The Heat Is Online

Extinction risk from climate change

Nature, v. 427, pp. 145-148, Jan. 8 2004

Extinction risk from climate change


1 Centre for Biodiversity and Conservation, School of Biology, University of Leeds, Leeds LS2 9JT, UK
2 Royal Society for the Protection of Birds, The Lodge, Sandy, Bedfordshire SG19 2DL, UK, and Conservation Biology Group, Department of Zoology, University of Cambridge, Downing Street, Cambridge CB2 3EJ, UK
3 National Institute of Public Health and Environment, P.O. Box 1, 3720 BA Bilthoven, The Netherlands
4 Department of Biological Sciences, Macquarie University, North Ryde, 2109, NSW, Australia
5 University of Durham, School of Biological and Biomedical Sciences, South Road, Durham DH1 3LE, UK
6 Animal, Plant and Environmental Sciences, University of the Witwatersrand, Private Bag 3, WITS 2050, South Africa
7 Centro de Referência em Informação Ambiental, Av. Romeu Tórtima 228, Barão Geraldo, CEP:13083-885, Campinas, SP, Brazil
8 School of Geography, University of Leeds, Leeds LS2 9JT, UK
9 Center for Applied Biodiversity Science, Conservation International, 1919 M Street NW, Washington, DC 20036, USA
10 Department of Zoology, University of Stellenbosch, Private Bag X1, Stellenbosch 7602, South Africa
11 Climate Change Research Group, Kirstenbosch Research Centre, National Botanical Institute, Private Bag x7, Claremont 7735, Cape Town, South Africa
12 Unidad Occidente, Instituto de Biología, Universidad Nacional Autónoma de México, México, D.F. 04510 México
13 Natural History Museum and Biodiversity Research Center, University of Kansas, Lawrence, Kansas 66045 USA
14 Cooperative Research Centre for Tropical Rainforest Ecology, School of Tropical Biology, James Cook University, Townsville, QLD 4811, Australia
* Present address: UNEP World Conservation Monitoring Centre, 219 Huntingdon Road, Cambridge CB3 0DL, UK

Correspondence and requests for materials should be addressed to C.D.T. (

Climate change over the past 30 years has produced numerous shifts in the distributions and abundances of species1, 2 and has been implicated in one species-level extinction3. Using projections of species' distributions for future climate scenarios, we assess extinction risks for sample regions that cover some 20% of the Earth's terrestrial surface. Exploring three approaches in which the estimated probability of extinction shows a power-law relationship with geographical range size, we predict, on the basis of mid-range climate-warming scenarios for 2050, that 1537% of species in our sample of regions and taxa will be 'committed to extinction'. When the average of the three methods and two dispersal scenarios is taken, minimal climate-warming scenarios produce lower projections of species committed to extinction (18%) than mid-range (24%) and maximum-change (35%) scenarios. These estimates show the importance of rapid implementation of technologies to decrease greenhouse gas emissions and strategies for carbon sequestration.The responsiveness of species to recent1-3 and past4, 5 climate change raises the possibility that anthropogenic climate change could act as a major cause of extinctions in the near future, with the Earth set to become warmer than at any period in the past 140 Myr (ref. 6). Here we use projections of the future distributions of 1,103 animal and plant species to provide 'first-pass' estimates of extinction probabilities associated with climate change scenarios for 2050.

For each species we use the modelled association between current climates (such as temperature, precipitation and seasonality) and present-day distributions to estimate current distributional areas7-12. This 'climate envelope' represents the conditions under which populations of a species currently persist in the face of competitors and natural enemies. Future distributions are estimated by assuming that current envelopes are retained and can be projected for future climate scenarios7-12. We assume that a species either has no limits to dispersal such that its future distribution becomes the entire area projected by the climate envelope model or that it is incapable of dispersal, in which case the new distribution is the overlap between current and future potential distributions (for example, species with little dispersal or that inhabit fragmented landscapes)11. Reality for most species is likely to fall between these extremes.

We explore three methods to estimate extinction, based on the speciesarea relationship, which is a well-established empirical power-law relationship describing how the number of species relates to area (S = cAz, where S is the number of species, A is area, and c and z are constants)13. This relationship predicts adequately the numbers of species that become extinct or threatened when the area available to them is reduced by habitat destruction14, 15. Extinctions arising from area reductions should apply regardless of whether the cause of distribution loss is habitat destruction or climatic unsuitability.

