昆明市林火驱动因子及火险区划研究

  目的  对昆明市2000—2015年的火点数据进行分析，确定主要林火驱动因子，建立昆明市林火预报模型并进行火险区划，为昆明市林火预防提供参考。  方法  基于林火数据，选取气象、地形、植被、人为等17个林火驱动因子构建Logistic回归林火概率模型，并划分5个中间模型选取全样本的显著变量因子，用受试者工作特征曲线(ROC)进行模型检验与评价，基于全样本的模型结果分析昆明市主要林火驱动因子，并计算得到林火发生概率的最佳阈值，根据Logistic模型结果划分五级火险区。  结果  海拔、距居民点距离、距铁路距离、归一化植被指数(NDVI)值、月均地表温度、月均气压、月均相对湿度、月均风速、人均国内生产总值(GDP)等9个因子与昆明市林火发生概率存在显著关系；Logistic模型的预测准确率高达81.7%；ROC曲线下面积(AUC)的值为0.905；划分的最佳阈值为0.342；火险区划的五级火险区面积比率分别为48.82%、35.17%、11.26%、2.55%、2.20%。  结论  昆明市林火主要驱动因子是气象因子；昆明市高火险区集中分布在五华区、盘龙区、官渡区、呈贡区、西山区、安宁市等西南部地区。图4表3参22

Driving factors of forest fire and fire risk zoning in Kunming City

  Objective  With an analysis of the data of fire spots in Kunming City from 2000 to 2015, this study is aimed to determine the main driving factors of forest fires, establish the forest fire prediction model in Kunming City, map the forest fire-risk zones so as to provide reference for the prevention and management of forest fires in Kunming City.  Method  First, with 17 forest fire driving factors selected, a Logistic regression forest fire probability model was constructed on the basis of the satellite forest fire data, including meteorology, terrain, vegetation, artificial, and so on. Then five intermediate models were employed to select the significant variable factors for whole samples before the model was tested using ROC curve. Based on the modeling results of the whole samples, an analysis was conducted of the main forest fire driving factors in Kunming City with the best threshold for probability of forest fire calculated. At last, five fire-risk zones were determined based on the results obtained via the Logistic model.  Result  Among the nine driving factors were altitude, distance from residential areas, distance from railways, NDVI value, monthly average surface temperature, monthly average air pressure, monthly average relative humidity, monthly average wind speed and per capita GDP. The prediction accuracy of Logistic model was as high as 81.7%. The area AUC under ROC curve is 0.905. The best threshold of division was 0.342, the area percent of the five-level fire danger zones were 48.82%, 35.17%, 11.26%, 2.55% and 2.2% respectively.  Conclusion  The main driving factor for forest fire in Kunming City is the meteorological factor with the high fire risk areas in Kunming City concentrated in the southwest region including Wuhua District, Panlong District, Guandu District, Chenggong District, Xishan District and Anning City. Ch, 4 fig. 3 tab. 22 ref.]