Greenhouse Gas Emissions from Biological Nutrient Removal During Wastewater Treatment

Since the 1900s, humans have been altering the global nitrogen (N) cycle by industrially fixing N for fertilizer production. This reactive N is often released back to coastal environments through many mechanisms, including wastewater treatment, where it can lead to numerous consequences such as fish kills and algae blooms. In many locations, wastewater treatment effluent is one of the largest sources of excess N to coastal environments. Although regulations limiting N loads in wastewater effluent in the U.S. were first developed in the 1970s, stricter regulations started to emerge in many states in the 2000s. In order to meet new discharge requirements, many centralized wastewater treatment plants (WWTPs) and onsite wastewater systems (OWTS) have been upgraded to include biological nitrogen removal (BNR) systems. These BNR systems make use of nitrifying and denitrifying bacteria to convert reactive forms of N (ammonium and nitrate) to nitrogen gas. Current BNR systems can reduce effluent total N loads to below 5 mg/L. However, nitrous oxide (N2O), a greenhouse gas (GHG) over 200 times more potent than carbon dioxide (CO2), may be produced along with or instead of nitrogen gas. Further, organisms that respire CO2 and produce methane (CH4) have been documented in BNR systems, making these systems potential sources of these additional potent GHGs. The BNR systems at WWTPs and OWTS can vary in many ways including the order and number of the different zones or compartments (aerated, anoxic, and anaerobic) and recycling arrangements. Therefore, although BNR systems at both WWTPs and OWTS may reduce N loads to coastal ecosystems, they may release GHGs that contribute to climate change. The central objective of this research was to examine the magnitude, variability, and potential production mechanisms of GHG emissions from a BNR system at a WWTP and advanced OWTS. This research is timely as BNR systems are increasingly used at both WWTPs and OWTS, but differences in the systems can result in different GHG emissions and N removal efficiency. Greenhouse gas emissions were measured using a cavity ring down spectroscopy (CRDS) analyzer (Picarro G2508) capable of measuring N2O, CO2, and CH4 nearly simultaneously in real time. To first evaluate this new technology, a comparison study was conducted (Chapter 1) to test the CRDS (Picarro G2508) relative to two alternative methods for measuring GHG emissions, Gas Chromatograph (Shimadzu GC 2014) and Los Gatos N2O analyzer. The results of the study indicated that the detection limit of the Picarro was an order of magnitude lower than that of the Gas Chromatograph, but an order of magnitude higher than that of the Los Gatos N2O analyzer. Although both the Picarro and Los Gatos analyzers offer efficient and precise alternatives to GC-based methods, the Picarro has the unique capability of measuring all three GHGs (N2O, CO2, and CH4) simultaneously. Therefore, the Picarro was deemed suitable for use in the WWTP and OWTS studies. Two major studies examining GHG emissions from a WWTP and OWTS were performed. The first was a yearlong study to determine the temporal (bi-monthly across annual cycle) and spatial (4 major zones: pre-anoxic, aerated IFAS, post-anoxic, and reaeration) variability of GHG (N2O, CO2, and CH4) emissions from an Integrated Fixed Film Activated Sludge (IFAS) BNR system at the Field’s Point WWTP in Providence, RI (Chapter 2). In addition, to understand environmental controls on the GHG emissions, potential relationships between the GHG emissions and water and tank parameters were examined. Finally, the emissions of all three GHGs were used to evaluate the importance of the BNR system to the overall GHG budget of the WWTP. The results of this study indicated that emissions of all 3 GHGs were highest from the aerated IFAS zone and all 3 GHGs varied by season (hourly variation was examined in Appendix 1). The N2O emissions were related to both ammonium and nitrate. When considering the emissions of all 3 GHGs in terms of CO2 equivalence, BNR is responsible for approximately 12% of the total GHG emissions for the Field’s Point WWTP (including emissions from: electricity, natural gas, liquid fuel, sludge disposal, and supplemental carbon). Generally, the BNR tank had higher emissions of all three GHGs than other parts of the treatment train (grit chambers, primary clarifiers, final clarifiers) (Appendix 2). However, the N2O emissions from the BNR tank represented only 0.01 – 0.34% of the influent N. Appendix 3 investigated the use of isotopomers to determine the mechanisms of N2O production from the BNR tank. The second major study compared N2O emissions from the BNR system at the Field’s Point WWTP to those from three common types of advanced OWTS used in RI to remove N (Advantex, SeptiTech, and FAST) (Chapter 3) (CH4 and CO2 emission measurements are reported in Appendix 4). The emissions were compared in terms of normalized per capita emissions and emission factors (% of N removed released as N2O). In addition, the specific abundance of a nitrification gene (ammonium monooxygenase, amoA) and denitrification gene (nitrous oxide reductase, nosZ) were quantified in order to determine the abundance of microorganisms that may be producing N2O in these systems. The results of this study (Chapter 3) indicated that in general N2O emissions from N removal during wastewater treatment were <1% of the N removed, except for one SeptiTech system (4%) and one Advantex system (21%). In general, N2O emissions (on a mole/area basis) from the WWTP were larger than those from OWTS and the OWTS with the largest N2O emissions was Advantex. However, when N2O emissions were normalized per capita and surface area of the treatment tank, they were similar between the WWTP and OWTS. Although there was no linear relationship between N2O emissions and amoA or nosZ abundances, amoA and nosZ abundances did differ between the WWTP and OWTS. The results of this dissertation allow us to focus future research efforts on the zones (aerated IFAS at WWTP) and systems (WWTP and Advantex OWTS) that produced higher emissions. In addition, future studies should try to develop a better understanding of the large temporal and spatial variability observed in these systems. The results of this research determined that N2O emissions were related to both ammonium and nitrate, indicating that both nitrification and denitrification likely play a role in N2O emissions. However, preliminary isotopomer results indicate that nitrification may be responsible for the N2O emissions. With additional studies on the mechanisms of production, suggestions to operators can be made so that emissions can be lowered while maintaining N removal.

removed), data was collected every 10 sec. and the average for the morning hours during which flux measurements were made is shown. WW = wastewater DO=Dissolved Oxygen SVI=Sludge Volume Index (B) Data that was collected from same locations as flux measurements (n=3 for each zone). For variables that included data below the detection limit, the non-parametric Kaplan-Meier method was used to determine the mean and standard deviation………………….…… 68      Table S2. Average production (kg of gas d -1 ) and standard error for the year long measurements for each gas by zone ……………………………………………...77 Chapter 3 Table S1. Primer sets used for qPCR analysis……………...…………………….111 Table S2. Average and standard error of wastewater properties from pre-anoxic,   Chapter 2 of this dissertation examines the temporal and spatial variability of N2O, CO2, and CH4 emissions from one of the integrated fixed film activated sludge BNR tanks at the largest centralized WWTP in RI. Appendix 1 shows the hourly variation in the emissions of all three GHGs from one zone of the BNR tank examined in Chapter 2 and Appendix 2 compares the GHG emissions from the BNR tank to other components of the treatment train (grit tanks, primary clarifiers, and final clarifiers). Appendix 3 investigates potential mechanisms of the N2O emissions from the BNR tank at the centralized WWTP using isotopomers.
In Chapter 3 and Appendix 4, the emissions of all three GHGs from the BNR tank at the centralized WWTP are compared to those from advanced OWTS designed to remove N. In addition, the abundance of a nitrifying (amoA) and denitrifying (nosZ) gene are compared among the systems in order to investigate potential mechanisms of the N2O emissions.
This research is timely as BNR systems will increasingly be used at both WWTPs and OWTS as the human population continues to grow. As a result, the need for efficient N removal systems that successfully remove N with minimal greenhouse gas emissions will continue to grow.

Materials and procedures
Objective 1: Minimum detection limits Gas fluxes were calculated from linear rates of change in gas concentrations within a closed chamber as described in Martin and Moseman-Valtierra (2015) and Supplemental Materials. We primarily report detection limits as the slope of gas concentration versus time in units of ppb s -1 to preserve generality and refer to them hereafter as "minimum detectable slopes."

Analyzers
Both the Picarro and LGR report gas concentrations (as dry mole fractions in ppm) roughly every 2 seconds. All default settings were maintained for the Picarro and more information about the CRDS technology used can be found in Fleck et al.
(2013). The LGR was factory calibrated by measuring known standards (NOAA CMDL primary standard for N2O and CO, and a LICOR 610 dewpoint generator for the water vapor calibration).

Monte Carlo simulations for detection limits of Picarro and LGR
To estimate the minimum detectable slope of each gas (CO2, CH4, and N2O for the Picarro, and only N2O for the LGR), we first measured and then modeled (using Monte Carlo simulations) Allan standard deviations based on instrument noise levels (Allan 1966

Shimadzu GC-2014 method quantification limit
The precision of the Shimadzu GC-2014 was determined as outlined in Christiansen et al. (2015). A low standard containing concentrations of CO2 (319.6 ppm), CH4 (2.625 ppm) and N2O (0.519 ppm) was read 20 times and the precision was defined as the method quantification limit (standard deviation x 3 x t 99%). The resulting precision was 265 ppm for CO2, 1.6 ppm for CH4, and 0.14 ppm for N2O.
To calculate the minimum detectable slope, the precision was divided by the chamber closure time (5 minutes).

Mesocosm Experiment A
To compare CO2, CH4, and N2O fluxes measured by the Picarro and Shimadzu GC-2014, an experiment using two distinct mesocosms (Mesocosm IDs: A-1 and A-2, Mesocosm A-1 in an effort to produce a wide range of N2O fluxes (Table 1). For more details on conditions of mesocosms prior to gas flux measurements see Table 1 and Supplemental Materials.
As the objective of this study was to compare the Shimadzu GC-2014 and Picarro analyzers, and not to specifically contrast the different soils, replication was obtained by making multiple gas measurements simultaneously with both instruments on each mesocosm. Each mesocosm constituted a time series of measurements each separated by one minute (sufficient time for the analyzer and open chamber to return to ambient concentrations). Therefore, each flux measurement in this series was considered a separate replicate.

