Performance and Reliability of the NOAA National Water Model Operational Forecasts for Water Resources Management

The NOAA Office of Water Prediction launched a new operational hydrologic prediction service based on the National Water Model (NMW) in the continental United States. Skillful hydrological forecasts at various lead times would be beneficial for decision-making in water resources management, especially in watersheds that play a role in providing water resources to lakes and reservoirs. However, for stakeholders to make the best use of the NWM, they need to know something about the overall skill of the forecasts in their watersheds of concern. The purpose of this study is to examine quantitative predictability of the NWM operational streamflow forecasts and to answer the following overarching questions: what is the forecast reliability depending on lead time? and what are the strengths and weaknesses of the forecasts compared to benchmark forecasts (persistence and climatology)? 9 basins in the western U.S. that flow into lakes and reservoirs are selected as test application areas: Taylor Park reservoir (CO), El Vado reservoir (NM), Rio Grande river (CO), Roosevelt reservoir (AZ), Lake Tahoe (CA and NV), Lake Sherburne reservoir (MY), and Buffalo Bill reservoir (WY). A wet season from September to March in 2017-2018 year is used to demonstrate the analyses. This study focuses on the short-range (out to 18 hours) and medium-range (out to 10 days) streamflow forecasts. In general, the NWM performed well in terms of providing streamflow forecasts that show similar trends with observed hydrographs. As expected, the NWM showed better performance than persistence forecasts at shorter lead times, but it showed a decreasing trend in skill at longer lead times. Results also showed that the reliability of the NWM forecasts varied with drainage size and terrain complexity.