Impact of Chronic Nitrogen Loading on Greenhouse Gas Fluxes in Coastal Wetlands

Coastal wetlands are valuable ecosystems that historically have not been protected and have been lost at rapid rates. Recently, they have gained attention for their potential role in climate change mitigation given that they have the ability to sequester and store large amounts of atmospheric carbon dioxide (CO2). While other ecosystems can sequester and store carbon as well, salt marshes have the unique ability to store vast amounts of carbon while emitting relatively negligible amounts of methane (CH4) and nitrous oxide (N2O). These are additional greenhouse gases (GHGs) that can be emitted from ecosystems, particularly CH4 in large quantities from freshwater wetlands. These gases are 45 and 270 times, respectively, greater at trapping heat in the atmosphere than CO2. However, anthropogenic nitrogen (N) inputs into coastal estuaries have the potential to shift biogeochemical cycling within coastal wetlands, possibly switching salt marshes from CO2 sinks to being sources of one or more of the three major GHGs. Excessive anthropogenic N inputs are a threat to overall ecosystem health on local, regional, and global scales. Salt marshes are natural and efficient filters of excess N entering into these systems, although they cannot filter out unlimited quantities. Excess N in coastal systems can lead to a suite of negative consequences including poor water quality as a result of over stimulation of primary productivity and overall habitat degradation. The threat of anthropogenic N to coastal areas is only increasing as populations grow and concentrate along desirable coastal locations. For coastal wetlands that already face threats of habitat loss from increased rates of sea level rise (SLR) and urban development, N inputs can exacerbate rates of marsh loss. As efforts expand to protect these valuable ecosystems through development of financial incentives, such as carbon trading markets, it is important to quantify how N loading impacts GHG fluxes within wetlands and rates of N transformation. Most research to date in salt marsh systems has focused on impacts from short-term N additions on GHG fluxes. The goal of this research was to examine the role of chronic N loading on GHG fluxes in Spartina alterniflora-dominated marshes and to assess quantities of N available for transformation through measurement of denitrification enzyme activity (DEA). To accomplish these goals, we first examined the role of chronic N loading on GHG fluxes using three salt marshes located along a historic N gradient (high, medium, low) within Narragansett Bay, RI. Narragansett Bay is an ideal location for this work since it has received chronic N loading, mainly from wastewater inputs, since the late 1800s. To asses impacts of N loading on GHG fluxes, CO2, CH4 and N2O were measured for one field season in 2016. Along with measured fluxes, plant properties, edaphic parameters, and nutrient availability were measured. Relationships of fluxes to these additional parameters were then explored. We then compared rates of DEA at the opposite ends (high, low) of the N gradient within Narragansett Bay in 2017, focusing on four marsh zones (creekbank, mudflat, low marsh, high marsh) at two sites to assess any differences in N availability and rates of N transformation. Additionally, GHG fluxes were measured at the high N site in 2017 to explore relationships with DEA rates. As a result of this work, we found that the site receiving the highest N loading experienced the highest CO2 uptake as well as the highest emissions of CH4 and N2O compared to the other two sites along the N gradient. However, these emissions were not on an order of magnitude to significantly offset CO2 uptake. This was as expected, however, the other measured parameters (plant properties and edaphic variables) and DEA did not necessarily fall along expected trends of the N gradient. There were no significant differences in DEA among sites or zones, suggesting each site had similar amounts of N available for transformation and that soil in each zone had equal ability to transform N. At the marsh with the highest historic N loading, GHG fluxes fell along expected trends among zones with increased uptake of CO2 within vegetated zones contrasting with CO2 emission in non-vegetated zones. CH4 fluxes were highest in the bare creekbank zone, but were similar among the three remaining zones. Surprisingly, no significant N2O fluxes were measured in any of the four zones, suggesting along with DEA results that most N inputs are completely reduced to N2 via denitrification. In an effort to strengthen research and policy aimed at protecting and restoring these valuable ecosystems, it is important to continue to explore the dynamics between N, DEA rates, and all three GHGs. Of particular importance is to measure GHG fluxes and DEA across longer temporal scales. Additionally, examining actual denitrification rates from these marshes and also discerning key factors that help maintain the capacity for N transformation even as marsh landscapes shift as a result of SLR are important future directions.

