摘要
对于森林火灾这种严重危害森林生态系统的自然灾害而言,预防大于扑救。为实现对吉林省延边地区的森林火灾进行预测,根据历史火灾资料、气象数据、地形地貌数据、植被数据和人为与社会经济数据建立了延边地区森林火灾发生概率预测模型。研究结果表明:建立的4个林火发生概率预测模型均具有较高的拟合度(模型精度值>0.9)和准确率(>80%);经最优模型筛选,随机森林模型(RF)的综合得分最高,最适用于延边地区的林火发生概率预测模型;最优模型中,最小相对湿度和人口密度的变量重要性最大;延边地区南部的林火发生概率明显高于北部,IV级和V级林火发生风险区也多分布于南部。
For the natural disaster of forest fires,which seriously harm the forest ecosystem,prevention is more important than suppression.To achieve the prediction of forest fires in the Yanbian area of Jilin Province,a forest fire occurrence probability prediction model was established based on historical fire data,meteorological data,topographic data,vegetation data,and human and socio-economic data.The research results showed that the four established forest fire occurrence probability prediction models have high goodness of fit(model precision value >0.9) and accuracy(>80%).After optimal model selection,the Random Forest model(RF) has the highest comprehensive score and is most suitable for predicting forest fire occurrence probability in the Yanbian area.In the optimal model,the variables of minimum relative humidity and population density have the highest importance.The probability of forest fire occurrence in the southern part of the Yanbian area is significantly higher than that in the northern part,and risk areas of grade IV and grade V forest fires were also more distributed in the southern part.
作者
曹丽丽
刘向宇
陈响
于淼欣
谢文君
单仔赫
高博
单延龙
于渤
崔晨曦
Cao Lili;Liu Xiangyu;Chen Xiang;Yu Miaoxin;Xie Wenjun;Shan Zihe;Gao Bo;Shan Yanlong;Yu Bo;Cui Chenxi(Science and Technology Innovation Center of Wildland Fire Prevention and Control(Beihua University),Jilin 132013,P.R.China)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2024年第3期90-96,共7页
Journal of Northeast Forestry University
基金
国家自然科学基金项目(32271881、31971669)。
关键词
延边地区
森林火灾
广义线性模型
机器学习模型
发生概率预测
Yanbian area
Forest fire
Generalized linear model
Machine learning model
Occurrence probability prediction
