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

Abstract

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.

Fig. 1. Coastal ecosystem responses to RSLR following the Last Glacial Maximum.

a, Present-day distribution of mapped coastal ecosystems and the location of case studies highlighted in b,c. b, Median rates of RSLR over time derived from GIA modelling (Methods). c, Timing of habitat advance and retreat for a selection of locations (Methods). Mangrove, tidal marsh and coral reef island development is predominantly associated with periods of RSLR of less than 7 mm yr⁻¹.

Fig. 2. Probability of vertical adjustment of mangrove and tidal marsh to rising sea levels.

a–c, Palaeo-stratigraphic assessments (a) of marsh adjustment or retreat for tidal marshes (b) and the probability of the initiation of sustained mangrove vertical adjustment (c), in relation to rates of RSLR encountered over the past 10,000 years. b, The red line represents minerogenic UK marshes and the orange line represents organic marshes of the Mississippi Delta. d–f, Results in a–c are compared with vertical adjustment as assessed by the surface elevation table (d), analysed for the probability of a deficit between vertical adjustment and RSLR for the same period of measurement, for 477 tidal marsh SET-MH installations (e) and 190 mangrove SET-MH installations (f). Adopting IPCC likelihood definitions (Methods), we indicate in each case the probability thresholds at which mangrove or marsh drowning becomes likely (P ≥ 0.66) or very likely (P ≥ 0.90). The corresponding histograms for each RSLR increment are shown in Extended Data Fig. 2. b,c,e,f, The line represents the median and the shaded region shows 90% confidence interval (CI).

Fig. 3. Projected exposure of coastal ecosystems to RSLR.

a–d, Coastlines with mapped mangrove, tidal marsh or reef habitat subject to >4 mm yr⁻¹ and >7 mm yr⁻¹ RSLR over 2080–2100 under the median projections for 1.5 °C (a), 2.0 °C (b), 3.0 °C (c) and 4.0 °C (d) warming scenarios relative to 1850–1900. 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.

Fig. 4. Wetland inland retreat potential.

The percentage of the current wetland area that could potentially be compensated for via inland retreat until 2100, calculated for the 3.0 °C warming scenario (Methods). a, The scenario for wetland inland retreat capacity possible with a population density below 20 people per km². b, The scenario for wetland inland retreat unimpeded by population density (the no barriers scenario; Methods). Scenarios restricting landward encroachment under lower populations density thresholds are shown in Extended Data Fig. 6.

Extended Data Fig. 1. Components of the Surface Elevation Table - Marker Horizon system.

A benchmark rod driven to the point of refusal serves as a vertical benchmark against which tidal marsh/mangrove elevation gain or loss is measured. At the time of installation, an introduced horizon (feldspar or similar) is placed on the wetland surface, against which sediment and organic accretion is measured. These measurements allow for the inference of shallow subsidence, and the relation between elevation gain and RSLR.

Extended Data Fig. 2. Details of individual studies from which the probabilities of marsh and mangrove vertical adjustment to RSLR were inferred.

(a) Analysis of tidal marsh retreat and advance recorded in sediment archives from across the United Kingdom. (b) Analysis of marsh drowning recorded in sediment archives from the Mississippi Delta. (c) Analysis of global mangrove accretion recorded in sediment archives from. (d) Analysis of surface elevation tables (SET) from global tidal marshes. (e) Analysis of surface elevation tables (SET) for mangrove SET sites in Extended Data Table 2. For a, b, d, and e probabilistic analysis shown in Fig. 2 follows and, where the probability of an elevation deficit/surplus, marsh retreat/drowning, or rapid drowning/terrestrial succession was modelled as a binary response variable, and the relationship of this response with rates of RSLR was estimated in a Bayesian framework. Details of the probabilistic analysis used in e to estimate the relationship between mangrove initiation and RSLR rates (c) can be found in. Numbers of observations for each RSLR increment are shown at the base of each column.

Extended Data Fig. 3. Probability of conversion to open water with RSLR.

Tidal marsh area change 1999-2019 within a 5 km² of tidal marsh SET-MH sites in relation to RSLR (a). Change in surface water occurrence at the tidal marsh SET-MH sites, comparing 20 years pre-2000 and post-2000 in relation to RSLR trends during the period of SET measurement (b). Reef island planform change in relation to RSLR (c: area change and RSLR data from studies in Supplementary Information Data 2). Numbers of observations for each RSLR increment are shown at the base of each column. Image data in (a) are from Google Earth.

Extended Data Fig. 4. Surface water area in tidal marshes increases with RSLR and elevation deficit.

A: Normalised surface water change (change in the % occurrence of surface water) comparing the two decades pre-2000 and post-2000 at the tidal marsh SET-MH monitoring sites (n = 476). The presence of surface water increases with RLSR (r² = 0.16, P<0.001). Grey shading indicates the 95% confidence level interval for linear model predictions. B: Results of Random Forest analysis predicting the normalised surface water change at the SET-MH monitoring sites, based on predictive variables sourced from (listed in Extended Data Table 3).

Extended Data Fig. 5. Marsh elevation, elevation deficit and surface water change.

Normalised surface water change (change in the % occurrence of surface water) comparing the two decades pre-2000 and post-2000 at the tidal marsh SET-MH monitoring sites, in relation to the deficit between elevation gain and RSLR. For marshes above the median marsh elevation of 14 cm above the lower limits of survival (a), the relationship is weak (linear regression r² = 0.03, P = 0.009; n = 298). For marshes below this median marsh elevation (b), normalised surface water change increases with the size of the elevation deficit (linear regression r² = 0.187, P < 0.001; n = 231). Elevation capital is the elevation of the marsh above the lower limits of survival, at which open water conversion would be expected.

Extended Data Fig. 6. Inland retreat potential of existing mangrove and tidal marsh.

Percentage of the current wetland area that could potentially be compensated for via wetland inland retreat until 2100, calculated for two sea-level rise scenarios. Median projections for 2.0 °C warming, allowing wetland retreat to a population threshold of (a) 5 people km⁻² (worst case scenario, WC) and (b) 20 people km⁻² (best case scenario, BC) Median projections for 3.0 °C warming, allowing wetland retreat to a population threshold of (c) 5 people km⁻²; and (d) 20 people km⁻².