This research presents a new multivariate statistical approach to downscale hydroclimatic variables associated with global climate signals, from low-resolution Global Climate Models (GCMs) to high-resolution grids that are appropriate for regional and local hydrologic analysis. The approach uses Principal Component Analysis (PCA) and Multichannel Singular Spectrum Analysis (MSSA) to: (1) evaluate significant variation modes among global climate signals and spatially distributed hydroclimatic variables within certain spatial domain; (2) downscale the GCMs' projections of the hydroclimatic variables using these significant modes of variation and (3) extend the results to other correlated variables in the space domain. The approach is applied to the Colorado River Basin to determine common oscillations among observed precipitation and temperature patterns in the basin and the global climate signals El Niño Southern Oscillation (ENSO) and Pacific Decadal Oscillation (PDO). These common oscillations serve as a basis to perform the downscaling of ENSO-related precipitation and temperature projections from GCMs, using a new gap-filling algorithm based on MSSA. The analysis of spatial and temporal correlations between observed precipitation, temperature and vegetation activity (represented by the Normalized Difference Vegetation Index, NDVI) is used to extend the downscaling of precipitation to vegetation responses in ten ecoregions within the basin. Results show significant common oscillations of five and 15-year between ENSO, PDO and annual precipitation in the basin, with wetter years during common ENSO and PDO positive phases and dryer years during common negative phases. Precipitation also shows an increase in variability in the last 20 years of record. Highly correlated responses between seasonally detrended NDVI and precipitation were also identified in each ecoregion, with distinctive delays in vegetation response ranging from one month in the southern deserts (in the fringe of the monsoon precipitation regime), to two months in the mid latitudes and three months to the north, affected by seasonal precipitation. These results were used to downscale precipitation and temperature from two GCMs that perform well in the basin and have a distinctive ENSO-like signal (MPI-ECHAM5 and UKMO-HADCM3) and to extend the downscaling to estimate vegetation responses based on their significant correlations with precipitation.