Impact of organics, aeration and flocs on N₂O emissions during granular-based partial nitritation-anammox

Abstract

Partial nitration-anammox is a resource-efficient technology for nitrogen removal from wastewater. However, the advantages of this nitrogen removal technology are challenged by the emission of N₂O, a potent greenhouse gas. In this study, a granular sludge one-stage partial nitritation-anammox reactor comprising granules and flocs was run for 337 days in the presence of influent organics to investigate its effect on N removal and N₂O emissions. Besides, the effect of aeration control strategies and flocs removal were investigated as well. The interpretation of the experimental results was complemented with modelling and simulation. The presence of influent organics (1 g COD g^-1 N) helped to suppress NOB and significantly reduced the overall N₂O emissions while having no significant effect on anammox activity. Besides, long-term monitoring of the reactor indicated that constant airflow rate control resulted in more stable effluent quality and less N₂O emissions than DO control. Still, floc removal reduced N₂O emissions at DO control but increased N₂O emissions at constant airflow rate. Furthermore, anammox bacteria could significantly reduce N₂O production during heterotrophic denitrification, likely via competition for NO with heterotrophs. Overall, this study demonstrated that the presence of influent organics together with proper aeration control strategies and floc management could significantly reduce the N₂O emissions without compromising nitrogen removal efficiency during one-stage partial nitritation-anammox processes.

1. Introduction

Partial nitritation-anammox (PNA) is a promising alternative for nitrogen (N) removal from wastewater, with considerable savings in energy, external organic carbon, sludge treatment, and associated carbon-footprint compared to conventional nitrification-denitrification (van Dongen et al., 2001). In this process, approximately half of the ammonium (NH₄⁺) is first oxidized to nitrite (NO₂^-) during nitritation by ammonium oxidizing bacteria (AOB), under aerobic conditions. In the following anaerobic ammonia oxidation (anammox) process, nitrite and ammonium are combined to form nitrogen gas (N₂). By 2014, there were more 100 full-scale applications of partial nitritation-anammox reactors, most of which were one-stage reactors, meaning that both reactions take place in a single reactor (Lackner et al., 2014).

However, as for any biological nitrogen removal process, nitrous oxide (N₂O) may be formed during the partial nitritation-anammox conversions. N₂O a strong greenhouse gas that is 298 times stronger than CO₂ over a 100 year time horizon (IPCC, 2013). Care should be taken that N₂O emissions do not counterbalance the carbon-footprint advantages from this technology. N₂O may be produced through three biological pathways, two of which are associated with nitritation: (1) the nitrifier nitrification pathway, where N₂O is formed as a by-product during hydroxylamine (NH₂OH) oxidation; (2) nitrifier denitrification, which involves the reduction of nitrite to N₂O (Wunderlin et al., 2012). N₂O is also an obligatory intermediate during heterotrophic denitrification, which can be either an N₂O source or sink (Ali et al., 2016; Conthe et al., 2019; Wan et al., 2019). N₂O can further be produced abiotically during the nitritation process, but its contribution is negligible (< 3% of total N₂O production) compared to biotic pathways (Su et al., 2019).

As only partial oxidation of ammonium to nitrite (nitritation) is required for the successful application of partial nitritation-anammox, the establishment of a balanced microbial community is essential. Granular sludge or biofilm reactors provides long sludge retention times (SRT) and make it possible to enrich the slow growing anammox bacteria (Vlaeminck et al., 2010). Oxygen-limited operation conditions are considered the most practical approach to limit NOB, as their oxygen affinity is lower than AOB (Blackburne et al., 2008; Wyffels et al., 2004). Therefore, controlling DO at a low setpoint is the most often implemented aeration strategies on partial nitritation-anammox reactors (Lackner et al., 2014). However, DO concentration alone might not always provide a good correlation with substrate depletion or biomass activity. Several simulations indicated that the optimal DO for nitrogen removal in a one-stage partial nitritation-anammox reactor decreased with smaller granules (Volcke et al., 2010;Wan et al., 2019) or more flocs (Hubaux et al., 2015) as the apparent oxygen affinities increased due to elevated oxygen penetration at smaller aggregates (Hubaux et al., 2015; Picioreanu et al., 2016; Volcke et al., 2010). As for N₂O, a previous simulation study also indicated that constant DO control failed in maintaining low N₂O emission at constant DO when flocs accumulated (Wan et al. 2020). However, these results still need to be validated with experimental data.

Selectively wasting sludge is another way to suppress NOB. This can be achieved by reducing the SRT in a suspended biomass system, as the growth rates of AOB are usually higher than those of NOB at elevated temperatures (> 30 °C). For granular sludge/biofilm systems, uncoupling the SRT of flocs and granules/biofilm by selectively removing flocs successfully suppressed NOB and sustained anammox bacteria (Han et al., 2020; Laureni et al., 2019; Wett et al., 2013). This corresponds with the knowledge that anammox bacteria mainly resides in granules or aggregates while AOB/ NOB dominate in flocs (Vlaeminck et al., 2010). Yet, the effect of floc removal in this system on N₂O emissions has not yet been reported.

The presence of influent organics is another challenge in anammox-based nitrogen removal processes. Heterotrophs grow inevitably on the organics from the influent and on endogenous soluble microbial decay products (Gilbert et al., 2014; Kindaichi et al., 2004). Heterotrophic denitrifiers are can compete for nitrite with anammox bacteria at high COD/N conditions (Chen et al., 2016). However, in the presence of relatively low organic carbon concentrations, heterotrophic denitrification can co-exist with anammox, providing nitrite for anammox by reduction of residual nitrate and increasing nitrogen removal efficiency in one-stage partial nitritation-anammox reactor (Mozumder et al., 2014 ; Jenni et al., 2014). Heterotrophic organisms also help to suppress NOB by competing for oxygen (Li et al., 2017). The growth of heterotrophs in the presence of influent organics also affects N₂O emissions.

