中文摘要：

由于洋面上常规观测资料严重不足，使得客观分析的结果难以准确地描写初始台风的热力结构和环流特征，这是造成台风数值预报误差较大的重要原因。传统的Bogus方案虽然对台风预报有很大改进，但是仍有许多不足。文中在前人研究工作的基础上，对Bogus方案进行了改进，提高了对台风强度的描述能力，并结合四维变分资料同化的方法，利用模式动力学约束，自动生成具有非对称环流结构的台风初值，最后进行了四维变分同化试验和数值模拟试验。数值结果表明：这种结合Bogus方法和四维变分同化方法的台风初值化方案对于台风的路径模拟和强度模拟均有较好的改进。

英文摘要：

The serious shortness of conventional observation data over ocean makes the initial thermodynamic and circumfluent structures of typhoon difficult to be accurately described by the objective analysis results. And it is also a quite significant reason for errors of typhoon numerical forecasting. Though typhoon numerical forecasting can be greatly improved by the Bogus methods, it still has a lot of problems. This paper introduces an improved initialization method of typhoon to reconstruct the initial fields. The new method can well describe the circumfluent structure and intensity of typhoon. The four dimensional variational data assimilation method（4DVAR） is also used in this paper to assimilate the Bogus data into the numerical model. Under the restriction of numerical model dynamics, the unsymmetrical circumfluent structure of typhoon can be automatically rebuilt. The numerical experiment results show that the new initialization method is valid for improving both the typhoon track simulating and intensity simulating.