expand to protect these valuable ecosystems through development of financial incentives, such as carbon trading markets, it is important to quantify how N loading impacts GHG fluxes within wetlands and rates of N transformation. Most research to date in salt marsh systems has focused on impacts from short-term N additions on GHG fluxes.
The goal of this research was to examine the role of chronic N loading on GHG fluxes in Spartina alterniflora-dominated marshes and to assess quantities of N available for transformation through measurement of denitrification enzyme activity (DEA). To accomplish these goals, we first examined the role of chronic N loading on GHG fluxes using three salt marshes located along a historic N gradient (high, medium, low) within Narragansett Bay, RI. Narragansett Bay is an ideal location for this work since it has received chronic N loading, mainly from wastewater inputs, since the late 1800s. To asses impacts of N loading on GHG fluxes, CO2, CH4 and N2O were measured for one field season in 2016. Along with measured fluxes, plant properties, edaphic parameters, and nutrient availability were measured. Relationships of fluxes to these additional parameters were then explored. We then compared rates of DEA at the opposite ends (high, low) of the N gradient within Narragansett Bay in 2017, focusing on four marsh zones (creekbank, mudflat, low marsh, high marsh) at two sites to assess any differences in N availability and rates of N transformation. Additionally, GHG fluxes were measured at the high N site in 2017 to explore relationships with DEA rates.
As a result of this work, we found that the site receiving the highest N loading experienced the highest CO2 uptake as well as the highest emissions of CH4 and N2O compared to the other two sites along the N gradient. However, these emissions were not on an order of magnitude to significantly offset CO2 uptake. This was as expected, however, the other measured parameters (plant properties and edaphic variables) and DEA did not necessarily fall along expected trends of the N gradient. There were no significant differences in DEA among sites or zones, suggesting each site had similar amounts of N available for transformation and that soil in each zone had equal ability to transform N. At the marsh with the highest historic N loading, GHG fluxes fell along expected trends among zones with increased uptake of CO2 within vegetated zones contrasting with CO2 emission in non-vegetated zones. CH4 fluxes were highest in the bare creekbank zone, but were similar among the three remaining zones. Surprisingly, no significant N2O fluxes were measured in any of the four zones, suggesting along with DEA results that most N inputs are completely reduced to N2 via denitrification.
In an effort to strengthen research and policy aimed at protecting and restoring these valuable ecosystems, it is important to continue to explore the dynamics between N, DEA rates, and all three GHGs. Of particular importance is to measure GHG fluxes and DEA across longer temporal scales. Additionally, examining actual denitrification rates from these marshes and also discerning key factors that help maintain the capacity for N transformation even as marsh landscapes shift as a result of SLR are important future directions.
v ACKNOWLEDGMENTS I owe the sincerest thank you to a wonderful and talented group of people who all inspired and helped me throughout the completion of my Master's. I would like to begin by thanking my advisor, Dr. Serena Moseman-Valtierra for her guidance and support. I can distinctly remember spending a day in the field with her and surprising her a bit when I announced towards the beginning of my degree that I had never presented at a conference before. Ever. She quickly helped me to remedy this situation. She helped me to gain many opportunities not only presenting at conferences, but at workshops in adjoining states. I cannot thank her enough for helping to push me out of my comfort zone. Additionally, I would like to thank my advisor for helping me to gain experience and confidence teaching and mentoring others. Her quiet patience while I found my footing in assisting to teach Salt Marsh Ecology helped me to grow in my teaching and mentoring abilities.
I would also like to thank Dr. Carol Thornber and Dr. Cathy Wigand for their guidance and support as my committee members. Without your additional support I would not have accomplished as much as I was able to during my time at URI. I am fortunate to have had the opportunity to work with you more and I look forward to continuing to do so throughout the years.
Additionally, I would like to thank my former coworkers at the Environmental Protection Agency: Dr. Cathy Wigand, Dr. Rick McKinney, Dr. Beth Watson, Dr. Autumn Oczkowski, Alana Hanson, as well as many others. Your love of wetland science was infectious and I am forever grateful for your support of my learning while at the EPA. Without it, I would not have thought to pursue a graduate degree. I could vi not have asked to work with a more talented and fun group of wetland scientists. I would also like to specifically thank Dr. Cathy Wigand and Dr. Rick McKinney for being the first to call me a wetland scientist and helping to give me the guidance and confidence to believe it myself.
I owe a large and sincere thank you to all the current and former labmates of the "MV lab"; Rose Martin, Melanie Garate, Liz Brannon, and Ashley Hogan. These past two years would not have been as fun and memorable without you all there along the way. I cannot thank you enough for your support and encouragement over the past two years. I would also like to thank the many people who helped in the field for this research, making this work possible and also making the long lab and field days more enjoyable. A sincere thank you to Ryan Quinn, Katie Lynch, Christina Curran, Claire Martel, Christine Sullivan, Emma Nassaney, and Danielle Perry for their help in lab and field analyses. You helped to make the long days enjoyable and I will miss #marshmadness adventures with all of you.

Introduction
Coastal wetlands (salt marshes, mangroves, and seagrass beds) have been gaining attention in recent years for their ability to play a significant role in climate change mitigation. While already well known for their beneficial abilities to buffer coastlines against storm events (Costanza et al., 2008), filter nutrients from runoff (Kennish, 2001), and provide wildlife habitat, they most recently have been gaining attention for their ability to sequester and store large amounts of carbon (Lu et al., 2017;McLeod et al., 2011). These systems sequester atmospheric carbon dioxide (CO2) through photosynthesis and then store it as carbon, termed blue carbon, in leaves, roots, and sediments. The vast majority of carbon in wetlands is stored within the sediment and is able to be retained for millennia due to the anaerobic conditions in their wet soils (Duarte et al., 2005;McLeod et al., 2011). Anaerobic conditions reduce decomposition rates, enabling wetlands to store more carbon per unit area than any other ecosystem (Chmura et al., 2003;Murray et al., 2011). This ability to remove and store large amounts of CO2, a greenhouse gas (GHG), from the atmosphere has led to increased interest for understanding and quantifying the role coastal wetlands play in climate change mitigation, especially in the face of global climate change.
Although coastal wetlands are extremely important ecosystems, they are currently being lost at rapid rates (McLeod et al., 2011;Pendleton et al., 2012). In the U.S. alone, 50 percent of original salt marsh habitat has been lost (Kennish, 2001). Direct and indirect stressors from human impacts threaten coastal marshes on both regional and global scales. Predominant threats include increasing rates of sea level rise (SLR) caused by climate change, excess nutrient inputs from wastewater and agriculture (Deegan et al., 2012;Diaz and Rosenberg, 2008;Wigand et al., 2014), diminished sediment supplies from dam installations (Weston, 2014), and habitat conversion.
Southern New England is a hotspot for SLR with rates roughly three to four times the global average (Sallenger et al., 2012;Watson et al., 2016a). Studies have found Rhode Island marshes are not maintaining elevation with SLR (Raposa et al., 2016), and as much as 17.3 percent marsh loss has occurred over the past four decades due to associated erosion (Raposa et al., 2015;Watson et al., 2016c). A recent analysis found of Rhode Island's 3,321 acres of salt marsh in Narragansett Bay, 13-87 percent may be lost under projections of three to five feet of future SLR (Narragansett Bay Estuary Program, 2017).