Gas flux measurements
Static flux chambers were used to simultaneously measure CO2, CH4, and N2O fluxes with the Picarro and Shimadzu GC-2014. For each measurement, an intact soil mesocosm was transferred in a pot to a 5 gallon bucket that was then covered with a transparent static flux chamber (Table 1). A closed-cell polyethylene foam collar and plastic wrap were used to make a gas-tight seal between the rim of the bucket and the chamber. The chamber contained two battery-powered fans to mix the interior gases.
A coiled stainless steel tube (inner diameter of 0.71 mm) attached to a port at the top of the chamber maintained equilibrium with atmospheric pressure. The duration of chamber deployments (5 min.) was based on observed periods of linear changes in gas concentrations (Table 1). Nylon tubing (0.46 cm inner diameter and approximately 5 m in total length) connected to the Picarro via two gas-tight ports in a closed loop.
The chamber also had an extra port with stopcock by which discrete gas samples were manually collected and analyzed on the Shimadzu GC-2014. Gas samples (35 mL) were drawn by hand into 60 mL nylon syringes equipped with Luer-Lok stopcocks at 0, 0.5, 1, 1.5, 2, 3, 4, and 5 minutes. Gas samples were transferred to For data collected with the Picarro, the first 30 seconds of measurements (4.5 minutes remaining) were not included in the flux calculations in order to account for gases passing through the length of the tubing between the analyzer and the chamber.
Since collection of discrete gas samples did not require tubing, the entire 5 minutes of data (8 data points) were included in calculations of fluxes from samples analyzed on the Shimadzu GC-2014. Nitrogen levels (ammonium chloride and ammonium nitrate) were applied iteratively in this experiment to each mesocosm in an effort to produce a wide range of N2O fluxes (Table 1).
Gas fluxes were measured as described above (Objective 2) except for the following changes: no discrete gas samples were collected and nylon tubing (approximately 7 m for each analyzer) ran from gas-tight ports at the top of the chamber to the Picarro and LGR analyzers in parallel so that measurements were made by the two analyzers simultaneously. The total system volume for the Picarro and LGR (chamber, tubing, analyzer, and bucket) was 3.74 x 10 -2 and 3.77 x 10 -2 m 3 respectively. Air temperature inside the chamber was monitored with a Hobo® pendant temperature logger (Onset Inc.). For the N addition, square steel collars (56 cm x 56 cm) were placed in two groups of three collars (6 collars total). Each collar was at least 1.3 meters from the next one in a given group and the different groups were spaced at least 11 m from each other in a line that ran parallel to the shoreline. These were installed 2 years prior to the gas flux measurements. In order to avoid cross-contamination of plots by N additions, all three plots in a given group were assigned one of the N treatments in the form of sodium nitrate ( analyzer, and bucket) was 1.95 x 10 -2 m 3 .

Statistics
The statistical significance of each gas flux was determined using a sequential three step approach based on (1) visual inspection of data for any obvious measurement errors, (2) a test of the significance of regressions for linear periods of gas changes over time, and (3) application of slope detection limits to all fluxes with statistically significant regressions. In this study, removal of points occurred for one flux. If the regression was not significant (p-value > 0.05), then the flux was classified as not determined (ND). If the regression was significant (p-value < 0.05) then we compared the flux to the slope detection limit determined in Objective 1. Fluxes with significant regressions and that exceeded the slope detection limit were defined as significant. Fluxes below the slope detection limit were classified as ND even if the regression was significant. Fluxes labeled as ND were excluded from statistical analysis.
In addition, the normalized root mean square error (NRMSE) was calculated for each significant flux as outlined in Christiansen et al. (2011) and used as a metric to compare the precision of analyzers. Although R 2 has been used in previous literature, the NRMSE is not subjective to the range of the data and can therefore be used to compare the precision of the analyzers more objectively.
A paired t-test was used to determine if there was a significant difference between Picarro and Shimadzu GC-2014 fluxes (Objective 2). This was possible only for N2O in mesocosm A-1 because in most cases the Shimadzu GC-2014 did not detect significant fluxes ( For both the Picarro and LGR, the averaging period has essentially no effect on the minimum detectable slope (Table 2A) (Table 4). At the lower range of N2O fluxes (Mesocosm A-2) all three Picarro N2O fluxes were significant (14 ±1 mol m -2 h -1 ) while none of the Shimadzu GC-2014 N2O fluxes for this mesocosm were above the detection limit (Table 4).
Unfortunately, the majority of the CH4 and CO2 fluxes were below the detection limit of the Shimadzu GC-2014 and as a result could not be determined (Supplementary Material Table 1). Methane fluxes detected by the Picarro ranged from 1 to 4604 µmol m -2 hr -1 but only one of these fluxes was above the detection limit of the Shimadzu GC-2014 (Supplementary Material Table 1). All of the CO2 fluxes were below the detection limit of the Shimadzu GC-2014 but the range measured by the Picarro was 1.8 to 31.6 µmol m -2 s -1 (Supplementary Material Table   1).
Similar to Mesocosm B-1, the fluxes from the Picarro were on average 12% higher than for the LGR (Figure 2A and 2C).

Discussion
Comparing the suite of three GHGs: CO2, CH4, N2O CRDS technology in the Picarro confers several advantages over GC approaches for the quantification of GHG fluxes in dynamic coastal ecosystems. First, the Picarro had 1-3 orders of magnitude lower analytical detection limits for CO2, CH4, and N2O (Tables 2 and 3) than the Shimadzu GC-2014 and greater precision as evident in the consistently lower NRMSE values of the Picarro (Table 4). Indeed, the Picarro was consistently able to detect CO2 and CH4 fluxes as small as 2 µmol m -2 s -1 and 1 µmol m -2 hr -1 respectively from the salt marsh mesocosms, which were below the detection limit of the Shimadzu GC-2014 over the chamber duration time that we employed ( Further, the short chamber closure periods offered by high-precision, in situ analyzers, such as the Picarro and LGR, enables researchers to limit many of the errors associated with longer chamber closure times, such as alterations of the gas diffusion gradient and increases in temperature and represents a significant technological advancement (Davidson et al. 2002).

Measurements of N2O-comparing Picarro and LGR
In both lab and field experiments, the N2O fluxes measured by the Picarro and LGR were generally similar despite the differences in technology (Figure 2 and 3).
However, in some mesocosms (first round of Mesocosm B-1 measurements and Mesocosm B-2) and in field plots with low N additions, when fluxes were relatively low (3-132 µmol m -2 hr -1 ), the Picarro fluxes were slightly larger than LGR fluxes (9 -13%). This discrepancy may have partially been due to the low sample size, as no difference was found between the analyzers for N2O fluxes from the high N field plots for which the range of N2O fluxes (18-43 µmol m -2 hr -1 ) overlap with those from Mesocosm B-1 (on first date), Mesocosm B-2, and the low N enriched plot. The differences in IR regions used by the analyzers (nearIR for the Picarro and mid-IR for the LGR) may also partially explain this discrepancy. In one of these mesocosms (B-1, Figure 2A) consecutive measurements resulted in increasing flux values, potentially due to a lag in response to N additions. However, this is unlikely to have altered the comparison of analyzers because there was no relationship between the difference in fluxes from the two analyzers and measurement number (data not shown). To further discern the cause of such small but consistent differences between the two analyzers, further work including direct inter-calibration would be helpful.
Based on published N2O fluxes in coastal marsh ecosystems, ranging from 1.4 to 14.8 µmol m -2 hr -1 (Allen et al. 2007;Hirota et al. 2007;Liikanen et al. 2009;Moseman-Valtierra et al. 2011), the Picarro and LGR will generally be able to detect low N2O fluxes. The minimum detectable fluxes for the field chamber used in this study for the Picarro was 1.7 µmol m -2 hr -1 while for the LGR it was 0.1 µmol m -2 hr -1 .
One tradeoff for the higher detection limit of the Picarro however is the unique ability of the Picarro to simultaneously measure all three important GHGs, which is particularly advantageous as these gases are highly variable in space and time (Bartlett et al. 1985;Robinson et al. 1998;Bange 2006) and disturbance-induced CH4 and N2O fluxes can potentially offset CO2 uptake (Liu and Greaver 2009).
The significant advantage of high precision IR GHG analyzers, such as the Picarro and LGR, in coastal biogeochemistry is that they allow for rapid quantification of real time GHG data and this comes at a time when there is strong need to develop better climate change models that can include potential climate feedbacks from coastal ecosystems. Analyzers like the Picarro and LGR are significantly advancing scientists' abilities to better understand how anthropogenic stressors have the potential to change the GHG budget of coastal ecosystems.

Comments and recommendations
Several practical benefits are obtained from the rapid, real-time data collection of in situ gas analyzers such as the Picarro and LGR. Disadvantages of the Shimadzu GC-2014 include long run times and limited numbers of samples as well as substantially higher detection limits. However, the real time measurements collected by analyzers such as the Picarro and LGR facilitate identification of experimental errors (such as rapid changes in gas concentration and pressure resulting from disturbance associated with chamber placement) allowing the user to repeat measurements when needed. This is a clear advantage over grab sample based GCmethods.
Both the Picarro and LGR are sensitive to water and therefore must be operated with caution in coastal environments. Even small amounts of moisture in the analyzers' cavities may condense on the mirrors and lead to costly repairs. Further, the user must be aware that on warm days humidity may increase rapidly in the chamber during deployment. Fortunately, the Picarro monitors moisture and alerts the user if the moisture reaches a set threshold. In addition, the Picarro has two hydrophobic membrane filters in the inlet sample system that traps stray water droplets before they reach the sensitive optical cavity. One solution to this problem is to switch the inlet and outlet tubing if the moisture begins to rise. Moisture traps may also be devised relatively simply and employed if more humid conditions require further intervention. With proper attention to basic logistical needs, the Picarro and LGR offer significantly improved capabilities for GHG measurements from coastal environments.

Gas Flux Calculations
Gas fluxes were calculated from the linear periods of change in gas concentrations in the chamber over time (dC/dt) using the ideal gas law (Eq. 1).
Where F is the calculated flux (moles per unit area per unit time), dC/dt (ppm s -1 ) is the slope of the linear regression of concentration vs. time, V is the chamber volume (m 3 ), T is the temperature (K), P is pressure (Pa) and A is the surface area (m 2 ) of the mesocosms or field plots that were measured. The Picarro measures gases on average every seven seconds but interpolated concentrations are reported for each gas approximately every two seconds. These raw interpolated data were used in the flux calculations. Fluxes calculated from Shimadzu GC-2014, Picarro and LGR concentration data will be referred to as Shimadzu GC-2014 fluxes, Picarro fluxes, and LGR fluxes, respectively throughout the manuscript.

Monte Carlo simulations of instrument noise
For the Picarro G2508 the noise of the instrument was first quantified by connecting it on a closed loop to a single bottle of compressed ambient air with approximately 0.33 ppm of N2O, 400 ppm of CO2, and 1,800 ppb of CH4 (Air Liquide America Specialty Gases). This single bottle was continuously measured for 30 hours.
The Allan standard deviation of the resulting data set was modelled (Allan 1966) for each of the three gases with a combination of a Gaussian white noise term that follows a square root law, a flicker noise (also called 1/f or pink noise) term that leads to a constant Allan standard deviation independent of averaging, and a random walk noise term (also called brown noise). A Monte Carlo simulation of the instrument noise was performed, using optimized parameters for each of these three noise sources for LGR analyzer exhibits a dramatically smaller (40X) white noise contribution, and a moderately smaller brown noise contribution (2.6X) than the Picarro analyzer.

