文献类型：期刊文献

中文题名：典型植被指数识别火烧迹地潜力分析

英文题名：Potential analysis of typical vegetation index for identifying burned area

作者：孙桂芬[1] 覃先林[1] 刘树超[1] 李晓彤[1] 陈小中[2] 钟祥清[2]

第一作者：孙桂芬

机构：[1]中国林业科学研究院资源信息研究所国家林业局林业遥感与信息技术实验室;[2]四川省林业信息中心

年份：2019

卷号：0

期号：1

起止页码：204-211

中文期刊名：国土资源遥感

外文期刊名：Remote Sensing for Land & Resources

收录：CSTPCD;;北大核心:【北大核心2017】；CSCD:【CSCD2019_2020】；

基金：国防科工局重大专项项目"高分森林灾害监测应用示范(一期)"(编号:21-Y30B05-9001-13/15)和"机载光学全谱段数据处理及林火预警技术研究"(编号:CAFYBB2018SZ009)共同资助

语种：中文

中文关键词：GF-1;WFV数据;Landsat8数据;火烧迹地;植被指数;分离性

外文关键词：GF-1 WFV data;Landsat8 data;burned area;vegetation index;separability

分类号：TP79;S762

摘要：植被指数法是利用卫星遥感影像识别火烧迹地的常用方法之一。植被因受火的干扰会形成火烧迹地,其光谱特征易与裸地、水体、道路、阴影和耕地等地物光谱混淆,使用遥感影像采用合适的植被指数提高过火区遥感监测精度仍是亟待解决的问题。以四川省2014年和内蒙古自治区2017年发生的4次森林火灾形成的火烧迹地作为研究区,利用高分一号16 m宽幅(GF-1 WFV)数据和Landsat8数据的波谱特性,选取归一化植被指数(normalized difference vegetation index,NDVI)、增强型植被指数(enhanced vegetation index,EVI)、全球环境监测植被指数(global environment monitoring index,GEMI)、过火区识别指数(burned area index,BAI)和归一化火烧指数(normalized burn ration,NBR)等5种典型植被指数,通过构建不同植被指数的分离指数M来定量评价这些植被指数识别火烧迹地的潜力。研究结果表明,基于近红外—短波红外波段的NBR和基于可见光—近红外波段的BAI对过火区的分离性较好,NDVI的分离性次之,EVI和GEMI的分离效果较差;基于GF-1 WFV和Landsat8数据采用BAI和NBR指数对内蒙古鄂伦春自治旗火烧迹地进行了识别(其中GF-1 WFV数据只用于BAI识别),并利用高分二号(GF-2)数据进行了精度验证,两者火烧迹地识别总体精度均大于87%,Kappa系数均大于0. 7。
Vegetation index is one of the commonly used method for adopting satellite remote sensing image to identify burned areas.Due to the disturbance of fire,vegetation becomes burned area,and its spectral characteristics are easily confused with the spectra of bare land,water body,road,shadow and arable land and some other factors.Therefore,the improvement of the accuracy of remote sensing monitoring for burned area using appropriate vegetation index remains an urgent problem.In this paper,four burned areas in Sichuan Province and Inner Mongolia where fire burning occurred in 2014 and 2017 were selected as the study areas.Based on the spectral characteristics of Gaofen-1 satellite 16 m wide width(GF-1 WFV)data and Landsat8 data,the authors chose normalized difference vegetation index(NDVI),enhanced vegetation index(EVI),global environment monitoring index(GEMI),burned area index(BAI)and normalized burn ration(NBR)and constructed separation index M of different spectral indices to quantitatively evaluate the potential of different spectral indices for burned areas identification.The results show that NBR calculated with near-infrared and short-wave infrared band and BAI based on visible light-near infrared band have a better capability for separating burned areas,the separability of NDVI takes the second,whereas EVI and GEMI have a poor separability.For GF-1 WFV data and Landsat8 data,BAI and NBR which have a good separate capability for burned area identification were used for the burned area in Oroqen Autonomous Banner of Inner Mongolia to separate burned areas(for GF-1 WFV data,only BAI was used to identify burned area),and Gaofen-2 satellite(GF-2)data which have higher spatial resolution combined with confusion matrix method were used to verify the accuracy.The overall accuracy of both methods were higher than 87%,and the Kappa coefficients were all higher than 0.7.