The use of a multidimensional satellite rainfall error models are very important to improve the satellite rainfall estimation products as well as hydrological applications. In this context, this paper investigates the error propagation of some satellite rainfall estimates algorithms and evaluates the impact of these errors on streamflow estimations through hydrological models. To evaluate the sensitivity of the satellite rainfall estimation was used a multidimensional error stochastic model named SREM2D (A Two-Dimensional Satellite Rainfall Error). This model was calibrated by nine errors metrics that serves as input to the SREM2D. The ensemble generated by the SREM2D model was used as input data for Large Basin Model of the National Institute for Space Research (MGB-INPE) developed for Brazilian basins. This methodology was applied to the Tocantins basin located in the eastern Amazon. Preliminary results show a good performance of the proposed stochastic model to generate ensembles from satellite rainfall fields. The streamflow estimated through the MGB-INPE is been evaluated in this context.