Global patterns and drivers of tidal marsh response to accelerating sea-level rise


 The vulnerability of the world’s tidal marshes to sea-level rise threatens their substantial contribution to fisheries, coastal protection, biodiversity conservation and carbon sequestration. Feedbacks between relative sea-level rise (RSLR) and the rate of mineral and organic sediment accumulation in tidal wetlands, and hence elevation gain, have been proposed to ameliorate this risk. Here we report on changes in tidal marsh elevation and shoreline position in relation to our network of 387 fixed benchmarks in tidal marshes on four continents measured for an average of 10 years. During this period RSLR at these marshes reached on average 6.6 mm yr-1, compared to 0.34 mm yr-1 over the past millenia. While the rate of sediment accretion corresponded to RSLR, the loss of elevation to shallow subsidence increased in proportion to the accretion rate. This caused a deficit between elevation gain and RSLR which increased consistently with the rate of RSLR regardless of position within the tidal frame, suggesting that long-term in situ tidal marsh survival is unlikely. While higher tidal range (>3m) conferred a greater stability in measures of shoreline change and vegetation cover, other regions showed a tendency towards instability and retreat.

Summary Paragraph 84 85 The vulnerability of the world's tidal marshes to sea-level rise threatens their substantial 86 contribution to fisheries, coastal protection, biodiversity conservation and carbon 87 sequestration. Feedbacks between relative sea-level rise (RSLR) and the rate of mineral and 88 organic sediment accumulation in tidal wetlands, and hence elevation gain, have been 89 proposed to ameliorate this risk. Here we report on changes in tidal marsh elevation and 90 shoreline position in relation to our network of 387 fixed benchmarks in tidal marshes on four 91 continents measured for an average of 10 years. During this period RSLR at these marshes 92 reached on average 6.6 mm yr -1 , compared to 0.34 mm yr -1 over the past millenia. While the 93 rate of sediment accretion corresponded to RSLR, the loss of elevation to shallow subsidence 94 increased in proportion to the accretion rate. This caused a deficit between elevation gain and 95 RSLR which increased consistently with the rate of RSLR regardless of position within the 96 tidal frame, suggesting that long-term in situ tidal marsh survival is unlikely. While higher 97 tidal range (>3m) conferred a greater stability in measures of shoreline change and vegetation 98 cover, other regions showed a tendency towards instability and retreat. Tidal marshes are amongst the most vulnerable of the world's ecosystems. Throughout 120 human civilisation tidal marshes have been reclaimed for agriculture and settlement, and the 121 pace of loss has accelerated in concert with burgeoning coastal populations on all inhabited 122 continents over the past century 1 . To this pressure has been added the threat of accelerating 123 sea-level rise. As tidal marshes occur within tightly defined elevation ranges relative to mean 124 sea level, they are sentinel ecosystems at the forefront of climate change impact. Their 125 potential loss with sea-level rise threathens a range of ecosystem services valued at ~$27 126 trillion per year 2 , extending to fisheries production, recreation, coastal protection, water 127 quality enhancement and carbon sequestration. 128 129 Sea-level rise can lead to in situ marsh loss through three mechanisms: landward retreat, 130 internal expansion of ponds and channels, and loss of marsh surface elevation relative to 131 mean tide level 3 . The fate of tidal marshes under accelerating sea-level rise will be 132 determined by opportunities for landward retreat, but also by the capacity of tidal marshes to 133 gain elevation through processes of vertical accretion (the accumulation of mineral sediment 134 and organic matter 4 ). Feedbacks between the rate of sea-level rise and the vertical 135 development of marsh substrates ameliorates the risk of conversion to unvegetated mudflat. 136 Modelling based on observations from US East Coast marshes has suggested an equilibrium 137 may emerge between the position of a marsh within the tidal frame, plant productivity, root 138 mass development, sedimentation and the elevation of the marsh in response to mean sea-139 level 5 (Fig 1) sustained under low rates of RSLR. How widely these controls, and their upper 140 thresholds, operate across marsh sites around the globe, has been a central and disputed 141 question in the regional-to global-scale modelling of tidal marsh responses to projected rates 142 of relative sea-level rise (RSLR, the combination of vertical land movement and sea level 143 change) under climate change 6-8 . 144 145 Several factors operating at regional and global scales may influence the efficacy of tidal 146 marsh vertical adjustment to sea-level rise. Tidal range in marshes can vary by two orders of 147 magnitude (less than 10 cm to more than 10 m) influencing susceptibility to drowning under 148 a given rate of RSLR 9 . Tidal hydrodynamics and river discharge contribute to sediment 149 delivery and accumulation 9 , and these may be modified by flow control structures 10 . Plant 150 productivity is influenced by climate (precipitation and temperature), atmospheric CO2 and 151 vegetation composition, as is soil organic carbon accumulation and decomposition. The rate 152 of RSLR varies across coastlines and continents, and millennial-scale variability in RSLR 153 may also confer a legacy of soil organic content 11 . Only by sampling across hydro-154 geomorphic settings and biogeographic gradients can the significance of these factors be 155 clarified, and the consistency of feedbacks between RSLR and position in the tidal frame be 156 determined. 