Climate variability in a low‐order coupled atmosphere‐ocean model

ABSTRACT The dynamical behavior of the climate system is investigated through the use of a low‐order coupled atmosphere‐ocean general circulation model. The goal is to gain some qualitative understanding of how non–linear interactions between the individual system components may affect the climate. Both the atmosphere and ocean models are fully dynamic: the former is defined by 3 ordinary differential equations derived from a truncated Fourier series expansion of the mean and perturbation components of the quasi‐geostrophic potential vorticity equation, while the latter is specified by 6 ordinary differential equations representing the time‐dependent variations of ocean temperature and salinity in a 3‐box model of the North Atlantic. Despite the existence of 2 basic equilibrium ocean model responses to perpetual atmospheric conditions, equilibrium states are never attained in the coupled system within 10000 years of integration; the deep ocean flow continually adjusts to the atmospheric regime changes associated with particular ocean circulations, which leads to new circulations and new atmospheric regimes. Low‐frequency quasi‐periodic oscillations about a single state of the thermohaline circulation result from an advective‐diffusive process, modulated by the correlation of the atmospheric behavior with the phase of the ocean cycle. The climate is strongly effected by interactions with the ocean, leading to distinct atmospheric patterns for different phases in the oscillations, and a conversion of some of the high‐frequency atmospheric signal to lower frequencies. This conversion also results in a measurable ocean response at high frequencies. Furthermore, owing to the richness of the atmospheric response to small modifications in the meridional and zonal gradients in diabatic heating, even modest adjustments in the ocean circulation resulting from interactions with the high‐frequency atmospheric component can also lead to climate change over relatively short time periods. The results of the model are applied to recent deductions of climate variability in the North Atlantic, obtained from Greenland ice‐cores.