Model-based evaluation of the contribution of partial denitrication (PD) on the one-stage autotrophic nitrogen removal process

The nitrate produced by the one-stage partial nitritation-anammox (PN/A) process can be removed through partial denitrication (PD) by adding carbon source. In this study, a 1D multi-population biolm model was developed to evaluate the contribution of partial denitrication on the one-stage autotrophic nitrogen removal process at inuent NH 4+ = 100 mg N/L. The dynamic simulation that was carried out to investigate the effect of nitrite-oxidizing bacteria (NOB) revealed that PD contributed to the reactor to obtain total nitrogen removal eciency (TNR) of above 90% and the euent nitrate was signicantly decreased with the absence of NOB. However, PD decreased TNR of the one-stage PN/A process with the presence of NOB. Increased inuent chemical oxygen demand (COD) widened the dissolved oxygen (DO) range required for high TNR whether NOB were present or not. The steady-state simulation results showed that NOB were always absent in the granules at high DO and COD levels and the optimum DO > 0.5 mg/L when inuent COD was over 50 mg/L. Besides, higher inuent COD/NH 4+ (C/N) and larger granule diameter (diameter > 1600 µm) were contributed to widening the range of DO required for high TNR. The nitrogen removal contribution of anammox bacteria (AMX) was signicantly higher than denitrication in the reactor.


Introduction
Traditional nitri cation and denitri cation method needs high aeration and COD consumption, so how to remove nitrogen with high e ciency and low consumption has always been one of the key issues in wastewater treatment. Mulder et al. (1995) discovered the anammox process, Strous et al. (1999) found that the anammox process can effectively remove ammonium and nitrite in wastewater with high ammonia nitrogen (such as sludge digestion e uent) through experimental research and consume low energy, so it is a sustainable wastewater nitrogen removal technology (Kartal et al. 2010). The world's rst full-scale anammox reactor with granular sludge was built in Dokhaven wastewater treatment plant (WWTP) in Rotterdam, the Netherlands (van Loosdrecht et al. 2004), marking the o cial commissioning of anammox into operation. The anammox process is shown in Eqs. 1 (Strous et al. 1999). Due to anammox process can be in uenced by completely nitri cation resulting in poor TNR (Pérez et al. 2014), the one-stage PN/A process which is used to achieve partial nitri cation has been proposed to improve autotrophic nitrogen removal (Ganigue et al. 2007).
One of the de ciencies of the one-stage PN/A process is that the e uent nitrate that is produced by AMX cannot be treated. The e uent nitrate can be removed by denitri cation bacteria through the PD process, so TNR can be removed up to 100% theoretically . Thus, controlling the nitrate only reduction to nitrite is considered to be a new technique called partial denitri cation (PD) or short-cut denitri cation and is arguably one of the most promising technologies for future nitrogen removal processes (Shi et al. 2019;Xu et al. 2020;Zhang et al. 2020). Eqs. 2 is a PD process in which acetate is used as a carbon source (Du et al. 2017a) by controlling the reduction reaction of NO 3 in the rst stage.
However, most experiments on the PD process were based on reactors such as SBR without aeration (Du et al. 2015;Wang et al. 2019) or UASB (Du et al. 2017b;Xu et al. 2020) and the contribution of PD on the one-stage autotrophic reactor with the presence of NOB has hardly been reported. Besides, one of the shortcomings of autotrophic nitrogen removal is to treat in uent COD (Sliekers et al. 2002) because of the high sensitivity of AMX activity to COD (Chamchoi et al. 2008;Jin et al. 2012), thus, the anammox process is only considered to be suitable for removing wastewater with no in uent organic carbon (Jin et al. 2012;Wang et al. 2016). To solve this problem, anammox coupled with the heterotrophic nitrogen removal process is proposed. For example, Hao and van Loosdrecht (2004) used a model that can simulate the activities of heterotrophic and autotrophic biomass in bio lms and relate reactions such as COD oxidation, denitri cation, nitri cation, and anaerobic ammonia oxidation to evaluate the heterotrophic and autotrophic nitrogen removal. When the in uent COD is too high, will the proliferation of denitrifying bacteria affect AMX activity? Therefore, two types of pressing problems remain about current research on the PD process. First, whether the PD process has an impact on AMX activity and the NO 3 generated by AMX can be treated effectively. Second, whether the presence of NOB enables the PD process to help to improve TNR in the one-stage autotrophic reactor.
