Rapid carbon accumulation at a saltmarsh restored by managed realignment exceeded carbon emitted in direct site construction

Increasing attention is being paid to the carbon sequestration and storage services provided by coastal blue carbon ecosystems such as saltmarshes. Sites restored by managed realignment, where existing sea walls are breached to reinstate tidal inundation to the land behind, have considerable potential to accumulate carbon through deposition of sediment brought in by the tide and burial of vegetation in the site. While this potential has been recognised, it is not yet a common motivating factor for saltmarsh restoration, partly due to uncertainties about the rate of carbon accumulation and how this balances against the greenhouse gases emitted during site construction. We use a combination of field measurements over four years and remote sensing to quantify carbon accumulation at a large managed realignment site, Steart Marshes, UK. Sediment accumulated rapidly at Steart Marshes (mean of 75 mm yr-1) and had a high carbon content (4.4% total carbon, 2.2% total organic carbon), resulting in carbon accumulation of 36.6 t ha-1 yr-1 total carbon (19.4 t ha-1 yr-1 total organic carbon). This rate of carbon accumulation is an order of magnitude higher than reported in many other restored saltmarshes, and is somewhat higher than values previously reported from another hypertidal system (Bay of Fundy, Canada). The estimated carbon emissions associated with the construction of the site were ~2–4% of the observed carbon accumulation during the study period, supporting the view that managed realignment projects in such settings may have significant carbon accumulation benefits. However, uncertainties such as the origin of carbon (allochthonous or autochthonous) and changes in gas fluxes need to be resolved to move towards a full carbon budget for saltmarsh restoration.

Yes -all data are fully available without restriction potential to accumulate carbon through deposition of sediment brought in by the tide and burial of 21 vegetation in the site. While this potential has been recognised, it is not yet a common motivating 22 factor for saltmarsh restoration, partly due to uncertainties about the rate of carbon accumulation 23 and how this balances against the greenhouse gases emitted during site construction. We use a 24 combination of field measurements over four years and remote sensing to quantify carbon 25 accumulation at a large managed realignment site, Steart Marshes, UK. Sediment accumulated 26 rapidly at Steart Marshes (mean of 75 mm yr -1 ) and had a high carbon content (4.4% total carbon, 27 2.2% total organic carbon), resulting in carbon accumulation of 36.6 t ha -1 yr -1 total carbon (19.4 t ha -28 1 yr -1 total organic carbon). This rate of carbon accumulation is an order of magnitude higher than 29 reported in many other restored saltmarshes, and is higher although more similar to values 30 previously reported from another hypertidal system (Bay of Fundy, Canada). The estimated carbon 31 emissions associated with the construction of the site were ~2-4% of the observed carbon 32

Introduction 38
Earth's ecosystems exhibit overall net carbon uptake, causing increases in atmospheric CO2 to be 39 smaller than expected from fossil emissions and land-use change [1]. They also contain substantial 40 carbon stocks, which currently store carbon out of the atmosphere but are sensitive to changes in 41 climate or land-use [2, 3]. Coastal 'blue carbon' ecosystems, including saltmarshes, are especially 42 carbon dense and sequester carbon at a rate an order of magnitude faster than terrestrial 43 ecosystems [4]. Both allochthonous and autochthonous carbon is sequestered in saltmarshes; 44 carbon accumulates in saltmarshes as sediment carried in by the tide is deposited, and this sediment 45 buries saltmarsh plant remains. Globally, the ~5.5 million hectares of saltmarshes [5] are estimated 46 to accumulate carbon at an average rate of ~2.4 t C ha -1 yr -1 [6]. Despite their importance, ~50% of 47 saltmarsh area has been lost, particularly through reclamation for agriculture or urbanisation, or 48 degraded [7], with annual losses of 1-2% [8,9]. 49 In response to losses of saltmarsh and its associated biodiversity, 'no net loss' policies have sought 50 to protect remaining wetlands and create new habitat [10], contributing to over 100,000 ha of 51 intertidal wetland creation over the last 30 years [11]. However, the pace of global wetland creation 52 is not sufficient to offset losses, where a key barrier is the availability of project financing [12]. 