Tracing N2O formation in full-scale wastewater treatment with natural abundance isotopes indicates control by organic substrate and process settings

Highlights • 15N and 18O abundances were applied to trace N2O transformations in full-scale WWTP.• Heterotrophic denitrification was identified as main N2O production pathway.• Seasonal N2O emission peaks occurred during winter when NO2− accumulates.• Dosage of reject-water and full aeration of biological treatment increased N2O emissions.


Introduction
Nitrous oxide is the third most important greenhouse gas and the dominant ozone depleting substance in the stratosphere (IPCC 2013; Ravishankara et al., 2009). Wastewater treatment plants are potent point sources and significant contributors to global anthropogenic N 2 O emissions (Tian et al., 2018;Vasilaki et al., 2019). N 2 O emissions from WWTP exhibit strong temporal dynamics (Gruber et al., 2020). The underlying drivers of these dynamics, however, remain partially unclear, and are likely linked to the complexity of the different nitrogen-cycle reactions involved in N 2 O production in wastewater treatment systems (Domingo-Félez and Smets 2020; Schreiber et al., 2012). Three main metabolic pathways performed by two different groups of bacteria have been identified in WWTPs: (i) hydroxylamine (NH 2 OH) oxidation (Ni) and (ii) nitrifier denitrification (nD) by ammonia oxidizing bacteria (AOB), as well as (iii) heterotrophic denitrification (hD) by heterotrophic denitrifying bacteria (HET) (Ren et al., 2019;Wunderlin et al., 2012). Multiple other microbial and abiotic N 2 O production pathways have been described in literature for specific ecosystems (Butterbach-Bahl et al., 2013) but are not discussed here, to focus on the most relevant processes. However, given a sufficient supply of organic carbon, HET are also able to reduce N 2 O to N 2 , the target product of N elimination in WWTP (Conthe et al., 2018;Pan et al., 2013).
The systematic and efficient mitigation of N 2 O emissions in WWTPs is a challenging task and requires both long-term monitoring of emissions to identify emission peaks, as well as a mechanistic understanding of N 2 O formation mechanisms in the wastewater treatment process (Vasilaki et al., 2019). A number of approaches have been applied successfully in full-scale WWTPs to reduce N 2 O emissions, such as the control of the dissolved oxygen (DO) through different aeration rates and timing (Rodriguez-Caballero et al., 2015;Sun et al., 2017), or different feeding regimes (e.g., step / intermittent feeding) maintaining low in situ ammonium concentrations (Hu et al., 2013). However, given the intricacy of N 2 O production and turnover, methods to quantify and to mechanistically understand the pathways involved are essential to explain emission dynamics and develop robust mitigation strategies (Duan et al., 2021).
Differences in stable isotopic substitution of the N 2 O molecule and the bulk isotopic composition of reactive nitrogen substrates ammonium (NH 4 + ), nitrite (NO 2 − ), and nitrate (NO 3 − ), provide valuable information on N 2 O transformation processes, since distinct microbial and/or abiotic pathways exhibit characteristic isotopic signatures (Sutka et al., 2006;Yoshida and Toyoda, 2000 (Denk et al., 2017;Toyoda et al., 2017;Yu et al., 2020). In addition, the O isotope ratio in the N 2 O pool is influenced by O-atom exchange reactions between water and N intermediate molecules, especially NO 2 − (Casciotti et al., 2002;Lewicka-Szczebak et al., 2016). SP is independent of the substrate isotopic composition and, therefore, an especially sensitive tool for distinguishing mechanisms of N 2 O production and consumption. A powerful way to use the isotopic composition of N 2 O to constrain its formation and processing is the dual isotope mapping approach, where SP values are plotted against either Δδ 15 N bulk (N 2 O, substrate) or Δδ 18 O(N 2 O, H 2 O) and compared to the isotope signatures known for a given process . Despite the potential that natural abundance N 2 O isotope measurements offer for pathway characterization, past applications have been almost exclusively limited to laboratory scale reactors (Wunderlin et al., 2013, Tumendelger et al. (2016).
In this study, we tested, for the first time, whether natural abundance stable isotope measurements in a full-scale WWTP can be applied to characterize N 2 O production pathways under changing inflow composition and process operation. In particular, we evaluated the influence of organic substrate availability and aeration strategies on the N 2 O formation pathways. To further support the estimated contributions of different production pathways and N 2 O reduction, we used measurements of the 15 N/ 14 N and 18 O/ 16 O isotope ratios of N substrates, NH 4 + , NO 3 − , and NO 2 − . Additionally, we performed both spatially and temporally resolved process monitoring of N 2 O emissions and aqueous nitrogen species to interpret the process dynamics during the experiments. Finally, we propose reduction strategies based on the observed emission patterns and attributed pathways.