Mesocosm conditions for Objective 2
Prior to gas flux measurements, both mesocosms were placed in a climate

Impacts of closure time and averaging period with Picarro and LGR data
For the Picarro, there is an improvement in the minimum detectable slope for each gas with increased chamber closure time, improving as 1 / T 1.5 (T = seconds).
Some of this improvement was due to the increased data contained in the measurement period (leading to an improvement with 1 / T 0.5 ), and the remainder was due to the larger time span of the fit, improving the determination of the slope (leading to a 1 / T improvement). For the LGR the minimum detectable slope improves with increased chamber closure time, as 1/T 0.75 . This is due to an increasing influence of the brown noise component for times greater than 100 seconds for the analyzer. The minimum detectable slope for N2O for the LGR analyzer is one to two orders of magnitude lower than for the Picarro (Table 1). The LGR also had a higher precision for N2O because the lines used by mid-IR (LGR) are about 10 5 times stronger than the lines used in near-IR (Picarro). However, the Allan variance for the Picarro extends to an hour for N2O (rather than 100 seconds for the LGR) ( Figure 1) and is evidence of the reduced sensitivity of the Picarro to environmental factors.  needed to further discern the mechanisms responsible for the GHG fluxes.

Introduction
Wastewater treatment plants (WWTPs) have the potential to be significant sources of greenhouse gas (GHG) emissions on a national scale (US EPA, 2013).
Although the U.S. Environmental Protection Agency estimated in 2011 that anthropogenic N2O and CH4 emissions from WWTPs accounted for 1.5% and 2.8 % respectively of U.S. GHG emissions, these may be underestimates due to the large temporal and spatial variability reported in recent studies of GHG emissions from a range of wastewater treatment processes Czepiel et al., 1993;Ren et al., 2015;Tomaszek and Czarnota, 2015;US EPA, 2013;.
One major advance in wastewater treatment that may affect GHG emissions is the use of biological nutrient removal (BNR) to remove nitrogen (N) . This type of BNR is the practice of removing reactive N from wastewater using naturally occurring nitrifying and denitrifying bacteria under aerated, anoxic, and anaerobic conditions. Removing the N helps avoid conditions that can lead to eutrophication in receiving waterbodies .
Biological nutrient removal has recently been recognized as a potentially large source of N2O emissions from WWTPs because of the high concentrations of dissolved inorganic N undergoing rapid transformations and abundant microbial communities in wastewater (Grote, 2010;Tomaszek and Czarnota, 2015). During the BNR processes, two major sources by which N2O can be produced are microbial nitrification (predominantly in aerated zones) and denitrification (mainly in anoxic zones) (Tomaszek and Czarnota, 2015). A variety of organisms that produce CO2 through respiration or produce CH4 have been documented in BNR tanks Gray et al., 2002;Lens et al., 1995). However, the relative magnitude of production and emission of these GHGs is not well studied and likely to depend on conditions and methods employed to facilitate N removal.
BNR technology has advanced quickly and there are now over two dozen different BNR system designs used worldwide (Grote, 2010). The two main categories of BNR are suspended growth (ex. activated sludge and aerated lagoons) and attached growth (ex. moving bed reactor and trickling filters) (Eddy et al., 2013).
More recently, hybrid processes such as integrated fixed film activated sludge (IFAS) that combine both the suspended and attached growth designs have been developed (Eddy et al., 2013). Currently, there are 30 WWTPs utilizing the IFAS BNR method in the U.S (Carollo, 2012). Integrated fixed film activated sludge BNR systems utilize a plastic media designed to increase surface area for microbial growth without requiring additional tank volume and are commonly used to upgrade existing tanks to include BNR in order to meet new N discharge limits (Eddy et al., 2013).
Although several studies have measured N2O, CH4, or CO2 emissions from various BNR technologies, the reported emissions vary by at least 2 orders of magnitude Tomaszek and Czarnota, 2015).
Aerated zones of BNR systems thus far seem to have higher emissions than anoxic zones of all three GHGs due at least in part to the air stripping effect produced by the mechanical aeration Ren et al., 2015). Only two studies have examined seasonal variation in N2O emissions from BNR and report conflicting results (Sommer et al., 1998;Sümer et al., 1995). One study found that N2O emissions during the spring and summer were twice as high as those during the winter (Sommer et al., 1998) while the other did not find any seasonal variation (Sümer et al., 1995). For CH4 and CO2, some studies report a correlation between gas fluxes and wastewater temperature (Czepiel et al., 1993;) but others do not (Wang et al., 2011;. Only four studies have examined all three GHGs simultaneously from BNR systems in the field and laboratory, none of which were IFAS. These studies found that CO2 and N2O fluxes were larger than CH4 fluxes and that influent C/N ratio may impact the magnitude of all three GHG fluxes (Bao et al., 2016;Kong et al., 2016;Ren et al., 2015;. Water consumption, N intake, and wastewater generation and processing rates are influenced by several environmental conditions, such as weather and climate, that vary over time and also are likely to vary between each individual WWTP; these factors therefore influence variation in GHG fluxes (Brotto et al., 2015). The implementation of new methods such as IFAS BNR may further increase the heterogeneity of GHG emissions within and between BNR systems.
This is the first known study to examine N2O, CH4, and CO2 fluxes simultaneously from an IFAS BNR system in the U.S. Specifically, this study examines (1) temporal (bi-monthly across annual cycle) and spatial variability in GHG emissions from 4 major zones (Pre-Anoxic, Aerated IFAS, Post Anoxic, and Re-Aeration) of one IFAS BNR tank at the Field's Point WWTP in Providence, RI and (2) potential relationships between GHG fluxes and a suite of water and tank parameters to understand potential environmental controls of GHG fluxes. In addition we use the gas fluxes and concentrations of dissolved gases to discern in which zones the emitted gases are produced. Finally, we use the simultaneous measurements of all three major GHGs to estimate the total GHG emissions (and relative importance of each gas) from the IFAS BNR system, and evaluate the potential importance of this major component to the overall GHG budget of the WWTP.

Field Site
This study was conducted at the Narragansett Bay Commission's Field's Point The Aerated IFAS zone contains molded high-density polyethylene disc media (25 mm diameter; 10 mm length) at a fill rate of approximately 50%, providing an effective surface area of 500 m 2 /m 3 for biofilm to grow.

Sampling Campaign
In order to examine temporal and spatial variability of GHG emissions across an annual cycle (Objective 1), GHG fluxes were measured approximately twice a month for one year (June 2014 -June 2015) in one of the IFAS BNR tanks (Figure 1).
All measurements were collected between the hours of 8:00 am and 1:00 pm during weekdays. Three gas flux measurements were distributed approximately equally across each zone, except for the Aerated IFAS zone where all three measurements were collected in relatively the same location due to logistical constraints (Figure 1).
There was approximately 3 -30 minutes (on one occasion up to one hour) between each measurement within a zone.
In order to examine potential relationships between GHG fluxes and several water and tank parameters (Objective 2), water samples were collected within 3 hours (either before or after) of the GHG flux measurements and immediately stored on ice.
Water samples were collected from the same locations as the GHG flux measurements, except for the Aerated IFAS zone where the size of the zone prevented this (Figure 1). Water samples were analyzed for dissolved gas concentration and multiple other water and tank conditions (see sections 2.4 and 2.5).
Due to logistical constraints, gas measurements and water samples were not collected during August and September and water samples were not collected on one day in October (10/28/14). Gas fluxes and water samples were collected only on single dates (rather than bimonthly) in December, January, February, and April. In January, February, March, and April only two measurements/samples (rather than 3) were collected from each zone.

Greenhouse Gas Fluxes
To quantify N2O, CH4, and CO2 fluxes from the BNR tank, a real time GHG analyzer and pump (Picarro G2508, Santa Clara, CA) were connected to a transparent

Dissolved Greenhous Gases
A 35 mL subsample of the water samples collected as outlined in Section 2.2 was transferred to a 60 mL syringe and equilibrated with helium within 4.5 hours of collection  were calculated as outlined in Weiss and Price (1980) for N2O, Lammers and Suess (1994) for CH4, and Weiss (1974) for CO2.

Water and Tank Parameters
In order to examine potential relationships between water and tank process parameters and GHG fluxes, a subsample (approximately 15 mL) of the water sample collected as outlined in Section 2.2 was filtered (45 m) within 4 hours and frozen until analyzed for ammonium concentration using the phenolhypochlorite method (Solorzano et al. 1969) and nitrite using Hach Spectrophotometric Methods 102066.
Another subsample was filtered, acidified, and analyzed for nitrate using Hach Spectrophotometric Methods 102066. Due to logistical constraints, water samples from 7/21/14 were not analyzed for ammonium, nitrite, and nitrate. In addition the following data was provided by the Narragansett Bay Commission: wastewater temperature (HACH Model 57900-00, Loveland, CO), influent flow rate, dissolved oxygen (DO) (HACH Model 57900-00, Loveland, CO) in the Aerated IFAS zone, internal mixed liquor return flow, return activated sludge flow, and sludge volume index (SVI).

Statistics
To examine differences in GHG fluxes between zones (Pre-Anoxic, Aerated IFAS, Post Anoxic, and Re-Aeration) and over seasons (summer 2014, fall, winter, and summer 2015) during the yearlong measurement period, a two-factor ANOVA (zone x season) was performed for each gas. The time of day of the measurement was not included as a factor in this analysis because differences in fluxes between dates were not significantly related to the time the measurements were collected (checked with linear regressions, data not shown). Assumptions of homoscedasticity and normality were checked using residual plots and data was log transformed when necessary. In all cases, a post hoc Tukey test was performed to determine which zones and seasons were significantly different. The same statistical approach was used for dissolved gas concentrations.
Potential inter-relationships between GHG fluxes were examined using linear regressions; one regression for each zone and gas combination was performed. To determine potential mechanisms of N2O fluxes, relationships between N2O fluxes and dissolved N species (ammonium, nitrate, and nitrite) were examined using a principal components analysis (PCA). The PCA included data from all four zones (comprised of 3 replicates per zone on each date over an annual cycle). Ammonium, nitrate, and nitrite data that were below the detection limit were censored using the ranks of the u-scores prior to being included in the PCA (Helsel, 2011). Separate regressions were completed to test relationships of each GHG flux with DO in the Aerated IFAS zone.
A multiple regression was performed for each gas flux for each zone with the following predictors: water temperature, water flow rate, and SVI. Only significant regressions are reported. All statistical analyses were performed in R (R Core Team, 2013).

