May 23 - July 29
Ensemble Forecasts and Verification of the May 2015 Multi-Hazard Severe Weather Event in Oklahoma
Austin Coleman and Nusrat Yussouf
What is already known:
- The current prototype Warn-on-Forecast (WoF) system is a storm-scale ensemble at 3-km horizontal grid spacing that has the potential to extend probabilistic low level mesocyclone forecast lead times for severe convective events.
- The science and technology being developed to achieve WoF goals can also be used to improve 0-3 hour extreme rainfall forecasts for convective systems.
- The 3-km horizontal grid spacing of the current prototype WoF system is too coarse to resolve tornadic circulations.
What this study adds:
- The 0-3 hour heavy rainfall forecasts from the prototype system verifies better with NCEP’s Stage IV analyses and Mesonet observations in terms of locations compared to that from the operational HRRR forecasts for this case study.
- The ensemble forecasts systematically underestimate precipitation amounts, which indicates further investigation of microphysics scheme sensitivities as well as grid-spacing sensitivities are needed.
- Ensemble forecasts at 1-km horizontal grid spacing from the downscaled 3-km prototype system introduces many spurious cells with embedded spurious mesocyclones.
Abstract:
Dual threat severe weather events in which both tornadoes and flash floods affect the same area within a short time frame pose a complex problem since the life-saving actions for these two events are contradictory. One such event is the 6-7 May 2015 tornado and flash flood event over Oklahoma. This study explores the capability of a rapidly-updating 3-km horizontal grid spacing convective-scale ensemble data assimilation and prediction system developed as part of the Warn-on-Forecast initiative to forecast features of this dual threat severe weather event. Results indicate that the 0-1 h probabilistic forecasts of reflectivity verify reasonably well with the observations. However, beyond the 1 hour forecast period, the forecast accuracy is degraded, including biases in storm motion as well as spurious cell generation. The ensemble probability matched mean quantitative precipitation forecasts capture the placement of most intense areas of precipitation very well, but underestimate the amount of accumulated precipitation. These quantitative precipitation forecasts are found to outperform the deterministic quantitative precipitations forecasts of the operational High-Resolution Rapid Refresh model as well. Additional ensemble forecast experiments from simple downscaling to 1-km grid spacing from the 3-km ensemble do not significantly reduce the storm motion bias found in the original results and introduce more spurious cells.