Was the 2015 North Atlantic subpolar cold anomaly predictable?
Was the 2015 North Atlantic subpolar cold anomaly predictable?
复制标题
DOI：
10.1175/jcli-d-20-0750.1
复制
发表时间：
2021-07
影响因子：
4.9
通讯作者：
E. Maroon;S. Yeager;G. Danabasoglu;N. Rosenbloom
中科院分区：
文献类型：
--
作者：
E. Maroon;S. Yeager;G. Danabasoglu;N. Rosenbloom
2015年，亚极地北大西洋（SPNA）经历了极端寒冷，这一事件通常被称为“冷团”。对2014年11月初始化的社区地球系统模式第1版10年代际预测大集合（CESM1-DPLE）后播中该事件的演变进行了对比。尽管SPNA初始条件已经很冷，尽管其他年份的SPNA海面温度技能都很高，但CESM1-DPLE后播未能预测2015年的寒冷天气。本文的目的是了解导致这种预测失败的原因，以便为未来的十年预测工作提供见解。我们的分析表明，2015年冬季和春季强烈的北大西洋涛动（NAO）条件可能持续了冷团，但没有在任何CESM1-DPLE成员中模拟。我们使用CESM1- dple的未初始化对应，CESM1大集合（CESM1- le）来检查2015年事件的稀有性。CESM1-LE的结果表明，2015年冬季观测到的NAO异常状态至少部分解释了为什么该事件没有包含在CESM1-DPLE传播中。为了测试另一种可能性-初始条件的缺陷降低了预测-我们使用CESM1-DPLE协议但不同的初始条件进行了额外的预测。改变初始条件并没有改善2015年冷团的模拟，在某些情况下，还降低了模拟效果。鉴于预测该事件的难度，该案例可以作为未来预测系统开发的有用测试平台。
The subpolar North Atlantic (SPNA) experienced extreme cold during 2015, an event often called the “cold blob”. The evolution of this event in the Community Earth System Model version 1 Decadal Prediction Large Ensemble (CESM1-DPLE) hindcast initialized in November 2014 is compared to observations. This CESM1-DPLE hindcast failed to predict cold conditions during 2015 despite already cold SPNA initial conditions and despite having high sea surface temperature skill in the SPNA in all other years. The goal of this paper is to understand what led to this prediction failure in order to provide insight for future decadal prediction efforts. Our analysis shows that strongly positive North Atlantic Oscillation (NAO) conditions during winter and spring 2015 likely sustained the cold blob but were not simulated in any CESM1-DPLE members. We examine the rarity of the 2015 event using the CESM1-DPLE’s uninitialized counterpart, the CESM1 Large Ensemble (CESM1-LE). Results from the CESM1-LE indicate that the exceptional state of the observed NAO in the winter of 2015 is at least part of the explanation for why this event was not encompassed in the CESM1-DPLE spread. To test another possibility — that deficiencies in the initial conditions degraded the prediction — we performed additional hindcasts using the CESM1-DPLE protocol but different initial conditions. Altering the initial conditions did not improve the simulation of the 2015 cold blob, and in some cases, degraded it. Given the difficulty of predicting this event, this case could be a useful testbed for future prediction system development.