Stratifying by Vegetation and Hydrology Improves Tidal Marsh Methane Emission Accounting

Methane emissions must be directly measured or estimated using methods such as proxies when managing wetlands for greenhouse gas offset activities. Salinity is a useful proxy for tidal marsh CH 4 emissions when comparing across a wide range of salinity regimes but does not adequately explain variation in brackish and freshwater regimes where variation in emissions is large. We sought to improve upon the salinity proxy in a marsh complex on Deal Island Peninsula, Maryland, USA by identifying four strata based on hydrology and plant community composition. Mean CH 4 chamber-collected emissions measured as mg CH 4 m -2 hr -1 ranked as S. alterniora (1.2 ± 0.3) >> High-elevation J. roemerianus (0.4 ± 0.06) > Low-elevation J. roemerianus (0.3 ± 0.07) = S. patens (0.1 ± 0.01). Sulfate depletion generally reected the same pattern with signicantly greater in the S. alterniora stratum (61 ± 4%) than in the S. patens stratum (1 ± 9%) with the J. roemerianus strata falling in between. We attribute the high CH 4 emissions in the S. alterniora stratum to sulfate depletion likely driven by limited connectivity to tidal waters. Low CH 4 emissions in the S. patens stratum are attributed to lower water levels, higher levels of ferric iron, and shallow rooting depth. Moderate CH 4 emissions from the J. roemerianus strata were likely due to plant traits that favor CH 4 oxidation over CH 4 production. We concluded that stratication by hydrology and plant community composition can be an effective proxy to estimate CH 4 emissions at the site scale. Tidal wetland restoration and conservation projects have the potential to mitigate greenhouse gas emissions to the atmosphere and generate carbon credits, but a better understanding of the factors inuencing wetland CH 4 emissions in brackish and freshwater systems (salinity < 18 ppt) is required to lower carbon crediting project costs and estimation uncertainty at site-specic scales. We found signicantly different methane emission rates across four strata dened by hydrology and plant community composition that otherwise had similar salinity regimes. We inferred that they deviated from the rates predicted by a salinity proxy due to processes that regulate the availability of competing terminal electron acceptors such as O 2 , Fe(III), and SO 42- and due to plant traits the regulate CH 4 emissions. Low CH 4 emission rates in the high-elevation S. patens stratum was attributed to relatively high inputs of Fe(III) through tidal inputs and oxidation of reduced iron and O 2 through diffusion across the soil surface and root O 2 loss, both of which maintain high [SO 42- ] by suppressing microbial reduction of SO 4 . By contrast, the S. alterniora stratum was relatively isolated from tidal inputs of Fe(III) and SO 4 , so it produced large amounts of CH 4 once SO 42-had been suciently depleted. The greater CH 4 emissions from the low-elevation S. alterniora stratum than the low and high elevation J. roemerianus strata could not be explained by water table depth, salinity, pH, and [Fe 2+ ], suggesting an important role for plant traits such as root O 2 loss regulating CH 4 emissions at local scales. The mechanisms driving these patterns were not measured directly but likely involve variations in rates of SO 42- diffusion from tidal oodwater based on site hydrologic connectivity; rates of sulde oxidation as inuenced by O 2 diffusion or root O 2 loss; and differing plant primary productivity. Our results suggest that stratication by vegetation and hydrologic setting improves estimates of CH 4 emissions from tidal marshes. Our ndings illustrate the need to better understand controls over CH 4 emissions at site-specic scales to improve carbon sequestration offset estimates in brackish and freshwater coastal wetland ecosystems.


