引用本文:
【打印本页】   【下载PDF全文】   View/Add Comment  Download reader   Close
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 93次   下载 98 本文二维码信息
码上扫一扫!
分享到: 微信 更多
基于Logistic回归模型的浙江省林火发生驱动因子分析
蔡奇均, 曾爱聪, 苏漳文,等
福建农林大学 林学院,海峡两岸红壤区水土保持协同创新中心
摘要:
【目的】对浙江省2000-2016年林火发生的主要影响因素进行分析,确定驱动因子,建立林火预测模型并划分火险等级,为我国东南林区林火预防与管理提供科学支撑。【方法】以卫星火点数据为基础,运用Arcgis 10.2等软件,对气候因素(气温日较差、月平均气温、月平均降水量、月平均相对湿度、上年月平均降水量、上年月平均相对湿度、日累积降水量、日相对湿度、日平均风速)、地形因素(海拔、坡度)、植被数据(植被覆盖度)、社会基础设施因素(人口密度、人均GDP、距居民点距离、距铁路距离、距公路距离)进行空间信息提取,并与随机点结合后通过Logistic回归模型分析浙江省林火发生的驱动因子;使用“标准化系数”方法检验各林火驱动因子对林火发生的相对重要性;利用ROC曲线方法对模型预测能力进行拟合检验,并计算划分林火发生概率的最佳阈值。【结果】气温日较差、日相对湿度、月平均相对湿度、上年月平均降水量、日累积降水量、植被覆盖度、上年月平均相对湿度、月平均降水量、距居民点距离、距公路距离、距铁路距离等11个因素与林火发生存在显著相关;模型总体预测准确率达到79.1%;计算出浙江省划分林火发生概率的最佳阈值为0.458。【结论】浙江省高火险区主要位于东部和南部地区,东部地区的林火管理重点应放在人为活动的管理与防火宣传教育上,而南部地区应当在林火高发区增设瞭望塔和视频监测设备。
关键词:  林火发生;Logistic回归模型;林火驱动因子;火险区划;卫星火点数据  浙江省
DOI:
分类号:
基金项目:国家自然科学基金项目(31770697);福建农林大学杰出青年基金项目(XJQ201613);福建农林大学国际科技合作与交流项目(KXB16008A)
Driving factors of forest fire in Zhejiang province based on Logistic regression model
CAI Qijun, ZENG Aicong, SU Zhangwen,et al
Abstract:
【Objective】This study analyzed main influencing factors of forest fire occurrence in Zhejiang province from 2000 to 2016,determined the driving factors,established the forest fire prediction model and classified the fire risk rating to provide effective management for forest fire prevention.【Method】The Logistic regression model was used to analyze the driving factors of forest fires in Zhejiang based on satellite fire data and spatial information of climatic factors (daily temperature difference,average monthly temperature,average monthly precipitation,average monthly relative humidity,average precipitation in the previous year,average relative humidity in the previous year, daily average precipitation,daily relative humidity,and daily average wind speed),topographic factors (elevation and slope),vegetation factors (vegetation coverage) and infrastructure factors (population density,per capita GDP,distance to settlement,distance to road,and distance to railway) extracted by Arcgis 10.2.The “normalized coefficient” method was performed to test the relative importance of selected driving factors (variables).The ROC curve method was applied to test the predictive ability of the model and calculate the optimal threshold (cut-off value) for forest fire risk division.【Result】Eleven factors including daily temperature difference,daily relative humidity,average monthly relative humidity,average precipitation of the previous year,daily average precipitation,vegetation coverage,average relative humidity of the previous year,average monthly precipitation,distance to settlement,distance to road,and distance to railway were significantly related to fire occurrence.The overall prediction accuracy of the model was 79.1% with the cut-off value of 0.458.【Conclusion】The high fire risk zones in Zhejiang were mainly located in the eastern and southern regions.In the eastern,forest fire management should focus on the management of human activities, fire prevention publicity and education.In contrast,more watchtowers and video monitoring equipment should be allocated in the southern region.
Key words:  fire occurrence  Logistic regression model  forest fire driving factors  fire danger zone  satellite fire data  Zhejiang province

You are the NO.4583066
Copyright©2009:Editorial Department of Journal of Northwest A&F University (Natural Science Edition)
Designed by Beijing E-Tiller Co.,Ltd