Introduction

Climate change has accelerated sea level rise (SLR) since the 1970s and is now more rapid than the mean SLR rate recorded during the previous two millennia1,2,3. By the end of this century it is projected that SLR will reach 82 cm2 and—in extreme scenarios—could exceed 2 m3 with the early onset of Antarctic ice sheet instability4. Regional variation in predictions in SLR show that tropical regions and small islands are among the most vulnerable5, threatening species that depend on these coastal habitats, such as sea turtles6,7,8. Sea turtle species exhibit natal philopatry, returning to the beach where they were born9 with exceptionally high precision for returns to island rookeries10. However, climate changes might be too rapid for sea turtles to respond through their ability to disperse or colonize new habitats11. These biological traits and their reliance on sandy beaches make them particularly vulnerable to changes in coastal areas, like those resulting from SLR.

As a result, concern exists on the potential impacts of SLR on sea turtles, however only a dozen studies to date have projected how SLR will impact them12. These previous studies have been mainly regionally-focused, including assessments from only one or two species7,13,14,15,16. This regional focus is likely a result from the challenges inherent in successfully assessing shoreline response to SLR17,18. Although most sea turtle assessments have been obtained from field survey methods, such studies of estimations of stream reach water surface slopes often have low accuracy19. Other approaches which couple LiDAR with biological data20,21 have higher accuracy, but are also more costly22. However, new methodologies, such as the use of open Digital Elevation Models (DEMs) might be a good proxy broadly applicable to assess SLR by satellite images23.

Considering that most sea turtle rookeries across the globe are located in remote areas in low and middle-income countries, less costly approaches for field surveys are often preferred and can provide baseline data to identify areas at most risk. Indeed, the few studies assessing the impacts of SLR on sea turtles to date discuss the challenges inherent in successfully predicting shoreline response to SLR and storm activities17,18 and the inability to couple projections with biological information such as sex ratios and reproductive success12. Here we present an assessment of the potential impact of SLR on sea turtle rookeries by applying a low-cost methodology to estimate the probability of flooding of nest locations under multiple IPCC SLR scenarios. This approach combines turtle nest locations, freely available DEMs and Climate Central maps under Coastal DEM predictions22. The study considers seven sea turtle rookeries with five nesting species, categorized from vulnerable to critically endangered24 including leatherback turtles (Dermochelys coriacea), loggerhead turtles (Caretta caretta), hawksbill turtles (Eretmochelys imbricata), olive ridley turtles (Lepidochelys olivacea) and green turtles (Chelonia mydas) (Fig. 1; Extended Data Table 1). Our study sites encompass some important nesting sites for sea turtles globally (e.g., Raine Island, Australia, the largest green turtle rookery worldwide25) and have different characteristics (i.e., beach width, slope, size), which will allow us to estimate SLR effects on a wide range of nesting rookeries and highlights the broad applicability of our approach.

Figure 1
figure 1

Vulnerability of sea turtle nests under sea level rise at IPCC’s RCP 4.5. Seven study sites at sea turtle rookeries spanning countries in the Caribbean Sea, Gulf of Mexico and Pacific Ocean, with five species represented. In the Caribbean and Gulf of Mexico: Mondonguillo beach, Costa Rica; Guanahacabibes peninsula, Cuba; Saona Island, Dominican Republic; Zeelandia, Turtle, Kay bay, Tumbledowndick, Crooks and Oranjebaai beaches, St Eustatius; and St George Island, Florida, USA. In the Pacific Ocean: Coast of Ecuador; and Raine Island, Australia. Map Data: Google Earth and free images.

Full size image

Vulnerability of sea turtle nests to SLR

From GPS locations of 2835 marine turtle nests belonging to five different species from seven study areas across the globe, we estimated the vulnerability of nests to flooding considering available projections of SLR caused by climate change from 2010 to 2100 (Fig. 1). Firstly, we used data from GPS nest locations + available DEMs + IPCC projections for the seven studied populations across the globe to assess and compare the suitability of the different DEMs to make flooding predictions. The comparison of freely available elevation data shows that the mean estimated elevation for nest sites (and consequently mean proportion of flooded nests) differed substantially between DEM sources for each rookery (Extended Data Fig. 2). It was expected that the highest resolution (30 m) DEMs would be most accurate (compared with lower resolutions of 90–100 m) but there were similarities in accuracy amongst DEMs for each study site (Fig. 2; Extended Data Figs. 3 and 4). In addition, Cohen’s kappa results showed weak relationships between DEM model predictions and the in situ data (Fig. 3; Supplementary Table 3).

