Long‐term nutrient addition increases respiration and nitrous oxide emissions in a New England salt marsh

Abstract Salt marshes may act either as greenhouse gas (GHG) sources or sinks depending on hydrological conditions, vegetation communities, and nutrient availability. In recent decades, eutrophication has emerged as a major driver of change in salt marsh ecosystems. An ongoing fertilization experiment at the Great Sippewissett Marsh (Cape Cod, USA) allows for observation of the results of over four decades of nutrient addition. Here, nutrient enrichment stimulated changes to vegetation communities that, over time, have resulted in increased elevation of the marsh platform. In this study, we measured fluxes of carbon dioxide (CO 2), methane (CH 4) and nitrous oxide (N2O) in dominant vegetation zones along elevation gradients of chronically fertilized (1,572 kg N ha−1 year−1) and unfertilized (12 kg N ha−1 year−1) experimental plots at Great Sippewissett Marsh. Flux measurements were performed using darkened chambers to focus on community respiration and excluded photosynthetic CO 2 uptake. We hypothesized that N‐replete conditions in fertilized plots would result in larger N2O emissions relative to control plots and that higher elevations caused by nutrient enrichment would support increased CO 2 and N2O and decreased CH 4 emissions due to the potential for more oxygen diffusion into sediment. Patterns of GHG emission supported our hypotheses. Fertilized plots were substantially larger sources of N2O and had higher community respiration rates relative to control plots, due to large emissions of these GHGs at higher elevations. While CH 4 emissions displayed a negative relationship with elevation, they were generally small across elevation gradients and nutrient enrichment treatments. Our results demonstrate that at decadal scales, vegetation community shifts and associated elevation changes driven by chronic eutrophication affect GHG emission from salt marshes. Results demonstrate the necessity of long‐term fertilization experiments to understand impacts of eutrophication on ecosystem function and have implications for how chronic eutrophication may impact the role that salt marshes play in sequestering C and N.