Introduction
Methane is a potent greenhouse gas produced under the dominantly anaerobic conditions found in wetland soils. The global warming potential of methane (CH 4 ) gas is 32-45 times greater than an equivalent amount of carbon dioxide (CO 2 ) over a 100-year period (Neubauer and Megonigal 2015). While the majority of CH 4 emissions come from anthropogenic sources, wetlands produce most of the naturally emitted CH 4 (Wang et al. 1996;Solomon et al. 2007) and are the most important source of uncertainty in current global CH 4 budgets (Saunois et al. 2020). Coastal wetland CH 4 emissions were recently estimated at 5.3-6.2 Tg CH 4 yr -1 , amounting to 60% of the global marine CH 4 budget (Al-Haj and Fulweiler 2020), < 7% of global wetland CH 4 budget (Saunois et al. 2020), and the largest source of uncertainty in the coastal wetland greenhouse gas budget . There is emerging interest in using tidal marsh restoration and conservation to mitigate greenhouse gases in the atmosphere and as a source of carbon credits (Crooks et al. 2011;, Needelman et al. 2018, Emmer et al. 2020a, Emmer et al. 2020b, but the high carbon sequestration rates characteristic of tidal wetlands soils (Nahlik and Fennesy, 2016;Chmura et al. 2003) can be partly or completely offset by CH 4 emissions (Poffenbarger et al. 2011). The uncertainty introduced to greenhouse gas offset activities by CH 4 emissions is especially large for coastal wetlands ecosystems with freshwater-to-brackish salinity <18 ppt (Poffenbarger et al. 2011). The sources of this variability remain elusive as there has been relatively little research designed to partition variation. A better understanding of the factors that regulate coastal wetland CH 4 emissions is needed to improve global CH 4 budgets and to support the implementation of carbon credit methodologies in freshwater and brackish coastal wetlands.
Methane is produced in wetlands by methanogenic archaea and bacteria. The production of CH 4 occurs when there is an excess of electron donors over electron acceptors, depleting the availability of alternative electron acceptors such as ferric iron (Fe(III)) and sulfate (SO 4 2-) . Electron donors are produced from labile organic materials that undergo fermentation to low molecular weight carbon compounds and H 2 . Electron donors can be present in the soil (e.g. Fe(III)), supplied from external sources such as oodwater (e.g. SO 4 2-), or provided by plants (e.g. molecular oxygen or oxidized compounds generated by radial oxygen loss). The availability of SO 4 from seawater suppresses CH 4 emissions from polyhaline (salinity > 18 ppt) marshes to consistently low rates (0.2 to 5.7 g CH 4 m −2 yr −1 ) (Poffenbarger et al. 2011). Methane emissions from mesohaline brackish systems (5-18 ppt salinity) are greater and more variable (3.3 to 32.0 g CH 4 m −2 yr −1 ). The process is also regulated by such physiochemical factors as pH (Walker et al. 1998;Garcia et al. 2000) and temperature (Megonigal and Schlesinger 2002;Whalen 2005).
Plant species composition affects CH 4 emissions through several mechanisms (Koebsch et al. 2013;Moor et al. 2017;Mueller et al. 2020). The availability of electron donors is largely determined by primary productivity which varies with species composition . Species composition also regulates electron acceptor availability through rhizosphere processes such as root oxygen loss (Calhoun and King, 1997;Colmer, 2003;Jespersen et al. 1998) and rhizosphere regeneration of ferric iron (Neubauer et al, 2005;Sutton-Grier and Megonigal, 2011). Methane can be transported to the atmosphere via aerenchyma tissue, bypassing the emission barriers caused by slow CH 4 diffusion rates through soils and soil-surface CH 4 oxidation zones (Ding et al. 2005;Le Mer and Roger, 2001;Sorrell et al. 2013;Villa et al. 2020). This is important because CH 4 oxidation can consume 70% or more of the CH 4 produced in tidal wetland soils (Megonigal and Schlesinger 2002).
In tidal marshes, water table position and periods of soil inundation are controlled by hydrologic factors such as soil hydraulic conductivity, distance from open water, and soil surface elevation relative sea surface elevation. Elevation zonation subdivides tidal marshes into low marsh areas that are frequently inundated by the tides, and high marsh areas that are infrequently in uenced by the tides. Water table depth in uences soil oxygen availability (Epp and Chanton, 1993;Gilbert and Frenzel, 1995), and hence the potential for aerobic processes such as CH 4 oxidation (Grün eld and Brix, 1999;Megonigal and Schlesinger 2002). Soils in low marsh areas that are permanently or frequently inundated experience low rates of O 2 diffusion and sustain anaerobic environments where methanogenesis can occur (Ding et al. 2010). Tidal wetland studies have documented correlations between elevation, water level, and CH 4 emissions (Altor and Mitsch, 2006;Audet et al. 2013;Ding et al, 2010;Grün eld and Brix, 1999), implicating hydroperiod as a dominant in uence on wetland CH 4 emissions.
Both plant community composition and water table depth have proven to be effective proxies for predicting CH 4 emissions in wetland ecosystems such as peatlands (Audet et al. 2013;Bubier et al. 1995;Couwenberg et al. 2001;Dias et al. 2010). Wetland plant species exhibit different tolerances to inundation (Sorrell et al. 2000;Vann and Megonigal, 2003), leading to varying plant community composition across elevation gradients (Perry and Hershner 1999). Wetland vegetation is well suited for serving as a proxy to predict CH 4 uxes due to direct and indirect in uences of plant species on labile soil organic carbon (i.e. root exudates), soil moisture, and CH 4 gas transport via plant aerenchyma tissue (Couwenberg et al. 2001). Previous studies have established direct links between plant species composition and CH 4 uxes (Audet et al. 2013;Bhullar et al. 2014;Shäefer et al. 2011) and have used plant species composition to accurately predict CH 4 uxes from peatlands (Couwenberg et al. 2001;Dias et al. 2010).
Water table depth, as in uenced by relative elevation, has also proven to be a good proxy for predicting CH 4 emissions in peatlands as water table level determines aerobic/anaerobic zones and redox states in the soil pro le (Ding et al. 2010).
The objective of this study was to advance our understanding of the effects of water level and plant species composition on CH 4 emissions in brackish marshes at a site scale. We measured CH 4 uxes in two brackish marshes on the Deal Island Peninsula on the Eastern Shore of Maryland, USA across four different strata de ned by water level and plant community composition (Needelman et al. 2018). We collected eld data on elevation, water level, soil temperature, and soil pore water SO 4 2-, sul des, pH and salinity, and laboratory soil incubations using eldcollected soil cores to assess potential CH 4 production.

