Widespread retreat of coastal habitat is likely at warming levels above 1.5 °C

Several coastal ecosystems—most notably mangroves and tidal marshes—exhibit biogenic feedbacks that are facilitating adjustment to relative sea-level rise (RSLR), including the sequestration of carbon and the trapping of mineral sediment1. The stability of reef-top habitats under RSLR is similarly linked to reef-derived sediment accumulation and the vertical accretion of protective coral reefs2. The persistence of these ecosystems under high rates of RSLR is contested3. Here we show that the probability of vertical adjustment to RSLR inferred from palaeo-stratigraphic observations aligns with contemporary in situ survey measurements. A deficit between tidal marsh and mangrove adjustment and RSLR is likely at 4 mm yr−1 and highly likely at 7 mm yr−1 of RSLR. As rates of RSLR exceed 7 mm yr−1, the probability that reef islands destabilize through increased shoreline erosion and wave over-topping increases. Increased global warming from 1.5 °C to 2.0 °C would double the area of mapped tidal marsh exposed to 4 mm yr−1 of RSLR by between 2080 and 2100. With 3 °C of warming, nearly all the world’s mangrove forests and coral reef islands and almost 40% of mapped tidal marshes are estimated to be exposed to RSLR of at least 7 mm yr−1. Meeting the Paris agreement targets would minimize disruption to coastal ecosystems.

Nature | Vol 621 | 7 September 2023 | 113 deficit is sustained for a sufficiently long period, elevation capital is exhausted. For wetlands, retreat and a transition to open water may occur, and in reef islands, submergence of reef crests will increase wave exposure and wave over-topping frequency. Whether the areal extent of the habitat expands or contracts over time depends on the rate of loss and the rate of new habitat formation, both both of which are influenced by RSLR 14 .
Contemporary observations of high accretion rates in coastal ecosystems have indicated resilience under current and projected RSLR rates, prompting reassessment of their vulnerability in modelling studies 3,11,14 . Conversely, studies emerging from the palaeo record show a comparatively high vulnerability of mangroves 12 and tidal marshes 15,16 to rates of RSLR that are anticipated in coming decades under moderate and high emissions scenarios 17 . Palaeo records show that most coral reef islands formed during the later stages of the Holocene epoch under conditions of stable or falling relative sea level 2 (RSL). The upper limits of resilience to projected RSLR remains an important knowledge gap, with wide ranging implications for coastal zone protection and management.
Here we analyse three independent lines of evidence to assess the vulnerability and exposure of coastal ecosystems to the higher rates of sea-level rise (4 mm yr −1 to more than 10 mm yr −1 ) projected under global warming scenarios. We focus on intertidal and supratidal ecosystems that undergo vertical adjustment from biogenic feedbacks, facilitating resilience to RSLR: mangroves, tidal marshes and coral reef island systems. We exclude beaches, rocky reefs and rock platforms, for which biogenic feedbacks with RSLR are largely absent, and subtidal vegetated ecosystems (seagrass meadows and kelp forests) for which thermal stress is likely to be the primary driver of change, rather than RSLR 18,19 . First, we review the behaviour of these ecosystems over the range of sea-level histories encountered following the Last Glacial Maximum 19 thousand years ago (ka), and particularly since 10 ka. Second, for mangroves and tidal marshes, we document elevation trends in relation to contemporary rates of RSLR using a global network of survey benchmarks, the surface elevation table-marker horizon (SET-MH) network. Third, we analyse the extent to which contemporary coastal ecosystems show conversion to open water (hereafter referred to as 'retreat') under a range of settings with varying rates of RSLR. From these three lines of evidence, we estimate the probability of coastal ecosystem retreat in relation to RSLR rates and model the response of the world's existing coastal ecosystems under the RSLR projections of the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report 17 , including potential compensation through conversion of terrestrial uplands (hereafter 'landward migration'). From these analyses a picture emerges of the narrowing boundaries of the 'safe operating space' 20 for coastal ecosystems: the climate futures expected to be of low risk to existing ecosystems.

