Decadal Variability in Coupled Sea-Ice–Thermohaline Circulation Systems*

An interdecadal oscillation in a coupled ocean–ice system was identified in a previous study. This paper extends that study to further examine the stability of the oscillation and the sensitivity of its frequency to various parameters and forcing fields. Three models are used: (i) an analytical box model; (ii) a two-dimensional model for the ocean thermohaline circulation (THC) coupled to a thermodynamic ice model, as in the authors’ previous study; (iii) a three-dimensional ocean general circulation model (OGCM) coupled to a similar ice model. The box model is used to elucidate the essential feedbacks that give rise to this oscillation and to identify the most important parameters and processes that determine the period. Numerical experiments in the 2D THC–ice model show that the model stability is sensitive to the ocean–ice coupling coefficient, the eddy diffusivity, and the strength of the thermohaline-circulation feedback per unit surface-polar density perturbation. The coupled model becomes more stable toward low coupling, greater diffusion, and weaker THC feedback. The period of the oscillation is less sensitive to these parameters. Nonlinear effects in the sea-ice model become important in the higher ocean–ice coupling regime where the effective sea-ice damping associated with this nonlinearity stabilizes the model. Surface Newtonian damping is also tested. The 3D OGCM, which includes both wind stress and buoyancy forcings, is used to test this coupled ocean–ice mechanism in a more realistic model setting. This model generates an interdecadal oscillation whose characteristics and phase relations among the model variables are similar to the oscillation obtained in the 2D models. The major difference is that the oscillation frequency is considerably lower. This difference can be explained in terms of the analytical box model solution in which the period of the oscillation depends on the rate of anomalous density production by melting/cooling of sea ice per SST anomaly, times the rate of warming/cooling by anomalous THC heat advection per change in density anomaly. The 3D model has a smaller THC response to high-latitude density perturbations than in the 2D model, and anomalous velocities in the 3D case tend to follow the mean isotherms so the anomalous heat advection is reduced. This slows the ocean–ice feedback process, leading to the longer oscillation period.