This paper presents a regional net primary productivity (NPP) model parameterized with remote sensing information of MODIS, and tests it in areas of Inner Mongolia semi-arid grassland. The model uses the concept of Light Use Efficiency (LUE), while change of air temperature, plant-water deficits and respiration reduce the efficiency of light usage, and the model is modified to couple with ecology model to better represent regional environmental characteristics of the semi-arid grasslands. In this study, the two different ways to simulate NPP in the process of vegetation growth were introduced to calibrate model parameters. The accuracy of model products was assessed by comparing them with field measurement and other models. The NPP simulated has good linear relation with the measured (R2=0.948). The promising results suggest this model could provide spatial and temporal distribution of net primary productivity for semi-arid grassland management.