Responses to past sea-level rise
RSL varies globally in response to both water and land vertical movement 21 . Coastlines continue to adjust to the loss of ice sheets after the previous glacial period, particularly in higher latitudes, a process called glacial isostatic adjustment 21,22 (GIA). GIA modelling provides insights into sea-level trends since the last deglaciation. These observations and models have been applied to interpret rates of RSLR associated with the timing of mangrove and tidal marsh retreat and/or advance in stratigraphic successions, with results showing broad consistency among settings 12,15 . Rapid global mean sea-level (GMSL) rise (over 10 mm yr −1 ) during several periods since the Last Glacial Maximum has drowned mangrove forests and tidal marshes (Fig. 1). Periods of rapid GMSL rise include: (1) meltwater pulse 1A (14.6-14.3 ka) which drowned mangroves, tidal marshes and coral reefs, the remains of which have been found at water depths of around 90 m; and (2) a rapid rise in GMSL 11.3-11.0 ka, leaving relict features at water depths of around 50 m. Interspersed with these phases have been periods of slower GMSL rise, allowing extended periods of coastal ecosystem expansion ( Fig. 1; Methods). For example, mangrove sediments associated with preserved coastal palaeo channels on the Sahul Shelf (northwest of mainland Australia) and the Sunda Shelf (western South China Sea) dating to 16.0-14.5 ka were probably drowned during meltwater pulse 1A (Methods). Mangrove forests re-established in India from around 10 ka on the former delta of the Ganges-Brahmaputra River (Fig. 1) and the northeast Australian continental shelf (Queensland). In both locations, RSLR declined to between 6 and 7 mm yr −1 , and rates of sedimentation were high. These forests were subsequently drowned during a period of more rapid RSLR of 7 to 8 mm yr −1 between 9.0 and 8.5 ka (Methods). Widespread mangrove forest development commenced around 8.5-7.5 ka in Southeast Asia, northern and eastern Australia, South America and Africa (Fig. 1) as the rate of RSLR declined 12 below 7 mm yr −1 (Fig. 2c), and mangroves associated with large rivers were able to maintain their intertidal position by trapping sediment and accumulating root mass.
Tidal marshes in Great Britain were 9 times more likely to retreat than advance during the Holocene when RSLR exceeded 7.1 mm yr −1 , based on more than 780 reconstructions of tidal marsh evolution 15 (Fig. 2b). In the Mississippi Delta, only short-lived and rapidly retreating fringing tidal marshes existed before 8.2 ka, with retreat occurring in approximately 50 years before RSLR slowed 16 to less than 6 to 9 mm yr −1 . Tidal marsh retreat took longer (centuries), as RSLR dropped below 6 mm yr −1 after around 8.2 ka, but marshes did not stop retreating in the Mississippi Delta 16 until RSLR was less than 3 mm yr −1 (Fig. 2b).
Sea level stabilized in the mid-Holocene in those parts of the world that were distant from former centres of glaciation, and many coral reefs-especially in the Pacific-reached sea level with diversification of reef habitats 23 . Subsequent fall of sea level relative to these regions resulted in emergent reef platforms, some of which became suitable habitat for mangroves, and on which it became possible for reef islands to form 24 . Infilling of estuaries resulted in development of extensive coastal plains 12 , reducing intertidal areas previously covered by mangroves, including in coastal Northern Australia (South Alligator River, Fitzroy River, Ord River, Cleveland Bay and Richmond River), Thailand (Great Songkhla Lakes) and Vietnam (Mekong River and Red River).
The palaeo record therefore indicates a capacity for vertical adjustment to rates of RSLR similar to those encountered in the instrumental period. If these rates of RSLR are sustained, coastal lowlands may be re-occupied by tidal wetlands where migration is permitted, and in many places this encroachment has already commenced 25 . The potential for increased extent in these regions under higher sea level is captured in global wetland adjustment models 14,26 . However, there is consistent evidence that vertical adjustment and habitat extent are greatly reduced 12,23 as RSLR approaches 7 to 8 mm yr −1 .

Elevation trends under current sea-level rise
Mangrove and tidal marsh accretion can increase with the rate of RSLR. Increased inundation depth and duration can facilitate both mineral deposition 27 and higher plant productivity and root mass accumulation 11 . Rates of accretion measured against artificial marker horizons and radiometric markers often correspond to high rates of RSLR encountered in settings where land is subsiding 3,28,29 . High rates of accretion (10 to 20 mm yr −1 ) have been observed in contemporary mangroves and tidal marshes on active deltas 29,30 , and accretion in the intertidal zone increases with increased depth and duration of inundation 3 . The assumption that accretion enables vertical adjustment to RSLR is the basis of projections of possible resilience under projected future rates 31 , but the assumption requires testing against fixed elevation benchmarks.
To assess the relationship between accretion of surface sediment, vertical adjustment and sea-level rise in contemporary coastal wetlands, we used the SET-MH method ( Fig. 2d; Methods). The SET-MH 114 | Nature | Vol 621 | 7 September 2023 Article uses a benchmark survey rod coupled with an introduced sediment horizon to assess the relationship between accretion and elevation gain 32 . SET-MH data in tidal marshes show that shallow subsidence (the difference between sediment accretion and elevation gain) increases with accretion rate and the rate of RSLR 28,29 . A previous analysis of a globally distributed network of 477 tidal marsh SET-MH stations showed that the increase in the subsidence rate with increasing accretion was non-linear 29 . For this reason, elevation deficits emerged under rates of RSLR similar to those inferred from the stratigraphic record 29 (Fig. 2e and Extended Data Fig. 2). We repeated this Bayesian analysis for 190 SET-MH installations in mangrove forests (Methods), estimating the cumulative probability of vertical adjustment at or exceeding the rate of RSLR at the SET-MH stations ( Fig. 2f and Extended Data Fig. 2). The results were consistent with those for tidal marshes. We found that an elevation deficit at mangrove sites is very likely (P > 0.9) at RSLR between 7 and 8 mm yr −1 (Fig. 2f), consistent with tidal marshes monitored using the same method (Fig. 2e). These observations concur with the limits of tidal marsh and mangrove stability in relation to palaeo-RSLR as inferred from the stratigraphic record 29 and described previously (Fig. 2a-c).

