数值预报发展的新方向——集合数值预报

A Brief Overview of Ensemble Forecasts

数值预报走过了百年的历史(V.Bjerknes,1904)[1],取得了了不起的成就,预报准确率有了很大的提高。但是,考虑到大气的非线性作用以及初值和模式造成的数值预报不确定性,要提高传统的单一确定性预报的技巧越来越难。基于大气初值的敏感性试验,Lorenz提出了数值预报发展的新方向——集合数值预报。通过介绍集合数值预报的基本思想和方法;并针对初值和模式的不确定性,概述现有集合数值预报成员的产生方法及国内外的研究进展;同时解释怎样检验集合预报的结果的论述,如何从众多的集合预报产品中提炼有用的信息及预报产品的释用。

More than a century pasted since the approach of numerical weather prediction was firstly proposed by V. Bjerknes （1904）. A lot of dramatic progresses have been made to increase the skills of numerical weather prediction. However, it seems to be more and more difficult to improve the accuracy of the single deterministic forecast due to the intrinsic uncertainties （originated from both initial conditions and model approximations） and the chaotic atmosphere. Based on the initial condition uncertainties, a pioneering work has been made by Lorenz （1965） to orientate the numerical weather forecasts--ensemble forecast. This paper sums up the basic Concepts, outstanding issues and recent development of ensemble technique, which includes （1）what＇ s the ensemble weather prediction （2） the recent development of ensemble prediction concerning the generation methods of initial perturbation and model perturbation. （3） How to validate ensemble prediction and interpret and evaluate its products.