In this study, a 1D multi-population bio lm model was developed for a one-stage autotrophic reactor to evaluate the nitrogen removal performance of the PD process at high DO levels (presence of NOB) versus low DO levels (absence of NOB) and the main TN removal contributor in each of the conditions. The simulation results enabled the PD process to operate more reliably under a variety of different in uent conditions in a one-stage autotrophic reactor.

Granular sludge reactor model
Stoichiometric matrix, kinetic expressions and model parameter values The nitrogen removal process in the one-stage autotrophic granular sludge reactor was simulated by a 1D multi-population bio lm model (Wanner and Reichert 1996). The stoichiometric matrix, kinetic expressions, and model parameter values are summarized in the Supplementary Material (Tables S1-S3). Five particulate components (ammonium-oxidizing bacteria (AOB), NOB, heterotrophic bacteria (HET), AMX, and inert biomass (INT)), and ve soluble components (soluble biodegradable COD (SS), DO, NH 4 + , NO 2 -, and NO 3 -) were included in the model. Nitri cation was described as a two-step process: oxidation of NH 4 + to NO 2 by AOB, followed by NO 2 oxidation to NO 3 by NOB. Heterotrophic growth reactions were included using DO, NO 2 or NO 3 as an electron acceptor. As in most denitri cation models, denitri cation was divided into two separate reactions in which both nitrate and nitrite are involved, as anammox bacteria competed with denitri cation bacteria for nitrite (Sin et al. 2008). Therefore, the model consisted of heterotrophic and three types of autotrophic bacteria in the granules. The competitions for oxygen, ammonium, nitrite, and nitrate between the three types of autotrophic organisms and the heterotrophic bacteria present were considered in the model. Detailed process stoichiometry and rate expressions are listed in Tables S1 and S2 of the Supplementary Material.

Simulation parameters and initial conditions of the reactor
The model only considering radial gradients in spherical biomass granules was set up to describe the autotrophic and heterotrophic interaction. Spherical granules were grown from an initial diameter of 7.5 × 10 -4 μm to a prede ned steady-state granule diameter (750 μm by default, and then varied from 400 to 2000μm to examine the effect of granule size) and were equally divided into 31 layers to simulate the distribution of substrates and microorganisms. The oxygen levels were assumed from 0.1 to 5 mg/L to evaluate the performance of the one-stage reactor at different in uent COD levels, while the temperature was set at 20 ℃. The in uent ammonium concentration was xed at 100 mg N/L at any simulation with no nitrite or nitrate presenting in the in uent and the initial mixed liquor volatile suspended solids (MLVSS) was assumed at 3000 mg/L in the granules. The in uent COD concentration was varied from 0 to 600 mg/L to evaluate the effect of organic carbon on the reactor performance. The density in the granules was set to 93333 g COD m -3 . The bulk liquid was assumed to be well mixed, and the external mass transfer limitations were neglected to simplify the evaluation. The hydraulic retention time (HRT) was set to 1/4 d and the sludge retention time (SRT) was equal to HRT. Besides, the porosity of the granules was chosen to be 80% and initial biomass fractions in the bulk liquid were all 0.01.
