Biodiversity indicators reveal contrasting landscape associations across scales and studies

Abstract Context. Landscape–biodiversity associations depend on response and spatial extent. Taxon counts can remain stable as community composition changes. Objectives. We compared five biodiversity indicators on identical sites across five natural-cover radii. We assessed associations with natural cover and infrastructure accessibility among studies. Methods. We analysed 1,656 terrestrial PREDICTS sites from 31 studies in 18 countries; 1,641 had complete data for all five indicators. Models compared 1–50-km buffers with fixed accessibility, Human Footprint and ecological controls. Results. Indicator choice changed association direction and magnitude. At 10 km, accessibility was negatively associated with richness and positively with dominance shift and effective-taxa loss. Natural cover was positively associated with reference-relative abundance redistribution. Across radii, natural-cover coefficients remained positive for redistribution, negative for richness and near zero for total abundance. Scale profiles changed modestly, while study slopes varied widely in sign and magnitude. Spatially conditioned and held-out checks showed weak transfer of the pooled relationship. Conclusions. Response, extent and study context jointly shaped landscape inference. Complete indicator-by-scale profiles show these dependencies within a common sampling frame.