基于对象方法的南岭山区雨雪冰冻灾害遥感检测与空间分析  

Forest Damage Detection from an Freezing Rain and Snow Using Remote Sensing Based on Object-oriented Method and Spatial Analysis in Nanling Mountain Areas

作　　者：朱俊凤[1,2,3] 钱峻屏[2] 廖广社 王芳[5] 何克军[4] 张运英[2] 

机构地区：[1]中国科学院广州地球化学研究所,广东广州510640 [2]广州地理研究所,广东省遥感与GIS应用重点实验室,广东广州510070 [3]中国科学院研究生院,北京100049 [4]广东省林业局,广东广州510173 [5]广州大学,广东广州510006

出　　处：《山地学报》2010年第4期478-486,共9页Mountain Research

基　　金：广东省科技计划项目(2007B020500002-9);国家自然科学基金项目(40801034);广州市高校科技计划项目(08C025)资助~~

摘　　要：针对2008年初的南方雨雪冰冻灾害,以受灾严重的广东南岭山区为研究区域,采用面向对象的方法进行森林植被破坏遥感检测研究,首先得到林地分布图,然后用K均值聚类法进行基于对象NDVI差值的森林受灾程度分级,得到灾害等级地图,最后用分带统计法对灾害的空间分布形态进行了分析。林地提取的总体精度为95.33%,灾害分级的总体精度为87.27%,Kappa系数为80.74%。灾害空间分析表明:高程上,600～1600m主要为重度、中度受灾;600～1300m,高程越高,受灾比例越大;1300～1600m,比例有所下降;400m以下和1600m以上主要为轻度受灾;坡度上,随着坡度的增大,重度、中度受灾比例有所增加,但差异并不十分明显;坡向上,各级灾害比例几乎没有差别。Mapping forest damage quickly and accurately using remote sensing technology is great significance for forestry disaster assessment and disaster recovery and reconstruction, as well as lofig-term monitoring of forest eco- systems. An extreme freezing rain and snow in early 2008 resulted in significant forest damage across South China. This paper took Nanling mountain areas in Guangdong Province which was hard-hit as study area to detect forest damage using remote sensing based on object-oriented method. First, we got the distribution of forest land, then got the disaster level map using K-means cluster based on object NDVI difference, finally, we analyzed spatial patterns of forest disaster using zonal statistic. The overall accuracy of forest land extraction is 95.33 %, the overall accura- cy of disaster classification is 87.27% and Kappa coefficient is 80. 74%. Spatial analysis of forest disaster showed that the damage at 600 - 1 600 m was mainly major and moderate disaster. At 600 - 1 300 m, the higher the eleva- the greater the proportion of major and moderate disaster, at 1 300 - 1 600 m, the proportion decreased. The damage at 0 - 400 m and 1 600 - 1 900 m was mainly minor disaster. The proportion of major and moderate disaster increase as the slope increased, but the difference was not very obvious. The proportion of all kinds of disasters was almost no difference on aspect.

关 键 词：雨雪冰冻 南岭 面向对象 遥感 归一化植被指数(NDVI) 空间分析 

分 类 号：P429[天文地球—大气科学及气象学]