The North Atlantic Ocean undergoes pronounced basin-wide, multi-decadal variations. The corresponding fluctuations in sea surface temperature (SST) have become known as the Atlantic Multidecadal Oscillation (AMO) or Atlantic multidecadal variability (AMV). AMV is receiving increasing attention for three key reasons: (1) it has been linked to climate impacts of major socio-economic importance, such as Sahel rainfall; (2) it may temporarily mask anthropogenic global warming not only in the North Atlantic Sector, but over the Northern Hemisphere (NH); and (3) it appears to be predictable on decadal timescales. This chapter provides an overview of current understanding of AMV, summarizing proposed mechanisms, our ability to simulate and predict it, as well as challenges for future research.

A multi-model analysis of Atlantic multidecadal variability is performed with the following aims: to investigate the similarities to observations; to assess the strength and relative importance of the different elements of the mechanism proposed by Delworth et al. (J Clim 6:1993–2011, 1993) (hereafter D93) among coupled general circulation models (CGCMs); and to relate model differences to mean systematic error. The analysis is performed with long control simulations from ten CGCMs, with lengths ranging between 500 and 3600 years. In most models the variations of sea surface temperature (SST) averaged over North Atlantic show considerable power on multidecadal time scales, but with different periodicity. The SST variations are largest in the mid-latitude region, consistent with the short instrumental record. Despite large differences in model configurations, we find quite some consistency among the models in terms of processes. In eight of the ten models the mid-latitude SST variations are significantly correlated with fluctuations in the Atlantic meridional overturning circulation (AMOC), suggesting a link to northward heat transport changes. Consistent with this link, the three models with the weakest AMOC have the largest cold SST bias in the North Atlantic. There is no linear relationship on decadal timescales between AMOC and North Atlantic Oscillation in the models. Analysis of the key elements of the D93 mechanisms revealed the following: Most models present strong evidence that high-latitude winter mixing precede AMOC changes. However, the regions of wintertime convection differ among models. In most models salinity-induced density anomalies in the convective region tend to lead AMOC, while temperature-induced density anomalies lead AMOC only in one model. However, analysis shows that salinity may play an overly important role in most models, because of cold temperature biases in their relevant convective regions. In most models subpolar gyre variations tend to lead AMOC changes, and this relation is strong in more than half of the models.

A long-standing problem in climate models is the large sea surface salinity (SSS) biases in the North Atlantic. In this study, we describe the influences of correcting these SSS biases on the circulation of the North Atlantic as well as on North Atlantic sector mean climate and decadal to multidecadal variability. We performed integrations of the Kiel Climate Model (KCM) with and without applying a freshwater flux correction over the North Atlantic. The quality of simulating the mean circulation of the North Atlantic Ocean, North Atlantic sector mean climate and decadal variability is greatly enhanced in the freshwater flux-corrected integration which, by definition, depicts relatively small North Atlantic SSS biases. In particular, a large reduction in the North Atlantic cold sea surface temperature bias is observed and a more realistic Atlantic Multidecadal Variability simulated. Improvements relative to the non-flux corrected integration also comprise a more realistic representation of deep convection sites, sea ice, gyre circulation and Atlantic Meridional Overturning Circulation. The results suggest that simulations of North Atlantic sector mean climate and decadal variability could strongly benefit from alleviating sea surface salinity biases in the North Atlantic, which may enhance the skill of decadal predictions in that region.