GHG Fluxes from Wetlands
While CO2 is often the GHG that gains the most attention for contributing to global warming due to its relative abundance, methane (CH4) and nitrous oxide (N2O) are additional potent GHGs that also significantly contribute to warming.
Concentrations of all three GHGs have been rapidly increasing in the atmosphere as a result of human activities since the Industrial Revolution (IPCC Climate Change Report, 2014), resulting in myriad consequences that pose significant threats to natural and human-made systems. Although less abundant in the atmosphere, CH4 and N2O are 45 and 270 times, respectively, more efficient at trapping heat in the atmosphere than CO2 (Neubauer and Megonigal, 2015). All three of these GHGs are present within coastal wetlands and wetlands can act as both sinks and sources of GHGs. The production of all three GHGs is sporadic both temporally and spatially from salt marshes, fluctuating with seasons as temperature and plant dynamics shift (Dalal et al., 2003;Le Mer and Roger, 2001;Moseman-Valtierra et al., 2016). GHGs within wetlands are produced through biogeochemical processes involving plant and soil microbial respiration (Picek et al., 2007;Reddy and DeLaune, 2008). Processes that generate GHGs within marshes are continuously cycling, consuming or producing each gas, and whether a wetland is a sink or source at a given time depends on the overall net exchange (i.e., if more gas is being consumed or produced).
In coastal wetlands, CO2 is sequestered from the atmosphere during salt marshes inhibits methane production. A controlling factor on CH4 production is the availability of sulfate. If more sulfate is available, methanogenic bacteria are outcompeted for energy sources by sulfate-reducing bacteria, thus limiting the production of CH4 (Bodelier and Laanbroek, 2004;Poffenbarger et al., 2011). Salinity is a proxy for the amount of sulfate available; thus, with higher salinities less CH4 will be produced, and with lower salinities, more CH4 will be produced.
Coastal wetlands are typically sinks of N2O. N2O is consumed or produced in salt marshes through the soil microbial processes of nitrification and denitrification (Dalal et al., 2003;Liu and Greaver, 2009). Coastal wetlands have the potential to be N cycling hotspots and to act as a filtering buffer, preventing land derived N additions from reaching coastal waters (Bowen et al., 2009). Critical to the process are nitrification and denitrification that filter out excess N (ammonium and nitrate) by converting it into inert N2. N2O is consumed during denitrification and may be produced as a byproduct during both nitrification and denitrification. The regulation of N2O production depends on temperature, oxygen and sulfide availability, as well as plant and microbial community composition (Alvarez-Rogel et al., 2016). Of particular importance on the regulation of N2O is nitrogen (N) availability (Liu and Greaver, 2009).

Impacts of Nitrogen on Coastal Wetlands
N additions from human activities can alter biogeochemical cycling within coastal wetlands, resulting in both positive and negative effects. Salt marshes are naturally N limited (Mendelssohn, 1979;Valiela and Teal, 1974), but human activities have been generating an ever-increasing supply of N through agricultural, industrial, and wastewater practices, altering the amount of reactive N entering into these systems.
This is resulting in a problematic overabundance of the nutrient on a global scale (Diaz and Rosenberg, 2008;Galloway et al., 2003). When nutrient additions alter biogeochemical cycling in coastal wetlands, rates of CO2 uptake can be altered. Initially N inputs can be beneficial, increasing above and belowground biomass (Fox et al., 2012). However, excess nutrients within coastal systems can lead to eutrophication (Diaz, 2001) as well as increased rates of respiration and decomposition (Wigand et al., 2009). Deegan et al. (2012) found high levels of experimentally added N additions on an ecosystem level led to increased levels of decomposition and significant marsh slumping and loss along creekbanks over the course of nine years. When marsh is eroded, the carbon that has been stored within the soils is released, switching the system from a net sink to source of CO2, and simultaneously diminishing any future carbon burial within the system (Murray et al., 2011;Pendleton et al., 2012). N inputs also can switch marshes from being sinks to sources of CH4 and N2O ( However, if denitrification conditions are not ideal, N can fail to undergo the full conversion to inert N2, leading to the termination of denitrification at N2O, resulting in increased emissions (Liu and Greaver, 2009;Moseman-Valtierra et al., 2015). Since N additions can alter marsh structure and function, rates of denitrification may be altered as marsh landscapes shift in response. Wigand et al. (2004) found that rates of potential denitrification enzyme activity (DEA) increased as N loading increased along the N gradient within Narragansett Bay. N additions can shift microbial communities within salt marshes, decreasing densities of bacteria responsible for completing denitrification (Kearns et al., 2015). The filtering out of N by denitrification is critical to water quality.
If this function is lost, water quality degrades, negatively impacting coastal communities.
Despite growing research on the impacts of N loading to coastal systems, there is a lack of information on how chronic nitrogen loading impacts GHG fluxes within salt marshes. Few studies have been conducted to examine the impact of N on all three GHGs (Liu and Greaver, 2009). The majority of the research has focused on impacts from experimentally applied, short-term N additions to marsh systems on smaller landscape scales (Liu and Greaver, 2009 The goal of this research was to examine the role of chronic nitrogen loading on GHG fluxes in Spartina alterniflora-dominated New England salt marshes, and to test relationships of those fluxes to plant properties (height, biomass) and nutrient availability. We examined fluxes of CO2, CH4, and N2O from three marshes along an established N gradient in Narragansett Bay, RI. This work was conducted over the course of six months (June-November) in 2016. We hypothesized that the marsh experiencing higher N loading would have altered biogeochemical cycling and increased aboveground biomass productivity, resulting in increased uptake of CO2, but higher emissions of CH4 and N2O compared to the marshes experiencing lower N loads due to microbial respiration. Overall between the sites we expected biomass and nutrients (ammonium) concentrations to be negatively related to CO2 uptake and positively related to CH4 and N2O emissions.
To provide context for fluxes observed during the 2016 field season, we also examined potential DEA rates by conducting DEA assays in the late-spring and midsummer of 2017. The aim of this work was to use the DEA assays as a means of indirectly measuring N loading to the sites to support observed N2O fluxes. We additionally examined DEA within four different marsh zones. Since there is heterogeneity within marsh landscapes, examining multiple zones allowed for determining any differences within and among sites for potential denitrification and N2O fluxes. We hypothesized that higher N loading would result in higher rates of potential DEA and that each zone would experience different potential DEA rates as a result of varied biogeochemical cycling between zones. While not discussed in detail here, this research will contribute to a larger collaborative project called Bringing Wetlands to Market aimed at examining the feasibility of creating a carbon market to provide an economic incentive to protect and restore wetlands (Emmer et al., 2014).
Understanding the dynamics between nitrogen, blue carbon, DEA rates, and all three GHGs will help strengthen efforts aimed at protecting and restoring these valuable coastal systems.

