The greenhouse gas offset potential from seagrass restoration

Awarding CO2 offset credits may incentivize seagrass restoration projects and help reverse greenhouse gas (GHG) emissions from global seagrass loss. However, no study has quantified net GHG removal from the atmosphere from a seagrass restoration project, which would require coupled Corg stock and GHG flux enhancement measurements, or determined whether the creditable offset benefit can finance the restoration. We measured all of the necessary GHG accounting parameters in the 7-km2Zostera marina (eelgrass) meadow in Virginia, U.S.A., part of the largest, most cost-effective meadow restoration to date, to provide the first seagrass offset finance test-of-concept. Restoring seagrass removed 9,600 tCO2 from the atmosphere over 15 years but also enhanced both CH4 and N2O production, releasing 950 tCO2e. Despite tripling the N2O flux to 0.06 g m−2 yr−1 and increasing CH4 8-fold to 0.8 g m−2 yr−1, the meadow now offsets 0.42 tCO2e ha−1 yr−1, which is roughly equivalent to the seagrass sequestration rate for GHG inventory accounting but lower than the rates for temperate and tropical forests. The financial benefit for this highly successful project, $87 K at $10 MtCO2e−1, defrays ~10% of the restoration cost. Managers should also consider seagrass co-benefits, which provide additional incentives for seagrass restoration.

be directly attributed to a restoration project in a recognized carbon pool (i.e., negative emissions over time), minus any GHG emission increases. It is important to emphasize that this enhanced sequestration equals the CO 2 sequestered by the restoration project (i.e., the 'with project' scenario) minus the background sequestration that would occur if the project did not exist (i.e., in the status quo baseline: the 'without project' scenario) 24,25 . The former can be measured directly; the latter must be estimated by extrapolating pre-project conditions or by comparing project and control sites over time.
For seagrass restoration projects, the net GHG benefit equals CO 2 sequestrated as enhanced sediment C org (see Fig. 1: gross meadow sediment stock minus an equivalent area bare sediment stock) and the long-term average C org sequestered in above-and belowground biomass within the project area, minus any enhanced GHG production 24,25 -specifically CH 4 , N 2 O, and CO 2 evasion associated with CaCO 3 buried in seagrass sediment [20][21][22]26 . Community respiration does not affect the GHG offset benefit for meadow restoration projects, because CO 2 fixed through photosynthesis and then returned to the atmosphere through respiration is not a net flux of CO 2 to the atmosphere. Enhanced respiration could, however, adversely affect a seagrass conservation project attempting to avoid the remineralization of sequestered C org stocks. As noted above, the offset benefit from seagrass biomass sequestration over interannual timescales corresponds to the average, annual standing biomass stock, not peak biomass. This average reflects loss and turnover due to herbivory, senescence, export, and, in some cases, harvest or other disturbances. Some of the exported seagrass carbon may remain sequestered at deep ocean depositional sites 27 , and some is deposited along the coastline as wrack on beaches, marshes, and on tidal flats. The VCS and other offset crediting standards conservatively assume that exported biomass is decomposed and returns to the atmosphere as CO 2 .
The offset-credit methodology recommends measuring the sediment C org stock repeatedly over time to quantify sequestered C org enhancement (i.e., stock change), rather than measuring the C org stock to an arbitrary depth on a single occasion or estimating C org accumulation from burial rates 25 . This is because seagrass sediment C org stock estimates 15,28,29 and burial rates 2,30,31 likely overestimate net CO 2 removal from the atmosphere due to uncertainties with dating techniques for sediment accretion over relatively short time scales (decades) 18 . These estimates also include allochthonous carbon (C org fixed outside the project area) that is excluded from GHG offset accounting methodologies and background C org that would be sequestered in the area in the baseline scenario (see Supplement) 18,32 . This study shows how repeated stock change measurements can provide a more reliable approach for assessing how meadow presence enhances sediment C org accumulation and how Seagrass meadow sediment C org concentrations are typically highest below the surface in a region corresponding with the rhizosphere and approach the background concentration observed at unvegetated sites with increasing depth (data adapted from Greiner et al. 12 and used with permission). The seagrass-enhanced sediment C org stock (light gray) can be quantified by integrating the area under the profile and subtracting the background C org stock that one would expect to find absent the meadow (dark gray); note that this approach does not require establishing a reference plane or quantifying bed accretion (black gradient) attributable to the meadow by sediment dating.

