Tidal marsh resilience to sea level rise controlled by vertical accretion and landward migration under nature-based human adaptation scenarios Tidal marsh resilience to sea level rise controlled by vertical accretion and landward migration under nature-based human adaptation scenarios

: Tidal marshes are not only lost to human disturbance but also face the threat of sea 1 level rise (SLR). However, current earth system models used to estimate future changes in 2 wetland extent omit wetland’s real responses to SLR without field observations. We 3 synthesised global data on sediment accretion rate (SAR) and surface elevation change (SEC) 4 for tidal marshes and developed a mathematical model to assess their resilience to future 5 SLR. Sediment loadings and precipitation largely explain the variance of marsh SAR and 6 SEC. Human disturbance resulted in less sediment accretion and existing conservation 7 activities were inefficient in promoting sediment accretion. Under the representative 8 concentration pathways and nature-based human adaptation scenarios, tidal marshes will gain 9 up to 63% of the current area by 2100 if sufficient sediment loadings and accommodation 10 space allow landward migration. If current accommodation space maintains, net areal losses 11 of > 30% are possible and hotspots of future marsh loss are largely in North America, 12 Australia and China. Projections for most SLR scenarios see marsh area peaking in the mid 13 rather than late 21 st century. This implicates that tidal marshes may contribute to achieve a 14 climate neutral world by 2050. We highlight the importance of nature-based adaptation in 15 enhancing the resilience of tidal marshes to future SLR. 16

marsh loss has not been accounted for. Tidal marshes now also face an even more perilous 26 future because of sea level rise (SLR). Loss of tidal marshes to SLR is partly mitigated by 27 vertical sediment accretion, resulting from the balance of below-ground root production and 28 decomposition (Ouyang et al., 2017), and sediment input from marine (Chmura and Hung, 29 2004) and/or riverine sources (Craft 2007). 30 31 Further, there are feedbacks and interactions among plant growth, geomorphology and 32 hydrodynamics that allow tidal marshes to resist submergence due to SLR (Gedan et al.,33 2011; Kirwan and Megonigal, 2013;Marani et al., 2006). Tidal marsh sediments also show 34 different deposition patterns dependent on morphodynamics, which shape tidal marshes 35 through combined ecogeomorphology and hydrodynamics (Friedrichs and Perry, 2001). In 36 addition to sediment accretion, tidal marsh surface elevation change depends on erosion and 37 subsurface processes, including compaction, shrink-swell, subsidence, decomposition and 38 tectonic adjustments (Cahoon et al., 2011; Rogers et al., 2006). Sediment accretion may 39 alleviate the impact of SLR on tidal marshes when surface elevation increases at or exceeds 40 Table 1 A comparison of studies evaluating the response of coastal wetlands to sea level rise  76 Meta-analysis/Quantitative reviews Land building via vertical sediment accretion Lateral landward migration via accommodation space Controls on sediment accretion/surface elevation change

References
Using real data on surface elevation change to assess the response of mangroves to sea level rise

Not considered
Assessed the impact of controls such as climatic, geomorphic, geographic factors on mangrove sediment accretion/surface elevation change Indo-Pacific Lovelock et al. 21 Comparing real data on sediment accretion and relative sea level rise to show marsh submerging or aggrading without considering shallow or deep subsidence Not considered but stated the limitation

Not considered
Global Kirwan et al. 23 Approximate land building using a wetland adaptation score that relies on suspended sediment concentration to describe sediment supply, which is conflated with the former Assessed accommodation space available for wetland landward migration

Not considered
Global Schuerch et al. 24 Using real data on surface elevation change to model the response of tidal marshes to sea level rise

Analysis of controls on SAR and SEC 171
We examined the relationship between SAR/SEC and influential factors, including latitude, 172 precipitation, sea ice concentration, total suspended matter, tidal range and tidal frequencies 173 in a statistical model. The categorical variables are number coded, with one level of each 174 variable selected as the reference. The influences of other factors (e.g. species richness, plant 175 type, location, and elevation) on SAR and/or SEC were not included in the model due to 176 limited available data, which would have substantially reduced the overall degrees of 177 freedom if included. We started the model with multiple regression, but the hypothesis of 178 normality was grossly violated and the residuals were highly heterogeneous. relationship between SEC and SAR, and the relationship between SEC/SAR and total 194 suspended matter, using a linear regression. 195 196 In addition to the above analyses, we examined the difference in SEC/SAR among 197 categorical variables that were significant in the BRT models, and other factors using the 198 Kruskal-Wallis rank sum test. Where significant differences were found, non-parametric 199 Mann-Whitney U tests were used to identify significant differences among the groups. Paired We examined how SAR and SEC are driven by tidal frequencies and other potential drivers 302 that were excluded in the above models owing to limited data but enough for univariate 303 analyses. There are significant differences in both SAR and SEC among tidal marshes 304 inundated at different frequencies ( Fig. 5c and d), and among geomorphologic settings (

The impact of human activities on SAR 355
The impact of different human activities on SAR was compared for sites where conservation 356 and/or human disturbances are reported in conjunction with sediment accretion. SAR are 357 significantly lower in human-disturbed than in undisturbed sites (Fig. 5g, paired t test, t=-358 2.02, p=0.048, df=50), while in-situ conservation (e.g. inclusion of tidal marshes in reserves) 359 had no significant impact on SAR, probably because of the significance of ex-situ drivers, 360 e.g. allochthonous sediment supply. Among sites affected by human disturbances, 361 impoundment is the most frequently reported activity. SAR at impounded sites are 362 significantly higher than those of natural sites (Fig. 5g, paired t test, t=-2.1, p=0.047, df=23). 363

