Understanding AMOC changes resulting from varying historical radiative forcings

A potential Atlantic Meridional Overturning Circulation (AMOC) slowdown, possibly caused by external forcings, is widely debated, and its historical drivers and future evolution remain uncertain. Here we disentangle the effects of greenhouse gases and anthropogenic aerosols on the AMOC and on other relevant processes in the high-latitude North Atlantic (NA) over 1850–2014. We analyze a multi-model ensemble of experiments from the Large Ensemble Single Forcing Model Intercomparison Project, specifically: hist-GHG (varying concentrations of greenhouse gases, other forcings constant) and hist-aer (same as hist-GHG, but for anthropogenic aerosols), and we compare these to the respective CMIP6 historical simulations (all forcings varying) and observational datasets.

Robust AMOC weakening under hist-GHG and strengthening under hist-aer is found across the respective multi-model ensembles with various accompanying changes, exhibiting a high degree of spatial antisymmetry. In both sets of experiments, the same causal pathway (yet with opposite sign) occurs. We describe the key role of subpolar upper-ocean salinity and connect its variations to changes in sea ice and air–sea heat fluxes. Our results indicate that the primitive radiative forcing directly impacts sea-ice mass, and thereby drives upper-ocean salinity variations, while accompanying changes in surface freshwater fluxes further modulate salinity. The resulting variations in salinity induce changes in upper-ocean density and stratification in the subpolar NA that, in turn, determine the simulated AMOC trends. We further discuss key mechanisms in play, including the positive AMOC–salinity and AMOC–evaporation feedbacks, describing the dominant processes of the causal pathway.

By offering insights onto the respective roles of external forcings in the context of climate change and by advancing our understanding of key NA ocean–atmosphere interactions, our results also highlight models limitations in the representation of coupled processes that are critical for reliable projections.