Incomplete heterotrophic denitrification, caused by insufficient organic carbon (Chung and Chung, 2000; Schulthess et al., 1994), or inhibition by free nitrous acid (Zhou et al., 2008) or DO (Wunderlin et al., 2012) has been reported to cause N₂O emissions for one-stage partial nitritation-anammox reactors (Ali et al., 2016; Li et al., 2017). However, heterotrophic denitrification was also speculated to be a net sink of N₂O in several full-scale one-stage partial nitritation-anammox reactors (Castro-Barros et al., 2015; Domingo-Félez et al., 2014; Wan et al., 2019). It is still unclear whether heterotrophic denitrification serves as a net source or sink in one-stage partial nitration anammox reactors. The effect of influent organics on nitrogen removal and N₂O emissions definitely needs to be studied further.

In this study, a lab-scale one-stage partial nitritation-anammox reactor, with granular sludge, was operated for 337 days to study the effect of influent organics, as well as the impacts of different aeration control strategies and flocs removal on both nitrogen removal and N₂O production. Along with the long-term operation, batch tests were carried out to track the maximum activities of the functional groups and to study the N₂O formation mechanism. Besides, a mathematical model was developed and simulations were performed to provide a mechanistic interpretation of the experimental observations.

2. Materials and methods

2.1. Reactor operation

A one-stage PN/A bubble column reactor (6.5 L) was inoculated with granular sludge from a full-scale one-stage PN/A reactor in Tilburg, Netherlands. The N₂O monitoring started from day 1 (Figure 1), after a start-up period (which lasted 17 days, reactor DO 2 g m^-3, influent with 500 g NH₄⁺-N m^-3 and 500 g COD m^-3). The reactor was continuously fed with synthetic wastewater with an influent ammonium concentration of 500-800 g N m^-3, and influent organic carbon present in the form of CH₃COONa, as listed in Table 1. The COD/N ratio was kept at 1, except for a period without organic carbon. The reactor was operated at an HRT of 1 day. The pH was controlled at 7.5 ± 0.1 by dosing HCl (0.1 mol L^-1) and NaOH (0.5 mol L^-1). The temperature was maintained at 38 ± 1 °C with a water jacket bath.

Figure 1. Results from long-term operation of the one-stage partial nitritation-anammox reactor.

(A) Influent and effluent composition; (B), reactor pH, DO and airflow rate; (C) effluent nitrite concentration, N₂O emissions (shaded blue: high aeration - with organics, shaded green: high aeration - no organics, shaded red: low aeration - no organics.) and their correlation on days 50-337; D, Maximum nitrogen conversion rates in batch tests and the total NH₄⁺ removal rate of the reactor at normal operation; E, Visualization of events: floc removal, sensor cleaning and batch test

Table 1. Influent composition and aeration control strategies of the reactor.

Phase	Time	Influent NH4+ (g N m-3)	Influent organics (CH3COONa, gCODm-3)	Aeration control	Description
DO-a	d1-d46	500	500	DO control	Low N load
DO-b	d47-d89	800	800		Elevated N load
DO-c	d90-d177	800	800		With sensor cleaning
Qair-a	d178-d264		800		High Qair with influent organics
				Qair: 0.27 m3 d-1
Qair-b	d265-d280				High Qair without influent organics
Qair-c	d281-d314	800	0	Qair: 0.22 m3 d-1	Low Qair without influent organics
Qair-d	d315-d330				High Qair without influent organics
				Qair: 0.27 m3 d-1
Qair-e	d331-d337		800		High Qair with influent organics

Different aeration control strategies were applied (Table 1): During the DO control phase (day 1-177), the airflow rate was manipulated to reach a desired oxygen concentration. This DO setpoint was adjusted manually according to the effluent nitrite concentration measurement: the DO setpoint was lowered in case two subsequent measured effluent nitrite concentration exceeded 20 g NO₂^--N m^-3 (Figure 1A-B). The DO sensor was cleaned at days 89, 110, 139, 149, 167, 194, 223, 257 and 324 (figure 1 E). During Q_air control phase (day 178-337), the reactor was operated at a constant airflow rate. The airflow rate was adjusted manually to keep the effluent nitrite and ammonium concentrations sufficiently low (Figure 1A-B). From day 60 onwards, flocs (diameter < 0.25 mm) were removed manually from the reactor on a regular basis with a sieve, aiming to suppress the NOB (Figure 1E, days 63, 78, 89, 101, 105, 110, 127, 139, 194, 212, 223, 230, 246, 257). Each time, 1 L of total 6.5 L liquid in the reactor (15%) was taken and sieved for floc removal, the remaining liquid was returned to the reactor.

2.2. Batch tests

In-situ batch tests (Table 2) were conducted on 15 distinct days (indicated on Figure 1E). Each time, the maximum activities of anammox bacteria, denitrifiers, AOB and NOB (from 110 onwards) were quantified consecutively. Anoxic conditions were maintained by bubbling nitrogen gas (constant airflow of 0.24 L/min) to quantify the anammox activity. The maximum activity of anammox was expressed as the maximum ammonium consumption rate. Once ammonium was consumed, organics (CH₃COONa) and nitrate were dosed to stimulate denitrification while still bubbling nitrogen gas. The maximum activity of denitrifiers was quantified as the maximum nitrate consumption. The nitritation test was conducted once nitrate was consumed, applying an aeration airflow rate of 1.5 L/min, which guaranteed fully aerobic conditions (> 4 g O₂ m^-3). The maximum activity of AOB was estimated as the maximum ammonium consumption rate. From day 110 onwards, nitratation was measured simultaneously with the nitritation test by adding nitrite together with ammonium. The maximum activity of NOB was calculated as the maximum nitrate production rate. N₂O emissions were also monitored during the batch tests from day 110 onwards.

Table 2. Batch tests conditions.