Results
Enhanced carbon sequestration. With repeated stock change measurements, we observed significant C org stock enhancement at the meadow scale resulting from increasing C org concentrations within the bed, seagrass-enhanced bed accretion, and meadow expansion. The meadow-wide, net sediment C org sequestration attributable to the restoration increased from 1,130 t C org in 2013 to 2,010 t in 2016 (Table 3; Fig. 3). Note that these values are stocks relative to a known baseline that represents the 'without restoration project' scenario, not rates, which can be obtained by dividing the stock by a time interval. Approximately 280 t of this 880 t C org stock increase occurred in the top 2 cm of the bed, which was likely deposited between 2013 and 2016 (see Supplement discussion of accretion); the remaining 600 t accumulated within the bed between 2013 and 2016. The 2013 meadow stored an average of 196 g C org m −2 and the 2016 meadow stored an average of 292 g C org m −2 . The 2013 enhanced stock took 12 years to accumulate. Between 2013 and 2016, the enhanced sediment C org stock almost doubled, indicating that the sequestration rate also increased. Meadow C org sequestration in sediments was 346 t CO 2 yr −1 from 2001-2013 and 1070 t CO 2 yr −1 from 2013-2016.
The average aboveground biomass standing stock over three years was 109 gdw m −2 , equivalent to approximately 40.5 g C org m −2 . This reflects seasonal fluctuations that ranged from 330 g dry weight (gdw) m −2 in August (201.4 ± 29 g live plus 129.7 ± 15 g dead) to 38.5 gdw m −2 in March (19.58 ± 4.8 g live plus 18.86 ± 2.4 g dead) (see Supplement). All reported errors relate standard errors (SE), unless otherwise stated. The average aboveground biomass shoot −1 was 0.4 ± 0.07 gdw. Multiplying the average annual biomass per shoot by the interpolated average annual density values and integrating over the meadow area yielded an aboveground biomass standing stock of 710 t CO 2 in 2013 and 810 t CO 2 in 2016, due to meadow expansion. This standing stock is the average amount of C org held in seagrass biomass throughout the year and is less than a third of the peak biomass in summer. Live belowground biomass ranged from 35.51 ± 7.3 gdw m −2 in January to 95.26 ± 13 gdw m −2 in August; the average annual live belowground biomass was 47.1 gdw m −2 (Supplement). Dead belowground biomass ranged from 91.03 ± 17 gdw m −2 in June 2016 to 131.91 ± 12 gdw m −2 in March, yielding an average, annual dead belowground biomass of 119 gdw m −2 (Supplement). Average, annual unit area estimates for live and dead belowground biomass were 16.0 and 40.4 g C org m −2 , respectively. Multiplied by the respective meadow areas, the combined belowground biomass stock sequestered 1,200 t CO 2 in 2013 and 1,520 t CO 2 in 2016.
Sediment C org represented the largest sequestered carbon pool in the meadow in both 2013 and 2016, accounting for 68.5% of the total GHG benefit in 2013 and more than three-quarters of the total GHG benefit in 2016 (Table 3). Annual belowground biomass (live + dead) accounted for 14.7% of the total 2016 sequestered stock, and aboveground biomass represented 8.4%. Enhanced sediment C org and the average, annual seagrass stock sequestered a combined 6,060 t CO 2 in 2013 and 9,590 t CO 2 in 2016 ( Table 3).
