Fig. 6: Geographical projection of DIM occurrence probability, model’s performance (test AUC), variables contribution and their permutation importance.
a–c, Predictors included in each case were: model type a, biome and PFT (a); model type a plus continuous edaphoclimatic variables (maximum temperature, Tmax, aridity index, sand and clay content) (b); and model type b plus the number of species with HSM < 0 (N spp. HSM < 0) and HSM variance included in place of the two edaphoclimatic variables with the lowest contribution in model type b (sand and clay content), to keep the same number of predictive variables in model types b and model c (c).
