Rapid Development of NASA SPoRT's Streamflow-AI using a R2O/O2R Paradigm

River flooding is a concern for emergency managers and other decision makers throughout the United States. Accurate medium range (> 2 days) forecasts of river flooding events are particularly challenging for National Weather Service (NWS) forecasters because routine streamflow forecasts provided by the River Forecast Centers may not incorporate quantitative precipitation forecast (QPF) data beyond one or two days. Because of this and in collaboration with NWS forecasters, the NASA Short-term Prediction Research and Transition Center (SPoRT) developed a deep learning model (Streamflow-AI) to help fill the operational gap, which produces routine 7 day river level forecasts using a suite of QPFs. The SPoRT Streamflow-AI product was initially tested on a small subset of river basins primarily located within the Tennessee Valley in late 2019 in collaboration with the Lower Mississippi River Forecast Center and four NWS Forecast Offices (Huntsville; Nashville, Morristown, and Sterling). After a successful initial test and validation of the product, the coverage has since expanded to nearly 200 locations across the eastern U.S. Throughout the project, research efforts have been tailored to meet the needs of the end users with additions including: lengthening the forecast time from 5 to 7 days, adding the National Blend of Model (NBM) QPF scenario, addition of a snow-aware model, and the creation of river level forecasts output that can be displayed directly in our collaborator’s decision support software. This presentation will provide an overview of NASA SPoRT’s Streamflow-AI product with a focus on the research-to-operations/operation-to-research (R2O/O2R) aspects of the project and will highlight ongoing research due to those efforts.