The SDM-randomForest dataset comprises species distribution models for 20,567 terrestrial vertebrate species (2,739 amphibians, 8,115 birds, 5,711 reptiles, and 4,002 mammals) developed using the Random Forest algorithm. Models were calibrated using species range maps from IUCN and BirdLife together with bioclimatic variables derived from a 30-year climatology (1981–2010) based on the GSWP3-W5E5 observational climate dataset (ISIMIP3a). The calibrated models were subsequently applied to historical and future climate-related forcing data from ISIMIP3b, including CMIP6-based simulations under SSP1-2.6, SSP3-7.0, and SSP5-8.5 scenarios. Projected distributions were constrained using a 1,000-km dispersal buffer. Two types of outputs are provided: (i) naturalized runs (nat), representing buffered climatic suitability without direct human forcing, and (ii) land-use-filtered runs, in which habitat suitability is estimated from ISIMIP3a/3b land-use and irrigation datasets and combined with climatic suitability to derive buffered probabilities of occurrence for each species.
