Evaluating the 3D EnKF - VAR Hybrid Data Assimilation in GSI for Surface and Upper Level Analyses

This study examines the 3 dimensional analysis produced using the Hybrid Ensemble Kalman Filter (EnKF) Variational (VAR) Data Assimilation in the Gridpoint Statistical Interpolation (GSI) System. The data assimilation ingests the 1 hour forecast High-Resolution Rapid Refresh (HRRR) and The Global Ensemble Forecast System, as the background and ensemble member set, respectively. Also, the conventional and satellite radiance observations are assimilated. The analysis covers a CONUS domain and has a 3 km horizontal resolution with 50 vertical native levels. The experiments focus on the advantages of using the flow dependentbackground error in the hybrid scheme to dynamically characterize the model background error based on the flow of the day. From the case study results, the hybrid scheme has a higher accuracy in 2m temperature and 10m winds speed than the background and the 3D VAR scheme, especially in regions of weather systems such as frontal boundaries and low pressure centers. Statistical comparisons of the surface analysis indicated the hybrid scheme outperformed the background and 3D VAR, but is unable to surpass the results from theReal TimeMesoscale Analysis (RTMA). Also, the impact of the flow dependentbackground error covariance in the hybrid scheme was compared with the terrain following background error covariance in the RTMA. Upper-level analysis comparison suggests the hybrid has a lower RMSE than the background and the 3D VAR for the lower and mid atmosphere but have similar results for the upper atmosphere. A brief sensitivity test on the vertical localization showed little impact on the upper-level analysis. Lastly, the benefit of assimilating satellite radiance observation and the performance of the enhanced radiance bias correction in GSI was examined.