基于Logistic回归模型的吉安市森林火灾发生概率预测OA
Prediction of forest fire occurrence probability in Ji'an City based on a Logistic regression model
[目的]拟预测江西省吉安市森林火灾发生概率,为吉安市森林火灾精准防控提供科学依据.[方法]基于 2001-2020年MODIS火点数据,结合气象、地形、植被及人类活动等多维因子,分析江西省吉安市森林火灾的时空分布特征及其驱动机制.采用多重共线性诊断和相关性分析筛选关键影响因子,构建了 Logistic 回归(binary logistic regression)模型,预测森林火灾发生概率,并利用混淆矩阵和曲线下面积(area under the curve,AUC)评估模型性能.[结果](1)吉安市森林火灾年际变化呈 5 年周期性波动,主要发生在 9 月至次年4 月,空间分布呈现北多南少、西多东少的特征;(2)人口密度、上月植被指数、海拔、本月降雨量、上月温度和灯光指数是火灾发生的主要驱动因子,其中本月降雨量和灯光指数与火灾风险呈正相关,其余因子呈负相关;(3)火灾发生概率在 0.2～0.7,永丰县、安福县、永新县、吉安县和遂川县为高风险区;(4)模型AUC值为 0.748,具有较好的预测能力.[结论]研究可为吉安市森林火灾风险管理提供科学依据,建议在高风险区域加强监测预警,并针对不同驱动因子采取差异化防控措施.
[Objective]This study aimed to predict the probability of forest fires in Ji'an City,Jiangxi Province,providing a scientific basis for precise prevention and control of forest fires in the region.[Method]Based on MODIS fire point data from 2001 to 2020,combined with multi-dimensional factors such as meteorology,topography,vegetation,and human activities,the spatio-temporal distribution characteristics and driving mechanisms of forest fires in Ji'an City,Jiangxi Province,were analyzed.Multiple collinearity diagnosis and correlation analysis were used to screen key influencing factors,and a Logistic regression(Binary Logistic Regression)model was constructed to predict the probability of forest fires.The performance of the model was evaluated by using a confusion matrix and the area under the curve(AUC)value.[Result](1)The interannual variation of forest fires in Ji'an City exhibited a 5-year cyclical fluctuation,with most fires occurring between September and April of the following year,and the spatial distribution showed a pattern of higher incidence in the north than in the south,and more in the west than in the east.(2)Population density,vegetation index of the previous month,altitude,rainfall of the current month,temperature of the previous month,and nighttime light index were the main driving factors of fire occurrence,among which current-month rainfall and nighttime light index were positively correlated with fire risk,while the other factors showed negative correlations.(3)The probability of fire occurrence ranged between 0.2 and 0.7,with Yongfeng County,Anfu County,Yongxin County,Ji'an County,and Suichuan County identified as high-risk areas.(4)The model achieved an AUC value of 0.748,demonstrating good predictive performance and providing a scientific basis for targeted forest fire prevention and management in Ji'an City.[Conclusion]A scientific basis for forest fire risk management in Ji'an City can be provided by this study.It is recommended that monitoring and early warning systems be strengthened in high-risk areas,and that differentiated prevention and control measures be implemented according to the various driving factors.
焦鸿渤;叶清;彭佳慧;郑育桃;曹子琪;陈俊松
新余市林业局,江西 新余 338000江西农业大学 林学院,江西 南昌 330045||江西农业大学 鄱阳湖流域森林生态系统保护与修复国家林业和草原局重点实验室,江西 南昌 330045江西农业大学 林学院,江西 南昌 330045||江西农业大学 鄱阳湖流域森林生态系统保护与修复国家林业和草原局重点实验室,江西 南昌 330045江西省林业科学院 园林规划设计研究所,江西 南昌 330013江西农业大学 林学院,江西 南昌 330045江西省林业科学院 园林规划设计研究所,江西 南昌 330013
农业科技
森林火灾发生概率森林火险MODISLogistic回归
forest fire occurrence probabilityforest fire riskMODISLogistic regression
《生物灾害科学》 2026 (1)
75-86,12
江西省林业局林业科技创新专项([2022]12号)
评论