Improving the accuracy of satellite-derived precipitation estimates is very important to extend the extent and depth of their application. Existing calibration methods calibrated satellite-derived precipitation estimates based on a strong relationship between gauge measurements and estimated precipitation. However, such a relationship is always disturbed by variation of land surface characteristics, elevation or climate types. Existing methods always assume the study area as homogeneous and seldom consider its spatial heterogeneity. This study proposed a spatial calibration method for the Tropical Rainfall Measuring Mission (TRMM) 3B43 precipitation data, assuming that the relationships between gauge measurements and satellite-derived precipitation estimates vary spatially. These relationships were further explored using geographical weighted regression; consequently the derived relationships were used to construct a spatial calibration model. The monthly, seasonal, and annual precipitation data over Mainland China in 2011 was used as an example. The experiment results show that the proposed method can produce calibrated results accurately, with Nash-Sutcliffe efficiency (NSE) values of 0.72–0.91, 0.83–0.93, and 0.92 for monthly, seasonal, and annual precipitation, respectively. Furthermore, compared with the calibration methods based on regression analysis (RA) and geographical difference analysis (GDA), the proposed method increases NSE values by 1–12% and 0 −16%, respectively, and reduces root mean square error values by 1.7–26.4% and 1.8–36.4%, respectively.