Study Area
Our study area was located on the Deal Island Peninsula in Somerset County, Maryland, USA (38.185172N, 75.906279W) (Fig. 1). It consisted of two brackish tidal marshes, one unditched (Unditched) and one that had been ditched then restored (Ditched) located in the same marsh complex (referred to as Unditched-2 and Ditched-2 in . Ditch plugs were installed at the Ditched site in April of 2014 by inserting a plastic polyethylene sheet vertically across the ditch approximately 50 m upstream from the tidal source and securing the plug using sediment sourced from the ditch upstream of the plug. The Ditched site had an overall lower elevation than the unditched site, and was primarily composed of Juncus roemerianus (black needlerush). The Unditched site had a more diverse species community including J. roemerianus, Spartina patens (salt marsh hay), Spartina alterni ora (smooth cordgrass), Phragmites australis (common reed), and Iva frutescens (marsh elder). Plant productivity in tidal marshes in this region include a period of senescence during the late fall through the early spring, with peak plant productivity occurring in late July through August. Soils on site consist of thick moderately to highly decomposed organic horizons overlying loamy mineral horizons; within Soil Taxonomy they classify as the Mispillion series, Loamy, mixed, euic, mesic, Terric Sul hemists, which are common estuarine marsh soil in this area. This microtidal marsh had a diurnal tidal range of approximately 0.6 meters as measured in the adjacent tidal creek.

Design
Four strata that differed in their plant community composition and elevation, both of which are closely associated with water levels, were identi ed prior to the study from onsite observations and overhead satellite imagery. The strata corresponded to geographic units that may be used to estimate CH 4 emissions when engaged in site-speci c carbon crediting accounting methodologies . Water level variability was primarily controlled by elevation in these marshes, with lower elevations having higher water levels. Two of the strata had a plant community composition dominated by J. roemerianus, but differed in elevation; with one site at a "High" elevation and the other at a "Low" elevation. The High J. roemerianus stratum was located at the unditched site and had a mean elevation of 0.334 m relative to NAVD88, while the Low J. roemerianus stratum was located at the ditched site with a mean elevation 0.305 m. The two additional strata consisted of one dominated by S. alterni ora at a relatively low mean elevation of 0.299 m, and one dominated by S. patens at a relatively high mean elevation of 0.409 m. Both the Low S. alterni ora and High S. patens strata were located at the Unditched site. A representative area was selected within each stratum that included a range of elevation and plant diversity. Five sampling plots were randomly established within the representative area in each stratum, for a total of 20 plots. It should be noted that our ux measurements covered a small inference space, since they were only in representative areas of each strata and not randomly distributed across the entire marsh. Three of the strata were located within 25-50m of a tidal creek, while one (S. alterni ora) was located in a more central location in the marsh complex at approximately 100 m from a tidal creek. The 20 plots occupied an approximate area of 0.06 km 2 , with each 5-plot strata encompassing an approximate area of 1,000 m 2 .