Tested process	Dosed substrates	Aeration Conditions	Substrates to evaluate the activities
Anammox	NH4+ and NO2-	Bubbled with N2-	Maximum consumption rate of NH4+
Denitrification	CH3COONa and NO3-	Bubbled with N2	Maximum consumption rate of NO3-
Nitritation (and nitratation – from day 110 onwards)	NH4+ (and NO2- – from day 110 onwards)	Aerated with air	Maximum consumption rate of NH4+ Maximum production rate of NO3-
Interaction between anammox and heterotrophic denitrification	1st dosage: CH3COONa and NO3- 2nd dosage: NH4+ 3rd dosage: CH3COONa and NO3-	Bubbled with N2	-

On day 337, additional batch tests were carried out to investigate the interaction between anammox and heterotrophic denitrification. The reactor was bubbled with nitrogen gas to ensure anoxic conditions. Nitrate and organics were added to the reactor to stimulate denitrification. Once N₂O emissions were observed, an excess amount of ammonium was dosed into the reactor to initiate simultaneous anammox conversion. Following complete nitrate consumption and N₂O emissions decreasing to zero, organics and nitrate were dosed again to the reactor.

2.3. Reactor performance monitoring

The biomass concentration was determined as total suspended solids (TSS) and volatile suspended solids (VSS) according to standard methods (APHA, 1998). Liquid samples were filtered through 0.45 μm disposable Millipore filter. Ammonium, nitrite, nitrate and COD were determined with standard test kits (Macherey-Nagel, Germany). The N₂O emissions in the off-gas were continuously measured online in the periods of day 1 to day 34, day 48 to day 143, and day 234 to day 337 (X-stream Gas analyser, EMERSON, Germany).

The granules were sampled to monitor the evolution of morphology on day 1, 195, 269, and 317 with digital microscope camera (Olympus SZX9 equipped with Olympus DP21 camera). Granule size distributions were determined with biological-image analysis on day 1 and day 269 with Fiji software (Schindelin et al., 2012). 200 ml of granules (triplet) were harvested from the reactor when the reactor was aerated and well mixed. The granules were placed on a clean plastic plate and photographed without attaching to each other. The images were converted to binary images to calculate the calliper diameter in pixels. The pixel-diameters were then converted to the actual diameter with a scaling factor.

2.4. Simulation study

An existing 1D granular sludge partial nitritation-anammox model (Wan et al., 2019) was adapted to describe the design and operating conditions of the reactor set-up under study. The N₂O formation through nitritation was modelled according to a two-pathway model (Pocquet et al., 2016), including both nitrifier nitrification and nitrifier denitrification pathway. The heterotrophic denitrification was modelled in four steps with N₂O as an intermediate (Hiatt and Grady Jr., 2008). The details of the biconversions, reactor configuration and mass transfer processes described in the model were summarized in supplementary information (section S1.1 to S1.3). During operation at constant DO, the air flow rate was manipulated to reach the DO set point. Details on the gas-liquid transfer model are summarized in supplementary information (S1.3.1). The DO setpoint was maintained at 0.5 g O₂ m^-3 (optimal DO setpoint for N removal). It can be noted that the setpoint in the simulation was not the same as in the experiment (between 0 and 2.5 g O₂ m^-3 at different phase). Still, it must be noted that the DO measured by the sensor may not have been representative for the whole reactor (see further). Moreover, the model was not calibrated, since the primary goal was not to quantitatively predict the N₂O emissions, but to qualitatively understand the impacts of the operating conditions (flocs, organics, aeration) on the competition of each organism and the N₂O formation mechanisms. For the simulation of constant airflow rate control, the airflow rate was set at 0.98 m³ d^-1, corresponding to the airflow rate at optimal DO.

The initial conditions for all scenarios were obtained by steady-state simulation over 5000 days. This was followed by dynamic simulations for the different dynamic scenarios, corresponding with different operations or batch tests from experiments (Table 3), in order to validate and interpret the experimental results. The different floc fractions were modelled by implementing a bifurcation from the effluent to recycle part of the detached sludge into the inlet, as described by Hubaux et al. (2015) (detailed in supplementary information S1.4). The amount of the flocs retained in the reactor was controlled by adjusting the percentage of the recycled sludge. The model was set up and run in the Aquasim 2.1d software (Reichert et al., 1996).

Table 3. Simulation set-up.

Scenarios	Removed flocs (% of initial flocs)	Influent NH4++ g N m-3)	Influent COD (g COD m-3)	Aeration strategy	Pulse dosage
Pre-simulation (5000 days) to set initial conditions for all scenarios	-	800	800	Constant DO (*) (=0.5 g O2 m-3)	-
Stop influent COD at constant Qair (cf. day 264)	-	800	from 800 to 0	Constant Qair (*) (=0.98 m3 d-1)	-
Floc removal at constant DO (cf. Phase DO)	15%	800	800	Constant DO (*) (= 0.5 g O2 m-3)	-
Floc removal at constant Qair (cf. Phase Qair-a)	15%	800	800	Constant Qair (*) (= 0.98 m3 d-1)	-
Batch test 1: Effect of additional COD dosage	-	-	-	No aeration	At t = 0 h: add NO3- and COD At t = 0.5 h: add COD
Batch test 2: interaction between anammox and denitrification	-	-	-	No aeration	At t = 0 h: add NO3- and COD At t= 0.5 h: add NH4+
Reference: denitrification	-	-	-	No aeration	At t = 0 h: add NO3- and COD

^(*)Value corresponding with maximum nitrogen removal efficiency at initial conditions, namely 90%

3. Results

3.1. Long-term reactor performance

The one-stage partial nitritation-anammox reactor was operated for 337 days, under different influent conditions and aeration strategies. The reactor was first operated with DO control (phase DO: day 1-177). Nitrate concentrations above 100 g N m^-3 (more than 20% of the influent ammonium) were observed during the reactor start-up (Figure 1 A). In order to remedy this, the DO setpoint was gradually reduced from 2 to 0.1 g O₂ m^-3 (Figure 1 B). Besides, the influent ammonium concentration was increased from 500 g N m^-3 (phase DO-a) to 800 g N m^-3 (phase DO-b). Since nitrate formation persisted (above 100 g N m^-3), part of the flocs were removed regularly (from day 60 onwards, Figure 1 E). Regular cleaning of the oxygen sensor combined with reducing the DO setpoint (Figure 1 E) finally resulted in keeping the nitrate concentration as low as 30.3 ± 21 g N m^-3 (phase DO-c). Nevertheless, still some uncontrollable nitrite peaks occurred (e.g., at day 89, 109, 119, 147, 161 and 175), leading to a high average nitrite concentration (35.9 ± 42.8 g N m^-3). During the DO control phase, the nitrogen removal efficiency amounted to 75.5 ± 10.8 % of N load (Figure 2 A), the N₂O emissions being 4.1 ± 2.8 % of N removed.

