The prediction of Persian Squirrel Distribution Using a Combined Modeling Approach in the Forest Landscapes of Luristan Province

Habitat destruction is the most important factor determining species extinction; hence, the management of wildlife populations necessitates the management of habitats. Habitat suitability modeling is one of the best tools used for habitat management. There are several methods for habitat suitability modeling, with each of having  some different advantages and disadvantages. In this study, we used 15 modeling methods along with 9 environmental factors including Bio1, Bio2, distance to roads, distance to residential areas, distance to agricultural lands, distance to streams, the percentage of slope, geographic aspect, and NDVI to model the Persian squirrel’s habitat suitability in the forests of Luristan Province. The AUC of each model was computed and the models with an AUC higher than 0.9 were selected. Finally, the output maps resulted from the selected models were multiplied by their AUC and the average of them was considered as a combined model. In this study, Maximum Entropy, Boosted Regression Tree, Generalized Linear Model, and Random Forest were the only models with an AUC higher than 0.9. Based on the combined model, 66% of the forest areas in Luristan Province could be suitable for the Persian squirrel, of which 32.1%, 18.4%, and 15.5% have low, moderate, and high suitability, respectively. Among the 9 environmental factors used in this study, distance to roads, distance to agricultural lands and NDVI showed the highest contribution in the habitat suitability of  the Persian squirrel. This study indicated that the combination of high-accuracy models could  yield more reliable results, as compared to  their separate use.