Spatial Distribution Analysis of index Mammals in the Eshkevarat no hunting area Using the Random Forest Tree Learning Method

Abstract

Introduction: In Iran, no hunting areas are managed for maintaining the biodiversity for 3 to 5 years to join the four conservation areas managed by the Department of environment in the event of improved habitat and wildlife. Habitat suitability models can provide a rapid assessment of the status of index species as a tool. So with strong support, protection level of no hunting areas is enhanced. One of these areas with high-diversity is no hunting area of Eshkevarat. The purpose of this study was to evaluate the status of four species in the region including wild goat (Capra aegagrus), Roe deer (Capreolus capreolus), brown bear (Ursus arctos) and leopard (Panthera pardus).
Materials & Methods: For this purpose, random forest (RF) model was used and the validity of the model was evaluated by using area under curve (AUC), root mean square (RMSE) and crossed mean entropy (MXE) statistics.
Result: Model evaluation results showed that RF method has been successful in implementation. Elevation was evaluated as a factor in the suitability of bear, goat, and roe deer habitats while proximity to prey had the greatest impact on leopard distribution. The optimal habitat area for bears, leopards, roe deer and goats was 34.79%, 2.12%, 3.48% and 11.34% of the total area, respectively.
Conclusion: Spatial distribution of habitat patches of the mentioned species indicates high habitat diversity in the area, which reveals the strong role of this area in biodiversity conservation of the mentioned species and upgrading its level of protection can eliminate many of the challenges facing the area.