A Testbed To Develop, Apply, and Evaluate Advances In In Process-Based Streamflow Prediction For Water Management In The Columbia River Basin

Water management depends on short range to seasonal and interannual hydroclimate predictions, and it has been a long-standing scientific and technical challenge to advance their skill. The research community has invested substantially in developing better predictions through improved land process modeling and subseasonal to seasonal climate forecasting, yet few of these advances have been adopted by the water management world. USACE partnered with the National Center for Atmospheric Research to implement a system for short-range to seasonal hydrologic forecasting for the Columbia River Basin (CRB) based on watershed modeling and ensemble streamflow prediction (ESP) methods. A testbed-like platform for introducing supporting methods that have been shown to increase the reliability and skill of hydrologic analyses and forecasts, and also aid in the development of actionable over-the-loop (more automated) hydrological predictions, has been developed. This system produces real-time ensemble forecasts for both seasonal and short-range (0-30 day) lead times. Multiple techniques including post-processing, climate conditioning, and data assimilation are applied and evaluated with respect to predictive skill. The potential strengths and weaknesses of more recent techniques are identified through comparisons of hindcasts and real-time forecasts of other operational methods that take the form of statistical models, conceptual hydrological models, and machine learning applications. The goal of this work is to identify actionable forecasting methods that could enhance predictability of streamflow, to support in-season decision making as water managers face recent increases in hydrological variability and uncertainty.