Internal Atlantic Multidecadal Variability mechanism at two model resolutions

The Atlantic Multidecadal Variability (AMV) is a key factor in modulating climate change and its impacts around the world. Therefore, understanding of its physical mechanism will be crucial to achieving predictability on decadal timescales. However, details of the mechanism are not fully understood. This is evident in the wide range of simulated AMV timescales and spatial patterns exhibited by climate models in both pre-industrial and historical simulations.

In this study, we assess the impact of model resolution on the internal AMV mechanism by taking advantage of the close physical similarities between the medium- and low-resolution versions of the HadGEM3 models. Here, we present results from analysing the N96ORCA1 (~135km atmosphere, 1° ocean) and N216ORCA025 (~60km, 0.25°) pre-industrial simulations.

At both resolutions, we found that the internal AMV has a timescale of 70-100 years, comparable to the observed record. The processes driving decadal SST variability varies by latitude. Ocean heat transport changes associated with the AMOC drive subpolar variability, while surface fluxes associated with cloud and wind changes are more important in the subtropics. The AMOC strengthening is induced by density forcing from two sources. First, a Labrador Sea surface cooling driven by low-frequency positive NAO leads the AMOC by 5 years. Second, a source of anomalously saline Arctic water flowing into the subpolar North Atlantic also leads the AMOC by 5 years. Interestingly, the two resolutions disagree on the relative importance of these AMOC drivers. In the lower resolution model, the Arctic contribution is more important. However, the NAO dominates in the medium resolution model, and decadal NAO variability is more strongly associated with the AMV. Differences between the models are likely due to mean state differences including the strength and position of ocean currents such as the Gulf Stream, and their impacts on upper ocean properties.