Uniform and Anisotropic Flow-Dependent Localization within JEDI EnVar for Direct Radar Reflectivity Data Assimilation in the Next-Generation Rapid Refresh Forecast System (RRFS)

Past studies commonly employed a uniform localization radius of 5-20 km in the ensemble-based radar data assimilation (DA) to alleviate sampling errors caused by the limited ensemble size. However, many studies (Destouches et al. 2020; Montmerle and Berre 2010) highlighted the necessity of flow-dependent and spatially varying localization for convective-scale DA. In the context of multiscale DA, Wang and Wang (2023ab) also suggested that the flow-dependent localization method outperforms the uniform localization method in the assimilation of radar observations. In this study, we further developed the uniform and anisotropic flow-dependent localization (Ménétier et al. 2015) in JEDI Background error on an Unstructured Mesh Package (BUMP) for direct radar reflectivity assimilation. We first demonstrate that JEDI-based EnVar produces comparable analyses and forecasts with GSI-based EnVar using the same uniform localization method for direct assimilation of radar reflectivity observations (Wang and Wang 2017, Duda et al. 2019). Experiments are further conducted to compare the performance in analysis and forecast of the 3 May 2018 squall lines between uniform and flow-dependent localization methods. Detailed results will be presented at the conference.