RESEARCHING NEW WAYS TO REDUCE N 2 O EMISSION FROM A GRANULAR SLUDGE SEQUENCING BATCH REACTOR TREATING DOMESTIC WASTEWATER UNDER SUBTROPICAL CLIMATE CONDITIONS

- N 2 O emissions from wastewater treatment plants have become an important issue, since this compound is a significant greenhouse gas that affects the sustainability of sewage treatment. This work aimed to investigate and to reduce N 2 O emission from a pilot-scale aerobic granular sludge sequencing batch reactor (AGS-SBR) operated for carbon and nitrogen removal from domestic wastewater under subtropical climate condition. Three operational strategies (S-I, S-II and S-III) with different anoxic phase durations were compared regarding treatment efficiency and N 2 O emission. For all the studied strategies, volatile suspended solids (VSS) was between 1.0 and 1.2 g/L. S-III, with the longest anoxic phase, obtained the highest biological oxygen demand (BOD) and NH 4+ –N removal efficiencies (86% and 84%, respectively), the lowest N 2 O emission factor (16.99 gN 2 O-N/person·year) and the lowest total nitrogen (TN) to N 2 O conversion ratio (0.47%). The results indicated that the extension of the anoxic phase was an effective way to significantly reduce N 2 O emission and to improve treatment efficiency .


INTRODUCTION
Production and emission of greenhouse gases (GHG) from wastewater treatment plants (WWTP) is a very important issue, which is becoming increasingly significant (Foley et al., 2010;Mannina and Cosenza, 2015). In particular, N 2 O emission in wastewater treatment systems with biological nutrient removal deserves special attention, since this compound is one of the main GHG, and its global warming potential is about 300 times higher than that of CO 2 (Yang et al., 2013).
For over a century, the conventional activated sludge process has been a standard model of biological wastewater treatment systems. However, this technique has as main drawbacks the poor settling properties of the biomass, which is in the form of flocs and can compromise the quality of the final effluent, and the requirement of large areas to place the secondary clarifiers (Liu and Tay, 2004). Sequencing batch reactors (SBR) with aerobic granular sludge (AGS) are presented as a promising option for the biological treatment of domestic and industrial effluents, due to the efficiency and robustness of this type of system (Di Bella and Torregrossa, 2013;Moreira et al., 2015;Pronk et al., 2015). With this technique, it is possible to obtain high removal rates of organic matter and nutrients in a single reactor, producing a final effluent with high quality (De Kreuk et al., 2005;Liu et al., 2015). However, the dynamics of nitrogen removal in this process, including nitrous oxide emission, are not completely understood; thus, further research is required.
In the current scientific literature, there are few studies reporting N 2 O emission by biological treatment systems using aerobic granular sludge. The few studies that address this topic are generally related to systems fed with synthetic wastewater (Rathnayake et al., 2013;Wei et al., 2014), being even more scarce reports concerning the emission of N 2 O by granular sludge fed with real wastewater (Castro-Barros et al., 2015), especially under tropical and subtropical climate conditions.
Wide variations concerning N 2 O emissions have been reported in scientific literature, as it can be affected by many factors. Conversions of influent nitrogen to N 2 O reported by Kampschreur et al. (2009) ranged from 0.001 to 14.6%. Moreover, a national survey conducted by Ahn et al. (2010) in the United States listed conversions ranging from 0.01 to 1.8% of the influent nitrogen. The N 2 O emission factor from a municipal activated sludge WWTP reported by Daelman et al. (2013) was 163.2 g N 2 O-N/person·year, which is 80 times higher than the EF proposed by the IPCC (2006). Thus the importance to estimate a N 2 O emission factor for each specific condition.
With this work, the aim was to advance the knowledge concerning the behavior and the rates of N 2 O emission in a pilot-scale SBR with AGS fed with domestic wastewater under subtropical climate conditions. The objectives of this research were: i) to develop aerobic granular sludge in a pilot SBR fed with municipal sanitary wastewater; ii) to evaluate the performance of the treatment; and iii) to monitor, to quantify and to compare N 2 O emissions from the SBR operated under three cycle configurations.