Figure 2.

Nitrogen removal efficiency (% of N load) (A) and N₂O emissions (% of nitrogen removal) (B) at different operating strategies: DO control (phase DO) versus airflow rate (Q_air) control (phase Q_air); Comparison between high Qair with organics (phase Q_air-a, phase Q_air-e); high Qair without organics (phase Q_air-b and phase Q_air-d) and low Qair without organics (phase Q_air-c); The box plots indicate the interquartile range, the crosses indicated the average, the middle lines indicated the medians.

The air flow rate was kept constant from day 178 onwards (phase Q_air). This resulted in a more stable effluent quality, with much lower nitrite (12.0 ± 12.3 g N m^-3) and nitrate concentrations (33.8 ± 27.5 g N m^-3) (Figure 1 A). As a result, the nitrogen removal efficiency also showed less fluctuation (76 ± 6.6% of N load, Figure 2 A). In addition, the N₂O emission was much lower and stable (2.4 ± 2.8% of nitrogen removal, Figure 2 B) at airflow rate control compared to operation with DO control.

Within the phase with constant airflow, the reactor was operated with and without influent organic carbon, as well as at different airflow rates. N₂O emissions varied at different influent compositions but the nitrogen removal was relatively stable (Figure 2). In phase Q_air-a, the reactor was fed with both ammonium and organics, leading to a nitrogen removal of 79.0 ± 6.1% and a N₂O emission of 1.8 ± 0.9% of nitrogen removal (values of days 234-241 and days 254 - 264, where N₂O were measured). The fluctuating and high N₂O emissions from day 242 to day 253 (Figure 1 C) were caused by the batch test and failures in pH control. Therefore, these days were excluded in calculation of average N₂O emissions in Figure 2. Operating the reactor without organic carbon in the influent (phase Q_air-b, days 265-330) at the same airflow rate caused a higher DO concentration (Figure 1 B) and more ammonium oxidation, resulting in lower effluent ammonium concentrations (from 149 ± 58 g N m^-3 in phase Q_air-a to 87 ± 50 g N m^-3 in phase Q_air-b) and higher nitrate concentrations (from 10.5 ± 7.2 g N m^-3 to 48.8 ± 10 g N m^-3, Figure 1 A). Although the nitrogen removal only slightly increased (to 81.8 ± 6.6% of N load), the N₂O emissions almost doubled (3.5 ± 0.8% of nitrogen removal) compared to operation with influent organic carbon in phase Q_air-a.

Lowering the airflow rate while still operating without influent organics (at phase Q_air-c, days 283-d314) reduced the DO concentration and the ammonium oxidation rate, leading to higher effluent ammonium concentrations (161 ± 29 g N m^-3). The effluent nitrate concentration (45 ± 13g N m^-3) gradually increased to around 80 g N m^-3 at the end of this period, corresponding to a nitrogen removal efficiency of 73.8 ± 4.1% of N load. The N₂O emissions were lower (1.9 ± 0.5% of nitrogen removal) than in phase Q_air-b which had a higher airflow rate but higher than in phase Q_air-a when the influent contained organics. To confirm these observations, the airflow rate was increased again at phase Q_air-d (days 317 to 330), leading to a drop in the effluent ammonium concentration and a slight increase in the effluent nitrate concentration, corresponding with a nitrogen removal efficiency of 76.1 ± 5.2%. The N₂O emissions increased to 4.0 ± 0.9% compared to operation at a lower airflow rate (phase Q_air-c). Again, adding organic carbon to the influent (phase Q_air-e, days 330-337), led to a decrease in the N₂O emissions (1.0 ± 0.6% of nitrogen removal). The nitrogen removal efficiency amounted to 71.5 ± 8.9%.

On top of the aeration and influent organics, the N₂O emission dynamics were also influenced by floc removal (Figure 3) and pH variations (Figure S2). During operation with DO control, floc removal resulted in a drop in the N₂O emission and a pH decrease led to a decrease in N₂O emissions (Figure S2 A). However, at fixed airflow rate, N₂O emissions were slightly higher after floc removal and a pH drop led to a short peak of N₂O emissions, which was more pronounced in the presence of influent organics (Figure S2 B) than in the case without influent organics (Figure S2 C).

Figure 3. Three-day average N₂O emissions (top) and nitrite concentrations (bottom) before and after floc removal.

3.2. Biomass activity and N₂O emissions in batch tests

In-situ batch tests were carried out to track the dynamics of the maximum activity of each functional guild as proxy for their relative abundance (Figure 1 D). At the beginning (phases DO-a and DO-b), both the maximum rates of anaerobic ammonium oxidation and nitrate reduction decreased as the reactor was over-aerated at that period, which was also evidenced by the high effluent nitrate concentration at that period (Figure 1 A). From phase DO-c onwards, the DO sensor was cleaned regularly (Figure 1 E) so the applied airflow rate was lower (Figure 1B). As a result, the maximum rates of nitrate reduction and anaerobic ammonium oxidation increased at the beginning of the phase DO-c (Figure 1 D), corresponding to a lower effluent nitrate concentration (Figure 1 A). The absence of influent organics (in phase Q_air-b) resulted in a significant decrease in maximum nitrate reduction rate and increase in maximum nitrate production rate but did not significantly affect the maximum rates of anaerobic ammonium oxidation and aerobic ammonium oxidation. Under DO control, at day 128, a batch test was carried out immediately after floc removal, showing a significant drop in the maximum aerobic ammonium oxidation rate and nitrate reduction rate but not affecting the maximum anaerobic ammonium oxidation rate. However, at constant airflow rate (phase Q_air-a), floc removal (d246) resulted in a slight drop in maximum rates of aerobic ammonium oxidation, nitrate reduction, anaerobic ammonium oxidation and nitrate production.