Characteristics and Performance of IFAS BNR System
The characteristics and performance data of Tank 1 of the IFAS BNR system at Field's Point during this study are summarized in Table 1. There was about a 6°C difference in water temperature between the warmest and coldest day during the study.
Air flow and DO in the Aerated IFAS zone were lowest in the fall. The average percent N removal was 74% and was lowest in the winter. There was a decrease in ammonium concentrations from the BNR system influent to the Pre-Anoxic zone.
Nitrate and nitrite concentrations were generally low relative to ammonium. The highest nitrate concentrations (averaging 2.04 ± 1.49) were in the Aerated IFAS zone, while nitrite concentrations were often below the detection limit, but were also generally highest in the Aerated IFAS zone.

Spatial and Temporal Variability of Greenhouse Gas Fluxes
During the yearlong study, N2O fluxes ranged from -7.6 x 10 -4 to 2.6 µmol m -2 s -1 (Figure 2A). For N2O fluxes, the zones, seasons, and the interaction between zones and seasons were significant (Table 2, Figure 2A). Nitrous oxide fluxes from the Aerated IFAS zone were highest, followed by those from the Re-Aeration zone ( Table   2, Figure 2A). On average N2O fluxes from the Aerated IFAS and Re-Aeration zones represented 75% and 21% respectively of the total N2O fluxes from all four zones.
Nitrous oxide fluxes from the two anoxic zones (Pre and Post Anoxic) were significantly different from each other and represented a small proportion of the N2O fluxes, on average a combined 4% of the total N2O fluxes from all four zones (Table   2, Figure 2A). Nitrous oxide fluxes in the summer of 2014 were significantly lower than those in the fall and winter (Table 2, Figure 2A). Exceptionally large N2O fluxes were measured on two dates, 7/21/2014 (summer 2014) when fluxes from the Aerated IFAS zone were approximately 3.5 times larger than the yearly average from that zone and 1/14/15 (winter), the only date when fluxes from the Re-Aeration zone were larger than those from the Aerated IFAS zone (Figure 2A). These dates illustrated the complex nature of the temporal variability and interaction between season and zone, as they were distinct from other dates within their respective seasons.
During the course of this study, CH4 fluxes ranged from 0.01 to 10.8 µmol m -2 s -1 ( Figure 2B). As with N2O, there was a significant difference in CH4 fluxes between zones, seasons, and the interaction of zones and seasons (Table 2, Figure 2B).
The Aerated IFAS zone had significantly higher CH4 fluxes than all other zones (Table 2, Figure 2B). On average 74% of the total CH4 fluxes from the BNR system were from the Aerated IFAS zone. Methane fluxes in the fall were significantly larger than those from other seasons (Table 2, Figure 2B). The significant interaction in CH4 fluxes between season and zone (Table 2, Figure 2B) is illustrated in the fall and one date in the winter (3/3/15) when CH4 fluxes from the Post Anoxic zone surpassed those from the Pre-Anoxic and Re-Aeration zones and even the Aerated IFAS zone on 3/3/15 ( Figure 2B).
Carbon dioxide fluxes ranged over three orders of magnitude from 2 to 2493 µmol m -2 s -1 ( Figure 2C). Similar to both N2O and CH4, there were significant differences in CO2 fluxes between the four zones, with the largest fluxes from the Aerated IFAS zone (Table 2, Figure 2C). Similar to N2O, the second largest CO2 fluxes were from the Re-Aeration zone with minimal fluxes from the two anoxic zones (Table 2, Figure 2C). There were also significant differences in CO2 fluxes between seasons and the interaction of zone and season (Table 2, Figure 2C). The lowest CO2 fluxes were in the winter (Table 2, Figure 2C).
In general, although the three GHGs exhibited different temporal trends, the highest fluxes of all three GHGs were from the Aerated IFAS zone (Figure 2 and Table 2). There was a significant positive linear relationship between CH4 and CO2 in the Aerated IFAS (p<0.001, r 2 =0.40, Figure 3A) and Re-Aeration zones (p<0.001, r 2 =0.22, data not shown). There was also a significant positive linear relationship between N2O and CO2 fluxes in the Pre-Anoxic zone (p<0.001, r 2 =0.47, Figure 3B).

Dissolved Greenhouse Gas Concentrations and Production
In order to better understand whether GHG fluxes were produced in the same zone they were emitted from or if they were produced upstream, dissolved gas concentrations were measured across all four zones and the influent and effluent of the BNR system. The dissolved N2O, CO2, and CH4 concentrations displayed distinct temporal and spatial patterns from the fluxes of the same GHGs (Figures 2 and 4).
Dissolved N2O concentrations were relatively low and ranged from 0.01 to 3.22 µM, except for one date (1/14/15) when concentrations were the highest (up to 9.2 µM) in all zones except the Inflow ( Figure 4A). Unlike N2O fluxes, there were not significant differences in dissolved N2O concentrations between zones or the interaction between zones and seasons (Table S1). However, dissolved N2O concentrations did significantly differ across seasons with the highest concentrations measured in winter, but this was partly driven by the single winter date (1/14/15) with exceptionally high values ( Figure 4A, Table S1).
In general, dissolved CH4 concentrations decreased as the water flowed through the BNR treatment process except on a few dates (6/30/14, 7/21/14, 11/20/14, 5/7/15) when concentrations in the BNR tank were higher than in the influent ( Figure   4B). Although dissolved CH4 concentrations exhibited significant differences between zones, seasons, and the interaction of zones and seasons, the patterns were distinct from those of CH4 fluxes ( Figure 4B, Table S1). In contrast to CH4 fluxes, which were highest in the Aerated IFAS zone, dissolved CH4 concentrations were highest in the Pre-Anoxic zone ( Figure 4B, Table S1). Further, unlike CH4 fluxes which were highest in the fall, dissolved CH4 concentrations in the summer of 2015 were higher than the other seasons ( Figure 4B, Table S1).
There were significant differences in dissolved CO2 concentrations between zones and seasons, but not the interaction of zones and seasons (Table S1). However, different trends were observed in the dissolved CO2 concentrations than the CO2 fluxes. While CO2 fluxes were largest from the Aerated IFAS zone, dissolved CO2 concentrations were high in all zones except the Pre-Anoxic and Inflow (Table S1). Further, while CO2 fluxes were lowest in the winter, dissolved CO2 concentrations were significantly higher in the winter and summer 2014 than the fall and summer 2015 ( Figure 4C and Table S1).
Given the strong mismatch between gas emissions and dissolved concentrations, we used both types of data along with flow rate in a set of mass balance equations to estimate the production of each gas in the water column for each zone. These production values were calculated as outlined in  and assumed solids were inert (Diagram S1). In general, the Aerated IFAS zone had the highest production estimates of all three gases (Table S2). This was also the zone that had the highest emissions of all three gases (Table 2). One exception was the January date (1/14/15) when the highest N2O fluxes from the Re-Aeration zone were observed.
On this date, high N2O production was estimated in the Aerated IFAS, Post-Anoxic and Re-Aeration zones.

Relationships between Gas Fluxes and Water and Tank Parameters.
A PCA was used to examine relationships between N2O fluxes and dissolved N concentrations across all zones and dates ( Figure 5). The first component explained 48% of the variance and the second component explained an additional 24% of the variance for a total of 72% explained by the first two principal components ( Figure 5).
Nitrate loaded on the first axis while ammonium, nitrite, and N2O flux loaded equally on both axis ( Figure 5). The opposite orientation of N2O and ammonium suggests that N2O fluxes had a strong inverse relationship with ammonium concentrations ( Figure   5). The small angle between N2O and nitrate suggests that N2O flux and nitrate had a moderate positive relationship ( Figure 5). The nearly 90° orientation between N2O and nitrite suggests that there was weak linkage between the two ( Figure 5).
There was not a significant relationship between DO concentration and any of the gas fluxes from the Aerated IFAS zone (data not shown). In the Post Anoxic zone there was a significant relationship between water temperature and CO2 fluxes (p = 0.04). There was a significant relationship between temperature and CH4 flux in the Aerated IFAS zone (p = 0.04).

Discussion
The new IFAS BNR system at Field's Point successfully removed total N during the course of this study (Table 1A). The lower percentage of N removal observed in the winter was likely because the N discharge limit is only in effect from May through October. At the end of the discharge limit season the air flow rate is often decreased by plant operators to lower expenses, which explains the low DO in the Aerated IFAS zone in the fall (Table 1). The decrease in microbial activity during colder months is also another key factor in the lower N removal in observed in the winter. The decrease in ammonium concentrations between the BNR system influent and the Pre-Anoxic zone was unexpected because the low DO concentration in this zone should limit nitrification (conversion of ammonium to nitrate) (Table 1). This decrease in ammonium is due to the fact that the Pre-Anoxic zone also receives internal mixed liquor recycle water from the Aerated IFAS zone. This water will be low in ammonium, therefore diluting the ammonium concentration in the Pre-Anoxic zone. Higher nitrate and nitrite concentrations in the Aerated IFAS zone were expected because the DO concentration in this zone is designed to favor nitrification (Table 1).

Overview of GHG Emissions
This study found measurable fluxes of all three GHGs from both aerated and anoxic zones of the IFAS BNR system. Comparisons to other studies indicate that direct GHG fluxes from IFAS BNR may be lower than from other methods of BNR.
Nitrous oxide fluxes in this study accounted for 0.01 to 0.34% of influent N. This falls at the lower end of the range (0.001 -8.2%) reported by previous studies on N2O emissions from various types of BNR processes (Tomaszek and Czarnota, 2015). The percent of influent chemical oxygen demand (COD) released as CH4 ranged from 0.02 to 0.13% kg CH4/kg influent COD. This was also at the low end of the range of reported CH4 emissions (0.07 -1.13%) from other BNR processes . Carbon dioxide fluxes represented 0.2 -1.1 kg CO2/kg influent COD, which is similar to the reported range from other BNR systems of 0.58 -0.97 kg CO2/kg COD .