Nutrient enrichment alters salt marsh plant community composition (Rogers, Harris, & Valiela, 1998). The elevation constraints of most salt marsh plant species, and the tight interaction of species with sediment conditions, including redox and relative elevation, generate a series of interactions. Plant species present on marsh swards occur in well-constrained bands determined to a large degree by submergence regime (Bertness, 1991), but marsh plants release O 2 and oxidizing compounds to sediment (Lovell, 2005). Plant taxa in marshes hence both depend upon and affect redox within sediments, and effects vary with relative elevation in the tidal frame.
Studies have demonstrated the effects of over four decades of fertilization (Valiela, 2015) on plant communities and elevation at Great Sippewissett Marsh in Cape Cod, MA, USA (Fox, Valiela, & Kinney, 2012;Rogers et al., 1998;Valiela, 2015). Large decadalscale changes within vegetated plots subject to high nutrient enrichment resulted in increased S. alterniflora height and biomass, as well as changes in species cover, with Distichlis spicata becoming more dominant (Valiela, 2015). Changes in elevation and subsequent feedbacks on vegetation communities likely affect associated rhizosphere microbial processes and soil biogeochemistry, and C and N cycles that drive GHG fluxes.
Nitrous oxide production in salt marsh sediment may result from microbial processes including denitrification and nitrification (Kool et al., 2010;Koop-Jakobsen & Giblin, 2010;Wrage, Velthof, van Beusichem, & Oenema, 2001). Nitrous oxide production is dependent on oxygen availability in sediments: under anaerobic conditions, nitrate undergoes complete denitrification to N 2 gas. In salt marshes not subjected to eutrophication, N is limiting (Valiela & Teal, 1974) and so N 2 O fluxes generally are small. N 2 O emissions were found to be negligible at high and low marsh elevations of several southern New England salt (Martin & Moseman-Valtierra, 2015;Moseman-Valtierra et al., 2011) and brackish (Martin & Moseman-Valtierra, 2015, 2017a marshes receiving low N loads. Nutrient enrichment drives changes in salt marsh structure and function via a number of mechanisms and may ultimately increase GHG emissions. Nutrient enrichment accelerates salt marsh litter decomposition and CO 2 emission (Deegan et al., 2012;Wigand, Brennan, Stolt, Holt, & Ryba, 2009). Where changes in vegetation communities have driven elevation increases, less saturated sediments may support increased aerobic metabolism and greater CO 2 emission due to respiration. Coupling of oxic and anoxic conditions where soil is less saturated may support biogeochemical pathways that, in combination with N-replete conditions, result in N 2 O emission. Therefore, elevation gains may facilitate N 2 O production by incomplete denitrification (Koop-Jakobsen & Giblin, 2010) and nitrifier denitrification (Kool et al., 2010;Wrage et al., 2001) and are likely also suboptimal for N 2 O uptake resulting from denitrification (Chapuis-Lardy et al., 2007). Chronic enrichment, such as that characteristic of coastal waters worldwide (Rabalais et al., 2009), could produce a sustained N 2 O emission response (Murray, Erler, & Eyre, 2015), potentially increasing radiative forcing of eutrophic salt marshes as N 2 O has 263 times the global warming potential of CO 2 (Neubauer & Megonigal, 2015).
There is a need for long-term experiments to test effects of chronic nutrient enrichment on salt marsh GHG fluxes. In this experiment, we tested effects of ~45 years of nutrient enrichment (using an organic NPK fertilizer containing multiple N species) at Great Sippewissett Marsh on CO 2, CH 4 and N 2 O fluxes measured with opaque chambers in situ. We related GHG fluxes in experimental plots to vegetation communities and elevation. As N retention in the Great Sippewissett plots is high and exports are minor (<7%) (Brin, Valiela, Goehringer, & Howes, 2010), this experiment allowed relative certainty of N loading rates to our experimental plots. We hypothesized that (1) N enrichment will stimulate N 2 O emission, especially at higher elevations where soil saturation is decreased, coupling oxic, and anoxic conditions that drive higher ratios of N 2 O/N 2 emissions from denitrification; (2) CO 2 emissions will be greater due to increased community respiration in the nutrient enriched treatment and at higher elevations where decreased sediment saturation and increased O 2 availability support more rapid decomposition, and (3) CH 4 emissions will vary with elevation, with smallest emissions at higher elevations where sediment O 2 availability is greatest.

| Site description and experimental design
This study was conducted at plots experimentally nutrient-enriched for over 45 years (and continuing) in Great Sippewissett marsh, Cape Cod (Valiela, 2015) (Figure 1). Organic fertilizer is hand-broadcast every 2 weeks, March-October, each year, with treatments that include an unfertilized control and three doses of nitrogen (N) using Milorganite, a commercially available mixed NPK fertilizer (10% N, 6% P, 4% K by weight) containing ammonium nitrate, potassium nitrate, ammonium sulfate, and urea, among other constituents (www.milorganite.com).
For this study, we focused on GHG fluxes from two replicate control (C) plots (no fertilization, 12 kg N ha −1 year −1 from external sources) and two replicate plots that received high doses of N (XF) (1,572 kg N ha −1 year −1 ). We measured GHG fluxes monthly from July-October 2016 at three elevation zones (Creekbank, Mid and High) in XF plots and 2 (Creekbank and Low) in C plots (Table 1), with six or eight replicate plots per zone. Due to distinct topography responses to nutrient addition, high and medium elevation zones have no counterpart in C plots, and the low elevation zone has no counterpart in XF plots. Elevation and vegetation in XF plots display striking responses to four decades of fertilization, as previous studies have described in detail (Fox et al., 2012;Rogers et al., 1998). In XF plots, relative elevations ranged from 0 to −69 cm relative to MHW. Tall S. alterniflora dominates along the creekbank, D. spicata, and intermediate ht. S. alterniflora dominate at Mid elevations, and I. frutescens dominates high elevations (Table 1). In C plots, the elevation range is much narrower (from −36 to −70 cm relative to MHW). Short S. alterniflora dominates at low elevations and tall S. alterniflora dominates creek banks (Table 1).

