We compiled an extensive database of erosion and runoff measurements on erosion plots under natural rainfall in China. We used this database to analyse how soil loss by sheet and rill erosion and runoff in China were affected by land use, slope gradient, slope length and mean annual precipitation. Our results show that land use dominates the variation of soil loss and runoff: Soil loss and runoff rates on land covered by grass and trees are one to three orders of magnitude lower than rates on cropland. Slope gradient and slope length affect soil loss and runoff rates on cropland but there is no statistically significant effect on either soil loss or runoff on plots with a permanent vegetation cover. Runoff rates consistently increase with mean annual precipitation. The relationship between soil loss and mean annual precipitation is, on the contrary, nonlinear for all land use types, with a clear increase of soil loss with precipitation up to a mean annual precipitation of ca. 700 mm yr-1, a subsequent decline and a second rise when the mean annual precipitation exceeds ca. 1400 mm yr-1. We attribute this non-linear response to the interplay of an increasing rainfall erosivity and an increasing protection due to vegetation cover with increasing mean annual precipitation. This non-linear response implies that the effect of precipitation changes induced by climate change on the erosion risk depends on how both rainfall erosivity and vegetation cover change with changing climate. Our study provides important insights as to how soil loss and runoff in China are related to controlling factors and this will allow improving assessments of total soil erosion and runoff rates over the entire territory of China.