To improve the analysis quality through incorporating the flow-dependent ensemble covariance into variational data assimilation system, the new GRAPES hybrid-3Dvar system is built, based on GRAPES regional 3DVar system which uses the statistic covariance, by augmenting the state vectors with another set of control variables preconditioned upon the ensemble dynamic covariance. The new Hybrid-3Dvar system and localization method has been verified through the single observation assimilation experiment with ensemble samples produced by 3D-Var’s control variable perturbation method. The real observation assimilation and forecast experiment for typhoon Soudelor come to the conclusions: (1) the background covariance which is represented by ensemble samples are flow-dependent and the root mean square (RMS) spread in the ensemble of momentum field and mass field are largest near the center of typhoon; (2) the analysis increments of the new Hybrid-3DVar have more detailed structure and more medium and small-scale information; (3) The analysis and 24h prediction qualities of model variables in the new Hybrid-3DVar are obviously improved in comparison with the 3DVar system, and the precipitation position predictions are more accurate; (4) The 24h forecast track of typhoon Soudelor is closer to observational one and the 48h-predicted intensity approaches the real observation as well.