Vegetation phenology is a sensitive indicator of ecosystem response to climate change, and plays an important role in the terrestrial biosphere. Improving our understanding of alpine vegetation phenology dynamics and the correlation with climate and grazing is crucial for high mountains in arid areas subject to climatic warming. Using a time series of SPOT Normalized Difference Vegetation Index (NDVI) data from 1998 to 2013, the start of the growing season (SOS), end of the growing season (EOS), growing season length (GSL), and maximum NDVI (MNDVI) were extracted using a threshold-based method for six vegetation groups in the Heihe River headwaters. Spatial and temporal patterns of SOS, EOS, GSL, MNDVI, and correlations with climatic factors and livestock production were analyzed. The MNDVI increased significantly in 58% of the study region, whereas SOS, EOS, and GSL changed significantly in < 5% of the region. The MNDVI in five vegetation groups increased significantly by a range from 0.045 to 0.075. No significant correlation between SOS and EOS was observed in any vegetation group. The SOS and GSL were highly correlated with temperature in May and April–May, whereas MNDVI was correlated with temperature in August and July–August. The EOS of different vegetation groups was correlated with different climatic variables. Maximum and minimum temperature, accumulated temperature, and effective accumulated temperature showed stronger correlations with phenological metrics compared with those of mean temperature, and should receive greater attention in phenology modeling in the future. Meat and milk production were significantly correlated with the MNDVI of scrub, steppe, and meadow. Although the MNDVI increased in recent years, ongoing monitoring for rangeland degradation is recommended.