根据广州市林火历史数据,在林火发生时间、分布范围、火源因素及与树种、地形之间关系分析的基础上,构造林火泰森多边形,利用泰森多边形面积反应林火发生概率的大小,结合火点与道路距离、海拔、坡度、林分类型等因素,建立林火条件熵模型,通过条件熵衡量林火的易发程度,并对林火易发程度进行区划,分为高发区、多发区、一般区、偶发区和无林火区5个等级,此方法能更加精确地确定林火发生的热点地区,直观地反映林火空间分布规律,使得林业管理部门在森林防火工作上更具针对性,有效防范和控制林火的发生,最大程度地减少林火损失。研究结果表明:1)林火多发月为10月至次年4月,与地形、道路、林分类型等因素密切相关,分布于离道路近,地势起伏的中、低丘陵地区,针叶林和针叶混交林发生林火次数多;2)林火空间分布非平稳性,林火高发区和多发区占4.23%,呈现斑状聚集分布,受地形和人为因素影响明显。偶发区占52.96%,分布范围广,具有突发性和随机性的特点。

According to the fire historical data of Guangzhou,the paper analyzes the fire time,distribution range,ignition factors,relationship with tree species and terrain,and creates forest fire voronoi,and then using voronoi polygon area,shortest distance between the road and fire,elevation,slope and forest stand types,establishes the fire conditional entropy. Finally,it used the inverse distance weighting method to spatial interpolation and divided forest fire susceptive zones into high incidence,frequently occurring,general,accidental and no fire zones.The zones reflect spatial distribution rule of the forest fire,the forestry management can effectively prevent and control the occurrence of forest fires,minimize fire losses.The results show that forest fire happens from October to April of following year.Forest fire is closely related to road,forest stand types,the terrain factors,distribution in low hill zone near the road and upslope,coniferous forest and coniferous mixed forest fire occurred more frequently by human factors.Non-stability fires spatial distribution,high incidence and frequently occurring zone present the patchy gathered accounted for 4.23 percent,accidental zone reflects the fire characteristics of sudden and randomness accounted for 52.96 percent.