Many environmental models provide different results when applied at different scales. A methodology for detecting and overcoming these scaling problems is developed and applied to a global-scale soil erosion model. The method involves applying a sensitivity analysis of the soil erosion model to scale in order to determine the causes of the scaling problems. To overcome these problems, multiple scale data sets of model parameters are used to develop statistical and fractal techniques that explain the spatial variability of model parameters at different scales. These scaling methods allow us to employ scale explicitly as a parameter in the model equations. When soil erosion is calculated at the global scale using these techniques and validated using available data, there is generally a good agreement.