HIGH-RESOLUTION STORM-SCALE ENSEMBLE FORECASTS OF THE.pdf

23.6 HIGH-RESOLUTION STORM-SCALE ENSEMBLE FORECASTS OF THE 28 MARCH 2000 FORT WORTH TORNADIC STORMS 1Nicki L. Levit,1,2Kelvin K. Droegemeier and 1Fanyou Kong 1Center for Analysis and Prediction of Storms and 2School of Meteorology University of Oklahoma Norman, Oklahoma 73019 1. INTRODUCTION Ensemble forecasting – or the creation of multiple, concurrently valid forecasts from slightly different initial conditions, from different models, or by using different options within the same model – has become the cornerstone of operational global numerical weather prediction (e.g., Kalnay 2003). Extensions to the regional scale have been underway for some time (e.g., Brooks et al. 1995; Hamill et al. 2000), though as for global models, mostly in a framework of hydrostatic dynamics and model grid spacings that do not explicitly resolve convective clouds. During the past decade, increasing emphasis has been placed on the explicit prediction of storm-scale weather, especially deep convection, with the WSR-88D Doppler radar serving as the foundational observing system from which fine-scale wind, thermodynamic, and moisture fields can be observed directly or retrieved (e.g., Droegemeier 1997). Although storm-resolving (grid spacings of 1 km or less) non-hydrostatic models show considerable promise for practicable NWP (e.g., Xue et al. 2003), probabilistic, rather than deterministic forecasts, likely will be required (e.g., Brooks et al. 1992; Hou et al. 2001; Elmore et al. 2002a,b, 2003). To begin exploring ensemble forecasting of deep convective storms in the context of full NWP (i.e., as opposed to a single-sounding cloud model), we apply herein the scaled lagged average ensemble forecasting (SLAF) technique (Ebisuzaki and Kalnay 1991) to the prediction of thunderstorms that occurred in north central Texas on 28 March 2000, and that produced a tornado in the Fort Worth metropolitan area (Xue et al. 2003). Our goals