Field Methods
We sampled monthly from April to December 2015; samples were not taken from January until March under the assumption that CH 4 production would be negligible due to low temperatures (Marsh et al. 2005). Methane ux, air temperature, and pore water concentrations of pH, SO 4 2-, hydrogen sul de, and CH 4 were measured at each plot. Soil temperature at 10 cm was recorded at two plots per stratum hourly during the sampling season using HOBO 8k Pendant sensors (Onset Corp., Bourne, MA). Soil temperature and water level data were not collected during the month of April because loggers were not ready for deployment until May.
Each of the 20 sample plots received a custom-fabricated square aluminum metal collar that was permanently inserted into the marsh to a depth of 10 cm nine months prior to the rst sample. Flux chambers were constructed of an aluminum frame made of 2.5-cm wide angle stock covered with transparent polycarbonate plastic lm. Chambers were placed on top of the collar about 10 minutes prior to sampling. Chambers were equipped with a closed-cell neoprene strip on the top and bottom, which when clamped to the collar assured an airtight seal (Yu et al. 2013). The taller plants in the J. roemerianus strata were accommodated without damaging plant stems by stacking chambers.
Opaque chamber lids with a sampling port were clamped to the top of the chamber to complete the seal. Chambers had a height of 69.5 cm and an interior length and width of 49.5 cm, yielding a total volume of 0.17 m 3 for single chambers and 0.34 m 3 for double chambers. In order to prevent the weight of the observer from causing ebullition due to soil compression (Sorrell et al. 2013), each plot had a 3m wooden boardwalk suspended above the soil surface by PVC legs for approaching the ux collar.
Methane ux samples were collected over a 1-hr period from the 5 replicate ux plots in a given strata. An initial sample was taken immediately after each chamber was sealed with four subsequent samples taken at approximately 15-min intervals for a total of 25 samples (5 per plot) over the 1-hr period. Using a 30 mL syringe, the sampling port was opened and then expelled back into the chamber three to ve times before each sample was taken. Each 18 mL air sample was withdrawn from the chamber and injected into a N 2 -ushed 12-mL Exetainer vial with rubber septum until analysis (Yu et al. 2013). Air temperature within the sampling chamber was recorded upon the collection of each ux sample from thermometers a xed to the interior of each chamber.
Porewater samples were taken at 10 cm depth using a porewater sipper and syringe (Fisher et al. 2013) and analyzed for pore water CH 4 , hydrogen sul de (un ltered), pH (un ltered), salinity (un ltered), and SO 4 2-( ltered through a 0.45-µm lter) as described by Keller et al. 2009. Porewater CH 4 was collected by withdrawing 15 mL of pore water, after which 15 mL of ambient air was drawn into the syringe and the syringe capped. The sample was then agitated for 1-2 minutes for the CH 4 to be stripped into the drawn air, the stripped water was expelled, and the gas sample was stored in N 2 -ushed Exetainers for analysis (Keller et al. 2009). Hydrogen sul de samples were diluted in a 1:1 ratio of sample to sul de antioxidant buffer in the eld to prevent sul de volatilization and oxidation (Koch et al. 1990). Hydrogen sul de and pH samples were analyzed the same day as sample collection; salinity was analyzed within two weeks in the laboratory using a YSI Model 3100 conductivity meter; and all other pore water samples were frozen and analyzed during the winter of 2016.
Additional data were collected during the July 2015 sampling event, which was predicted to be during a peak CH 4 emission period. We collected porewater at 20 cm depth in addition to 10 cm and analyzed it for the same analytes excluding CH 4 but including ferrous iron (Fe 2+ ).
Water level was measured at each stratum in order to determine water levels at the time of sampling and antecedent water level conditions during the two-week period leading up to the sampling period. Water level recorders (HOBO U20-L, Onset Corp, Bourne, MA) were installed adjacent to the chamber transects to continuously record water levels in the marsh; one was also installed in the tidal creek adjacent to the eld site during the eld season. We deployed two water level loggers in each stratum, except for the low water table J. roemerianus stratum, which had one water level logger due to its small area relative to the other strata. Barometric pressure was collected onsite to correct the unvented loggers (HOBO U20-L, Onset Corp, Bourne, MA). We surveyed the elevation of all 20 plots and water level logger locations using a Real-time Networking Global Positioning System (RTN GPS) unit, which provides elevation data with approximately 2 cm accuracy (http://www.keynetgps.com).
Soil cores were collected during the July sampling event and analyzed for potential anaerobic CH 4 and CO 2 production. Cores were collected from approximately 0-40 cm depth using a circular metal gouge corer. The corer was inserted into the marsh, with careful attention paid to minimize compaction of the soft peat. The core was removed and cut at a depth of 20 cm, yielding two depth increments per plot. Cores were placed into sample bags in which as much air as possible was removed. The cores were then placed in a cooler with ice and transported back to the laboratory, where each bag was ushed three times with nitrogen gas to remove oxygen, stored on ice during transport, and placed in a 4 °C cold room until processing. Water for these incubations was collected from the bore hole from which the core was removed, stored on ice for transport back to the lab, purged with nitrogen gas to remove oxygen before being sealed and placed in a cold room at 4 °C. Soil cores and water samples were stored in the cold room within 8 hours of their collection and incubated within 5 days.