53 Payments for ecosystem services, such as flood protection or biodiversity, offer potential financial 54 mechanisms for saltmarsh creation or restoration [13]. Carbon accumulation (and thus climate 55 mitigation) has been recognised as a potential benefit of saltmarsh restoration, and could therefore 56 provide a further motivation for site creation [14,15]. 57 Robustly quantifying the rate of carbon accumulation on restored saltmarshes will be necessary if 58 carbon finance mechanisms are to be developed [16] and is also important to enable saltmarsh 59 restoration to be properly included in national carbon budgets [17]. Furthermore, rising sea levels 60 threaten existing saltmarshes, and the climate sensitivity of their carbon stocks and fluxes needs to 61 be quantified [18]. While saltmarsh restoration could potentially compensate for loss of natural 62 saltmarshes, given known differences in topography and ecology [19,20], it may not be appropriate 63 to assume that restored or created marshes will ultimately store carbon at a rate comparable to 64 natural saltmarshes [21]. Thus it is also important to determine any differences between carbon 65 accumulation and sequestration in natural and restored saltmarsh. 66 Previous attempts to quantify actual or potential carbon accumulation following saltmarsh 67 restoration have used a variety of techniques: (a) spatially explicit models to predict landscape-scale 68 carbon accumulation based on observed carbon accumulation in natural habitats [22]; (b) 69 measurements at a single time-point to take a snapshot of carbon stocks [23]; (c) restored saltmarshes of different ages as a space-for-time substitution to estimate the rate of carbon 71 accumulation [24]; and (d) repeat measurements of the elevation of sediment surface to quantify 72 sediment deposition rates [25]. While all approaches highlight the potential for saltmarsh 73 restoration to lead to carbon accumulation, each has limitations when used in isolation. A further 74 challenge is that previous studies have either assessed only total carbon (which does not distinguish 75 organic carbon from inorganic carbon such as biogenic or lithogenic carbonates), or have quantified 76 organic carbon using loss on ignition, which is known to have poor accuracy and large uncertainties 77 [26]. 78 A further consideration when evaluating the net carbon benefit of a saltmarsh restoration or 79 creation project is the balance between the carbon costs of constructing the site (e.g. building new 80 flood defences inland and breaching the existing embankments, termed "managed realignment") 81 and the carbon accumulation provided by the site [e.g. 27]. If project carbon costs are high relative 82 to the rate of carbon accumulation, it may take years for the site to pay off the debt of construction 83 [28]. 84 This research aims to evaluate carbon costs and benefits from saltmarsh creation through managed 85 realignment, using a novel combination of techniques. Over the course of several annual cycles we 86 use remote sensing, field measurements and robust laboratory techniques to quantify total and 87 organic carbon accumulation in an evolving saltmarsh in the first years after restoration. This allows 88 us to reliably quantify the amount and rate of carbon accumulation following restoration. We then 89 assess the carbon emissions incurred during site construction before identifying additional 90 requirements for producing a full carbon budget for saltmarsh restoration. 91 92

Study site 94
Steart Marshes (Somerset,UK;51.20 N,3.05 W) is a 250-ha managed realignment site, forming part 95 of a larger 400 ha complex of restored wetland habitats managed by the Wildfowl and Wetlands 96 Trust. It was constructed to create new intertidal habitat in compensation for previous losses, and to 97 provide enhanced flood defences [29]. Prior to site construction, the land was under a mix of 98 agricultural uses, including permanent pasture (i.e. pasture had been the land use over many years), 99 grass ley (part of cyclical arable land management) and arable (winter wheat, barley, oilseed rape 100 and maize) (Fig. 1a). The site lies near the mouth of the River Parrett which drains a catchment of 101 interbedded limestone and mudstone [30] and flows into the Severn Estuary. Hydrodynamic processes in the Parrett are dominated by a large tidal range which gives rise to strong tidal flows 103 and large intertidal areas. At Hinkley, just to the west of the Parrett Estuary mouth, the mean spring 104 tides have a high water height of 5.6 mODN and a low water height of -5.1 mODN, giving a range of 105 approximately 11m [31,32]. 