3.3. N₂O during heterotrophic denitrification

During the denitrification tests, N₂O emissions showed different patterns before and after floc removal and their dynamics depended on the initial organic concentrations. In a batch test before floc removal (Figure 4 A), N₂O emissions were observed immediately after dosing organics and nitrate. However, after floc removal (Figure 4 B), a lag between dosages and N₂O emissions was observed. In the batch test with higher initial organics concentrations (Figure 4 C), the lag was extended. In the batch test after 50 days without influent organics, an extremely long lag was observed before the N₂O peak (Figure 4 E). The dosage of additional organics upon the occurrence of N₂O occurred in the batch tests of day 248 (Figure 4 D) and day 315 (Figure 4 E) stimulated N₂O emissions. In all heterotrophic denitrification batch tests, N₂O emissions were in accordance with the total nitrite reduction rate ((C_NO2-(t+Δt) - C_NO2-(t))/Δt - (C_NO3-(t+Δt) - C_NO3-(t))/Δt, Figure 4, purple line).

Figure 4. Batch test results: substrate concentrations, N₂O emissions and nitrite reduction rate dynamics.

In another batch test to study the interaction between anammox and heterotrophic denitrification and its effect on N₂O formation by heterotrophic denitrification, ammonium was added upon the observation of N₂O emissions (66 min after dosage of organics and nitrate, Figure 5). Although, the N₂O emissions remained increasing, the acceleration in N₂O emissions (d_emissions/dt) dropped immediately after the addition of ammonium, indicating the reduced N₂O production rate by heterotrophic denitrification. In the following denitrification test in the presence of ammonium (from 140 min to 250 min, Figure 5), the N₂O emissions were significantly reduced.

Figure 5. N₂O emissions and substrates concentrations during denitrification when ammonium were added.

3.3. Simulated N₂O emissions and microbial composition dynamics

Model simulations were carried out to provide a mechanistic interpretation of the experimental observations at different scenarios, namely floc removal at different aeration control strategies, stopping influent organics feeding and batch tests.

The simulation results confirmed that floc removal led to a drop in N₂O emissions and nitrite concentrations at constant DO but stimulated N₂O emissions and nitrite accumulation at constant airflow (Figure 6 A). Before floc removal, flocs account for 14.8% of the total biomass in the reactor and 38% of AOB, 84% of heterotrophs and 3.2% of anammox were located in flocs. Removing 15% of flocs wasted 2.8% of total biomass, including 8.2% of AOB, 12% of heterotrophs and 0.9% of anammox from the reactor (Figure S4). This indicated that floc removal mainly reduced the presence of AOB and heterotrophs as they are more prevalent in flocs than anammox bacteria (Vlaeminck et al., 2010).

Figure 6. Simulated N₂O emissions and substrate concentrations at different dynamic scenarios.

A: N₂O emissions after floc removal; B: N₂O emissions after stopping influent organics dosing; C&D: N₂O production from heterotrophic denitrification in batch tests; E&F: Effect of dosing ammonium on N2O production during denitrification; G&H: Effect of dosing organics on N₂O production during denitrification.

Stopping influent organics feeding at constant airflow rate led to higher DO concentration, as well as increased nitrite formation and N₂O emissions (Figure 6 B). The absence of influent organics also increased the fractions of AOB, NOB and anammox (Figure 53 A). On the contrary, the fraction of heterotrophs declined, leading to lower N₂O consumption rate (Figure 53 B). The N₂O production rate by AOB showed a peak immediately after the absence of influent organics and decreased gradually due to the competition for nitrite by NOB.

During the denitrification batch test, an N₂O emissions increased rapidly when organics became limiting (Figure 6 C). Replenishing organics led to clearly reduced N₂O emissions (Figure 6 E), which dropped to zero once nitrite was depleted (Figure 6 F). Adding ammonium during the denitrification immediately decreased the nitrite concentration (Figure 6 H) and N₂O emissions (Figure 6 G).

4. Discussion

4.1. Influent organics can reduce N₂O emissions without sacrificing nitrogen removal efficiency

The presence of organics did not significantly affect the overall nitrogen removal efficiencies at constant airflow rate (Figure 2 A), although the effluent ammonium and nitrate differed at different influent organics conditions. On the one hand, the presence of organics stimulated the growth of heterotrophs, competing for oxygen with AOB thus reducing the ammonium oxidation rate. This was evidenced by the higher effluent ammonium concentration in the presence of organics in phase Q_air-a. On the other hand, the presence of organics reduced the residual effluent nitrate and improved the overall N removal. The competition for oxygen by heterotrophs helped to suppress NOB in this reactor, decreasing the nitrate produced by NOB. This is confirmed by the lower maximum nitrate production rate in the presence of influent organics (figure 1 D). In addition, the presence of influent organics stimulated the heterotrophic denitrification, consuming nitrate produced by anammox. Therefore, although the presence of influent organics led to higher effluent ammonium, the overall nitrogen removal did not decrease significantly as the effluent nitrate was diminished by stimulated denitrification and competition for O₂ on NOB.

While nitrogen removal efficiency was not affected, the N₂O emissions decreased significantly in the presence of influent organics. As indicated in Figure 2 B, N₂O emissions in all the periods with organics in the influent were lower than for the periods without influent organics.

