Assessing the Skill of Convection-Allowing Ensemble Forecasts of Severe MCS Winds from the SSEO

Convection-allowing models (CAMs) have shown utility in forecasting the general location, mode, and evolution of thunderstorms. Significant questions remain, though, in the ability to diagnose the explicit threat of individual severe hazards (tornadoes, large hail, and damaging winds) from CAMs. Model diagnostics, such as updraft helicity and various hail proxies, have been developed to assess the potential for tornadoes and large hail. Little work, however, has been done in developing techniques to diagnose severe winds from mesoscale convective systems (MCSs). It is hypothesized that a combination of forecast radar reflectivity and 10-meter wind fields should provide an indication of the threat of severe winds from MCSs. Verification of CAM forecasts of 10 meter winds associated with thunderstorms is therefore crucial, but is heavily complicated by the quality and coverage of observations available for verification, requiring the use of innovative techniques.

This study examines forecasts from 2012-2014 from the Storm Prediction Center Storm Scale Ensemble of Opportunity (SSEO) of 24-hour neighborhood probability of 10-meter wind speed greater than 30 knots. Forecasts are verified with practically perfect hindcasts generated from wind reports. The forecast and practically perfect fields are filtered using forecast and observed radar reflectivity, respectively, to ensure that they are associated with an organized MCS. Verification is performed using the Method for Object-Based Diagnostic Evaluation (MODE), treating both the forecast and observed wind probability swaths as objects, allowing for forecast assessment through their comparison. Overall performance of the SSEO will be assessed, as well as the relative importance of the individual (WRF-ARW, WRF-NMM, and NMMB) and time-lagged members through a step-wise removal of individual members from the ensemble.