Field site
The Hofen WWTP (Switzerland, 47 • 27 ′ 57.3 ′′ N 9 • 23 ′ 49.1 ′′ E) treats the wastewater of roughly 70,000 population equivalents. After mechanical treatment by screening, grit chambers, and primary clarification, the wastewater enters the biological treatment stage, consisting of six activated-sludge plug-flow reactors, each comprising three cascaded stirred reactors (3 × 530 m 3 , Fig. 1). While organic compounds and N are removed biologically, phosphorus is removed through chemical precipitation using iron(III). This biological treatment scheme represents a standard activated sludge configuration (Tchobanoglous et al., 2014). The average COD and nitrogen load of the treatment plant are 9700 kgCOD/d and 860 kgN/d with average removal rates of 95% and 65%, respectively.
The biological treatment is equipped with multiple liquid-phase sensors for continuous DO (LDO sc, Hach, USA) monitoring (Fig. 1 The wastewater is evenly distributed over the six treatment lanes. The N removal process is anoxicoxic, i.e., anaerobic denitrification to N 2 and aerobic NH 4 + oxidation. The DO concentration is controlled at distinct set-points for each compartment. The first zones are generally operated anoxically and stirred, but can be aerated, as soon as the air consumption in Zone 3 exceeds a defined threshold. This primarily happens during wet weather conditions and in the winter seasons at low temperatures. The second and the third zone are obligatory oxic, i.e. are continuously aerated. Even under aerated conditions, denitrification can proceed within anoxic microsites/microaggregates (Daigger et al., 2007). After the biological treatment to eliminate fixed N, the wastewater enters the secondary clarifiers. Two activated sludge lanes share one secondary clarifier, respectively, and therefore receive the same return sludge (Fig. 1). The biological treatment is operated with a fixed total-solids retention time (SRT) of 13 days. Excess activated sludge is treated in an anaerobic digestion process (not shown in Fig. 1), delivering ammonium-rich reject water to the biological treatment.
Increasing the ammonium load in the inflow, reject water is dosed into the primary clarifier to make sure that the N load is equally distributed among the lanes. Typically, reject water from sludge treatment is added overnight from 11 pm to 7 am in batches, every 30 min.

. Continuous N 2 O monitoring
Continuous N 2 O emission monitoring was done using the flux chamber approach, as described in Gruber et al., (2020). A part of the monitoring results (November 2019 -December 2020) has already been presented by Gruber et al. (2021). Flux chambers were installed in Zone 1, 2 and 3 according to Fig. 1. Additionally, 1.5-meter-long columns, called anox tubes ( Fig. S.1), were installed in Zone 1 of selected lanes (1.1, 2.1, 2.2, 3.2) to sample N 2 O from the mixed liquor during non-aerated operation by gas stripping with a blower. This technique provides qualitative information on temporal fluctuations of dissolved N 2 O concentrations for Zone 1. N 2 O concentrations from the anox tubes are not quantitative, since the efficiency of the stripping process can only be roughly quantified ( Fig. S.2). However, anox tubes provide a temporal trend of dissolved N 2 O concentrations, relevant for interpretation of N 2 O production/consumption processes. A small share of the off-gas from the chambers and anox tubes was diverted to a central N 2 O measuring unit, consisting of an automated valve system, preceding a dehumidifier and a non-dispersive infra-red sensor (X-stream, Emerson, St. Louis MO, USA). The N 2 O monitoring system was installed in October 2019, and since then is running continuously.

. Campaigns with isotope measurements
In 2019, 2020, and 2021 three intensive sampling campaigns supported by N 2 O isotopic measurements were performed on two selected lanes (2.1 and 2.2, Table 1). Campaigns were conducted on days with rather dry weather conditions on the day of sampling, since rain weather reduces emissions substantially (Gruber et al., 2020). Gaseous and aqueous samples of specific zones were collected for isotopic analyses and concentrations measurements during the experiments. Details on the experiments are given in Table 1.

. Collection of gaseous and aqueous samples and isotopic analyses
Gas samples for N 2 O isotopocule analyses were collected from the sampling lines of the N 2 O monitoring system. For this, the respective line was disconnected from the automated multiport inlet system (Gruber et al., 2020) of the off-gas monitoring device, and the sample gas was extracted with a membrane pump (model PM25032-022, KNF Neuberger AG, Switzerland). Gas samples were integrated over 15 to 20 min to ensure representativeness, dehumidified by permeation drying (model PD-50T-72MSS, Perma Pure LLC, USA) and stored in 40 L aluminum coated gas bags (model GSB-P/44, Wohlgroth AG, Switzerland) until analysis at the Laboratory for Air Pollution / Environmental Technology, Empa. For every gas sample a duplicate was collected to check integrity during transport and prevent sample loss; duplicate samples agreed within 0.5 ppm N 2 O for all gas bags.
