We rely on numerical models to project the ocean's response to climate change. On decadal and shorter timescales, their output can often be validated against state-of-the-art observations, whereas our theoretical understanding of the steady-state ocean supports conclusions on multi-millennial timescales. In between, there is a missing timescale on which model output cannot be compared either with a thorough oceanographic instrumental record or with a mature theory. Unfortunately, this is the timescale of anthropogenic climate change.
The immediate solution to this problem is the development of a robust conceptual understanding of the global ocean on these timescales, supported by model-data comparison of palaeoceanographic proxies. Here I borrow concepts from thermodynamics to connect global circulation, on multi-decadal and longer timescales, to the climate forcing over the dense water formation regions in the North Atlantic and the Southern Ocean. I will use this to understand palaeoclimate phenomena such as the bipolar seesaw, and to show how mass loss from Antarctic ice sheets has the potential to mitigate against a declining Atlantic meridional overturning circulation. It is therefore essential for global overturning to be evaluated even in modelling studies focused on the Atlantic, especially in high resolution simulations which typically have short spin-ups.