Laboratory Analyses
Flux chamber headspace samples were measured on a Varian 450 gas chromatograph using a Combi-Pal autosampler and corrected for dilution of 18 ml of sample into 12 ml of N 2 in the Exetainer. Flux rate was calculated as the linear increase in headspace [CH 4 ] over time based on measurements of chamber temperature and volume and assuming atmospheric pressure (n=147 uxes). The linear slope was calculated in Excel using the Regression function. Data points were excluded from the regression if they indicated an ebullition event (large spike in [CH 4 ]) or an Exetainer leak (large drop in [CH 4 ]). Most uxes were calculated from ve points, but never from fewer than three points. No ux measurements were excluded based on arbitrary regression R 2 or p-value limits but uxes were excluded in several cases where an ebullition event or leak was large compared to the CH 4 ux rate rate (n=17). In cases where there was no signi cant trend in headspace [CH 4 ] and no evidence of ebullition or leaks the ux was assigned a value of zero. Because most of the excluded uxes were collected during periods of low CH 4 ux they had relatively little in uence on the annual ux calculation.
To estimate annual emissions averaged rates from each measurement campaign. We assumed that the uxes in the unsampled months of January, February, and March were equal to our observed values from April. The twelve monthly values were averaged and converted to annual units. While this method likely overestimated CH 4 emissions, overestimation is the conservative and therefore preferable approach for carbon credit accounting (Needelman et al. 2018).
Hydrogen sul de was determined with a Lazar Laboratory model 146S sul de electrode. Sul de antioxidant buffer was prepared the day before sample collection with deoxygenated (N 2 -stripped) distilled water, sodium salicylate, sodium hydroxide, and ascorbic acid according to Koch et al. (1990). A standard curve created from a serial dilution of a Na 2 S/buffer solution prepared on the day of each sampling event and readings were complete within 4 hours of sample collection. Soil cores collected for incubations were removed from cold storage within 5 days of collection and placed into an anaerobic hood containing a N 2 /H 2 mixture (Megonigal and Schlesinger 2002). Two sections were removed from each core, yielding a 8-12 cm depth sample and a 28-32 cm depth sample. The outer 10 mm (approximately) of the resulting disks were removed to expose the center of the core, which was assumed to have had minimal O 2 exposure from collection to processing. We then removed as many live roots as feasible. Five grams of wet soil material was placed in a 35-mL serum bottle with 5 mL of the degassed water from the core hole. Headspace samples of 0.5 mL were injected directly into a Shimadzu gas chromatograph with a ame ionization detector for CH 4 or a LI-COR LI-7000 (LiCor, Lincoln, NE) for CO 2 . Methanogenesis generally slowed dramatically after 5 days, so our calculations of potential anaerobic CH 4 production rates are based on incubation days 1-5.

Statistical Analysis
Data were analyzed using SAS 9.3 (SAS Institute Inc. Cary, NC). Regression analyses were performed on ux data using Proc Reg to determine the slope of CH 4 or CO 2 concentration change over time. All variables were evaluated for normality using PROC UNIVARIATE and those that required transformation were log transformed to improve normality. Parameters transformed were: CH 4 ux, pore water hydrogen sul de concentration, pore water SO 4 2concentration, and pore water CH 4 concentrations. All parameters were analyzed using PROC MIXED with strata and month in the model statement with repeated measures. Post-hoc Tukey mean comparisons were used with α = 0.05 used to indicate signi cance.

Antecedent Water Levels
Water level data collected during the 2 weeks prior to and during sampling events varied signi cantly between strata (p<0.0001) and month (p<0.0001). Water levels of the S. patens stratum was signi cantly lower than all other strata, with a mean of 9 cm below the soil surface, while the other strata had similar mean water levels approximately 1 cm below the soil surface (Table 1). We were unable to test for a strata by month interactive effect because only two wells were deployed in each strata (and only one in the High J. roemerianus stratum); however, S. patens had a lower mean water level in all months (Derby 2016). Mean water levels were highest in July, August, and October and lowest in May, June, September, and December.

Methane Emissions and Porewater Chemistry by Strata
Average CH 4 ux over the study varied signi cantly between strata (p<0.0001) and month (p<0.0001), and had a signi cant interactive effect (p=0.018). Methane emissions from the four strata ranked S. alterni ora >> High J. roemerianus > Low J. roemerianus = S. patens (Table 1). Mean CH 4 emissions from the S. alterni ora stratum was 2.72 times greater than the next highest CH 4 emitter ( were due to high rates of SO 4 2consumption. Indeed, SO 4 2was depleted by 61% in the S. alterni ora stratum. Sulfate depletion was signi cantly different between strata (p<0.0001) and month (p<0.0001) and ranked S. alterni ora = High J. roemerianus > Low J. roemerianus > S. patens. Sulfate depletion rates in the S. alterni ora stratum were over two times greater than those seen in low J. roemerianus (Table 1). Low SO 4 2concentrations in the S. alterni ora stratum were accompanied by signi cantly higher amounts of hydrogen sul de (72 mg L -1 ) as compared to the other three strata (all < 20 mg L -1 ).
The S. patens stratum had the lowest mean porewater CH 4 concentrations and SO 4 2depletion rates, with only 1.4% SO 4 2depleted. S. patens also exhibited the highest mean concentrations of reduced iron (i.e. ferrous iron, Fe 2+ ) during the single campaign when it was measured, with 72 mg L -1 of ferrous iron in porewater collected 10 cm below the soil surface (Table 1). None of the other three strata had reduced iron porewater concentrations exceeding 0.8 mg L -1 . S. patens also had a signi cantly lower mean porewater salinity (12.3 ppt) than the other three strata, which ranged in mean salinity from 14.2 to 14.8 ppt. The High and Low J. roemerianus and S. alterni ora strata were not signi cantly different in porewater iron concentrations or salinity (Table 1).
The difference in elevation between the two J. roemerianus strata was not highly apparent in the water level and CH 4 -related attributes we measured. The two strata were not signi cantly different from one another in, salinity, SO 4 2concentrations, sul de concentrations, percent SO 4 2depleted, porewater CH 4 concentrations, and reduced iron concentrations ( Table 1).
The highest CH 4 emissions were observed in July, August, and September; the lowest were observed in April, November, and December (Fig. 2). Signi cant strata by month interactions were observed in May, June, and September. In May and June, S. alterni ora was not signi cantly different from any strata other than Low J. roemerianus; all other strata were not signi cantly different from one another. In September, S. alterni ora was not signi cantly different from any strata other than S. patens; all other strata were not different from one another. No signi cant within-month differences were observed in April, July, August, October, November and December (Derby 2016). Porewater CH 4 exhibited a similar seasonal trend as CH 4 emissions, with the highest concentrations in the months July through November (Derby, 2016).

