关键词: 毛乌素沙地, 荒漠化, 遥感信息提取, 监督分类, 景观格局

Abstract: This experiment focuses on the fragile zone in Mu Us Desert where farming, forestry, animal husbandry interlacing ecologically in desertification research as the object. It studies the optimal computer automatic classification of Mu Us Desert based on CBERS and TM remote sensing image data type, and investigates the evolution process of Mu Us Desert (20002013) combined with the evolution of GIS spatial analysis and landscape index quantitative. The experiment shows that, through the maximum likelihood supervised classification method, the highest overall precision of sand type obtained is above 86.21%, which is the ideal means of desertification land classification. A nonlinear relationship exists between environment changes and shifting sands, semishifting sands, semifixed sands，fixed sands. Those four types of sand plaques present to be instability, in which shifting and semishifting sands change significantly in position, going on gradually from southwest to northwest and south. Succession between patches will still occur frequently in a period of time in the future, while the percentage of the four types of sands accounts from 82.29% to 75.07% of the study area, proving the ecological environment is getting better by the governance of Mu Us Desert.

Key words: Mu Us Desert, Desertification, Sand land information extraction, Image classification method, Landscape pattern