Study Locations
Narragansett Bay is one of the most densely populated estuaries in the USA. As a result, some areas of the bay have historically received high N loads since the late 1800s, mainly from wastewater originating from point sources. The largest point source within the bay has been wastewater treatment facilities (WWTFs) (Nixon and Pilson, 1983;Oczkowski et al., 2008). Beginning in 2005, policy decisions and upgrades to WWTFs in the area have lessened the overall amount of N entering into the bay and as of 2012 it has been reduced by 50% from previous levels (Krumholz, 2012). While the reduction of N loading into the bay has significantly reduced N loads within bay water, N loading at sites directly adjacent to wastewater inputs from septic systems or agricultural inputs likely has not changed.
We chose three field sites along an established and historical N gradient within Narragansett Bay, RI (Figure 1), representative of high (Mary's Creek, 4.5 ha), medium (Mary Donovan, 24.1 ha), and low (Nag Marsh, 28.8 ha) N loads with an overall 200fold difference (Table 1) (Watson et al., 2016c;Wigand et al., 2003). The N loads entering into Mary's Creek are dominated by wastewater inputs from surrounding residential development while the majority of N inputs at Mary Donovan are agriculturally derived from surrounding farms. Nag Marsh, which is part of the Narragansett Bay Estuarine Research Reserve, lacks significant development adjacent to the field site and is located on an island within the middle of the bay, overall receiving limited anthropogenic N inputs (Wigand et al., 2003). Each site is predominantly composed of the common marsh grass, Spartina alterniflora, which is the most inundation tolerant species of marsh vegetation. At each of the sites, and the majority of marshes in New England, S. alterniflora has largely displaced high marsh vegetation as SLR has increased the extent and duration of inundation in the area (Smith, 2009;Watson et al., 2016b).

Experimental Design
To examine the influence of chronic N loading on GHG fluxes, six plots were established in May 2016 equidistant from each other within the low marsh zone at each of the three field sites. Plots were established parallel to the dominant creek at each site.
At Mary Donovan and Nag Marsh plots were spaced roughly 15 m apart, and at Mary's Creek plots were spaced roughly 5 m apart due to the smaller scale of the site (Table 1) (Watson et al., 2016c). An aluminum flashing collar (10 cm tall x 28 cm diameter) was installed in each plot, inserted to a depth of 6 cm. Each collar had six evenly spaced holes (3 cm tall x 0.5 cm diameter) directly below the sediment surface for drainage after rainfall or inundation. Collars were installed at least two weeks prior to any Soil temperature, salinity, redox, pH, soil moisture, and photosynthetic active radiation (PAR) were also measured at each plot (  (Table 2).

GHG Flux Measurements
Greenhouse gas fluxes were measured in situ using two different real-time analyzers with closed, transparent chambers. CO2 and CH4 were measured with a cavity ring-down spectroscopy (CRDS) analyzer (Model G2508; Picarro Inc., Santa Clara, California, USA). N2O fluxes were measured using an off-axis integrated cavity output spectroscopy (OA-ICOS) analyzer (Model N2O/CO, Los Gatos Research (LGR), Mountain View, California, USA). While the CRDS analyzer is capable of measuring all three GHGs, the OA-ICOS analyzer has a lower detection limit and is better able to detect small concentrations of N2O .
Measurements were taken biweekly, or monthly, alternating between the two analyzers and sampling followed previously outlined methods ( Over the course of the study, Nag Marsh was sampled less frequently due to logistical constraints. As a result of the previously described plant stress some flux measurements were omitted. Additionally, two flux measurement dates from Mary's Creek were omitted in two months where it was sampled twice with the Picarro G2508 to allow for more equal comparisons of one flux measurement day per month for all three sites. This was done as those two dates fell at different times of the month than the dates for the other two sites. Due to poor weather hampering flux measurements, Mary's Creek measured on August 3 rd , 2016 was counted as a July measurement day.