The total, cumulative gross primary production (GPP) in the meadow from 2001-2013 was calculated to be 39,700 t CO 2 . By 2016, this estimate had increased to 84,900 t CO 2 , due to meadow expansion. Total, enhanced C org sequestration was, therefore, 15.3% of cumulative GPP in 2013 and 11.3% in 2016.
Bulk porewater CH 4 concentrations measured at seagrass and bare sites in August and October yielded a negligible diffusive flux (Fig. 5). The highest average CH 4 porewater concentration was 0.30 ± 0.25 µmol L −1 at 1.5 cm below the sediment water interface at seagrass sites in October. The highest average concentrations in August were observed at 10.5 cm below the sediment water interface, 0.18 ± 0.14 µmol L −1 at the bare sites and 0.19 ± 0.06 µmol L −1 at the seagrass sites (Fig. 5). Assuming a sediment diffusivity of 0.1 × 10 -4 cm 2 s −1 and using Fick's first law of diffusion, a CH 4 concentration of 0.02 nmol cm −3 gave a diffusive flux of -0.007 µmol m −2 hr −1 . This flux was negligible compared to CH 4 emissions captured in the water column and was therefore excluded from subsequent GHG accounting.
Average  (Table 3; Fig. 6). www.nature.com/scientificreports www.nature.com/scientificreports/ We did not find a significant difference between average C inorg concentrations by paired depth horizon in bare and seagrass sediment cores (t = -0.287, df = 13, p > 0.389). Inorganic carbon concentrations in the top 12 cm of the bed were similar throughout the meadow (site n = 16), averaging 0.11 ± 0.04 mg C inorg cm −3 . Scaling our average concentration from the top 6 cm of the bare site, 0.13 ± 0.04 mg C inorg cm −3 , by meadow area gave estimated CO 2 emissions from CaCO 3 formation of 450 t CO 2 in 2013 and 623 t CO 2 by 2016. However, the absence of a significant difference in CaCO 3 between bare and seagrass sites meant that there was no net CO 2 evasion attributable to the seagrass restoration (Table 3), so seagrass-enhanced CO 2 evasion from CaCO 3 between 2001-2016 was zero.
Integrating both stock changes and fluxes, this seagrass meadow restoration generated a net GHG benefit, which increased from 0.21 t C ha −1 yr −1 between 2001-2013 to 0.42 t C ha −1 yr −1 from 2013-2016, 12-15 years after restoration started.

Discussion
By applying the VCS GHG accounting methodology for the first time to an actual seagrass restoration project 24 , this study confirms the generally accepted but essentially untested hypothesis that seagrass restoration results in net GHG removal from the atmosphere-a GHG offset benefit that can potentially finance restoration. We also found that seagrass presence increased both CH 4 and N 2 O release, but these increases had a relatively small effect on the net GHG benefit. Although other studies have reported increases in gross seagrass bed sediment carbon concentrations following seagrass restoration (e.g. 9,12 ,), these reports do not translate directly to an offset benefit 18 . As we demonstrate in this study, gross C org stocks determined in previous studies overestimate the GHG offset benefit, because they do not account for background C org sequestration that would occur in the absence of seagrass or GHG flux increases due to meadow restoration. All of these parameters must be known to determine the GHG offset benefit provided by seagrass restoration. This study also demonstrates the utility of the stock change approach for seagrass GHG offset accounting and addresses questions about stock change feasibility 18 . We observed considerable variability in CH 4 and N 2 O fluxes at seagrass sites, especially during spring and summer months. More work is needed to understand site-specific drivers of CH 4 and N 2 O production to better constrain annual fluxes 48 . This includes determining whether CH 4 production varies with sediment C org concentrations, whether CH 4 and CO 2 interactions affect CH 4 release, and whether microbial community differences affect CH 4 and N 2 O enhancement. We also note that using benthic chambers may have moderated release rates for both trace gases by inhibiting flow-induced efflux and that using experimentally cleared control sites, rather than bare sites outside the meadow, may have reduced the apparent seagrass enhancement effect. We advise other seagrass blue carbon studies to measure both trace gases directly, until a sufficient number of additional studies suggest conservative release rates for seagrass GHG accounting that are generally applicable.