The model simulation time was set to 5000 days to assure that a steady-state has been reached in the granules. Besides, to evaluate the dynamic simulations of the granules with the presence and absence of NOB, we rstly simulated biomass fractions in the steady-state granules at different COD levels (NOB was hardly present in the granules). Then, we chose the steady-state biomass fractions as the initial conditions for dynamic simulations of the granules with the absence of NOB. Finally, the initial AOB, NOB, AMX, HET, and INT fractions in the granules were multiplied by 0.8, and then added 0.2 times of the initial NOB fraction of steady-state simulation to assess the effect of the contribution of the PD process on the reactor with the presence of NOB.

Results And Discussion
Contribution of PD to nitrogen removal with the presence and absence of NOB A series of dynamic simulations were performed for the in uent contained only ammonium and no organic carbon, and the in uent contained ammonium and organic carbon with the presence and absence of NOB.

Contribution of PD to nitrogen removal with NOB
The contribution of AMX and denitri cation to nitrogen removal under different in uent COD are shown in Fig. 1. It was clear that the PD process was not bene cial for nitrogen removal with the presence of NOB. Fig. 1a displays that the removal of nitrogen was entirely dependent on AMX and the optimum DO was 0.3 mg/L with maximum TNR at 80% when the in uent had no organic carbon. TNR was deteriorated at high DO levels. When the in uent COD was increased to 100 mg/L, Fig. 1b shows that the higher the in uent COD, the higher the optimum DO level (DO =0.7 mg/L) because both COD oxidation and nitri cation consumed oxygen and a higher oxygen ux towards the granule was needed at increasing COD. In addition, the metabolism of N could be divided into denitri cation and anammox pathways due to the presence of COD. The optimal TNR decreased to around 50% that meant the in uent COD leading to poor nitrogen removal when considering NOB presenting in the reactor. Due to a part of NO 2 was reduced to N 2 by one-stage denitri cation bacteria (DB) and a part of NO 2 was consumed by NOB, so a large accumulation of nitrite could not be achieved inside the granule ) and AMX activity became lower. Although the in uent COD was high, we found that AMX was the major contributor to nitrogen removal at any DO level, which was consistent with previous studies (Cui et al. 2017;Du et al. 2019). When the in uent COD was increased to 300 mg/L (Fig. 1c), the DO level required for optimal TNR had not changed, and the contribution of denitri cation was increased compared to the in uent COD = 100 mg/L situation, while the contribution of AMX was reduced. Furthermore, the increasing contribution of denitri cation was not a good phenomenon which meant that the short-cut denitri cation we expected to achieve was reduced, so we seemed not to achieve an e cient PD process in a one-stage reactor with the presence of NOB.
The above three scenarios were simulated under the assumption of MLVSS = 3000 mg/L in the granule. When we increased biomass and assumed MLVSS = 6000 mg/L under in uent COD = 300 mg/L, we found that TNR and the contribution of AMX increased much (Fig. 1d) compared to MLVSS = 3000 mg/L, indicating that the short-cut denitri cation process and the activity of AMX were enhanced when biomass in the granule was increased. The simulation result in Fig. 1d shows that after doubling biomass in the granule, the optimum DO was decreased from 0.7 mg/L to 0.5 mg/L. It can be inferred that the more MLVSS in the granule, the more bene cial to remove nitrogen under high in uent COD conditions.
The impact of COD on the PD process is important, which had been con rmed by large experiments (Du et al. 2014;Le et al. 2019;Ma et al. 2020). Comparing to the case of in uent COD = 100 mg/L (Fig. 1b), the higher in uent COD (Fig. 1c) resulted in higher heterotrophic biomass, which led to a higher contribution of denitri cation and was in line with the simulation results of Hubaux et al. (2015). Meanwhile, the simulation results also demonstrated that the contribution of denitri cation was 60% when in uent COD = 900 mg/L. Chamchoi et al. (2008) suggested that when in uent COD was more than 300 mg/L without aeration, AMX were completely inactivated and replaced by DB.