Habitat change under current sea-level rise
As a third line of evidence, we assessed whether changes in the extent of tidal marsh and open water were consistent with RSLR and/or the deficit between RSLR and marsh vertical adjustment (Extended Data Fig. 3; Methods). Previous surveys of contemporary North American tidal marshes in low-to-moderate tidal range settings 33 found that habitat retreat commenced at a RSLR of 4 to 6 mm yr −1 . For example, the Maryland Eastern Shore is retreating 34 under a long-term RSLR trend of around 6 mm yr −1 . In a comprehensive analysis of tidal marshes in the contiguous USA, gains in tidal marsh were found to be inversely related to RSLR, with some marsh loss associated with short-term perturbations, notably hurricanes 34 . RSLR was also associated with reduced normalized difference vegetation index (NDVI) values for vegetation adjacent to the marsh 34 , possibly resulting from saline water intrusion. We used high-resolution global mapping of surface water change 35 and tidal wetland extent 36 in the immediate vicinity of tidal marsh SET-MH stations globally to determine the influence of contemporary RSLR, elevation capital and elevation deficit on marsh loss. Canopy cover obscured observations of surface water in mangroves. We found that tidal marsh sites were likely to show a trend towards increased presence of surface water (P > 0.66) once RSLR exceeded 2.3 mm yr −1 (Extended Data Fig. 3). The frequency of surface water observations at marsh sites increased with both the rate of RSLR (r 2 = 0.16, P < 0.001; Extended Data Fig. 4) and marsh elevation deficit (r 2 = 0.14, P < 0.001; Extended Data Figs. 4 and 5). The relationship between surface water change and marsh elevation deficit was evident in lower elevation marsh sites (r 2 = 0.20) rather than higher elevation sites (r 2 = 0.03; Extended Data Fig. 5), illustrating the temporary resilience conferred by elevation capital. We also found a significant relationship between the proportion of tidal marsh conversion to open water habitat and RSLR (P = 0.018). Tidal marshes were as likely as not (P = 0.5) to be retreating as RSLR increased above 5.4 mm yr −1 (Extended Data Fig. 3), with relatively few marshes advancing. This estimate of retreat may be conservative because patches of interior marsh break-up may not have been identified (Methods). The ameliorating influence of elevation capital was also evident in the extent of marsh retreat. Where marshes had higher than the median elevation capital, there was no relationship between marsh retreat and RSLR (P = 0.850). At lower than the median elevation capital, the relationship was highly significant (P = 0.002).
There are relatively few data on the change to reef-top habitats. Surveys of reef island planiform change in the tropical western Pacific and Indian Oceans have shown a remarkable degree of stability under rates of RSLR up to the contemporary GMSL rate 37,38 . Our collation of existing data on reef island morphometric changes (n = 872) from the Indian and Pacific Oceans shows a higher probability of island contraction at rates of RSLR above the rate of contemporary GMSL rise (Methods; Extended Data Fig. 3c). Island size reduction is likely (P ≥ 0.66) at RSLR above 6.2 mm yr −1 . The rate of RSLR in the Solomon Islands has averaged between 7 and 10 mm yr −1 since 1994 (ref. 39), and in the exposed northern Isabel Province, five of the twenty vegetated reef islands have completely eroded, leaving dead mangrove trunks on hard coral 40 . A further six islands contracted by more than 20% in the

Projected response to future sea-level rise
Modelling of spatial variability in RSLR was completed in the IPCC AR6 for each warming scenario 42 . We compared regional RSLR projections to 2080-2100 with the distribution of mangroves, tidal marshes and coral reefs across the globe ( Fig. 3; Methods). For each of the modelled scenarios, we determined the proportion of mangrove, tidal marsh and coral reef island habitat occurring where RSLR is projected to rise to levels for which eventual retreat of mangroves and tidal marshes is likely (4 mm yr −1 ) or very likely (7 mm yr −1 ), the best estimate from our combined palaeo and instrumental observations. For reef islands (a subset of mapped reefs), contraction or increasing island instability by RSLR of 7 mm yr −1 is likely (Extended Data Fig. 2c), although we cannot yet specify a rate of RSLR at which contraction is highly likely, given the scarcity of contemporary observations at higher rates of RSLR, and because this threshold will vary with the rate of surrounding reef vertical growth, reef flat width, wave exposure, island size and height, and reef-derived sediment supply.
In the 1.5 °C scenario, the likely (P ≥ 0.66) rate of GMSL rise at 2080-2100 is between 2.4 and 6.4 mm yr −1 . Coastlines subject to rates of RSLR of 4 to 7 mm yr −1 correspond to centres of contemporary mangrove development, notably Southeast Asia and the Caribbean. Under this rate of RSLR, elevation deficits are likely (P = 0.66-0.90; Fig. 2). The probability of reaching a rate of RSLR at which elevation deficits are very likely (7 mm yr −1 ) remains low (<11%), although coastlines subject to high rates of land subsidence-including, for example, the US Gulf Coast and Southeast Asian deltas 28,43 -are projected to exceed this rate. Median projections for the 2 °C warming scenario suggest that one third of global mangroves are subject to ≥7 mm yr −1 and nearly all exposed to ≥4 mm yr −1 of RSLR, although there is comparatively little change in the proportion of tidal marshes and reefs exposed to ≥7 mm yr −1 of RSLR (Table 1). Under 3 °C of warming, nearly all tropical and subtropical latitude coastlines are exposed to ≥7 mm yr −1 of RSLR, and these are the locations of most of the world's mangroves and coral reefs. Median RSLR projections along the world's coastlines therefore show the probability of elevation deficits in mangroves shifting from likely to very likely between 2 °C and 3 °C of global warming ( Fig. 3 and Table 1).
At high latitudes, portions of coastline have declining RSLR owing to gravitational, rotational and elastic deformational effects resulting from mass loss of glaciers and the Greenland Ice Sheet offsetting GMSL rise. For this reason, proportional loss of existing tidal marsh with RSLR is expected to be lower than for mangroves with increased   Note that projected rates of RLSR rely to a considerable extent on tide gauge records that may capture local anomalies (for example, due to fluid extraction) that could produce locally higher rates. e-g, The proportion of global tidal marsh (e), mangrove (f) and coral reef (g) habitat subject to 7 mm yr −1 of RSLR by 2100 in the scenarios shown in a-d, as well as the 5 °C scenario. Error bands show the 17-83% likely range. These projections do not take into account the possibility that ice sheet instabilities substantially increase RSLR in warming scenarios exceeding 2 °C.
Nature | Vol 621 | 7 September 2023 | 117 warming. At 2 °C warming, the high-latitude European and North American west coasts remain below 4 mm yr −1 RSLR under median estimates, and at 3 °C the Baltic Sea and Gulf of Alaska remain below 4 mm yr −1 . Tidal marsh habitat is likely to expand in extent in northern Siberia under higher RSLR owing to limited topographic and human development impediments ( Fig. 4 and Extended Data Fig. 6). Far northern coastlines therefore emerge as important future habitats for tidal marsh-as also projected for seagrass meadows and kelp forests 19,30,44 -under warmer temperatures and reduced ice cover and ice scour, increasing their relative contribution to blue carbon capture and storage at high latitudes.

