Incorporation of NASA/SPoRT, 7-day Streamflow-AI River Levels to Complement the Physical Hydrologic Model Forecasts in Operations

Ensemble forecasts via Artificial Intelligence (AI) models allow forecasters rapid updates to see a range of possible outcomes as a complement to operational, physical modeling systems. When used in conjunction with physical, hydrologic forecasts from National Weather Service (NWS) River Forecast Centers (RFCs), the gauge height forecasts from NASA’s Short-term Prediction Research and Transition (SPoRT) machine learning model (i.e. Streamflow-AI) adds valuable flooding assessment information in the extended period of days 3-7. The SPoRT Streamflow-AI product uses a Long Short-Term Memory (LSTM) network which is trained to predict river stage height at a 6-hr temporal resolution out to 7-days based on time-lagged inputs of gauge height, relative soil moisture, and forecasted precipitation (i.e. QPF). Currently, the Streamflow-AI provides three forecast scenarios based on different QPFs from the NWS Weather Prediction Center, the Global Forecast System deterministic model and the National Blend of Models. With these QPF amounts from different forecast models, you have an AI-based ensemble approach to hydrologic forecasting that is generated 4x/day. This presentation will focus on the NWS Weather Forecast Office experience integrating the NASA SPoRT Streamflow-AI product into operations, the SPoRT Streamflow-AI value beyond the normal 0-48 hr QPF forcing used within RFC forecasts, and its use for Decision Support Services (DSS) and flood operations.