Development of Weather-dependent Background Error Structures within a Convective Scale Variational Radar Data Assimilation System

How to create weather-dependent error structures for better assimilating high-density WSR-88D radar observations in convective scale NWP models has become an increasingly active research topic. Yet the implementation and assessment of such structures in a convective-scale data assimilation system are still not well studied. In this study, a weather-dependent binning approach is adopted and further extended to separately construct background error structures for the clear sky and several weather classes based on strengths of radar composite reflectivity derived from an ensemble product. The background error statistics are generated using an ensemble of 3-hour forecasts at 3-km horizontal resolution from 25 severe weather events during the 2019 Hazardous Weather Testbed (HWT) spring forecast experiment. Following that, the sampled weather-dependent background error profiles are integrated into the NSSL Warn-on-Forecast variational analysis and forecast system (WoF-Var). The impact of the weather-dependent background error structure on the WoF-Var analyses is firstly assessed with a single observation experiment. This experiment highlights the heterogeneous nature of the analysis increment in the control variables, which is corresponding to different weather situations. This updated WoF-Var is then evaluated with one severe weather events: 10 Dec 2021 long-lived supercell, which produced significant property damage and fatalities. Comparisons of experiments with and without weather-dependent background errors reveal that the new scheme using weather-dependent error structure is able to better tune the in-storm structure. Consequently, the better analyzed storm structure has a positive impact on the forecast of the long-lived supercell storm several hours in advance.