The influence of global change drivers
The behaviour of future ecosystems may not always be anticipated by palaeo and contemporary analogues. Processes influencing vertical adjustment of coastal wetlands and reefs to sea-level rise may be modified by climate change, though often the influence is to supress vertical adjustment. Land-use change driven by population growth may increase sediment supply by rivers, subsidizing sediment accumulation in coastal deltas 45,46 . Counteracting this is the association between economic development and dam construction, an intervention that retains sediment within catchments. Sediment yields to coastal environments in the global north are nearly half those prior to such hydrological modifications 45 . Major hydrological developments in Southeast Asian rivers have negative implications for the resilience of mangroves to sea-level rise 47 . Elevated concentrations of atmospheric CO 2 and associated climate change may modify biotic feedbacks to sea-level rise. Long-term field mesocosm experiments in Chesapeake Bay, USA have shown that root growth and marsh vertical adjustment was enhanced by the atmospheric CO 2 fertilization effect 48 and moderate warming (approximately 1.7 °C above ambient 49 ). However, as observed RSLR increased above 7 mm yr −1 , water stress negated the benefit of elevated CO 2 (ref. 50), and temperatures above 1.7 °C increasingly promoted organic carbon remineralization, lowering elevation gain 50 .
Ocean acidification and thermal stress will supress reef vertical growth due to impacts on coral cover, unless rapid adaptation occurs. Recent estimates identify low accretion potential (averaging 1.8 ± 2.2 mm yr −1 ) across many tropical western Atlantic reefs 51 , compared with rates derived from palaeo-reef core records 51 . Currently  For mangroves and tidal marshes, loss is the proportion of existing area exposed to 4 mm yr −1 (likely loss; P > 0.66) and 7 mm yr −1 (very likely loss; P > 0.9) of RSLR based on probability distributions presented in Fig. 2, and the RSLR modelling in Fig. 3. For coral reef islands, the proportion refers to numbers of reefs, and uses the conservative estimate of likely vulnerability to RSLR at 7 mm yr -1 (the full dataset with uncertainties is presented as Extended Data  Article less than half of reefs in the western Atlantic and Indian Ocean have maximum accretion potential rates matching altimetry-derived rates of sea-level rise 51 . Recent modelling of the impacts of climate change on reef accretion potential to 2100 suggest that increasingly severe and frequent bleaching events will further limit reef accretion potential 52 (even in the absence of other confounding local disturbance pressures). The potential of reef-top habitats and reef islands to accrete will therefore be influenced by increasing water depths above the surrounding fringing reefs and probable shifts in the abundance and production rates of biota from which sediment is derived. Both may negatively affect future reef-top habitats and will almost certainly impinge upon cultural use and sustainability 14 .

