全球变化知识资源中心

Many studies have implied significant effects of global climate change on marine life. Setting these alterations into the context of historical natural change has not been attempted so far, however. Here, using a theoretical framework, we estimate the sensitivity of marine pelagic biodiversity to temperature change and evaluate its past (mid-Pliocene and Last Glacial Maximum (LGM)), contemporaneous (1960–2013) and future (2081–2100; 4 scenarios of warming) vulnerability. Our biodiversity reconstructions were highly correlated to real data for several pelagic taxa for the contemporary and the past (LGM and mid-Pliocene) periods. Our results indicate that local species loss will be a prominent phenomenon of climate warming in permanently stratified regions, and that local species invasion will prevail in temperate and polar biomes under all climate change scenarios. Although a small amount of warming under the RCP2.6 scenario is expected to have a minor influence on marine pelagic biodiversity, moderate warming (RCP4.5) will increase by threefold the changes already observed over the past 50 years. Of most concern is that severe warming (RCP6.0 and 8.5) will affect marine pelagic biodiversity to a greater extent than temperature changes that took place between either the LGM or the mid-Pliocene and today, over an area of between 50 (RCP6.0: 46.9–52.4%) and 70% (RCP8.5: 69.4–73.4%) of the global ocean.

Many studies have suggested that climate influences local species abundance, community structure and biodiversity, phenology and species range in the marine environment^{1, 2, 3, 4, 5}. To understand the magnitude of these changes, we need not only to understand the sensitivity of species and communities to temperature on a global scale, but also to give them a historical perspective. Here, to address this, we have first postulated that the arrangement of life in the oceans is the result of the interaction between the ecological niche and the regional environmental regime^{6, 7, 8, 9, 10}. By implementing fundamental ecological principles (for example, Hutchinson’s niche¹¹, Gause’s principle of competitive exclusion¹²) into a theoretical model, we can create pseudo-communities for any given region of the global ocean⁷. Each pseudo-community results from the aggregation of pseudo-species, each characterized by a unique niche. By focusing exclusively on the thermal niche, it is possible to see how marine biodiversity and its organization in space and time are influenced by climate-induced changes in temperature^{7, 8, 9, 10} (Methods). We test our framework against observed data for foraminifers, crustaceans (copepods and euphausiids), fish (oceanic sharks and tuna/billfish) and cetaceans. This approach is different from previous analyses that applied ecological nichemodels^{13, 14} and also from more recent studies that examined isothermal changes^{15, 16}. These studies were limited at the community level by our poor understanding of the spatial distribution of many species⁸, or due to a lack of biological knowledge, respectively. Having modelled the arrangement of life in the ocean, we then compare biodiversity vulnerability to past (LGM and mid-Pliocene) and contemporary (1960–2013) changes in temperature with future climate change scenarios (2081–2100) to set climate-induced changes in biodiversity into context.

To estimate biodiversity sensitivity, we used a framework based on the MacroEcological Theory on the Arrangement of Life^{6, 7, 8} (Methods). This theory proposes that the arrangement of life results from the interaction between the ecological niche of species and changes in their environment^{6, 7, 8, 10, 17}. A large number of pseudo-species can be generated, each having a unique ecological niche (here a one-dimensional thermal niche), and the interactions of the pseudo-species with the fluctuations in the local environmental regime (here the thermal regime) reconstruct the arrangement of biodiversity in space and time^{7, 8} (Methods). We therefore allowed pseudo-species to colonize any given region of the global ocean provided they could withstand the local annual sea surface temperature (SST). Locally, these pseudo-species collected into pseudo-communities. We found that the biodiversity resulting from this model based on annual SST values (Fig. 1a) was very similar to large-scale biodiversity patterns modelled previously⁷ at a weekly temporal resolution using rectangular niches (r = 0.99; p < 0.01; n = 9,927, n^∗ = 4); this indicates that biodiversity patterns are unaffected significantly by either the absence or consideration of seasonality (annual versus weekly SST), or the niche shape (Gaussian versus rectangular). Correlations between expected and observed global biodiversity patterns for foraminifers were r = 0.74 (p < 0.01; n = 1,040, n^∗ = 7) and r = 0.88 (p < 0.01; n = 8,649, n^∗ = 5) using the Brown University Foraminiferal Data Base¹⁸ (Supplementary Fig. 1) or the gridded data of ref. 19, respectively. The same correlation was r = 0.57 (p < 0.01; n = 433, n^∗ = 13) for copepods; (Supplementary Fig. 1), r = 0.76 for euphausiids¹⁹ (p < 0.01; n = 8,644, n^∗ = 7), r = 0.77 for oceanic sharks¹⁹ (p < 0.01; n = 7,961, n^∗ = 7), r = 0.76 for tuna/billfish¹⁹ (p < 0.01; n = 8,182, n^∗ = 7), and r = 0.54 for cetaceans¹⁹ (p < 0.01; n = 8,649, n^∗ = 13). High-biodiversity regions coincided with areas where communities were composed of more thermophilic species and in those areas, biodiversity was maximum in regions where communities were more eurythermic (Fig. 1a–c). Biodiversity was high in areas where exposure, the magnitude of climate change in a given region (measured by changes in SST), was low (Fig. 1d). Low-biodiversity regions corresponded to areas where communities were more psychrophilic and exhibited a high degree of eurythermy, corresponding to areas where exposure was elevated (Fig. 1b–d).

