GIS-based forest fire susceptibility mapping in Iran: a comparison between evidential belief function and binary logistic regression models

Abstract

The aim of this research was to produce forest fire susceptibility maps (FFSM) based on evidential belief function (EBF) and binary logistic regression (BLR) models in the Minudasht Forests, Golestan Province, Iran. At first, 151 forest fire locations were identified from Moderate-Resolution Imaging Spectero Radiometer data, extensive field surveys, and some reports (collected in year 2010). Out of these locations, 106 (70%) were randomly selected as training data and the remaining 45 (30%) cases were used for the validation goals. In the next step, 15 effective factors such as slope degree, slope aspect, elevation, plan curvature, Topographic Position Index, Topographic Wetness Index, land use, Normalized Difference Vegetation Index, distance to villages, distance to roads, distance to rivers, wind effect, soil texture, annual temperature, and rainfall were extracted from the spatial database. Subsequently, FFSM were prepared using EBF and BLR models, and the results were plotted in ArcGIS. Finally, the receiver operating characteristic curves and area under the curves (AUCs) were constructed for verification purposes. The validation of results showed that the AUC for EBF and BLR models are 0.8193 (81.93%) and 0.7430 (74.30%), respectively. In general, the mentioned results can be applied for land use planning, management and prevention of future fire hazards.