中文摘要：

火灾风险模拟是探究火灾风险机制和揭示影响火灾主要因素的重要途径。使用可解释火灾及其影响因素间局部变异关系的地理加权泊松模型为建模工具,选取坡度、人类活动可达性、太阳辐射、地表温度、植被状态等火环境因素,对河南省2002—2012年的火灾风险进行空间模拟,并分别从可靠性和区分能力两方面对模型进行检验。结果表明：1与广义泊松模型相比,地理加权泊松模型的模拟精度和模型性能均显著提高;2人类活动可达性对火灾的影响在河南省内未呈现显著空间变化,地形、太阳辐射、地表温度和植被状态对火灾影响在河南省内表现出显著空间变化,邻近居民区和远离主要道路的区域更易发生火灾;3火灾影响因素的估计系数图可用于防火管理。

英文摘要：

Fire danger modeling is one of the most important ways to explore the mechanism of fire occurrence and to discover the main fire-influencing factors. Geographically Weighted Poisson Model（ GWPM） can account for local variation of the relationships between fire hazards and fire-influencing factors was applied to model monthly fire frequency cell by cell for fire season of Henan province（ i. e. September and October）. Several fire-influencing factors involving slope,accessibility,solar radiation,land surface temperature and vegetation conditions were employed and potential multicollinearity among the independent variables was excluded in modeling. The reliability and discrimination capacity of the developed GWPM fire models were respectively evaluated by using the inner testing subset and the independent validation subset. The results indicate that： 1 the fitting accuracy and the model＇s performance are greatly improved by GWPM compared to Generalized Poisson Model（ GPM）. 2 Influences of factors present significant spatial variability,accounting for slope,solar radiation,land surface temperature and vegetation conditions on fire hazards but the influences of factors accounting for accessibility of human activity on fire hazards exhibit insignificant spatial variability. In addition,those areas which were located far away from main roads and near villages are fire-prone zones. The estimated coefficients maps have important implications for fire prevention management.