As part of a strategy to test a nested modeling system for South America, a series of sensitivity studies are performed which examine (1) the effects of remote SST forcing during two extreme rainfall seasons, (2) effects of local forcing (e.g., the Amazon forest) on regional circulation and rainfall, and (3) effects of domain size and location of boundaries on model performance, sensitivities and internal variability of the nested system. The models employed are the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM3) and the Regional Climate Model (RegCM2). Sensitivity results from the regional model driven by NCEP/NCAR Reanalyses boundary forcing will be compared with results from the same model driven by an ensemble of the GCM realizations. The limitations and potential of this dynamical downscaling method will be discussed.