157 158 Accurate measures of tidal marsh vertical adjustment in relation to sea level require a fixed 159 benchmark against which elevation gain or loss can be measured. To this end, the Surface 160 Elevation Table -Marker Horizon (SET-MH) method has been developed as a global 161 standard 12 for monitoring tidal marsh responses to sea-level rise (Fig 1). A benchmark rod is 162 driven into the marsh to form a stable benchmark against which elevation change can be 163 measured. Vertical accretion is also measured at most sites above an artificial soil horizon 164 (e.g., typically white feldspar or sand) introduced at the time of the first reading against the 165 benchmark (Methods). Comparison between the rate of vertical accretion and elevation gain 166 using the SET-MH method and the rate of RSLR measured at local tide gauges has indicated 167 the vulnerability of mangroves across the Indo-Pacific to sea-level rise and the importance of Here we analyse tidal marsh elevation adjustment in relation to sea-level rise from our 178 network of 387 SET-MH monitoring stations spanning four continents. Vertical adjustment 179 in marsh accretion and elevation at SET-MH stations were monitored for an average of 10.9 180 years (range 3.5 -20.0 years) in a network encompassing a broad range of tidal amplitude, 181 geomorphic settings, rates of RSLR and spanning 70 degrees of latitude north and south of 182 the equator. We analyse marsh elevation gain and accretion in relation to candidate predictive 183 variables collected for each site, including position within the tidal frame, modelled 184 suspended sediment concentration in adjacent water bodies, and climate. RSLR was derived 185 for three time-scales: (1) modelled for each site over century to millennial timescales; (2) 186 calculated from nearest tide gauges over the past 50 years; and (3) calculated from nearest 187 tide gauges over the period of SET measurements at each site (hereafter contemporaneous 188 RSLR). The centuries over which the tidal marshes formed were characteristed by gradually 189 falling sea-level at the southern hemisphere sites, and RSLR at the northern hemisphere sites 190 of less than 1mm yr -1 on average (Table 1; Data S1). During the past 50 years, RSLR at these 191 tidal marshes has increased to 4.1 mm yr -1 per year, and duringthe period of SET observation 192 to an average of 6.6 mm yr -1 , the latter rate consistent with threshold rates for tidal marsh 193 failure and retreat found in the palaeo-stratigraphic record 16,17 .  (Table S1; Data S1). All Previous modelling has stressed the importance of suspended sediment concentrations in 220 conferring resilience to wetlands subject to RSLR 6,23,13 and modelled total suspended 221 sediment, derived from the MERIS satellite, has been used to project tidal wetland responses 222 to RSLR scenarios at a global scale 7 . While total suspended matter (TSM) proved to be an 223 important determinant of accretion rate (Fig S1) at the regional scale (particularly for Europe 224 and Atlantic North America where previous studies have been focussed 23 ), only 11 percent of 225 global variation in accretion was explained by TSM. Random Forest models suggest the 226 strongest controls on accretion at the global scale are RSLR (both for the past 50 years and 227 contemporaneous), and position within the tidal frame (Fig 3; Fig S1). That is, the accretion 228 rate is a function of tidal inundation depth and duration, and the rate at which this increases 229 with RSLR. Spain, where RSLR declined) (Table 1). 243

244
There is a tendency for wetlands lower in the tidal frame to be increasing in elevation at a 245 higher rate (Fig 3b), as predicted by models 5,6 , though we found this feedback to be biased 246 towards sites close to retreating shorelines ( Fig S3). The mean rate of elevation gain in low 247 marshes (D>0) showing shoreline stability or progradation was 3.06 ± 3.11 mm yr -1 , similar 248 to the average for the dataset (2.94 ± 3.86 mm yr -1 ). For low marshes where the shoreline was 249 retreating, the rate of elevation gain was higher (Table S3) (Table S4). The elevation subsidy provided by proximity to eroding shorelines 258 20,26 does not confer resilience over broader spatial or temporal scales 27 . 259 260

Regional trends in vulnerability 261
