Adaptive Radial Velocity Assimilation in the Warn-on-Forecast System

The Warn-on-Forecast (WoF) system has provided experimental real-time short-term convective-scale forecasts over the last few years. This ensemble-based system ultimately delivers probabilistic forecasts of severe convective hazard potential including tornadoes, hail, and damaging winds. Forecast initial conditions are created via frequent assimilation of radar (reflectivity and radial velocity), satellite, and in-situ observations. From 2016-2018, 5-km radial velocity Cressman analyses were created prior to assimilation to thin the observation number. However, such analyses reduce observation magnitude and may fail to depict important storm-scale features and their intensities (e.g., mesocyclones). This study retrospectively assimilates denser, 3-km radial velocity analyses in lieu of the 5-km analyses for eight case studies during the spring of 2018. Although there are improvements in early forecasts, these initial eight cases suggest simply altering the analyses to 3-km negatively impacts later forecasts since noise initiates spurious convection and even creates model imbalance. Therefore, the 3-km analyses are thinned using an updraft variance mask. This limits radial velocity assimilation to areas where the observations are more likely to be impactful while reducing spurious convection. Masking the 3-km analyses can also reduce the number of observations assimilated (compared to the real-time 5-km analyses) which is computationally advantageous.