中文摘要：

采用2002年MODIS 1km的全年NDVI时序数据对新疆及周边地区进行了土地覆盖分类,在分类的过程中重点强调了稀疏植被覆盖区域,这些区域具有潜在荒漠化的趋势.介绍了一种针对不同土地覆盖类型并能重点突出稀疏植被的分类方法,这种方法较好地综合了季节性影响因素和多变的自然条件影响因素.从16天合成的优化过的时序NDVI图像上,通过分析物候变化,可以获得较好的分类结果.将具有潜在荒漠化趋势的区域模型化研究以后,研究结果表明新疆及周边地区40万km^2的土地有潜在荒漠化的趋势.由于MODIS NDVI数据覆盖范围较大,并且对植被的生长变化有较高的敏感度,所以它可以被有效地应用于监测大尺度环境变化和荒漠化进程.

英文摘要：

Detecting desertification with remote sensing is an important task for the North west of China and other arid areas in the world. In this paper, MODIS 1 km NDVI time series data of 2002 was used to classify land cover type in and around Xinjiang Province. During the classification, sparse vegetation areas which are at risk of deserti- fication are emphasized. A classification method, which is adapted to various land cover type and makes sparse vegetation emphasis, is introduced in this paper. This method better synthesizes seasonal influence factors and vari- able factors of natural conditions. From optimized time-series NDVI images from 16-day composites, better classification results can be achieved through analyzing seasonal changes. After modeling areas at risk of desertification, research results indicate that 0.4 million km2 around and in Xinjiang are potential desertification regions. Because of the larger cover extent and sensitivity to vegetation growth of MODIS NDVI imagery, it can be effectively applied to detect large scale environmental variances and desertification.