Abstract: On the basis of the dynamic retrieval of sea level pressure and data assimilation， a new approach to vortex initialization is proposed and examined in this study for typhoon prediction. Two of the boundary layer numerical models designed respectively for midlatitude and tropical regions are introduced to retrieve the sea level pressure （SLP） in dynamic consistency with the QuikSCAT sea winds. The SLP retrieval is then assimilated into a meso-scale numerical model in two steps with the three-dimensional variational (3DVAR) data assimilation technique. Sensitivity experiments for two typhoons made landfall on the east coast of China show that the status of boundary layer stability can be influenced by the assimilation, which is associated with the vertical wind shear within the boundary layer inflow. The vortex initialization also affected the outflow structure of vortex near the tropopause. Most importantly, the coupling between the boundary layer physics and the data assimilation favored the simulation of typhoon intensity, in which the positive effects of assimilation become more evident with the increase of model hours. The maintenance and rapid weakening phases of intensity are more reasonably reproduced in the sensitivity experiments than that in the control run. Finally, the new technique enhanced the simulation of the typhoon track through the adjustment of steering flow on the various pressure levels. 