Unwanted mainstream nitritation-denitritation causing massive N2O emissions in a continuous activated sludge process

Nitrous oxide emissions can contribute signi ﬁ cantly to the carbon footprint of municipal wastewater treatment plants even though emissions from conventional nitrogen removal processes are assumed to be moderate. An increased risk for high emissions can occur in connection with process disturbances and nitrite (NO 2 (cid:1) ) accumulation. This work describes the ﬁ ndings at a large municipal wastewater treatment plant where the levels of NO 2 (cid:1) in the activated sludge process ef ﬂ uent were spontaneously and strongly increased on several activated sludge lines which was suspected to be due to shortcut nitrogen removal that stabilized for several months. The high NO 2 (cid:1) levels were linked to a dramatic increase in nitrous oxide (N 2 O) emissions. As much as over 20% of the daily in ﬂ uent nitrogen load was emitted as N 2 O. These observations indicate that highly increased NO 2 (cid:1) levels can occur in conventional activated sludge processes and result in high nitrous oxide emissions. They also raise questions concerning the risk of increased greenhouse gas (GHG) emissions of the nitritation – denitritation processes – although the uncontrolled nature of the event described here must be taken into consideration – and underline the importance of continuous monitoring and control of N 2 O emissions.


INTRODUCTION
Nitrous oxide (N 2 O), a gas formed in varying amounts in nitrogen transformations and wastewater treatment, strongly promotes global warming (IPCC ).N 2 O emissions have been recognized as major contributors to the carbon footprint of municipal wastewater treatment plants (WWTPs) in a study of 16 Scandinavian treatment plants (Gustafsson & Tumlin ).In a long-term study by Daelman et al. (), 75% of a treatment plant's carbon footprint was due to direct N 2 O emissions.At the Viikinmäki WWTP, the measured direct N 2 O emissions were calculated to contribute over 60% of the total carbon footprint (Mölsä ).
Shortcut nitrogen removal processes have been thoroughly investigated in recent years due to energy and chemical savings they offer.Highly varying N 2 O emission factors have been observed in nitritation-denitritation processes (Vasilaki et al. ; Li et al. ).In addition to differences between processes, emissions can vary in single processes in different process conditions.In a study by Peng et al. (), the N 2 O emission factors in a nitritationdenitritation process treating co-digested pig slurry centrate could be decreased from 17 to 0.12% by the choice of aeration and feeding strategies.Vasilaki et al. () observed N 2 O emissions varying from 1.3 to 19% of the influent NH 4 -N load on a SCENA-process which combines nitritation-denitritation and biological phosphorus removal.
Currently, altogether three pathways have been identified as contributing to N 2 O production in a wastewater environment.Two pathways for N 2 O production in nitrificationthe hydroxylamine pathway and the nitrifier denitrification pathwayhave been recognized and either one or both can be used to model processes, depending on the process conditions, such as NO 2 À and DO (Peng et al.

).
In heterotrophic denitrification N 2 O is an intermediate and can therefore be consumed or accumulated.Elevated N 2 O emissions are typically linked to various different factors such as low dissolved oxygen (DO), high nitrite (NO 2 À ) or nitrous acid concentration, varying NH 4 þ concentrations, low chemical oxygen demand (COD) alternating anoxic and aerobic conditions and abrupt changes (Vasilaki et al. ) as different operational conditions promote different pathways.
Earlier studies at the Viikinmäki WWTP have shown large variations of N 2 O emissions (Kosonen et al. ).Some level of correlation of emissions with total nitrogen removal in AS has been observed but the earlier, lower peak values have remained unexplained (Blomberg et al. ).Based on a study of the denitrifying filters at the Viikinmäki WWTP (Heinonen et

