Change detection based on remote sensing images is a hot research topic, which can provide decision support for regional ecological conservation, resource management and development planning, etc. However, the current studies mostly focus on bio-temporal change detection, which cannot adequately reflect the continuous change characteristics in time dimension. In this paper, through the introduction of information theory, we propose a novel method of temporal information entropy to characterize the long time change features of vegetation comprehensively. Taking Yanhe watershed as the study area and based on Normalized Difference Vegetation Index (NDVI) data, we used this method to calculate the change information of vegetation coverage in Yanhe watershed from 2000 to 2010 and identified its spatio-temporal change characteristics. The results showed that the change level of vegetation coverage in Yanhe watershed was mostly increased in this period, accounted for 80.7% of the total area. Obviously increased area accounted for 13.9% and mainly distributed in northeast and southeast of the watershed. Vegetation coverage decreased area accounted for 2.4%, mainly distributed in the western and northwestern; Severely decreased area accounted for 1.1%, mainly distributed in the central and southwest of the study area, where needs strict environmental management. Comparing with the regression analysis method, temporal information entropy can be more objectively in reflecting the continuous change intensity and trend of vegetation coverage in long time series, which can provide scientific theoretical references for the protection and management of regional ecological environment.