Predicting species extinction risk from phylogenetic, spatial and remotely sensed environmental information

Background/Question/Methods

Human impacts on the biosphere and the resulting loss of species and function are profound. Only a very small portion of the world’s biota have been assessed for their risk of extinction and tools that allow rapid large-scale predictions of the potential taxonomic and geographic distribution of threatened species are therefore vital. Here we examine 483 species of mammals for which data-deficiency has to date prevented an assessment of their threat status, and use readily available or derivable characteristics in a spatial-phylogenetic framework to predict their risk of extinction. We compiled data on body mass and the sizes as well as remotely-sensed measures of human encroachment of species geographic ranges. We combined this with information on the phylogenetic and spatial association of species to construct statistical models predicting the threat status of 3,703 IUCN-assessed species across 16 orders.

Results/Conclusions

The models effectively discriminate threatened from non-threatened species and predict the global geographic variation in threat. Applying these models to data-deficient species we predict highly elevated threat probabilities, with potentially 75% ‘at-risk’ species, compared to 29% among assessed, non data-deficient species. These results show that current assessments substantially underestimate the number of mammals that are threatened. Specifically, our predictions substantially increase the number of at-risk species among rodents, bats and shrews, and add to the importance of specific areas (e.g. Columbia, Sulawesi, Philippines) as priorities for conservation. More generally, our findings demonstrate the future potential for phylogenies and remotely sensed data, together with information on species distributions, to help identify additional species and regions of conservation concern.