Our species distribution models for 21,146 tetrapods and vascular plants across twelve megadiverse countries project considerable changes in species richness (ΔSR) and composition (βSØR, see Methods) at three consecutive temporal horizons: 2030–2060, 2050–2080, and 2070–2100 (hereafter referred to as 2030, 2050, and 2070, respectively). These time horizons should be interpreted as relative to the onset of the melting simulations in 202017,22. As previously observed for amphibians15, the largest differences in species responses to climate change are observed between the global warming scenario (RCP 8.5) and any of the four melting scenarios. Given that differences between melting scenarios across the groups are small – even under the less drastic melting scenario we project drastic alterations to biodiversity –, we only show in the following the results obtained under scenarios RCP 8.5 and Melting 0.5 (Supplementary table S1), in order to gain clarity.
Our results agree with previous analyses showing a general decline in species richness under scenarios of global warming9,24,25. Here we show that melting has a large added impact on the observed patterns of biodiversity, probably as a result of induced major changes to temperature and precipitation22. The expected loss in species richness (negative ΔSR) appears to be more widespread with the added effects of melting than with global warming alone12, with some exceptions such as Indonesia, India, Philippines and China (figure 1). These results support the heterogeneous impacts of Greenland’s melting and associated AMOC weakening across the world, where melting can sometimes mitigate the impacts of changes induced by global warming22. However, excepting China, the above-mentioned countries have scarce biological data, thus geographical patterns of ΔSR should be interpreted with caution. Importantly, summarizing changes across plant and animal groups shows that there are differences in the geographic patterns of ΔSR between groups (Supplementary figures S1–S2). Overall, negative ΔSR are enhanced under melting scenarios but a contrasting pattern emerged for China, where the gain in species richness (positive ΔSR) increases by 2030 (figure 1); notwithstanding, even this trend of increasing species richness is reversed by 2050 and 2070.
Based on our results, we suggest that on average, regions with more humid and temperate climates (e.g., mountain ecosystems) are predicted to be more vulnerable to climate change than regions with more seasonal and warmer climates26-29. These results coincide with the predicted reduction in Equatorial and warm temperate climates (A and C in the Köppen classification) across the world under melting scenarios22. For example, areas with positive ΔSR, both for tetrapods and vascular plants, are generally characterized by higher (or similar) mean annual temperatures and less (or similar) seasonality than areas with negative ΔSR, with the exception of Chinese tetrapods that show the opposite pattern (Supplementary figure S3–S6). Differences in precipitation between areas of loss and gain of species richness are heterogeneous across countries. In general, areas with negative ΔSR for vascular plants appear to be more (or equally) humid (i.e., high annual precipitation and low seasonality) than areas with positive ΔSR; this pattern, however, appears to be less consistent, or even opposite, for tetrapods (Supplementary figures S4, S6).
The vast majority of the species modeled are distributed within or around biodiversity hotspots9, including important mountain regions around the world. In this context, we show that the geographic extent of potential species hotspots (PSHs) across countries decreases – relative to their present extent – under global warming (median reduction, 48–88%) and are magnified with the added contribution of Greenland’s melting (median reduction, 74–95%) (figure 2; Supplementary table S1). Although most countries show decreasing trends of PSH extent, China shows marked increases under global warming (153–250%). This trend is reverted under melting scenarios, in which all twelve countries show sharp declines in the extent of PSH shortly after the onset of freshwater release into the North Atlantic (Supplementary table S1). In addition, our results indicate that PSHs would also be subjected to moderate to high changes in species composition (Extended Data figure 2), with a median temporal dissimilarity (βSØR) of 0.26–0.89 across scenarios. We observe an increased impact of Greenland’s melting – and the ensuing weaker AMOC – on species composition (Supplementary table S2); for instance, the median temporal dissimilarity is considerably lower under global warming alone (median dissimilarity, 0.18–0.51) than with the added contribution of melting scenarios (median dissimilarity, 0.20–0.90).
As expected, the potential species hotspots we defined coincide with globally important biodiversity hotspots10, which harbor climatically vulnerable species not found anywhere else in the world30. Thus, although our modeled species represent a small fraction of global diversity, the alterations to geographic extent of PSHs and the alteration of species composition is an alarming possibility. Based on our models, we suggest a dramatic decline and alteration of biodiversity across hotspots within a relatively short period of time (10–40 years; see figure S2 in Velasco et al.15) after the onset of Greenland’s melting, and the ensuing weakening of the AMOC. More than twenty years ago, Myers et al.10 estimated that effectively protecting these biodiversity hotspots, collectively encompassing less than 2% of the Earth’s surface, would translate into the protection of 44% of vascular plants and 35% tetrapods. However, our results indicate that, even if climate mitigation and ecosystem protection and restoration take place, the world’s biodiversity hotspots are highly vulnerable in the face of tipping points pushing the climate system into a new state16,31.
The projected reduction in species richness and alteration to species composition highlight the threat for biodiversity posed by global warming and the additional contribution of Greenland’s melting15. Furthermore, the declines in species richness and alteration to species composition are associated with projected reductions of the geographic ranges of individual species, which are magnified under melting scenarios (figure 3). We project that under global warming alone, half of the vascular plant and tetrapod species will experience range reductions of at least 31–83%, relative to present-day distributions (Extended Data figure 3; Supplementary table S3–S4), with reductions across individual countries ranging from 10–38% (Peru) to 54–79% (Brazil). However, we estimate range expansions in more than half of the species in China, Colombia, and Venezuela (median expansion, 102–120%), albeit only for 2030 (Supplementary table S4). Nearly all range expansions are reversed towards range reductions with the added contribution of Greenland’s melting (median range loss, 95–99%). Under melting scenarios, species with a projected extreme to complete loss of geographic ranges (more than 80% reduction) increase in number relative to the RCP 8.5 scenario (Extended Data figure 4). More specifically, species ranges for the eight taxonomic groups are reduced drastically after the onset of freshwater release (Extended Data figures 3–4), with a median range reduction of 58–99% by 2030, increasing to 67–100% by 2070 (Supplementary tables S3); these projections mean that in the worst-case scenario, half of the species will suffer complete range reductions.
