The good performance of hydrological models depend on the reliability and availability of precipitation data. Thus, if the satellite precipitation data are critically evaluated, you can get better accuracy in predicting the flooding of large basins with a sparse network of measures of precipitation. For that was developed by Hossain and Anagnastou (2006) model called multidimensional stochastic error SREM2D (A Two Dimensional Satellite Rainfall Error Model), which was used as study methodology in the 2008-2010 period, simulating the propagation of errors for models: HYDROE, GSMAP, 3B42RT CMORPH on Tocantins River basin, for the period 2008 to 2010. Thus, this paper studies the propagation of errors of estimate of the precipitation by satellite to assess the behavior of these uncertainties in hydrological models. Results preliminary show that the model SREM2D has the potential to create realistic ensemble precipitation with greater accuracy. This study has followed up on hydrologic modeling of the MGB-IPH model (Model of Large Basin to assess the uncertainty of rainfall-runoff.