One reason is that influent organics lead to a reduced ammonium oxidation rate, implying less nitrite accumulation and N₂O formation via nitrifier denitrification pathway. This was evidenced by both experiments (figure 1 C, phase Q_air-b) and simulations (Figure 6 B, Figure S5 B). Besides, influent organics can also stimulate the N₂O consumption by heterotrophic denitrification by alleviating the carbon limitation to the N₂O reduction step. This is confirmed by the simulation results, where the N₂O reduction rate by heterotrophic denitrification is much higher in the presence of influent organics (Figure S5 B). This is also in line with the results of the batch tests (Figure 4 B, D, E), where no N₂O emission was observed until the COD concentration became limiting (around 150 g COD m^-3, note that part of the residual COD was slowly biodegradable COD accumulated from cell decay during the bath test) and the lag between the dosage and observation of N₂O emissions was extended with a higher initial COD concentration (Figure 4 D). The positive effect of influent organics on N₂O mitigation was also reported in previous simulation studies (Chen et al., 2019; Wan et al., 2019) and was experimentally confirmed in this study.

However, adding organics at high nitrite concentrations did not reduce but stimulate N₂O emissions in the batch tests (Figure 4 C and E). This can be explained by the inhibition of free nitrous acid (HNO₂) on the N₂O reduction step (Zhou et al., 2008) caused by accumulated nitrite (Figure 4 D and E). Therefore, adding organics when N₂O reduction step was inhibited only increased nitrite reduction rate (Figure 4 C and E) and temporally stimulated the N₂O accumulation. Contrary to the experimental observation, replenishment of organics in the middle of denitrification led to the drop of N₂O emissions in the simulation (Figure 6 E). This is because the inhibition of free nitrous acid was not considered in the model, and the N₂O consumption by denitrification was stimulated without free nitrous acid inhibition. In an additional simulation where the N₂O reduction step was inhibited by free nitrous acid, adding COD in the middle of denitrification stimulated N₂O production (Figure S6). Overall, it is clear that nitrite accumulation needs to be avoided to reduce N₂O emissions. This condition is typically fulfilled in well-operated one-stage partial nitritation-anammox reactors: nitrite concentrations are typically low (less than 10 g N m^-3 under normal operating conditions in this study) and the pH is about neutral, thus the inhibition of free nitrous acid is negligible.

High concentrations of influent organics have been reported to be inhibitory to anammox and lead to its out competition (Chen et al., 2016; Leal et al., 2016). In this study, the anammox activity and the overall nitrogen removal efficiency were relative stable during the long-term operation with an influent organics concentration of 800 g COD m^-3 (COD/N ratio 1 g COD g N^-1). Successful application of partial nitritation-anammox reactor at high influent organics (COD/N ratio 1.4 g COD g N^-1) with an improved N removal efficiency was also reported before (Jenni et al., 2014). Likely, one-stage partial nitritation-anammox reactors allow a higher influent organics concentration than anammox reactors, as part of the influent organics were oxidized by ordinary heterotrophic organisms. Still, influent organics increased the amount of flocs and filamentary structures around the granules in this study (Figure S1), as also reported in an anammox reactor (Pijuan et al., 2020). This could lead to a reduced SRT and more efforts on sludge retention would be required. The variations in granule morphology could further affect the N₂O emissions, as discussed in the next sections.

4.2. Airflow rate control leads to stable reactor performance and low N₂O emissions

Controlling the oxygen supply is crucial in operating a one-stage partial nitritation-anammox reactor. The oxygen supply needs to be low enough to have nitrite oxidizers outcompeted by anammox bacteria, but still high enough to enable full ammonium conversion. Keeping the DO setpoint low is considered a practical approach to not only suppress NOB due to lower oxygen affinity of NOB than AOB (Blackburne et al., 2008; Wyffels et al., 2004), but also to achieve partial nitritation by limiting the oxygen availability for AOB.

In this study, two aeration strategies were tested experimentally, namely constant DO and constant airflow rate. Controlling the bulk DO concentration at a constant setpoint, as commonly applied in practice (Lackner et al., 2014), resulted in a more fluctuating nitrogen removal efficiency and higher N₂O than when applying a constant airflow rate (Figure 2). DO setpoint control resulted in accumulation of nitrate and nitrite even when DO was decreased to 0.1 g O₂ m^-3 (Figure 1 A). This could partly be attributed to the growth of biofilm on the DO probe, causing a deviation in the DO measurement resulting in over-aeration. Regular cleaning of the DO sensor combined with decreasing DO setpoint successfully suppressed NOB (Figure 1 D). Still, fluctuations in nitrite and nitrate were observed, probably caused by random detachment and attachment of biofilm on the DO sensor. Therefore, frequent adjustments to the DO setpoint were required in phase DO-c (Figure 1 B) even if the DO sensor was cleaned periodically.

Apart from the offsets in the DO measurements, variations in floc fractions and granule sizes in the reactor could be additional reasons for the fluctuating effluent nitrite and nitrate concentrations at constant DO. A decreasing granule size or increasing floc fraction decreases the apparent oxygen affinity constant of AOB and NOB as well as the apparent inhibitory coefficients on anammox bacteria, due to the higher oxygen penetration in small aggregates (Picioreanu et al., 2016; Volcke et al., 2012). Therefore, even if DO concentration could be maintained constant in practice, the occurrence of peaks in nitrite and nitrate concentration are likely to occur with varying granule size and floc content. The effect of granule size or floc contents on the optimal DO concentration in one-stage partial nitritation-anammox reactors was also reported in several simulation studies (Hubaux et al., 2015; Mozumder et al., 2014; Wan et al., 2019).

In the case of constant airflow, variations in apparent oxygen affinities are compensated by the changes in the DO concentrations. Therefore, the effluent nitrite and nitrate concentrations show less fluctuation at constant airflow rate. The experimental results from this study thus confirm earlier simulation results indicating that constant airflow rate control leads to a better effluent quality than constant DO control under fluctuating floc amounts (Wan et al., 2021).