Plant and Environmental Properties
Plant heights (10 random stems) and stem density were measured within each collar monthly. Above and belowground plant biomass samples were collected at the beginning of September 2016. To determine aboveground biomass at each site, a 25 cm x 25 cm quadrat was placed within one meter of each replicate flux measurement plot.
All aboveground biomass within the plot was clipped at the shoot base, stored in a plastic bag on ice, and refrigerated until processing (within five days). The biomass was rinsed in a two mm sieve, dried at 70 °C for 48 hours, and then weighed. To determine belowground biomass, soil cores were taken from within the clipped quadrat plot with a peat corer (50 cm tall x 5 cm diameter). These cores were cut in half. One half was used for belowground biomass and the other for bulk density analysis. As we were only interested in the active root zone (Valiela et al., 1976) for belowground biomass due to its potential impact on gas transport (Picek et al., 2007), this half for biomass was sectioned to a 20 cm depth. The half for bulk density was sectioned to 4 cm, as this is the depth most likely influencing fluxes since microbial activity and diffusion of gases decrease with depth (Ball et al., 1997;Smith et al., 2003). The portion for belowground biomass was rinsed over a 2 mm sieve, separating out macro organic matter from soil.
Macro organic matter was then separated out into roots and rhizomes with rhizomes distinguished as greater than 2 mm in diameter. Any macro organic biomass was saved, placed in paper bags, dried for 48 hours at 60 °C, and weighed. To calculate bulk density the soil samples were dried at 60 °C for 48 hours, weighed, and then the weight was divided by the known soil volume.
Porewater samples for nutrient analysis were taken outside of each gas flux measurement plot, within 25 cm, and to a depth of 10 cm during flux sampling events in June and September. Rhizon samplers (Soil Moisture Co., Goleta, CA) connected to 60 mL nylon syringes were used to collect 10 mL of porewater. Each sample was filtered, placed into a plastic vial, and frozen at -20 °C until analysis. Each sample was run for ammonium (NH4 + ) on an Orion TM Aquamate 7000 spectrometer following Solorzano (1969).

Experimental Design
The difference in potential DEA for Mary's Creek and Nag Marsh was examined in 2017. Small soil cores (5.1 cm diameter x 10 cm depth) for DEA analysis were collected at each site to capture the opposite ends of the N gradient. In each marsh, cores were collected from the following four zones: bare creekbank, die-back area converted to mudflat, low marsh (S. alterniflora-dominant), and high marsh (Spartina patensdominant). Each zone had four replicate plots, spaced one meter apart. Samples from the four zones at the two sites were collected at low tide in late spring and mid-summer.
To compare GHG fluxes across the marsh landscape, measurements were taken approximately monthly (June-September) 2017 at Mary's Creek in each of the four zones where DEA rates were studied. Four collars were installed at least two weeks prior to GHG flux measurements, spaced one meter apart within homogenous sections of these zones and oriented parallel to the shoreline. These measurements were taken with the Picarro G2508 as described above. Flux sampling was conducted in the low marsh and creekbank zones monthly, while the mudflat and high marsh zones were sampled twice over the course of the study. The total number of fluxes taken per zone were as follows; creekbank: 13, mudflat: 8, low marsh: 15, high marsh: 8. Due to logistical constraints with the analyzer exceeding maximum operating temperature in September, only one flux from the creekbank zone was able to be measured.
Additionally, only three fluxes were able to be collected from the low marsh zone in August due to retreating daylight.

Denitrification Enzyme Activity (DEA)
All cores were processed within three days of collection and refrigerated until processing. DEA was measured using short-term anaerobic assays following previously described methods (Gardner, 2008;Groffman and Tiedje, 1989;Smith and Tiedje, 1979). The top 3 cm of each core was used for DEA analysis, split into two sections from which a homogenized 6 g soil sample was taken from each half, generating two replicates per core. Each soil sample was placed into a clear, glass 125 mL flask. Ideal conditions for DEA were then created for each soil sample following the procedure detailed in Wigand et al. (2004), adding 1.4 mg KNO3, 1 mg glucose, and 0.25 mg chloramphenicol per g of sediment wet weight. These were then combined with ultrapure water, generating a solution in which 12 mL was added to each 6 g of sediment. Flasks were then flushed with ultrapure helium (Airgas, Radnor, Pennsylvania) for two minutes to begin the nonlimiting conditions for DEA. The soil slurries were then allowed to incubate under anaerobic conditions with gas samples collected at 0 and 90 minutes. Gas samples were analyzed for N2O by electron gas chromatography, using a Shimadzu 2014 Greenhouse Gas Analyzer.

GHG Flux Comparisons Across Zones
To complement the DEA assays, GHG flux measurements were taken approximately monthly using the Picarro G2508 for four months in the summer of 2017.
All fluxes were taken at Mary's Creek as this is the site where we expected the highest DEA rates. Sampling was conducted in the low marsh and creekbank zones monthly, while the mudflat and high marsh zones were sampled twice over the course of the study.

