Sustained Wind Forecasts from the High-Resolution Rapid Refresh Model: Skill Assessment and Bias Mitigation

We examine the skill associated with sustained wind forecasts in the High-Resolution Rapid Refresh (HRRR) model, extending and enhancing previous work. Some utilities use numerical weather prediction models like the HRRR to anticipate electrical transmission line shutdowns for public safety reasons, increasing the importance of forecast accuracy and motivating the need to understand sources of bias and differences among observation networks. We demonstrate that the HRRR forecasts for airport stations are very good albeit with a tendency to underpredict the highest wind speeds and at the windiest locations. Forecasts for non-airport networks are much less accurate owing to a variety of factors, including differences in the way winds are measured and the environments they are measured in, and this results in predictions with excessive temporal variation relative to observations. We demonstrate a practical approach to modifying sustained wind forecasts so that they are more useful proxies for conditions being observed in the field.