The abundances of N and O stable isotopes in aqueous or gaseous samples were reported relative to a standard in the δ-notation in per mil (‰) (Werner and Brand, 2001): ), mixed liquor samples from the wastewater reactors collected in parallel with gas samples, were filtered with 0.45 and 0.2 µm single-use membrane filters, and stored refrigerated until further processing (Magyar et al., 2021). Nitrogen and oxygen isotope analyses of NO 3 − , NO 2 − , and NH 4 + were conducted at the Department of Environmental Sciences, University of Basel, Switzerland. δ 18 O and δ 2 H in wastewater were measured at the Stable Isotope Laboratory of the Department of Environmental System Sciences, ETH Zurich.

N 2 O isotope measurement (gas phase)
N 2 O sample gas concentrations were determined with a nondispersive infrared spectrometer (X-stream, Emerson, St. Louis MO, USA). Thereafter, sample gases were diluted to ambient N 2 O concentrations (approx. 330 ppb) with high-purity synthetic air using mass flow controllers (Vögtlin Instruments GmbH, Switzerland), and the  dilution ratio adjusted after CRDS analysis (G5131-i, Picarro Inc., USA). The isotopocule abundances in the samples were measured using quantum cascade laser absorption spectroscopy (QCLAS), preceded by preconcentration (TREX), as described in Ibraim et al. (2018). All samples were analysed in triplicate and standard deviations for repeated analyses was around 0.5 ‰ for all delta values. For calibration a two-point delta calibration approach was implemented (CG1: δ 15 N α = 2.06 ± 0.05 ‰, δ 15 N β = 1.98 ± 0.20 ‰, δ 18 O = 36.12 ± 0.32 ‰; CG2: , and instrumental drift, as well as differences in N 2 O concentration corrected .

Isotope analysis in dissolved N species
The N and O isotopic abundances in NO 2 − were determined using the azide method, where NO 2 − is chemically converted to gaseous N 2 O at low pH (4 to 4.5) (Magyar et al., 2021;McIlvin and Altabet, 2005). For the conversion, a sample volume equivalent to 40 or 10 nmol of NO 2 − (depending on the concentration in the sample) was added to 3 ml of nitrite-free seawater in a 20 ml headspace vial, and crimp-sealed. The seawater is used to maximize N 2 O yield and minimize oxygen exchange during the reaction (Granger et al., 2020). Then, 300 µl of acetic acid-sodium azide solution (1:1 mixture of 2 M NaN 3 with 20% acetic acid) were injected in the vial, and the mixture was shaken. The reaction was stopped using 200 µl 10 M NaOH after at least 30 min. The pre-processing was conducted on the sampling day, and the samples were stored upside-down at room temperature until analysis. The N and O isotopic composition in the concentrated and purified N 2 O samples were measured using a Delta V Plus gas chromatograph isotope ratio mass spectrometer (GC-IRMS, Thermo Scientific, Germany) interfaced with a customized purge-and-trap system and a GC PAL autosampler (CTC, Switzerland), and standardized using the nitrite reference materials N-7373 and N-10,219 (Casciotti et al., 2007) (Zhang et al., 2007). Briefly, a sample volume equivalent to 40 nmol of NH 4 + was converted to NO 2 − by reaction with 0.5 mL of a 50 µM alkaline hypobromite in a 20 ml headspace vial. Then, this NO 2 − sample was converted to N 2 O by reaction with sodium azide, and the N 2 O was analysed as described in the preceding section. In addition to the nitrite standards N-7373 and N-10,219, international ammonium reference materials (IAEA-N1 and USGS26) were prepared, measured alongside the samples and used to standardize the measurements. The isotopic composition (N, O) of NO 3 − was measured by conversion to N 2 O with the denitrifier method (Casciotti et al., 2002;Sigman et al., 2001). Prior to the NO 3 − isotope analysis, 1 ml of the filtered sample was pre-treated with 40 µl 0.6 M sulfamic acid in 2 ml Eppendorf tubes for NO 2 − removal. The preparation was neutralized by adding 9 µl 2.5 M NaOH after at least 15 min and before the end of the day. Until further processing, the samples were stored at -20 • C. Then, NO 3 − sample equivalent to 20 nmole was converted to N 2 O by a pure culture of denitrifying bacteria (Pseudomonas chlororaphis ATCC 13,985) lacking the NosZ enzyme for N 2 O reduction. The N and O isotopic composition in the concentrated and purified N 2 O samples were measured using a Delta V GC-IRMS (Thermo Scientific, Germany) interfaced with a customized purge-and-trap system and a GC PAL autosampler (CTC, Switzerland), and standardized using international nitrate reference materials (IAEA-N3, USGS32, and USGS34) prepared and measured alongside the samples.

