This is a review about the Atlantic Meridional Overturning Circulation (AMOC), its mean structure, temporal variability, controlling mechanisms, and role in the coupled climate system. The AMOC plays a central role in climate through its heat and freshwater transports. Northward ocean heat transport achieved by the AMOC is responsible for the relative warmth of the Northern Hemisphere compared to the Southern Hemisphere and is thought to play a role in setting the mean position of the Intertropical Convergence Zone north of the equator. The AMOC is a key means by which heat anomalies are sequestered into the ocean's interior and thus modulates the trajectory of climate change. Fluctuations in the AMOC have been linked to low-frequency variability of Atlantic sea surface temperatures with a host of implications for climate variability over surrounding landmasses. On intra-annual timescales, variability in AMOC is large and primarily reflects the response to local wind forcing; meridional coherence of anomalies is limited to that of the wind field. On interannual to decadal timescales, AMOC changes are primarily geostrophic and related to buoyancy anomalies on the western boundary. A pacemaker region for decadal AMOC changes is located in a western "transition zone" along the boundary between the subtropical and subpolar gyres. Decadal AMOC anomalies are communicated meridionally from this region. AMOC observations, as well as the expanded ocean observational network provided by the Argo array and satellite altimetry, are inspiring efforts to develop decadal predictability systems using coupled atmosphere-ocean models initialized by ocean data.

The subpolar North Atlantic (SPNA) is subject to strong decadal variability, with implications for surface climate and its predictability. In 2004–2005, SPNA decadal upper ocean and sea‐surface temperature trends reversed from warming during 1994–2004 to cooling over 2005–2015. This recent decadal trend reversal in SPNA ocean heat content (OHC) is studied using a physically consistent, observationally constrained global ocean state estimate covering 1992–2015. The estimate's physical consistency facilitates quantitative causal attribution of ocean variations. Closed heat budget diagnostics reveal that the SPNA OHC trend reversal is the result of heat advection by midlatitude ocean circulation. Kinematic decompositions reveal that changes in the deep and intermediate vertical overturning circulation cannot account for the trend reversal, but rather ocean heat transports by horizontal gyre circulations render the primary contributions. The shift in horizontal gyre advection reflects anomalous circulation acting on the mean temperature gradients. Maximum covariance analysis (MCA) reveals strong covariation between the anomalous horizontal gyre circulation and variations in the local wind stress curl, suggestive of a Sverdrup response. Results have implications for decadal predictability.

A recent state estimate covering the period 1992–2010 from the Estimating the Circulation and Climate of the Ocean (ECCO) project is utilized to quantify the upper-ocean heat budget in the North Atlantic on monthly to interannual time scales (seasonal cycle removed). Three novel techniques are introduced: 1) the heat budget is integrated over the maximum climatological mixed layer depth (integral denoted as H), which gives results that are relevant for explaining SST while avoiding strong contributions from vertical diffusion and entrainment; 2) advective convergences are separated into Ekman and geostrophic parts, a technique that is successful away from ocean boundaries; and 3) air–sea heat fluxes and Ekman advection are combined into one local forcing term. The central results of our analysis are as follows: 1) In the interior of subtropical gyre, local forcing explains the majority of H variance on all time scales resolved by the ECCO estimate. 2) In the Gulf Stream region, low-frequency H anomalies are forced by geostrophic convergences and damped by air–sea heat fluxes. 3) In the interior of the subpolar gyre, diffusion and bolus transports play a leading order role in H variability, and these transports are correlated with low-frequency variability in wintertime mixed layer depths.

The Atlantic Meridional Overturning Circulation (AMOC) (Box 1) is a system of ocean currents in the Atlantic that move warmer, upper waters northwards and cooler, deeper waters southwards. Accordingly, the AMOC is a major source of northward heat transport, accounting for 20-30% of total atmospheric and oceanic heat transport into the mid-latitudes 1 . The AMOC, therefore, has a key role in governing the climate of the North Atlantic region and beyond, influencing European air temperatures and precipitation, the frequency of Atlantic hurricanes and winter storms, spatial patterns of sea level and tropical monsoons 2,3 , and the global carbon budget 4 .The strength of the AMOC is typically 17 Sverdrups (Sv; 1 Sv = 10 6 m 3 s −1 ) 5 . However, both observations and models indicate that the AMOC exhibits substantial variability on daily to multi-decadal timescales. Coupled climate models suggest decadal variability can arise naturally due to internal interactions within the climate system [6][7][8] . The AMOC is also expected to respond to external forcing, including anthropogenic aerosols, volcanic eruptions and solar changes 9,10 , as well as anthropogenic greenhouse gas emissions 11 . Indeed, observations, reanalyses, models and proxies [12][13][14][15][16] indicate substantial contemporary decadal-scale changes in AMOC strength. The RAPID array at 26.5° N (refs 5,17 ), for example, revealed a statistically significant weakening from 2004 (refs 18,19 ), probably representing decadal variability rather than ongoing long-term weakening [20][21][22] . There are indications that the AMOC might be recovering in strength 12 .Despite evidence of decadal variability, many questions remain. For example, high-quality continuous observations, like the RAPID array, are short and sparse, making it difficult to assess longer-term AMOC variability and determine whether decadal changes are representative of those across the wider Atlantic. Moreover, there is uncertainty about the relative roles of internal variability and forced variability, owing to diverse AMOC variability 8,23 and externally forced AMOC trends 10,24 in models. Indeed, the AMOC might have already weakened over the twentieth century 25,26 , potentially implying that it is more sensitive to external forcing than previously thought. Understanding how and why the AMOC has changed on decadal timescales is thus crucial not only to understand the AMOC's role