Understanding the uncertainty in simulated AMOC changes to historical greenhouse gas emissions in CMIP6

There is substantial model uncertainty in how the AMOC evolves in CMIP6 historical simulations. A significant part of this uncertainty has been related to the substantial uncertainty in the strength of the historical Anthropogenic Aerosol forcing. However, there is also significant uncertainty in how models simulate the AMOC response to historical greenhouse gas forcing, which is not well understood. Therefore, this raises the question, how sensitive is the real-world AMOC to historical greenhouse gas forcing? Here, we use simulations from the Large Ensemble Single Forcing Model Intercomparison Project (LESFMIP) to isolate the historically forced signal, and to understand the spread in AMOC response between models. By 2014, AMOC declined significantly in hist-ghg simulations as expected, but there is a very large spread in AMOC decline that is approximately equivalent to the uncertainty due to aerosol forcing. We find that the decline appears to be strongly related to changes in turbulent heat loss in the subpolar North Atlantic, which is itself related to the change in air-sea temperature and humidity contrasts (i.e., it is thermodynamically driven). The spread in hist-ghg simulations is also consistent with the spread in abrupt 4xCO2 simulations, and further analysis of those simulations supports a causal relationship between the spread in the initial forced changes in the air-sea temperature contrasts and the resultant AMOC decline.