From Fig .1 we can derive that the presence of COD also promoted the DO range required for high TNR becoming broader. Hubaux et al. (2015) compared the contribution of AMX with that of denitrifying bacteria to nitrogen removal and they found that increasing COD appropriately led to an increase in the robustness of the granular sludge reactor to changes in DO.
A special situation aroused when the in uent COD was very low, as shown in Fig. 2. The simulation result demonstrated that TNR was higher in the case of low in uent COD when NOB was present, in agreement with the results of Mozumder et al. (2014). When the in uent COD was reduced to only 10 mg/L, we were surprised to nd that the PD process resulted in increasing the activity of AMX and optimal TNR compared to the absence of COD, suggesting that the PD process helped to improve nitrogen removal under very low COD concentration, even in the presence of NOB.
Therefore, the PD process was bad to improve nitrogen removal under high in uent COD conditions but slightly increasing TNR at very low COD levels with the presence of NOB. Moreover, although the PD process was not good for nitrogen removal at high COD levels, it widened the optimum DO range so that making operation conditions easier to achieve high TNR.
Contribution of PD to nitrogen removal without NOB As shown in Fig. 3, we analyzed the contribution of PD to nitrogen removal assuming NOB was absent and it was reassuring to discover that PD improved nitrogen removal under different in uent COD conditions. When the in uent had no organic carbon (Fig. 3a), the optimum DO was also at 0.3 mg/L compared to the presence of NOB, with the difference that the optimal TNR was higher than that with the presence of NOB. When COD = 100 mg/L (Fig. 3b), the optimal TNR was achieved a little increasing compared to no COD condition. Not only the optimum DO level (DO = 0.7 mg/L) had not been changed compared to the situation with the presence of NOB, but also the DO range for TNR above 80% was widened, indicating that the PD process helped to improve nitrogen removal with the absence of NOB. When COD was increased to 300 mg/L (Fig. 3c), the contribution of denitri cation was increased. Fig. 3c shows that although AMX activity was inhibited, the optimal TNR was not affected if we made the reactor at a high DO level (DO = 1.5 mg/L). After increasing MLVSS to 6000 mg/L, AMX activity was restored, and the optimal TNR was further increased at a lower DO level which was also similar to the simulation results with the presence of NOB.
Without considering NOB, the PD process not only helped to widen the optimum DO range but also increased the optimal TNR slightly. However, if the in uent COD was high and DO was low, HET utilized su cient COD for a complete denitri cation process and was therefore unable to achieve a signi cant accumulation of NO 2 in the granule, which was not conducive to increasing the activity of AMX. From the simulation results, we could conclude that the PD process could be implemented for e cient nitrogen removal with the absence of NOB even at high COD levels.
The e uent NO 3 concentration at different COD and DO levels were compared for the cases with and without NOB (Fig. 4). From the simulation results, when the in uent COD and the bulk oxygen was appropriately increased (Fig. 4b, 4c and 4d), the e uent NO 3 concentration was considerably reduced (almost no NO 3 in the e uent) compared to the no organic carbon in uent with the absence of NOB because the heterotrophic bacteria reduced the NO 3 produced by AMX to NO 2 through partial denitri cation, and this conclusion was similar to the result of the model simulations by Mozumder et al. (2014). However, when considering NOB, we found that the e uent NO 3 concentration was still high because a large amount of NOB produced much NO 3 at high DO levels. Hence, TNR was not satis ed when NOB was present. In conclusion, the PD process without NOB can solve the problem of a one-stage autotrophic reactor.

