Using climate forcings runs to attribute the decadal variability and predictability of the AMOC

The Atlantic Meridional Overturning Circulation (AMOC) is a key factor in global and North Atlantic decadal variability and decadal climate predictability. In order to understand and trust decadal predictions of the AMOC, we need to understand the origins of its variability and attribute the variability to key external forcings. Robson et al (2022) attributed most of the decadal variability of the AMOC between 1850-1985 to anthropogenic aerosols. The mechanism was a cooling over the American continent which led to colder winds over the North Atlantic and larger turbulent heat loss increasing dense water formation and the strength of the AMOC. However, the models with the most advanced aerosol schemes produced a simulation of the North Atlantic that did not agree with observations, leaving the question of the forcings important to the real-world AMOC open.

 

We take advantage of large ensembles of historical single- and multi-forcing runs covering 1850-2030 created as part of the Large Ensemble Single Forcing Model Intercomparison Project (LESFMIP) under CMIP6. These also include "all-but" runs, where all but one forcing is used. The single-forcing runs can take the model far away from historical climate. For example, a run with only anthropogenic aerosol changes becomes much colder than the present-day climate, with impacts on sea-ice cover and dense water formation sites. Attribution studies rely on linearity, all the individual single-forcing runs added up should give the same result as one run using all the forcings at once (as in the case of the historical all-forcings run). It follows that the historical run with the all-but-aerosol run removed, should be the same as the single-forcing aerosol-only run. However, given what we know about the changes to dense water formation sites in the latter run, it is not surprising that this is not the case. It is not even true when considering the volcanic forcing, which only alters the model climate for a few years after an eruption.

 

In this work, we investigate the non-linearities of the impact of climate forcings on the AMOC. The deviations from linearity help us understand how useful single-forcing runs are for attribution work and how different forcings combine to change the response of the AMOC. Hopefully, this can bring us one step closer to understanding how useful our decadal predictions of the AMOC are and which forcings are important to the real-world AMOC.