Intermediate glacial states were characterized by large temperature changes in Greenland and the North Atlantic, referred to as Dansgaard–Oeschger (D–O) variability, with some transitions occurring over a few decades. D–O variability included changes in the strength of the Atlantic meridional overturning circulation (AMOC), temperature changes of opposite sign and asynchronous timing in each hemisphere, shifts in the mean position of the Intertropical Convergence Zone and variations in atmospheric CO₂. Palaeorecords and numerical studies indicate that the AMOC, with a tight coupling to Nordic Seas sea ice, is central to D–O variability, yet, a complete theory remains elusive. In this Review, we synthesize the climatic expression and processes proposed to explain D–O cyclicity. What emerges is an oscillatory framework of the AMOC–sea-ice system, arising through feedbacks involving the atmosphere, cryosphere and the Earth’s biogeochemical system. Palaeoclimate observations indicate that the AMOC might be more sensitive to perturbations than climate models currently suggest. Tighter constraints on AMOC stability are, thus, needed to project AMOC changes over the coming century as a response to anthropogenic carbon emissions. Progress can be achieved by additional observational constraints and numerical simulations performed with coupled climate–ice-sheet models.

中期冰川状态的特征是格陵兰岛和北大西洋的温度发生较大变化，这被称为Dansgaard-Oeschger（D-O）变异性，并且在几十年中发生了一些转变。D–O的变化性包括大西洋经向翻转环流（AMOC）的强度变化，相反符号的温度变化和每个半球的异步时间，热带辐合带平均位置的变化以及大气CO _2的变化。古记录和数值研究表明，AMOC与北欧海冰紧密耦合，是D-O变异的核心，但是，一个完整的理论仍然难以捉摸。在本综述中，我们综合了气候表达和提议的解释D–O周期性的过程。通过涉及大气，冰冻圈和地球生物地球化学系统的反馈，出现了AMOC-海冰系统的振荡框架。古气候观测表明，AMOC可能比目前的气候模型更敏感。因此，需要对AMOC稳定性施加更严格的限制，以预测下一个世纪AMOC的变化，以应对人为碳排放。