Efficient Regional Hybrid Ensemble-Variational Data Assimilation using the Global-Ensemble-Model-Augmented Error Covariance for Numerical Weather Prediction over Eastern China

Abstract

An efficient regional hybrid ensemble-variational (EnVar) data assimilation method using the global-ensemble-model-augmented error covariance is proposed and preliminarily tested in this study. This method uses the global ensemble error covariance as the complementary low-resolution regional ensemble error covariance. The high-resolution dynamic ensemble mean is used as the first guess in hybrid EnVar and then re-centered to the updated high-resolution dynamic ensemble perturbations after minimization analysis. In this study, the proposed method is implemented into the Weather Research and Forecasting Model’s (WRF) data assimilation system coupled with the ensemble transform Kalman filter (ETKF) and preliminarily tested for numerical weather prediction during the Mei-Yu season over eastern China. It is found that the experiment containing fewer regional dynamic ensemble members but augmented with global ensemble error covariance obtains similar results to the experiment containing many more regional dynamic ensemble members. However, the former experiment only takes up one third of the latter experiment’s computational cost. The method proposed in this study also outperforms the 3DVar, hybrid EnVar using the pure global ensemble error covariance, as well as the hybrid EnVar using regional ETKF ensemble with a smaller size. The method proposed in this paper effectively combines the contributions of the ensemble error covariance from both the global and the regional models to produce better initial conditions for the regional WRF data assimilation system.