106 The construction of the managed realignment site started in early 2012, comprising the excavation 107 of a creek network and pools, the construction of new flood defence embankments and the raising 108 of a small length of existing embankment. The creek network (7.6 km total length) was designed to 109 meet the geomorphological requirements of the scheme (see [33] Table S1 for full 138 details). Cores of 30-50cm were collected using a soil auger and sectioned into 5-10 cm lengths for 139 later analysis. In total, we collected 78 cores, resulting in 596 samples. Depending on site conditions, 140 surface silts deposited post-breach were sometimes difficult to sample using a soil auger as they 141 were either prone to compression or highly friable. In these cases, we collected undisturbed surface 142 samples using adapted syringe tubes and/or collected the full surface sediment plates (down to base 143 of mud cracks), and then sampled deeper sediments by taking a core between mud cracks. The 144 horizon between the deposited silts and the underlying agricultural soils was determined through 145 visual inspection of the cores (prior to sub-sampling) and the depth (in core and from surface) was 146 recorded. The horizon was readily identifiable through a change in colour and texture of the soils, 147 and by the presence of remnant vegetation and roots. Samples with a defined volume of 5 cm 3 were 148 taken from the above-horizon section of the core or directly from surface sediments for dry bulk 149 density measurements [34]. 150 All samples were stored at 4°C prior to analysis. Dry bulk density was determined by drying the 151 samples of a known volume to a constant weight at 105°C. The remaining core samples were dried 152 at 60°C, covered, in aluminium trays/glass jars for approximately 96 hours, then ground using a 153 pestle and mortar to ensure a homogeneous sample for further analysis. from UK National Tide Gauge Network)). We obtained LiDAR DTMs for eleven further time points 189 after breaching (see Table 1). Downloaded DTMs were processed in Rv4.02 [37] using the "raster" 190 package [38]. Tiles were merged before being clipped by the site area. The site area was defined by 191 manually drawing a polygon around the crest of the flood embankment to remove areas outside of 192 the site. We then restricted analyses to locations subject to tidal inundation which were taken to be 193 those areas below 7.07 m ODN, which is the level of the highest astronomical tides at the nearest 194 port, Burnham-on-Sea [32]. The first DTM available after the breach (31 October 2014) was clipped 195 to locations below 7.07 m and the resulting polygon (with an area of 244.7 ha) used to clip the 196 remaining DTMs. 197 Filtered DTM data should represent the ground elevations, but filtering does not completely remove 198 dense, relatively short vegetation. Vegetation cover at the site in the first three years was sparse 199 (Fig. S1, H Mossman pers. obs.) and so we do not consider this an issue for those years; in the latest 200 year, vegetation cover was denser and extensive, but unvegetated areas remained (H Mossman 201 pers. obs.). The 50 cm resolution cannot account for surface morphology smaller than this (e.g. 202 surface desiccation cracking, which was observed during summer months). We also observed 203 sediment dewatering and shrinkage during dry periods, but these changes were small compared to 204 interannual changes in elevation (Table 1) (Table S1). We found good agreement between LiDAR and ground-based 220 measurements of sedimentation rates (Fig. S2). 