Persian leopard and wild sheep distribution modeling using the Maxent model in the Tang-e-Sayad protected area, Iran

Abstract

The maximum entropy (Maxent) model was used to predict the distribution of Persian leopards and wild sheep in the Tang-e-Sayad protected area in Iran. For this purpose, eight variables, as well as 30 occurrence points of leopard and 98 points of wild sheep, were used. Two techniques, density-based occurrence points thinning and performance-based predictor variables selection were used to improve the results of the model. The model results were analyzed based on four threshold limit-based statistics (sensitivity, specificity, kappa and true skill statistics) and area under the curve (AUC), followed by determining the relative importance of variables based on the jackknife procedure. The results of threshold limit-based statistics revealed that the success of the model for distribution prediction of leopard and wild sheep were good and relatively good, respectively. According to the jackknife procedure, for wild sheep and for leopard, slope and distance to road, respectively, were the most important predictor variables. The results also indicated that the efficiency of the model did not improve by reducing the density of occurrence points for the wild sheep (AUC=0.784–0.773). However, the selection of predictor variables slightly improved the performance of the model (AUC=0.794–0.819). The results of the study also showed overlapping habitat for two species due to both human and ecological reasons for which we proposed some conservation actions such as excluding domestic grazing, controlling illegal poaching and restoration of old migratory corridors.