增长模叠加方式及其对集合预报初始扰动的影响

The Effect of Superposition Manners of the Bred Modes on the Initial Perturbations in Ensemble Forecasting

用增长模繁殖法(BGM)有效地构造集合预报初始扰动的一个前提是,任两个扰动之间应尽可能的发散。本文从增长模发展和饱和的角度分析了增长模叠加方式与集合预报初始扰动“取样性”的关系,无论是从扰动形势演变得出的定性结论还是定量计算的结果都显示:在相同初始模的情况下,通过改变扰动叠加方式分别构造多个扰动样本的做法是近似等效的。用这种方法增加的集合样本从“取样性”的角度来看是没有多大意义的,更多的是在浪费有限的计算资源。构造多个初值的有效方法应该是从不同初始模出发进行增长模的繁殖。

Generating initial perturbation in ensemble forecasting by using the breeding growing modes should be based on the premise that either perturbation should spread from other possibly. Focusing on the saturation of growing modes, the relation between the way of superposing the perturbation upon the analysis and sampling of initial perturbations in ensemble forecasting is discussed in this paper. The result shows that the ensemble samples generated by changing the manner to superposing on growing modes have the identical effect if the same initial modes are given, which is meaningless as the sampling is concerned. Therefore, the effective method to create more initial perturbations should begin with more different initial random modes.