Fire Risk Modeling using Discriminant Analysis and Adaptive Network Based Fuzzy Inference System in the Golestan National Park

Nowadays، fires are the most common damaging factor of natural ecosystems after urban and agricultural human activities. The aim of this study is to identify the susceptible areas to fires using Discriminant Analysis and Adaptive Network Based Fuzzy Inference System in the Golestan National Park as one of the most important biosphere reserves. The required data and digital layers were prepared from associated sites، organizations and field surveys in the area of the study. After preparing the data by assuming the occurred fire، the most significant effective factors were identified using Discriminate Analysis. These factors were calculated as a weighted average and models were implemented by using K-Fold Algorithm with trapezoidal، Gaussian، triangular and bell membership functions in six stages. The best model was used for simulation. The fire hazard map was prepared with five classifications، very low، low، medium، high and very high. An accuracy assessment was performed using the relative operating characteristic. The results of the Discriminant Analysis showed the important factors، including presence of hunters and shepherds، distance from roads، average temperature، distance from the springs، rainfall during the growing season and slope. Validation trapezoidal membership functions showed best results with value of R2= 0. 534 and RMSE= 0. 283. The results of the accuracy assessment obtained with value of ROC= 0. 875. While the very High-risk area was 972 hectares، and High-risk area was 16879 hectares of 91895 hectares area. According to accuracy of the proposed map، it can be used to control the region fires in Golestan national park.