中文摘要：

本文将ENSO预测的目标观测敏感区与多模式集合预报方法相结合,提出了一种能够有效提高预报技巧且又具有较小计算成本的多模式集合预报方法.该方法在目标观测敏感区内采用模式不等权的多模式超级集合预报方法（SUP）,而在其他区域采用相对简单的等权的多模式消除偏差集合平均方法（BREM）.利用CMIP5中15个气候系统模式的工业革命前参照试验（pi-Control）数据,针对热带太平洋海温的长期演变开展了理想预报试验.将新集合预报方法与现有的多模式集合预报方法进行了比较.结果表明,在所考察的预报期内（即1～20年）,新集合预报方法与整个热带太平洋区域使用SUP方法具有相当的预报技巧,但前者的计算成本明显小于后者,计算时间仅为后者的1/4.可见,新方法是一个具有较高预报技巧且计算成本较小的多模式集合预报方法.同时,其较高的预报技巧强调了热带太平洋SST预测对ENSO目标观测敏感区内的模式误差也是极端敏感的,也正因如此,多模式集合预报方法才能够有效过滤模式误差的影响,具有较高的预报技巧.

英文摘要：

In this study,we attempted to combine the sensitive area for target observation associated with E1 Ni（n）o-Southern Oscillation predictions with multimodel ensemble forecast methods and proposed a new method that provides robust forecast skill with smaller computational cost.Specifically,this new method allows for superensemble prediction （SUP） with unequal weighting and higher skill to be used in perturbation sensitive areas and for the bias-removed ensemble mean with equal weighting to be used in other regions.By using the monthly mean data of pre-industrial control simulations of 15 models in Phase 5 of the Coupled Model Intercomparison Project experiments,we compared this new method with the existing multimodel ensemble forecast methods preliminarily under the ideal forecast experiments of tropical Pacific sea surface temperatures （SSTs） in various forecast periods.The results show that in the forecast period of 1-20 years,the proposed method has high forecast skill similar to that when using the SUP with SSTs in the entire tropical Pacific.The new method greatly reduces the computational cost,and the computation time is only one-fourth of that in the SUE Therefore,the new multimodel ensemble forecast methods are efficient and have more accurate forecast skill,which further demonstrates that the tropical Pacific SST forecasts are extremely sensitive to model errors in the sensitive region.Thus,the new method has a good performance in prediction skill.