多源资料循环同化在台风“潭美”预报中的应用

为了检验不同观测资料在台风预报中的作用，以美国NCEP (National Centers for Environmental prediction)业务同化系统GSI (Grid Statistical Interpolation)为平台，选取2013年路径较复杂且登陆后降水持续较强的“潭美”台风过程为例，分别加入常规地面和高空观测资料、极轨卫星NOAA18、NOAA19、METOP-A和METOP-B资料，以及多普勒雷达基数据资料，探讨不同观测资料同化对台风的预报效果。同时，对台风采用Bogus初始化方案以及循环资料同化对台风路径和强度预报效果进行了对比分析。结果表明:常规观测资料对台风路径预报改善效果最明显，卫星资料的融入对海上台风路径的修正较好，而雷达资料则对台风登陆后的路径预报有改善；并且多源资料的融入效果最好。同时，采用Bogus方案可有效调整初始台风的位置和强度，从而对后期台风路径和强度预报有正效应。采用间隔6 h资料循环同化方法，可有效利用各时段资料，对台风路径和强度预报有较好的改善。 

THE STUDY OF MULTI-SOURCE DATA CYCLING ASSIMILATION USING GSI-3DVAR IN TRAMI TYPHOON FORECAST

The assimilation of multi-source data and its impact on the prediction of tropical cyclone Trami(2013) is discussed in the GSI system. The conventional observations, satellite data and radar data are assimilated respectively and the results are compared. It reveals that the assimilation of conventional observations greatly improves the TC track forecast. The assimilation of satellite data shows positive impact on the prediction of TC track before its landing, while the assimilation of radar data improves the prediction of TC track after its landing. It proves that the assimilation of multi-source data presents the best TC track forecast. The TC initialization based on BOGUS and the cycle assimilation are also discussed and their respective impact on TC prediction is also investigated. It shows that BOGUS initialization can adjust the location and intensity of initial vortex according to the observed data and thus improve the TC forecast. The cycling assimilation, which has a 6 h cycle period, is proved to have positive impact on the TC forecast by effectively assimilating more data.