We study the characteristics of wintertime Arctic amplification (AA) and the inter-model spread in model output from the sixth phase of the Coupled Model Intercomparison Project (CMIP6) and the Community Earth System Model Large Ensemble Project (CESM-LE) . The CMIP6 models exhibit a large spread in AA intensity (as measured by an AA index, ranging from about −0.2 to about 1.1 °C/decade in the Historical simulations) and in its spatial/vertical structure (no AA to AA extending to upper-troposphere). A large model spread in Arctic sea-ice trends (from about −3.1 to about 0.3 %/decade) is also found. Intra-model spread in the ensemble simulations of six selected CMIP6 models, which we interpret as a measure of internal climate variability, is similarly large. We further find that CMIP6 model integrations where AA extends to the upper-troposphere, show weaker warming in the midlatitudes. The observed recent midlatitude cooling trends are generally not seen in the one-realization CMIP6 models but are seen in some ensemble members in the CMIP6 ensemble simulations; They are likely not a response to Arctic sea ice decline according to our analysis and are more likely caused by internal climate variability. By comparing the CMIP6 inter-model spread to the intra-model spread in the six selected CMIP6 ensembles and CESM-LE simulations, we argue that a large part of the CMIP6 inter-model spread in AA index, surface air temperature (SAT) trends and Arctic sea ice trends can be attributed to internal climate variability: about 67% for AA index, 58% for Arctic SAT and 22% for Arctic sea-ice. However, the inferred internal climate variability has been relatively stable in the recent several decades, while the inter-model spread in Arctic temperature trends and particularly Arctic sea-ice trends has been increasing in the CMIP6 models. This highlights the importance of constraining CMIP6 model discrepancies in simulating recent climate variations, which will potentially benefit the future multimodel climate projections.

我们研究了耦合模型比对项目（CMIP6）和社区地球系统模型大型集合项目（CESM-LE）第六阶段的模型输出中冬季北极放大（AA）和模型间扩展的特征。CMIP6 模型在 AA 强度（由 AA 指数测量，在历史模拟中范围从约 -0.2 到约 1.1 °C/十年）和其空间/垂直结构（没有 AA 到 AA 延伸到上部-对流层）。还发现了北极海冰趋势的大型模型（从大约 -3.1 到大约 0.3 %/十年）。六个选定的 CMIP6 模型的集合模拟中的模型内传播，我们将其解释为内部气候变率的度量，同样很大。我们进一步发现 CMIP6 模型集成，其中 AA 延伸到对流层上层，在中纬度地区显示出较弱的变暖。观察到的近期中纬度降温趋势通常在单一实现的 CMIP6 模型中看不到，但在 CMIP6 集合模拟中的一些集合成员中可以看到；根据我们的分析，它们可能不是对北极海冰减少的反应，更有可能是由内部气候变化引起的。通过在六个选定的 CMIP6 集合和 CESM-LE 模拟中比较 CMIP6 模型间扩展与模型内扩展，我们认为 CMIP6 模型间扩展的很大一部分在 AA 指数、地表气温 (SAT) 趋势和北极海冰趋势可归因于内部气候变化：AA 指数约为 67%，北极 SAT 约为 58%，北极海冰为 22%。然而，推断的内部气候变率在最近几十年相对稳定，而在 CMIP6 模型中，北极温度趋势，特别是北极海冰趋势的模型间传播一直在增加。这突出了在模拟近期气候变化中限制 CMIP6 模式差异的重要性，这可能有利于未来的多模式气候预测。