On the use of neighboring habitats as predictors of species distributions

Choosing the appropriate scale for measuring environmental predictors is needed for accurately modelling species distributions. This need is becoming increasingly important with the use of high‐resolution species distribution models (SDMs), emphasizing the challenge of aligning predictors with the spatial and ecological scales at which species interact with their environments. Focal predictors, which summarize landscape information within a spatially moving window, are powerful to account for neighboring information and scale dependency but have remained overlooked in SDMs. Using an automated selection procedure to identify the best predictors and measurement scales from a high‐dimensional pool of candidates, including 13 nested circular focal sizes from 25 m to 5 km radius for each landscape feature, this study evaluated the use of focal predictors through a set of national‐scale, high‐resolution SDMs for more than 7000 species across 17 major taxonomic groups. It further examined whether focal selection depended on species' mobility or body size. Among all species, focal predictors were selected at least once in ≥ 94% of the SDMs, highlighting their important role. For mobile species, larger focal windows were selected for the land use and land cover category, whereas sessile species were associated with larger focal windows for topographic predictors. For small species, predictors with smaller focal windows were more often selected. Given the importance of focal predictors across all studied taxa, adjusting the optimal scale for each predictor and species is of utmost importance to improve model performance and account for species' scale dependency.