基于光谱混合分析的干旱荒漠区植被遥感信息提取研究——以古尔班通古特沙漠西缘为例  

Extraction of Vegetation Information in Arid Desert Area based on Spectral Mixture Analysis——a Case in the Western Gurbantunggut Desert

作　　者：崔耀平[1,2] 王让会[1] 刘彤[3] 张惠芝[1] 

机构地区：[1]中国科学院新疆生态与地理研究所,新疆乌鲁木齐830011 [2]中国科学院研究生院,北京100049 [3]石河子大学生命科学学院,新疆石河子832003

出　　处：《中国沙漠》2010年第2期334-341,共8页Journal of Desert Research

基　　金：国家林业科技支撑项目(2006BAD26B0902);国家重点基础研发规划(2006CB705800);"十一五"科技支撑子项目(2007BAC17B03)共同资助

摘　　要：干旱荒漠区植被的盖度和空间分布特征是评判该区生态环境状况及荒漠化程度的重要指标。以LandsatTM影像为数据源,利用归一化植被指数(NDVI)和线性光谱混合分析模型(LSMM)两种方法对研究区——古尔班通古特沙漠西缘进行分析。在运用LSMM过程中,通过多种方法选择出最佳端元,并利用实测数据对混合像元分解结果进行验证。结果表明:①NDVI提取植被的方法受到很多限制,不适合在干旱荒漠区应用;②基于最小法非受限光谱混合分解结果较为理想,植被、盐碱地、裸沙和黑色砂粒等4种端元地物被选取出来;③LSMM提取的植被分量与实测植被盖度显著相关,线性相关系数为0.858,表明干旱荒漠区的植被盖度可以通过遥感影像提取的植被分量间接得到。The coverage and spatial distribution of vegetation is a fundamental index to estimate the ecological environment and desertification in arid desert area.Comparison among deserts in the same latitude region,Gurbantunggut Desert is the unique which teems with lots of plant species,although the distribution of vegetation is more sparse than that of some typical desert steppe.In this paper,vegetation coverage of the western Gurbantunggut Desert was selected as the target indicator,which was extracted using normalized difference vegetation index （NDVI） and linear spectral mixture model （LSMM）,based on Landsat TM images.In order to get better result,we chose the best endmember in the LSMM process through different methods,and utilized the measured vegetation coverage data to validate the vegetation fraction that came from remote sensing images.The results showed that：the NDVI method was subjected to many restrictions,and wasn＇t suitable for application in the arid desert area.While,the LSMM showed a better result,thus vegetation,alkaline soil,bare sand and dark sand had been successfully selected for further analyses.A significant linear relationship also was found between vegetation fraction and vegetation coverage,with a correlation coefficient of 0.858.All those reflected that vegetation coverage in the arid desert area can be extracted indirectly through remote sensing images.

关 键 词：光谱混合分析 端元 干旱荒漠区 NDVI 古尔班通古特沙漠 

分 类 号：X87[环境科学与工程—环境工程]