Abstract: For the demand of fine prediction and based on the features of high resolution models, the "non-bogussing" initialization was tested to make fine prediction of tropical cyclones (TC), i. e., the initial fields from 4-dimensional assimilation was adopted, without adding the TC bogussing, to avoid illusive information induced from idealized TC structure, and to acquire finer or more accurate prediction to the track, intensity and structure of TCs. The feasibility of the scheme was evaluated by using all TC cases which affected the South China in 2001. The results show that the scheme has a better capability to predict the genesis of TCs, and higher accuracy for track prediction, with prediction errors unremarkably increasing with integral time, and especially reflects better the complicated track and local effects. While it shows a higher skill for TC intensity prediction in terms of intensity change. In addition, it reflects well the asymmetric and spiral structures of cloud, rainfall and winds for the TCs, showing well practicability. All of these revealed the advantages and developing potentials of the scheme, which implies that "non-bogussing" initialization should be one of the main approaches for the TC numerical prediction techniques in the future. However, at the current condition, the scheme possesses some insufficiencies. The main problem is that due to the lack of observational information, larger errors may often occur in the initial TC position or intensity, which may cause larger prediction errors, when the TC is too weak, newly arising or far away from the mainland.