In uence of combined COD and DO on TNR and the analysis of microbial communities
The in uence of combined COD and DO on steady-state reactor performance is shown in Fig. 5. For a previous study, it was considered that more e cient coupling of PD to anammox would be to keep the DO below 0.7 mg/L, however, the author just considered the effect of DO on the reactor without COD in the wastewater (You et al. 2020). The simulation results in Fig. 5 showed that when the in uent COD > 50 mg/L, the optimum DO > 0.5 mg/L, which meant that once the in uent COD was increased, we could achieve high TNR at high DO levels. As mentioned previously, it could also be inferred from Fig. 5 that the DO levels corresponding to high TNR became wider at moderate in uent COD. Numerous experiments and model simulations have also demonstrated that increased COD could lead to signi cantly higher TNR and a wider DO range (Giustinianovich et al. 2016;Hubaux et al. 2015) at steady-state. As a result, we could better control the DO concentration to achieve maximum TNR. Previous studies of the two-stage reactors or one-stage reactors without aeration have been considered to be unsuitable for removing wastewater that contained residual COD, however, the simulation results of this model could give a good direction that even in the presence of COD in the wastewater, we could increase the concentration of DO so that most of the COD was consumed in the outer layer of the granule by aerobic heterotrophs (Fahmi et al. 2020) It is necessary to analyze the distribution of biomass in the granule that can be helpful for us to explain why the PD process improved nitrogen removal at steady-state. There have been many previous documents on the analysis of microbial communities of heterotrophic and autotrophic bacteria in granule/bio lm in the one-stage or two-stage reactors Shi et al. 2019;Xu et al. 2020), for the PD process, we wished to be able to enrich Thauera (denitrifying heterotrophic bacteria with NO 3 as an electron acceptor) in the granule/bio lm to achieve high TNR. So, the reason why PD process is more e cient and economical than conventional nitrogen removal is mainly due to the coexistence of a variety of micro-organisms which results in a greater accumulation of NO 2 -. As autotrophs and heterotrophs can coexist in the granule, this means that it is important to study both COD and DO parameters to balance them to achieve a satisfactory TNR. Fig. 6 gives us a good overview of the simulation results of autotrophic and heterotrophic bacteria proportion at different COD and DO levels. Aeration in the onestage reactor resulted in the presence of AOB and NOB in the outer layer of the granule and it has been demonstrated that the sequence of biological reactions in the granule/bio lm started with the aerobic oxidation of COD followed by the nitri cation process of AOB Zhou et al. 2018). Fig. 6a showed that AOB steady-state fraction had a lower share in the granule at high in uent COD levels. This was because the aerobic heterotrophic bacteria located in the outermost layer of the granule had signi cantly higher biomass and AOB were outcompeted by them at high COD levels (Fig. 7b). From Fig.  6a and Fig. 6b, even if there was su cient DO at high COD levels, AOB and NOB steady-state fraction were low because they were competed by HET for the spatial position. The result from Fig. 6b was the same as the simulation of Mozumder et al. (2014) in that in the presence of HET and COD, the steadystate NOB fraction was lower and only became progressively higher when the wastewater was almost free of COD and DO > 0.6 mg/L. When in uent COD > 100 mg/L, NOB were always absent within the granule regardless of the DO levels at steady-state. Therefore, when the reactor was in steady-state, we did not need to be concerned that the presence of NOB would have a negative effect on nitrogen removal at high COD levels. Based on the conclusion in the previous section, we could conclude that the PD process helped to improve nitrogen removal in a steady-state reactor even at high COD levels.
We can think of the heterotrophic bacteria in the granule as mainly of two types as aeration heterotrophs and anoxic heterotrophs. Aerobic heterotrophic bacteria are responsible for degrading most of the COD in the outermost layer of the granule (Zhou et al. 2018), allowing denitrifying bacteria (DB) located in the slightly inner anoxic zone of the granule, to carry out a partial denitri cation process with NO 3 as the electron acceptor. We could nd that at high DO and COD levels (Fig. 6c), a high proportion of heterotrophic bacteria (50%) could be found in the granule, which were mainly aerobic heterotrophs, not anoxic heterotrophs . The result of Fig. 6d showed that steady-state AMX fraction maintained high at low DO levels. If the in uent COD was increased appropriately, then the DO levels required for higher steady-state AMX fraction became wider that corresponding to the results in the previous section. When DO > 0.5 mg/L and COD > 50 mg/L, TNR could achieve above 90%, however, the steady-state AMX fraction was not the highest under these conditions. This was suggested that AMX was not most active when the steady-state fraction was the highest. In general, we should control the DO concentration so that the three types of functional bacteria, AOB, HET, and AMX, can better co-exist within the granule and achieve optimal TNR at different COD levels.