Seagrass-effects on CH
Identifying the net GHG benefit from seagrass restoration. Studies based on burial rates have suggested that seagrass meadows may sequester more carbon in soils than terrestrial forests 55 . The net sequestration rate based on sediment and plant stock changes and emissions of CH 4 and N 2 O that we measured in this study, 0.42 t C ha −1 yr −1 , is lower than the average rates for temperate and tropical forests, 2.6 and 5.3 t C ha −1 yr −1 , respectively 51 , but generally agrees with the IPCC sequestration rate for seagrass systems, 0.43 t C ha −1 yr −1 51 . www.nature.com/scientificreports www.nature.com/scientificreports/ Similar studies in other systems may also support the use of this default factor, but we note several reasons why this default factor may not be an appropriate rate for all seagrass systems at all times. First, the IPCC rate is double the sequestration rate that we calculated for the first decade of our restoration, 0.21 t C ha −1 yr −1 . Long-term research in this restored meadow has shown that it took about a decade for sediment carbon sequestration rates and plant biomass to be equivalent to natural meadows 12 . Second, sediment accretion may vary throughout the meadow. We assumed uniform sediment accretion, but actual accretion may be lower near the meadow edge, as evidenced by the grain size distribution and reported Reynolds stresses 56,57 (see the Supplement). Third, our system has negligible carbonate, because the sediment in the region is siliciclastic, and there are no nearby coral reefs. We did not expect to find a significant difference in CaCO 3 at seagrass and bare sites. This is not the case in other seagrass systems, where increased CO 2 evasion may be significant (see the comparison between this system and others in Sadrene et al. 22 ). Finally, the South Bay meadow also appears to be metabolically balanced on a decadal time scale, but studies in autotrophic systems may need to determine whether direct plant metabolism increases pCO 2 and results in a CO 2 flux back to the atmosphere 58,59 . These caveats point to areas where future research needs to be done to verify how generally the IPCC default factor applies to seagrass ecosystems worldwide.
The stock change approach indicates that the carbon sequestration rate for this meadow is increasing but that net CO 2 sequestration as a percentage of meadow-wide community GPP may be declining with meadow age. www.nature.com/scientificreports www.nature.com/scientificreports/ Cumulative GPP increased by 114% between 2013 and 2016, due largely to meadow expansion, but the enhanced sequestered stock only increased by 78% over this period. The fraction of GPP that is sequestered may increase over time if the meadow stops expanding and GPP reaches a long-term steady state. Recent work at this site has shown that GPP initially exceeded respiration in this meadow but later reached equivalence 59,60 , a finding that may pertain to eelgrass systems generally 61 . Studies need to determine whether carbon sequestration as a percentage of GPP changes over time in other systems, including those that appear to be net autotrophic 30 , and whether seagrass offset benefits continue to accumulate indefinitely.  and CaCO 3 gas exchange/reaction ratio may vary; we used 0.6, as discussed in the methods 26 . c Note that we did not observe seagrass-enhanced CaCO 3 burial in this system. Given that measuring sediment C org stock changes in a seagrass system is feasible, we recommend using this method to calculate seagrass net GHG benefits to avoid issues associated with using burial fluxes for this purpose 18 . Use of 210 Pb dating to calculate sedimentation rates in seagrass systems has been criticized where relatively short (decadal) time scales are addressed and where bioturbation could disturb sediment profiles 18 . A recent study used surface elevation tables (SET) to compare changes in surface elevation between bare and seagrass sites over short (<1 yr) time scales 62 , but the SETs and marker horizons used widely in salt marshes are generally problematic in seagrass meadows. Subtidal currents re-suspend surface sediments, scouring occurs around vertical objects, including SET pins, and the high-water content of surface sediment makes precise (mm-scale) measurements of surface elevation difficult 63 . Burial rate sequestration estimates also assume that surface deposition is the primary vector for transferring C org to the sediment, but we observed considerable C org accumulation within the bed. This may be due to sediment C org accumulation from root C org exudates or from increased preservation of benthic microalgae migrating up and down within the sediment 64 . The sediment C org stock increase that we observed, 874 t C org , exceeded the increase we would have estimated by scaling the Greiner et al. 12 surface burial flux reported for this system by meadow area and by the three-year time period, 755 t C org . However, we also observed sediment C org declines in 2016 at particular sites, which affected the 2016 sediment C org spatial distribution (Fig. 3). Random disturbance events will likely affect long-term (i.e. decadal) sediment C org accumulation rates by periodically removing sequestered sediment C org stocks. A stock change approach captures these changes. Burial flux rates derived from dated sediment cores may need to be reconsidered, given the magnitude of the within bed C org accumulation that we observed.
