We perform a series of experiments to test different strategies for downscaling global scale information from a general circulation model (GCM), from the NASA Goddard Institute for Space Studies (GISS), using a regional mesoscale model (Regional Atmospheric Modeling System - RAMS) for the Mid-Atlantic states of the U.S. First we compare the performance of RAMS driven by past atmospheric data (NCEP reanalysis) to observations and its performance when driven by GISS GCM data. Then, the sensitivity to different nesting techniques is analyzed. In particular, we compare conventional Davies nudging to spectral nudging, in which the longwaves of the solution are nudged to those of the driving data throughout the domain. The potential benefits of the different techniques are described for both the case when RAMS is driven by real atmospheric solutions (NCEP reanalysis) and for the case when it is driven by GCM data.