Relationship between predictability and forecast skill of ENSO on various time scales
Relationship between predictability and forecast skill of ENSO on various time scales
复制标题
DOI：
10.1029/2011jc007249
复制
发表时间：
2011-12
影响因子：
--
通讯作者：
Yanjie Cheng;Youmin Tang;Dake Chen
中科院分区：
文献类型：
--
作者：
Yanjie Cheng;Youmin Tang;Dake Chen
[1]利用LDEO 5模式，基于一个发展良好的厄尔尼诺-南方涛动（ENSO）系综系统，对1856 - 2003年期间的厄尔尼诺现象进行了预报。利用集合预报和最近发展的可预报性框架，探讨了ENSO的基于信息和基于集合的潜在可预报性措施。这些潜在的可预报性措施和实际的可预报性措施的关系进行了研究，从年际到年代际的多个时间尺度。结果表明，在3种基于信息量的潜在可预测性指标中，相对熵（RE）在量化基于相关性的预测能力方面优于预测信息量（PI）和预测能力（PP），而PI和PP在估计均方误差（MSE）的预测能力方面更优。这些关系的主要原因进行了分析，并确定了潜在的可预测性措施的控制因素。结果表明，在弱ENSO期间，信号分量对RE起主导作用，而色散分量对RE的贡献也相当大。
[1] Ensemble predictions are performed using the LDEO5 model for the period from 1856 to 2003 based on a well developed El Nino–Southern Oscillation (ENSO) ensemble system. Information-based and ensemble-based potential predictability measures of ENSO are explored using ensemble predictions and the recently developed framework of predictability. Relationships of these potential predictability measures and actual predictability measures are investigated on multiple time scales from interannual to decadal. Results show that among three information-based potential predictability measures, relative entropy (RE) is better than predictive information (PI) and predictive power (PP) in quantifying correlation-based prediction skill, whereas PI and PP are better indicators in estimating mean square error (MSE)-based prediction skill. The primary reason for these relationships is analyzed and the control factors of the potential predictability measures are identified. It is found that RE is dominated by the signal component, but the dispersion component has a comparable contribution during weak ENSO periods.