Aerated IFAS zone as a hotspot for GHG emissions
Despite low overall emissions, the Aerated IFAS zone of the BNR system emitted the largest fluxes of all three GHGs compared to other zones, 75%, 74%, 82% of total N2O, CH4, and CO2 fluxes from the BNR tank respectively. Higher N2O, CH4, and CO2 fluxes from aerated zones vs. anoxic zones was expected as it has been reported by previous studies conducted in other BNR processes Ren et al., 2015;Wang et al., 2011).
In aerated zones, mechanical air stripping is thought to be the main process contributing to GHG emissions . However, oxygenated conditions may also lead to greater production rates of GHGs. For example, increased N2O fluxes from aerated zones may be due to increased nitrification, incomplete denitrification, or altered environmental properties, such as temperature, that indirectly affect N2O production and consumption (Aboobakar et al., 2013).
Our study did not definitively discern whether these GHGs were produced in the aerated zone or previous zones, however, our estimates of production rates for this study suggested that there was high production of all three GHGs in the Aerated IFAS zone (Table S2). Although these production estimates assume solids are inert and dissolved concentrations in the effluent of BNR are similar to those in the final clarifiers, they should provide a good indication of general trends in production in consumption. While N2O and CO2 production in the Aerated IFAS zone is expected, CH4 production in aerobic zones such as the Aerated IFAS zone may not typically be expected as methanogenesis occurs under strict anaerobic conditions . In this study, it is likely that anaerobic micro-sites occurred in biofilms on the inner portions of the floc and/or plastic media present in the Aerated IFAS zone leading to the high CH4 production in this zone. Other studies have documented strictly anaerobic bacteria and archaea (including methane producers) in oxic BNR reactors (Lens et al., 1995).

Temporal variability of GHG emissions
All three GHGs demonstrated temporal variability that was not a simple function of seasonality. N2O fluxes exhibited the largest range within a single zone (4 orders of magnitude) over the course of the year. Temporal differences in N2O fluxes were largely driven by two dates, one date in July when fluxes from the Aerated IFAS zone were large and one date in January when fluxes from the Re-Aeration zone were large ( Figure 2A). Unfortunately, water samples were not collected on the July date making it difficult to determine the cause of the increased fluxes. The January measurement was the date with the highest dissolved N2O concentrations overall and a consistent increase was observed in these values from the inflow to the re-aeration zone on this single date (Figure 2A and 3A). It is possible that the colder wastewater temperatures in January (Table 1) increased the solubility of N2O, such that it was not completely stripped in the first aeration zone (Aerated IFAS) but rather remained dissolved until the second aeration zone (Re-Aeration zone). This day had the highest N2O production in the Aerated IFAS, Post Anoxic, and Re-Aeration zones (data not shown). This was also the date with the highest inflow nitrate concentration, which suggests that denitrification may have been a source of N2O production. Out of the three gases, CO2 exhibited the lowest temporal variability ( Figure   2C). The lower CO2 fluxes in winter may have been a result of reduced microbial activity, although there was not a consistent significant relationship between water temperature and CO2 emissions. Another study of the N removal tanks at a nearby WWTP in New Hampshire also found that CO2 did not have a significant correlation with temperature (Czepiel et al., 1993). Lower air flow rates in the IFAS zone in the winter may be another explanation for the lower CO2 fluxes in the winter. CO2 production in the winter was not lower than other seasons. It is possible that the CO2 was still produced in the winter but remained in the dissolved form, supported by the higher dissolved CO2 concentrations and lower fluxes observed in the winter.

Relationships between GHG Fluxes and Water and Tank Parameters
The PCA indicated that N2O fluxes were related to both ammonium and nitrate. Since data from all zones were included in the PCA, it is possible that denitrification is an important contributing factor in anoxic zones whereas nitrification is important in aerated zones.
The lack of significant consistent relationships between any of the GHGs and DO, water temperature, SVI, and water flow rate was surprising. However, GHGs are known to result from complicated combinations of many biological and mechanical processes. The inter-relationship of the gases to each other, in contrast, suggest there may be another unmonitored environmental variable that affects them all. For example, the positive relationship between CO2 and CH4 fluxes in the Aerated IFAS zone suggests that these gases may have similar release mechanisms in this zone, likely striping due to mechanical aeration. More extensive data sets and alternative modeling techniques may improve the ability to predict which critical environmental factors govern GHG emissions from WWTPs.

Overall Greenhouse Gas Emissions
Simultaneously examination of the fluxes of all three GHGs (N2O, CH4, and CO2) enables us to evaluate the emissions in terms of CO2 equivalence (using global warming potentials of 265 for N2O and 28 for CH4). Doing so reveals that, the majority of the fluxes from BNR were CO2 (86%), followed by N2O (11%), and CH4 (3%). When comparing GHG fluxes from BNR to other sources of GHGs at the Field's Point WWTP in terms of CO2 equivalence, BNR is responsible for approximately 12% of the total GHG emissions (Table 3).
While it was not the intent of this study to distinguish the source of influent carbon as biogenic or anthropogenic, one recent study using stable radiocarbon isotope signatures to determine the origin of CO2 has shown that up to 6% of influent total organic carbon may be released as CO2 emissions of fossil origin (Law et al. 2013).
As the largest emissions in this study were from CO2, future studies that investigate the source of influent carbon to the WWTP will be important.
The IPCC emission factor is the current accepted method for estimating N2O emissions from WWTPs. However, the large variation in GHG emissions from different BNR methods makes it difficult to apply a single emission factor. The IPCC reports an N2O emission factor of 7 g N2O person -1 yr -1 for BNR processes Doorn et al., 2006). The average for our study was 9 g N2O person -1 y -1 but ranged from 1 to 32 g N2O person -1 y -1 . Another study that measured N2O fluxes from 12 BNR systems throughout the U.S. reported N2O per capita emission factors ranging from 0.28 to 92 g N2O person -1 y -1 , up to an order of magnitude higher than the one suggested by the IPCC . While the average N2O emission factor of this study was similar in magnitude to that reported by the IPCC, it is clear that the factor can vary widely even within the same WWTP. Therefore, further studies are needed to determine what environmental factors may be important in constraining this variation so that N2O emissions from WWTPs can be properly estimated.

Conclusions
This was the first study to examine N2O, CH4, and CO2 emissions simultaneously from an IFAS BNR system in the U.S. Although large temporal and spatial variability of all three GHG fluxes was observed, the N2O and CH4 fluxes were small compared to those reported for other types of BNR methods and relative to the influent N and COD. Further, efforts to reduce emissions should focus on the Aerated IFAS zone where the highest fluxes and estimated production was observed. As the majority of the fluxes were from CO2, future studies will need to discern the portion of CO2 emissions that are biogenic or anthropogenic.
 Fluxes of all three GHGs (N2O, CH4, and CO2) varied by 3 orders of magnitude over the course of the one year study  On average in terms of CO2 equivalence, the majority of the fluxes were from CO2 (4312 tonne CO2 y -1 ) rather than N2O (522 tonne CO2 y -1 ) and CH4 (159 tonne CO2 y -1 )  Only 0.01 to 0.34% of influent N is released as N2O and 0.02 to 0.13% of influent COD is released as CH4 (kg CH4/kg influent COD)  There were significant positive linear relationships between CH4 and CO2 fluxes in the Aerated IFAS and Re-Aeration zones and N2O and CO2 fluxes in the Pre-Anoxic zone  The largest emissions and estimated production were from the Aerated IFAS zone  BNR is responsible for approximately 12% of the total GHG fluxes for the Field's Point WWTP

Introduction
Humans substantially modify global nitrogen (N) cycles by industrially fixing N for fertilizer and ultimately releasing reactive N back to the environment through various mechanisms, including wastewater treatment. The continued growth of human population will lead to further increases in excess reactive N, increasing the need for N remediation . In recent years, remediation has focused on upgrading centralized wastewater treatment plants (WWTPs) to include biological nitrogen removal (BNR). Since one in five homes in the U.S. are serviced by conventional onsite wastewater treatment systems (OWTS) (United States Environmental Protection Agency (US EPA), 2013) they can also be large sources of N US EPA, 2015). The use of OWTS can be advantageous relative to centralized WWTPs, as they recharge groundwater supplies, require less infrastructure and have lower energy costs (US EPA, 2013). In order to ameliorate N inputs to the environment, conventional OWTS are also being upgraded to advanced OWTS that include BNR.
Although BNR systems at WWTPs and OWTS vary in design, all employ nitrifying (conversion of ammonium to nitrate) and denitrifying (conversion of nitrate to nitrogen gas) bacteria in oxic and anoxic environments, respectively . The systems are designed to remove N mainly in the form of N2 gas, the final product of denitrification. However, in addition to N2, the BNR process may produce substantial quantities of nitrous oxide (N2O), a greenhouse gas 265 times more potent than CO2 that can also deplete ozone in the stratosphere (Core Writing Team et al., 2014;Tomaszek and Czarnota, 2015). Nitrous oxide is produced by 89 microbial N transformations including nitrification and denitrification. Nitrification can produce N2O as a by-product and denitrification can be both a source and sink of N2O . Therefore, the abundance and biological activity of nitrifying and/or denitrifying bacteria is likely a key factor influencing the rates of these N transformations associated with N2O emissions.
Previous studies have documented the magnitude of N2O emissions relative to N removal rates from various types of BNR systems at centralized WWTPs, with emission factors (% of N load released as N2O) varying by over four orders of magnitude, 0.001 to 25.3 % (Tomaszek and Czarnota, 2015). In contrast, only one study published values for N2O emissions from advanced OWTS designed to remove N (Todt and Dorsch 2015). Biological nitrogen removal at both WWTPs and OWTS will become increasingly important as the human population and wastewater production, continues to increase. Therefore, the magnitude of N2O emissions from BNR of both WWTPs and OWTS should be determined in order to evaluate the effectiveness of these systems in N remediation and their potential impacts on greenhouse gas emissions. In addition, insights regarding the microbial sources of N2O emissions will help to discern the potential mechanisms by which they may be mitigated through technological and operational changes to wastewater treatment systems, while striving to maximize N removal.
We quantified and compared N2O emissions from BNR at a centralized WWTP and three types of advanced OWTS (Advantex, SeptiTech, and FAST) in terms of instantaneous emissions, normalized per capita emissions, and emission factors (% of N released as N2O). We also quantified and compared amoA (nitrification) and nosZ (denitrification) gene abundances and ratios from the same treatment systems to examine potential relationships between abundances of nitrifying and/or denitrifying bacteria and N2O emissions. A positive relationship between amoA abundance and N2O emissions would indicate that nitrification was likely responsible for the N2O emissions. A negative relationship between N2O emissions and nosZ would indicate that complete denitrification was a sink for N2O emissions.
Understanding the mechanism (nitrification or denitrification) responsible for the N2O emissions may allow for operational changes to reduce N2O emissions while maintaining N removal.