| Relative elevation and vegetation measurements
We measured relative elevation within vegetation zones using wooden laths (1 m height; n = 6 or 8 per vegetation zone, within 0.5 m of flux measurement locations) which were painted with a mix of food dye and water-soluble glue and deployed just before high tide (Smith & Warren, 2012). We recorded heights of watermarks to determine elevation of vegetation zones relative to mean high water (Table 1). We estimated percent cover of each elevation zone within each of the XF and C plots using established vegetation cover survey methods for the dominant species in each defined elevation zone (Braun-Blanquet, 1932).

| Greenhouse gas flux measurement
To define the effect of elevation, vegetation, and nutrient enrichment on GHG fluxes, we designed 0.3 m diameter × 0.6 m tall cylindrical static flux chambers to encompass vegetation without damage. Chambers sit snugly in PVC collars installed about 6 cm F I G U R E 1 Experimental plot locations in Great Sippewissett Marsh. This study focused on XF (fertilized; 1,572 kg N ha −1 year −1 ) and C (Control; 12 kg N ha −1 year −1 ) plots (labeled to match previous publications on the experimental plots) TA B L E 1 Average relative elevations, percent cover, and dominant plant species of elevation classes for C and XF treatments into the marsh surface, and we achieved a gas-tight seal by adding a small amount of water to the gap between collar and chamber. To prevent water pooling or shading effects within collars, we installed collars approximately 1 hr prior to gas flux measurements. We repeated measurements at the same points, marked with small flags, during each visit. We performed flux measurements at n = 6 or eight locations for Creekbank, low and high elevation classes. For the Mid elevation classes, where the dominant species (D. spicata and S. . alterniflora) formed patches, we increased replication to n = 12-16.
We used a cavity ring down spectroscopy (CRDS) in situ analyzer (Picarro G2508) to measure CO 2 , CH 4 , and N 2 O concentrations in real time (Martin & Moseman-Valtierra, 2015). We covered chambers with white, light-blocking fabric to exclude photosynthetic CO 2 uptake and measured community respiration and GHG gas fluxes. We

| Statistical analyses
We used linear mixed effects models for analysis of the effect of vegetation zone on CH 4 , N 2 O, and CO 2 fluxes from each treatment type (XF and C plots). We treated vegetation type as a fixed effect and sampling month and plot as random effects. To test for effects of treatment (XF or C) for Tall S. alterniflora (present in plots of both treatments), we used linear mixed effects models with treatment as the fixed effect and sampling month and plot as random effects.
To obtain p-values to assess significance of the effect of vegetation zone and treatment on GHG fluxes, we performed likelihood ratio tests of full models against models with the fixed vegetation zone or treatment effect removed. We performed all statistics in R (Team, 2014) and interpreted significance at α = 0.05.

| N 2 O fluxes
Detectable N 2 O emissions were restricted to XF plots at medium and high elevation zones (relative elevation to MHW > −25 cm) (  Figure 2). This finding is consistent with the hypothesis that less saturated sediment and N-replete conditions may favor microbial transformations that produce N 2 O. In C plots, N 2 O emissions were minimal, with the majority of fluxes below detection of the Picarro G2508 analyzer (Table 2, Figure 3a). N 2 O fluxes in the XF plots differed significantly between vegetation zones (Table 3), with large emissions in the Mid and High zones, and negligible emissions in Creekbank zone. Emissions in the Creekbank zone were below detection in both XF and C plots.

| CH 4 Fluxes
Methane emissions varied with elevation, with negligible emissions at high elevations (Table 2). While CH 4 emissions differed significantly between elevation zones in XF plots, short and Creekbank S. alterniflora zones in C plots had similar emissions (Table 3). Creekbank CH 4 emissions did not differ between XF and C plots, demonstrating that nutrient enrichment did not affect CH 4 emissions in this zone. Since elevations are lower in the C plots, generally more CH 4 was produced in C than XF plots (Table 2). One very large flux (1,351.8 μmol m −2 hr −1 ), seemingly due to ebullition based on the "stepwise" pattern of increase evident from plots of CH 4 concentration over time (Middelburg et al., 1996), was measured in the short S. alterniflora zone in a C plot in August, signifying that short-term measurements may not capture ebullition events and therefore may underestimate total CH 4 emissions.