Pilot reactor operation and cycle configurations
The pilot reactor studied was an acrylic bubble column 2.18 meters high and with a 25 cm internal diameter, which worked as a SBR. The working volume was 98 L and the reactor outlet pipe provided the exchange of 56% of the liquid volume, in order to treat 55 L of influent during each cycle. The reactor was operated at room temperature (21-26°C) without pH control.
The reactor was operated under three different strategies: Strategy I (S-I), Strategy II (S-II), and Strategy III (S-III), each one with different anoxic phase duration. The cycle phases configuration of the three strategies are shown in Figure 1. During the anoxic phase of S-III, in order to maintain the sludge dispersed through the liquid column when aeration was off, there were aeration pulses of 10 seconds each 15 minutes, which were short enough to keep a very low dissolved oxygen (DO) concentration in the mixed liquor. The variation of the anoxic and aeration phase durations was tested to verify its influence on the treatment efficiency and on the N 2 O emission. The SBR cycle configuration corresponded to a hydraulic retention time (HRT) of 7.13 hours for S-I and S-II, and 10.69 hours for S-III.
These 3 cycle configurations were adopted in order to favor granular sludge formation and accumulation in the reactor, and to compare the SBR performance under different regimes. The intention was to obtain aerobic granules with a gradient of oxygen through them, providing aerobic/anoxic layers and the occurrence of simultaneous nitrification-denitrification (SND).
In this study, real domestic wastewater was used to feed the system (influent characterization is in Table  1). The wastewater was collected from an inspection chamber of the public sewage system, located in the city of Florianópolis, state of Santa Catarina, south of Brazil (27°36'12.7"S; 48°31'14.9"W). During the SBR start-up period, wash-out conditions were applied to favor the aerobic granules formation. The selective conditions applied were the shear force caused by the aeration, maintained at 32 L air /min, and the reduced settling time (15 minutes). The settling velocity imposed to the biomass to be retained inside the system was 7.47 cm/min.

Analytical procedures
The sludge settleability was assessed through sludge volume index (SVI), which was calculated at times of 5, 10 and 30 minutes (SVI 5 , SVI 10 and SVI 30 , respectively) (Schwarzenbeck et al., 2004). Granulometric analyses were performed through laser diffraction (Mastersizer 2000 -Malvern Instruments, UK), and optical microscopy (Olympus BX40, Japan) was used to observe granules formation. According to Liu et al. (2010), the sludge is considered granular when at least 50% of the biomass presents diameter over 200 µm.
The performance of the wastewater treatment was analyzed in terms of carbon and nitrogen removal. Biochemical oxygen demand (BOD), chemical oxygen demand (COD) and total suspended solids (TSS) were quantified according to Standard Methods (APHA, 2005). Nitrite and nitrate were quantified by ion chromatography (DIONEX ICS-500). Total nitrogen (TN) and ammonium (N-NH 4 + ) were quantified using Hach ® test kits (High Range Nitrogen-Ammonia Reagent Set: #2606945; High Range Total Nitrogen Reagent Set: #2714100). A multiparameter probe (YSI 6820, USA) was used to measure dissolved oxygen (DO) concentration regularly.
To perform N 2 O measurements, the upper part of the reactor was sealed with an airtight lid, which would only allow the air flow through a PVC hose connected to it. The airflow was continuously monitored by a gas analyzer (Guardian SP -Edinburgh) based on dual wavelength infrared sensor technology.

N 2 O emission quantification
For purposes of quantifying N 2 O emission, a graph of N 2 O concentration versus time was plotted, and the area under the curve was calculated using a definite integral. The total amount of N 2 O released in one SBR cycle was calculated according to Eq. 1.

( )
where m Nd is the amount of released nitrogen in one SBR cycle (g N); TN I the total nitrogen influent concentration (mg N/L); TN E the total nitrogen effluent (mg N/L); V te is the volume of wastewater treated in one SBR cycle (L).