Figure 2
figure 2

Summary of model predictions. Predictions of the models for different DEMs and empirical data (in situ) indicate the best fit of probability of nest flooding for (a) Costa Rica and (b) Ecuador turtle nests. IPCC sea level rise predictions are included for RCP 4.5 (blue) and RCP 8.5 (red). For Ecuador, the RCP climate change scenario did not influence the probability of nest flooding. Cohen’s kappa (κ) values for prediction (or not) of flooding for each nest compared in situ data with respective DEM datasets. Bold values correspond to a small but significant relationship between the in situ and DEM data.

Full size image
Figure 3
figure 3

Summary of actual vs. estimates of slope and elevation by DEM. Comparison of the relationship between slope (radians) and elevation (m) using empirical data (in situ measurements) and DEMs for turtle nesting beaches in Costa Rica.

Full size image

Then, we estimated the probability of flooding from GPS nest locations georeferenced on CoastalDEM maps (e.g., Extended Data Fig. 5) for five of the seven study areas, since CoastalDEM did not provide reliable estimates in Costa Rica and Ecuador (e.g., Extended Data Fig. 6). For the other five study areas, predictions from Coastal DEM maps (Fig. 4) and from basic in situ data (Supplementary Tables 8–15) were reliable to predicting nest flooding under SLR. Raine Island (Australia) and Saona Island (Dominican Republic) are the most vulnerable populations with 100% nest flooding predicted under moderate emissions scenarios for 2050 (Fig. 1; Supplementary Table 4). For Florida, flooding probability greatly increases after 2050 (Fig. 2a; Supplementary Table 5). Cuba, probably thanks to their elevated beaches, showed the lowest vulnerability to flooding throughout the twenty-first century (Fig. 2b; Supplementary Table 6). For St. Eustatius, main differences in flooding vulnerability arise from differences between turtle species (see Fig. 4c and next section).

Figure 4
figure 4

Predictions of sea turtle nest flooding in a subset of rookeries. Sites include (a) Florida (USA) (2050 and 2100), (b) Cuba (2050), and (c) St Eustatius (three species, 2050). Column 1: Proportion of nests likely to be flooded at each location. Column 2: Heatmap of nests predicted to be free from flooding by 2050. Column 3: Heatmap of nests predicted to be flooded by 2050, identifying the areas at highest risk. Probability values of 1 (yellow), 0.5 (blue) and 0 (purple) represent the density of unflooded or flooded nests within respective heatmaps. A Climate Central moderate scenario was adopted for these scenarios (Kopp et al. 2017).

Full size image

Overall, our results suggested that flat islands and cays may be highly vulnerable to sea level rise under the moderate IPCC scenario (Fig. 1; Supplementary Table 4)7,26 predominantly those in the Caribbean and Pacific27,28. Some of these locations are important rookeries25,29 for species which return every 1–4 years to nest at the beach where they were born9. Furthermore, some of them host nestings for more than one species and present several nesting environments and beach characteristics such as different slopes, width, or sand grain size, among others30,31,32.

A number of rookeries subjected to beach erosion have already been assessed as vulnerable due to loss of beaches used for nesting8,33, nest loss21,32 and changes in nesting behaviour16. The philopatry of leatherbacks34 and loggerheads is not quite strict35 and they can move great distances and nest further up the beach in response to SLR depending on future beach availability. However, it has already been reported that 20% of Costa Rican leatherbacks nest in flooded areas when scarp barriers were present30. The expected habitat loss rates found in the study areas could have important effects on nesting success since philopatry could lead many individuals to nest on inundated beaches. An unknown variable is the potential for sea turtles to adapt to new scenarios. Therefore, their survivorship will depend on their resilience and adaptability to rapid changes within their nesting habitats36.