WWTP characteristics
All data presented are from on-line measurements and laboratory analyses at the Viikinmäki municipal WWTP in Helsinki, Finland, which has a population equivalent (PE) of over one million.The treatment plant is situated underground with all exhaust air from the wastewater treatment process passing through one exhaust air pipe enabling the measurement of total N 2 O emissions.
Viikinmäki WWTP is a conventional AS plant with a mechanical pre-treatment, seven primary settling lines and nine AS lines (lines 1-9) and ten post-denitrifying filters.The AS lines are pre-denitrifying, comprising anoxic and aerobic zones and nitrate recirculation (100-300% of influent flow).Phosphorus removal is based on chemical precipitation with ferrous sulfate.Calcium hydroxide is used for pH control in AS and methanol is used as a carbon source in the post-denitrifying filters.
Each 11,500 m 3 AS line is divided into six equally sized zones.The aerated volume of each line is controlled automatically to optimize ammonia and total nitrogen removal.The first zone is anoxic, zones 2 and 3 can be mixed or aerated, controlled by on-line ammonia measurement in zone 6, and zones 4-6 are aerated.The oxygen set-points can be constant or controlled by ammonia measurement in zone 6.Each AS line has a separate return sludge recirculation (approximately 70% of influent flow), but excess sludge is pumped to the beginning of the process and removed in the primary settling and there can thus be some exchange of bacteria between the AS lines.
The treatment plant has mesophilic digestion of sludge.15-20% of reject waters from sludge centrifuges, equivalent to 300 kg NH 4 þ -N/d, are treated in a pilot deammonification moving bed biofilm (MBBR) process (ANITAMox™).The treated reject water is normally led to primary settling line 1, where the wastewater flows mainly to AS line 1 and partly to AS line 2. The untreated reject water is distributed equally to all treatment lines.Viikinmäki WWTP has an environmental permit that states the requirements for effluent quality for total phosphorus, total nitrogen, biological and chemical oxygen demand (BOD and COD) and suspended solids (Table 1).There are no limit values for greenhouse gas emissions, but the yearly total emissions are reported.
In 2019 the average influent flow was 290,000 m 3 per day.The wastewater temperature varied mainly from 10 to 20 C. The influent loads, environmental permit limit values for effluent concentrations and removal rates and the treatment results in 2019 are presented in Table 1.The total nitrogen removal after AS was 69% as a yearly average.

