Vegetation phenology is one of the most prominent and sensitive indicators of the responses of terrestrial ecosystems to climate change. To understand and predict the changes in terrestrial ecosystems, it is important to investigate phenological changes. In this study, we investigated vegetation phenology in grasslands in Xinjiang, which is a typical arid and semiarid region in China. We analyzed the spatiotemporal changes in grassland phenology in response to climate change from 2001 to 2014 using the MODIS MCD12Q2-Land Cover Dynamics global phenology product and concurrent meteorological data. The effects of monthly temperature and precipitation on the grassland start of the growing season (SOS) in spring were analyzed with the partial least squares (PLS) method. We found that the MCD12Q2 phenology data could successfully reflect the SOS of the grassland vegetation in Xinjiang. The perennial mean grassland phenology in Xinjiang exhibited significant differences among elevation zones. For each 1000 m increase in the elevation gradient, the SOS was delayed by 13 days. The SOS of the different grassland types in Xinjiang varied from day of year (DOY) 100 to DOY 136, and grassland SOS advanced at a rate of 0.11 d/a. The relative percentage of advanced area (57.5%) was larger than that of delayed area (42.5%), and the percentages of significantly advanced area (4.6%) and significantly delayed area (2.6%) were low. Temperature plays a decisive role in grassland SOS. Warmer temperatures in early spring and lower temperatures and less precipitation in January were more favorable for grassland SOSs in this region than were other temperature and precipitation regimes, and increased precipitation during the early spring had a more favorable effect on grassland SOS in low-altitude areas than in high-altitude areas.