The article primarily experiments a new method for estimating reservoir submergence loss from landsat. First

satellite image of experimental region

transformed to Gaussian coordinate

is registered to digital terrain image of the same region. Secondly

boundaries of reservoir submergence are determined with elevation and slope of digital terrain model （DTM）. It develops a statistical method for modifying the error caused by calculating an area with grids. Land use of this region is classified by fuzzy pattern recognition

hybrid pattern （fuzzy cluster and fuzzy pattern recognition together）

cluster and maximum-likelihood pattern. The classification accuracy of hybrid pattern is about 79.2 percent.The experiment shows that above method is able to satisfy the request of submergence loss survey for reservoir planning.The article paies special attention to the importance of incorporation of geography information with satellite data. Specially

the region

generally

speaking

is mountainous and study contents include many social and environmental problems which can’t reflected by remote sensing image. Therefore

this incorporation may greatly improve the level of estimating reservoir submergence loss by space renote sensing.