Implications for management
The committed loss of coastal habitats under high warming scenarios should not discourage conservation and restoration efforts. Under small elevation deficits, centuries may elapse before the elevation capital of a wetland is exhausted, and this will provide sufficient time for the supply of ecosystems services, including those critical for well-being and sustenance. Over the current century, landward migration driven by sea-level rise may compensate wetland loss, or even facilitate wetland expansion and associated carbon burial potential 53 . Extensive mangrove forest development in the mid-Holocene, coupled with high rates of vertical accretion under 4 to 7 mm yr −1 RSLR promoted blue carbon capture and storage at a scale that may have contributed to an observed decline in global atmospheric CO 2 concentrations for this period 12 . In the near-term, increased GMSL potentially allows for the recolonization of these coastal floodplains, expanding mangrove area while promoting higher rates of organic carbon accumulation than currently encountered 53 . Although intensive coastal development in Asia has reduced coastal wetland extent in former biogeographic centres 36 and is likely to restrict landward retreat (Fig. 4), extensive coastal floodplains provide viable opportunities for mangrove landward migration and even aerial expansion in northern and northwestern Australia 54 , the northern Gulf of Mexico 55 , Siberia and-depending on opportunities for restoration in more populated areas-Central America, Colombia and the western Mediterranean (Fig. 4.). In the Gulf of Mexico and northern Australia, mangrove forest and tidal creek encroachment under higher rates of RSLR is already being observed 25,56 . The implication of the gap between the Paris Agreement aspiration (2 °C with an aim of 1.5 °C) and the pathways consistent with the implementation of current policies (2.4 °C to 3.5 °C by 2080-2100, medium confidence 57 ) is profound for coastal ecosystems. Warming above 2 °C would restore the conditions faced by mangroves and tidal marshes under previous high RSLR periods and would likely expose most of the world's mangrove and two thirds of the world's tidal marsh to elevation deficits ( Table 1). Warming of 3 °C by 2100 would accelerate GMSL rise to rates consistent with a high probability of eventual tidal marsh and mangrove retreat and increased reef island instability for much of their geographic extent. Once reached, these rates of RSLR are projected to persist for centuries to millennia 58 . The thermal inertia of ocean waters is likely to drive irreversible ice sheet grounding line retreat where bedrock slopes away from the coast 58 , ensuring ongoing marine ice sheet instability 59 . Projected elevation deficits therefore define committed losses upon the exhaustion of elevation capital. Our analysis therefore suggests that the long-term contribution of blue carbon to climate mitigation is compromised under higher emissions scenarios. While preserving organic carbon in situ in many settings 12,53 , narrower, younger and more transitional wetlands would predominate 23 . As a result, coastlines and reef islands that are currently protected will be increasingly exposed to erosion and retreat, consistent with palaeo observations 16,23 .
Coastal ecosystems represent another of the numerous tipping elements for climate change impacts and rank among the more vital to human well-being and vulnerable to imminent warming levels 60 . The non-linear response to external forcing as seen in a wide range of ecosystems is closely associated with the concept of safe operating space, which promotes planetary boundaries being maintained a safe distance from critical thresholds of unacceptable environmental change 20 . Our findings demonstrate that the boundaries for a safe operating space for coastal ecosystems are approaching, and will be set by near-term emissions pathways. They also highlight the importance of mitigating against local environmental stressors (such as pollution in coral reefs) and restoring cleared and degraded wetlands to enhance resilience against climate change and coastal recession. In the face of irrevocable disruption under high rates of RSLR, the most effective means of promoting the continued survival of widespread mangrove forests, tidal marshes and coral reef islands is to achieve the Paris Agreement goal of net zero emissions by 2050. To this end, a contribution will be made by the preservation, restoration and landward accommodation of coastal blue carbon ecosystems.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06448-z. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Palaeo wetland response to RSLR
To estimate RSLR following the Last Glacial Maximum (Fig. 1), we use a revised numerical simulation of GIA 61 , which adopts the ICE-6G global ice reconstruction from the Last Glacial Maximum to the present 62,63 . We use an ensemble of 300 combinations of rheological parameters in the GIA model to estimate RSL at 500-year time steps on a 512 × 260 global latitude × longitude grid, resulting in predictions of RSL at >130,000 points in space for each time step.
Post-glacial coastal habitat development and retreat prior to the Holocene are inferred from relict features that include the following: (1) drowned mangroves, tidal marshes and coral reefs, the remains of which have been found at around 90 m water depth 64 , corresponding to meltwater pulse 1A (14.6-14.3 ka); and (2) relict features at around 50 m water depth 65,66 , corresponding to a rapid rise in GMSL dating to 11.3-11.0 ka. Mangrove vertical development in relation to Holocene RSLR is based on 78 observations of the timing of the initiation of sustained mangrove peat development 12 . Evidence of post-glacial mangrove expansion is evident on the Sahul Shelf, Western Australia 67 and the Sunda Shelf, Southeast Asia 68 ~12-14.5 ka, a phase ceasing during meltwater pulse 1A (Fig. 1). A relatively brief (~300-year) period of mangrove expansion and vertical development prior to 9 ka is documented in the western Ganges-Brahmaputra Delta 69 , and the Queensland continental shelf 70 , locations of high sediment delivery at the time 69,70 . Our GIA modelling (Fig. 1) suggests RSLR dipped to around 6 mm yr −1 at this time. Mangroves in both sites were drowned during a period in which RSLR increased to circa 7 mm yr −1 ~9 ka. A pan-tropical expansion in mangrove development and sustained vertical adjustment 54,71-75 commenced from ~8.5 ka as RSLR declined 12 to <6 mm yr −1 according to GIA modelling (Fig. 1). A subset of these observations representative of the global dataset 12 are provided in Fig. 1 81 . Reef island development commenced in the Pacific during the mid-Holocene corresponding to RSLR stabilization and fall. The example provided in Fig. 1 is Bewick Cay, Queensland, Australia 24 .
Tidal marsh vulnerability to Holocene RSLR presented in Fig. 2 is based on the data from two studies utilizing multiple proxies across the UK 15 and the Mississippi Delta 16 . Holocene RSL data was compiled for 54 regions from Great Britain with the rate of RSL varying in relation to proximity to the centre of the Last Glacial Maximum British-Irish Ice Sheet 15 . RSLR rates estimated from GIA model predictions were compared to sea-level tendencies for 781 tidal marsh index points 15 (positive n = 403; negative n = 360; no tendency n = 19). For the Mississippi Delta, the Holocene RSL history was inferred from 72 sea-level index points, with marsh tendency assessed using 334 boreholes showing a well-defined Pleistocene-Holocene transition overlain by at least 2 m of sediment 16 .

