Scientific evidence guides public policy for improving the management of soil and water resources. With stronger scientific evidence, more informed public policy will lead to desired outcomes. The studies described in this dissertation use the Water Erosion Prediction Project (WEPP) computer model to address two soil erosion modeling issues. First, a statewide analysis of soil erosion in Iowa resulting from different corn stover removal rates is modeled to produce maps of soil erosion risk under various management scenarios. The results indicate that no--till is an effective practice for soil erosion control on sloping soils when maximum amounts of corn stover are removed from the field. However, maintaining adequate levels of soil organic carbon may be more of a constraint for stover harvesting than soil erosion on flat soils. Modifications to the WEPP user interface are needed to simplify soil erosion modeling with corn stover removal and site specific conditions. The second soil erosion modeling issue addressed in this dissertation is uncertainty of soil erosion and sediment load delivery predictions. The paper in Chapter 3 of this dissertation demonstrates a novel stochastic approach for explicitly quantifying prediction uncertainty using WEPP. Uncertainty of sediment load predictions is explicitly calculated using Monte Carlos simulation with stochastic climate variable inputs. Scientists, environmentals and farmers commented in focus group interviews that the stochastic analysis results helped them to better understand the uncertainty of soil erosion and sediment delivery predictions used to design control measures.