On-line measurements, sampling and analysis
The data presented are based on measurements in the AS process and of the total exhaust air of the whole treatment plant.Parameters are mainly measured from all individual AS lines.However, liquid phase N 2 O is measured only on AS lines 5 and 9 and until April 2019 NO 2 À -N data are available only from AS lines 1 and 2. The data presented in Figures 3, 5 and 6 are based on these limitations.The AS process is monitored with both on-line measurements (Table 2) and laboratory analyses.
The total N 2 O emissions of the wastewater treatment process together with CO 2 , NO, NO 2 , NH 3 and CH 4 have been measured since 2012 using the Gasmet CEMS Fourier transform infrared (FT-IR) system at the only exhaust air pipe where all the air from the underground treatment plant is collected.The measurements represent the total emissions of the process excluding only possible leakages in the above-ground biogas treatment.At the end of January 2017, after mass balance calculations of carbon dioxide production and smoke tests, the sampling point was moved upwards in the exhaust air pipe for a more representative sample.
All the above gas phase and liquid phase on-line measurements, together with wastewater temperature, flows and air flows etc. are recorded in the process control system (Valmet DNA).
Laboratory analyses are carried out twice a week from flow-weighted 24-hour samples collected automatically (Endress þ Hauser Liquistation CSF33).Sludge concentrations are analyzed from grab samples.AS effluent   In April 2019 the N 2 O emissions at Viikinmäki WWTP exceeded many times all previous values since the beginning of monitoring (Figure 1).Also, exceptionally high NO 2 À -N concentrations were observed in several treatment lines, up to 80% of NO X -N (Figures 2-4).Typical NO 2 À -N levels since the beginning of measurements in 2016 have been at highest 11% and mainly below 5% of NO X -N, with concentrations mostly below 1 mg/L (Figure 2(a), 2(c) and 2(d)).
The highest daily average N 2 O emissions observed during the peak emission period were 5,540 kg N 2 O/d (3,520 kg N 2 O-N/d), 20 times higher than the yearly average emissions of 2018.The total yearly emissions, which were 93.6 tn N 2 O in 2018 were more than doubled to 227 tn in 2019.Also, nitrogen monoxide (NO), normally under detection limit, could be observed in the exhaust air in April and May 2019, with the highest daily average emission value of The NO 2 -N values before April 2019 are the average of AS lines 1 and 2. They are considered to be representative for the whole AS process until the beginning of the abnormally high N 2 O emissions, based on essentially similar nitrification results between treatment lines.In April 2019, the values from only two treatment lines are deemed to be unrepresentative for the whole AS process (second marked period in Figure 2(a)).In July 2018 there was a two-week period of very poor nitrification due to an unknown    until 25 April 2019 are included as there are no NO 2 À -N data available from lines 5 and 9 before that date.The AS lines 5 and 9 are both equipped with two N 2 O probes, and a clear difference in their liquid phase N 2 O could be seen from April to August 2019 (Figure 3).
It can be seen in Figure 3 that after the peak flows both the NO 2 À -N concentrations in lines 1 and 2 and the liquid phase N 2 O concentrations in lines 5 and 9 were increasing.
It can be assumed that the NO 2 À -N concentrations were rising in a similar way in all treatment lines, which is supported by the high NO 2 À -N values observed after the NO 2 À -N monitoring was expanded.
After the initially similar rise of N 2 O concentrations in treatment lines 5 and 9, high peaks and a deterioration of ammonia removal (Figure 5(a) and 5(b)) could be observed only in line 9.After mid-May the peaks were diminished but the N 2 O base level concentration increased in line 9, possibly indicating a shift in the bacterial composition or behavior, before decreasing rapidly in August, together with the NO 2 À -N concentrations (Figure 3).
Most N 2 O emission and concentration peaks could be linked with a higher aerated volume, but the peak in 24.6.(Figures 2(a) and 3) could not.The influent load was lower than usual due to holidays and a low COD/N ratio was suspected to be the cause.
The different fractions of nitrogen in AS effluent and exhaust air during sampling days in 2019 are presented in Figure 4.The organic N in AS effluent represents the difference between the measured values of total nitrogen and NO X -N and ammonia nitrogen.N 2 O in AS effluent is included in the organic N, as there are liquid phase N 2 O data only from two AS lines.The amount of nitrogen The observed total nitrogen removal in AS during the period of increased NO 2 À concentrations was high, which is in accordance with a lower consumption of organic carbon in nitritation-denitritation.However, particularly during the highest peak emissions, a considerable portion of the observed nitrogen removal was due to the stripping of N 2 O (Figure 4).The highest daily N 2 O emission on a sampling day was 4,570 kg N 2 O (2,910 kg N 2 O-N) on 15 April 2019, equivalent to 20% of the day's AS influent load of 14,100 kg N. The all-time highest daily emission of 5,540 kg N 2 O (3,520 kg N 2 O-N) took place on the previous day, which was not a sampling day and thus not included in in Figure 4.

