Understanding the spatial pattern of the potential natural vegetation (PNV) is import for identifying the response of spatial pattern to the climate change and carbon cycling and for predicting net primary production (NPP) distribution of PNV in ecological restoration projects. The dynamic characteristics of spatial distribution, the potential vegetation of NPP, the gross and cumulative NPP and the relationship between NPP and hydro-thermal factors on the upper research of Heihe River were investigated by using meteorological data including rainfall, temperature and accumulative temperature ≥0℃ (1960-2009) in conjunction with the location and altitude record from 12 climate stations in western China covering the period from 1960 to 2009. Based on these data, we used a comprehensive sequential classification system (CSCS) method, validated at regional and global scales, to estimate the NPP variation of grassland ecosystems and the responses to climate change. PNV's fell into 8 classes based on vertical zones. Increasing altitude in the mountainous region resulted in NPP initially tending to rise but subsequently decline. The predictive distribution of NPP was dependent on hydro-thermal factors. The relationship between NPP and annual accumulated temperature ≥0℃ was negative in cold areas, but was positive in cool temperate areas. There were positive correlations between NPP and humidity in arid, semiarid, sub-humid and humid areas. The spatial pattern of PNV reflected the ability of vegetation to adapt to different habitats and climate change.