The Atlantic Meridional Overturning Circulation (AMOC) is important for the climate system as it redistributes heat, and sequesters heat and carbon from the atmosphere. The AMOC is believed to export freshwater across the southern boundary of the Atlantic at 34°S, potentially giving rise to a feedback loop that could make the AMOC vulnerable to collapse. Here we search for evidence of that so-called salt advection feedback in the internal variability from two climate models.
Our analysis shows that several links in the salt advection feedback are active, in particular, a strong control of North Atlantic density on the strength of the AMOC, and an impact of AMOC variability on the freshwater transport across 34° in the South Atlantic. However, we found that the AMOC-induced variability of the freshwater flux across 34°S is too weak to significantly affect the stratification in the Atlantic. We conclude that the salt advection feedback cannot be diagnosed from internal variability alone.
The stability of the Atlantic Meridional Overturning Circulation (AMOC) is associated with the so-called salt advection feedback, which links the strength of the AMOC to the AMOC-induced freshwater transport in the South Atlantic. If the AMOC exports freshwater, as observations suggest, this feedback may be positive, possibly indicating that the AMOC is vulnerable to collapse. Here we study whether we can identify elements of the salt advection feedback by analyzing the internal variability of the AMOC in climate models, namely CESM1 and ESM2M. We find that several elements of the feedback loop are indeed active, but AMOC-induced variability of the freshwater flux across 34°S is too weak to significantly affect the stratification in the Atlantic. We conclude that internal AMOC variability in climate models is not strong enough to activate the salt advection feedback. This means that it is unlikely that the internal variability of the AMOC alone could trigger a collapse if the AMOC were to be in a bi-stable regime. But we will not be able to draw any conclusions on AMOC stability in models by studying control integrations alone.