关键词: 森林火灾, 风险评估, 博弈协调, 组合赋权, 风险区划, 随机森林(RF)算法

Abstract:

To scientifically identify and assess forest fire risk, a multi-dimensional evaluation system and model was established based on"Hazard-Exposure-Vulnerability" framework. Taking Zhengzhou City as the study area, Analytic Hierarchy Process (AHP) and RF algorithm were combined to obtain subjective and objective weights, followed by consistency testing and game-theoretic coordination using the Kendall coefficient to derive combined weights. Forest fire risk assessment and zoning were conducted in Geographic Information System(ArcGIS) software, and model accuracy was validated through spatial overlay analysis with nearly a decade of historical fire point data. Results indicate that stable combined weights are obtained after game-theoretic coordination of subjective and objective weights. The areas with high and relatively high forest fire risk in Zhengzhou City account for 5.2% and 24.2% of the total area, respectively. Furthermore, 83.3% of the Sentinel-2 historical fire points are concentrated in high and relatively high-risk zones, demonstrating the good applicability of the proposed model.

Key words: forest fire, risk assessment, game-theoretic coordinated, combined weighting, risk zoning, random forest (RF) algorithm