The conversion of total influent nitrogen to N 2 O was calculated by relating the amount of N 2 O-N released in a SBR cycle, the TN influent concentration and the volume of wastewater treated in a SBR cycle (Eq. 3). (1) (2) where ConversionN inf → N 2 O -N is the fraction of influent nitrogen converted to N 2 O (%); m N 2 O-N is the amount of N 2 O released in a SBR cycle, expressed in terms of nitrogen (g N 2 O-N).
The N 2 O Emission Factor (EF) was calculated by relating the amount of released N 2 O, the daily per capita wastewater generation and the volume of wastewater treated in one SBR cycle, according to Eq. 4. BLASTN 2.2.28 against GreenGenes 13.8 database.
To attribute taxonomy, only sequences with hits higher than 99% of identity in alignment were considered.
The fluorescent in situ hybridization (FISH) technique was used to investigate the microbial community of the granular sludge samples from S-I, S-II and S-III. Granular sludges sampled from the reactor were fixed in 4% paraformaldehyde solution for 2-3 h at 4°C. The sludge samples were rinsed twice with phosphate-buffered saline (PBS) and then dehydrated by successive 50%, 80%, and 98% ethanol washes (Amann, 1995). In situ hibridizations were performed using the specific probes NSO190, Ntspa662 and PAE997 for ammonium-oxidizing bacteria (AOB), nitrite-oxidizing bacteria (NOB) and Pseudomonas genus, respectively. Oligonucleotides were synthetized and fluorescently labeled with a hydrophilic sulfoindocyanine dye (Cy-3) at the 5' end. Details on oligonucleotide probes are available at probeBase.

Biomass characteristics and composition
AGS was successfully cultivated in the SBR fed with domestic wastewater containing low concentration of organic substrate, without adding an external carbon source and without biomass inoculation. The average biomass concentrations in the mixed liquor during S-I, S-II and S-III were 1100 mg/L, 1200 mg/L, and 1050 mg/L, respectively. Considering granules to be the particles with diameters between 0.2 and 5.0 mm (Liu et al., 2010), granulometric analyses indicated average granular biomass fractions of 66%, 32% and 59% during S-I, S-II and S-III, respectively.
Initially, before the granules formation, all particles presented diameters below 200 µm. On the 16th day, 83% of the biomass reached diameters higher than 200 µm, and 46% of the particles were greater than 600 µm. The granular biomass varied throughout the studied strategies, showing mean diameters of 427±89, 265±51 and 292±54 µm for S-I, S-II and S-III, respectively.
According to de Kreuk et al. (2007), the biomass is considered to be predominantly granular when at least 50% of the biological aggregates present diameters superior to 200 µm. Therefore, the sludge was considered predominantly granular during S-I and S-III. However, the same authors noted that other characteristics, such as SVI, must also be considered in granular systems evaluations.
During S-I, SVI 5 , SVI 10 and SVI 30 presented the highest variation. This fact can be attributed to biomass instability during the granules formation and stabilization, as granulation is a gradual process comprising three stages: (i) sludge acclimation, (ii) expressed in terms of nitrogen (g N 2 O-N); Q pe is the daily per capita wastewater generation (L/person·d). The N 2 O Flow-Based Emission Factor (FBEF) was calculated by relating the amount of released N 2 O and the volume of wastewater treated in one SBR cycle, according to Eq. 5.