Individual seagrass projects should also take care to avoid overestimating the GHG offset benefit by failing to account for allochthonous C org . The VCS carbon-offset protocol conservatively requires that carbon fixed outside the project area (allochthonous carbon) be excluded from the GHG offset benefit, because this cannot be unequivocally attributed to the seagrass restoration project 18,24 . We conservatively deducted the background C org concentration from the entire seagrass C org profile to account for possible deposition of allochthonous carbon (see Fig. 1 Table 4 for log-likelihood ratio test results for assessing the treatment effect.  www.nature.com/scientificreports www.nature.com/scientificreports/ Offset-credit finance as an incentive for seagrass restoration. Had this restoration project been able to apply for VCS offset-credits in 2001, it would now receive up to 8,630 credits. The actual allocation of credits would be slightly lower to account for CO 2 emissions from project activities (i.e. travel to restoration sites, etc.) and 'buffer pool' set aside credits to account for the risk of GHG offset gain reversals 24 . Investors do not typically consider GHG offset projects viable unless they sequester at least 50,000 tCO 2 e over the project lifetime (typically 30 years) 65 . Reaching 50,000 credits by 2031 would require a further increase in the C sequestration rate by this meadow. Future work, including repeated carbon stock change measurements and bed accretion measurements, will be necessary to determine whether the sequestration rate continues to increase.
Given current market prices, carbon offset-credits currently provide a marginal incentive for seagrass restoration. At a price of $10 ton −1 , offset-credits would finance approximately 10% of the approximately $800 K South Bay restoration cost 17,66 . Fully financing a seagrass restoration project with a unit cost equivalent to this South Bay Z. marina restoration would require a voluntary offset price greater than $95 per MtCO 2 e. This cost-benefit comparison excludes project development costs, which may exceed $100 K, and net present value discounting. We note that the carbon burial rates measured in South Bay are on the low end of those documented for other seagrass meadows globally 5 . Other species and locations may generate larger sediment C org stocks than we measured for Z. marina over time (e.g. 67) . However, the South Bay restoration was accomplished at a unit cost of only $1,200 ha −1 17 , and the range for other seagrass projects is $1,900-4,000,000 ha −1 54 .
Rather than rely solely on carbon offset-credits to finance meadow restoration, coastal managers should think holistically about the other values that seagrass systems provide, including fisheries support, nutrient removal, and reduced marsh erosion, among other services. Quantifying these values, even absent markets for co-benefit 'credits, ' would provide further incentive for seagrass restoration, in addition to carbon sequestration.

