Where did they not go? Considerations for generating pseudo-absences for telemetry-based habitat models.

摘要: Habitat suitability models give insight into the ecological drivers of species distributions and are increasingly common in management conservation planning. Telemetry data can be used habitat to describe where animals were present, however this requires use presence-only modeling approaches or generation ‘pseudo-absences’ simulate locations did not go. To highlight considerations for generating pseudo-absences telemetry-based models, we explored how different methods pseudo-absence affect model performance across species’ movement strategies, types, environments. We built marine terrestrial case studies, Northeast Pacific blue whales (Balaenoptera musculus) African elephants (Loxodonta africana). tested four commonly models: (1) background sampling; (2) sampling within a buffer zone around presence locations; (3) correlated random walks beginning at tag release location; (4) reverse last location. using generalised linear mixed additive boosted regression trees. found that separation environmental niche space between presences was single most important driver explanatory power predictive skill. This result consistent habitats, two with vastly syndromes, three types. The best-performing method depended on which created greatest separation: elephants. However, despite fact greater performed better according traditional skill metrics, they always produce biologically realistic spatial predictions relative known distributions. may positively biased cases sampled from environments dissimilar presences. emphasizes need carefully consider extent domain heterogeneity samples when developing highlights importance scrutinizing ensure fit objectives.