Microbiological procedures
Next-generation sequencing and fluorescent in situ hybridization (FISH) techniques  were used to investigate the microbial community of the granular sludge samples from S-I, S-II and S-III. DNA sequencing was performed using MiSeq ® Illumina technology for sequencing by synthesis (SBS). The DNA was extracted from biomass samples using a MoBio PowerBiofilm™ DNA extraction kit (MoBio Laboratories, USA). The rRNA 16S V3/V4 region was amplified using the 341F (CCTACGGGRSGCAGCAG) and 806R (GGACTACHVGGGTWTCTAAT) primers, with Illumina adapters, required for sequencing. The amplification was performed in 35 cycles at 50ºC of annealing temperature, where each sample was amplified in triplicate. The sequencing was performed in Illumina MiSeq ® , using the V2 kit, with a single-end 300 runs. The system guaranteed the reading of 100,000 sequences with sampling taxonomic identification and quantification of the number of sequences obtained from each taxon. OTU Picking was performed using sludge aggregation and (iii) granules maturation (Wang et al., 2005). Low SVI values and SVI 30 /SVI 10 ratios close to 1 are associated with a denser and more compact biomass, with good settling properties. The mean SVI 30 values were 126, 118 and 70 mL/g for S-I, S-II and S-III, respectively. In another study with AGS in SBR operated under similar conditions, Wagner and Costa (2013) verified that SVI 30 decreased gradually and stabilized at 53 mL/g after 100 days of operation. During S-II, SVI 5 , SVI 10 and SVI 30 values were closer than under S-I, indicating an improvement in the settleability. The SVI 30 /SVI 10 ratio remained 0.8 from the 149 th until the 230 th day, indicating an improvement in granular structural stability and biomass compactness, even with biomass concentration variations in the reactor.
S-III, which presented the longest anoxic phase, showed the closest values among SVI, and the average SVI 30 /SVI 10 ratio was 0.88±0.09, reaching 1.0 on the 286 th and 356 th days. According to de Kreuk et al. (2007), the SVI 30 /SVI 10 ratio gives excellent information regarding the granular fraction of the biomass. The higher the ratio, the better the granule settleability and compactness. Furthermore, the SVI 30 / SVI 10 ratio also indicates the granulation process status (Liu and Tay, 2007). These authors consider the granulation process to be completed when the SVI 30 /SVI 10 ratio reaches 0.9. Therefore, even with the diameter decrease verified in S-II and S-III, the granule settleability and compactness improved. According to Liu and Tay (2007), a higher granule size does not guarantee a better settleability, while the SVI is directly related to sludge density. This means that aerobic granulation should not be restricted only to the granule size increase, but also to the improvement in the biomass compactness and settleability.

Wastewater treatment performances
The effluent concentrations and the removal efficiency in terms of carbon and nitrogen verified for S-I, S-II and S-III are presented in Table 2. Although there were fluctuations in the soluble COD influent, the effluent concentrations did not show considerable variations. Under S-I, the COD soluble removal efficiency was 79%, with a mean effluent concentration of 52 mg/L. During S-II, the removal efficiency was 70%, with the SBR effluent presenting an average concentration of 50 mg/L. In S-III, the COD soluble removal rate was 68% and the effluent concentration was 58 mg/L.
Organic matter removal was also analyzed through BOD concentration. As observed with COD, influent BOD varied since the system was fed with real wastewater. An improvement in the BOD removal was observed over time, from 69 to 86%, and effluent concentrations from 106 to 31 mg/L. Regarding Brazilian national regulations CONAMA 430/2011, which require at least 60% removal or an effluent BOD concentration below 120 mg/L, all strategies met the quality criteria for carbon removal. The BOD removal improvement that was observed in S-III might also be related to the change in the HRT, which went from 7.13h (S-I and S-II) to 10.69h (S-III).
In terms of nitrogen removal, under S-I and S-II the NH 4 + -N average removal efficiency was below 60%. The longest aeration phase, applied in S-III, favored ammonium removal, achieving a stable and effective removal of 80% under this operational condition. In fact, as can be seen in Figure 2, which shows the pH profiles during the GSBR cycles, S-III presented the highest decrease in pH during the aerobic phase, indicating the occurrence of a more intense nitrification process in relation to S-I and S-II. As with the BOD removal, the NH 4 + -N removal improved with the increase in the HRT from S-I and S-II to S-III. This fact is consistent with the results presented by Wagner and Costa (2013), who observed a significant increase in the NH 4 + -N removal when the HRT went from 7.5h to 10h, when operating a SBR under conditions similar to the present study.