Because climate change can affect the distributional area of each species independently, classical community-level approaches need to be modified (see Methods). In method 1 we use changes in the summed distribution areas of all species. This is consistent with the traditional speciesarea approach: on average, the destruction of half of a habitat results in the loss of half of the distribution area summed across all species restricted to that habitat. However, this analysis tends to be weighted towards species with large distributional areas. To address this, in method 2 we use the average proportional loss of the distribution area of each species to estimate the fraction of species predicted to become extinct. This approach is faithful to the speciesarea relationship because halving the habitat area leads on average to the proportional loss of half the distribution of each species. Method 3 considers the extinction risk of each species in turn. In classical applications of the speciesarea approach, the fraction of species predicted to become extinct is equivalent to the mean probability of extinction per species. Thus, in method 3 we estimate the extinction risk of each species separately by substituting its area loss in the speciesarea relationship, before averaging across species (see Methods). Our conclusions are not dependent on which of these methods is used. We use z = 0.25 in the speciesarea relationship throughout, given its previous success in predicting proportions of threatened species14, 15, but our qualitative conclusions are not dependent on choice of z (Supplementary Information). As there are gaps in the data (not all dispersal/climate scenarios were available for each region), a logitlinear model is fitted to the extinction risk data to produce estimates for missing values in the extinction risk table (Table 1). Balanced estimates of extinction risk, averaged across all data sets, can then be calculated for each scenario.

For projections of maximum expected climate change, we estimate species-level extinction across species included in the study to be 2132% (range of the three methods) with universal dispersal, and 3852% for no dispersal (Table 1). For projections of mid-range climate change, estimates are 1520% with dispersal and 2637% without dispersal (Table 1). Estimates for minimum expected climate change are 913% extinction with dispersal and 2231% without dispersal. Projected extinction varies between parts of the world and between taxonomic groups (Table 1), so our estimates are affected by the data available. The speciesarea methods differ from one another by up to 1.41-fold (method 1 versus method 3) in estimated extinction, whereas the two dispersal scenarios produce a 1.98-fold difference, and the three climate scenarios generate 2.05-fold variation.

Given its role in conservation planning, we also use a different approach to estimate extinction, modified from the IUCN Red Data Book listing procedure16: this is semi-numerical and includes components of expert judgement. Species are assigned to different threat categories based on distribution sizes and declines, with each category carrying a specified probability of extinction16 (see Methods and Supplementary Information). For scenarios of maximum expected climate change, 33% (with dispersal) and 58% (without dispersal) of species are expected to become extinct (Table 1). For mid-range climate change scenarios, 19% or 45% (with or without dispersal) of species are expected to become extinct, and for minimum expected climate change 11% or 34% (with or without dispersal) of species are projected to become extinct.

We can compare these values with the proportions of species projected to become extinct as the result of global habitat losses, currently the most widely recognized extinction threat. We apply the speciesarea relationship to changes in global land use that have taken place since human land conversion began17. Estimates of extinction range from 1% to 29%, depending on the biome (considering only species restricted to single biomes; Table 2). Given that a high proportion of the world's species reside in tropical forests (extinction estimate 4%; Table 2), global extinction related to habitat loss would be expected to be in the lower half of the range, and thus lower than the rate projected for scenarios of mid-range climate change (24%; average of area methods). Projected conversion of humid tropical forest at an annual rate of 0.43% (ref. 18) from 1990 to 2050 predicts a further 6.3% of species committed to extinction.

Regional differences are expected, so we also compare the relative risks during 20002050 associated with land use and climate change (using area approaches) for the three regiontaxon combinations that correspond most closely to single habitat or biome types. First, for montane Queensland forests12, extinction risk is dominated by climate change (713% and 4358% predicted extinction for minimum and maximum climate scenarios, respectively; 0% predicted on the basis of further habitat destruction, given its legal protection). Second, for cerrado vegetation in Brazil, high rates of habitat destruction19 make it possible that only current reserves will survive. Making this pessimistic assumption, an estimated additional 34% of all original species will be committed to extinction due to habitat destruction during 20002050, a value lower than the 4856% of woody plant species projected to be committed to extinction for mid-range climate warming (3845% for minimum warming). Last, for South African Proteaceae, 27% of all original species are projected to become extinct as a result of land use changes during 20002050 (for a pessimistic linear extrapolation of land use scenarios after 2020)20, falling between the 3040% (without dispersal) and 2127% (with ubiquitous dispersal, which is unlikely for these plants) projected extinction for mid-range climate scenarios.

