Vegetation phenology changes is one of the most direct and sensitive indicators of global climate change and it has become the focus problem of the word studies. The Qinghai-Tibetan Plateau is a unique geographical unit covered by alpine vegetation types so that it is very important to study the remote sensing monitoring model of these vegetation types' phenology. Firstly, Based on MODIS Normalized Difference Vegetation Index (NDVI) data from 2003 to 2012, we reconstructed the long-term time-series datasets through the combination of Inverse Distance Weighted Interpolation and Savitzky-Golay fitting method. After filtering, the obvious noise is removed but the detail information of vegetation growth is kept well so that the time-series curve is definitely suitable for the extraction of phenology paramethers. Then, we studied the extraction models of the typical vegetation phenology in the Qinghai-Tibetan Plteau with dynamic threshold value method, biggest change slope method and logistic curve fitting method. We compared and analyzed the monitoring results based on the nearly ten years NDVI dataset using the relationship between vegetation growing characteristics and daily mean temperature and then selected the dynamic threshold value method as the best model for typical vegetation phenology extraction in the Qinghai-Tibetan Plateau. Finally, we extracted the phenology information of grassland in the plateau with dynamic threshold value model. After the analysis of nearly ten years vegetation phenology, the results showed that the alpine grassland in the plateau experienced the trend of start of season (SOS) advancing (the ratio is 0.248d/a) as the end of season (EOS) following a more complex rule. The andvanced SOS manily caused by the rise of spring temperature and the influence of precipation is not significant. What's more, the vegetation phenology and variation trends in the plateau showed obvious spatial distribution rule from the southeast to the northwest.