AMOC Variability in Climate Models and Its Dependence on the Mean State

Understanding internal variability of the climate system is critical when isolating internal and anthropogenically forced signals. Here, we investigate the modes of Atlantic Meridional Overturning Circulation (AMOC) variability using perturbation experiments with the Institut Pierre-Simon Laplace's (IPSL) coupled model and compare them to Coupled Model Intercomparison Project Phase 6 (CMIP6) pre-industrial control simulations. We identify two characteristic modes of variability-decadal-to-multidecadal (DMDvar) and centennial (CENvar). The former is driven largely by temperature anomalies in the subpolar North Atlantic, while the latter is driven by salinity in the western subpolar North Atlantic. The amplitude of each mode scales linearly with the mean AMOC strength in the IPSL experiments. The DMDvar amplitude correlates well with the AMOC mean strength across CMIP6 models, while the CENvar mode does not. These findings suggest that the strength of DMDvar depends robustly on the North Atlantic mean state, while the CENvar mode may be model-dependent.