Many unknowns remain in projecting extinctions, and the values provided here should not be taken as precise predictions. Analyses need to be repeated for larger samples of regions and taxa, and the selection of climate change scenarios need to be standardized. Some of the most important uncertainties follow (see also Supplementary Information). We estimate proportions of species committed to future extinction as a consequence of climate change over the next 50 years, not the number of species that will become extinct during this period. Information is not currently available on time lags between climate change and species-level extinctions, but decades might elapse between area reduction (from habitat loss) and extinction14. Land use should also be incorporated into analyses: extinction risks might be higher than we project if future locations of suitable climate do not coincide with other essential resources (such as soil type or food resources). There is also uncertainty over which species will inhabit parts of the world projected to have climates for which no current analogue exists6. Equally importantly, all parts of the world will have historically unprecedented CO2 levels6, which will affect plant species and ecosystems21, 22 and herbivores23, resulting in novel species assemblages and interactions.

Despite these uncertainties, we believe that the consistent overall conclusions across analyses establish that anthropogenic climate warming at least ranks alongside other recognized threats to global biodiversity. Contrary to previous projections24, it is likely to be the greatest threat in many if not most regions. Furthermore, many of the most severe impacts of climate-change are likely to stem from interactions between threats, factors not taken into account in our calculations, rather than from climate acting in isolation. The ability of species to reach new climatically suitable areas will be hampered by habitat loss and fragmentation, and their ability to persist in appropriate climates is likely to be affected by new invasive species.

Minimum expected (that is, inevitable) climate-change scenarios for 2050 produce fewer projected 'committed extinctions' (18%; average of the three area methods and the two dispersal scenarios) than mid-range projections (24%), and about half of those predicted under maximum expected climate change (35%). These scenarios would diverge even more by 2100. In other words, minimizing greenhouse gas emissions and sequestering carbon25 to realize minimum, rather than mid-range or maximum, expected climate warming could save a substantial percentage of terrestrial species from extinction. Returning to near pre-industrial global temperatures as quickly as possible could prevent much of the projected, but slower-acting, climate-related extinction from being realized.Methods

Climate-envelope modelling The statistical match between climate variables and the boundaries of a species' distribution (climate envelope) represents conditions in which a species (normally) shows a positive demographic balance (rarely the absolute physical limits of a species, but the set of conditions under which it survives in at least some multi-species communities). The statistical approach is generic, but specific methods vary between studies (
Supplementary Information). The approach has been validated by successfully predicting distributions of invading species when they arrive in new continents and by predicting distributional changes in response to glacial climate changes; its scope has been discussed widely (see, for example, refs 12, 2629). Dispersal is assumed to be universal or zero (main text), except for the Mexican study in which 'universal dispersal' is movement through contiguous habitats11.Climate scenarios Climate projections for 2050 were divided into three categories: minimum expected change resulting in a mean increase in global temperature of 0.81.7 °C and in CO2 of 500 p.p.m. by volume (p.p.m.v.); mid-range scenarios with temperature increases of 1.82.0 °C and CO2 increases of 500550 p.p.m.v.; and maximum expected scenarios with temperature increases of >2.0 °C and CO2 increases >550 p.p.m.v. (ref. 30). Projections for the year 2100 were allocated to 2050 scenarios according to their end temperatures and CO2 levels (Supplementary Information).Species Within each region we use only data for endemic species (near-endemic in two cases). Near-endemics are defined as >90% of the distribution area known to occur (European birds) or thought to occur (cerrado plants, given incomplete data) within the region modelled. For European birds, near-endemics are included only if their extra-European distribution is similar to climate space within Europe. The focus on endemics permits us to model all range boundaries of each species (Supplementary Information).Speciesarea approaches Method 1 analyses overall changes in distribution areas, summed across species. The proportion of species in a region going extinct (E1) is estimated as

where Aoriginal is the area initially occupied by a species, and Anew is the future area projected for the same species, with summation carried out across species.

Method 2 is based on the average proportional change in distribution area, averaged across species. Regional extinction risk (E2) is

where n is the number of species and Anew/Aoriginal is the proportional distribution change for each species in turn.

