Number of species presence records, AUC-values, and training gain for the selected bioclimatic variables.

AUC-values are a threshold-independent model quality criterion and range from 0 to 1 (perfect fit). Useful models yielded in AUC-values above 0.7, where excellent models achieve at least AUC-scores above 0.9. Training gains are determined by Jackknife test for the selected variables by using only this variable for the model (upper value) and if the specific variable is removed for the rest of the variable set (lower value). For the species with current (south-) western focus of distribution (P. ariasi, P. mascittii and P. perniciosus) BIO 11 (Mean temperature of the coldest quarter) represents the most important variable. This is indicated by the highest training gain of the model by using only this variable and the lowest training when this variable is removed from the set of variables. The drop of BIO 10 (Mean temperature of the warmest quarter) from the set of variables instead seems to lower training gain most for the species with (south-) eastern focus of distribution. BIO 13 (Precipitation of the wettest month) is identified as most important variable when used in isolation for P. neglectus, while BIO 11 seems to be most influencing factor in isolation regarding the occurrences of P. perfiliewi.