Anaerobic incubations
Sur cial soils (8-12 cm) from the S. patens stratum had the lowest CH 4 production, highest CO 2 production and a signi cantly higher ratio of CO 2 :CH 4 production (ratio=993) that the other strata. At the other extreme was the S.
alterni ora stratum which produced substantially more CH 4 and less CO 2 than the other strata, and therefore had the lowest CO 2 :CH 4 ratio (ratio=40) among the four sites ( Table 2). The two J. romerianus strata fell in between these extremes with a CO 2 :CH 4 ratio of about 200 (Table 2), though there were no signi cant differences in CO 2 :CH 4 ratio between these strata and S. alterni ora. The 10 cm incubations produced signi cantly more CH 4 and CO 2 than those from 30 cm depth (p=0.04, data not shown, Derby 2016).

Discussion
Methane emissions varied by plant community type and hydrogeomorphic setting, suggesting that these variables are useful for dividing tidal marshes into strata to optimize the costs of sampling effort with the need to reduce parameter uncertainty when estimating CH 4 emissions. Mean emissions across strata ranked as S. alterni ora > High-elevation J. roemerianus > Low-elevation J. roemerianus >S. patens (Table 1). There is a need to understand the mechanisms that lead to such differences among strata in order to advance proxies and models of CH 4 emissions from tidal wetlands.
We attribute the low emissions from the S. patens stratum to a combination of relatively low water levels, shallow rooting depth, and higher mineral inputs, all of which have the capacity to suppress CH 4 production and promote CH 4 oxidation. It is well established that CH 4 production is suppressed by alternative electron acceptors such as O 2 , Fe(III) and SO 4 2- (Holm et al. 2016;Neubauer et al. 2005;Poffenbarger et al, 2011;Roden and Wetzel, 2003). Relatively high O 2 ux into the upper soil surface (0-10 cm depth) would be favored by both the relatively thick aerobic zone (i.e. deeper water-table) and shallow distribution of root biomass that is characteristic of S. patens communities (Bernal et al. 2016). Mean antecedent water depth was 9 cm deep in this stratum compared to other strata with water levels near the soil surface.
Because the majority of S. patens roots are in the top 10 cm of the soil pro le (Windham 2001), it is likely that root oxygen loss was also an O 2 source in this community. O 2 suppresses methanogenesis via electron donor competition by two mechanisms, directly as an electron acceptor for aerobic bacteria and indirectly by regenerating poorly crystalline iron oxides on the root surface (i.e. iron plaque). Root-deposited iron plaque is rapidly consumed by iron-reducing bacteria (Weiss et al. 2003;Weiss et al. 2004), suppressing both SO 4 2reduction and methanogenesis (Neubauer et al. 2005). This mechanism is consistent with our observation that the S. patens community had dramatically higher concentrations of reduced iron (measured only in July) and SO 4 2than the other strata, and the lowest SO 4 2depletion ( Table 1). The close proximity of this site to the tidal creek may have also allowed for greater mineral inputs during ooding events to support iron cycling. Finally, if rates of methanotrophy are O 2 -limited as studies suggest (King, 1996;Lombardi et al, 1997;Megonigal and Schlesinger 2002), then relatively high rates of CH 4 oxidation would be expected to further decrease the amount of CH 4 emitted through passive diffusion through plants, as documented in other tidal wetlands (Megonigal and Schlesinger 2002).
The S. alterni ora stratum had the highest average CH 4 emissions and porewater chemistry that differed from the S. patens community in several respects (Table 1). The S. alterni ora stratum was in the center of the marsh complex (Fig. 1). Because hydrologic uxes generally decrease with increasing distance from the tidal creek (Jordan et al. 1985), it is likely that soils in the S. alterni ora stratum had relatively slow rates of advection compared to the other strata. Indeed, water table depths decreased relatively slowly after oods in the S. alterni ora stratum (Derby 2016).
This hydrologic difference likely decreased rates of advective transport of O 2 and SO 4 2to the soil pro le to replenish these electron acceptors. We propose that the relatively low inputs of SO 4 2to the S. alterni ora stratum led to SO 4 2limitation of SO 4 2reduction rates, allowing methanogens to compete more effectively with sulfate-reducing bacteria for electron donors. Porewater evidence supporting this interpretation includes low concentrations of Fe 2+ , high SO 4 2depletion, and high concentrations of both hydrogen sul de and CH 4 . This interpretation is also consistent with the results of the July anaerobic incubations showing that the CO 2 :CH 4 ratio was lowest in S. alterni ora soils and highest in S. patens soils. Because aerobic respiration, sulfate reduction, and iron reduction generate CO 2 rather than Page 10/19 CH 4 , these data suggest that methanogens had relatively little competition for organic carbon and H 2 in the S. alterni ora stratum.