On the Ebro Delta in Spain sea-level stabilised over the measurement period, and here tidal 262 marshes were high in the tidal frame, shorelines were stable, and elevation increasing ( Table  263 1). Though RSLR increased in the macro-tidal marshes of the North Sea (Essex, Norfolk, The  (Table 1). This is likely due to their relatively high position 277 in the tidal frame, stable shorelines and lower RSLR than the global average (Table 1). 278 Mangroves occupy low marsh positions and tidal marsh loss has been associated with a 279 consistent trend of landward encroachment by mangrove over the past seventy years 28 280 consistent with the increasing hydroperiod within these tidal marshes. Africa and the Atlantic and Pacific coasts of North America were lower in the tidal frame and 284 subject to higher rates of RSLR than in Australia (Table 1). These marshes had a lower 285 proportion of vegetated marsh than is considered stable 21 and in 83% of cases are retreating 286 (Table 1; Data S1). Tidal marsh elevation gain in these settings was comparable with the 50-287 year average RSLR but not contemporaneous RSLR, against which a pronounced elevation 288 deficit emerges for South African (2.05 mm yr -1 ), North American Pacific-coast tidal marshes 289 (~5 mm yr -1 ), and to a lesser extent North American Atlantic-coast tidal marshes (< 1mm yr -290 The most vulnerable marshes in our global network are associated with the Mississippi River 293 deltaic plain. The active delta sites recorded the highest sediment accretion in the global 294 network (13.28± 7.15 mm yr -1 ) translating into the highest elevation gain (6.45 ± 6.09 mm yr -295 1 ), yet still 7.73 mm yr -1 below contemporaneous RSLR. Marshes were already low in the 296 tidal frame, and the ratio of unvegetated to vegetated marsh was the highest in the global 297 network (Table 1). Shorelines adjacent to monitoring sites retreated at a mean rate of 21 ± 35 298 cm per year. Marshes in the chenier plain to the west of the active delta are even more 299 vulnerable. Chenier plain marsh elevations and position in the tidal frame are close to the 300 lower survival limit of the dominant genus (Spartina) 29 , adjacent shorelines retreated at a 301 mean rate of 66 ± 102 cm per year, and the mean deficit between elevation gain and 302 contemporaneous RSLR is 15.95 ± 4.09 mm yr -1 (Table 1). Despite having high sediment 303 accretion, marsh elevation gain is still too low to counter the shallow and deep subsidence 304 experienced by Delta wetlands as they respond to contemporaneous RSLR; the sediment 305 accretion experienced during periods of higher legacy sediment erosion from the vast 306 Mississippi River watershed in the past 30 is no longer sufficient. 307 308 Concluding paragraph 309 310 Tidal marshes have been subject to relatively low rates of RSLR over the past few millennia, 311 although this is changing rapidly 11 . Our estimation of RSLR trends across the network 312 suggests local RSLR rates increased from 0.34 mm yr -1 (averaged for the past 1000 years), to 313 4.1 mm yr -1 averaged over the past 50 years, to 6.6 mm yr -1 averaged over the period of SET Mississippi Delta marshes due to accelerated sea-level rise. Science Advances 6, 402 eaaz5512 (2020  We conceptualise surface elevation trends as a function of elevation gains (through sediment 488 accumulation, and soil volume expansion, including root mass gain) and losses (through 489 sediment erosion, and soil volume losses such as subsidence and compaction). These 490 processes are driven by hydrological, geomorphological and biological processes (Fig 1). 491 Hydrological processes influence the accumulation of sediment through the mechanism of 492 tidal inundation. Tides define the lateral limit of tidal marshes and the space available for The Surface Elevation Table-Marker Horizon (SET-MH 2 ) technique is regarded to be the 507 global standard in measuring wetland responses to sea-level rise in real time 3 . It combines a 508 benchmark rod against which marsh elevation change is monitored (the SET), with an 509 artificial soil marker horizon against which marsh vertical accretion is measured (the MH) 4 510 ( Figure 1). Prior to installation, a platform is usually constructed to minimise disturbance 511 and compaction. In our network two types of benchmark rod were used: an "original" design 512 consisting of a hollow aluminium pole up to 8 metres in length, and an "rSET" design, 513 consisting of a solid stainless steel rod capable of insertion to greater depths (up to ~30 514 metres). In both cases benchmark rods serve as a fixed point against which marsh elevation 515 change is measured. A portable arm is attached to the benchmark at each visitation and 516 supports 9 replicate pins that are lowered to the marsh surface at four fixed compass 517 directions; measurements of the height of each pin above the portable arm are taken at each 518 visit. At commencement, replicate (3 to 4) marker horizons (feldspar or clay) are laid on the 519 soil surface over 0.25 m 2 square plots adjacent to each SET and are subsequently buried by 520 the accumulation of tidally borne sediment and root growth. A shallow core is extracted and 521 the depth of the marker horizon in each replicate plot recorded at each visit. The difference 522 between surface accretion, as measured from cores extracted from the MH, and surface 523 elevation change, as measured using the SET, is a measure of shallow subsidence or 524 expansion occurring between the bottom of the marker horizon and base of the SET 525 benchmark 4 (Figure 1). 