Constant DO control also leads to much more pronounced fluctuations in N₂O emissions than constant airflow, with a much higher average N₂O emission level for the same influent compositions. The associated fluctuations in the oxygen transfer rate may indeed lead to fluctuations in ammonium oxidation rate and related N₂O production by nitrifier nitrification. Besides, high nitrite peaks caused by fluctuations in ammonium oxidation at constant DO control can stimulate the N₂O production by nitrifier denitrification (Chandran et al., 2011). In addition, the nitrite peaks result in high free nitrous acid (HNO₂), which is inhibitory to the N₂O reduction by heterotrophic denitrification. Zhou et al. (2008) reported inhibitory concentrations of 0.0007–0.001 g HNO₂-N m^-3 corresponding to a total nitrite concentration of 17 to 24 g NO₂^--N d^-1 under the condition of this study (pH 7.6 and T 37 °C) – which were indeed reached here. Moreover, airflow rate fluctuations as such are known to cause increased N₂O emissions in one-stage partial nitritation-anammox reactors: Castro-Barros et al., (2015) reported N₂O emissions peaking at the transient from low aeration periods to high aeration periods during a full-scale monitoring campaign. The experimental results from this study thus confirm previous simulation results indicating that constant airflow rate leads to a better effluent quality than constant DO control (Wan et al., 2021).

In full-scale reactors, the application of a constant airflow may not be sufficient to reach stable nitrogen removal, since the influent nitrogen load may vary. Airflow rate control based on effluent nitrite and ammonium concentrations rather than constant DO control is then recommended to reduce N₂O emission and suppress NOB (Joss et al., 2011), as it can strictly control the oxygen transferred into the reactor and avoid over-aeration.

4.3. Effect of floc removal on N₂O dynamics depends on aeration control strategy

During the long-term operation, an accumulation of flocs and decreasing of granule size were observed in the reactor (Figure S1), which was likely due to the different hydraulic conditions between the lab-scale reactor and the full-scale reactor where the granules were taken from. The relatively high amount of easily biodegradable organics in the influent, which stimulated the growth of fast growth organisms (heterotrophs), could be another reason for the smaller granule size in this reactor (de Kreuk and van Loosdrecht, 2004). Smaller granules and flocs contain a higher aerobic fraction than larger aggregates, resulting in a relatively higher abundance of AOB and heterotrophs than anammox bacteria. This is in line with the experimental results showing that anammox bacteria mainly resides in granules while AOB and NOB dominate in flocs (Siegfried E. Vlaeminck et al., 2010). Therefore, the flocs were removed regularly from the reactor in the long-term operation to avoid NOB accumulation.

Operating at constant DO, floc removal led to a drop in N₂O (Figure 3 A). Floc removal implies a relatively higher waste of AOB than anammox from the reactor (simulation results, Figure S4), which led to a clear drop in maximum ammonium oxidation rates, without affecting the maximum rates of anammox (batch tests, Figure 1 D d125). The decreased ammonium oxidation rate upon floc removal leads to lower effluent nitrite concentrations and N₂O emissions (Figure 3) at DO control. Besides, removing flocs decreased the oxygen consumption and thus a somewhat lower airflow rate was required to maintain the bulk DO, decreasing the stripping and N₂O emissions slightly. These observations are in agreement with previous simulation studies indicating that more flocs led to higher N₂O emissions at constant DO (Liu et al., 2020). In contrast, when aeration airflow rate is kept constant, floc removal slightly increases the N₂O emissions (Figure 3 A and Figure 6 A). Floc removal at constant airflow rate results in a higher DO concentration (Figure 1B and Figure 6 A) because of reduced total oxygen consumption. The elevated DO concentration stimulated the ammonium oxidation, despite the fact that some AOB were removed, and increased DO inhibition on anammox. As a result, nitrite accumulates and stimulates the N₂O emissions after floc removal at constant airflow rate. The results of this work thus provided experimental confirmation of previous simulation-only studies (Liu et al., 2020; Wan et al., 2021) that reported the presence of flocs do affect N₂O emissions.

4.4. Effect of pH drop on N₂O emissions

The effect of a pH drop (from 7.6 to 6.5) on N₂O emission dynamics depends on the aeration strategy. At fixed DO, a pH drop leads to a continuous decrease in N₂O emissions (Figure S2 A). This is in line with the observation in several nitritation reactors where lower pH decreased N₂O production when DO were maintained constant (Law et al., 2011; Rathnayake et al., 2015). The effect of pH can likely be attributed to reduced catalytic activities of enzymes involved in N₂O production at lower pH (Blum et al., 2018). Even though pH changes also causes variations in free ammonia (FA) and free nitrous acid (FNA) concentrations, the latter do no significantly affect N₂O formation by AOB (Law et al., 2011).

In contrast, under a constant airflow rate, a pH drop led to a short-term peak in N₂O (Figure S2 B and C). These result contrast with the ones for a nitritation reactor (Su et al., 2019b), where a pH drop (from 8 to 6.5) led to decreased N₂O emissions at constant airflow rate. In this study, the pH drop at constant airflow resulted in a higher reactor DO increased the DO in the reactor due a lower ammonium oxidation rate at lower pH. Probably the higher DO concentration caused the inhibition on anammox and led to a short-term nitrite accumulation that triggered N₂O production by nitrifier denitrification pathway. Unfortunately, the nitrite concentration at the drop of pH is not available in this study as nitrite was not monitored online. The effect of pH on N₂O emissions in one-stage partial nitritation-anammox reactors were never reported in literatures. Further studies are required to validate our assumptions.

With organics in the influent, the N₂O emissions following a pH drop are more pronounced. Probably the inhibition of free nitrous acid on N₂O reduction by heterotrophic denitrification (Zhou et al., 2008) at lower pH and accumulated nitrite led to high N₂O accumulation during heterotrophic denitrification in the presence of organics. This is also confirmed by batch tests where adding organics at high nitrite concentrations stimulates N₂O productions by heterotrophic denitrification (Figure 4 D and E). The stimulated N₂O production by heterotrophic denitrification at lower pH was also reported in partial nitritation reactors (Lv et al., 2016; Rathnayake et al., 2015).

