Forecasting stream temperature using data assimilation in support of water management decisions

Abstract:

The Delaware River Basin (DRB) is an ecologically diverse region and a societally important watershed along the East Coast of the United States. Managers of reservoir operations within the DRB need to balance supplying enough drinking water to over 13 million people while also timing the releases of cool water to maintain suitable cold-water habitat for economically important fisheries downstream. Accurate forecasts of water temperature are needed to aid managers making decisions about when and how much water to release. To this end, we have developed a modeling framework capable of assimilating water flow and temperature observations to forecast water temperature throughout the stream network in support of management decisions. This framework leverages extensive monitoring networks from multiple agencies as well as a process-based streamflow and temperature model, the coupled Precipitation-Runoff Modelling System and Stream Network Temperature model (PRMS-SNTemp). Our stream-temperature forecasting framework uses the ensemble Kalman filter to assimilate observations at hundreds of DRB locations into PRMS-SNTemp to predict streamflow and water temperature 10 days into the future driven by the National Oceanic and Atmospheric Administration Global Ensemble Forecast System. The ensemble Kalman filter iteratively incorporates information from the observations into our understanding of the system, updating model states and parameter estimates, which have been calibrated with older observations, as new observations are collected and assimilated. We evaluate model performance for predictive accuracy and ability to predict exceedances of thermal thresholds at select management-relevant locations within the DRB. These results quantify current model skill and will serve as benchmarks against which future modelling efforts can demonstrate advancement toward a valuable and novel stream temperature forecasting capability.