Landfalling Cyclone Forecast Sensitivity to Varying Data Assimilation Methods in a Mesoscale Model

Abstract

During PACJET 2001, an intense extratropical cyclone rapidly developed off the Southern California coast and produced substantial rainfall as it interacted with coastal topography. Operational models tended to misplace the position and underforecast the intensity of the storm, which was due in part to a lack of observations over the cyclogenesis region off the coast of Southern California. PACJET made dropsonde and other insitu observations in the area, which were available for data assimilation tests using the Navy's COAMPS model. Previous work has suggested that orographic rainfall predictions by a mesoscale model are sensitive to details in the specification of initial conditions. Nuss and Miller (2001) found significant differences in mesoscale precipitation for a landfalling front interacting with coastal topography that was rotated by 1 degree relative to the large scale wind direction. Their results suggest that in some situations the terrain forced precipitation can be sensitive to small differences in the synoptic-scale structure. In this study, the sensitivity of the cyclogenesis and subsequent orographic rainfall to the choice of data assimilation method is examined. The range of forecast errors and character of the forecast differences are examined to highlight crucial aspects in the initial state that must be faithfully represented by the assimilation system to accurately predict the cyclogenesis and orographic rainfall.