H 2 O isotope measurement
In experiment 3, aqueous samples were analyzed for δ 18 O-H 2 O using the high-temperature carbon reduction method. For that purpose, a high-temperature elemental analyzer (TC/EA; Finnigan MAT, Germany) was coupled to a Delta plus XP isotope ratio mass spectrometer via a ConFlo III interface (Finnigan MAT, Germany; (Werner et al., 1999)).
The TC/EA was additionally equipped with a custom-made Nafion-trap followed by a 4-port valve (Werner, 2003) between the carbon reduction tube and the GC column. The set-up of the carbon reduction tube follows the "MPI-BGC method" described by Gehre et al. (2004). Water was injected automatically with a GC PAL autosampler (CTC, Switzerland) equipped with a 10 μl gas-tight syringe. Preparation for injection of 0.5 μl of water was made with three washing cycles (3 μl) and five pull-ups.
All results were normalized to VSMOW and SLAP, assigning consensus values of 0 and 55.5 ‰ for δ 18 O and 0 and 428 ‰ for δ 2 H to VSMOW and SLAP reference waters, respectively (Coplen, 1988).

. N 2 O emissions at the Hofen WWTP
The average N 2 O emissions of lane 2.1 and 2.2 at the Hofen WWTP were 0.8 kg N 2 O-N/d during the monitoring campaign ( Table 2). The resulting emission factor (0.2% of the total nitrogen load) is low compared to other WWTPs with full-year nitrification and denitrification (median: of 0.4%) (Gruber et al., 2021). Emissions from both lanes displayed similar temporal patterns, with high emissions in winter, and lower emissions during the summer season (Fig. 2). However, the emission pattern is not reproducible in different years. By far the highest N 2 O emissions were observed over several weeks starting in January 2021. The emission peak occurred in parallel with increased NO 2 − concentrations in the effluent of the WWTP, which is known to enhance N 2 O emissions via both nD and hD pathways and has been linked to emission peak phases in other WWTPs (Gruber et al., 2021b, Ren et al., 2019, Kuokkanen et al., 2021. In fact, all lanes were fully aerated during the peak emission phase to increase NO 2 − oxidation capacities of the biological treatment, which in turn favours N 2 O stripping and strongly lowers NO 2 − as well as N 2 O reduction capacities during denitrification. Consequently, during full aeration of Zone 1, emissions in all zones of both lanes increase. However, the major share of the emissions occurs in Zone 2 (Fig. 2), where likely most of the nitrogen turnover happens in case of full aeration of a lane. The detrimental effect of aeration of Zone 1 (in terms of N 2 O production) compared to anoxic operation was also shown in Campaigns 1 and 3, where the first zone of lane 2.1 or 2.2 were aerated (Table 2). Similarly, in April 2020 only Zone 1 of lane 2.1 was aerated, which led to substantially higher net N 2 O emissions as compared to lane 2.2 (Figs. 2, and 5).

. Identification of N 2 O production pathways using dual isotope mapping approaches
The isotope sampling campaigns at the Hofen WWTP were conducted during different seasons and day times, and under either oxic or anoxic operation of Zone 1 (Tables 1 and 2). The mean SP value for N 2 O emitted from oxygen-replete zones in all three experiment was -1.7 ± 2.7 ‰, which is somewhat lower than results (4.5 ‰) from a previous full-scale WWTP study (Toyoda et al., 2011) and literaure results for N 2 O from Ni, which yields consistently higher SP values (+32.0 to +38.7 ‰). However, values are fully in the range of isotopic signatures measured for nD and hD at a lab-scale WWTP (Wunderlin et al., 2013), as well as in pure culture studies (hD: -7.5 to +3.7 ‰, nD: -13.6 to +1.9 ‰) (summarized in Denk et al. (2017), Ostrom and Ostrom (2017), Yu et al. (2020)). In contrast, N 2 O liberated from Zone 1 under anoxic operation, using the anox tube, displayed significantly higher SP values of 12.3 ± 2.2 ‰. .
