This study aimed to present a risk analysis of the effects of different land use scenarios on reducing soil and water losses, using a physically based hydrological and soil erosion model. We quantified the effect of various land use scenarios on the expected rate of discharge and sediment loss during a single rain event in a small agricultural watershed on the Loess Plateau in China, using the geometric mean and stochastic distributions of measured field saturated conductivity (Ks) values. Land use scenarios were based upon physical, economic and agricultural interests, and effects on farmers' incomes were evaluated using empirically derived equations. A physically based hydrological and soil erosion model was used to quantify the effects of land use on discharge and soil loss. Using geometric mean values of Ks as the model input resulted in higher values for runoff coefficients and total soil loss compared with the use of stochastic Ks values. The use of stochastic Ks-distributions resulted in a range of model outcomes reflecting the effect of spatial heterogeneity on simulated discharge and soil loss. The conservation-driven scenario was most effective in reducing water and sediment losses by runoff and erosion, followed by the soil-driven scenario and the agriculture-driven land use scenario. Only the agriculture-driven scenario resulted in a small increase in household income, while a serious loss of income is predicted for the other scenarios. The use of variability of parameters and a Monte Carlo analyses allows statistical analyses and comparison of computed results for alternative land use scenarios, and leads to a more balanced judgement.