Understanding AMOC Stability: The North Atlantic Hosing Model Intercomparison Project
The North Atlantic Hosing Model Intercomparison Project (NAHosMIP) aims to understand the sensitivity of the AMOC in current GCMs to hosing. Using the experimental setup of Jackson and Wood (2018b), in which additional freshwater input is applied to a larger area of the subpolar North Atlantic, we designed a set of experiments to understand whether different GCMs show a similar AMOC hysteresis, with a particular focus on models participating in the current Coupled Model Intercomparison Project (CMIP6). Using a large, idealized hosing allows us to understand the model sensitivities and how they differ, and from analyzing the similarities and differences between the model responses, we may be able to understand what controls this AMOC response and how the real world may behave. Although the first set of experiments is very idealized, we also include a set of similar experiments which are less idealized in that they apply the freshwater around the coasts of Greenland only and use a more realistic (though still large) amount of freshwater input. These experiments help us to understand how to apply our understanding of the AMOC sensitivity from the idealized experiments to a less-idealized scenario.
Initial results are shown from eight climate models participating in the Sixth Coupled Model Intercomparison Project (CMIP6). The AMOC weakens in all models as a result of the freshening, but once the freshening ceases, the AMOC recovers in half of the models, and in the other half it stays in a weakened state. The difference in model behavior cannot be explained by the ocean model resolution or type nor by details of subgrid-scale parameterizations. Likewise, it cannot be explained by previously proposed properties of the mean climate state such as the strength of the salinity advection feedback. Instead, the AMOC recovery is determined by the climate state reached when hosing stops, with those experiments where the AMOC is weakest not experiencing a recovery. Future studies will examine the mechanisms involved in the AMOC recovery to improve our understanding of the important feedbacks involved, as well as to examine the impacts of a sustained AMOC weakening. We also hope to use this protocol for future experiments with higher-resolution climate models, which improve the resolution of eddies and boundary currents in the subpolar North Atlantic. Understanding how the models' responses to freshwater forcing evolve in the presence of warming is also a future research direction.
We have presented the experimental protocol for the NAHosMIP project, which aims to understand the sensitivity of the AMOC to additional freshwater in the North Atlantic. We show that about half the CMIP6 models which run this protocol find states where the AMOC does not recover after hosing of 0.3 Sv. The difference in model behavior cannot be explained by the ocean model resolution or type, by details of subgrid-scale parameterizations, or by aspects of the mean climate state such as the strength of the salinity advection feedback, the location or depth of deep convection, or the position of the intergyre boundary. Instead, the AMOC behavior appears to be related to the state the model reaches after hosing finishes; specifically, those experiments where the AMOC has reached the weakest states, where March mixed-layer depths are the shallowest, and where the eastern subpolar gyre and Nordic seas are the coldest and freshest with the greatest sea ice extent are those where the AMOC subsequently does not recover. An important question for further analysis is why different models reach different states during hosing.