Reliable estimates of the uncertainty associated with projections of future Colorado River flow need to draw on robust sampling of large ensemble climate model simulations as well as credible runoff sensitivities from high resolution process-based studies. Computational limitations still impede modeling studies that seamlessly integrate across this wide range of spatial and temporal scales. As a more practical approach, we revisit previous studies with water balance models for the Upper Colorado River Basin (UCRB) and its main reservoirs, applying an exhaustive set of climate model large ensembles and the latest sensitivity estimates. We show-case a new set of validation metrics that aim to scrutinize the models’ sensitivity biases rather than their mean state and flux biases, as this is shown to be more predictive of model behavior under climate change. We then discuss a range of new observational constraints on both the physical and socio-economic sources of uncertainty in regional climate projections and how they can help reduce uncertainties in UCRB projections.