Evaluating Fire Hazard Potentials using Fuzzy Analytic Hierarchy Process and Logistic Regression Approaches in Golestan National Park

Nowadays, the creation of a spatial distribution map for assessing fire risk is widely recognized as a crucial management tool at various levels. This tool helps monitoring natural resource sustainability and the effective control of this environmental hazard. Through the integration of field operations, remote sensing data, geographic information system techniques, and diverse statistical methods, it becomes feasible to develop a dependable spatial prediction of fire hazard potential for different regions. In this research, nine factors were identified to be effective in fire risk modeling, including altitude, slope, aspect, distance from the road, NDVI, LST, TWI, and TPI. Fuzzy Analytic Hierarchy Process (FAHP) and Logistic Regression (LR) were employed to identify risk areas and determine the most significant factors influencing the occurrence and spread of fire. Historical fire areas were identified using Google Earth Engine and MODIS images. The initial results showed that both models assign the highest coefficients to NDVI, LST, and distance from the road. However, during the verification phase, the performance characteristic curve of both models was relatively similar (0.847 for FAHP method and 0.837 for LR method). Upon examining historical fire pixels, it was found that FAHP method correctly identified approximately 87% of the pixels belonging to classes with high and very high risk, whereas LR method only overlapped with 22% of these pixels. This suggests that FAHP method is better at identifying areas with a high risk potential compared to LR method. While it is important to acknowledge that the creation of risk prediction maps using various models cannot completely eliminate all fires, it can greatly diminish their frequency and help their management by offering effective solutions.