Vulnerability of nests across species

To determine differences between modelled nest flooding probabilities by species we assessed rookeries containing multiple species (i.e., St Eustatius, Dominican Republic and Ecuador rookeries; Fig. 1). Overall, we identified that leatherback turtle nests may be at a higher risk from flooding compared to the rest of the studied species because they tend to nest in open areas near to the high tide line (Fig. 4c). In St Eustatius, significant variation exists between nest flooding probability by species (Supplementary Table 7): leatherback turtle nests are at the highest risk from flooding compared with green turtles (ß = 1.892, p < 0.001; Fig. 4c), and there was no significant difference between flooding risk for hawksbill and green turtle nests (ß = 0.388, p = 0.339; Supplementary Table 3). The model predicted that, on average, 50.0% leatherback, 18.2% hawksbill and 13.1% of green turtle nest locations would be flooded by 2050 (Fig. 1; Supplementary Table 4). The predicted model accuracy for estimation of likelihood of nest flooding was 79.4%. Leatherback turtle nests in open areas of beaches have been found to already been subject to occasional flooding30, whilst hawksbill and green turtles tend to nest at higher elevations closer to dunes and steep cliffs along the coastline37 and olive ridleys nest at open beaches with low slopes38. This difference in nesting location might explain those differences in nest location flooding potential between species. In consequence, leatherback turtle populations as well as other turtles nesting at open beaches (e.g., loggerhead turtles) may be at greater risk from SLR than other species.

In Ecuador, no differences in modelled nest flooding severity were found between olive ridley, greens and hawksbill species, potentially due to beaches at this location being steeper32 (Supplementary Table 15). Similarly, no difference in nest flooding was predicted between nests of different species in the Dominican Republic, although only because all locations were estimated to be inundated by 2050 (Fig. 1; Supplementary Table 3; Supplementary Table 4). It is likely that the potential effect of SLR will vary between species, potentially linked to variables such as nesting beach characteristics (e.g., slope, aspect39), nesting habitat preferences and suitable nesting areas33.

Conclusions—conservation concerns

Our models of nest flooding validated by field data considered that even a moderate increase in greenhouse emissions (RCP 4.5)3 might impact the reproductive output of sea turtles at the rookeries included in our study. Recent predictions of accelerating global SLR due to rapid melting of ice in Greenland40 and the Antarctic41 in combination with ocean currents42 indicate that pessimistic scenarios could be more accurate than conservative scenarios43. Such scenarios support our projections by indicating that sea turtle nesting populations could be vulnerable to flooding under even moderate scenarios over the next decades.

Relatively recent methods of remote sensing and modeling including DEMs4,43,44, drones, photogrammetry and GPS have been adopted to assess impacts of SLR on sea turtle populations12. However, most highly accurate methodologies entail high costs (e.g., 1500–15,000€ per satellite image) limiting their use to more localised studies. Considering that most sea turtle nesting populations around the world are located in low and middle-income countries, local conservation projects cannot afford the costs of these intensive methodologies to assess the vulnerability of nesting beaches. We have demonstrated that a methodology based on low-cost technological models can be a useful tool for predicting possible future SLR scenarios in important sea turtle nesting areas. We highlighted the utility of global open DEM data with high accuracies for remote areas that could assist with estimation of the vulnerability of sea turtle nesting populations worldwide.

Scientific assessments are essential for prediction of the impacts of future climate scenarios and to assist stakeholders and managers in anticipating extreme scenarios of coastal erosion or flooding, and to predict areas at higher risk of flooding45. Such assessments will help identify conservation refugia and nesting beaches that have greater resilience to climate change36. Although sea turtles have been around for millions of years and would be present in several climate change events, we do not know how their populations might be affected by these projected rapid changes of high loss of nesting sites in the study areas by 2050. Thus, this demonstrates the urgency of developing a multi-species assessment at a global scale in order to develop conservation plans for the most vulnerable populations while there is still time. Conservation management strategies are already in place to enhance resilience to SLR at some nesting beaches, including sand refilling of nesting beaches46 such as in Raine Island, relocation of nests to safe places47 or the protection of hatcheries for rookeries with extreme erosion and flooding31. In addition, we highlighted the need for climate change adaptation measurements to be implemented in management plans considering estimated projections under moderate SLR scenarios.

However, if the world maintains current carbon dioxide emission rates, worst-case scenarios might be vastly underestimated by 3–4 times48 and existing management strategies may then be insufficient to protect the future of many sea turtle populations worldwide. In summary, our study predicts massive flooding at important rookeries in Australia, Dominican Republic, Costa Rica and the USA. These critical areas will face the effects of SLR in the next few decades, meaning that it is now urgent to reduce anthropogenic emissions to safeguard the future of sea turtle populations against climate change and associated sea level rise.