摘要
在分析影响森林可燃物含水率关键气象因子的基础上,利用多元回归、分类与回归树(CART)等方法分别构建了云南松(阴坡)、云南松(阳坡)、华南松和侧柏4种不同林型的活、死可燃物含水率预测模型。结果表明:多元回归模型对不同林型森林活、死可燃物含水率预测的平均相对误差介于5.90%~6.60%、20.1%~36.9%;CART模型在基于气象因子的森林可燃物含水率预测中适用,其对活可燃物含水率预测的最优平均相对误差(5.38%~7.00%)明显低于死可燃物(22.88%~26.64%),这与多元回归模型一致且精度普遍更高,同时解决了多元回归模型无法预测云南松(阳坡)活可燃物含水率的问题。
The prediction of forest fuel moisture content is of great significance to the forecast of forest fire risk and the protection of forest ecosystem.Based on the systematic analyses of the key meteorological factors affecting the forest fuel moisture content,the prediction models of the live and dead fuels moisture content of four different forest types,namely,Pinus yunnanensis(shady slope),Pinus yunnanensis(sunny slope),Pinus armandii and Platycladus orientalis,were established utilizing the methods of multiple regression,classification and regression tree(CART),etc.The results showed that the average relative errors of the multiple regression model in predicting the live and dead fuels moisture content in different forest types were between 5.90%~6.60%and 20.1%~36.9%,respectively.CART model was found applicable for the prediction of forest fuel moisture content based on meteorological factors.The optimal average relative errors of the prediction of live fuels moisture content(5.38%~7.00%)were significantly lower than that of dead fuels(22.88%~26.64%),which was consistent with the multiple regression model and generally had higher accuracies.Besides,the problem that the live fuel moisture content of Pinus yunnanensis(sunny slope)cannot be predicted was also solved.The research results are expected to provide some theoretical supports for the establishment and accuracy improvement of further prediction model of forest fuel moisture content and even the forecast model of forest fire risk.
作者
袁晓玉
杨晓丹
王中玉
YUAN Xiaoyu;YANG Xiaodan;WANG Zhongyu(Public Weather Service Center,China Meteorological Administration,Beijing 100081,China;Department of Mathematical and Physical,North China Electric Power University,Beijing 102206,China)
出处
《科技导报》
CAS
CSCD
北大核心
2023年第16期124-135,共12页
Science & Technology Review
基金
国家重点研发计划项目(2020YFC1511602)。
关键词
气象因子
森林可燃物
含水率预测
多元回归模型
分类与回归树
meteorological factors
forest fuel
moisture content prediction
multiple regression model
classification and regression tree