It is safe to assume a direct relationship between the loss of suitable areas and species’ extinction risks32, where species with reduced geographic ranges are more prone to extinction than wide-ranging species. In the present case, the complete disappearance of suitable areas for species within megadiverse countries, which is exacerbated by melting, is of paramount concern because these losses entail high probabilities of species being fully extirpated from their native range and going extinct globally. Our results indicate that, in general, plant species have slightly higher risks of extinction (median range loss, 37–100%) than animal species (median range loss, 30–96%) across all scenarios (Supplementary table S5). Our models show a complete disappearance of climatically suitable areas for 1,239–4,483 species (6–21% of the total) under global warming, which increases to 7,728–10,312 species (36–49% of the total) with the added contribution of Greenland’s melting (Extended Data figure 4, Supplementary figures S7–S8).
The general pattern of complete range reduction is ubiquitous across countries, but with some variation (figure 4, Supplementary table S6). For instance, Brazil and Australia show the highest proportion of species (relative to the species modeled per country) projected with complete range loss (proportion of species, 8–60% and 7–59%, respectively), which is consistent with substantial alterations to regional climates predicted under melting scenarios22. On the other hand, India and Philippines show the lowest proportion of species with complete range loss (proportion of species, 3–18% and 2–19%, respectively) (Supplementary table S6). The estimated changes in our evaluated variables of mean annual temperature and annual precipitation under the different climate models cannot account for the spatial and temporal heterogeneity observed in species responses across countries (Supplementary figures S9–S11); this despite that these variables (mean annual temperature and annual precipitation) are two of the most important contributors to our species distribution models (Supplementary figure S12–S13). This highlights the idiosyncratic response of species to climate change, yet our results suggest that high global warming and ice sheet melt can have an overarching impact on biodiversity and the climate system15,22, leading to worldwide drastic alterations to climate and biodiversity loss.
The projected range losses would have a considerable impact for the future of the worlds’ biodiversity and, in the case of South American countries, further increase the global concerns on the region’s deforestation trends14,33, an alternative large-scale singular event. On the other hand, island countries, such as Indonesia and the Philippines, are projected to have smaller reductions in species ranges. In part, we think this is the result of a lower number of modeled species, which we believe to be related to a lack of publicly available biodiversity data in these countries, rather than an inherent lower climatic vulnerability15,22. Data limitations in some of these countries limit the assessment capabilities of species vulnerability, thus hindering conservation and mitigation planning34. Sources of uncertainty in prediction, including modelling uncertainty, need to be taken into account when assessing vulnerability of regional biodiversity under climate change35-37 and to define conservation priorities38. In this respect, although the species distribution modelling is based on a single global circulation model and inter-model variability is not considered, the resulting species models take into account modelling uncertainty by relying on an ensemble approach with seven different algorithms. Thus, our species distribution models allowed us to assess for the first time the probable consequences of Greenland’s melting on biodiversity.
Based on our results, we suggest an additional domino effect might be triggered in the potential cascading effects of anthropogenic climate change, where biodiversity will be crossing a tipping point in response to Greenland ice sheet thawing. In this context, alterations to species composition and decreasing species diversity are expected to have further cascading effects on biological interactions and ecosystem functioning, further increasing the probability of species’ extinction37. In this context, we show that on average, vascular plants are expected to be the most vulnerable (Extended Data figure 3), but this might partially be the result of the small number of species modelled for ferns, gymnosperms, and lycophytes. Nonetheless, flowering plants are by far the most abundant group in our dataset – 15,162 species – and apart from the three aforementioned groups, it is the most vulnerable group in terms of complete loss of species geographic ranges. The fact that flowering plants are projected to suffer substantial negative impacts increases the concerns about the vulnerability of biodiversity in general, due to possible future alterations to ecological interactions and ecosystem functioning8, 39,40. In turn, the two most vulnerable tetrapod groups are amphibians and reptiles15,41-43, yet mammals and birds also appear to be highly vulnerable in some countries (figures 4). These groups of animals may be further impacted due to their dependence on unaltered and diverse forest ecosystems. In this context, the climatic vulnerability of flowering plants – which are the ecological basis of most terrestrial ecosystems – can potentially increase further due to the collapse of tetrapod diversity, which includes many pollination and dispersal vectors. The collapse of flowering plant diversity will likely increase the extinction risks of other ecologically linked groups8,39, even when these appear to be less vulnerable to climate change, such as birds in Andean countries (figure 4).
Based on the projected changes to the geographic range of thousands of animal and plant species, we suggest a likely tipping point in the collapse of biodiversity across the world’s most megadiverse countries in response to unrelenting global warming. These impacts would be magnified and be reached relatively quickly (10–40 years), if a substantial amount of freshwater from Greenland’s ice sheets is released into the North Atlantic. In light of recent observations of a substantial ice sheet loss and of the AMOC being currently at its weakest point in millennia17,18, our projections provide reasons of major concern for the future of species across major biodiversity hotspots. The effects of global warming and Greenland’s melting would be reflected in substantial reductions to species diversity and alteration to species composition. If current trends of climate change continue – resulting from unrelenting greenhouse gas emissions – these effects have the potential to lead to substantial alterations to ecological interactions and ecosystem functioning.