Electronic Supplementary Material

Table S1. Pearson correlation coefficients (r) matrix among predictors.

Altitude / Distance to human settlements / Forests / Grasslands / Shrublands / Slope / Rocky areas / Roughness
Altitude / —
Distance to human settlements / –0.27 / —
Forests / 0.13 / –0.06 / —
Grasslands / –0.14 / –0.11 / 0.57 / —
Shrublands / 0.07 / –0.11 / 0.52 / 0.43 / —
Slope / –0.27 / –0.06 / –0.29 / –0.11 / –0.21 / —
Rocky areas / –0.33 / –0.05 / 0.46 / 0.47 / 0.35 / –0.09 / —
Roughness / –0.08 / –0.03 / –0.08 / –0.03 / –0.08 / –0.27 / –0.12 / —

Figure S1. Multi-panel Cleveland dotplots for eight predictor variables. Axis labels were suppressed to improve visual presentation. Roughness, altitude, distance to human settlementsand slope have few unusually small or large values.


Figure S2. Moran's I correlogram of spatial distributions carried out with 10,000 permutations for bearded vultures in the study area. The black dot represents significant (P< 0.05) Moran's index.


Figure S3. Boxplot summarizing model validations of ten species distribution models (maximumentropy,MAX; factorial decomposition of Mahalanobis distances, MAD; generalised linear models, GLM; boostedregression trees, BRT; generalized additive models, GAM; classification tree analyses, CTA; artificial neural networks, ANN; flexible discriminant analyses, FDA; multivariate adaptive regression splines, MARS; random forests, RF) and their ensemble prediction (EP) developed for three age classes (adults, sub-adults and young). Area Under the Curve (AUC) varies from 0 and 1, Boyce’ Index from −1 to 1.For all panels, boxes closer to oneindicate high predictive accuracy.White boxes indicate values often k-fold cross-validationscarried out with original data, light grey boxes with new field data and dark grey boxes with the external dataset.

1