Contemporary wetland response to RSLR
Assessment of mangrove and tidal marsh vertical adjustment was conducted using the SET-MH technique. This globally distributed network of monitoring stations 82 combines a stable benchmark rod against which measurements of elevation change are made, with an artificial marker horizon introduced at the time of benchmark rod installation, against which sediment accretion is measured (Extended Data Fig. 1). Pins extended from a portable arm (the surface elevation table) extend to the marsh surface, measuring surface elevation change in relation to the base of the benchmark rod. Comparison with elevation gain can then be made against water level changes measured at nearby tide gauges 29 .
This technique was previously used 29 to test how elevation gain at 477 SET-MH monitoring stations compared to RSLR changes measured over the same period. To this analysis (presented as Fig. 2e) we have added a mangrove SET-MH network of 190 SET-MH stations (Fig. 2f), the location of which are provided in Extended Data Table 2. These data combine published rates of elevation gain with new measurements reported here (Extended Data Table 2). RSLR for the period of SET-MH measurement was extracted from tide gauge records provided by the National Oceanic and Atmospheric Administration (https:// tidesandcurrents.noaa.gov/sltrends/sltrends.html) and, for Australia, the Australian Baseline Sea Level Monitoring Project (http://www.bom. gov.au/oceanography/projects/abslmp/abslmp.shtml).

Contemporary habitat distribution
The contemporary distribution and extent of mangroves 83 (https://doi. org/10.34892/07vk-ws51), tidal marshes 84 (https://doi.org/10.34892/ w2ew-m835) and coral reefs 85 (Figs. 1, 3 and 4) was accessed from the Ocean Data Viewer (https://data.unep-wcmc.org), hosted by the UN Environment World Conservation Monitoring Centre. An important caveat in relation to the representation of tidal marshes is the poor coverage of their possible extent at high northern latitudes. For Fig. 3, the coral reef dataset was complemented by additional data on the global distribution of atolls, which was sourced from the World Atolls database 86 (https://www.arcgis.com/home/item.html?id=1c18adf04 d9e47669281061ff60167e1).

Surface water and marsh change analysis
Because mangrove canopy cover obscured surface water observations, we report changes in surface water occurrence, and conversion of wetland to open water only for tidal marshes. We used two earth-observation derived global datasets to estimate tidal marsh conversion to open water across the SET-MH monitoring network. The Global Tidal Wetland Change (GTWC) dataset 36 depicts losses and gains of tidal marshes, tidal flats and mangroves (collectively termed 'tidal wetlands') at 30-m resolution over a 20-yr period (1999-2019). The data were developed through a machine learning classification of more than 1.1 million Landsat scenes acquired over the global coastal zone since 1999. GTWC data layers include tidal wetland losses, gains, and the probability of occurrence of tidal wetlands for the first (1999) and last (2019) time steps of the analysis. The Global Surface Water dataset depicts the location and temporal distribution of surface water from 1984 to 2020 at 30-m resolution 35 . The data were generated from >4.4 million Landsat scenes by individually classifying each Landsat pixel into water and non-water using an expert system. Although the two datasets are developed using Landsat data, the datasets differ in their temporal spans (2 to 4 decades), methodological approaches to mapping change dynamics, post-processing methods and minimum mapping unit. We therefore used both datasets to estimate the extent of tidal marsh conversion to open water in relation to observed RSLR (Supplementary Data 1).
To estimate net tidal wetland change and the extent of conversion to open water at each SET-MH monitoring site, we developed a buffer feature around the SET installation with an area of 5 km 2 . For global tidal wetland change, the area of losses and gains of each tidal wetland ecosystem type (tidal marshes, tidal flats and mangroves) was computed, yielding a net change estimate of tidal wetlands associated with each SET site. For global surface water, we used the water occurrence change intensity layer, which is computed as the absolute difference in the per pixel mean water occurrence between two distinct epochs 35 (1984-1999 and 2000-2020). The average surface water change in each SET buffer feature was computed (Supplementary Data 1).
The relationships between surface water and tidal wetland change versus contemporaneous RSLR and elevation deficit were tested using multiple linear regression. Predictive variables are provided in Extended Data Table 3, and consist of climatic, hydrological and edaphic properties associated with each SET-MH station, and are sourced from ref. 39. Potential collinearity of predictors was assessed using variance inflation factor from the car package 87 . The variance inflation factor was found to be below the level usually considered problematic (3.22). The overall relative importance of the key predictors was assessed using random forest regression analyses 88 , a machine learning approach which tallies the results of small classification trees (n = 20,000) while retaining a bootstrapped subset of all observations for out-of-bag (internal) error testing. Analyses were performed in R version 4.1.3 and presented as Extended Data Fig. 4b.