The resilience of tidal marshes to SLR 365
Significant relationships exist between tidal marsh SEC and RSLR under SLR scenarios of 366 SLR RCP2.6 (R 2 =0.33, p<0.05, F test), RCP4.5 (R 2 =0.21, p<0.05, F test) and RCP8.5 367 (R 2 =0.34, p<0.05, F test) (Fig. 6). These relationships are significant to estimating global 368 SEC since we cannot collect SEC from all the tidal marshes while global RSLR data are 369 available at relatively high precisions. The relationships were incorporated into the modelling 370 framework (Fig. 2) to simulate the response of tidal marshes to SLR. We estimated global 371 We further evaluated the resilience of tidal marshes to SLR by estimating spatially-explicit 436 tidal marsh area changes by 2100 under different RCP scenarios (Fig. 8). At the lower 437 boundary of the business-as-usual scenario (5 people km -2 ), we estimated the hotspots of tidal 438 marsh loss (>100 km 2 ) would account for 66.8%, 67% and 68.2% of total areal marsh losses 439 under RCP 2.6, RCP4.5 and RCP8.5, respectively. These hotspots of marsh loss mainly occur 440 in USA, China, Australia and some European countries (including Russia, Italy and France) 441 (Fig. 7). account for the highest number of sampling sites in the collated references (Fig. 5a). This 458 probably explains why precipitation is the dominant driver of SAR and SEC in our models. 459

460
Our results show that SAR predicts SEC (R 2 =0.98), more closely than has been reported for 461 mangroves (Lovelock et al., 2015), which is consistent with an earlier finding that showed the 462 approximation of tidal marsh SAR to SEC concluded from a smaller database (Kirwan et al., Tides bring allochthonous mineral sediments and contribute to tidal marsh sediment 466 accretion, and help redistribute sediments within tidal marshes (Chmura et al., 2004). Larger 467 tidal ranges correspond to stronger tidal flow and wider intertidal regions (Rogers et al.,468 2019), including the upper intertidal region where sediment accretion is most significant and 469 noticeable, and larger accommodation space available for increasing elevation. Further, 470 smaller tidal range means lower tidal energy, which limits inorganic sediment input from 471 marine sources, whereas macro-tides generate strong currents, which resuspend nearshore 472 sediments and transport them onto tidal marsh surfaces (Hensel et al., 1999). Thus, 473 macrotidal marshes (tidal range> 4m) are more resilient to changes in the rate of relative SLR 474 (RSLR) than microtidal marshes (tidal range< 2m) 1 . In winter, the presence of sea ice affects 475 tidal inundation and thus tidal marsh SAR in high-latitude regions (Ward et al., 2014). Paris Agreement. This means that tidal marshes will contribute to the climate neutral target 530 and our study may facilitate future estimates on the contribution of tidal marshes to the target. 531 532 A previous assessment of countries with coastlines most vulnerable to SLR identified very 533 different areas of vulnerability to our modelled hotspots of marsh loss (Nicholls and 534 Cazenave, 2010). This highlights the impact of our findings -tidal marshes in countries 535 experiencing the most serious SLR will not necessarily be submerged if sufficient 536 accommodation space allows lateral migration. In contrast, at the upper boundary of the high 537 level of nature-based adaptation scenario (300 people km -2 ), the proportion of hotspots of 538 tidal marsh loss is 2.8-3.5% lower, accounting for 64-64.7% of total marsh areal losses under 539 different SLR scenarios (Supplementary Material Fig. S6). This finding highlights the 540 effectiveness of nature-based adaption for shrinking hotspots of tidal marsh loss. 541 542 Our method cannot precisely account for the fate of tidal marshes, since ecogeomorphic 543 feedbacks tend to increase rates of sediment accumulation when marshes become more 544 flooded (Kirwan et al., 2016) under future SLR scenarios. However, it is difficult to estimate 545 the extent of SAR increase due to ecogeomorphic feedbacks in the future, in particular 546 vegetation type changes at local scales (Reed et al., 2020). Moreover, geomorphologic 547 systems often react with relatively long time lags, with the response interval depending partly 548 on the magnitude, frequency and duration of energy factors (Wright and Thom, 1977). Allen 549 (1990) demonstrated a clear lag between SAR and RSLR for immature marshes. The lag was 550 analysed through a numerical model (Kirwan and Murray, 2008), and an analytical model 551 which captures the role of governing factors such as suspended sediment concentrations and 552 plant productivity. We used a 5-year lag, which may not be enough to account for feedbacks 553 at the local scale, adding uncertainty to our analysis of tidal marsh resilience to future SLR. 554 Additionally, the resolution of the satellite data in our models may have limited the precision 555 of the estimate on suspended matter, precipitation and sea ice. However, our mathematical 556 model provides new insights on assessing the resilience of tidal marshes to SLR with real 557 elevation change data by incorporating both vertical sediment accretion and lateral landward 558 migration. 559 560

Data availability 561
The satellite data that support the findings of this study are downloaded from 562 http://hermes.acri.fr/index.php?class=archive (total suspended matter) and   Table S1 578