Characteristics of short-term predictability of land-falling cyclones along the North American west coast

Land-falling North Pacific storms often impact the west coast of North America with strong winds, heavy precipitation and large mountain snowfall. These storms are frequently poorly predicted by operational models despite continued improvements in model resolution, model physics and data assimilation. Previous studies have shown that short-term (24-72 h) forecast errors of sea level pressure along the North American west coast are statistically related to large-scale upper-level flow regime. We expand on this prior work by using other methods to define predictability and relate the predictability to atmospheric flow patterns that range from mesoscale to large scale. Specifically, we examine 24-48 h forecasts of land-falling North American west coast cyclones over 2 winter seasons. The tools used to assess predictability of the cyclones are 1) ensemble sensitivity produced from a WRF-model ensemble Kalman filter (EnKF), 2) adjoint sensitivity produced with an MM5 adjoint model, and 3) spread growth in the cyclone sea-level pressure field within the same WRF-model EnKF. We use these tools to further distinguish why certain regimes may exhibit poor predictability, and whether large intrinsic potential for error growth or slower-growing, larger errors present at initialization are more important. The goal of this project is to determine some of the factors that contribute to cyclone predictability during a variety of weather regimes on multiple scales. Initial results over the 2-year period are shown, and plans to extend this work beyond this initial examination are discussed. Determining these relationships could be used to improve operational modeling/data assimilation systems during specific weather regimes associated with poor prediction of cyclones.