221 222 Data analysis: variation in sediment carbon 223 Variation in sediment carbon composition was assessed as a function of depth using locally weighted 224 polynomial regression (loess function in R), fitted separately for above and below the agricultural 225 soil-new sediment horizon. We assessed whether there was a difference in the percentage carbon in 226 the newly accreted sediment, natural sediment (pooling locations, time points and depths for both) 227 and the pre-restoration soils from the four land uses using Anova with a Tukey HSD post hoc test. 228 Post-restoration samples from at or below the agricultural horizon were not included in this analysis 229 because (1) we were interested in the carbon accumulating after the restoration in the newly 230 accreted sediment and (2) elevated carbon contents were observed due to the burial of remnant 231 agricultural vegetation as opposed to saltmarsh processes. We calculated the mean and standard 232 deviation of TC in newly accreted sediment, and also calculated the mean and standard deviation of 233 the ratio of TOC and TC. We considered it justified to treat the carbon content of new sediment as 234 coming from a single population (i.e. not varying between years) as (1) the TC of new sediment did 235 not vary with depth in cores and (2)  total carbon accumulation (t). This was divided by site area to obtain tC/ha. This calculation was 243 repeated with the additional step of multiplying by the ratio to TOC to TC to estimate site level total 244 organic carbon accumulation. 245 As each stage in this calculation involves measurements made with error, we used Monte-Carlo 246 resampling to estimate site-level carbon accumulation while propagating errors from each step. If 247 elevation measurement errors were independent for each DTM pixel in each time point then errors 248 largely cancel out. A more conservative approach is to assume that measurement errors apply 249 systematically to a survey. We do the latter, and take mean elevation change between surveys as 250 coming from a normal distribution with a mean equal to the measured change in elevation, and a 251 standard deviation of 0.04 m based on measurements of control points [31]. Bulk density of newly 252 accreted sediment was sampled from a normal distribution with mean 1.11 and SD 0.27 t.m -3 . 253 Sediment TC was sampled from a normal distribution with mean 4.37% and SD 0.50%, and the ratio 254 of TOC to TC was sampled from a normal distribution with mean 0.53 and SD 0.08. We took 100,000 255 samples from these distributions to obtain a distribution of carbon accumulation estimates. The change in elevation measured from DTMs was consistent with field measurements of 286 sedimentation (Fig. S2). Comparison of successive DTMs indicated that the net elevation of the site 287 increased over time (Fig. 2), and by September 2018 714,513 m 3 of sediment had accumulated 288 across the site, with an average depth of 0.292 m (Table 1). There was no clear trend in 289 sedimentation rate with time since breach (regression: slope < 0.001, F1,8 = 0.35, P = 0.568), 290 although the most rapid sedimentation was noted immediately following the breach ( Table 1) The bulk density of newly accreted sediment ranged from 0.553 to 1.568 t m -3 (mean = 1.110 ± 0.267 306 SD). There were significant differences in TC between pre-restoration soils of different land uses, 307 newly accreted sediments from within the restoration site and sediments from the existing natural 308 saltmarsh (F5,249=48.7, p<0.001, Fig. 3). Soils collected prior to restoration from all land uses had 309 significantly lower TC than the newly accreted sediment and the natural saltmarsh sediments, with 310 those from the pre-restoration disturbed (A) and arable (D) areas having the lowest carbon contents 311 (Fig. 3). Sediments from the natural saltmarsh had significantly higher TC (4.72 ± 0.58 %) than the 312 newly accreting sediment on the restoration site (4.37 ± 0.50 %). The ratio of TOC to TC was similar 313 in natural saltmarsh and newly accreting sediment on the restoration site (natural = 0.524, restored 314 = 0.529, Fig. 3), giving a TOC of 2.24 ± 0.33% in newly accreted sediment on the restoration site and 315 2.44 ± 0.31 % on the natural saltmarsh. 316 There was some spatial variation in the TC of new sediment (significant difference between sampling 317 sites (F3,44 = 5.1, P = 0.004)) but no difference between years (F1,46 = 0.369, P = 0.547). The TC of 318 newly accreted sediment was consistent with depth (Fig. 4). Some samples taken at the horizon with 319 the underlying agricultural soils had very high carbon content, reflecting the terrestrial vegetation 320 burried by the initial inundations of sediment. Below the horizon, TC was lower than in newly 321 accreted sediment and directly comparable to the pre-breach measurements of the agricultural soils 322 (Fig. 4).  (Table S2). An estimated additional 20% of fuel 337 consumption was assumed for the construction of earthworks in other areas of the Steart Marshes 338 complex, giving total emissions associated with machinery fuel usage of 1,772 tCO2e (483 tC). 339 Combining these figures with the estimated emissions from personnel travel, energy use in portable 340 accomodation, and embodied emissions of construction materials from the Environment Agency 341 carbon calculator, gives estimated total construction emissions of 2,762 tCO2e (753 tC). These 342 emissions are equivalent to ~2% of the of the estimated TC accumulation, or ~4% of the estimated 343 TOC accumulation in the sediments over the 4 year study period, with a carbon payback period on 344 the order of 1 (TC) to 2 (TOC) months. 345

Discussion 347
We find that Steart Marshes managed realignment has rapidly accumulated carbon since the 348 fronting flood defence embankment was breached, and that this carbon accumulation is two orders 349 of magnitude greater than the carbon costs incurred during site construction. The rate of carbon 350 accumulation at Steart Marshes (TC = 36.6 t C.ha -1 .yr -1 , TOC = 19.4 t C.ha -1 .yr -1 ) is considerably higher 351 than has been found at other sites. In the Bay of Fundy, which like the Severn Estuary is hypertidal, 352 carbon accumulation is lower but within the same order of magnitude at 13.29 t C ha -1 yr -1 [25], but 353 rates at other sites are an order of magnitude lower than at Steart Marshes. For example, 354 saltmarshes in eastern England were reported to accumulate carbon at a rate of 1.04 t C ha -1 yr -1 for 355 the first 20 years following creation [24], while a recovering saltmarsh in Australia accumulates at a 356 rate of 0.5 t C ha -1 yr -1 [40]. 357 The rate of carbon accumulation in a restored saltmarsh is a product of the rate of sediment 358 accumulation and the carbon content of that sediment, and we can look at both these elements to 359 see if Steart Marshes is unusual. Steart Marshes has experienced rapid sediment accumulation since 360 it was breached (mean rate of increase in elevation = 75 mm yr -1 , Table 1 The Wash (Freiston Shore) [53,54]. 379 Sediment carbon content at Steart Marshes is higher than in some managed realignment sites, but is 380 close to the range of values reported elsewhere. Comparison with values from other managed 381 realignments indicates bulk density varies from 0.74 -1.4 t m -3 [24, 55-57] (cf 1.1 in this study) and 382 carbon content varies from 1.8-4.23% (cf TC 4.4% and TOC 2.2% in this study). Combining all 383 combinations of sediment carbon content and accretion rates gives the space of potential carbon 384 accumulation rates in saltmarsh restored by managed realignment (Fig. 5). This indicates that Steart 385 Marshes has high rates of carbon accumulation because it experiences both high rates of accretion 386 and has relatively high sediment carbon content; thus while neither variable is exceptionally high 387 If the high rate of carbon accumulation at Steart Marshes is unusual, is the conclusion that 399 saltmarshes restored by managed realignment rapidly pay off their carbon construction costs 400 applicable to other sites? We can evaluate this by mapping our estimates of construction carbon 401 costs onto the potential carbon accumulation space (Fig. 5). Most combinations of accretion rate 402 and sediment carbon content would pay off construction costs of a 250 ha site within a year, and 403 even a site with low carbon accumulation rates (Tollesbury managed realignment, eastern England) 404 would be close to breaking even with its carbon construction costs over one year. 405 Our analysis assumes that soil properties (soil carbon, bulk density) come from a single statistical 406 population across the site and over time. However, there were small differences in the carbon 407 content of new sediment across the site. The reasons for this are unclear, but could relate to spatial 408 variation in algal films and vegetation establishment across the site. If the drivers of variation in 409 sediment carbon across the site were known this could be used to scale-up and refine estimates, but 410 this is not currently possible. The bulk density of sediment would be expected to exhibit temporal 411 variation, with lower bulk density (but greater sediment volume) when sediment is waterlogged (e.g. winter, spring), and higher bulk density (but lower sediment volume) when sediment is dry (e.g. 413 summer, early autumn). Our bulk density measurements come from spring and summer, so should 414 capture this temporal variation in bulk density. However, explicitly quantifying temporal variation in 415 bulk density would allow temporal coupling with sediment accumulation data and thus refined 416 quantification of intra-annual variation in carbon accumulation -apparent reductions in carbon 417 stocks over the summer when sediment volume reduced may not occur in reality because of a 418 concurrent increase in sediment bulk density. 