| CO 2 Fluxes
Carbon dioxide emission (respiration) was greatest in XF plots at the mid and high elevations (Tables 2, 3). Results support the hypothesized greater respiration rates at higher elevations, although Creekbank S. alterniflora emissions did not differ between XF and C plots, demonstrating that fertilization did not directly affect CO 2 emission in this zone. As higher elevations emitted more CO 2 , XF plots generally produced more CO 2 due to respiration than C plots ( Figure 3c). As was expected as measurements were performed using darkened chambers to exclude photosynthetic uptake, CO 2 emission was observed during all measurements. Results clearly demonstrate larger CO 2 emissions due to community respiration from higher-elevation vegetation zones in XF October 11.2 ± 4.9 19.5 ± 9.9 CO 2 flux (mmol m −2 hr −1 ) July 28.6 ± 6.6 30.3 ± 3.0 August 39.6 ± 6.7 37.4 ± 2.1 September 20.3 ± 3.2 28.5 ± 2.9

| D ISCUSS I ON
October 12.4 ± 3.2 9.0 ± 0.7 n = 6 to 8 for Creekbank, low and high elevation classes, and 12-16 for the mid elevation class. a Below instrument detection limit.
TA B L E 2 Average greenhouse gas fluxes (±SE) measured monthly from C and XF plot elevation classes plots as hypothesized, but there is a need for further investigation to determine whether increased productivity due to nutrient enrichment drives increased photosynthetic CO 2 uptake in this system. A number of studies report measurements of negative net ecosystem exchange (NEE), the net CO 2 flux resulting from plant photosynthesis and community respiration (performed using transparent flux chambers), for southern New England salt marshes. For example, in highly productive tall S. alterniflora (Moseman-Valtierra et al., 2011) and Phragmites australis (Martin & Moseman-Valtierra, 2015, 2017a marshes, photosynthetic CO 2 uptake exceeded respiration rates during the growing season, resulting in substantial net GHG uptake (up to 54 mmol m −2 hr −1 ). However, NEE measures indicated CO 2 emission or minimal uptake in less productive S. patens and D. spicata marshes (Martin & Moseman-Valtierra, 2015).
Nutrient addition has been shown to increase salt marsh aboveground biomass production (Adam Langley, Mozdzer, Shepard, Hagerty, & Patrick Megonigal, 2013;Deegan et al., 2012;Valiela, Teal, & Sass, 1975), and therefore, both community respiration and photosynthetic CO 2 uptake may increase. At Great Sippewissett Marsh, in the early years of the fertilization experiment, S. alterniflora biomass increased (Valiela et al., 1975), and greater S. alterniflora productivity was associated with higher root zone Eh in a positive feedback loop (Howes et al., 1981). Both photosynthetic CO 2 uptake and CO 2 produced by microbial metabolism are likely increased in XF relative to C plots. However, shifts in species composition may complicate productivity responses to fertilization (Langley & Megonigal, 2010). There is in particular a need for data on effects of nutrient enrichment on salt marsh vegetation communities dominated by deeper-rooting woody species including Iva frutescens, which may influence soil processes through rhizosphere processes such as ventilation and water uptake.
Although elevation gains mediated trade-offs in GHG emissions by supporting decreased CH 4 emission as well as increased CO 2 and N 2 O emission, CH 4 emissions in this system ranged from negligible to small relative to those from freshwater wetlands (Mitsch & Gosselink, 2000). Results of this study and others in simi-  (Wigand, McKinney, Charpentier, Chintala, & Thursby, 2003;Wigand et al., 2014). Nutrient enrichment is a widespread stressor of salt marsh ecosystems, as eutrophication of coastal waters continues to increase globally (Bricker et al., 1999;Rabalais et al., 2009).
The potential for nutrient enrichment to increase GHG emissions has implications for the role that salt marshes play in sequestering C and thereby countering climate change, an ecosystem service for which they are often valued (Mcleod et al., 2011