Statistical Analysis
GHG fluxes were calculated for each measurement period based on the chamber volume and linear changes in gas concentration over time using the Ideal Gas Law (PV=nRT). All fluxes were calculated using R (R Core Team, 2016). Only fluxes that were above detection limits   greater or equal to 0.9 and the p-value less than or equal to 0.05. N2O fluxes were determined to be statistically significant if the R 2 was greater than or equal to 0.7 (due to wide variability in N2O) and the p-value less than or equal to 0.05.
Linear mixed effects analyses were used to test for differences in GHGs as well as plant properties (height, density) and edaphic parameters ( One-factor ANOVAs were used to compare above and belowground plant biomass, as well as soil bulk density among sites. A two-factor ANOVA was used to examine the relationship of site and month on porewater ammonium concentrations. A Pearson's Correlation Analysis was used to test for a relationship between CO2 and plant height, and Spearman's Correlation Analyses were performed to examine the relationship between GHGs to plant properties and site parameters. DEA sample replicates were averaged per core and then examined using a three-factor ANOVA and a Tukey's HSD post-hoc test to explore differences among sites, seasons, and zones. Significance was assigned at α=0.05 for all statistical tests. Linear mixed effects analyses were performed in R, using the lme4 package (R Core Team, 2016). All other statistical analyses were run in JMP 13.0 (JMP 13.0).
Overall average N2O fluxes from the three sites were minimal (-0.07 to 0.62 µmol m -2 h -1 ) (Figure 2c). The largest individual N2O flux (1.23 µmol m -2 h -1 ) observed was measured from Mary's Creek in July and the lowest individual N2O flux (-0.10 µmol m -2 h -1 ; Mary Donovan in August) exhibited N2O uptake. Of the 60 fluxes taken for N2O over the course of the study, 46.7% were found to be significant. The remaining fluxes were either below the detection limit (0.02 µmol m -2 h -1 ) of the LGR N2O/CO analyzer (21.7%) or were above the detection limit, but were not significant and thus labeled a zero flux (31.7%). There was no significant correlation between redox and N2O (Spearman's r= -0.31, p=0.10).

Plant and Environmental Properties
Bulk density and salinity did not significantly differ among the sites (Tables 1,   2, 3). Redox significantly differed among the sites, with Mary Donovan having significantly higher average redox (Χ 2 2=78.71, p<0.0001; Table 3).
S. alterniflora stem height was significantly lower at Nag Marsh compared to the other two sites (Χ 2 2=7.81, p=0.02; Figure 3a). Stem density also significantly differed between the sites (Χ 2 2=39.26, p<0.0001), with Mary Donovan having significantly lower average density than the other two sites (Figure 3b). All correlation analyses between stem height and stem density for each of the three GHGs were not significant (p>0.05 for all). The only exception was a significant negative correlation between stem density and CO2 (Spearman's R 2 = -0.44, p<0.01). This correlation was found to be negative since all CO2 fluxes were negative. Aboveground biomass was significantly higher at Mary's Creek than at Mary Donovan (F2,15=4.64, p=0.03; Figure   4).
Belowground biomass did not significantly differ among sites, although there was a trend of lowest average biomass at Mary Donovan (F2,15=3.59, p=0.05). Roots and rhizomes also did not significantly differ among sites ( Nag Marsh had significantly higher average ammonium concentrations than the other marshes (F2,30=10.57, p<0.01) ( Figure 5). Ammonium porewater concentrations were higher in September than June (F1,30=5.19, p=0.03).

Denitrification Enzyme Activity (DEA)
Rates of potential DEA were not found to be significantly different among sites, zones, or dates, but there were significant interactions among these factors. There were significant interactions among site x month x zone (F3,48=5.98, p<0.01) as well as month x zone (F3,48=4.02, p=0.01) and site x zone (F3,48=5.46, p<0.01). The highest average rate of potential DEA (584.1 ng N g -1 h -1 ) was measured for the low marsh zone in June at Mary's Creek (Table 6), followed closely by the mudflat zone at Nag Marsh (423.1 ng N g -1 h -1 ; Table 6, Figure 6).

GHG Flux Comparisons Across Zones
At Mary's Creek, CO2 was significantly different among zones (Χ 2 2=32.02, p<0.0001), with the low marsh and mudflat zones exhibiting CO2 uptake while the creekbank zone exhibited on average emissions (Tables 3, 7). Only three of the zones (mudflat, creekbank, low marsh) had significant fluxes throughout the study and thus were the only zones included in the analysis (high marsh excluded). While the mudflat zone was included, it only had significant fluxes on one of the two dates measured.
The creekbank had significantly higher CH4 emissions than the low marsh zone, but not the high marsh and mudflat zones (Χ 2 2=9.05, p=0.03; Table 7). CH4 emissions were similar among the remaining three zones. There were no significant N2O fluxes (out of 48 total fluxes measured) to report for the 2017 field season from any of the four zones.

Discussion
Our findings support the hypothesis that high N loading would result in increased uptake of CO2 as well as increased emissions of CH4 and N2O. This pattern was observed for the marsh receiving the highest historical N load, Mary's Creek, relative to the other two sites with lower historical rates of N loading. However, several key parameters (plant biomass, porewater NH4 + ) did not fall along the expected trends of the N gradient, suggesting that N loading is likely not the only influencing factor driving GHG and DEA differences among these three sites.