To evaluate the N 2 O production pathways during the experiments in more detail, we applied the dual isotope mapping approach, where SP values are plotted against either Δδ 15 N bulk (N 2 O, substrate) or Δδ 18 O (N 2 O, H 2 O) and compared to the isotope signatures known from literature for a given process   (Fig. 3). Wunderlin et al., (2013) followed this approach relating SP to Δδ 15 N bulk (N 2 O) values to verify process conditions that are most conducive to distinct production pathways (e.g., hD, nD, Ni) during batch experiments in a laboratory-scale reactor with activated sludge. Since no elevated SP was observed in the aerated zones, no significant contribution of Ni to N 2 O production was anticipated. Moreover, Δδ 15 N(N 2 O, NH 4 + ) values, which considers ammonium as a possible substrate, did not co-vary with the SP values towards Ni source endmember signatures (Fig. S.3).
Alternatively, Lewicka-Szczebak et al. (2016) (Fig. 3a) displays a considerable decrease in both SP and Δδ 18 O(N₂O, H 2 O) during the change from anoxic (Zone 1) to oxic (Zone 2) conditions. This corresponds to a decline in partial N 2 O reduction for Zone 2, in relation to Zone 1, as reduction of N 2 O to N 2 by hD increases SP of the residual N 2 O pool, since the 15 N-O bond is more stable than 14 N-O (summarized in Denk et al. (2017), Ostrom and Ostrom (2017), Yu et al. (2020)). Additional support for the concurrent reduction of nitrite and N 2 O through hD comes from the concomitant increase in δ 18 O(NO 2 -) and δ 15 N(N 2 O) shown in Fig. 3b.
Interpreting the Δδ 18 O(N₂O, H 2 O) signatures of N 2 O emitted in the aerobic zone (i.e., in parallel with low SP values) requires a more nuanced interpretation, but yields additional information. The Δδ 18 O (N₂O, H 2 O) value is controlled by both equilibrium isotope effects during O-exchange of precursors with water and branching isotope effects during O-abstraction (Casciotti et al., 2007;Casciotti et al., 2010;Kool et al., 2007). Both effects depend strongly on the bacterial community that performs denitrification, and can differ substantially among systems (Kool et al., 2007;Martin and Casciotti, 2016 (Buchwald and Casciotti, 2013) (Fig. 3b). Complete exchange can be associated with nitrite produced in nitrification (Buchwald et al., 2012;Casciotti et al., 2010), but can also be mediated by the iron-containing nitrite reductase NirS, which is present in many heterotrophic denitrifiers (Casciotti et al., 2007;Casciotti et al., 2002 Fig. 3) and thus a major contribution of hD. This result is also consistent with the observation of  Yu et al. (2020). The expected change in isotopic composition during partial reduction of N 2 O to N 2 is indicated by black "reduction lines". For panels (a) and (c), all data points fall on one line, while for panel (c) data points of individual days present individual reduction lines for Campaigns 1 and 3. Numbers next to data points of Campaign 3 (squares) indicate the sampling sequence (t1: 6 -7 am, t2: 8 -9 am, t3 = 10 -11 am, t4 = 1 -2 pm, t5 = 2:30 -3:30 pm). Orschler et al. (2021) Lewicka-Szczebak et al. (2016) were derived from soil systems that likely differ significantly in terms of the active microbial communities and expressed enzymes, as compared to wastewater systems (Wu et al., 2019). The prevalence of anaerobic hD under oxic conditions can easily be rationalized by anoxic microsites in sludge flocs even in aerated zones (Sexstone et al., 1985;Daigger et al., 2007). Nevertheless, given the variability seen in Δδ 18 O(N₂O, H 2 O), we cannot exclude a variable contribution from nD under certain conditions, which could be what drives difference between aerobic samples in Fig. 3a. Slightly lower SP and lower Δδ 18 O(N₂O, H 2 O) values may be due to an increased contribution of nD. Alternatively, the higher values may also be caused by a partial reduction of N 2 O also during aerobic phases, assuming that the organic substrate is not fully consumed in Zone 1 and leaks into Zone 2. Furthermore, N 2 O with a high SP and Δδ 18 O(N₂O, H 2 O) might be transported, and mixed in, from Zone 1, as discussed in Section 3.3 in more detail.
Plotting SP values relative to Δδ 15 N(N 2 O, NO 3 − ) indicates a higher variability among the three intensive sampling campaigns (Fig. 3d) and 2) versus 41.1 ‰ (Campaign 3), were possibly caused by concentration-dependent variations (affecting cell-specific rates) in the isotope effects associated with denitrification (Kritee et al., 2012). More precisely, the higher NO 3 − concentrations during experiment 3 (10-18 mg NO 3 − -N/L) compared to experiment 1 and 2 (0-7 mg NO 3 − -N/L) may manifest in substantially higher isotope effects. The increased nitrate concentrations were due to the full aeration of all zones over multiple weeks before experiment 3. The operation led to reduced denitrification activity and NO 3 − accumulation in the biological treatment.
