The revised universal soil loss equation (RUSLE) is an empirical model to predict water erosion (soil erosion); however, the model calculation has limitations at large scale. We calibrated the rainfall erosivity factor (R factor) based on climate zones and the cover-management factor (C factor) based on land-cover types and agricultural crops. The support practice factor (P factor) was also revised based on crop types and land slope. Using the RUSLE, we estimated the annual soil erosion reference values in China in 2000, 2005, 2010 and 2015. The results indicate that: (1) in 2015, the hotspots with major erosion rates are predicted to occur in downstream Yangtze River, Yunnan-Guizhou Plateau, Loess Plateau, and the foothill area of Kunlun Mountains, accounting for 9.65% of the statistical area. (2) The hotspots with a rapid increase of erosion during the study period are located in the arable area of Xinjiang, Sichuan Basin, southeastern Yunnan-Guizhou Plateau, Yangtze Plain and Northeast Plain, which cover 10.36×10⁴ km². (3) The erosion areas with a significant decrease are distributed in the southern and eastern Loess Plateau, Qinling Mountains and Southeast coast of China. These regions cover an area of 13.60×10⁴ km². By calibrating the parameters of RUSLE model, this study obtained the spatial and temporal differentiation rules of soil water erosion at a national scale, which could provide a basis for formulating soil restoration measures in different regions.