The study proposes a new mathematical method, referred to as spectral combination, to downscale Gravity Recovery And Climate Experiment (GRACE) observations. The goal is to improve the spatial resolution of GRACE from 1̊ to 0.25̊, based upon available hydrological variables. First, a new approach based upon condition adjustment is proposed to estimate uncertainties related to hydrological variables. Second, a spectral-spatial estimator is developed to derive downscaled Total Water Storage Anomalies (TWSA) by optimally combining GRACE models and hydrological variables. Last, groundwater storage anomalies (GWSA) are derived from the downscaled TWSA. The proposed spectral combination approach was tested over the Canadian Prairies by considering GRACE data and required Global Land Data Assimilation System (GLDAS) variables for February 2003 to December 2016. The results reveal greater details in TWSA after spatial downscaling. Quantitatively, retrieved downscaled GWSA were validated using 75 unconfined in situ piezometric wells that were distributed across the Province of Alberta. A correlation of 0.80, with an RMSE of 11 mm, was obtained after downscaling with all wells over the validation area. These results are better than those obtained before downscaling (correlation of 0.42, with an RMSE of 21.4 mm), demonstrating that the proposed approach is successful.