419 420 Future changes in carbon accumulation 421 Although we found the fastest rates of accretion shortly following breaching, we did not find a 422 statistically significant reduction in accretion rates. However, a reduction in accretion rates would be 423 expected as the saltmarsh develops. This is because accretion rates tend to be faster at lower 424 elevations which experience more frequent tidal inundation [60], and as these lower areas increase 425 in elevation they experience fewer inundations, and thus slower accretion. Indeed, space-for-time 426 substitutions indicate that carbon accumulation rates slow over time [24]. It is likely that carbon Our results indicate that carbon accumulation at Steart Marshes greatly exceeds construction costs. 439 However, there are a number of uncertanties that, while highly likely not to affect this qualitative 440 conclusion, would need to be considered to refine the quantitive carbon budget ( Some assumptions, such as assuming the carbon content of sediment lost is the same as sediment 442 gained, are likely to mean our estimate of carbon accumulation is conservative (Table 2). Others, such as not accounting for greenhouse gas emssions following site flooding, will offset some of the 444 carbon accumulation benefit of the site ( carbonates should clearly be excluded from estimates of carbon storage (as they comprise fossil C 459 that has not been in recent contact with the atmosphere), the treatment of biogenic carbonates is 460 more complex. Carbonate production is a source of CO2 to the atmosphere, while carbonate 461 dissolution is a sink [65], where the relative balance of these processes and the impact on ecosystem 462 carbon budgets (which also depends on whether the carbonate is imported or produced in-situ, and 463 the role of carbonates in stabilising organic carbon stores) remains a significant uncertainty in blue 464 carbon science [18]. 465

Conclusions 466
Our results show that at Steart Marshes fast rates of sediment accumulation and high sediment 467 carbon content combine to result in exceptionally fast carbon accumulation rates. Carbon 468 accumulation at Steart Marshes over the first four years following reinstatement of tidal flow is two 469 orders of magnitude larger than the carbon costs of site construction. Thus qualitatively, it is clear 470 that the creation of the site by managed realignment has delivered benefits for carbon storage and 471 sequestration, and other sites with lower carbon accumulation rates are likely to rapidly pay off their 472 construction carbon debt. However, there are numerous uncertainties that would need to be 473 resolved in order to move to a fully quantitative carbon budget for restored saltmarshes. 474

Acknowledgements 476
We thank the Wildfowl and Wetlands Trust, particularly Alys Laver and Tim McGrath, for access to 477 the site and their ongoing enthusiasm and support. We thank Grace Biddle, Colin Hill and David 478 McKendry for their work in the laboratory. This study uses data from UK National Tide Gauge 479 Network, owned and operated by the Environment Agency, and provided by the British 480 Oceanographic Data Centre. 481 Table S1. Field sampling dates and information. Samples highlighted in bold are those selected for 663 the quantification of total organic carbon. Access issues prevented sampling at some locations in 664 March 2015 and September 2016. We assessed the consequence of the additional uneven sampling 665 by removing samples from these two time periods and recalculating the mean carbon content in 666 newly accreted sediment. The value differed by less than 1% of the original value (i.e. 4.367% vs 667 4.372%), so we retain all samples in the data presented in the manuscript. 668 Table S2. Summary of the fuel consumption and t.CO2 emitted by construction vehicles in the 669 construction of Steart Marshes. 670