Regional context improves prediction of local threats from global spatial impact maps

Global biodiversity is rapidly declining under the influence of human-induced environmental change, but our understanding of what is driving change at the local scale is often incomplete. Global datasets such as the Living Planet Database (LPD) provide invaluable site-based time-series data on wildlife populations but often lack detailed information on the local threats affecting these. Here, we test whether threat impact probability maps for species can predict the presence of threats recorded in the LPD for terrestrial birds and mammals. We tested several logistic regression models, ranging from those based solely on the mapped threat probabilities to those that incorporate species-level threat data, biogeographic realm, species richness, and the Human Footprint Index. Our results show that threat impact probabilities used in combination with contextual information, particularly biogeographic realm, can provide significant improvements in predictive power compared with models without them. In contrast, the Human Footprint Index contributed little to model performance. These findings underscore the importance of incorporating ecological and regional context into threat assessments and provide a promising start to increasing our understanding of threats at a population-level.