论文部分内容阅读
黑龙江大兴安岭是森林雷击火的高发地区,急需研发精确的火险预测模型对该区森林火灾进行预测.本文基于大兴安岭地区森林雷击火灾数据及环境变量数据,采用MAXENT模型进行森林雷击火的火险预测.首先对各环境变量进行共线性诊断,再利用累积正则化增益法和Jackknife方法评价了环境变量的重要性,最后采用最大Kappa值和AUC值检测了MAXENT模型的预测精度.结果表明:闪电能量和中和电荷量的方差膨胀因子(VIF)值分别为5.012和6.230,与其他变量之间存在共线性,不能用于模型训练.日降雨量、云地闪电数量及云地闪回击电流强度是影响森林雷击火发生的3个最重要因素,日平均风速和坡向的影响较小.随着建模数据比例的增加,最大Kappa值和AUC值均有增大趋势.最大Kappa值都大于0.75,平均值为0.772;AUC值都大于0.5,平均值为0.859.MAXENT模型的预测精度达到中等精度,可应用于大兴安岭地区的森林雷击火火险预测.
Daxinganling, Heilongjiang Province is a high incidence area of forest fires.Accordingly, a precise fire risk prediction model is urgently needed to forecast the forest fires.This paper uses the MAXENT model to forecast the fire risk of forest fires.This paper first, based on the forest lightning fire data and environmental variables data in Daxinganling area, Linearity diagnosis of each environmental variable, and then use the cumulative regularization gain method and Jackknife method to evaluate the importance of environmental variables, and finally use the maximum Kappa value and AUC value to test the prediction accuracy of the MAXENT model.The results show that: lightning energy and medium (VIF) of charge and charge were 5.012 and 6.230, respectively, which were in line with other variables and could not be used for model training. The daily rainfall, the number of lightning in cloud field and the flashback current intensity The three most important factors for the occurrence of lightning fire have little effect on daily average wind speed and aspect.With the increase of the ratio of modeling data, the maximum Kappa value and AUC value have an increasing trend.The maximum Kappa value is greater than 0.75, the average Value of 0.772; AUC values are greater than 0.5, with an average of 0.859.MAXENT model accuracy of the forecast to achieve medium accuracy, can be applied to Daxing Forest fire lightning fire Ridge area forecast.