Direct Assimilation of Radar Reflectivity within the NOAA Operational Hybrid EnVar System to Improve High-Impact Weather Forecasts: Development of the Convective-Scale Static Background Error Covariance

Past studies have shown that hybridizing the static and ensemble covariances in data assimilation can improve large scale global forecast compared to a standalone ensemble filter (e.g. Wang et al. 2013). However, studies on the impact of including the static covariance through a hybrid approach on convective scale prediction are rather limited. This is partially because the static covariance while well-tuned for large scale global NWP may not be well suited for convective scales. Given the deficiencies of the ensembles, it is hypothesized that a static covariance if well developed for convective scales and combined with the ensemble covariance would further improve the analysis and prediction at convective scales compared to using the pure ensemble covariance.

A convective-scale static B is extended and well-constructed in this study for convective scales including the capability of direct assimilation of radar reflectivity. The development includes 1) adding additional control variables for convective scales; 2) computing B statistics adaptively; 3) allowing for geographical variations. The impact of the newly developed static B is firstly examined on the 8 May 2003 Oklahoma City tornadic supercell. Relative to experiment using full ensemble covariance (EnVar, Wang and Wang 2017), incorporating static B (Hybrid) remarkably reduces the spin-up time and better initializes the reflectivity of storms during the DA period. Consequently, the predicted reflectivity distributions in Hybrid fits better to the observations than that in EnVar. This convective hybrid EnVar with the newly developed static B is also tested over high impact convective scale weather over the CONUS. It is found Hybrid overall performs better than the pure EnVar during almost entire 18-h forecast, especially for high reflectivity thresholds.