Qinghai Lake is the largest one of inland lakes in China. Some issues about water level drop, lake area shrinking and lake separateies have brought to broad attentions in recent years. Remote sensing researches for these issues were discussed in this paper. Landsat MSS, TM, ETM⁺ images and Chinese FY-1C, FY-1D, HY-1A satellitic data from 1975 to 2002 were applied to this study. Furthermore, we collected some useful data including relief map, water depth map and hydrological data. Those satellitic data, combined with useful data, were processed and analyzed, some results revealed that the lakeshore physiognomy have greatly changed from 1975 to 2000. The region changed mostly located in the west shore near the Bird Island and Buha River, in the north shore-Shaliu River and the east shore-Sha Island and Haiyan Bay. Due to rich land-carriage materials from Buha River and lake currents action, the west shore of lake is progressing towards lake. The famous Bird Island and Haisepi Island become the part of land; Also, the east shore was accumulated eroded by the removal of sand dune. Shadao Lake and Haiyanwan Lake were gradually separated from Qinghai Lake. On basis of lakeshore line change, we estimated that lake water level dropped about 2 m from 1975 to 2000, by the approach to matching two images and contrasting water depth data. This numeric value estimated was close to hydrological observation value 1.71 m. Also, this study estimated that the area of Qinghai Lake was 4405.64 km² in 1975, and was 4256.04 km² in 2000. FY-1D image on June, 2002 and field investigating results revealed that Haiyanwan Lake with an area or 100 km² has been separated from Qinghai Lake. Water level drop and lake variation were the main cause of lake shrinking. In addition, we carried through field investigating and lake water depth measuring in the south-easters lakeshore of Qinghai Lake in Sep., 2002. The study indicated that the relativity of the grey scale of TM1 band image and water depth was very high in this lake. As a result, the water depth retrieval model was established.