Estimates of future flow, including extremes, of the Columbia River headwaters

Streamflow projections, including extremes, for the 2050s for the Columbia River headwaters above Donald are obtained by downscaling four regional climate models of the North American Regional Climate Change Assessment Program (NARCCAP) suite and subsequent driving of a hydrologic model. We employ the entire model chain from global and regional climate models, station‐based statistical downscaling, and a fully distributed, physically based hydrologic model and verify the results against observed streamflow. The performance is model dependent but is generally encouraging enough to justify the application of the climate scenarios. A general warming of about 2°C is projected and, on average, slightly drier conditions, especially in late summer. We find evidence that the projected changes are elevation dependent and relatively small scale, with decreasing signals with higher elevations. All models project a shift of the hydrograph toward a more rain‐fed regime, with peak flows occurring in June instead of July. Annual peak flow is projected to not increase, and August low flow decreases in all four models. With nonshrinking (static) glaciers, relatively high melting rates are simulated for August and September that partly compensate for the shifted hydrograph; this enhanced glacier melt is also detected in simulated historic Columbia headwater flow. The static approximation is supported by a heuristic seasonal sensitivity analysis that suggests a moderate average areal glacier recession of about 10% for the midcentury. We discuss the need for a dynamic glacier component for a refined assessment of future drought risk.