During aeration, due to nitrification, there was a progressive increase of nitrite formation, with mean effluent concentrations of 5.9 (S-I), 9.7 (S-II) and 14.5 (S-III) mg NO 2 --N/L. Nitrate was formed in trace concentrations in S-I and S-II, remaining at low levels during these operational strategies, in a range between 0.1 and 0.4 mg NO 3 --N/L. During S-III, a higher nitrate formation was observed, with a mean effluent concentration of 4.15±1.29. These results indicate the occurrence of incomplete nitrification in all conditions Table 2. Effluent concentrations and removal efficiencies verified for S-I, S-II and S-III. tested, with higher nitrite accumulation and nitrate formation at the longest aeration period. The extension of the anoxic phase of the cycle did not show as great an influence on nitrite accumulation as the extension of aeration phase.
Nitrite accumulation in reactors with aerobic granules has been reported by some authors (Yang et al., 2013;Isanta et al., 2012;Coma et al., 2012), including conditions of low-strength wastewater (Wang et al., 2007;Figueroa et al., 2008). Although nitrite-oxidizing bacteria (NOB) have their activity decreased by low DO concentration, in the present study the cause of partial nitrification was not low DO in the mixed liquor, since DO remained close to 8 mg/L throughout the aeration phase of the operational cycle, due to the high level of aeration required. Typical pH and DO cycle profiles can be seen in Figure 2.
In this study, the temperature might have been one important factor that contributed to nitrite accumulation, since it reached values up to 27° C inside the reactor. The maximum specific growth rate of ammonium-oxidizing bacteria (AOB) is higher than that of NOB at temperatures above 15º C (Bérnet and Spérandio, 2009), which can favor nitrite accumulation. In fact, this is the basis of SHARON technology, which consists of a chemostat reactor operated at 30°C with a low HRT, to favor AOB growth and NOB washout, achieving partial nitrification (Hellinga et al., 1998). The present research showed that, when the HRT increased from 7.13h (S-I and S-II) to 10.69h (S-III), there was a higher nitrate production, i.e., a more complete nitrification process.
The solids retention time (SRT) is another parameter that can influence nitrite accumulation. The calculated SRT were 14, 15 and 9 days for S-I, S-II and S-III, respectively. However, in the present research, nitrite accumulation might have been more strongly associated with the cycle duration than with the SRT. The cycle duration, which was extended from 4h (S-I and S-II) to 6h (S-III), favored the occurrence of nitrification, as there was a longer time to allow ammonium oxidation. Besides temperature and cycle duration, another factor that could contribute to nitrite accumulation is AOB and NOB stratification and the existence of an oxygen gradient in the aerobic granules. The presence and predominance of AOB colonies in the outer layer of the granules could favor nitritation, since they are in a more beneficial position for oxygen consumption than NOB, which are present in the inner layers (Poot et al., 2016;Guimarães et al., 2017). Many studies have tried to achieve partial nitrification, while this study showed feasible SBR operational conditions in tropical and subtropical climate for this.
The nitrite route has several advantages, including a lower oxygen requirement for nitrification (25% less), lower organic carbon consumption in denitrification (40% less) and lower sludge production (Van Loosdrecht and Jetten, 1998). These advantages are even more notable in the case of nitrogen-rich wastewater with a low organic carbon content. However, one of the main concerns related to nitrite accumulation is N 2 O production. Several studies have shown that nitrite accumulation is usually accompanied by higher N 2 O emissions (Itokawa et al., 2001;Kampschreur et al., 2008).

N 2 O emissions
The variation of N 2 O emitted in one typical cycle of the SBR in each operational condition tested is shown in Figure 3. For all of the studied strategies, N 2 O emission was not constant during the cycle phases. During S-I and S-II, the emission of N 2 O started at the beginning of the aerobic phase, and the peak concentration of N 2 O occurred in the early moments of aeration, between 2' and 2'30", and then decreased until it ceased. During S-III, the N 2 O emission pattern was similar to S-I and S-II. However, some N 2 O emission was also observed during the aeration pulses applied in the anoxic phase of this strategy. The maximum emissions were 0.90, 0.36 and 0.12 mg N 2 O/s for S-I, S-II and S-III, respectively, showing that, when extending the anoxic phase, a lower N 2 O emission peak was observed.