The position of different microbial communities in the granule is determined by their competition for substrates and space. The previous study has been demonstrated that AMX, HET, and AOB can coexist in the granule ). It has been found that AOB are usually located on the outside of the granule and AMX are located deeper in the granule in terms of pro le biomass distribution (Mozumder et al. 2014), whereas aerobic heterotrophs are located on the outermost surface of the granule, DB are located in the anoxic zone of the granule (Zhou et al. 2018). Typical steady-state biomass distribution pro les without considering NOB in the model are displayed in Fig. 7. When in uent COD was 10 mg/L, Fig. 7a showed that the anaerobic zone in the granule was dominated by AMX, while only a small amount of AOB were present due to oxygen limitation. On the one hand, at R/R max = 0.7 of the granule, the growth rate of AOB accelerated due to the increased oxygen ux and increased as it approached the outer surface of the granule, while AMX were largely inactive due to its sensitivity to oxygen. On the other hand, aerobic heterotrophs had lower biomass because of the absolute dominance of AOB and less COD on the outer surface of the granule. When in uent COD was 100 mg/L (Fig. 7b), the result showed that AMX were still occupied the anaerobic zone in the interior of the granule, and the AMX biomass wherein the same location was slightly higher than the low in uent COD condition (Fig. 7a). This was because when the in uent COD concentration was increased, the biomass of denitrifying heterotrophic bacteria (DB) outside the granule increased, converting the NO 3 produced by AMX into NO 2 available to AMX, resulting in an accumulation of NO 2 in the interior of the granule (Shi et al. 2019;Zhang et al. 2020) and the higher AMX biomass. However, compared to the low in uent COD condition, AMX biomass started to decrease at granule R/R max = 0.6 because the optimum bulk oxygen concentration increased due to the increasing COD concentration, thus reducing the anaerobic zone thickness. Besides, in the outermost layer of the granule, where the aerobic heterotrophs dominated due to su cient bulk oxygen and COD (heterotrophic bacteria grew faster than autotrophic bacteria), AOB were outcompeted by the aerobic heterotrophs and AOB were present just below them Zhou et al. 2018). In general, COD and DO are almost completely depleted by the outer HET and AOB, thus forming an anoxic zone in the mid-granule, where DB reduced only NO 3 to NO 2 using limited COD (partial denitri cation), which is consistent with Hao et al. who found that the maximum concentration of HET occurred in the outer layer of the granule . Studies have shown that as COD increased too high, the total number of AMX decreased, and the thickness of heterotrophic bacteria in the granule increased. What we would like to see in the reactor s was a signi cant accumulation of NO 2 in the granule and an increase in AMX activity leading to higher TNR.

In uence of combined in uent C/N and DO on TN removal under stable operation
To evaluate the effect of in uent COD/NH 4 + on nitrogen removal at steady-state, further simulations with different C/N and DO levels were run at a xed NH 4 + concentration (Fig. 8). As mentioned before, simulations have shown that changes in in uent COD could in uence the distribution of microorganisms in the granule, TNR, and the optimum bulk oxygen concentrations on the PD process.
Previous experiments have demonstrated that in uent C/N has a signi cant impact on the PD process.