Two suspected periods of stable shortcut nitrogen removal
Two periods with prolonged nitrite accumulation have been observed at the Viikinmäki WWTP in recent years, suspected to indicate nitritation-denitritation in AS.In 2019 nitrite accumulation was observed on seven out of nine AS lines.At treatment line 9, which was the first one to exhibit elevated concentrations of liquid phase N 2 O and the last one to normalize, the suspected nitritation-denitritation with high ammonia removal lasted as long as over three months.
A corresponding period of nitrite accumulation was observed also earlier, on only one treatment line in 2016 (Figure 5(a) and 5(b)).
In April 2019 the ammonia removal rate in AS was initially decreased, in all AS lines except 4 and 5, but ammonia removal recovered to mainly well over 95% while up to 80% of the NO X -N was in the form of NO À is in the form free nitrous acid (FNA), which can have an inhibitory effect on several microbial processes in concentrations as low as app.0.02 mg N/L.The initial decrease in ammonia removal as well as the observed high total nitrogen removal rate compared to treatment lines 4 and 5 (Figures 3 and 5) indicate that the process was exhibiting shortcut nitrogen removal.Also, if NO 2 À was accumulated in denitrification and not due to NOB inhibition, it should mainly be transformed into NO 3 À during aeration and not be seen in the samples.
The NO 2 -N levels were high in all treatment lines except 4 and 5, when analyses began.An increase in effluent ammonia could also be observed in all other AS lines except 4 and 5, from 2 to 10 days after line 9 (data not included).When the regular laboratory analyses of NO 2 À -N from all AS lines were started on 25 April 2019, ratios of NO 2 À -N to NO X -N on individual AS lines varying between 58 and 70% were observed with NO 2 À -N concentrations mainly 4-6 mg/L.In treatment lines 4 and 5 only a modest and short-term increase of NO 2 À -N with a maximum NO 2 À -N to NO X -N ratio of 13% in both AS lines, as well as an increase in aqueous N 2 O in line 5 could be observed, and ammonia removal remained high.
A change in the sludge composition could be observed in all the treatment lines exhibiting nitritation-denitritation, observed as increased turbidity and effluent total phosphorus (Figure 6(a)) and decreased filament and sludge volume indexes (data not included).High turbidity could also be visually observed on line 1 in 2016.
Sludge transfer from lines 4 and 5 to others was started during summer to promote NO 2 À removal.In line 9, however, the NO 2 À -N concentrations began decreasing in late August, just before it was due for sludge transfer, and it was left to recover on its own.
In August 2016 the nitrification rate dropped suddenly in treatment line 1 (Figure 5(b)).NO 2 À -N analyses from AS effluent in lines 1 and 2 were started in October 2016.It was discovered that in AS line 1 effluent, roughly 75% of NO X -N was NO 2 À -N while this was only 5% in line 2 The process behavior normalized roughly a month after the start of NO 2 À -N measurements.When comparing ammonia and total nitrogen removal between treatment lines in 2016 and 2019 (Figure 5) it appears likely that the nitrite accumulation on line 1 in 2016 had begun in August and thus lasted in total almost three months.
There are not enough data to evaluate the effect of NO 2 À accumulation on N 2 O emissions in 2016 as there are no liquid phase N 2 O measurements in line 1.The peak values of total N 2 O emissions were considerably lower than in 2019, when seven out of nine AS lines were affected, and there had been peak values also earlier that year (Figure 1).The highest daily N 2 O emissions in 2016, however, occurred during NO 2 À accumulation in line 1.
Evaluating the impact of process conditions on NO 2

À accumulation
The process conditions, such as influent flow, temperature and pH in the AS process during the initiation of the nitritation-denitritation period had no apparent similarities in 2016 and 2019 (Figure 6).In general, the process conditions at the Viikinmäki WWTP should promote full nitrification.
As FNA is known to have an inhibitory effect on NOB (Zhou et al. ), it can speculated if once over a critical level, the accumulated NO 2 À and FNA sustained the nitritation-denitritation process via NOB inhibition when NO 2 -N was mainly 4-6 mg/L and pH below 6.5.However, the reasons for the initial triggering of the critical accumulation on some but not all treatment lines are not clear.The process conditions are scrutinized in more detail in the following chapters.

