Interdecadal variability and oceanic thermohaline adjustment

Abstract

Changes in the strength of the thermohaline overturning circulation are associated, by geostrophy, with changes in the east‐west pressure difference across an ocean basin. The tropical‐polar density contrast and the east‐west pressure difference are connected by an adjustment process. In flat‐bottomed ocean models the adjustment is associated with viscous, baroclinic Kelvin wave propagation. Weak‐high latitude stratification leads to the adjustment having an interdecadal timescale. We reexamine model interdecadal oscillations in the context of the adjustment process, for both constant flux and mixed surface boundary conditions. Under constant surface flux, interdecadal oscillations are associated with the passage of a viscous Kelvin wave around the model domain. We carry out experiments suppressing wave propagation along each of the model boundaries. Suppressing wave propagation along either the tropical or eastern boundary does not eliminate the oscillation, but increases both its period and amplitude. Suppressing wave propagation along either the polar or the western boundary eliminates the oscillation. Our results suggest the oscillations can be self‐sustained by perturbations to the western boundary current arising from the southward boundary wave propagation. Mixed boundary condition oscillations are characterized by the eastward, cross‐basin movement of salinity‐dominated density anomalies, and the westward return of these anomalies along the northern boundary. We suggest the latter is associated with viscous Kelvin wave propagation. Under both types of boundary conditions, the strength of the thermohaline overturning and the tropical‐polar density contrast vary out of phase. We show how the phase relationship is related to the boundary wave propagation. Box models and zonally averaged models assume that the east‐west and north‐south pressure gradients vary in phase and are proportional to one another. We suggest this assumption is valid only on timescales long compared to the adjustment timescale. The importance of boundary regions indicates an urgent need to examine the robustness of interdecadal variability in models as the resolution is increased, and as the representation of the coastal, shelf/slope wave guide is improved.