The fact that the peak emission of N 2 O occurred at the beginning of the aerobic phase does not mean that this is the moment of greatest N 2 O production. In fact, the emission pattern suggests that denitrification was possibly the major source of N 2 O generation, probably due to the occurrence of an incomplete denitrification process. Moreover, there seems to be no N 2 O formation during the occurrence of nitrification. Yang et al. (2013), who investigated N 2 O emission by a single stage reactor with partial nitrification/ anammox, suggested that N 2 O emitted during aeration is produced by microorganisms during the anoxic phase of the reactor cycle. Since there is no air flow during the anoxic phase, N 2 O is retained in the system and accumulates during this step. Mello et al. (2013) investigated the emission of N 2 O by an activated sludge treatment plant with intermittent aeration and found that less than 1% of the produced N 2 O was released in the absence of aeration. Therefore, the N 2 O that accumulated in the mixed liquor during the anoxic phase is released when aeration starts, causing a peak in the concentration of N 2 O emitted at the beginning of the aerobic phase. After a few minutes, all the accumulated N 2 O ends up being released to the atmosphere and the emission falls to values near zero.
There are several factors that suppress N 2 O production, and consequently the N 2 O emission, under the aeration phase of the reactor cycle. Kampschreur et al. (2008) found that high concentration of DO during nitrification prevented N 2 O production by microorganisms. In the present study, DO values during the aerobic phase were between 8.0 and 8.8 mg/L, which could prevent the formation of N 2 O in this period.
Conversely, there are aspects that contribute to N 2 O emission, such as the occurrence of nitrite accumulation during nitrification, which can later be converted to N 2 O during denitrification. Furthermore, if the denitrification process is incomplete, the N 2 O formed might not subsequently be converted to N 2 , causing a higher net N 2 O generation. Besides that, incomplete denitrification can lead to a higher nitrite concentration in the mixed liquor. Studies by Shaw et al. (2006) indicated that high concentrations of NO 2 can positively affect the emission of N 2 O in a nitrifier denitrification process, making NO 2 concentration one of the important variables that could be related to N 2 O generation during denitrification in activated sludge systems.
Some N 2 O emission parameters verified in this research are shown in Table 3. The average N 2 O emission was 0.181, 0.058 and 0.016 gN 2 O-N/cycle for S-I, S-II and S-III, respectively. In terms of volume of wastewater treated, the flow-based emission factors observed were 3.29·10 -3 , 1.05·10 -3 and 0.29·10 -3 gN 2 O/L for S-I, S-II and S-III, respectively. Considering the volume of effluent treated in a cycle, and assuming a per capita wastewater generation of 160 L/person·d, the EF observed during S-I, S-II and S-III were 192.2, 61.6 and 17.0 gN 2 O-N/person·year. These values are much higher than the guidelines proposed by the IPCC (2006), of 3.2 gN 2 O/person·year (i.e., 2.04 gN 2 O-N/ person·year) in the case of wastewater treatment systems with controlled nitrification and denitrification. However, this EF proposed by the IPCC was based on a single experiment of Czepiel et al. (1995), performed in a WWTP in Durham, in northern United States (temperate climate). Therefore, the present study suggests that the N 2 O emission by biological treatment systems located in subtropical/tropical regions is likely higher than the emission in temperate regions. Such higher EF values were also observed by Mello et al. (2013) in an intermittent aeration activated sludge system located in the highlands of Rio de Janeiro, Brazil, which is a subtropical region.
