中文摘要：

The GRAPES-TCM is used to make ensemble prediction experiments for typhoon track.Three kinds of ensemble schemes are designed for the experiments.A total of 109 experiments are made for the nine typhoons in 2011 and the integral time is 72 h.The experiment results are shown as follows.In the three ensemble schemes,on the whole,scheme 1 has the best track prediction.Its average absolute track error and overall deviations of typhoon moving speed and moving direction are all the smallest in the three schemes.For both scheme 1 and scheme 2,they are all smaller than those of their control predictions.Both of their ensemble predictions show superiority to their deterministic predictions.Overall,compared with the observations,the typhoon moving directions of the three schemes mainly skew to the right,and in the late integration they mainly tend to be relatively slow.In the three schemes,the track dispersion of scheme 1 is the largest and that of scheme 3 the smallest.In scheme 1 it is much larger than in schemes 2 and 3.The difference of dispersion between scheme 2 and scheme 3 is small.The track dispersions of the three schemes are all much smaller than their rational dispersions.Compared with the eight domestic and overseas operational numerical weather prediction(NWP) models,scheme 1 has better predictions than the other seven operational models except ECMWF NWP model.Scheme 1 has the value of operational application.

英文摘要：

The GRAPES-TCM is used to make ensemble prediction experiments for typhoon track. Three kinds of ensemble schemes are designed for the experiments. A total of 109 experiments are made for the nine typhoons in 2011 and the integral time is 72 h. The experiment results are shown as follows. In the three ensemble schemes, on the whole, scheme 1 has the best track prediction. Its average absolute track error and overall deviations of typhoon moving speed and moving direction are all the smallest in the three schemes. For both scheme 1 and scheme 2, they are all smaller than those of their control predictions. Both of their ensemble predictions show superiority to their deterministic predictions. Overall, compared with the observations, the typhoon moving directions of the three schemes mainly skew to the right, and in the late integration they mainly tend to be relatively slow. In the three schemes, the track dispersion of scheme 1 is the largest and that of scheme 3 the smallest. In scheme 1 it is much larger than in schemes 2 and 3. The difference of dispersion between scheme 2 and scheme 3 is small. The track dispersions of the three schemes are all much smaller than their rational dispersions. Compared with the eight domestic and overseas operational numerical weather prediction （NWP） models, scheme 1 has better predictions than the other seven operational models except ECMWF NWP model. Scheme 1 has the value of operational application.