Study Sites and Measurement Locations
The wastewater systems we examined were within the Greater Narragansett Bay watershed in Rhode Island, USA. Field's Point is a full-scale centralized WWTP serving 226,000 people in Providence, RI (Narragansett Bay Commission, 2017). The plant provides primary and secondary treatment for flows up to 77 million gallons per day (MGD) for combined sewer from domestic and industrial sources. Secondary treatment includes an Integrated Fixed Film Activated Sludge (IFAS) system for BNR.
The IFAS system consists of 10 identical tanks, each with the following four main zones: (i) pre-anoxic, (ii) aerated IFAS, (iii) post-anoxic, and (iv) re-aeration. The aerated IFAS zone provides additional surface area for biofilm growth with the inclusion of perforated high-density polyethylene cylinder media (25-mm dia., 10-mm length). Two N2O emission measurements and water samples were collected in each of the four zones of one IFAS tank. Water samples were collected from just below the water surface within 3 hours of the emission measurements.
We examined three of the most commonly used advanced OWTS technologies for BNR in RI: Orenco Advantex AX20 (textile media filter), BioMicrobics MicroFAST (fixed activated sludge treatment unit), and SeptiTech D Series (trickling filter). All OWTS were located in Jamestown, RI, with measurements made in three systems per technology (9 systems total). All systems have an anoxic compartment for denitrification (SP1) and an oxic compartment for nitrification (SP2). We made one N2O emission measurement and collected one water sample from each compartment (SP1 and SP2) in each system per sampling event. The access riser lid to the systems was removed to allow trapped gases to vent for approximately 10 minutes before the emission measurement was made. Water samples were collected from the middle of the water column immediately after emission measurements were made.
Nitrous oxide emission measurements and wastewater samples were collected from each system once in June and once in October, resulting in a total of 16 measurements for the WWTP and 36 for the OWTS. Logistical constraints prevented sampling from all sites on the same day. Thus, sampling of all systems took place within two weeks of each other during each round of measurements.

Nitrous Oxide Emission Measurements
At each study site N2O emission measurements were made using a closed chamber connected to a real-time cavity ring down spectroscopy analyzer (Picarro G2508, 92 Santa Clara, CA) capable of measuring N2O approximately every two seconds (detailed in Brannon et al., 2016). At the centralized WWTP we used a transparent (polypropylene) rectangular floating chamber (height: 0.3 m, width: 0.3m, length: 0.5 m). At the OWTS sites, an open-bottom PVC cylindrical chamber (i.d.: 0.13 m, length: 0.40 m) was placed on the water so that the bottom was submerged 7.5 cm below the surface. The chamber was kept level and at a constant depth using a stabilizing bar that rested across the top of the access port. The chamber was deployed for between 3 and 10 minutes at all sites.
Gas emissions from all zones at the centralized WWTP, except the aerated IFAS zone, and both compartments of all OWTS sites were calculated as outlined in  for non-aerated stages. Due to the high aeration rates used in the IFAS zone at the centralized WWTP (~1457 standard cubic feet per minute (scfm)), emissions from this zone were calculated using a method for aerated stages which accounts for the effects of air flow .
The statistical significance of each gas emission was determined following Brannon et al. (2016), with the exception that, if the p-value of the linear regression of concentration over time was not statistically significant, then the flux was reported as zero. There were four measurements, two each from two different Advantex systems, that we were not able to calculate the emission value for because the concentration of another gas (CH4) measured by the analyzer exceeded the upper range of the analyzer and interfered with analysis of the target species (N2O).
For comparison across systems, N2O emissions were normalized by population and area of the treatment tank (mg N2O capita -1 d -1 ) according to supplementary equations 1 (WWTP) and 2 (OWTS). Also, N2O emission factors (mass/mass) were computed by normalizing the flux to the quantity of N removed, according to supplementary equations 3 (WWTP) and 4 (OWTS). For the IFAS BNR system at the centralized WWTP one normalized emission value and one emission factor (mass/mass) was calculated for each date that included the total emissions for the IFAS system (all four zones of all 10 tanks). For the OWTS one normalized emission value and one emission fraction (mass/mass) was calculated for each house on each date (n = 6 per technology).

DNA Extraction
Genomic DNA was extracted from water samples from the WWTP and OWTS. For the centralized WWTP samples, approximately 50 mL of sample was centrifuged at 3,000 xg for 15 minutes and the solids were used for DNA extraction using a PowerSoil DNA Isolation Kit (MoBio Laboratories, Inc., Carlsbad, CA). For the OWTS approximately 100 mL of sample was vacuum filtered onto sterile 0.22μm-pore size nitrocellulose membrane filters (Millipore Corporation, Darmstadt, Germany). Non-sterile filters were used for 12 samples, but blanks were included to check for contamination. The filter was used for DNA extraction using a PowerWater DNA Isolation Kit (MoBio Laboratories, Inc., Carlsbad, CA). The quality and concentration (ng/μL) of all extracted DNA was determined with a NanoDrop 8000 UV-Vis spectrophotometer (Thermo Fisher Scientific, Wilmington, DE) and stored at -20°C or below until qPCR analysis.

Quantitative PCR (qPCR)
The concentrations of ammonia monooxygenase genes (amoA) and nitrous oxide reductase genes (nosZ) were quantified by real-time polymerase chain reaction (qPCR) using the primer sets developed by Geets et al. (2007) and Junier et al. (2009) (Supplementary Table S1). Individual standard curves were prepared for each gene from a sample that presented one clear band of the correct size after PCRamplification and was purified with a QIAquick PCR Purification Kit (Qiagen, Germantown, MD). The concentration (ng/μL) of purified products that served as standards was determined using an Invitrogen Qubit 2.0 fluorometer (Thermo Fisher Scientific, Carlsbad, CA) and converted to copies/μL. Ten-fold serial dilutions of the purified product were prepared from 10 7 to 10 1 copy numbers/μL. for 10 min, 45 cycles at 94°C for 10 sec, 61°C for 15 sec, and 72°C for 20 sec. The thermocycler settings for amoA were as follows: 94°C for 10 min, 45 cycles at 94°C for 10 sec, 54°C for 10 sec, and 72°C for 14 sec. Amplification efficiencies for both genes ranged from 78 to 100%. A melt curve was analyzed for every run and the qPCR product for one of each triplicate was examined on a 1% (w/v) ethidium bromide-stained agarose gel to confirm the amplification of a single product for both genes. In addition to concentration (copies/μL), the abundance of each gene (copies/ng nucleic acid) was calculated using the qPCR results and the total concentration of DNA.

Wastewater Properties
For WWTP samples, a subset of the water sample used for qPCR analysis was filtered (0.45-μm-pore-size) and the filtrate used to determine the concentration of NH4 using the phenolhypochlorite method (Solorzano, 1969) and nitrate using the dimethylphenol method (Hach Company, 2015). The surface pH (Seven Go Duo Pro, Metler Toledo, Columbus, OH) and DO (LDO Probe, HACH Model 57900-00, Loveland, CO) were measured within 2 hours of the emission measurements. The water temperature was continuously measured in the IFAS zone only with a LDO probe (HACH Model 57900-00, Loveland, CO). The average water temperature during the time of the flux measurements is reported in Table S2.
For the OWTS samples, a Hanna Instruments HI9828 Multiparameter Meter (Woonsocket, RI) was used to determine wastewater pH, DO, and temperature in the field in each compartment. A subset of the sample used for qPCR analysis was used to determine the concentration of ammonium, nitrate, and BOD5 as described in Lancellotti (2016).

Statistical Analysis
We used linear regressions to examine relationships between N2O emissions and gene abundances and amoA/nosZ ratios; between N2O emissions and the wastewater properties; and gene abundances and amoA/nosZ ratios and the wastewater properties. Two separate regressions were performed: one for nitrification zones (aerated IFAS and re-aerations zones for the WWTP and SP2 for the OWTS) and one for denitrification zones (pre-anoxic and post-anoxic zones for the WWTP and SP1 for the OWTS). Gene concentrations below the detection limit of 10 copies/µL were assigned a value of zero. Wastewater properties below the detection limit were assigned a value of zero. All data were checked for normality and transformed when necessary. All statistical analyses were performed using JMP (Version 13, SAS Institute, Inc., Cary, NC, 1989

Nitrous Oxide Emissions
The largest N2O emissions at the WWTP were from the aerated IFAS zone and the post-anoxic zone, while emissions from the pre-anoxic and re-aeration zones were relatively low ( Figure 1A). The emissions of N2O from the WWTP represented between 0.02 and 0.04% of N removed, which is in the lower end of the range (0.001 -25.3%) reported by studies from other types of BNR systems at WWTPs (Tomaszek and Czarnota, 2015).
Our study is the first to measure N2O emissions from advanced OWTS designed for N removal. The Advantex systems had the highest N2O emissions of the 97 three OWTS (Figure 1A), and emissions were similar between SP1 (denitrification) and SP2 (nitrification) for all OWTS systems ( Figure 1A). Similar to the WWTP, the N2O emissions from the SeptiTech and FAST OWTS represented a relatively small percent of the N removed (0.0 -4.4%). In contrast, the N2O emissions from the Advantex systems represented a much higher percent of the N removed (0.05 -21.00 %). This suggests that conditions within the Advantex treatment train favor N2O, rather than N2, production. For example, the Advantex systems had the lowest pH (6.4) (Supplementary Table S2). Previous studies have demonstrated that nosZ is sensitive to low pH (<6.5) resulting in reduced conversion of N2O to N2 .
The emissions of N2O from the aerated IFAS and post-anoxic zones at the WWTP were higher than those from all three OWTS ( Figure 1A). In contrast, emissions from the pre-anoxic and re-aeration zones at the WWTP were similar in magnitude to those from all three OWTS ( Figure 1A). It is not surprising that the highest N2O emissions in this study are from the aerated IFAS zone of the WWTP, since it uses high air flow rates (on average 1638 scfm) compared to the OWTS (FAST: 17 -25 scfm, SeptiTech: venture air intake, and Advantex: passive air diffusion). Higher air flow rates at the WWTP may cause higher N2O emissions due to mechanical stripping of dissolved N2O. There was not a significant relationship between N2O and any of the wastewater properties in either the nitrification or denitrification components of these systems (data not shown).
Although N2O emissions were observed from all systems, a negative N2O flux (indicating uptake or consumption) was observed on 2 occasions (2 measurements in 98 the WWTP re-aeration zone) out of 34 measurements total. While negative N2O fluxes have not been reported for BNR systems, they have been observed in soil (Chapuis-Lardy et al., 2007). It is generally assumed that heterotrophic denitrification is responsible for N2O consumption (Chapuis-Lardy et al., 2007) and that in those cases, the N2O is being reduced fully to N2. Since NO3is a preferred electron acceptor over N2O and nosZ is sensitive to oxygen, it is likely that N2O uptake is confined to N-limited systems with low DO (Chapuis-Lardy et al., 2007). However, the two N2O uptake events in this study did not coincide with excessively low NO3or DO levels in the wastewater. Therefore, circumstances resulting in N2O uptake are unclear.
We used the total surface area and estimates of the number of individuals served by each system to calculate normalized N2O emission values, which ranged from 0 to 624 mg N2O capita -1 d -1 (Figure 2). The average for the WWTP was 6.0 mg N2O capita -1 d -1 , at the lower end of the range (0.8 to 383.6 mg N2O capita -1 d -1 ) reported for other types of BNR systems at WWTPs . The average N2O emission from OWTS in this study (60 mg N2O capita -1 d -1 ) is the first to our knowledge to be reported for any advanced OWTS and is higher than that determined from one conventional OWTS (without BNR) (5 mg N2O capita -1 d -1 ) (Diaz-Valbuena et al., 2011). Another study measured N2O emissions from the roof vent (0.013 t CO2e capita -1 yr -1 ), sand filter (6.5 x 10 -4 t CO2e capita -1 yr -1 ), and leach field (2.4 x 10 -99 reported by Truhlar et al. (2016). Our results suggest that advanced OWTS designed for N removal may have higher N2O emissions than conventional advanced OWTS lacking N removal. The benefits of N removal at both WWTPs and OWTS may therefore come at the cost of increasing N2O in the atmosphere, which would transfer the N problem from one environment (wastewater) to another (the atmosphere). As more advanced OWTS are installed and/or WWTPs are upgraded to include BNR, they may become a significant source of N2O.