526 527 Our network consists of 387 SET-MH stations in tidal marshes installed using common 528 protocols in 89 locations on four continents (North America, Australia, Europe, South 529 Africa). From this network changes in surface elevation and vertical accretion were 530 determined from repeated measurements occurring across timescales ranging from 3.5 to 20 531 years (average 10.9 years: Data S1), and rates of surface elevation change and vertical 532 accretion were determined at each site. The network consists of seven regional clusters (Fig  533   2), being the Atlantic coast of North America (91 SETs; 23 of which were located in the in relation to tidal frame was not known, or where the SET-MH station was associated with a 540 hydrological restoration initiative. Some sites had not recorded accretion but were included in 541 analyses of elevation change. Sites spanned macrotidal settings (greater than 3 m tidal range: 542 Bay of Fundy, Canada; Gulf of Maine, USA; The Wash, UK) to microtidal settings (less than 543 1 metres tidal range: US Gulf Coast; Venice Lagoon) and were evenly distributed between 544 coastlines subject to relatively rapid RSLR (>5mm yr -1 ; 119 SETs), near average global 545 eustatic RSLR (2-5mm yr -1 ; 150 SETs), and low RSLR (<2mm yr -1 (114 SETs) (Fig 4)  We measured the elevation (Z) of each SET-MH station in relation to the local height datum 552 using either a real time kinematic GPS or differential GPS, and accessed mean high water 553 (MHW), mean low water (MLW) and mean sea level (MSL) in relation to the local height 554 datum for the nearest tide gauge (Table S2). We calculated tide range as the difference 555 between MHW and MLW. We described position within the tidal frame using 556 "dimensionless d" 5 (D; Equation 1), a metric commonly used in the interpretation of 557 intertidal position 6,7 , and found in a survey of US marshes 7 to be a useful approximation of 558 flooding duration.

Relative Sea-level rise 579
Contemporary rates of RSLR (for the past 50 years, and the period for each site 580 contemporaneous with SET-MH measures) were obtained from NOAA 581 (https://tidesandcurrents.noaa.gov/sltrends/sltrends.html), or local tide gauges as documented 582 in Table S1. We also considered longer-term (centennial to millennial) rates of RSLR given 583 their possible influence on upper marsh processes. Rates of local and regional RSL change to be equally likely. We linearly interpolated between grid and time points from these 604 ensemble members to predict RSL rates of change and their uncertainties for each site in this 605 study. Rates of historic change were provided for consecutive 500-year periods from 0-606 500BP (SLR250 in Data S1) to 3500-4000 BP (SLR3750 in Data S1). 607 608

Suspended sediments (total suspended matter TSM) 609 610
A remote sensing product that estimates the dry weight of particles suspended in the coastal 611 Mean annual temperature and mean annual precipitation were sourced from the nearest 635 meteorological station as documented in Table S3. Dry bulk density is the dry weight of both 636 organic and inorganic materials in a sample of known volume, and typically reported as 637 grams per cubic centimeter 21 . We measured the bulk density of the upper 10cm, the section of 638 profile most likely to correspond to sediment accreted during the period of record. Dominant 639 vegetation was classified to genus level (Data S1), and clustered into the following categories 640 by growth form and habit: 641 We used Google Earth Engine to locate the position of SET platforms. The platforms were 662 used as a fixed point in the landscape against which to assess shoreline change. The distance 663 between the SET platform and the nearest vegetated shoreline was measured over the period 664 for which available historic imagery corresponded most closely to the length of the SET 665 record. For Australian sites, where mangroves frequently occupy the lower intertidal zone, 666 the distance to the closest contiguous mangrove stand was also measured. Imagery was 667 discarded if high water level or cloud cover obscured the platform or vegetated shoreline. In 668 some cases georectification errors prevented meaningful comparison between images. 669 Results are shown in Table S1. For each SET, relative pin height was calculated by subtracting baseline pin height from all 681 subsequent readings. Relative pin heights were averaged hierarchically within each SET arm 682 position and then across positions to integrate small-scale variation in surface elevation. The 683 rate of elevation change was then calculated as a linear regression slope for the relationship 684 between the date of measurement and averaged relative pin height. A similar approach was 685 used to calculate accretion rates. Simple and multiple linear regression were used to test 686 relationships between quantitative variables. Generalized additive models (GAM) was used 687 to test the relationship between subsidence and accretion rate. Analyses of variance were 688    Distribution of tidal marsh SET-MH stations used in the analysis, and de cit between elevation gain and local RSLR. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. The increasing vulnerability of tidal marshes to RSLR. While accretion increases with RSLR over the same period of measurement (a), and with increasing depth in the tidal plane (b), the rate of shallow marsh subsidence increases with accretion rate (with an upward in exion as RSLR rises above ~7mm yr-1 (c). As a result, the de cit between elevation gain and RSLR increases with RSLR (d). In panels (b) and (c) points are coloured for the 50-year RSLR trend in mm yr-1