Global Predictability and Prediction Skill of Atmospheric Rivers

Atmospheric rivers (ARs) are responsible for the majority of horizontal water vapor transport in the midlatitudes and can intensify downstream precipitation and influence flooding, snowpack and water availability. Consequently, there is significant incentive to estimate the predictability and quantify our prediction skill of AR events in operational forecast models, especially weeks to months in advance. Understanding and exploiting the full extent of AR predictability is vital for watershed and hazard preparation and water resource management in areas that are sensitive to heavy precipitation events often associated with ARs. Though the predictability limits and prediction skill of AR-related quantities such as precipitation and integrated vapor transport have recently been quantified for very limited regional areas, a systematic assessment of AR events themselves (with explicit consideration of AR geometries/intensities) using contemporary operational forecast models on a global scale has not yet been made. In this study, we create new objective skill metrics for AR events and quantify predictability limits and prediction skill of ARs in two decades of AR hindcasts from several operational models at lead times ranging from 1 day to 1 month. Our methodology is performed globally, but also allows for regional investigation of observed AR events of interest. We also examine the sensitivity of these skill estimates to large scale modes of climate variability (e.g. the El Niño-Southern Oscillation) in an effort to leverage higher skill conditioned upon particular phases of these modes. This study contributes to the overarching goals of the international WMO Subseasonal to Seasonal Project by utilizing operational forecast and hindcast experiments with an emphasis on subseasonal forecasting applications.