4.4. N₂O turnover mechanisms

Consistently with previous modelling (Peng et al., 2015; Wan et al., 2019) and experimental (Blum et al., 2018a) results, the contribution of N₂O formation pathways strongly depended on the applied aeration strength. During phase DO-a, the high aeration strength (inferred from the low residual ammonium concentrations and high effluent nitrate concentrations) resulted in low effluent nitrite concentrations together with significant N₂O emissions likely dominated by nitrifier nitrification. High aeration results in higher ammonium oxidation rate and thus NH2OH production, substrate for nitrifier nitrification. In turn, nitrifier denitrification is linked to the availability of nitrite (Chandran et al., 2011; Okabe et al., 2011; Wrage et al., 2001) and inhibited by high DO (Harris et al., 2015; Ma et al., 2017). From phase DO-b onwards, when a more limiting aeration was imposed (indicated by the lower effluent nitrate and higher residual ammonium concentrations), the N₂O emissions strongly correlated with the effluent nitrite concentration (Figure 1 C). High nitrite concentrations favour nitrifier denitrification pathway while at the same time the resulting free nitrous acid concentration likely inhibited the N₂O reduction by heterotrophic denitrifiers (Zhou et al., 2008). The correlation between nitrite and N₂O was also reported in another lab-scale reactor with intermittent aeration without influent organics (J. Blum et al., 2018). Thus, limiting nitrite accumulation in one-stage partial nitritation/anammox reactors is paramount to limit N₂O emissions.

Incomplete heterotrophic denitrification has been also speculated to be, on the one hand, a N₂O source in anammox reactors primarily due to carbon limitation or inhibition of N₂O reduction by free nitrous acid (Jia et al., 2018; Pijuan et al., 2020). On the other hand, several simulation studies support the role of heterotrophic denitrification rather as net N₂O sink in one-stage partial nitritation-anammox reactors (Chen et al., 2019; Lang et al., 2019; Wan et al., 2019). While direct experimental evidence is currently lacking, net N₂O consumption has been observed in the anoxic period of intermittently aerated full (Castro-Barros et al., 2015) and lab-scale (Domingo-Félez et al., 2014) one-stage partial nitritation-anammox reactors. In our study, excess COD was present in the influent, and the relatively low effluent nitrite concentration (< 10 g N m^-3 under stable operation) was lower than the reported inhibitory threshold (Zhou et al., 2008). Also, in one-stage partial nitritation-anammox reactors, anammox reduced the availability of substrates for N₂O production by heterotrophic denitrifiers, probably leading to less N₂O formation than consumption (N₂O from other pathways) in heterotrophic denitrification. Therefore, heterotrophic denitrification was speculated as a net N₂O sink in this reactor. This was also supported by the decreased average emissions in presence of influent COD (Figure 2 B).

What's more, our simulations (Figure 6 G, H) and experiments (Figure 5) suggested that anammox activity reduced N₂O formation by heterotrophic denitrification under anoxic conditions. In the simulations, anammox competed for nitrite with denitrifiers (figure 6 H), limiting nitric oxide and subsequent N₂O production. In the experiments however, the addition of ammonium to activate anammox resulted in a reduction of N₂O accumulation during heterotrophic denitrification even at non-limiting nitrite concentrations of 10 g N m^-3 (affinity constant of denitrifiers is 0.2 g N m^-3, Table S4). Owing to the recently discovered ability of anammox to grow with NO and ammonium as sole substrates (Hu et al., 2019), it is here tempting to speculate that the reduced N₂O emissions resulted from the direct competition for NO between anammox and heterotrophic denitrifiers. Unfortunately, the kinetics of anammox metabolism on NO remain as yet unknown and thus the reaction could not be included in the model. To further understand the potential positive role of anammox in controlling N₂O emissions, more experimental characterizations of anammox NO turnover are warranted.

In sum, our results indicate that heterotrophic denitrification acted as net N₂O sink during stable operation and suggest a positive role of anammox in consuming NO and thus reducing heterotrophic N₂O production.

5. Conclusions

A one-stage granular sludge partial nitritation-anammox reactor, was operated for 337 days. Model simulations were performed to aid in the mechanistic interpretation of the experimental observations.

• The presence of influent organics (COD/N = 1) significantly reduced N₂O emissions at constant airflow rate, by decreasing N₂O formation by AOB and increasing consumption by heterotrophs. The overall nitrogen removal efficiency was not affected, even though the effluent nitrate concentration was somewhat lower at the expense of a higher ammonium concentration.

• DO control failed to maintain low effluent nitrite and nitrate concentrations and led to significant fluctuations in N₂O emissions, while airflow rate control not only improved the effluent quality but also reduced N₂O emissions.

• Floc removal had contradictory effects on N₂O emissions depending on the aeration control strategy: at constant DO, removing flocs reduced ammonium oxidation rate and associated N₂O emissions; if a constant airflow was applied, floc removal elevated DO and nitrite concentrations, slightly increasing N₂O emissions.

• A sudden pH decrease led to a drop in N₂O emissions at constant DO but caused an N₂O peak, which is more pronounced in the presence of organics, at constant airflow rate.

• Nitrite accumulation should be avoided, as it stimulated the N₂O production by nitrifier denitrification and heterotrophic denitrification, and likely inhibited N₂O reduction by heterotrophic denitrification.

• Anammox bacteria potentially play a positive role in consuming NO and thus reducing heterotrophic N₂O production.

Highlights.

• Influent organics reduce N₂O emissions without sacrificing nitrogen removal efficiency

• Constant airflow rate control reduces N₂O emissions compared to constant DO

• Floc removal reduces N₂O under constant DO but slightly increases N₂O under constant airflow rate

• Anammox effectively decreases N₂O production by heterotrophic denitrification