中文摘要：

内陆干旱区天然植物群落与当地居民生活质量、区域生态安全及地质环境关系密切。以敦煌盆地为研究区,选择Rapid Eye卫星遥感数据开展植物群落遥感制图研究。（1）根据1∶250 000地貌图及1∶200 000水文地质图设置了天然植被样方调查路线,据此实地调查了植物物种、高度、频度、长势等,区内共记录植物31种,分属16科,主要有9种优势植物,分别为红柳、胡杨、芦苇、骆驼刺、白刺、黑果枸杞、獐茅、蒙古韭和盐穗木;（2）结合地貌图及水文地质图,提出了以Rapid Eye数据为主、多源地学数据支撑的过渡带天然植物群落类型的确定方法：利用植物样方数据集确定单方向上一种特定植物类型到另一种植物类型之间的变化边界;分析Rapid Eye遥感图像的光谱、纹理和几何特征,基于Rapid Eye卫星遥感数据确定沿单方向变化的植物类型边界;在植物样方数据集和遥感图像无法精确确定边界的情况下,利用微地貌和地下水流动系统空间分布,确定区域空间上相邻的植物类型边界。基于这些数据集合方法建立了研究区植物群落的遥感解译标志;（3）采用优势种命名法,划分了研究区植物群落分类体系,完成其遥感制图,并分析了其空间分布分别与地貌及土壤盐渍化之间的关系。

英文摘要：

Natural plant communities in inland arid areas and the quality of life of local residents, regional ecologi- cal security and geo-environment are closely related. The Dunhuang Basin, located in the northwestern China, is world-famous for its natural oasis and cultural heritage and selected as the study area. The objective of this paper is to map the distribution of natural plant communities in the area. This not only can provide long-term monitoring data of changes for studying on vegetation and groundwater system, but also can provide data support for ecological resto- ration project in the area. Thus, the RapidEye satellite remote sensing data was used to carry out remote sensing mapping of plant communities. First, investigation routines of vegetation sampling were set out according to land- forms maps 1 ： 250 000 and 1 ： 200 000 hydrological maps. In each field site, the vegetation species type and number, vegetation coverage, height, and frequency were recorded. Plants were recorded 31 species belonging to 16 families, and there are nine kinds of dominant plants ： Tamarix ramosissima, Popolus euphratica, Phragmites australis, Alhagi sparsifolia, Nitraria tangutorum, Lycium ruthenicum Murr, Allium mongolicum, Aeluropus trin, Halostachys caspica. Then, combined with the landform and hydrological maps, a multi-source geological data type method was supported by transitional method in order to determine the natural plant communities. The remote sensing interpretation signs of plant communities were established. The dominant species nomenclature was used to divide communities＇ classi- fication system in the study area and its remote sensing mapping was completed. There are 8 categories single domi- nant species communities in the area, taking up 73.73% of total study ing up 22.22% of total study area; 12 categories multilayer structure, area; 11 categories dual dominant species, tak- taking up 4.05% of total study area. Finally, the relationship was analyzed between plant communities and landforms and soi