Application of artificial neural network ‎in Prediction of wildlife potential ‎habitats in desert areas‏ ‏‎(Case study: ‎Wild goat in Kouh-e-Bafgh)‎

Wild goat (Capra aegagrus) as the most vulnerable species is one of the mountainous mammals whose population has declined due to the destruction of the habitat at the national and international levels. This study was aimed to determine the suitably potential habitat area for wild goat in Bafgh protected area using multi-layer perceptron neural network. A total of 196 points including presence (111 points) and absence (85 point) of the species were collected in fieldwork. 18 variables such as slope, geographical aspect, elevation above sea level, rocky regions, mean temperatures, vegetation types, water resources, inhabited and uninhabited villages, and roads (dirt and asphalt) were used to determine the suitability of habitats. Results showed that Juniperus excelsa -Amygdalus scoparia vegetation type (11.23%), slope steepness (10.42%), distance to south direction (10.15%), distance to Cousinia desertii-Artemisia sieberi-Zygophyllum eurypterum vegetation type (9.9 %), elevation above sea level (9.63%), and distance to water source (9.09%) are the most effective variables in habitat suitability evaluation in the Kouh-e-Bafgh protected area. The model output efficiency of 0.97 was achieved in this study. Based on the results, 36% of the protected area was evaluated as the suitable for wild goat habitat. Results also reveald that by reducing the distance to the roads the suitability of habitats is reduced. Therefore, this study suggests that human activities close to the potentially suitable area is suggested to be avoided.