Nucleic Acid Concentration
The concentration of nucleic acids (a proxy for the size of the microbial community) in all zones at the WWTP was five times higher than those of the three OWTS ( Figure 3). This is interesting because it does not appear that the WWTP receives larger carbon inputs compared to OWTS. Although the BOD of the influent to the OWTS in this study was not measured, it typically ranges from 145 to 386 mg/L (Soil Science Society of America, 2014), which is similar to the average BOD of the WWTP influent in this study (200 mg/L) (Supplementary Table S2). The nucleic acid concentration was generally higher in SP1 (denitrification compartment) compared to SP2 (nitrification compartment) in all three of the OWTS (Figure 3). This is not surprising because SP1 of OWTS receive septic tank effluent with high BOD (Supplementary Table S2).

Nitrifier (amoA) and denitrifier (nosZ) specific abundance
In general, amoA specific abundance was higher at the WWTP than any of the three OWTS technologies, except SP1 of FAST and SP2 of Advantex ( Figure 1B). At the WWTP, the lowest amoA abundance was in the pre-anoxic zone, while the abundance in the other three zones (aerated IFAS, post-anoxic, and re-aeration) was similar in magnitude ( Figure 1B). Out of the three OWTS, the highest amoA abundance was in FAST systems ( Figure 1B). In addition, there was a trend of higher amoA abundance in the SP2 than SP1 in Advantex and SeptiTech systems but not FAST systems ( Figure 1B). There was a significant positive relationship between amoA abundance and DO in denitrification zones/compartments (p < 0.01, r 2 = 0.88).
The specific abundance of amoA in this study, 0 to 10 2 copies/ng DNA, was within the range reported from other BNR systems (10 1 to 10 5 copies/ng DNA) including an integrated anoxic/oxic reactor (Wang et al., 2014) and conventional activated sludge (Song et al., 2014).
The specific abundance of nosZ did not follow the same trends within and between system types as amoA abundance (Figure 1). The specific abundance of nosZ was generally higher in all three OWTS than in all four zones of the WWTP ( Figure   1C). At the WWTP, there was higher nosZ abundance in the aerated zones (aerated IFAS and re-aeration) compared to the anoxic zones ( Figure 1C). This was surprising, since we expected that the higher DO concentrations of the aerated zones would result in lower nosZ abundance, as it is part of an anaerobic pathway. However, it is possible that the high DO levels were maintaining a supply of oxidized N (as NO3 -) that supported the growth of denitrifiers (many of which contain nosZ). Another study of BNR systems at WWTPs found a similar trend of higher nosZ abundance in aerobic zones compared to anoxic zones (Wang et al., 2014). Further, in our study there was a significant, albeit weak, positive relationship between nosZ abundance and nitrate in the nitrification zones/compartments (p < 0.01, r 2 = 0.31). Some microorganisms can reduce nitrate even in the presence of relatively high DO concentrations (Robertson and Kuenen, 1984;Zhang et al., 2016). Although we do not know if the microorganisms in this study were actively reducing N2O, we do know that they had the genetic capacity to do so and were relatively abundant in the aerated zones.
The abundance of nosZ was similar among the three OWTS ( Figure 1C), which suggests it did not play a strong role in accounting for notable differences in N2O emissions from the systems ( Figure 1). As expected, there was a trend of higher nosZ abundance in SP1 than SP2 for FAST and SeptiTech systems ( Figure 1C). The specific abundance of nosZ ranged from 0 to 10 3 copies/ng DNA, and was larger and more variable than that of amoA, but was lower than reported from other types of BNR systems at WWTPs (10 4 -10 5 copies/ng DNA) (Song et al., 2014;Wang et al., 2014).
The ratio of amoA to nosZ was higher in all zones of the WWTP than all three OWTS technologies (Figure 4). In some instances the amoA/nosZ ratio at the WWTP was above one, indicating that there was a higher abundance of amoA than nosZ ( Figure 4). In contrast, the amoA/nosZ ratio for OWTS was only above one once ( Figure 4). The higher amoA/nosZ ratio at the WWTP seems to be related to the high N2O emissions observed there. However, there was not a significant relationship between N2O emissions and amoA/nosZ ratio among either the nitrification or denitrification zones/compartments of all systems (data not shown). The strongest relationship of amoA/nosZ was with BOD in nitrification zones/compartments (p = 0.01, r 2 = 0.43).

Relationships between gene abundance and N2O Emissions
In our study, there was no significant relationship between N2O emissions and amoA or nosZ abundance or wastewater properties for nitrification or denitrification zones/compartments (data not shown). This indicates that neither nitrification (amoA) nor denitrification (nosZ) are solely responsible for the N2O emissions. The lack of statistically significant relationships was not particularly surprising. First, gene abundance indicates population size of specific microbial groups but not gene expression. For example, other studies have found that although abundance of DNA (amoA and nosZ) did not differ between BNR trains at a WWTP, mRNA gene expression did (Song et al., 2014). Further, they found a strong negative relationship between nosZ expression and N2O emissions (Song et al., 2014). Secondly, we collected water samples from a single depth. The abundance and activity of nitrifiers and denitrifiers likely varies with depth as a function of DO concentration. In addition, the production mechanism of N2O emissions may be more complicated than simple production by autotrophic nitrification or heterotrophic denitrification. For instance, nitrifier denitrification, the reduction of NO2to N2O and N2 by nitrifiers, is another potential source of N2O . Although there were no linear significant relationships between N2O emissions and amoA and nosZ abundance, there were interesting trends. Generally, the centralized WWTP had larger microbial populations (indicated by nucleic acid concentrations), lower nosZ abundance and therefore higher amoA/nosZ ratios compared to OWTS. This indicates that the higher N2O emissions at the WWTP (compared to OWTS) may be due to a larger nitrifying population (N2O source) and smaller complete denitrifying population (N2O sink).

Conclusion
This preliminary evaluation of N2O emissions from three advanced OWTS technologies indicates that they are generally lower (on a mole/area basis) relative to an IFAS BNR system at a centralized WWTP. However, when the N2O emissions were normalized per population served and area of treatment tanks, they were similar between the WWTP and OWTS. Among the three technologies of advanced OWTS that were evaluated, the one with the highest N2O emissions was the Advantex system.
Overall, the BNR systems examined in this study do not produce large N2O emissions relative to the amount of N removed, mostly <1%. The WWTP had higher amoA abundance and lower nosZ abundance compared to the OWTS. However, N2O emissions were not directly related to amoA nor nosZ abundance or to the wastewater properties we evaluated.
Further evaluation of N2O emissions from emerging BNR technologies and their microbial sources should be conducted, particularly as they become increasingly numerous as wastewater treatment demands increase.    The amount of influent and effluent N removed for OWTS could not directly be measured. Instead it was assumed that all systems removed Table S2. Average and standard error of wastewater properties from pre-anoxic, aerated IFAS, post-anoxic, and re-aeration zones in the wastewater treatment plant and denitrification (SP1) and nitrification (SP2) compartments in Advantex, FAST, and SeptiTech (onsite wastewater treatment systems).