a–c, Biodiversity (pseudo-species richness; a), degree of thermophily (b) and eurythermy (c) of all pseudo-communities. d, Exposure as measured by the coefficient of variation of annual SST (1960–2013). 19,609 pseudo-species were created and biodiversity (a) is also represented as the percentage of created species. The degree of eurythermy is represented as category, with each category indicating a decile from 1 to 10.

We estimated biodiversity sensitivity to a uniform 2 °C increase in annual SST across the global ocean using different indices (Methods) and we investigated the theoretical sensitivity of biodiversity to temperature without the effect of exposure (that is, the local effect of climate change or variability on annual SST). In this way, local change results from the intrinsic property of the local community. The examination of quantitative biodiversity changes (that is, average of individual changes in expected species abundance) showed that polar communities are more sensitive to temperature; here, a 2 °C temperature rise brought about a change of ~20–21% in the average abundance of all pseudo-species (Fig. 2a). In contrast, subtropical communities are less sensitive to temperature (~17–18% of quantitative change), which may explain why more biodiversity is concentrated here⁷ (Fig. 1a). Equatorial regions have a sensitivity intermediate between the poles and the tropics (~19%), which may indicate why biodiversity diminishes equatorwards (the well-known, hump-shaped biodiversity pattern²⁰). The differences between quantitative biodiversity changes of the polar and the subtropical biomes were small (maximum difference of 4%, which may however represent a large quantitative difference at the community scale when many species are involved; Fig. 1a), suggesting that quantitative biodiversity changes may be more influenced by exposure (external component) than sensitivity (intrinsic component).

a–c, Net quantitative changes in biodiversity (a), local species extirpation (b) and invasion (c). d, Net qualitative biodiversity changes resulting from the difference between species invasion and extirpation. All changes are expressed in percentage.

We assessed biodiversity vulnerability to temperature changes by combining sensitivity with exposure (Methods; Fig. 3). First, we assessed vulnerability to mean year-to-year variability in annual SST for 1960–2013. Only 5.3% of oceanic areas had important (that is, above 5%) quantitative community changes (Fig. 3a and Supplementary Table 1). Mean year-to-year vulnerability was strongest over the Pacific Ocean owing to the El Niño/Southern²² and Pacific Decadal²³ oscillations. Vulnerability was also substantial in the western part of the North Atlantic Ocean and in the North Sea, where it may be affected by the North Atlantic Oscillation²⁴. The analysis showed that a total of 14.1% of the global ocean experienced important species turnover (that is, the sum of species invasion and extirpation per geographical cell standardized by initial pseudo-species richness; Supplementary Table 1). There was no important local extirpation across the global ocean, although species local extirpation was elevated in the equatorial part of the Pacific Ocean (Fig. 3b). A total of 18.8% of the global ocean saw important local species invasion, particularly elevated in the polar biomes (Fig. 3c). The resulting net qualitative biodiversity changes showed an increase in biodiversity polewards and a reduction over permanently stratified areas (Fig. 3d).

a–d, Expected vulnerability of biodiversity to average year-to-year changes in annual SST (1960–2013). e–h, Expected vulnerability of biodiversity to changes in annual SST between 2000–2009 and 1960–1969. Net quantitative changes in biodiversity (a,e), local species extirpation (b,f) and invasion (c,g) are shown. d,h, Net qualitative biodiversity changes that result from the difference between species invasion and extirpation. All changes are expressed in percentage.