Interestingly, Δδ 15 N(N 2 O, NO 2 − ) was more consistent than Δδ 15 N (N 2 O, NO 3 − ) between campaigns, i.e., isotope effects seemed less strongly affected by N substrate concentrations (Fig. 3c). Therefore, isotopic signatures for samples from aerated and anoxic compartments cluster significantly closer to the predicted reduction line (Fig. 3c). The observed correlation of delta values for individual campaigns hence supports the notion that the isotopic composition of NO 3 -, NO 2 − and N 2 O are mostly controlled by the sequential reduction of NO 3 − to N 2 during complete denitrification. In summary, the isotopic composition of N 2 O, NO 2 − , and NO 3 − consistently point towards a high contribution of hD to N 2 O production during aeration on all days. nD may be of variable relevance, yet Ni can be excluded as a significant contributor. hD was previously shown to govern N 2 O production during aeration under low C:N conditions (Domingo-Felez et al., 2016). Our data confirm that obligate anaerobic processes, such as hD, play an important role even during aerated reactor conditions, supported by strong oxygen gradients and anoxic microniches in sludge flocs (Daigger et al., 2007). For zones under anoxic process conditions, observed isotope patterns provide clear evidence for substantial N 2 O reduction. To diagnose the contribution of different production pathways, the relation of SP and Δδ 18 O(N₂O, H 2 O) turned out to be more sensitive than the Δδ 15 N(N 2 O, substrate) approaches. However, combining both approaches as shown here, has the benefit of being able to additionally validate interpretations, and to provide independent process information to assess the full complexity of concurrent N 2 O formation and reduction.

. Diurnal variation in N 2 O emissions and production pathways
The main focus of the third campaign was to investigate the effect of the diurnal patterns in N loading (controlled by reject water dosage) and COD substrate inflow on N 2 O emissions and variations in N 2 O reduction. For this, we analysed the isotopic signatures of N 2 O and nitrogenous substrates in Zone 1 and 2 for five different time points during one day at lane 2.1 (Fig. 4). N 2 O emissions exhibited a clear diurnal pattern, with a peak at 9 am, right before the reject water dosage was stopped (Fig. 4a). N 2 O concentration changes in the anoxic zone, measured with the anox tube, were consistent with changes in the N 2 O flux from Zone 2 and 3. While NH 4 + concentrations also exhibited a clear diurnal variation pattern, NO 3 − concentrations were relatively stable throughout the study period (Fig. 4c, Figs. S.4 and S.5, (SI)). NO 2 − was highest in Zone 1 and gradually decreased in Zone 2 and 3, respectively ( Fig. S.6 (SI)). The diurnal trend of the N 2 O site preference in Zone 1 indicates a decreasing importance of N 2 O reduction from 7 am to 9 am (sampling points 1 and 2), also shown in the dual isotope mapping approach, e.g., for SP vs. Δδ 18 O(N₂O, H 2 O) (Fig. 3a). After 10 am, SP and Δδ 18 O(N₂O, H 2 O) values for N 2 O from Zone 1 increased along the predicted reduction line, which suggests a return to an increasing relevance of N 2 O reduction for samples 3 to 5. NO 3 − concentrations remain stable in Zone 1 (Fig. 4c) despite an increase of NO 3 − inflow from the return sludge ( Fig. S.5 (SI)), confirming that heterotrophic nitrate reduction (hD) was very active after 9 am. We suggest two main causes for the strong daily variation in N 2 O emissions and N removal. First, the dosage of reject water and the morning peak in N inflow, typically seen in WWTPs, led to a NH 4 + concentration increase (Fig. 4c, t1 -t2), while the N 2 O reduction capacity of the WWTP was lower due to the increased supply of NO 3 − . Second, and more importantly, the availability of organic substrate typically exhibits daily fluctuations. Therefore, despite high NH 4 + loads from 10 am to 2 pm (t3 -t4), high availability of organic substrate led to increasing nitrogen removal and, in turn, increased fractional N 2 O reduction rates. Notably, COD concentrations were not measured during the campaign, but are expected to correlate with the inflow rate to the wastewater treatment plant, which exhibits reproducible daily variation ( Fig. S.7 (SI)). The N 2 O SP in Zone 2 is at its maximum between 6 and 9 am, probably due to transport of N 2 O produced in Zone 1, where both N 2 O production and reduction were high during this part of the diurnal cycle, as described above (Fig. 4b). This would imply that N 2 O emissions from Zone 2, before and during the peak phase, i.e., the end of the reject water dosage, comprise a substantial contribution of N 2 O from Zone 1. hD as the main source of this N 2 O is supported by the high Δδ 18 O(N 2 O, H 2 O) values (36.2 ± 2.3 ‰). Alternatively, high SP values in Zone 2 before 9 am can be explained by partial N 2 O reduction, but this is unlikely given COD limitation during reject water dosage. Moreover, transport of N 2 O produced in an anoxic zone to an aerobic zone has been reported earlier for other WWTPs (Mampaey et al., 2016). After 10 am, the difference in SP values between Zone 1 and 2 was increasing again, indicating that N 2 O transport and mixing was less important.
