Further Advancement of the GSI Based Hybrid EnVar System for Convective Scale Radar Data Assimilation in HRRR and WoF Contexts: Development of Static Covariance for Direct Assimilation of Radar Reflectivity Observations

The impact of direct radar reflectivity assimilation in pure ensemble-based data assimilation (DA) method (e.g., EnKF, EnVar) may be limited due to the low spread of ensemble perturbations for hydrometeor variables. For example, the initial ensemble downscaled from mesoscale analyses may have small or zero variance of hydrometeors where in reality it is precipitating. Such issue may lead to a long spin up time during the data assimilation cycles. Johnson et al. (2015) suggested that static covariance, if constructed properly for convective scale radar data assimilation, can quickly and efficiently add reflectivity that is completely absent from the first guess field. Although static covariance has been widely used for large scales. How to form it for convective scale is still an unknown. Wang and Wang (2017) developed a method to directly assimilate radar reflectivity observations in GSI EnVar where pure ensemble covariance (i.e. no static covariance) is used. In this study, we further developed the static covariance appropriate for convective scale data assimilation. This extended static covariance together with the ensemble covariance form a hybrid DA system that is robust for convective scale radar data assimilation. In other words, this hybrid DA system now fully takes advantage of both covariances for storm scale data assimilation.

To further develop the static Bwithin the GSI based hybrid DA system for direct assimilation of radar reflectivity, the static Bis extensively developed to include the physically reasonable cross-variable correlations. The impact of the newly extended static Bis firstly examined in the WoF context for the analysis and prediction of the 8 May 2003 Oklahoma City tornadic supercell storm. Our results suggest that incorporating the new staticBcan greatly reduce the spin up time during the DA cycles. In the subsequent forecast period, the inclusion of the newly constructed static Bimproves the reflectivity forecast of the tornadic supercell. Experiments are also ongoing to test the new radar data assimilation approach in HRRR context. Detailed results will be presented in the conference.