Methods
Study area. We measured all of the parameters required by the VCS methodology to quantify the GHG offset benefit from the Z. marina restoration in South Bay, VA 24 . The restoration history 68 , project cost 17 , sediment C org stock enhancement 12,57,69 , and net ecosystem metabolism 58-60 of this meadow have been documented and provide a baseline for stock-change assessment. The South Bay meadow area is shallow, with an average depth at mean sea level of 0.76 ± 0.28 (SD) m, and oligotrophic, with low nutrient loading (Fig. 2) 57 . For additional background on Sediment C org stock enhancement. Meadow sediment C org stock enhancement was determined for both 2013 and 2016 by subtracting baseline sediment (i.e., bare) C org stocks from the gross stocks measured within the meadow (Fig. 1). C org is generally present in subtidal sediment without seagrass meadows, and this background C org should not be attributed to a seagrass restoration project. The restored meadow was already in existence when we began sampling in 2013, so time = 0 values at sites within the meadow were not available. The sediment C org baseline scenario (the Emmer et al. 24 'without project' scenario) that would represent pre-restoration (time = 0) was, therefore, established by measuring C org concentrations at bare control sites outside the meadow. The average C org concentration in cores collected at four bare sites by Greiner et al. 12 in 2011 and by Oreska et al. 57,64 in 2013 and in 2014 was 3.67 ± 0.55 (SE) mg C org cm −3 (see Supplement). We verified that this background concentration remained unchanged by collecting new, replicate cores (n = 4) at two of these bare sites in 2016. We deducted this average background sediment C org concentration from the sediment C org concentrations measured within the meadow in 2013 and in 2016 to identify the C org attributable to the seagrass restoration (Fig. 1). This is in accordance with the stock change assessment recommended by the VCS methodology 24 .
We assessed C org changes at sites within the meadow in 2016 by resampling 16 randomly-selected meadow sites first sampled by Oreska et al. 57 in 2013 (the 'with project' scenario). Four 12-cm long, 2.7 cm diameter cores were collected at each site and subdivided into 3-cm intervals. Macroscopic roots and rhizomes were removed from each sample manually, using tweezers. Note that belowground biomass (BGB) was quantified separately, as described in the following section, to avoid double counting. All sediment samples were prepared according to methods used previously in this system 12,57,64 . We measured %C on a Thermo Scientific Flash 2000 Organic Element Analyzer; %C org was determined by subtracting %C inorg , which we determined using element analysis of samples ashed at 500 °C for six hours 71 . The element analyzer average percent error was 0.48%, based on analysis of lab standards.
Allochthonous C org may be deposited within the bed due to bed accretion (Fig. 1). Rather than deduct an arbitrary 'allochthonous compensation factor' from the meadow sediment C org stock 72,73 , we accounted for allochthonous C org that could have been deposited in the baseline scenario by deducting the bare site sediment C org average from the entire meadow carbon profile, including the part of the sediment profile that may have resulted from accretion facilitated by the meadow (see Fig. 1 and the Supplement for more explanation).
Total, meadow-enhanced sediment C org stocks in 2013 and in 2016 were quantified by interpolating the average 2013 and 2016 sediment C org enhancement at each site in ArcGIS 10.2 Geostatistical Analyst using Ordinary Kriging 74 . We fitted stable, circular, spherical, Guassian, and exponential semivariogram models to each dataset and selected the sediment C org distribution maps with the lowest root mean square errors (Supplement). The 2013 Figure 6. Cumulative background (A) and gross meadow (B) GHG stocks in the meadow areas over time; sequestration (i.e., GHG uptake from the atmosphere) in this figure is shown as positive, GHG release (i.e., a GHG flux to the atmosphere) is negative; CH 4 and N 2 O quantities were standardized to CO 2 e; 'CaCO 3 ' relates CO 2 evasion attributable to CaCO 3 ; background stocks were calculated by scaling average bare site values by total meadow area at each time step; net stock enhancement attributable to the meadow (see Table 3 www.nature.com/scientificreports www.nature.com/scientificreports/ data was