REMOVAL AT A CENTRALIZED WASTEWATER TREATMENT
Since the greenhouse gas emissions measured in Chapter 2 were only collected during the morning hours, additional measurements were made throughout the day in order to examine hourly variability of the emissions. These measurements were performed on five days during 2014 and 2015 and employed the same methods as those outlined in Chapter 2. Three measurements were collected every 1.5 hours from 9:30 am to 3:30 pm from the same location in the re-aeration zone.
In general, CH4 and CO2 fluxes did not vary with time ( Figure 1). However, on three dates (10/14/14, 6/17/14, and 6/30/15), N2O fluxes increased throughout the day. On one date (10/14/14), N2O fluxes were 50 times greater in the afternoon than morning. The results of this additional study supplement the large variability of N2O emissions observed in Chapter 2 (over 3 orders of magnitude). This continues to highlight the need for a better understanding of the large variation observed in emissions, especially N2O. Measurable fluxes of all three gases (N2O, CO2, and CH4) were recorded from the grit chamber ( Figure 2). However, only CO2 and CH4 fluxes were measured from the primary and secondary clarifiers (Figure 2). Out of the three additional components examined, the highest emissions of all three GHGs were from the grit chamber ( Figure 2).
Fluxes of all three GHGs from the additional components (grit chamber, primary clarifier, and secondary clarifier) were lower than the average emissions from all zones of the BNR tank measured the previous year, except for CH4 fluxes from the grit chamber ( Figure 2). Fluxes of N2O from the grit chamber were three orders of magnitude lower than the average from the BNR zone (aerated IFAS) with the highest N2O emissions and one order of magnitude lower than the average from the BNR zone (pre-anoxic) with the lowest N2O emissions (Figure 2A). Fluxes of CO2 from all three of the additional components were one and two orders of magnitude lower than the average CO2 fluxes from the re-aeration and aerated IFAS zones respectively of the BNR tank, but on the same order of magnitude as those from the pre-anoxic and postanoxic zones ( Figure 2B). Fluxes of CH4 from the secondary clarifier were at least an order of magnitude lower than those from the BNR tank ( Figure 2C). However, CH4 fluxes from the primary clarifier were on the same order of magnitude as those from the pre-anoxic and re-aeration zones of the BNR tank and fluxes from the grit chamber were on the same order of magnitude as those from the aerated IFAS zone ( Figure   2C).
It was not surprising that out of the three additional components examined, the grit chamber had the largest emissions of all three GHGs ( Figure 2). The grit chamber is aerated and mechanical stripping is likely leading to increased emissions. A study by  found a similar trend of higher N2O, CH4, and CO2 emissions from grit tanks compared to clarifiers. It is also not surprising that CH4 fluxes from the grit chamber were high. Wastewater entering the plant likely contains high concentrations of dissolved CH4 as anaerobic conditions have been documented in sewers . The grit chamber is the first component that contains aeration which likely results in the stripping of all the dissolved CH4 that has accumulated in the influent pipes. The fact that additional CH4 fluxes were observed in the aerated zones of the BNR tank suggests that either some dissolved CH4 was not stripped in the grit chamber or additional CH4 production occurred in the primary clarifiers (located after the grit chamber) and was stripped in the aerated zones of the BNR tanks. Future studies should include measurements of dissolved CH4 so that the location of CH4 production relative to emission can be determined. Low to no N2O emissions from the additional components was expected because the components are not designed to include nitrogen removal and therefore would not likely contain nitrifying and denitrifying organisms responsible for N2O production. Likewise, low CO2 emissions were expected because the components are not designed to contain large microbial populations.
In terms of CO2 equivalence (using global warming potential of 265 for N2O and 28 for CH4) and normalizing by tank surface area, the average total emissions (including N2O, CO2, and CH4) from the grit chambers, primary clarifiers, and final clarifiers are 12, 66, and 32 tonne CO2 eq. y -1 , respectively. This is compared to 6637 tonne CO2 eq. y -1 from all four zones and 10 tanks of BNR. It should be noted that measurements were only collected on one date from the grit chambers, primary clarifiers and final clarifiers. Additional studies are needed to determine the temporal variability of the emissions. However, the measured GHG emissions from the grit chamber, primary clarifier and secondary clarifier combined represented less than 0.5% of the WWTPs total GHG budget, while BNR represented 12%. Therefore, future efforts to reduce emissions should focus on the BNR tanks.  Nitrous oxide can be produced by several processes including, but not limited to, nitrification, denitrification, and nitrifier denitrification . To complicate matters, denitrification can both consume and produce N2O . Although several studies have attempted to determine the mechanism of N2O emissions from BNR processes, no consensus has been found (Gejlsbjerg et al., 1998;Schramm et al., 2000;Tallec et al., 2006;Wunderlin et al., 2012). A better understanding of the mechanism responsible for N2O emissions will help develop mitigation strategies.
Examination of isotopomers is one approach that can help determine the mechanisms responsible for N2O production. Isotopomers refers to the intramolecular distribution of 15 N within the N2O molecule, also called site preference (Wunderlin et measuring range of the analyzer. All the data is included in the analysis below but results should be evaluated with caution. Although there does not appear to be a large difference in site preference between zones, there was a difference between dates (Figure 1). In June site preference was negative in all four zones, indicating that N2O emissions in June (generally low, average: 3.9 x 10 -2 ) were from either denitrification or nitrifier denitrification ( Figure 1). Since, Δδ 15 N was not measured in this study it is not possible to distinguish between denitrification and nitrifier denitrification. However, in July and August, when N2O emissions were generally higher (average: 5.6 x 10 -2 ), site preference was positive in all zones, indicating the nitrification was the source of N2O (Figure 1).
Despite the complications (long storage time resulting in substantial gas loss) mentioned above, the results indicate that site preference may be useful in understanding temporal differences in N2O fluxes. Therefore, future studies that include adjusted methodology are warranted and should include bulk Δδ 15 N in order to differentiate the contribution of nitrifier denitrification and denitrification to N2O production.  The CO2 and CH4 fluxes measurements were made on the same dates (June and October) and collected from the same locations as those outlined in Chapter 3.
This was the first known study to directly compare all three GHG emissions between a WWTP and advanced OWTS designed to remove nitrogen. Although CO2 and CH4 emissions were observed from all systems, CH4 uptake was observed on one occasion (SP2 of FAST). In addition, several CO2 and CH4 fluxes were either zero or below the detection limit (WWTP = 0, Advantex = 0, Septi = 1, FAST = 4). Four fluxes from two Advantex (3 from SP1 and 1 from SP2) systems were above the analyzer's detection limit for CH4. Since the analyzer measures all three gases simultaneously, CO2 flux measurements could not be made for those Advantex systems.
Emissions of CO2 at the WWTP were an order of magnitude higher from the aerated zones (aerated IFAS and re-aeration) than the anoxic zones (pre-anoxic and post-anoxic) ( Figure 1A). The WWTP CO2 fluxes represented 0.25 to 0.40 kg CO2/kg influent chemical oxygen demand (COD). This is below the range (0.58 to 0.97 kg CO2/kg COD) reported by studies from other types of BNR systems at WWTPs . Similar to the N2O emissions, the CO2 emissions reported here (0.25 to 0.4 kg CO2/kg influent COD) were on the lower end of the range for the yearlong measurements conducted in Chapter 2 of this dissertation (0.2 -1.1 kg CO2/kg influent COD).
The largest CO2 emissions from the OWTS were from the Advantex system ( Figure 1A). The Advantex systems also had the highest biological oxygen demand (BOD) values. A higher BOD can result from increased activity of microorganism that respire CO2. There was a trend of higher CO2 emissions from SP1 than SP2 for all three OWTS systems ( Figure 1A). This was not surprising as SP1 receives influent water that has high BOD.
When comparing CO2 emissions between the WWTP and OWTS, CO2 emissions from the aerated IFAS and re-aeration zones at the WWTP were higher than those from all three OWTS ( Figure 1A). However, CO2 emissions from the preanoxic and post-anoxic zones were similar in magnitude to those from SP1 of Advantex systems ( Figure 1A). All other OWTS had CO2 emissions below those at the WWTP ( Figure 1A).
Methane emissions at the WWTP were highest from the aerated IFAS and post-anoxic zones ( Figure 1B). The CH4 emissions from the WWTP represented 0.05 to 0.09% kg CO2/kg COD. This is at the lower end of the range (0.07 to 1.13% kg CH4/kg COD) reported by studies from other types of BNR systems at WWTPs . Similar to the N2O and CO2 emissions, the CH4 emissions reported here (0.05 to 0.09 kg CO2/kg COD) were on the lower end of the range for the yearlong measurements conducted in Chapter 2 of this dissertation (0.02 to 0.13% kg CH4/kg influent COD).
The largest CH4 emissions from the OWTS were from Advantex SP1 ( Figure   1B). This was also the system that had CH4 emissions above the measuring range of the analyzer on four occasions. The Advantex SP1 also had the lowest dissolved oxygen (DO) out of all the systems which might explain the higher CH4 emissions.
Other studies have found a weak correlation between CH4 emissions and DO in other BNR systems (Wang et al., 2011;. There was a trend of higher CH4 emissions from SP1 than SP2 for all three OWTS systems ( Figure 1B).
This was not surprising as the SP1 compartment is designed to have lower DO than SP2.
When comparing CH4 emissions between the WWTP and OWTS, CH4 emissions were highest from the Advantex SP1 system followed by the aerated IFAS and post-anoxic zones at the WWTP and SeptiTech SP1 ( Figure 1B). All other systems had similar CH4 emissions (pre-anoxic and re-aeration at WWTP, SP2 Advantex, SP1 and SP2 of FAST and SP2 of SeptiTech). and nitrification (SP1) and denitrification (SP2) compartments in Advantex, FAST, and SeptiTech (onsite wastewater treatment systems). Solid line in middle of box represents the median, edge of box represents 1 st and 3 rd quartile, and whiskers extend 1.5 x the inter quartile range beyond the edge of the box.

DISCUSSION
As the human population continues to grow, biological nitrogen removal (BNR) systems at both wastewater treatment plants (WWTPs) and onsite wastewater treatment systems (OWTS) will become increasingly common. Therefore, it is important that we determine the impact of these BNR systems on greenhouse gas (GHG) emissions. This research is among the first to apply an analyzer (Picarro G2508) that uses cavity ring down spectroscopy technology (Chapter 1) to measure the emission of three major GHGs (N2O, CH4, CO2) from BNR systems.
The results of this dissertation indicate that although N2O emissions from both WWTPs and OWTS are generally low (<1% of N removed) they can be variable, resulting in high emissions at times (up to 21% of N removed for one OWTS) (Chapters 2 and 3). The large variability in N2O emissions was particularly clear in the re-aeration zone at the WWTP where emissions varied by over 4 orders of magnitude throughout the year (Chapter 2) and on one occasion N2O emissions from the re-aeration zone were 50 times greater in the afternoon than the morning (Appendix 1). Despite the large variation in N2O emissions, the results of this dissertation determined the zones (aerated IFAS at WWTP) and systems (WWTP and Advantex) where the highest emissions were observed, highlighting areas to focus emission reduction efforts (Chapters 2 and 3).
In addition to N2O emissions, this dissertation demonstrated the importance of measuring all three GHGs simultaneously as CH4 and CO2 emissions followed different trends than N2O emissions. While N2O emissions were generally low, this was not the case for CH4 and CO2 emissions. Methane emissions represented the largest proportion of the emissions from the BNR systems in the OWTS (Appendix 4).
This is particularly important due to the high global warming potential of CH4 (28).
However, at the WWTP, CO2 emissions were highest of the three gases (Chapter 2). This is significant because CO2 emissions are considered to be of biogenic origin by the Intergovernmental Panel on Climate Change and are therefore excluded from GHG budgets. However, a recent study concluded that part of the CO2 emissions from wastewater treatment may be of fossil (anthropogenic) origin and should therefore be included in budgets (detailed in Chapter 2).
In addition, when considering the emissions of all three GHGs at the centralized WWTP, the BNR system represented a significant proportion (12%) of the total emissions of the WWTP (Chapter 2). Further, when emissions from the BNR system are compared to other sources of direct emissions (grit tanks, primary clarifiers, final clarifiers) the BNR tanks generally had the highest emissions of all three GHGs (Appendix 2). In combination, these findings indicate that the BNR tanks represent a significant proportion of the WWTPs direct GHG emissions.
The large variability observed in GHG emissions, especially N2O, highlights the need to better understand the mechanisms responsible for emissions. Preliminary results from the isotopomer analysis indicate that nitrification is responsible for the N2O emissions at the WWTP and that there is not a difference in the source between the zones (Appendix 3). This is supported by the observation that amoA abundance (nitrification gene) was not significantly different between the zones (Chapter 3) and the inverse relationship between ammonium concentration and N2O emission (Chapter 2). Future research should focus on further understanding the mechanisms responsible for the emissions. Only then can operational changes be suggested in order to reduce emissions while maintaining N removal.
This research highlights the potential of BNR systems to be sources of N2O, CH4, and CO2 indicating that increased GHG emissions may be a tradeoff of reduced N loads to coastal ecosystems. Increases in the human population will only exacerbate this issue and future studies will need to evaluate the implications of this tradeoff.