Does the Atlantic meridional overturning cell really have more than one stable steady state?

Abstract

The prevailing view regarding the oceanic meridional overturning cell (MOC) in the Atlantic is that, for a given North Atlantic freshwater flux, it has at least two stable states, one with a large surface (northward) mass flux and the other with a small flux. It has been argued that some abrupt paleoclimatic changes which occurred in the North Atlantic and the regions surrounding it might be related to a shift between these two distinctly different states. Here, we argue that, although the Atlantic MOC can indeed collapse due to a large freshwater flux, the actual ocean does not have multiple states for the same freshwater flux. The two state scenarios has its origin in the analytical box model of Stommel [1961. Thermohaline convection with two stable regimes of flow. Tellus 2, 244–230] and in a series of numerical models starting with that of Bryan [1986. High latitude salinity effects and interhemispheric thermohaline circulations. Nature 303, 301–304]. Using hybrid global analytical models involving both wind and density variations we demonstrate here that the application of Stommel's model to the North Atlantic yields multiple solutions because it considers the origin of the MOC upper limb to be a box whose export of water depends on its temperature and salinity which are not known in advance. When this origination box is replaced by a (observationally supported) Southern Ocean box whose surface water export depends solely on the wind, and when, together with this choice, the diapycnal diffusivities and eddy viscosities are taken to be as small as the usually observed values, the multi-solution scenario disappears and one gets only a single solution. Using the Uvic climate model, we re-confirm earlier results and argue that numerical models have multiple stable states and a resulting hysteresis because of the spuriously high eddy diffusivity that is typically used explicitly or implicitly. This is so because the diffusivity artificially introduces dense-to-light water conversion analogous to Stommel's origination box. Since we used a level model rather than a layered or isopynic model, the small vertical diffusivity limit still retains significant cross-isopycnal mixing due to the horizontal diffusivity, which is not supported by observations. Consequently, while our runs shows a tendency to no-hysteresis in the limit of small cross-isopycnal flow, we cannot actually reach that limit.