Robust projections of climate change impacts on regional hydrology are crucial for water resources management, especially in data sparse regions. Impact projections have to: (1) be available at gauged and ungauged locations, (2) consider changes in watershed behaviour for different future climates and land uses, (3) include estimates of uncertainty. We apply a novel water resources modelling framework that combines signature regionalization with trading space-for-time to obtain hydrologic projections in ungauged basins (PUB) and under climate change for the Olifants basin in South Africa, a UNESCO HELP basin. We find that using statistically downscaled GCM versus observed climate data leads to only slight deterioration in reliability of runoff projections for the historical period. Our framework projects a decrease in regional runoff by -8.7% (-3.9%) for A2 (B1) scenarios by the end of the century, with 80% (67%) of GCMs agreeing on the decrease in runoff.