Contrasting biogeographical patterns of threatened vertebrates on islands emerge from disparities between expert-derived maps and Global Biodiversity Information Facility data

Abstract

Aim: The most popular sources of information on species distributions are the expert-derived maps and georeferenced occurrences, mainly those compiled by the Global Biodiversity Information Facility (GBIF). These sources have been constantly used with biogeographical and conservation goals. However, their degree of accuracy in representing geographical biodiversity patterns remains poorly understood. Here, we compared both sources of information on species distributions to estimate global patterns of richness and species composition of threatened vertebrates on marine islands. Location: Global. Taxon: Terrestrial vertebrates. Methods: We gathered distributional data of all threatened terrestrial vertebrate species inhabiting 22,471 marine islands worldwide from GBIF and expert derived maps. Then, to assess strengths and biases from each source, using geographical information systems, we calculated and compared: (a) species richness per island, (b) general patterns of richness and (c) the number of shared species from both sources per island. Results: There were dramatic differences between the information derived from both data sources. The species richness estimated with expert-derived maps resulted in 10 more species per island, on average, than the estimations obtained from GBIF data. The mean proportion of shared species per island (between both data sources) was very low (3.1% of the species), and the general patterns of richness were markedly different. The most significant differences occurred in tropical areas and Europe. Conclusion: Differences between the two sources emerged from intrinsic biases: expert-derived maps tend to overestimate species' counts, whereas GBIF occurrences tend to do the opposite, except for some well-sampled regions where both data types converge. Our findings suggest that previous global assessments performed with these information sources on species distributions, especially those focusing on protected areas, should be carefully considered.