Abstract: The Three River Headwater Region is one of the source areas of China’s major rivers. For better grassland management under future climate change scenarios and for the sustainable development of animal husbandry in this region, it is very important to understand the spatiotemporal dynamic changes of grassland yield and the relationship between grassland yield and local precipitation. In this study, we used MODIS-NDVI (Moderate Resolution Imaging Spectrometer-Normalized Difference Vegetation Index) data and 442 ground sampling data from 2006 to 2013 to construct empirical models for the correlation between NDVI and the yields of three types of grasslands. Then, we used these models to estimate the distribution of grassland yield in The Three River Headwater Region during 2006-2013. We also studied the response of grassland productivity to changes in precipitation patterns in the Three River Headwater Region based on the correlation between grassland yield and annual, monthly, and cumulative precipitation. The major conclusions were as follows: (1) Grassland yield showed an increasing trend during 2006-2013 in the Three River Headwater Region, and the yield of different grassland types showed different responses to annual precipitation. (2) The most important rainfall month for grassland yield was October of the previous year; rainfall in April and May of the current year also strongly contributed to grassland yield. (3) The cumulative precipitation from October (previous year) to May in the current year was the main cause of inter-annual variations in grassland yield. (4) The cumulative precipitation in autumn of the previous year and in spring of the current year was a key factor in current-year grassland output.