Water table depths in the High and Low J. roemerianus strata were similar to the S. alternifora stratum but metrics related to CH 4 emissions were consistently lower than the S. alterni ora and higher than the S. patens strata, namely CH 4 emissions, porewater SO 4 2concentrations, and anaerobic incubation CO 2 :CH 4 ratio. The difference in CH 4 emissions between S. alterni ora and the J. roemerianus strata cannot be explained by water levels, salinity, pH, or reduced iron because they were not signi cantly different between these strata. For example, mean water table depth in the S. alterni ora stratum and the two J. roemerianus strata were within 1 cm of each other, while emissions were 2.5 times greater in the S. alterni ora stratum. The J. roemerianus strata were closer to the tidal creek and presumably CH4 production was not limited by tidal inputs of SO 4 2supply as we suspect was the case in the S.
alterni ora stratum. However, the relatively low CH 4 emissions in the J. roemerianus strata compared to S.
alterni ora may also have been related to plant traits that regulate CH 4 emissions by in uencing the balance between CH 4 production and oxidation (Sutton-Grier and , Mueller et al. 2020, which itself is in uenced by traits that affect CH 4 transport through plants (Komiya et al. 2020). Mueller et al. (2020) proposed that plant traits vary across species such that some push the balance between these opposing processes toward net CH 4 production while others favor net CH 4 oxidation. We hypothesize that among the dominant species present at the site, S. alterni ora favors net CH 4 production while J. roemerianus favors net CH 4 oxidation, and that the lower CH 4 emissions rates in the J. roemerianus strata may have been due to relatively high rates of root oxygen loss by J.
roemerianus. This interpretation is supported in part by data on porewater [SO 4 2-], which mediates the outcome of competition between sulfate-reducing bacteria and methanogens. Porewater [SO 4 2-] was 4.5 mM in the S. alterni ora stratum but exceeded 6 mM in the J. roemerianus strata where CH 4 emissions rates were relatively low. Although this difference in porewater [SO 4 2-] seems small, the relationship between these variables is non-linear and displays a threshold value above which sulfate reduction dominates and below which methanogenesis dominates ). There is a general lack of CH 4 -relevant porewater data for coastal wetlands, but a robust record from a brackish marsh located 100 km from the study site in Chesapeake Bay observed that porewater [CH 4 ] declined abruptly when SO 4 2concentrations exceeded approximately 4 to 6 mM (Keller et al. 2009). We propose that the S.
alterni ora and J. roemerianus strata fell on opposite sides of this threshold.
Although oodwater salinity can be an effective proxy for porewater SO 4 2concentrations when comparing sites across large spatial scales (Poffenbarger et al. 2011), our data demonstrate the limitations of using the salinity proxy at local scales. Variation among strata in SO 4 2concentration and SO 4 2depletion may have been caused by variation in rates of SO 4 2transport from tidal oodwater into soils, sul de oxidation related to O 2 diffusion from water table depth or root O 2 loss, or primary production (i.e. carbon availability). We cannot distinguish among these mechanisms with the present dataset. Sulfate depletion was a better indicator of CH 4 ux than salinity or SO 4 2concentration alone and may prove to be a superior proxy for CH 4 emissions in tidal brackish marshes.
Within our strata, CH 4 production did not strictly follow the conventional interpretation that differences in the free energy yield among competing microbial respiration processes means that just one process dominates microbial respiration at a time, with methanogenesis expected to occur only when all other electron acceptors are fully (or nearly) depleted. Our data suggest that peak sulfate reduction activity was occurring concurrently with peak CH 4 production in the S. alterni ora stratum which had both the highest mean pore water hydrogen sul de concentrations and the lowest SO 4 2concentrations. We attribute this to spatial variation in electron donors and acceptors that creates microsites of high SO 4 2depletion and methanogenesis. Microsites have been shown to produce small amounts of CH 4 in upland forested systems, originally thought to be too dry and too aerobic to produce this greenhouse gas (Megonigal and Guenther 2008). Microsites also exist in tidal marsh soils due to local (i.e. rhizosphere) consumption of electron acceptors at rates faster than they can be replenished (e.g. Rabenhorst et al. 2010).

