Improving Short-Term Precipitation Forecasting with Radar Data Assimilation and a Multiscale Hybrid Ensemble–Variational Strategy

Abstract

Abstract This paper presents a multiscale hybrid ensemble–variational (EnVar) data assimilation strategy with an hourly rapid update aiming to improve analysis of convection via radar observations and of convective environment via conventional observations. In this multiscale hybrid EnVar strategy, the ensemble members are updated by assimilating conventional data using an EnKF to provide the hybrid EnVar with flow-dependent background error covariance (BEC). A two-step approach is employed in the hybrid EnVar to achieve improved multiscale analysis by assimilating radar data and conventional data, respectively, in two successive steps. This two-step procedure enables the applications of different BEC tuning factors and different hybrid weights for radar and conventional observations. In addition, this study also examines the impacts of the flow-dependent BEC generated with and without radar data assimilation in EnKF on the performance of hybrid EnVar analysis and ensuing convective forecasting. The multiscale hybrid EnVar strategy was first evaluated through a comparison with 3DVar and EnKF using a convective rainfall case. Quantitative verifications for both precipitation and environmental variables demonstrated that the hybrid EnVar system with an optimal multiscale configuration outperformed both the 3DVar and EnKF. The multiscale hybrid EnVar strategy was then evaluated through a series of sensitivity experiments. It was shown that the two-step assimilation strategy outperformed the one-step for both the precipitation and environmental variables, and the ensemble BEC generated without radar data assimilation led to improved hybrid EnVar analysis over that with radar data assimilation by better representing uncertainties in convective environment and reducing spurious spatial and multivariate correlations.