Surface Runoff and Streamflow Projections from Earth System Models.

Projections of surface hydrology (specifically, runoff and streamflow) at decadal to century scales over stakeholder-relevant spatiotemporal resolutions remain a large gap in earth system models (ESMs). Cloud physics, land surface, and land-atmosphere parameterizations contribute to uncertainties in precipitation, runoff generation, groundwater storage, and evapotranspiration. Uncertainties result from variability in forcing, model response, or internal variability, along with gaps in understanding or mathematical representations of key processes. Here, we examine a set of mutually supporting hypotheses regarding surface runoff and streamflow ESM-projections, based on historical skills (e.g., by comparing with observed data such as from GRACE satellites), model-ensemble consensus, and hydrologic consistency. Our first hypothesis is that the latest generation (CMIP6) of ESMs have improved over the previous generation (CMIP5) across all geographies and time horizons. Our second hypothesis (a special case) is that the modeling groups that incorporated new and potentially improved cloud parameterizations in CMIP6, witnessed statistically significant improvements over CMIP5. Our third hypothesis is that internal variability and model response are comparable at regional scales especially for the near term, but the variability owing to forcing dominates, especially at mid- to late-century time horizons. The overarching hypothesis is that despite gaps in process understanding and ESM representations, as well as intrinsic variability, the projected changes in runoff and streamflow at regional and seasonal scales are significant and credible enough to require re-evaluation of design curves, planning scenarios, operations practices, and technical standards, while the variability and uncertainties can be adequately captured to inform time-phased and flexible adaptation strategies across the food-energy-water and the infrastructures-lifelines sectors.