Method 3 estimates the extinction risk of each species in turn, averaging across species to derive regional estimates of extinction (E3):


Species for which Anew > Aoriginal were analysed as though Anew = Aoriginal; that is, zero extinction would be returned by each equation if every species was projected to expand

(Supplementary Information).

It is important to recognize that further work is required to establish empirically how the absolute and proportional area losses of individual species (in other words, the type of data from climate envelope projections) are related to extinction risk. As yet, no agreed standard method exists for such calculations: assumptions and uncertainties inherent in the three methods will be considered in detail elsewhere.

Extinction probability estimates were not available for all scenarios in every region/taxon, so means of scenarios were calculated after using a least-squares analysis of variance model to impute missing values. Region/taxon mean probabilities of extinction for each scenario were logit-transformed and a three-way analysis of variance was fitted (region/taxon climate scenario dispersal scenario; weighted by (Nspecies) per region/taxon study). The fitted model was used to impute expected values of the probability of extinction for those region/taxon and scenario combinations for which direct estimates were not available. Scenario means were then calculated from the combined direct estimates and imputed values, using (Nspecies) for each region/taxon as weights.Red Data Book criteria Each species is assigned to a threat category16, or classified 'Not Threatened' (0% risk), depending on the projected decline in area over 50 or 100 years (Supplementary Information) and the final distribution area. Existing areas were considered, so we present only the extra extinction attributable to climate change. Logit-transformed three-way analysis of variance was used to estimate extinction risks for empty cells, as with the speciesarea approaches.

Extinct: species with a projected future area of zero (100% of species assumed to be committed to eventual extinction).

Critically endangered: projected future distribution area <10 km2, or decline by >80% in 50 years (species assigned a 75% chance of extinction16).

Endangered: projected area 10500 km2, or 5080% decline in 50 years (species assigned a 35% chance of extinction16).

Vulnerable: projected area 5002,000 km2, or >50% decline in 100 years on the basis of linear extrapolation of 50-year projection (species assigned a 15% chance of extinction16).

Supplementary information accompanies this paper.

Received 10 September 2003;accepted 13 October 2003




Parmesan, C. & Yohe, G. A globally coherent fingerprint of climate change impacts across natural systems. Nature 421, 3742 (2003) | Article | PubMed | ISI | ChemPort |


Root, T. L. et al. Fingerprints of global warming on wild animals and plants. Nature 421, 5760 (2003) | Article | PubMed | ISI | ChemPort |


Pounds, J. A., Fogden, M. L. P. & Campbell, J. H. Biological response to climate change on a tropical mountain. Nature 398, 611615 (1999) | Article | ISI | ChemPort |


Overpeck, J., Whitlock, C. & Huntley, B. in Paleoclimate, Global Change and the Future (eds Alverson, K., Bradley, R. & Pedersen, T.) 81103 (Springer, Berlin, 2002)


Benton, M. J. & Twitchett, R. J. How to kill (almost) all life: the end-Permian extinction event. Trends Ecol. Evol. 18, 358365 (2003) | Article |


Houghton, J. T. et al. Climate change 2001: the Scientific Basis. Contributions of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change (Cambridge Univ. Press, 2001)


Bakkenes, M., Alkemade, J. R. M., Ihle, F., Leemans, R. & Latour, J. B. Assessing effects of forecasted climate change on the diversity and distribution of European higher plants for 2050. Global Change Biol. 8, 390407 (2002) | Article |


Beaumont, L. J. & Hughes, L. Potential changes in the distributions of latitudinally restricted Australian butterfly species in response to climate change. Global Change Biol. 8, 954971 (2002) | Article |


Erasmus, B. F. N., van Jaarsveld, A. S., Chown, S. L., Kshatriya, M. & Wessels, K. Vulnerability of South African animal taxa to climate change. Global Change Biol. 8, 679693 (2002) | Article |


Midgley, G. F., Hannah, L., Rutherford, M. C. & Powrie, L. W. Assessing the vulnerability of species richness to anthropogenic climate change in a biodiversity hotspot. Global Ecol. Biogeogr. 11, 445451 (2002) | Article |


Peterson, A. T. et al. Future projections for Mexican faunas under global climate change scenarios. Nature 416, 626629 (2002) | Article | PubMed | ISI | ChemPort |


Williams, S. E., Bolitho, E. E. & Fox, S. Climate change in Australian tropical rainforests: an impending environmental catastrophe. Proc. R. Soc. Lond. B 270, 18871892 (2003) | Article | PubMed |