CO2 Fluxes Among Marshes
Overall the magnitude of the CO2 uptake across sites was relatively large with Mary's Creek having the highest uptake among the three sites (Figure 2a). This site also had the greatest aboveground plant biomass among the sites (Table 5, Figure 4). In In our study, stronger relationships of CO2 uptake to aboveground biomass may be due to similarities in belowground biomass among sites at the depths that we sampled (Table   5). Further, a relationship between CO2 uptake and belowground biomass may have However, excessive N can also lead to lowered stem density (Valiela, 2015) and increased decomposition rates, resulting in reduced belowground biomass (Langley et al., 2009;Wigand et al., 2014). Deegan et al. (2012) found marked differences in biomass production with N additions, resulting in increased stem height, but decreased belowground biomass. In that study excess fertilization with nitrate added for a decade resulted in visible marsh deterioration with creekbank slumping, leading to marsh loss.
In our study, several of the plant parameters (above and belowground biomass, stem density) at Mary's Creek and Nag Marsh were surprisingly similar despite being at opposite ends of the historic N gradient, while those at Mary Donovan were significantly lower (Tables 3, 5, Figures 3, 4). It is especially interesting we did not find greater differences among our sites when visually the sites overall look vastly different in areas, with Mary's Creek having visible slumping and erosion of the back portion of the marsh.
The marsh has been documented as losing roughly 20% of its areal extent in the past 40 years (Watson et al., 2016c). In comparison, Mary Donovan and Nag Marsh have been documented as losing 15% and 2% areal extent (Watson et al., 2016c), respectively, but do not exhibit the same large areas of unstable platform. Although these losses have been attributed to increased rates of SLR, the obvious erosion at Mary's Creek may be a combination of increased N and inundation. Measuring fluxes or biomass even a short distance from where the platform is stable at Mary's Creek likely would have resulted in substantial differences with increased respiration and decreased biomass.
Flux measurements in 2017 were aimed at examining differences in CO2 fluxes among zones at Mary's Creek. Among the three remaining zones with significant fluxes (creekbank, mudflat, low marsh), there were obvious and expected differences with the relatively bare creekbanks consistently exhibiting emissions while the vegetated low marsh zone exhibited CO2 uptake. Overall, creekbank emissions were two to three times greater than similarly bare ponded area emissions measured in Moseman-Valtierra et al. (2016).

CH4 Fluxes Among Marshes
As expected for our study, CH4 emissions were highest at Mary's Creek.
However, given that salinities did not significantly differ among sites and a linear regression showed only a trend between CH4 and salinity, it is likely the higher N load was driving the higher emissions at Mary's Creek. N additions can increase CH4 emissions (Bodelier and Laanbroek, 2004;Irvine et al., 2012) through the production of organic matter (Liu and Greaver, 2009), which is a necessary component of methanogenesis (Le Mer and Roger, 2001). Together, organic matter, sulfate availability, and anaerobic conditions are the greatest controlling factors of CH4 production in wetlands (Poffenbarger et al., 2011). Here, a negative correlation was found with redox and CH4, with lower redox values corresponding to higher CH4 fluxes.
This finding is not surprising given that methanogenesis requires anaerobic conditions (Le Mer and Roger, 2001;Poffenbarger et al., 2011). However, since Mary's Creek and Nag Marsh had similar redox values (Table 2) it is likely the higher N influenced the higher emissions at Mary's Creek. In comparison to Moseman-Valtierra et al. (2016) our CH4 fluxes were an order of magnitude greater than the values they measured for the low marsh zone.
For GHG flux measures in 2017, the creekbank had the highest CH4 emissions of all four zones, while the lowest emissions were from the high marsh zone. This is likely due to more aerobic conditions in the high marsh zone, which can allow for the oxidation of methane and lower CH4 emissions than more anaerobic zones (Le Mer and Roger, 2001;Poffenbarger et al., 2011).

N2O Fluxes Among Marshes
We found a trend that Mary's Creek had the highest mean N2O emissions, by an order of magnitude over the other two sites (Tables 3, 4). This finding agrees with other studies that have found N additions can switch salt marsh systems and other ecosystems from being sinks to sources of N2O (Chmura et al., 2016;Liu and Greaver, 2009;Moseman-Valtierra et al., 2011). However, our lowest N site, Nag Marsh, did not exhibit the lowest N2O emissions, and overall the magnitude of our N2O emissions were low in comparison to other studies (Murray et al., 2015). It is surprising that we observed relatively low N2O emissions from our sites within Narragansett Bay considering the estuary has received chronic N loading since the late 1800s . Most studies documenting increased N2O emissions from experimentally fertilized plots have typically only added fertilizer for less than a decade. Moseman-Valtierra et al. (2011) found mean levels of N2O emissions from short-term, nitrate-fertilized plots to be 1.8 µmol m -2 h -1 , even reporting a flux as large as 6.8 µmol m -2 h -1 . In contrast, our mean flux (1.23 µmol m -2 h -1 ) measured was during July, which corresponds to the highest mean soil temperature (Table 2). It was the only day of measurements where all six fluxes measured were significant, averaging 0.62 µmol m -2 h -1 . It is likely that these higher fluxes were due to increased microbial activity, which is positively correlated with temperature (Reddy and DeLaune, 2008).
Zonation is an important factor to consider when examining N2O fluxes. In our study, our measurements were concentrated in the S. alterniflora-dominated low marsh zone. For the 2017 field season we then expanded to other marsh zones and tested for differences in N2O fluxes among them. However, there were no significant fluxes measured within any of the zones (creekbank, mudflat, low marsh, high marsh) as they were all below the detection limit (2.7 µmol m -2 h -1 ) of the Picarro G2508 (Table 7).
Both studies mentioned above (Chmura et al., 2016;Moseman-Valtierra et al., 2011) were conducted in Spartina patens-dominated marshes. Low marsh zones are flushed more regularly with daily tides and are more anaerobic than high marsh zones, thus perhaps more likely to retain less N than upland marsh portions (Childers et al., 2000;Wigand et al., 2004) as it either exports out after inundation or is transformed into inert N2 gas through complete denitrification (Hamersley and Howes, 2003). The relatively small N2O fluxes in 2016 and lack of significant N2O fluxes during the 2017 season, may be due to both export and complete denitrification. This is further supported by findings of similar rates of DEA among zones in 2017 discussed below.