First, the nature of the microbial community in the granule/bio lm at high C/N and low DO levels was complex and existed functional species such as AOB, heterotrophs, AMX, etc. (Zhou et al. 2018). Second, when C/N > 1, AMX was eliminated by DB, leading to a deterioration in nitrogen removal Tang et al. 2010), and AMX activity was inhibited when C/N > 2 (Tang et al. 2010). Some simulations have shown that when C/N was increased from 0.1 : 3 to 4 : 3, AMX activity and TNR were signi cantly reduced at low DO levels (Hubaux et al. 2015). However, AMX activity appeared to be improved at moderate DO levels, probably due to the consumption of COD by the outer growth of aerobic HET, which resulted in decreasing the C/N in the granule, enhancing AMX activity, and TNR. The experimental concluded that AMX could not compete with DB at high C/N levels (reactor operation up to the 139-153 d and C/N = 4, AMX has signi cantly inhibited with only 24.39% NH 4 + removal e ciency, compared to the contribution of 29.47% for AMX and 68.91% for DB in the nitrogen removal process), and the biggest reason was that there was an insu cient accumulation of nitrite in the granule. Fig. 8a showed the contribution of AMX to nitrogen removal under different DO and C/N conditions. Simulation results showed that even at higher in uent C/N levels, AMX was still the main contributor to TN removal when bulk oxygen concentration was high. Therefore, increasing bulk oxygen concentration allowed aerobic heterotrophs to proliferate faster and degrade COD that resulting in lower C/N in the granule without affecting AMX activity, and Lackner et al. (2008) got the same result. The result in Fig. 8b showed that in uent C/N also in uenced the contribution of denitri cation to nitrogen removal that was not the main contributor to TN removal (maximum contribution rate of denitri cation was only 7%), which was in line with the conclusions drawn from a large number of experiments (Cao et al. 2017;Cao et al. 2016). This meant that in uent C/N affected the partial denitri cation process, therefore controlling the in uent C/N allowed equilibrium between AMX and DB within the granule, making PD process nitrogen removal more e cient. If in uent C/N was high to allow complete denitri cation (a conventional biological denitri cation process can be complete at C/N > 5-8 theoretically), we could be able to inhibit complete denitri cation by increasing bulk oxygen concentration to decrease denitri cation contribution rate (Fig.  8b). Studies have also been carried out based on in uent C/N at 2-3 for the PD process Ge et al. 2012;Ma et al. 2017). Summing up the above analysis, we found that increasing in uent C/N also resulted in a wider range of optimum DO operating conditions for the reactor and the results (Fig. 8) once again indicated that the main contributor to TN removal was AMX.
In uence of combined granule size and DO on TN removal under stable operation The in uence of the granule size and DO on the overall steady-state reactor performance in terms of nitrogen removal with organic substrate present in the in uent is shown in Fig. 9. The e ciency of autotrophic nitrogen removal could be considered not affected by the in uent COD if the granule was large enough to allow little competition between AOB, AMX, and DB for space and there was su cient DO (Fig. 9a). Choi et al. (2018) assessed the contribution of AMX, AOB, and DB to nitrogen removal according to granule size and found that the contribution of microbial activity to nitrogen removal was signi cantly in uenced by granule size. In the presence of COD, it was shown that the contribution of AMX to TN removal was higher for granule diameter larger than 1600 μm, whereas, for granule diameter smaller than 1600 μm, the contribution of AMX to nitrogen removal was unsatisfactory at any DO level (Fig. 9a). A simulation analysis of the effect of granule size on the nitrogen removal performance of a reactor by xing bulk oxygen concentration at 1 mg/L and comparing the in uent water with and without COD was carried out by Mozumder et al. (2014). This simulation showed that the presence of COD reduced the optimum diameter of the granule (the diameter of the granule was 3000 μm with the presence of COD, compared to 4000 μm for the optimal nitrogen removal without COD), and the optimal TNR became higher. The simulation result in Fig. 9b showed that the main contributor to TN removal was not denitri cation at any granule diameter level in the steady-state reactor with the absence of NOB.
In summary, to achieve satisfactory nitrogen removal, we need to control the granule size for optimal AMX activity and the corresponding denitri cation contribution was low so that we can achieve the PD process.