Island contraction and expansion
Data on island contraction or expansion (Extended Data Fig. 3) were sourced from recent assessments and reviews 37,40,89,90 (total island n = 872: Supplementary Data 2). We compared the proportion of islands showing areal contraction or expansion, binned at 1 mm yr −1 RSLR increments, using the rate of RSLR cited for each reef island in the manuscripts. Islands were considered stable if change was less than 3% of the original area, following ref. 37.

Ecosystem stability under RSLR
Contemporary marsh and mangrove resilience to RSLR was inferred from data from the globally distributed mangrove and tidal marsh SET-MH networks. The elevation surplus or deficit of each site was estimated by comparing the rate of tidal marsh surface elevation change recorded by the SET to rates of RSLR over the period of operation of the SET. RSLR was sourced from the National Oceanic and Atmospheric Administration's Laboratory for Satellite Altimetry (https://tidesandcurrents.noaa.gov/sltrends/). The elevation surplus/deficit of each SET site was categorized in Extended Data Fig. 2d,e as in surplus if surface elevation change exceeded the RSLR rate by 1 mm yr −1 , stable if surface elevation change was within ±1 mm yr −1 of the RSLR rate, or in deficit if the RSLR rate exceeded surface elevation change by 1 mm yr −1 . The stacked histograms in Extended Data Fig. 2d,e show the proportion of elevation budget categories in relation to RSLR rates (1 mm yr −1 bin size) at each tidal marsh (Extended Data Fig. 2d) and mangrove (Extended Data Fig. 2e) SET site.
The resilience of palaeo tidal marsh to RLSR is represented in Extended Data Fig. 2a,b for the UK and Mississippi Delta, respectively. A 'negative' sea-level tendency-indicating tidal marsh advance-is identified by decreasing marine influence (that is, regressive contact), whereas a 'positive' sea-level tendency-which indicates tidal marsh retreat-is identified by increasing marine influence (that is, transgressive contact) in sediment archives. In the example core (Extended Data Fig. 2a), the contact between an intertidal mud and tidal marsh peat, which represents a negative tendency and marsh advance, was dated to ~8,439-8,956 years ago. The thin accumulation of tidal marsh peat is overlain by an intertidal mud, representing a positive tendency and marsh retreat; this event was dated to 8,501-8,959 years ago. RSLR rates were estimated for the timing of these marsh advance and retreat events recorded in the stratigraphy using a GIA model. The stacked histogram (Extended Data Fig. 2a) shows the proportion of these events from sediment archives across the UK in relation to RSLR rates (0.5 mm yr −1 bin size). The facies succession identified in sediment cores from the Mississippi Delta (Extended Data Fig. 2b) were categorized based on the following criteria: a 'terrestrial' succession-indicating no evidence of marsh 'drowning'-is associated with the presence of terrestrial (marsh) mud or peat throughout the core and an absence of lagoonal facies; 'gradual drowning'-indicating marsh drowning that occurred over centuries-identified by at least a 30-cm-thick unit of marsh mud or peat occurring beneath lagoonal mud; 'rapid drowning'-indicating marsh drowning that occurred over about half a century-associated with less than a 30-cm-thick unit of marsh mud or peat occurring beneath lagoonal facies. The contact between marsh and lagoonal facies representing gradual or rapid marsh drowning was radiocarbon dated to determine the timing of the event, and RSLR rates at that time were estimated from an RSLR record obtained from compaction-free basal peats from the Mississippi Delta. The proportion of each type of facies succession is shown in comparison to estimated rates of RSLR (0.5 mm yr −1 bin size).
The 'initiation' of sustained mangrove accretion (Extended Data Fig. 2c) (at least 2 m of mangrove sediment) was radiocarbon dated and RSLR rates at that time interval were estimated from an ensemble of GIA model predictions 12 . The histogram shows the probability density (distribution) of initiation events in relation to RSLR rates (1 mm yr −1 bins).
We summarized the probability thresholds at which marsh or mangrove elevation deficit becomes likely (P ≥ 0.66) or very likely (P ≥ 0.90), adopting IPCC likelihood language 91 . To estimate the probability of a negative tendency (Fig. 2b,c) or elevation deficit (Fig. 2e,f) conditional on rates of RSLR, we follow ref. 15 by modelling the elevation budget or facies successions as binary response variables (elevation deficit or drowning, 1; elevation surplus or terrestrial, 0) in a Bayesian framework. We chose the bin widths for histograms and the number of segments in the Bayesian analysis by visual inspection for best fit. Details of the probabilistic analysis used to estimate the relationship between mangrove initiation and RSLR rates (Fig. 3f) can be found in ref. 4.