Aerated volume and the possible inhibition of AOB and NOB
In normal process conditions, the aerated volume of each AS line -50%, 67% or 83% of the total volume, equivalent to 3, 4 or 5 aerated zonesis controlled by automation based on the on-line ammonia measurement.The daily average air flows in aeration vary mainly between 6 and 13 nm 3 /s according to variations in load and nitrification result, with the highest air flows associated mainly to high aerated volume, peak flows and cold influent, and the lowest values to late summer and early autumn with warmer wastewater and high nitrification efficiency.After the high NO 2 À -N concentrations were observed in 2019, it was noted that ammonia removal was not improved when the aerated volume was increased from 67 to 83% and the N 2 O emissions increased significantly (data not included).The maximum aerated volume was manually limited to 67%, to both decrease NO 2 À -N concentrations in hope to mitigate inhibition by increasing the consumption of NO 2 À in the unaerated zones, and to reduce stripping of N 2 O.A stricter limitation on the aerated volume to 50% of total volume was also tested on some treatment lines but it resulted in an increase in effluent ammonia.The total sludge age was increased to prevent a decrease in the aerobic sludge age.Before the sharp increase in N 2 O emissions and NO 2 À -N concentrations in 2019, the measured NO 2 À -N concentrations were already somewhat elevated, most likely in all treatment lines (Figures 2(a) and 3).When the effluent ammonia of an AS line is increased, the following automated increment of the aerated volume both increases the NO X concentrations and decreases pH.It can be speculated that when combined with the automated control, initially moderate differences in the nitrification efficiencies and concentrations between AS lines, may have been enough to result in an ultimately completely different process performance.

Dissolved oxygen
Low DO has a role in both promoting nitritation-denitritationwhere a range of 1.0-1.DO was <1.5 m/L but not when DO was >1.5 mg/L.The measured oxygen concentrations in the constantly aerated zones 4-6 of AS line 9 in 2019 and AS line 1 in 2016 were mainly well over 2.5 mg/L (Figure 6(c) and 6(d)).Temporary low oxygen concentrations can occur at the beginning and end of aerated periods in zones 2 and 3 and also occasionally during peak loads.However, as separate aerated periods in zones 2 and 3 occur mainly one to three times per day and they last for several hours, frequent or long-term DO values below 1.5 mg/L would require a DO measurement error, which is unlikely to happen in several treatment lines simultaneously.Thus, low DO does not appear to be an explanatory factor for either triggering or sustaining the nitritation-denitritation. The pH values in the last aerated zones of the AS process vary normally mainly between 6.0 and 6.3 and the daily average alkalinity values are normally above 1 mmol/L.In acidic conditions the portion of FNA is higher and thus the inhibitory effect of NO 2 À accumulation is more pronounced (Zhou et al. ).

Alkalinity and pH
In 2016, the pH values temporarily dropped below 6.0 (Figure 6(d)) and the daily average alkalinity to 0.6 mmol/l due to an on-line measurement error which had led to insufficient dosage of calcium hydroxide, and there was a clear temporal connection between the decreased pH and alkalinity and the deterioration of nitrification in treatment line 1.
In 2019, such a connection with decreased pH or alkalinity before the rise in NO 2 À -N and N 2 O levels could not be made (Figure 6(c)).

Impact of side-stream deammonification effluent
The maximum direct impact of the side-stream deammonification process on the N 2 O emissions is limited, due to its size, but a possible connection with the NO 2 À accumulation in 2016 has been suspected.During the 2016 NO 2 À -N accumulation in treatment line 1, the only distinct difference observed between line 1 and other lines was that most of the deammonification process effluent was led via primary settling mainly to AS line 1.At the time, the pilot was in the start-up stage and treating roughly half of its design load.Primary settling should, however, decrease considerably the transfer of bacteria to AS.In 2019 AS line 1 was neither the first nor the last to switch to nitritationdenitritation.

CONCLUSIONS
Two separate incidents of prolonged NO 2 À accumulation and suspected nitritation-denitritation were observed in part of the AS lines at the Viikinmäki WWTP, while other lines kept on performing full nitrification in seemingly similar process conditions.At least one of the incidents could be linked with a significant increase in N 2 O emissions.Longterm data show a correlation between NO 2 À -N and N 2 O emissions also in normal process conditions but particularly during peak emissions.
It was observed that once initiated, NO 2 À accumulation was able to stabilize for a long period in process conditions that normally promoted full nitrification, and it was suspected to be due to NOB inhibition by NO 2 À or FNA.
The N 2 O emissions observed at the Viikinmäki WWTP indicate a risk of considerable emissions in nitritation-denitritation processes, if they are not properly controlled, which can outweigh the positive environmental impacts of energy and chemical savings.Implementing shortcut nitrogen removal processes should include monitoring the N 2 O emissions to avoid adverse environmental impacts.
In addition to measuring gas and liquid phase N 2 O, NO 2 À -N analyses can be used as an indicator of the longterm variations in N 2 O emissions.NO 2 À -N analyses can also reveal abnormal process conditions and provide new insight into the mechanisms of deterioration of nitrification and sludge composition.