In addition, the contribution of nD to N 2 O formation might have increased after 10 am in Zone 2, which could further explain the lower SP and Δδ 18 O(N₂O, H 2 O) here. Nevertheless, we believe that hD also contributed a major part to the emissions in the aerobic zones between 11 am and 4 pm, given the still-high Δδ 18 O(N 2 O, H 2 O) values.

. N 2 O emissions depend on process operation
The seasonal dynamics in N 2 O emissions indicate that phases when the air consumption in Zone 3 exceeded a defined threshold, and thus when Zone 1 was aerated, were generally characterized by high net N 2 O production (Fig. 2). To better understand the effect of aerobic conditions in the first zone on overall N 2 O formation, we compared the isotopic signatures of N 2 O produced along a fully aerated lane (2.1) and a lane under standard operation, i.e., with anoxic conditions in the Zone 1 (2.2) ( Table 1). The episodes of reject-water dosage in the morning had a high impact on the emissions (i.e., high N 2 O emissions in Campaign 3), but N 2 O emissions were even higher from the fully aerated lane ( Table 2). The difference between lanes was primarily driven by emissions in Zones 1 and 2, while emissions in the third zone were comparable (Fig. 5a).
The explanation for increased N 2 O emissions from the fully aerated lane 2.1 can be assessed when comparing isotopic signatures of the N 2 O released from Zone 1 of both lanes (Fig. 5b, Campaign 1). The N 2 O isotopic signature measured in the Zone 1 of lane 2.2, with conventional operation, i.e., Zone 1 mostly anoxic, indicates a substantial reduction of N 2 O. In contrast, for lane 2.1, with Zone 1 aerated, the share of N 2 O reduction was substantially lower. The proportion of N 2 O reduction can be estimated quantitatively by the expression ΔSP = εSP x ln f (Jinuntuya-Nortman et al., 2008;Lewicka-Szczebak et al., 2017;Mariotti et al., 1981), with ε being the enrichment factor (-8.2 to -2.9 ‰, ), and f the fraction of unreacted N 2 O. The isotopic enrichment factor between product P and substrate S is defined as εX P/S = αX P/S -1 = δX P / δX S -1, where α is the isotopic fractionation factor. Applying this approach yields an estimate of 92% of N 2 O (84 to 99% using max and min fractionation factors) reduced for the anoxic Zone 1 of lane 2.2, while only 68% (56 to 90% using max and min fractionation factors) is reduced in the aerated Zone 1 of lane 2.1 (assuming that the SP values for N 2 O from Zone 2 are representative for the N 2 O production process). As during Campaign 3, N 2 O production was very likely driven by hD, given the increased Δδ 18 O(N 2 O, H 2 O) values (35.2 ± 0.6 ‰) in the aerobic zones.
Campaigns 1 and 3 revealed that organic carbon availability, aeration of Zone 1, and reject-water N dosage are the most important modulators of N 2 O emissions during standard operation at the Hofen WWTP, and at a given time of the year. Notably, emissions were lowest in Campaign 2 (Table 2), with anoxic conditions in Zone 1 of both lanes, without reject-water dosage and sampling times in the afternoon, where increased organic substrate concentrations are expected. While it seems relatively clear that aerobic conditions in Zone 1 and low organic substrate availability both lead to higher emissions by impairing a more efficient N 2 O reduction, the mechanism behind the increased production of N 2 O caused by elevated reject-water dosage (which leads to an increase in NH 4 + concentrations) is not fully understood (Gruber et al., 2020 it needs to be noted that the gas phase in the anox tubes applied in Zone 1 is not in equilibrium with the liquid phase. We anticipate a three times higher concentration under equilibrium conditions (Fig. S.2). (b) N 2 O SP in Zone 1 and 2, indicating a minimum in N 2 O reduction in Zone 1 around 9 am, while N 2 O SP in Zone 2 is generally low but increased at high concentrations in Zone 1 due to transport. (c) NH 4 + and NO 3 − concentrations in Zone 1 and 2 of lane 2.1 are stable despite higher NO 3 − inflow (Fig. S.5), pointing towards high denitrifying activity at 11 am. The gray shaded area shows the period of reject water dosage. The timing of gas and liquid sampling is indicated by markers in Fig. 4b and c: t1: 6 -7 am, t2: 8 -9 am, t3 = 10 -11 am, t4 = 1 -2 pm, t5 = 2:30 -3:30 pm. reject water is somehow unfavorable for heterotrophic denitrifiers and nitrifiers. Further research is needed to unravel underlying mechanisms, e.g., by comparing the effects of dosages of reject-water NH 4 + versus (NH 4 ) 2 SO 4 solution in activated sludge. Nevertheless, our results already yield important information regarding efficient strategies to reduce N 2 O emissions during normal