基于Himawari-8卫星数据的林火判别  

Discrimination of forest fire based on himawari-8 satellite data

作　　者：冯豁朗 张贵[1] 谭三清[1] 吴鑫[1] FENG Huolang;ZHANG Gui;TAN Sanqing;WU Xin(College of Forestry,Central South University of Forestry&Technology,Changsha 410004,Hunan,China)

机构地区：[1]中南林业科技大学林学院,湖南长沙410004

出　　处：《中南林业科技大学学报》2021年第8期75-83,共9页Journal of Central South University of Forestry & Technology

基　　金：湖南省科技创新平台与人才计划项目(2017TP1022);湖南省应急管理科技项目(2020YJ007);中南林业科技大学青年科学基金项目(2018YC006)。

摘　　要：【目的】Himawari-8卫星数据具有高时效性,综合影像在时间和空间两个维度的亮温差异,研究林火判别方法,对林火进行实时的监测,有利于准确及时发现林火并实施应急处置。【方法】根据Himawari-8影像的中红外7通道(3.9μm)和热红外14通道(11μm)的亮温差异,在空间上利用影像像元值、影像均值和影像标准差之间比值关系进行潜在火点判别,时间上采用阈值法对前后时刻影像进行持续火点判别。采用中国森林草原防灭火网数据、日本气象厅WLF火点产品数据对判别结果进行精度验证。【结果】1)通过对102起发生林火时潜在火点的判断因子Z_(BT07)和Z_(BT74)值进行统计分析,得到Z_(BT07)的判断阈值为0.85,Z_(BT74)的判断阈值为1.6。2)通过对2020年4月1日22时10分发生在云南省永仁县的一场林火进行连续观测,完整地监测到林火从发生到结束的全过程。结果表明该方法能够对林火进行连续性监测,同时也能反映火势大小、火灾蔓延方向和火灾持续时间等信息。3)2020年11月7日05时06分湖南省醴陵市发生一场小型林火,通过与前人火点算法进行比较,结果表明前人算法在火灾初期没有识别到林火,而本研究方法识别到了林火。4)中国森林草原防灭火网的数据显示2020年4月1日云南省境内发生林火22起,判别结果24起林火中有19起与其一致,正确率达到0.79,漏检率为0.14,准确率和漏检率的综合评价值为0.82;日本气象厅WLF daily产品数据显示2020年4月1日云南省境内火点总计181个,判别结果中163个火点中有137个与其一致,正确率达到0.84,漏检率为0.24,准确率和漏检率的综合评价值为0.8。【结论】利用Himawari-8卫星数据,通过林火判别方法对林火进行监测,能够及时准确地识别林火,可为林火监测和应急处置提供科学依据。【Objective】Himawari-8 satellite data has high timeliness characteristics,and the integrated image bright temperature differences in both time and space dimensions,research forest fire discrimination methods,near real-time discrimination monitoring of forest fires,is conducive to timely detection of forest fires and implementation of emergency response.【Method】Based on the bright temperature difference between the 7 channels(3.9μm)and 14 channels(11μm)of mid-infrared and thermal infrared of Himawari-8 images,the potential fire points were discriminated spatially using the ratio between image element value,image mean value and image standard deviation,and temporally using the threshold method to discriminate continuous fire points in the before and after time images.The accuracy of the discrimination results was verified by using the data of China forest and grassland fire prevention and suppression network and the WLF product of Japan Meteorological Agency.【Results】1)By statistically analyzing the values of Z_(BT07) and Z_(BT74) when 102 forest fires occurred,the judgment thresholds of Z_(BT07) was obtained as 0.85 and that of Z_(BT74) was 1.6.2)By conducting a statistical analysis of a fire that occurred in Yongren County,Yunnan Province,at 22:10 on April 1,2020 fire with continuous observation,the whole process of the fire from the occurrence to the end was monitored completely.It shows that the method is capable of continuous monitoring of forest fires,and also reflects information on fire size,fire spread direction and fire duration.3)A small forest fire occurred in Liling city,Hunan province,at 05:06 on November 7th,2020,and the sensitivity of the method was verified by comparing it with the original time-series fire point algorithm.The results showed that the original time-series fire point algorithm did not identify forest fires at the early stage of the fire while the method in this paper identified forest fires.4)The data of China forest and grassland fire prevention and suppression network

关 键 词：Himawari-8 林火判别 潜在火点 持续火点 

分 类 号：S762.32[农业科学—森林保护学]