数值模式误差订正方法初探

Study on Numerical Model Error Correction Scheme

鉴于大气系统的复杂性,实际的大气运动方程包含许多数学简化。这些简化的方程组虽能模拟大气运动的主要特征,但无法揭示天气演变的所有细节,这将导致数值模式的预报能力受到极大限制。只有完善大气运动方程本身的确定性框架,才能从根本上减少预报误差。若无其他手段的帮助,仅依靠推导方法获取大气运动过程中的所有微分表达式将十分困难。演化算法具有自适应、自组织特征,它能处理复杂的数学问题,可从观测资料所蕴含的丰富信息中挖掘系统的动力学。本文结合演化建模的智能算法,提出了一种数值模式的误差订正方案,并以一维典型混沌系统Logistic模型为例,探讨了误差订正方案的有效性。数值试验结果表明该方法对模式误差有较强的反演能力,值得进一步深入研究。

Numerical climate model is composed of the equations with much mathematical simplification due to the complex of atmospheric motion.The simplified equations only can describe main features of climate system,but not to reflect all the details in atmospheric cycles.So,the predictivity of current numerical model is limited always.However,with an adaptive and self-organizing ability,evolutionary algorithm can be used to deal with some complex mathematical problems and extract dynamical information of a dynamic system from the observations.On basis of the intelligent characteristics of evolutionary algorithm,a dynamical method for model error correcting has been presented in this paper.To a typical one-dimensional chaotic system-Logistic model,an error correction scheme is provided and tested.Numerical results show that the scheme can be used to correct model error;it has a potential application prospect,which is worth to be further researched.