Knowledge on rainfall erosivity (R) of particular sites is fundamental for soil loss estimation by the Universal Soil Loss Equation (USLE) and therefore highly important in conservation planning. In order to obtain the R value estimates for places where it is unknown, an artificial neural network (ANN) was developed for the state of São Paulo, and its accuracy compared with the Inverse Distance Weighted (IDW) interpolation method. The developed ANN presented a smaller mean relative error in the R estimation and a confidence index classified as "excellent", better than the IDW. ANN can therefore be used to estimate R values for soil use planning, management and conservation in São Paulo state.
soil conservation; water erosion; universal soil loss equation