中文摘要：

采用基于WRF模式的集合变分混合同化方法（Ens-3DVAR），对2013年双台风“菲特”和“丹娜丝”的路径、强度和降水进行模拟，结果表明：对双台风路径和强度的模拟，无论是模拟效果还是稳定性，Ens-3DVAR方法72h模拟效果最优；三种试验方法对降水都有一定的模拟能力，SAL评分表明无论是对降水结构、强度，还是降水位置的模拟，Ens-3DVAR方法模拟效果最好；从Ens-3DVAR和3DVAR方法得到的初始时刻的同化增量场来看，同化卫星资料后，两种方法均改变了初始场信息，但Ens-3DVAR试验与3DVAR试验的增量无论是大小还是分布范围明显不同，说明预报系统的局地信息改变对模拟效果有很大的影响；Ens-3DVAR方法采用集合背景场和流依赖性背景误差协方差，弥补了传统3DVAR中采用均匀、各向同性、准定常的背景误差协方差所带来的局限，提供了更接近实际大气的背景场；同时该方法采用了多个不同时刻的输入资料，说明Ens-3DVAR方法是数值预报中利用历史资料的一种可行途径。

英文摘要：

A hybrid ensemble three-dimensional variational data assimilation （Ens-3DVAR）system was studied for the forecasting of the tracks, intensity and typhoon rainfall of two binary typhoons of FITOW and DANAS in 2013. The results show that for both the effect and stability of track and intensity of simulated binary typhoons, Ens-3DVAR is the best of the three methods compared. All of the three kinds of test have some ability for precipitation simulation. SAL score results show that for the structure, strength, and precipitation position, the Ens-3DVAR method has the best effect. Detailed diagnostics further revealed that the increments produced by Ens-3DVAR and 3DVAR were different for both the original information of typhoon itself and its environment. In particular, 3DVAR changed the information about the whole simulation area, while Ens-3DVAR changed the information about the typhoon. This suggests that the change in local information greatly influences the simulation effect. Ens-3DVAR, unlike WRF 3DVAR, assumes that the background forecast-error covariances are static and nearly homogeneous and isotropic. The flow-dependent ensemble covariance and effective background provided by Ens-3DVAR improve the result of assimilation and numerical prediction. At the same time, the input data of different time used in Ens-3DVAR proves that the use of historical data in numerical prediction is feasible.