchanged (
