Optimal excitation of AMOC decadal variability: Links to the subpolar ocean

•We computed optimal initial perturbations of SST and SSS for the AMOC in an OGCM.•Subpolar region surface density changes are precursors by 8–9 yr of AMOC changes.•Transient changes of the AMOC are linked to the stimulation of the deep ocean.•Upper ocean density gradients are critical to the AMOC transient changes.•The AMOC transient changes give way to a 24 yr-period oscillation.

This study describes the excitation of variability of the Atlantic Meridional Overturning Circulation (AMOC) by optimal perturbations in surface temperature and salinity. Our approach is based on a generalized stability analysis within a realistic ocean general circulation model, which extends the conventional linear stability analysis to transient growth. Unlike methods based on singular value decomposition, our analysis invokes an optimization procedure using Lagrangian multipliers, which is a more general approach allowing us to impose relevant constraints on the perturbations and use linear measures of the AMOC (meridional volume and heat transports).We find that the structure of the optimal perturbations is characterized by anomalies in surface temperature or salinity centered in the subpolar regions of the North Atlantic off the east coasts of Greenland and Canada, south of the Denmark Strait. The maximum impact of such perturbations on the AMOC is reached after 7–9 yr. This is a robust result independent of the perturbations type, the optimization measures, the model surface boundary conditions, or other constraints. The transient growth involves the following mechanism: after the initial (positive) surface density perturbation reaches the deep ocean, it generates a cyclonic geostrophic flow that extracts a zonally-varying temperature anomaly from the mean temperature field in the upper ocean. In turn, the anomalous zonal temperature gradient induces, by thermal wind balance, a northward flow in the upper ocean and a southward flow in the deep ocean, thus strengthening the AMOC. Subsequently, the transient growth gives way to a decaying oscillation corresponding to a damped oceanic eigenmode with a period of about 24 yr. This mode is controlled by westward-propagating large-scale “thermal” Rossby waves, modifying the density field in the North Atlantic and hence the AMOC. Simple estimates show that realistic changes in salinity or temperature in the upper ocean (such as those due to the Great Salinity Anomaly) can induce AMOC variations of several Sverdups via this mechanism, or 10–20% of the mean overturning. An idealized model is formulated to investigate the transient growth and highlight the role of mean convection in communicating surface density anomalies to the deep ocean.