best fit using a circular model, the equivalent 2016 data was best fit using a Gaussian model, and the uppermost 2-cm interval in 2016, which may be the result of accretion and is shown separately in Fig. 3, was best fit using an exponential model. Biomass CO 2 sequestration. The carbon sequestered in seagrass tissue is periodically lost to export, herbivory, and decomposition, so we calculated and reported the average, annual standing biomass stock based on seasonal measurements from 2014-2016 (see Supplement). This represents a running average that reflects periodic export and other fluctuations, rather than peak observed biomass. This is the same general approach that reforestation GHG offset projects use to address the cyclical harvest and replanting of aboveground biomass (AGB), and it is permitted for seagrass GHG accounting 24 . Shoot densities ranged from approximately 250 to 617 shoots m −2 in South Bay due to seasonal thinning and export, and biomass ranged from 0.26 to 0.781 gdw shoot −1 . We accounted for variability in AGB using existing density measurements (shoots m −2 ) taken at sites throughout this meadow over time to account for seasonal changes 57,75 . The average density over the course of a year was approximately half of the peak density observed during July (48%) 57,75 We quantified average AGB per shoot and BGB by collecting additional replicate (n ≥ 4) 15.2-cm diameter biomass cores seasonally from June 2014 to June 2016 to a depth of 15 cm at five central meadow sites (see Supplement), following methods employed by past studies in this system 69,76 . We also collected biomass cores (n ≥ 3) at four additional, systematically located sites during the summer of 2016 (see Supplement). Samples were sieved using a 1-mm mesh, separated the same day into live and dead fractions, and then dried to a constant weight at 60 °C. Biomass data-both live and dead-was averaged by site and then by month to generate seasonal averages, which were used to calculate the average, annual standing stocks. The average, annual shoot densities were multiplied by the average biomass shoot −1 , 0.41 ± 0.09 gdw shoot −1 (this study), and by 37.1% C gdw −1 biomass 76 . The resulting aboveground biomass values (C org m −2 ) were interpolated using Ordinary Kriging in ArcGIS 10.2 Geostatistical Analyst and a Gaussian semivariogram to generate average, annual AGB stocks for the 2013 and 2016 meadow extents. Average live and dead BGB values (g m −2 ) were multiplied by the average C org fraction in belowground biomass, 33.8% C org gdw −1 biomass 77 , and scaled by the 2013 and 2016 meadow areas to generate C org stocks. GHG fluxes. We deployed clear plastic, bell-shaped benthic chambers over vegetated and experimentally cleared 2 m x 2 m bare plots at the five central meadow sites to identify changes in benthic CH 4 and N 2 O fluxes attributable to Z. marina presence. Each chamber sat on the sediment surface, covering a 0.046 m 2 circular area and enclosing a 10.5 L volume. Comparing fluxes at cleared, central meadow plots allowed us to control for confounding factors at bare sites outside of the meadow. These areas are generally deeper with more sand-sized sediment and experience greater Reynolds stresses, because of area geomorphology 56 , factors that may affect sediment:water gas exchange. We cleared the bare plots during spring 2015, installed plastic lawn edging to a depth of 8 cm to prevent seagrass rhizome re-colonization, and allowed plots to equilibrate for five months. Comparing seagrass and cleared bare plots to assess a seagrass enhancement effect on CH 4 and N 2 O was conservative, because some seagrass BGB potentially remained at the cleared plots and may have contributed to microbial production of these trace gases. Eight chambers were deployed at each site during each observation, four replicates over seagrass and four over bare sediment. Every deployment exactly bracketed low tide, such that gas accumulation time captured equal parts falling-and rising-tide. Deployment durations ranged from 1 to 5 hours. Trace gases were collected on multiple days per month in October 2015, April 2016, June 2016, July 2016, August 2016, and October 2016. Using chambers allowed us to conduct a controlled experiment in situ to test for a seagrass presence effect, but we acknowledged that using benthic chambers may have introduced container effects that affected release rates, including the elimination of hydrodynamic flow-induced efflux.
