Development and Use of Ensemble Hydroclimate Projections in the Columbia Basin: A Practitioner’s Perspective

The Federal Columbia River Power System (FCRPS) River Management Joint Operating Committee (RMJOC) has closely collaborated with the University of Washington and Oregon State University to update climate change temperature, precipitation, and streamflow projections for the Columbia Basin. The new projections incorporate two main enhancements: using the latest temperature and precipitation information from the Fifth Coupled Model Intercomparison Project (CMIP-5), and incorporating several advances in hydroclimate modeling since a previous University of Washington and RMJOC study in 2009-2011. In addition to employing an objective method for selecting the best-performing climate models in the Columbia Basin, the streamflow projections explicitly account for uncertainties arising from methodological modeling choices, including temperature and precipitation downscaling, hydrologic model characteristics, hydrologic model calibration uncertainties, and to some extent the uncertainties in regional historical temperature, precipitation and streamflow datasets. This comprehensive effort has yielded 172 different hydroclimate scenarios which better capture the range of possible hydroclimate outcomes, and in turn will assist regional planners in better understanding the range of possible impacts to their long range planning efforts.

The emerging results generally confirm results from previous regional studies. Temperatures have already warmed about 1.5°F in the Columbia Basin since the 1970s climatological period, and are expected to warm another 1 to 4°F by the 2030s, with warming likely to be greatest in the interior compared to more coastal regions. Future precipitation trends are more uncertain, but a general upward trend is more likely than not, particularly in the winter months, while already dry summer months could become drier. Average winter snowpacks are very likely to decline over time as more winter precipitation falls as rain instead of snow, especially on the US side of the Columbia Basin. By the 2030s, higher average winter flows, earlier peak spring runoff, and lower average summer flows are very likely. The greatest streamflow changes are likely to be in the Snake River Basin, although that is also the basin with the greatest modeling and forecast uncertainly.

While the updated projections will be useful in informing follow-up modeling work and policy decisions, one of several unintended benefits from this project has come from the close collaboration and engagement between university principal investigators, hydropower planning practitioners, regional stakeholders and outside subject-matter experts. Frequent interactions through workshops, quarterly updated through the Columbia River Forecasting Group, monthly check-ins and many informal conversations throughout the project, has allowed technical staff and stakeholders to gain a firm understanding of the complex hydroclimate modeling process as a whole, manage expectations and become better educated on the appropriate use for this new data in planning studies. RMJOC technical staff also served as climate projection and dataset peer reviewers for the principal investigators, which gave the research team in-depth quality control and insight that is not typically available when large hydroclimate datasets are produced. Meanwhile, principal investigators and subject matter experts gained firm insights as to how the results will be used by technical planners and gained a better understanding of where the region’s changing climate could have a more significant impact on system operations than others. These insights helped the principal investigators to make a dataset that was useful for subsequent operational studies while maintaining scientific rigor. Principal investigators also were able to offer additional peer review and support during the modeling process.. This collaborative approach enhanced communication between researchers, stakeholders and policy makers. The process created not only a better dataset, but also a better understanding of the hydroclimate modeling process itself, and the impacts these datasets could have for other planning processes.

As one of several examples, the interactive learning between scientists and stakeholders contributed to the next step in the modeling chain: using the datasets in a scientifically sound manner for hydropower scenario planning. Because hydropower planning is a time-intensive and complex process on its own, the RMJOC technical team had to downselect a subset of the new datasets so that meaningful hydropower modeling results can be available to inform upcoming, long range regional planning decisions in a timely manner. Because the technical team acquired a firm understanding of the hydroclimate modeling chain and where the greater uncertainties lie within the dataset, they were able to employ a mix of objective and subjective techniques to make downselecting decisions which remain faithful to the original intent of the research, and obtain scientifically-focused peer review from both external stakeholders and subject matter experts to ensure GCM, downscaling, and hydrologic model uncertainties were fully captured in the scenario planning process.