Monitoring the changes of Dongting Lake water, as well as making a comprehensive understanding of the evolution rule has important practical significance for the lake district management, flood control and disaster alleviation. Based on SPOT-VGT NDVI data, adopting dynamic threshold method, this paper extracted every ten-day''s water information from 1998 to 2010. Combined with historical data, the analysis on the variation characteristics of water area was done. Main conclusions are as follows:①For long time series and large amount of remote sensing data, the dynamic threshold method shows significant advantages in information extraction. ②The water area in Dongting Lake presents obvious seasonal change in years and a interannual reduced trend (in rain season). From 1825 to 2010, the water reduction could be divided into four stages, and currently Dongting Lake is in a new round of sharply shrinking period. ③The water area prediction model that is established based on cumulative rainfall data gives satisfactory prediction accuracy.