Rosenzweig, M. L. Species Diversity in Space and Time (Cambridge Univ. Press, 1995)


Brooks, T. M., Pimm, S. L. & Oyugi, J. O. Time lag between deforestation and bird extinction in tropical forest fragments. Conserv. Biol. 13, 11401150 (1999) | Article | ISI |


Brooks, T. M., Pimm, S. L. & Collar, N. J. Deforestation predicts the number of threatened birds in insular Southeast Asia. Conserv. Biol. 11, 382394 (1997) | Article | ISI |


IUCN Red List Categories and Criteria, version 3.1. (IUCN Species Survival Commission, Gland, Switzerland, 2001).


Gaston, K. J., Blackburn, T. M. & Goldewijk, K. K. Habitat conversion and global avian biodiversity loss. Proc. R. Soc. Lond. B 270, 12931300 (2003) | Article | PubMed |


Achard, F. et al. Determination of deforestation rates of the world's humid tropical forests. Science 297, 9991002 (2002) | Article | PubMed | ISI | ChemPort |


Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca, G. A. B. & Kent, J. Biodiversity hotspots for conservation priorities. Nature 403, 853858 (2000) | Article | PubMed | ISI | ChemPort |


Roget, M., Richardson, D. M., Cowling, R. M., Lloyd, J. W. & Lombard, A. T. Current patterns of habitat transformation and future threats to biodiversity in terrestrial ecosystems of the Cape Floristic Region, South Africa. Biol. Conserv. 112, 6385 (2003) | Article |


Woodward, F. I. Potential impacts of global elevated CO2 concentrations on plants. Curr. Opin. Plant Biol. 5, 207211 (2002) | Article | PubMed | ISI | ChemPort |


Bond, W. J., Midgley, G. F. & Woodward, F. I. The importance of low atmospheric CO2 and fire in promoting the spread of grasslands and savannas. Global Change Biol. 9, 973982 (2003)


Whittaker, J. B. Impacts and responses at population level of herbivorous insects to elevated CO2. Eur. J. Entomol. 96, 149156 (1999)


Sala, O. E. et al. Biodiversityglobal biodiversity scenarios for the year 2100. Science 287, 17701774 (2000) | Article | PubMed | ISI | ChemPort |


Lackner, K. S. A guide to CO2 sequestration. Science 300, 16771678 (2003) | Article | PubMed | ChemPort |


Beerling, D. J. The impact of temperature on the northern distribution limits of the introduced species Fallopia japonica and Impatiens glandulifera in north-west Europe. J. Biogeog. 20, 4553 (1993) | ISI |


Baker, R. H. A. et al. The role of climatic mapping in predicting the potential geographical distribution of non-indigenous pests under current and future climates. Agric. Ecosyst. Environ. 82, 5771 (2000) | Article |


Peterson, A. T. & Vieglais, D. A. Predicting species invasions using ecological niche modeling. BioScience 51, 363371 (2001) | ISI |


Pearson, R. G. & Dawson, T. P. Predicting the impacts of climate change on the distribution of species: are bioclimate envelope models useful? Global Ecol. Biogeog. 12, 361371 (2003) | Article |


Intergovernmental Panel on Climate Change. Climate Change 2001: The Scientific Basis. (2001).

Acknowledgements. We thank the following for many contributions: E. Bolitho, V. Perez Canhos, D. A. L. Canhos, S. Carver, S. L. Chown, S. Fox, M. Kshatriya, D. Millar, A. G. Navarro-Sigüenza, R. S. Pereira, B. Reyers, E. Martínez-Meyer, V. Sánchez-Cordero, J. Soberón, D. R. B. Stockwell, W. Thuiller, D. A. Vieglais and K. J. Wessels, researchers involved in the Projeto de Cooperação Técnica Conservação e Manejo da Biodiversidade do Bioma Cerrado, EMBRAPA Cerrados, UnB, Ibama/DFID e RBGE/Reino Unido, and the European Bird Census Council. We thank G. Mace, J. Malcolm and C. Parmesan for valuable discussions, many funding agencies for support, and B. Orlando and others at IUCN for bringing together many of the coauthors at workshops. Comments from J. A. Pounds and S. Pimm greatly improved the manuscript.Authors' contributions The fourth and subsequent authors are alphabetically arranged and contributed equally. Competing interests statement. The authors declare that they have no competing financial interest