集合成员个数对集合预报技巧影响的试验分析

THE INFLUENCE OF THE NUMBERS OF THE MEMBERS ON THE PREDICTION SKILL OF ENSEMBLE FORECAST

利用全球中期预报模式T63L9,选取2004年6月4日至13日10d作为试验个例进行了集合预报试验,分析了不同集合成员个数对于预报结果的影响。结果表明,集合预报的技巧都明显高于单个控制预报。在集合成员较少时,随集合成员教的增加,集合预报的技巧提高明显,当集合成员数多于11个时,集合预报的效果提高缓慢。在中期预报时段内。集合成员数11为集合预报效果随集合成员教趋于饱和的临界值,如果继续增加成员数.预报效果提高较少,但计算量却大大的增加。本文只是单个试验个例的分析结果。为验证结论的普适性,还需要进行更多的试验。

The use of global medium-term forecasting model T63L9, ten days of June 4, 2004 to 13 as a test case of the ensemble prediction test, analysis of a number of members of different sets for forecasting is carried out. The results indicate that ensemble skills are significantly better than forecast individual control. Set in fewer members, with a few members of the pool, ensemble skills improve significantly, when the set number of members more than 11, the ensemble effect is improved slowly. In the medium-term forecast period, set the number of members of the ensemble forecasting 11, the results tend to the threshold saturation, if the number of members continued to increase, forecasting results raise less, but the cost is increased substantially. In this paper it is just a single example of the test, to verify the universal conclusion; more tests are needed.