Various vegetation indices and environmental parameters derived from remote sensing were used to investigate the spatiotemporal dynamics of grassland aboveground net primary productivity (ANPP) in Northern China from 2000 to 2011. An integrated model for grassland ANPP estimation was proposed based on 209 field observations and the corresponding MODIS reflectance products. Five samples for each of five grassland types were extracted randomly to validate the remote sensing model. The model was then used to derive the spatial and temporal patterns of ANPP for the five grassland types. Additionally, the correlations of ANPP to land surface temperature (LST) and land surface water index (LSWI) were analyzed to explore the response of ANPP to changes in hydrothermal conditions. Results show that different vegetation indices have different sensitivities to diverse grassland types. A notable gradient of ANPP in grasslands from west to east is obvious, and a slight increase in annual ANPP for each type of grassland during 2000-2011 is observed. The spatial pattern of ANPP in grasslands is found to correlate with the gradients of LST and LSWI. The temporal correlation of ANPP in grasslands is significantly negative to LST, but positive to LSWI. For alpine steppe and desert steppe, the ANPP is most sensitive to the change of LST: for desert steppe, the ANPP has the highest sensitivity to the change of LSWI. With this improved understanding of the spatiotemporal dynamics of ANPP as well as the effects of climate change, an efficient and sustainable use of China's grasslands could be possible.