Abstract:The evaluation results of ECMWF and GFS on the track and strength forecasts of typhoons affecting Zhejiang from 2016 to 2019 show that ECMWF is better than GFS for track forecasting, and GFS has more advantages for strength forecasting. On this basis, the paper proposes a feasible typhoon initialization scheme to improve the performance of typhoon forecasting skill. Based on ECMWF and GFS analysis fields and oceanic typhoon observation data, as well as typhoon vortex separation, the vortex field is obtained by separating the GFS analysis field, and after vortex relocation and maximum wind speed adjustment procedure, it is superimposed with the large-scale environmental field separated from the ECMWF analysis field, realizing the reconstruction of typhoon initial field. For 14 typhoons affecting Zhejiang in recent years, the hindcasting results of the mesoscale numerical weather forecast model which uses the new reconstruction scheme demonstrate that the new scheme gives full play to the ECMWF’s advantages in track forecasting and GFS’s advantages in strength forecasting, improving the forecast accuracy of typhoon track and strength effectively. The absolute error of track forecast is reduced by 21 km and the standard deviation by 26.6 km compared with the prediction of the GFS-driven mesoscale model. Compared with the forecast result of the ECMWF driven mesoscale model, the absolute error of strength forecast is reduced by 1.7 m·s-1 and the standard deviation by 2.3 m·s-1. The analysis of the typical case of Super Typhoon Lekima (2019) further indicates that the initial field reconstruction technology has a better forecasting ability for atmospheric circulation characteristics and typhoon’s warm core structure.