Calibrated thunderstorm probabilities over Australia derived from two 5-member lag ensembles

We coupled the calibrated thunder technique first described in Bright et al. (2005) to two separate 5-member lag ensembles at the Australian Bureau of Meteorology. The first ensemble is constructed from runs of the Australian Community Climate and Earth System Simulator (ACCESS), a modified version of the UK Met Office model on a 0.11o grid with parameterized convection. The second ensemble is based on a convection-allowing radar-assimilating 1.5 km version of the same model run hourly over a part of southeastern Australia.

We tested two storm-environmental predictors using the ensemble output; the cloud physics thunder parameter and total model precipitation. Reliable thunderstorm probabilities are then derived on a 40 km grid through calibration against cloud-to-ground lightning observations across Australia.

This study explores the performance of the calibrated thunder approach with respect to the time of day (nocturnal versus afternoon thunderstorms), predictor thresholds used, and the input ensemble employed. We will particularly focus on whether the convection allowing ensemble provides any advantages over the parameterized ensemble for the probabilistic prediction of CG-producing thunderstorms out to 12 hours.