Document Type : Research Article
Department of Nature Engineering, Shirvan Faculty of Agriculture, University of Bojnord
Research Department of Natural Resources, Golestan Agricultural and Natural Resources Research and Education Center, AREEO, Gorgan, Iran
Nowadays, preparation of fire risk spatial distribution map is one of the essential management tools at different levels to monitor the sustainability of natural resources and control this environmental hazard. The integration of field operations, remote sensing data, geographic information system techniques and various statistics can create a reliable spatial prediction of fire hazard potential for different regions. Therefore, in this research, among the nine effective factors in fire risk modeling, including altitude, slope, aspect, distance from the road, and NDVI, LST, TWI, and TPI, with two methods of fuzzy analytic hierarchy process (FAHP) and logistic regression (LR) used to identify risk areas and determined the most important factors affecting the occurrence and spread of fire. Google Earth Engine and MODIS images were also used to identify historical fire areas. The initial results showed that in both models, NDVI, LST and distance from the road have the highest coefficients. Although, in the verification phase, the performance characteristic curve of both models was relatively the same (0.847 in FAHP method and 0.837 in the LR method), it was determined by examining the historical fire pixels that in the FAHP, about 87% of the pixels of the classes with high and very high risk, and in the LR method, only 22% of the pixels of the said classes overlapped with areas with a history of fire. Therefore, it seems that the FAHP method was able to identify areas with high risk potential better than the LR method. Although the preparation of risk prediction maps by different models will not prevent the occurrence of all fires, but, its occurrence can be reduced and its control facilitated by providing management solutions.