Development and Research of GSI-based EnKF/Var/Hybrid System to Assimilate Convective-scale Radar Observations for Warn-On-Forecast (WoF) Applications

To achieve the ‘Warn-on-Forecast (WoF)' strategic goal (Stensrud et al. 2009), one key component is to provide sufficiently accurate initial conditions through assimilation of radar and other high-resolution data to initialize the convective-scale NWP models. In this project, we further developed the US National Weather Service's operational GSI-based EnKF/Var/Hybrid data assimilation system to include convective-scale radar data assimilation capability targeted for WoF applications.

Vertical velocity and hydrometeor related control variables were developed for both the EnKF and EnVar variational components of the hybrid system. Radar reflectivity observation operator was included. Various options of hydrometeor-related control variables were derived, developed and explored. Experiments were conducted with the newly extended GSI EnKF/Var/hybrid system for an 8 May 2003, Oklahoma City, tornadic supercell storm. A 45-member ensemble was used. The 3km convection allowing domain was one-way nested inside a 12km domain, which provided inhomogeneous mesoscale environment for the inner domain radar data assimilation. It was found that after 1 hour of radar data assimilation, GSI hybrid system using flow-dependent ensemble covariance produced realistically analyzed vorticity, reflectivity and updraft for the tornadic supercell. The subsequent ensemble forecast predicted the probability of a strong low-level vorticity track of the tonadic supercell that correlates well with the observed tornado track. The deterministic forecast was able to maintain a strong updraft and vorticity for a one-hour period. In contrast, using static covariance in GSI 3DVar, the analyzed updraft was rather weak and the tornadic supercell dissipated quickly during the subsequent forecast.

The impact of newly developed reflectivity control variable in GSI hybrid for reflectivity assimilation was also explored for the same case. It was found that the newly developed control variable produced the best analysis and forecast for the tornadic supercell. Detailed diagnostics suggested that this better performance was because the graupel and ice mixing ratio was better analyzed. The better analyzed hydrometeor fields produced a realistic cold pool, which led to the constructive interaction between surface gust front and the updraft aloft associated with the mid-level meso-cyclone.