Diurnal Variations in Forecasting Skill of UH-Derived Surrogate Severe Reports

Recent increases in computational resources have allowed for the broad use of real-time numerical weather prediction at resolutions that allow for explicit forecasts of convection as opposed to more traditional convection-parameterizing NWP simulations . These convection-allowing models (CAMs) are frequently used in ensemble systems (CAMEs) which provide a forecast mean and spread, analyses that are useful in creating probabilistic forecasts of convection. In forecasting severe convection using CAMEs, updraft helicity (UH) is frequently used as a "surrogate" for severe reports (SSRs), indicating in these forecasts where severe convection, and thus real severe reports, might be most likely. While many recent studies have found UH as a surrogate for severe reports to be quite skillful, few investigations have considered how this forecasting skill changes throughout the diurnal cycle. Nocturnal convection, including nocturnal supercells and overnight mesoscale convective systems, poses a significant threat to society, and its mechanisms of formation and maintenance are frequently different than those of daytime severe storms. Hence, it is reasonable to assume that CAME forecasting skill might vary over the course of a 36- or 48-hr forecast. To investigate this possibility, the current study will evaluate the diurnal variation in forecasting skill of all NSSL-WRF ensemble forecasts from 2016 (approximately 200 36-hr forecasts) using various thresholds of UH as the SSR parameter. Forecasting skills at various UH-field smoothing lengths, in addition to the variation in skill across forecasts of different categories of severe reports, will be evaluated. This assessment will utilize various traditional metrics, including contingency table statistics (bias, false alarm ratio, etc.) and their derivatives (receiver operating characteristic curves and areas), along with direct measures of skill such as the Brier and Factor skill scores.