Sea-level rise projections
Sea-level rise projections (Figs. 3 and 4) were those used in the IPCC AR6 and were sourced from https://doi.org/10.5281/zenodo.5914710 42 . Sea-level rise scenarios to 2100 were converted to point shapefiles for the median, 17th and 83rd percentile projections for the following warming-level-based scenarios: 1.5 °C; 2.0 °C, 3.0 °C, 4.0 °C and 5.0 °C. The 17th-83rd percentile ranges are associated with the assessed IPCC likely range; the IPCC assessment is that there is at least a 66% chance that the true value will fall within this range. From the AR6 sea-level rise scenarios, sea-level rise rates at 2100 were converted to raster format (cell size 1 degree) for the median, 17th and 83rd percentile projections for the above-listed temperature-limited scenarios. All land-based pixels (defined as pixels where sea-level rise rates were zero for all percentiles of one temperature scenario) were converted to NoData.
Ecosystem exposure to projected RSLR For Fig. 3, available polygons for tidal marshes, mangroves and coral reefs were converted to point files based on each polygon's centroid coordinates. Where polygon features consisted of multiple polygons, polygon features were split into single-polygon features before converting them to centroid points. All resulting polygon centroids were merged with the available point data for mangroves, tidal marshes and reef islands into a dataset containing 1,885,466 entries. To visualize the spatial variability of wetland exposure to local sea-level rise (Fig. 3a-d), local RSLR rates (incorporating vertical land movement) 42 were extracted from the median projections of the temperature-limited scenarios 1.5 °C, 2.0 °C, 3.0 °C and 4.0 °C and classified into the following RSLR rate exposure categories: <0 mm yr −1 (blue), 0-4 mm yr −1 (yellow), 4-7 mm yr −1 (orange) and >7 mm yr −1 (red).
To calculate proportional changes of exposure to local RSLR rates for all five temperature-limited scenarios (Fig. 3e-g), only the available polygons for salt marshes, mangroves and coral reefs were utilized, as those included accurate aerial information. As above, polygon data were converted to point files, based on their centroid locations, but to preserve the accurate aerial information multi-polygon features were not split up. All local RSLR rates of each scenario (five temperature scenarios, with three percentiles each) were extracted for each ecosystem category to calculate proportional exposure to RSLR rates <0 mm yr −1 , 0-4 mm yr −1 , 4-7 mm yr −1 and >7 mm yr −1 . For each temperature scenario, the respective uncertainty range was defined by the lower (17%) and upper (83%) percentiles respectively.
Modelling retreat potential AR6 RSLR data up to 2100 were utilized to model the inland retreat space of coastal wetlands available for two RSLR scenarios: the 2 °C and 3 °C warming levels (Extended Data Fig. 6). The 3 °C warming level, representing the greater potential landward retreat, is also presented in Fig. 4. This modelling relies on the global coastal wetland model, which assumes inland retreat can occur where local population densities are below a pre-defined population density threshold and the coastal topography provides sufficiently flat inland areas 14 .
AR6 RSLR data (RSLR between 2020 and 2100) were retrieved from https://doi.org/10.5281/zenodo.5914710 42 with a spatial resolution of 1 degree. Trajectories of RSLR for each coastal segment, the spatial unit that the global coastal wetland model is based on ref. 92, were derived from the data point located closest (Euclidean distance) to the center of the respective coastline segment. Inland retreat space was calculated as the area additionally inundated during mean high water spring conditions, under future RSLR scenarios, and expressed as percentage of current wetland extents 14 . High water spring levels were thereby assumed to rise at the same rate as mean sea level. Local topographical profiles were calculated based on global Shuttle Radar Topography Mission data 93 and on the method first presented in ref. 94.
Taking into account the widespread obstruction that human coastal infrastructure imposes on coastal wetland inland retreat 95 and assuming that the extent of obstruction is a function of population density, wetland inland retreat was accounted for only where population densities within the local 1-in-100 year floodplain are below a threshold of 20 people per km 2 as a best case of these scenarios, a threshold of 5 people per km 2 as a worst case scenario. This range has previously been estimated to represent current conditions for the existence of barriers to coastal wetland inland retreat 14 . Meanwhile, population density has been subjected to estimated population growth following the 'middle-of-the-road' shared socio-economic pathway (SSP2) 96 . We also modelled potential landward space available for ecosystem redistribution ignoring the potential impediment of population density, the no barriers scenario (Fig. 4).

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Extended Data Table 2 | Location and timing of mangrove SET-MH measurements, with corresponding rate of RSLR and elevation gain *Original SET Data; **RSLR for the period of SET-MH measurements, from the nearest tide gauge.

Extended Data Table 3 | Identifiers and Variables used in the RF regression analysis (Data 29 )
Corresponding author(s): Last updated by author(s):

Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.
Please do not complete any field with "not applicable" or n/a. Refer to the help text for what text to use if an item is not relevant to your study. For final submission: please carefully check your responses for accuracy; you will not be able to make changes later.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated Our web collection on statistics for biologists contains articles on many of the points above.

Software and code
Policy information about availability of computer code Data collection

Data analysis
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy  Tick this box to confirm that the raw and calibrated dates are available in the paper or in Supplementary Information.

Ethics oversight
Note that full information on the approval of the study protocol must also be provided in the manuscript. Policy information about dual use research of concern

Hazards
Could the accidental, deliberate or reckless misuse of agents or technologies generated in the work, or the application of information presented in the manuscript, pose a threat to:

Experiments of concern
Does the work involve any of these experiments of concern: No Yes Confirm that both raw and final processed data have been deposited in a public database such as GEO.
Confirm that you have deposited or provided access to graph files (e.g. BED files) for the called peaks.

Data access links
May remain private before publication.

Files in database submission
Genome browser session The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots. This checklist template is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/