operation at the Hofen WWTP. The guiding principle for the mitigation of N 2 O emissions is to maximize N 2 O reduction by avoiding aeration of Zone 1, and dosing reject-water primarily during periods with high organic carbon load, e.g. in the afternoon. The adaptation of the feeding strategy to optimize organic carbon utilization towards most efficient N 2 O reduction has been successfully applied in side-stream treatment (Peng et al., 2017). However, changing reject-water dosage operation strategies should be critically evaluated, as the effects of the NH 4 + loading are multifaceted. That is, besides potential impacts of the NH 4 + dosage on net N 2 O emissions, other constraints need to be considered. For example, increased NH 4 + peak concentrations can lead to NH 4 + breakthrough, and load equilibration in the diurnal pattern is beneficial for the nitrification performance (Meyer and Wilderer 2004). We propose to apply conventional activated sludge modeling and full-scale testing, combined with extensive process monitoring, to optimize reject-water dosage in terms of effluent quality and maximized reduction capacities for N 2 O mitigation (Henze et al., 2000). Isotopic technologies were successfully applied to analyze the contribution of N 2 O production pathways at the Hofen WWTP, and provided mechanistic understanding to support mitigation strategies. Still, long-term monitoring of the isotopic composition of N 2 O and other nitrogen species is needed in future studies to evaluate the consistency and robustness of the approach. A major advantage to characterize contributions of N 2 O reduction and production pathways at the Hofen WWTP involved the cascaded lanes, with clearly defined redox conditions in each zone. We expect that the application in flow-through, noncompartmented activated sludge systems can be more challenging due to increased mixing over a whole lane, leading to a higher exchange of the nitrogen pools. Furthermore, continuous long-term monitoring is important for the extrapolation and interpretation of the data and the characterization of the seasonal emission peaks. The lion's share of the total annual N 2 O emissions can be attributed to the January peak emission period ( Fig. 2; 50% of the total emissions) in association with elevated NO 2 − concentration levels. Seasonally impaired NO 2 − oxidation in WWTPs, leading to NO 2 − accumulation, has been linked to low abundances of nitrite oxidizing bacteria (NOB) and drastic changes in the whole activated sludge microbial community (Gruber et al., 2021). However, the NOB loss observed by Gruber et al. (2021) at the Uster WWTP led to NO 2 − accumulation over a periods of 1-2 months, and it is unclear whether similar process were also responsible for the accumulation of nitrite over a few weeks at the Hofen WWTP.

Summary and conclusions
-Measurements of relative 15 N and 18 O abundances in nitrogenbearing molecules were successfully applied to characterize dynamics of N 2 O formation pathways under normal operation in a fullscale activated sludge WWTP. N 2 O was mainly produced by heterotrophic denitrification, while nitrifier denitrification appeared to be less important and of rather variable influence; NH 2 OH oxidation was negligible. -Seasonal emission peaks occurred during winter when NO 2 − accumulates, and when the biological treatment is operated at full aeration, but NOB activity is still impaired. -Based on N 2 O isotopic measurements, N 2 O reduction was identified under anoxic conditions, and to lesser extent also under oxic conditions, when it is restricted to anoxic micro-niches. Fractional N 2 O reduction was most pronounced under organic-substrate-replete conditions, while N 2 O accumulation in the anoxic zone was primarily observed when organic substrate was limiting. Hence, the daily variation of organic substrate has a strong impact on the reduction of N 2 O, and in turn, diurnal N 2 O emission fluctuations. -The dosage of reject-water and full aeration of the biological treatment significantly increased N 2 O emissions, since N 2 O reduction was strongly impeded. Hence, an efficient mitigation strategy towards optimized N 2 O reduction may involve shifting reject-water dosage to periods with high organic substrate availability, as well as avoiding full aeration of the biological treatment. -Coupling isotopic technologies with continuous long-term monitoring of N 2 O emissions is a powerful tool for qualitative N 2 O pathway identification and the development of N 2 O mitigation strategies in full-scale WWTPs. However, clearly defined conditions in a reactor system are required to interpret the data.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.