欢迎访问中国科学院大学学报,今天是

中国科学院大学学报 ›› 2010, Vol. 27 ›› Issue (3): 420-431.DOI: 10.7523/j.issn.2095-6134.2010.3.017

• 优秀博士论文 • 上一篇    

一个新的大样本ENSO集合预报系统的发展与检验

郑飞, 朱江   

  1. 中国科学院大气物理研究所,北京 100029
  • 收稿日期:2009-12-08 发布日期:2010-05-15
  • 通讯作者: 郑飞
  • 作者简介:郑飞:获2008年度中国科学院优秀博士学位论文奖和2009年度全国百篇优秀博士学位论文奖. 现主要从事短期气候预测、集合资料同化与集合预报研究.
    导师 朱江 研究员:主要研究领域为资料同化的方法以及在海洋、大气和环境中的应用.
  • 基金资助:

    Supported by the Natural Science Foundation of China (40805033), the Chinese Academy of Science (KZCX2-YW-202), the Chinese COPES Project (GYHY-200706005), and the National Basic Research Program of China (2006CB403600) 

Developments and verifications of a new large size ENSO ensemble prediction system

ZHENG Fei, ZHU Jiang   

  1. Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029, China
  • Received:2009-12-08 Published:2010-05-15
  • Supported by:

    Supported by the Natural Science Foundation of China (40805033), the Chinese Academy of Science (KZCX2-YW-202), the Chinese COPES Project (GYHY-200706005), and the National Basic Research Program of China (2006CB403600) 

摘要:

详细介绍了一个新的大样本集合预报系统. 为了减小ENSO(厄尔尼诺-南方涛动)预报中的预报不确定性,该集合预报系统首先基于一个中等复杂程度的耦合模式,利用集合卡尔曼滤波资料同化方法同化有效的海洋观测资料为集合预报系统提供集合初始场;同时,一个发展的用于12个月预报的一阶线性马尔可夫(Markov)随机误差模式被嵌套到集合预报系统中来模拟模式不确定性. 基于1992年11月~2008年10月100个样本的集合回报试验,从确定性预报技巧和概率预报技巧2个方面对集合预报系统的预报水平进行了检验. 该集合预报方法能够很有效地将传统的确定性预报扩展到概率预报领域,且检验结果表明,预报样本均值的预报水平要优于单一的确定性预报. 对于概率预报而言,集合预报样本能够很好地跟随观测的变化,并且能够提供单纯确定性预报所不能够提供的额外信息.

关键词: ENSO, 集合预报, 中等复杂程度耦合模式, 集合卡尔曼滤波, 马尔可夫随机模型

Abstract:

A new large size ensemble prediction system (EPS) is introduced in this paper. To minimize the forecast uncertainties for ENSO (El Ni o-Southern Oscillation) predictions, the EPS is firstly based on an intermediated coupled model (ICM), and then the ensemble Kalman filter (EnKF) data assimilation method is adopted to generate the initial ensemble conditions for the EPS through assimilating available oceanic observations. Meanwhile, a developed linear, first-order Markov stochastic model-error model is embedded in the EPS to represent the model uncertainties during the twelve-month ensemble forecast process. The prediction skill of the EPS is verified based on the (100-members) retrospective ensemble forecast experiment covering the period between November, 1992 and October, 2008 in both deterministic and probabilistic senses. This ensemble technique provides a successful method of extending the standard deterministic forecasts to the probabilistic domain. The verification results show that the prediction skill of the ensemble mean is better than that of one single deterministic forecast using the same ICM. For the probabilistic perspective, those ensemble forecasts have their ensembles following observational variations well, and provide additional information that could not be gleaned from a purely deterministic approach.

Key words: ENSO, ensemble prediction, ICM, EnKF, Markov stochastic model

中图分类号: