A dynamical downscaling methodology for use over Brazil has been developed, using the regional spectral model (RSM). The RSM was originally developed at NCEP (the National Centers for Environmental Prediction). The simulation area covers 85W to 25W and 10N to 40S, centered over the entire country of Brazil, with 50-km grid space increment. NCEP 28 vertical levels T62 spectral resolution (~200km grid space at the center of the simulation domain) reanalysis are used to initialize and force the regional model for the periods covering March 1, 1997 through May 31, 1999.

It was found that there is a drying trend in the model soil moisture after the initial 3 months of continuous integration. The drying trend is a response to the positive feedback from the dry bias of precipitation, due to the imperfect model parameterization. By utilizing the available observed precipitation, we will show that a soil moisture correction scheme can help improve the regional simulation. We will also discuss how well the soil moisture-corrected RSM simulation compares to available station precipitation observations, in terms of the annual cycle and the inter-annual variability.