The total nitrogen removal, the fraction of nitrogen denitrified to N 2 O and to N 2 , and the total influent Nitrogen conversion to N 2 O are shown in Table 4. The results show that most of the total denitrified nitrogen was converted to N 2 , while only a small fraction was converted to N 2 O. Although the complete denitrification process did not occur, N 2 generation predominated over N 2 O generation in 84:16, 95:5, and 99:1 ratios for S-I, S-II and S-III, respectively. These results indicate that S-III, with the longest anoxic time, resulted in the lowest N 2 O emission among the studied strategies, both in terms of emission factor and TN to N 2 O conversion. The ratios of denitrified nitrogen emitted as N 2 O observed in this study are within the range of values observed by Foley et al. (2010) in a study involving seven WWTPs in Australia. The authors reported a wide oscillation between the percentages of nitrogen denitrified to N 2 O compared to the total nitrogen denitrified, ranging from 0.06 to 25.3%.
The fraction of TN converted to N 2 O ranged from 0.47% in S-III to 5.28% in S-I. These values are below the value reported by Sun et al. (2013), who recorded a conversion of total nitrogen to N 2 O of 6.52% in a full scale SBR. The same authors reported a conversion of 1.95% of the influent TN to N 2 O in a real scale A 2 O system, this value being within the range observed in the present study. The conversion rate verified in S-II is also very similar to the conversion presented by Castro- Barros et al. (2015), who noted that 2.0 % of the incoming nitrogen load was converted to N 2 O. Kong et al. (2013) analyzed the emission of N 2 O by a biofilm SBR under intermittent aeration, with intentional nitrite accumulation to favor the anammox process. The fraction of influent nitrogen converted to N 2 O was 1.50 ± 0.22%. This percentage of conversion is very close to what was observed in S-II, in which the occurrence of partial nitrification was also observed. By using molecular biology techniques, the authors found that Nitrosospira bacteria were the dominant gender of AOB responsible for N 2 O emissions via nitrifier denitrification.
A review study done by Kampschreur et al. (2009) listed conversions of influent nitrogen to N 2 O ranging from 0.001 to 14.6%. A wide variation of conversion to N 2 O was also found in a national survey conducted by Ahn et al. (2010) in the United States, where conversions ranging from 0.01 to 1.8% of the influent nitrogen were observed. Although there are variations between the values obtained by different authors, it is noted that the total nitrogen fraction converted to N 2 O tends to stay Table 3. N 2 O emission parameters for S-I, S-II and S-III. within the range of values between 0.4 and 6.5%, which covers the results obtained in this research.
Although the N 2 O EF obtained in this research is superior to the EF proposed by the IPCC (2006), it is consistent with several studies reporting N 2 O emissions by activated sludge systems. Daelman et al. (2013), studying the N 2 O emission by a municipal activated sludge WWTP, reported an emission factor of 163.2 g N 2 O-N/person·year. This EF is 80 times higher than the EF proposed by the IPCC, and even much higher than the EF obtained in S-II and S-III. Mello et al. (2013), investigating the emission of N 2 O by an activated sludge WWTP with intermittent aeration, observed an EF of 8.76 g N 2 O/person·year, i.e., 5.57 g N 2 O-N/person·year. This EF is also higher than the EF proposed by the IPCC (2006), although it is lower than the values observed in the present research.
The average FBEF ranged from 3.29·10 -3 gN 2 O/L in S-I to 0.29·10 -3 gN 2 O/L in S-III. These values are much higher than the factor reported by Mello et al. (2013) of 8.0·10 -5 g N 2 O/L, referring to an activated sludge system with intermittent aeration. However, this study was carried out in a region of humid subtropical climate, located at 600 meters of altitude, during the winter, unlike the conditions of the present study. In addition, the occurrence of nitrite accumulation in the system was not reported by the authors. These conditions may help to explain the low N 2 O emission verified by the authors, in relation to the values obtained in the present research.
Castro- Barros et al. (2015) studied the emission of N 2 O by a full-scale partial nitritation-anammox granular sludge reactor. The nitrogen load applied to the reactor was 1.75 kg NH 4 + -N/m 3 ·d, this load being around 10 times higher than the load applied in the present research. The authors verified that the conversion of influent nitrogen to N 2 O presented an average value of 2.0%, very similar to the conversion observed in S-II, which corroborates the results verified in the present study. In spite of the wide variation between the applied nitrogen loads, the percentage of nitrogen converted to N 2 O was quite similar in both cases.
