Abstract: Taking three major river basins in China as the study areas, this study tries to reveal the difference between downscaled results of TRMM 3B43 and GPM IMERG precipitation based on different Geographically Weighted Regression (GWR) models. Normalized Difference Vegetation Index (NDVI) and Digital Elevation Model (DEM) were selected as the basic ancillary variables of GWR models. Different combinations of NDVI and DEM were employed to downscale TRMM and GPM precipitation of 2014 summer, which were then evaluated based on rain gauge observations. The results demonstrate that, （1） Overall accuracy of the original GPM precipitation data is better than TRMM. Accuracy indices imply that both data have higher accuracy in the Yellow River Basin than in the Yangtze River Basin and the Pearl River Basin. （2) After downscaling, the spatial resolutions of both precipitation data have been significantly improved. The downscaled results demonstrate more details and trends of precipitation variation, but their accuracy is more dependent on the original data, which makes the accuracy improvement not significant. （3) Using GWR model with both NDVI and DEM as ancillary variables, difference between downscaled results of TRMM and GPM is significant in high altitude areas of the Yangtze River Basin and Pearl River Basin. （4) The downscaled results with different auxiliary variables in the Yellow River Basin are generally similar in spatial distribution. Downscaling TRMM using terrain as the only auxiliary variable generates the best results, while for GPM, GWR models with different auxiliary variables show no obvious difference.