Process-based future hydrological predictions are often based on the simulations of distributed hydrology models that are driven with the statistically disaggregated hydrometeorological fields from Global Climate Model (GCM) integration. More sophisticated methodologies include dynamic downscaling of GCM data through a regional climate model (RCM) before its statistical disaggregation and use in hydrological simulations. While RCM-based downscaling improves the spatial variability of hydrometeorological fields through better representation of fine-scale processes, it is not well known that how much an effect spatial variability of data has on statistical disaggregation and hydrological model simulation. In this study, we use two sets of Variable Infiltration Capacity (VIC) model ensemble simulations, one based on the daily data from NCAR Community Climate System Model (CCSM) ensemble and one based on the daily data from downscaled CCSM ensemble through Abdus Salam International Centre for Theoretical Physics Regional Climate Model (RegCM), to investigate the effect of spatial variability of climate model data in hydrological predictions. The VIC model ensemble simulations cover 1960-1999 in the baseline period and 2000-2039 in the future period and cover the continental United States at 1/8th degree horizontal grid spacing. In the baseline period, we compare the results from two model-based VIC ensembles with an observation-based VIC simulation, as well as with the snow course and USGS observations. In particular, we focus on the hydrological processes in the Western United States where topographic complexity influences both cold and warm season hydrological processes. Further, we investigate how spatial resolution of driving hydrometeorological fields effects the distribution of hydrological extremes, and their role in the simulated hydrological response in the future period.