The uncertainty in streamflow simulations and soil water estimates associated with satellite rainfall forcing is investigated for the Upper Des Moines River basin in the midwestern United States. Synthetic series of satellite rainfall estimates were produced with a rain gauge-satellite stochastic model and 10 years of daily rain gauge data (1979--1988) for three basins with drainage areas ranging from 2,000 km2 to 14,000 km2. The synthetic satellite rainfall series was based on observed satellite visible and infrared data which provided estimates of satellite rainfall for 180 randomly selected days in the period 1980--1987. Streamflow and soil water estimates were obtained with a rainfall-runoff-routing model (3R), based on soil water balance and accounting for snowmelt and frozen ground effects. Sensitivity of flow prediction with respect to rainfall was examined for three different conditions: (1) 3R calibrated and forced with rain gauge data, (2) 3R calibrated with rain gauge data and forced with satellite rainfall, and (3) 3R calibrated and forced with satellite rainfall. The most important results regarding the effect of satellite rainfall on flow simulation and soil water estimation for climate studies are as follows: (1) Flow simulation accuracy is sensitive to the basin scale, yielding higher correlation of simulated with observed streamflow for larger scales, (2) the hydrologic model forced with satellite data possesses skill during the period May--July for the midwestern United States, (3) derived upper soil water estimates are similar to the ones obtained using rain gauge forcing, and derived lower soil water estimates are lower than those obtained from rain gauge forcing. ¿ American Geophysical Union 1996