Ammonium (NH4 + ) concentrations
We found surprisingly overall high NH4 + concentrations, particularly within Nag Marsh with a collective mean that surpassed Mary's Creek by an order of magnitude (139.7 µM and 85.2 µM, respectively; Figure 5). Although higher levels of NH4 + can be linked to fertilization (Valiela et al., 1978), these high levels may be indicative of other stressors. RI marshes have been documented as not keeping pace with SLR in the region (Raposa et al., 2016) and as a result they are experiencing vegetation shifts from S. patens to the more inundation tolerant S. alterniflora as well as increased erosion (Raposa et al., 2015;Watson et al., 2016a,b). It is possible these NH4 + levels reflect increased decomposition of organic matter as a result of decomposing peat due to increased inundation. While Nag Marsh has the highest median elevation of the three sites (Table 1), and thus may experience less inundation in comparison, these NH4 + values may still be linked to decomposition since rates of SLR in the area are three to four times the global average (Sallenger et al., 2012) and Nag Marsh as well as Mary's Creek have been documented as experiencing significant erosion within the past 40 years (Watson et al., 2016c). Although NH4 + was found to be similar between these two sites, nitrate may not follow the same pattern and could further help constrain overall N loading.
Denitrification enzyme activity (DEA) Wigand et al. (2004) found rates of DEA increased with watershed N loading in Narragansett Bay salt marsh soils; this continues to be true, even after the decrease in N loading into the bay from WWTFs upgrades (Fulweiler and Heiss, 2014). However, rates of DEA may shift as time goes on, as evidenced by Fulweiler and Heiss's (2014) examintation of subsurface sediments in Narragansett Bay. Within our study, we did not observe significant differences in DEA between Mary's Creek and Nag Marsh (Table   6, Figure 6), which may suggest that they now both have equal amounts of N available for transformation if the N gradient is no longer as strong as it was historically.
Considering that the N loads were calculated for the bay from surrounding land use in Wigand et al. (2003) it is possible N loading to the sites has shifted since then. A recent study has shown the recent upgrades to the WWTFs have decreased nutrient standing stocks by 60% within the bay (Oviatt et al., 2017). However, sites such as Mary's Creek which are directly adjacent to septic systems or those with direct agricultural run-off are not as likely to be impacted by the bay-wide reduction in N loading. N loading to soils can have a legacy effect, remaining in the system decades even after reductions have been made (Grimvall et al., 2000). Future work is needed to constrain the current N loading in salt marshes within the bay.
Similarity in DEA rates between the bare creekbank and mudflat zones as well as the vegetated zones of the marshes was surprising since denitrification has been linked to organic matter availability as a controlling factor and plants are a major source (Eyre et al., 2013). But given the organic rich soils in New England marshes, there may be ample substrates among even unvegetated zones to support heterotrophic denitrifies.
It is also surprising to not have found differences in DEA rates between the high and low marsh zones, as previous studies have found significant, but variable, differences in DEA between these zones (Koop-Jakobsen and Giblin, 2009;Wigand et al., 2004).
Koop-Jakobsen and Giblin (2009) found higher rates of DEA within the high marsh zone, while Wigand et al. (2004) found the opposite trend. While similarly high rates of DEA among the zones highlights the large potential ability for N transformation across the marsh landscape, areas that do not have vegetation, such as creekbanks and mudflats, are more susceptible to erosion. With less marsh area, there is less available habitat to fully transform N and this could potentially lead to increased emissions of N2O and decreased water quality as land-derived nutrients export out into coastal waters.

Comparison of GHGs
In order to equally assess and compare all three GHGs measured over both field seasons, we converted mean values of both and CH4 and N2O emissions into CO2 equivalents using their respective global warming potentials, 45 and 270, respectively (Neubauer and Megonigal, 2015). Comparing these values to the mean CO2 uptake over the course of both growing seasons we found that only two percent of it was offset by CH4 and N2O emissions and that this small offset was driven in larger part by CH4 emissions.

Conclusion
Our study highlights the robust functional role of salt marshes in mitigating harmful GHGs and excess nutrients from anthropogenic sources. While we did observe increased N2O and CH4 emissions at our high N site, they were not on an order of magnitude to significantly offset CO2 uptake. Comparing all three GHGs by converting them into CO2 equivalents revealed the amount of CO2 uptake over the course of the study was only offset by two percent by N2O and CH4 emissions. This work highlights the large capacity of S. alterniflora-dominated marshes to filter out excess N while releasing relatively small amounts of more potent GHGs. Additionally, similar rates of DEA among zones and across marshes at the opposite ends of the N gradient highlight the large capacity for these ecosystems to filter out excess N, despite differences in plant communities or chronic N loading and deterioration.
Since the problems of excess N entering into coastal ecosystems and increasing GHGs overall in the atmosphere will only increase in the upcoming years, it is important to quantify the beneficial services coastal marshes provide in order to foster incentives for protecting these valuable systems. This will be particularly important for      Mary's Creek Nag Figure 6. Mean potential denitrification enzyme activity (DEA) rates (± SE) from the high and low N sites across four zones in the 2017. Potential DEA in the months of (a) June and (b) August.