Exploring the predictability of mesoscale cyclogenesis using ensemble data assimilation

The proliferation of high-resolution numerical weather prediction over the last decade has been widely seen as a great achievement for forecasting mesoscale phenomena. However, many aspects regarding the predictability of the mesoscale are still unknown. In this presentation, initial-condition sensitivities of short-term mesoscale forecasts of cyclogenesis and related snowstorms over the Pacific Northwest of the United States are explored with a 100-member ensemble generated with an ensemble Kalman filter. Two examples will be presented from December 2008 highlighting the role of initial condition uncertainties in short term forecasts.

Snowstorms in the Puget Sound region of Washington State are often characterized by mesoscale cyclogenesis off the Pacific Northwest Coast and 850 hPa temperatures less than -4 C. For forecast lead times between 24 and 36 hours substantial variability develops in both the 850 hPa temperature and the track of the developing cyclone. The ensemble members are ranked according to the predicted 850 hPa temperature and subsets of the 17 warmest and 17 coldest members are composited. In both cases considered the difference between the mean 850-hPa temperature for the warm and cold subsets grows larger than 6 C during the 36-hr forecast. Furthermore, the location of the mesocyclone differs by nearly 400 km between the warm and cold subsets. Such large error growth within the ensemble suggests that the predictive time scale for a deterministic forecast of these mesoscale cyclogenesis events can be less than 36 hours. Nevertheless, in both cases the ensemble mean forecast is remarkably consistent at increasing lead times highlighting one benefit of probabilistic mesoscale data assimilation and forecasting.