Carbon offset implications
Salinity is a useful proxy for CH 4 emissions from tidal marshes with salinity regimes > 18 ppt because CH 4 emissions are low compared to their soil carbon sequestration rates and variation among marshes is small. However, tidal brackish marshes at lower salinities may emit enough CH 4 to offset a signi cant fraction of their radiative cooling and variation in CH 4 emissions among marshes within a given salinity regime is large. Such uncertainty is accommodated in carbon offset programs such as the Veri ed Carbon Standard by requiring the project to estimate CH 4 emissions by direct monitoring or by using published data, a model, or a proxy that can be demonstrated to be valid for the project site (Needelman et al. 2018). Strati cation by hydrology and vegetation characteristics may provide a more effective proxy than salinity to estimate CH 4 emissions. Strati cation also offers a means to constrain variability within direct monitoring schemes.
We compared the CH 4 offset estimates produced through our direct measurements to those predicted by the salinitybased proxy equation of Poffenbarger et al. (2011). For this comparison we assumed a single soil carbon sequestration rate of 1.46 Mg C ha -1 yr -1 for all strata in the marsh complex, which is the default rate in the carbon crediting methodology of , derived as the median value from Chmura et al. (2003). In three of the four strata the salinity proxy overestimated emissions by about 20%, 40%, and 80% (Table 3). Overestimation is preferable to underestimation to avoid awarding carbon credits that exceed actual greenhouse gas bene ts, but overestimation decreases the nancial viability of carbon offset projects. These errors caused the positive radiative balance to be underestimated by just 4-8% in the J. roemerianus strata, suggesting that incorporating vegetation and hydrology proxies would not be a substantial improvement over the salinity proxy alone. In addition, the cost of in situ emission measurements would not be rewarded by a meaningful increase in carbon credits in the J. roemerianus strata. However, the positive radiative balance of the S. patens stratum was underestimated by >20%, suggesting that a proxy based on vegetation and hydrology would improve CH 4 emission estimates, and that the expense of collecting in situ data may be worthwhile. The salinity proxy overestimated CH 4 emissions in the S. alterni ora statum where the positive radiative balance was overestimated by about 25% (Table 3), indicating that a salinitybased proxy would award carbon credits exceeding actual climate bene ts in this stratum. Improved proxies are needed to incentivise carbon offsets projects by reducing monitoring costs while ensuring that projects achieve estimated climate bene ts.
Our results suggest that proxies for CH 4 emissions from tidal wetlands with salinity regimes < 18 ppt can be improved by incorporating metrics related to hydrology such as ooding frequency and duration or the position of the soil surface relative to the tidal frame, and metrics related to plant traits such as species identity, plant functional type, or biomass. Ideally these metrics would be identi able at high spatial resolution for low cost, such as through analysis of freely available remote sensing data. Currently there is no widely accepted method to remotely sense surface water salinity, but robust methods exist for remote sensing of plant cover, biomass, primary production, and elevation (

Conclusions
Tidal wetland restoration and conservation projects have the potential to mitigate greenhouse gas emissions to the atmosphere and generate carbon credits, but a better understanding of the factors in uencing wetland CH 4 emissions in brackish and freshwater systems (salinity < 18 ppt) is required to lower carbon crediting project costs and estimation uncertainty at site-speci c scales. We found signi cantly different methane emission rates across four strata de ned by hydrology and plant community composition that otherwise had similar salinity regimes. We inferred that they deviated from the rates predicted by a salinity proxy due to processes that regulate the availability

Declarations
Funding This work was supported by the USDA National Institute of Food and Agriculture, Hatch project 1013805, the NSF Research Coordination Network Program of the Ecosystems Science Cluster (DEB-1655622), the Smithsonian Institution, and the Garden Club of America.

Con icts of interest
None.
Availability of data and material (data transparency) All data from this manuscript will be made available through the Coastal Carbon Research Coordination Network.
Code availability (software application or custom code) Not applicable  Table 3. Field-measured salinity and methane ux mean values from four strata in a tidal marsh complex as compared to predicted ux (based on observed salinity) from Poffenbarger et al. 2011, with percent differences of actual and predicted carbon sequestration offsets.