Conditional Probabilities of Significant Tornadoes from RUC-2 Forecasts

Abstract

Several previous studies have established statistical relationships between the severity of convection and environmental conditions determined from rawinsonde observations. Here, the authors seek 1) to determine whether similar relationships are observed when severe weather reports are associated with gridded short-term numerical forecasts, and 2) to develop and demonstrate a prototypal probabilistic model to forecast the likelihood a thunderstorm will be tornadic. Severe weather reports and lightning network data from 1 January 1999 through 30 June 1999 were used to classify the weather at a set of Rapid Update Cycle (RUC-2) grid points into four weather categories. These were no thunderstorms, nonsupercellular thunderstorms, supercellular thunderstorms without significant tornadoes, and thunderstorms with significant tornadoes (F2 or greater). RUC-2 forecast convective available potential energy (CAPE), helicity, and 0?4-km mean wind shear from the same period were associated with this gridded classification of the weather. In general similar relationships were found between environmental parameters and storm categorization as others have previously documented. The Bayesian probabilistic model used here forecasts the likelihood that a thunderstorm will produce a strong or violent tornado, given a certain value of CAPE and helicity (or CAPE and wind shear). For two selected cases when significant tornadoes occurred, this model reasonably located the high-threat areas many hours in advance of the severe weather. An enhanced version of this prototypal tool may be of use to operational severe weather forecasters.