中文摘要：

针对黑河流域植被类型空间分布的垂直地带性特征,本文基于支持向量机算法构建黑河流域植被类型空间分布的模拟分析模型,并运用Kappa系数和混淆矩阵检验方法对模拟精度进行检验。验证结果显示,模型总体精度（OA）值为75.54%,Kappa系数值为0.66,表明了该方法在植被分布模拟上具有较好的结果,适用于区域尺度下植被类型分布的空间模拟。模拟结果表明,该方法对半灌木-矮半灌木荒漠和温带禾草-杂类草草甸草原类型的模拟精度最高,分别为（90.20%和90.02%）;分布面积最大的植被类型（如半灌木-矮半灌木荒漠,灌木荒漠、嵩草-杂类草高寒草甸等）相比于其他面积较小的植被类型具有显著优异的模拟结果;人工经济作物、荒漠植被类型以及草原草甸等植被类型对于所选环境因子的敏感性更强,而灌丛类型和乔木类型的模拟结果在不同类型间的波动较大;空间分布上,环境要素差异性明显、植被类型丰富的上游地区具有更好的模拟结果,优于地势平坦、气候差异性小的黑河中下游地段,但模拟结果在景观形态上具有更高的破碎度。

英文摘要：

According to the vertical zonality characteristic of vegetation type in Heihe River Basin, we established an analysis model of the vegetation distribution of Heihe River Basin at large scale based on Support Vector Machine algorithm. Kappa coefficient and the confusion matrix were used to validate the accuracy and performance of the model. The Overall Accuracy（OA）is 75.54% and Kappa coefficient is 0.66, indicating that this method was qualified to simulate vegetation distribution at regional scale. The results show that, semi-shrub,dwarf semi-shrub desert and temperate grasses-forbs meadow steppe have the highest simulation accuracy with OA of 90.20% and 90.02%, respectively. Vegetation types with large area such as semi-shrub and dwarf semishrub desert, shrub desert and Kobresia spp-forb high-cold meadows have much better accuracy than other vegetation types with small area. Artificial economic crops, desert vegetation types, and grassland and meadow are more sensitive to the chosen environmental factors. For shrub and arbor, simulation results differ among vegetation types. In the aspect of spatial distribution, upstream area with obvious distinctions in both vegetation types and environmental factors, has a better simulation results than middle and downstream area of Heihe River Basin, which are flat in terrain and have a small climate variation. Also, the simulation results of the upstream area have a higher degree of fragmentation in the landscape pattern.