Ubiquitous increases in flood magnitude in the Columbia River Basin under climate change

The USA and Canada have entered negotiations to modernize the Columbia River Treaty, signed in 1961. Key priorities are balancing flood risk and hydropower production, and improving aquatic ecosystem function while incorporating projected effects of climate change. In support of the US effort, Chegwidden et al. (2017) developed a large-ensemble dataset of past and future daily streamflows at 396 sites throughout the Columbia River basin (CRB) and selected other watersheds in western Washington and Oregon, using state-of-the art climate and hydrologic models. In this study, we use that dataset to present new analyses of the effects of future climate change on flooding using water year maximum daily streamflows. For each simulation, flood statistics are estimated from generalized extreme value distributions fit to simulated water year maximum daily streamflows for 50-year windows of the past (1950-1999) and future (2050-2099) periods. Our results contrast with previous findings: we find that the vast majority of locations in the CRB are estimated to experience an increase in future streamflow magnitudes. The near ubiquity of increases is all the more remarkable in that our approach explores a larger set of methodological variation than previous studies; however, like previous studies, our modeling system was not calibrated to minimize error in maximum daily streamflow and may be affected by unquantifiable errors. We show that on the Columbia and Willamette rivers increases in streamflow magnitudes are smallest downstream and grow larger moving upstream. For the Snake River, however, the pattern is reversed, with increases in streamflow magnitudes growing larger moving downstream to the confluence with the Salmon River tributary and then abruptly dropping. We decompose the variation in results attributable to variability in climate and hydrologic factors across the ensemble, finding that climate contributes more variation in larger basins, while hydrology contributes more in smaller basins. Equally important for practical applications like flood control rule curves, the seasonal timing of flooding shifts dramatically on some rivers (e.g., on the Snake, 20th-century floods occur exclusively in late spring, but by the end of the 21st century some floods occur as early as December) and not at all on others (e.g., the Willamette River).