On the skill of numerical weather prediction models to forecast atmospheric rivers over the central United States

Abstract Flooding over the central United States is responsible for large socioeconomic losses. Atmospheric rivers (ARs), narrow regions of intense moisture transport within the warm conveyor belt of extratropical cyclones, can give rise to high rainfall amounts leading to flooding. Short‐term forecasting of AR activity can provide basic information toward improving preparedness for these events. This study focuses on the verification of the skill of five numerical weather prediction models in forecasting AR activity over the central United States. We find that these models generally forecast AR occurrences well at short lead times, with location errors increasing from one to three decimal degrees as the lead time increases to about 1 week. The skill (both in terms of occurrence and location errors) decreases with increasing lead time. Overall, these models are not skillful in forecasting AR activity over the central United States beyond a lead time of about 7 days.