ÀFigure 1
Figure 1 presents the monthly variations in total N 2 O emissions (kg/d), calculated from FT-IR measurements from the exhaust air starting from the year 2013.The exhaust air flow rate has been measured starting from August 2017, with values varying mainly between 100 and 130 m 3 /s.An estimated constant air flow of 120 m 3 /s was used in emission calculations for February 2017 to August 2017 and a 17% reduction of the air flow was used before February 2017, to compensate for the estimated error caused by the placement of the N 2 O analyzer sampling point.The data presented in Figure 1 represent an overview of the longterm monitoring.Even though data before and after February 2017 are not completely comparable, the possible error is insignificant in comparison of the magnitude of the peak emissions in 2019.
Varying zones, lines 5 and 9 Liquid phase N 2 O b 0.01-500 μM Unisense Wastewater System a Sum of NO 3 -N and NO 2 À -N.b Two probes per line, AS line 9 since April 2016 and AS line 5 since November 2018.

Figure 1 |
Figure 1 | The monthly variations of the daily average N 2 O emissions (kg/d) at the Viikinmäki WWTP since January 2013 as a boxplot.Average values are marked with an X. Horizontal lines indicate the median values, and dots indicate the outliers (values exceeding the interquartile range with more than 50%).

Figure 3 |
Figure 3 | Plant influent flow, NO 2 À -N concentration in AS effluent in treatment lines 1, 2, 5, and 9 and soluble N 2 O in the last aerated zones in lines 5 and 9.

Figure 4 |
Figure 4 | The measured and calculated fractions of nitrogen in AS, the average of all treatment lines and total exhaust air, during laboratory sampling days in 2019 (N ¼ 96).
2 À -N (AS line 9 results are shown in Figure 5(a)).Faster recovery of ammonia oxidizing bacteria (AOB) compared to nitrite-oxidizing bacteria (NOB) was also observed by Chen et al. (), who used heat shocks to initialize nitritation-denitritation in an SBR-process, with both AOB and NOB initially suppressed but AOB recovering faster resulting in NO 2 À accumulation.According to Zhou et al. () nitrite accumulation can also be due to inhibition or unbalance of denitrification and a pH-and temperature dependent portion of NO 2

Figure 5 |
Figure 5 | The portion of NO2 À -N in AS effluent and ammonia and total nitrogen removal (a) on AS lines 5 and 9 in 2019 (data available from 25.4.2019) and (b) on lines 1, 2 and 4 in 2016.(N 2 O data available from 12.10.2016on line 1 and 26.10.2016 on line 2.) Line 2 was emptied for maintenance from mid-July to mid-September.

Figure 6 |
Figure 6 | Liquid phase N 2 O in 2019 and NO 2 À -N in 2019 and 2016 together with different process parameters.All values other than NO 2 À -N are on-line measurements.
Both low and high pH and the presence of FNA in low pH values and of free ammonia (FA) in high pH values have been linked to NO 2 À accumulation in different studies according to Philips et al. ().
al. ), the activated sludge (AS) process was identified as the main source of N 2 O emissions, and a modelling study of the N 2 O emissions produced in the AS (Blomberg et al. ) supported these observations.While NO 2 technologies in promoting nitritation-denitritation and suggest controlling pH and DO, as many of the other possible control measures require sequencing operation, high temperature (>25 C) or high ammonia load, none of which occur at the Viikinmäki WWTP during normal operation.

Table 1 |
Viikinmäki WWTP influent loads, environmental permit limit values and yearly average treatment results in 2019

Table 2 |
On-line measurements in AS 5 mg/L has been suggested (Peng & Zhu )and in N 2 O production (Kampschreur et al. ; Vasilaki et al. ).