The gas that collected in each chamber was syringe extracted and injected into an exetainer filled with 12 ml N 2 and 0.2 ml 0.01 M ZnC 4 H 6 O 4 to prevent microbial activity resulting from the syringe transfer. The total gas volume collected within each chamber was noted and used to calculate the gas flux as a function of time and bed surface area. We also measured bulk CH 4 concentrations in replicate porewater samples collected at bare and vegetated sites in August (site n = 6) and October (site n = 4) 2016 to determine the magnitude of the diffusion flux relative to the ebullition flux. We extracted 7 ml of porewater through mini-piezometers (inner diameter 1.8 mm) at 3-cm intervals, from 1.5 cm down to 13.5 cm. The water samples were syringe injected into exetainers filled with 12 ml N 2 and fixed with 0.2 ml ZnCl 2 . The diffusive flux was calculated using Fick's first law of diffusion: where the sediment diffusivity, Ds, was assumed to be 0.1 × 10 -4 cm 2 s −1 .
All exetainer samples were analyzed on a Varian 450-Gas Chromatograph with a Bruker GC/MS workstation at the Smithsonian Environmental Research Center. We determined sample CH 4 and N 2 O concentrations using onsite standards and corrected for differences in atmospheric temperature and pressure during each GC analysis. Standard curve R 2 values ranged from 0.992 to 0.996.
We tested for an effect of seagrass presence on CH 4 and N 2 O fluxes using linear mixed effect models in R 78,79 . Replicate results were averaged by site. Seagrass presence/absence and month were treated as fixed effects; individual sites were randomly selected. Tests were run on each GHG dataset using the lmer function (lme4 package version 1. [1][2][3][4][5][6][7][8][9][10][11][12][13][14]. We expected to find increased GHG fluxes attributable to seagrass presence, as well as a seagrass*month interaction effect. Both the CH 4 and N 2 O datasets required transformation due to heteroskedasticity and the presence of outliers. The optimal transformation (identified using the optim.boxcox function in the boxcoxmix package version 0.14) for the averaged data was λ = 0.133 (Maximum log-likelihood = −77.608). Model p-values were obtained from likelihood ratio tests on the full model and a reduced model without the fixed effects.
Net GHG benefit accounting. Total meadow CO 2 sequestration was calculated for both 2013 and 2016 by summing the above-and belowground biomass (both live and dead) and meadow-enhanced sediment C org stocks measured in each year. Cumulative, enhanced CH 4 and N 2 O emissions attributable to the meadow were estimated by multiplying the average enhanced (i.e., net) fluxes (g m −2 yr −1 ) by meadow area over time. Meadow area changes were calculated in ArcGIS 10.2 by georeferencing the Virginia Institute of Marine Science aerial photographs for every year after initial reseeding in 2001 and delineating the meadow perimeter 74,80 . Meadow area was interpolated for the three years where photographs were unavailable. These cumulative, net GHG emissions calculated for 2013 and for 2016 were subtracted from the respective meadow-enhanced CO 2 sequestration results to determine the net GHG benefit in each year (note that seagrass-enhanced CO 2 emissions from CaCO 3 were not observed).
We compared the total meadow sequestration in 2013 and in 2016 with the total, cumulative GPP within the meadow in each of those years to estimate the percentage of total GPP sequestered by the meadow. Cumulative GPP was estimated as a function of shoot density and meadow area. The relationship between meadow age and density was determined by fitting a polynomial regression to existing data from this meadow collected as part of the annual VCR-LTER seagrass survey 81 . This relationship was observed by Berger et al. 59 to be: 3 2 where Y was shoot density in shoots m −2 , and x was the meadow age in years (R 2 = 0.91). GPP was calculated using the following relationship observed in this meadow by Berger et al. 59 : where x was density (shoots m −2 ) and Y was GPP in mmol O 2 m −2 d −1 (R 2 = 0.69). Calculated GPP values for meadow areas of different age were summed and integrated over time to generate cumulative values.

Data availability
Data reported and analyzed in this study is available in the Supplement and on the LTER Network Data Portal (https://portal.lternet.edu/nis/home.jsp).