Development of Optimized Radar Data Assimilation Capability within the GSI-based Fully Coupled EnKF-EnVar Hybrid System for Convective-Permitting Ensemble Forecasting
Dr. Chengsi Liu
Center for Analysis and Prediction of Storms
University of Oklahoma, USA
When directly assimilating radar data within the variational data assimilation framework, if hydrometeor mixing ratios (q) are used as control variables (CVq), the extremely large cost function gradient with respect to very small background q may prevent efficient convergence and lead to inefficient assimilation of radial velocity and reflectivity observations. Another hydrometeor control variables is using logarithmic hydrometeor mixing ratios as control variables (CVlogq), which can alleviate the above problems of CVq, but produces spurious analysis increments due to the high nonlinearity of the logarithmic transformation. In this study, power transformed hydrometeor mixing ratios are used as control variables (CVpq), and the nonlinearity degree of the transformation can be adjusted by tuning a parameter p. Based on the GSI-based En3Dvar data assimilation system, the impact of CVpq on the direct radar radial velocity and reflectivity data assimilation is examined through the analysis and forecast experiments of five convective-scale storm cases in 2017. Results show that, in terms of the smallest root mean square error, using CVpq with 0.4 parameter p (CVpq0.4) gives the best 60-miniute forecasts of reflectivity. Compared with CVq, CVpq0.4 has much faster convergence speed of minimization and better radial velocity analysis. In addition, because CVpq0.4 uses the more linear transform than the logarithmic transform, the spurious analysis increments occurred in CVlogq are mostly alleviated.