Mechanisms of interdecadal climate variability and the role of ocean–atmosphere coupling

Abstract

Climate variability and mid-latitude mechanisms of ocean–atmosphere interactions are investigated with coupled and uncoupled integrations of a three-dimensional ocean–atmosphere–land–ice climate model of intermediate complexity. We focus on the decadal and interdecadal variability of the system and give a statistical and dynamical description of its oceanic and atmospheric signatures. In our coupled control integration, an oceanic oscillation of a period of around 20 years is found to be associated with variability of the meridional overturning circulation and is manifested by surface anomalies of temperature and salinity. On such timescales the oceanic oscillation is able to imprint itself on the atmosphere, which then covaries with the ocean at the oscillation period. The essentially slaved atmospheric pattern helps maintain the oceanic oscillation by providing large-scale anomalous heat fluxes, so catalyzing the oscillation. That is to say, because the atmosphere covaries with the ocean the damping felt by the ocean is less than what would be felt with a fixed atmosphere, so broadening the parameter regime over which such variability occurs. In addition to the presence of an atmosphere, the period and amplitude of the oscillation are found to be influenced both by the oceanic vertical diffusivity κ v , by geometrical factors, and by the presence of stochastic heat fluxes. In general, oscillations occur most readily for large values of κ v , when the mean state of the ocean is characterized by a strong meridional overturning circulation. If κ v is sufficiently strong, the ocean will oscillate even in the absence of a dynamical atmosphere. However, for more realistic values of κ v , the presence of an interacting atmosphere is required for significant oscillations. If the ocean is forced by imposed stochastic heat fluxes, instead of a fully interacting atmosphere, then decadal-scale oscillations can be produced suggestive of a damped oscillator. However, the parameter range over which oscillations occur is smaller than when the ocean is coupled to full atmosphere. More generically, the ability of comprehensive coupled ocean–atmosphere models to produce multi-decadal variability, realistic or otherwise, will depend on the oceanic mean state, and so on the diapycnal diffusivity of the modelled ocean, as well as on the ability of the atmosphere to reduce the damping felt by the ocean and so on the atmosphere’s ability to respond to persistent sea-surface temperature anomalies.