The nitrite accumulation that occurred in this study, as a result of partial nitrification, may have been one of the factors which favored N 2 O production during denitrification (Kampschreur et al., 2009). However, since nitrite accumulation was not directly related to N 2 O emission, there might be other factors influencing N 2 O emission. As noted by Quan et al. (2012), N 2 O emission could also be related to the granule constitution, since the spatial structure of the granules may induce incomplete denitrification, which may also lead to significant N 2 O generation.
The results obtained in the present study indicate that S-III, with the longest anoxic phase, promoted the lowest N 2 O emission and the highest ammonium removal rate among the studied strategies, probably due to a higher consumption by a better developed anoxic community. The emission factor and the conversion to N 2 O verified in S-III were 11 times lower than in S-I, and 4 times lower than in S-II.

Microbial Communities
The AGS composition was dominated by the genera Pseudomonas sp. (17%), Comammonas sp. (19%) and Pseudoxanthomonas sp. (45%) under S-I, S-II and S-III, respectively. The microbial dynamics characterized by new generation sequencing showed fluctuations along the operational strategies. Under S-I (64 th and 83 rd days), the most abundant families were Caulobacteraceae, Sphingomonadaceae, Pseudomonadaceae and Rhodocyclaceae. Under S-II (189 th day), the family Comamonadaceae was highlighted with relative abundance of 22%, whereas under S-III (293 rd and 391 st days), the Xanthomonadaceae predominated with 45%. A decrease in the relative abundance of Pseudomonadaceae (genus Pseudomonas sp.), as well as an increase of Xanthomonadaceae (genus Pseudoxanthomonas sp.) was observed over time during the studied strategies. It is important to point out that both populations are denitrifying. In addition, Pseudoxanthomnas sp. is a relevant community for granule structure, since they are EPS producers (Weissbrodt et al., 2014). This result corroborates the better stability of the system in terms of granular biomass characteristics obtained under S-III.
In terms of microorganisms related to the nitrogen cycle, DNA new generation sequencing underestimated Nitrosomonas and Nitrospira sequences, not detecting them. However, they were identified with FISH analysis. AOB hybridizing to probe Nso190 was identified in low abundance in S-I, which is in accordance with low ammonium oxidation activity at this period. Lower AOB activity could arise from competition with heterotrophic bacteria for oxygen (Okabe et al., 1999). In S-II and S-III, corresponding to the 189 th and 293 rd days, there was a gradual increase in the abundance of AOB, correlating with a higher nitrification rate. Hybridization with Ntspa662 probes was carried out to identify the presence and distribution of Nitrospira in the granular sludge. Nitrospira was present as small clusters in low abundance during all strategies. Since nitrite accumulated in the reactor at most times, it was directly available to NOB from the bulk liquid (Kim et al., 2006), but conditions (discussed above) were not completely favorable to their high proliferation. Regarding denitrifiers, a positive hibridization signal was observed for Pseudomonas sp. in samples of S-I and S-III, indicating this genus as a feasible community for nitrite denitrification processes. However, the availability of sufficient organic carbon is the key factor in NO and N 2 O consumption activities, as previously reported (Kampschreur et al., 2009). It can be concluded that the longer anoxic phase in the SBR operation cycle (S-III) promoted a BOD removal improvement by denitrifiers such as Pseudomonas, ensuring lower emissions of N 2 O in the granular sludge reactor under these conditions.

CONCLUSIONS
The treatment performance and the N 2 O emission from a pilot-scale SBR with AGS operated under three cycle configurations and fed with domestic wastewater under subtropical climate conditions were monitored and quantified. The nitrification process was incomplete, with nitrite accumulation occurring, which was mainly attributed to the temperature and to the cycle duration. There was a significant reduction of 91% both in the TN to N 2 O conversion and in the EF verified for the studied strategies, which were associated with the extension of the SBR cycle anoxic phase and with the higher HRT. Furthermore, the anoxic phase extension and the HRT increase were also associated with higher BOD and ammonium removal rates and with a better biomass stability.