In the last several decades, the responses of vegetation to global changes at regional and global scales have been studied with many mathematical models primarily driven by point meteorological observations. In this study, the net primary productivity (NPP) of Xinjiang, China is simulated using the GLObal Production Efficiency Model (GLO-PEM) which is a semi-mechanistic model of plant photosynthesis and respiration and driven entirely by satellite observations. With the available satellite observation data acquired from NOAA’s Advanced Very High Resolution Radiometer (AVHRR), the seasonal and inter-annual changes of NPP in the Xinjiang area are analyzed for the time period of 20 years from 1981 to 2000. Large spatial variability of NPP is found in this area. The temporal trends of NPP in different regions of the area differed significantly. However, for the whole area the mean annual NPP decreased in the 1980s and increased in the 1990s. Seasonal variations of NPP are large and inter-annual changes are moderate. The correlations between the simulated NPP and the precipitation and temperature suggested that precipitation and temperature played major roles in the variations of NPP.