We assessed biodiversity changes associated with global warming using 4 Representative Concentration Pathway (RCP) scenarios and 5 different atmosphere–ocean general circulation models between 2081–2100 and the reference period 2006–2013 (Methods). We found that between 42.1 ± 26.5% (RCP2.6) and 94.4 ± 9.8% (RCP8.5) of the global oceans are likely to show important (that is, above 5%) net quantitative biodiversity changes (Supplementary Table 1); these changes appeared in the tropical and polar oceans and were especially pronounced in the Northern Hemisphere, although becoming quasi-global when the warming becomes severe (Fig. 4a–d). Important species turnover may concern between 33.2 ± 20.4% and 94.5 ± 13.4% of the global ocean (Supplementary Table 1). Between 16.5 ± 6.9% (RCP2.6) and 32.3 ± 7.3% (RCP8.5) of the global ocean is expected to exhibit an increase in biodiversity whereas between 6.3 ± 4.8% and 44 ± 9.4% (RCP8.5) of the global ocean is expected to reduce their local biodiversity (Fig. 4h–k and Supplementary Table 1). Although variations of these rates of biological change are substantial among atmosphere–ocean general circulation models, estimations become less variable when the intensity of the warming increases (Supplementary Table 1).

a–f, Net quantitative biodiversity changes between 2081–2100 and 2006–2013 for scenarios RCP2.6 (a), RCP4.5 (b), RCP6.0 (c) and RCP8.5 (d), between the LGM and 1960–1969 (e) and 2006–2013 and thermal conditions corresponding to the mid-Pliocene (f). h–m, Net qualitative biodiversity changes for scenarios RCP2.6 (h), RCP4.5 (i), RCP6.0 (j) and RCP8.5 (k), between the LGM and 1960–1969 (l) and 2006–2013 and thermal conditions corresponding to the mid-Pliocene (m). g,n, Both net quantitative (g) and qualitative (n) changes were latitudinally averaged between 60° S and 60° N, including for comparison expected mean vulnerability changes in biodiversity corresponding to year-to-year variability (YY) in temperature (1960–2013; see Fig. 3a–d) and mean changes in temperature between 2000–2009 and 1960–1969 (Fig. 3e–h). All changes are expressed as percentage.

Large perturbations in temperatures took place between glacial and interglacial periods in the Quaternary and were responsible for major changes in the spatial distribution of many marine species^{29, 30}. It is therefore interesting to measure the extent to which future biodiversity changes may compare with natural changes that occurred between the LGM and today. Consequently, we assessed biodiversity changes between the LGM and the 1960s. During the LGM when CO₂ concentrations were around 190 ppm, global temperatures and mean sea level were 3–5 °C and 125 m lower than they were in the 1960s^{31, 32, 33}, respectively. Biodiversity reconstructions for the LGM were highly correlated with observed foraminifera LGM biodiversity³⁴ (Methods). For the three reconstructions (1–3), correlations ranged from r = 0.92 to r = 0.93 and were highly significant (p < 0.01): r₁ = 0.93 (n₁ = 101, n₁^∗ = 5), r₂ = 0.92 (n₂ = 101, n₂^∗ = 5) and r₃ = 0.93 (n₃ = 93, n₃^∗ = 5). Quantitative biodiversity changes observed between the LGM and the 1960s were substantial (>30%) in some extratropical regions and on the eastern margin of the Atlantic and Indian oceans, as well as the western part of the North Pacific Ocean (Fig. 4e). Local invasion took place in extratropical regions and local extirpation mainly occurred in the tropics (Fig. 4l). Important net quantitative biodiversity changes between the LGM and the 1960s concerned 84.6 ± 0.9% of the ocean. At that time important species turnover affected 77.9 ± 2.4% of the ocean and local species invasion was more important (30.5 ± 1.6% of the ocean) than local extirpation (22.5 ± 1.7%; Supplementary Table 1).

The mid-Pliocene is also an interesting period because it is thought that global temperatures may be close to those predicted at present by the Intergovernmental Panel on Climate Change for the end of the century³⁵. During the mid-Pliocene when CO₂ concentrations were ~400 ppm, global temperature and mean sea level were 2–3 °C and ~20 m higher than today, respectively^{36, 37,}