Hysteresis and Resilience of the AMOC in an Eddy-Permitting GCM

We show a quasi-irreversible shutdown (hysteresis) of the Atlantic meridional overturning circulation (AMOC) in an unfluxadjusted global climate model with an eddy-permitting ocean. This global climate model is a prototype of one submitted to the sixth Coupled Model Intercomparison Project and is the most comprehensive model to show this behavior. The AMOC has a resilience timescale for a given hosing: If the hosing exceeds the critical hosing for collapse for only a limited time, the AMOC recovers when the hosing is stopped. Beyond the critical time, when the hosing stops the AMOC stays in a weak state with no recovery within 200 years. We show that simple, observable metrics such as AMOC strength and mixed layer depth can indicate whether this critical resilience time has passed and that the mechanism controlling the presence of recovery is a positive advective feedback. Plain Language Summary The Atlantic meridional overturning circulation (AMOC) is an important part of our climate system, which keeps the North Atlantic relatively warm. It is predicted to weaken under climate change. The AMOC may have a tipping point beyond which recovery is difficult, hence showing quasi-irreversibility (hysteresis). Although hysteresis has been seen in simple models, it has been difficult to demonstrate in comprehensive global climate models. We use a comprehensive global climate model to examine the AMOC response to the addition of freshwater (hosing) in the North Atlantic, such as from melting glaciers. The freshening weakens the AMOC, and if the AMOC is weakened sufficiently it remains in a weak state even when the hosing is stopped. We also introduce and demonstrate the concept of a timescale of resilience: Although a particular hosing might shut down the AMOC if experienced for long enough, the AMOC can recover if the hosing is only experienced for a limited time. Indicators of whether the AMOC can recover are discussed, with implications for monitoring. Results also have relevance for the modeling and policy communities in providing results that can be used to understand the uncertainties around an AMOC shutdown.