基于决策树方法的青藏高原温泉区域高寒草地植被分类研究  

Vegetation classification of alpine grassland based on decision tree approach in the Wenquan area of the Qinghai-Tibet Plateau

作　　者：张秀敏[1] 盛煜[1] 南卓铜[2] 赵林[3] 周国英[4] 岳广阳[3] 

机构地区：[1]中国科学院寒区旱区环境与工程研究所冻土工程国家重点实验室,甘肃兰州730000 [2]中国科学院寒区旱区环境与工程研究所,甘肃兰州730000 [3]中国科学院寒区旱区环境与工程研究所青藏高原冰冻圈观测试验研究站,甘肃兰州730000 [4]中国科学院西北高原生物研究所,青海西宁810001

出　　处：《草业科学》2011年第12期2074-2083,共10页Pratacultural Science

基　　金：国家重点基础研究发展计划(973)项目(2010CB951402);科技部基础性工作专项"青藏高原冻土本底调查"(2008FY110200)

摘　　要：植被指数作为植被生长状态的最佳指示因子,已成为植被分类的重要手段之一。为了解青藏高原温泉区域高寒草地植被的分布状况,利用野外样方调查获得植被点数据,结合MODIS/EVI遥感影像数据及数字高程(DEM)数据,综合分析各种高寒草地植被类型的EVI时序曲线特征及其生长环境的高程、坡度和坡向等地形特征,建立知识库并采用决策树分类算法对该区域的高寒草地植被分类进行研究。结果表明,总体分类精度为72%,Kappa系数是0.6,决策树方法能有效地分类和识别具有相似EVI时序特征的高寒草地植被。Vegetation index as an important physical parameter indicating the plant growth was widely used in vegetation mapping and vegetation classification.In order to understand the distribution patterns of alpine vegetation in the Wenquan area of the Qinghai-Tibet Plateau,the field survey data from 283 samples and MODIS EVI and DEM was applied to determine EVI temporal characteristics and terrain characteristics of various vegetation types（including elevation,slope grade and slope direction）,and these characteristics data were used to establish the relative knowledge database.The established database was used to map vegetation classification of the Wenquan area by using decision tree approach.This study showed that the overall accuracy and kappa coefficient was 72% and 0.60,respectively,and this implied that the decision tree approach was effective to classify and identify the alpine grassland type by EVI data with similar phenological characteristics.

关 键 词：温泉区域 EVI 地形特征 决策树 植被分类 

分 类 号：S812.3[农业科学—草业科学]