Millennial-scale oscillations between sea ice and convective deep water formation

Abstract

During the last ice age there were several quasiperiodic abrupt warming events. The climatic effects of the so-called Dansgaard-Oeschger (D-O) events were felt globally, although the North Atlantic experienced the largest and most abrupt temperature anomalies. Similar but weaker oscillations also took place during the interglacial period. This paper proposes an auto-oscillatory mechanism between sea ice and convective deep water formation in the North Atlantic as the source of the persistent cycles. A simple dynamical model is constructed by coupling and slightly modifying two existing models of ocean circulation and sea ice. The model exhibits mixed mode oscillations, consisting of decadal-scale small-amplitude oscillations and a large-amplitude relaxation fluctuation. The decadal oscillations occur due to the insulating effect of sea ice and leads to periodic ventilation of heat from the polar ocean. Gradually, an instability builds up in the polar column and results in an abrupt initiation of convection and polar warming. The unstable convective state relaxes back to the small-amplitude oscillations from where the process repeats in a self-sustained manner. Freshwater pulses mimicking Heinrich events cause the oscillations to be grouped into packets of progressively weaker fluctuations, as observed in proxy records. Modulation of this stable oscillation mechanism by freshwater and insolation variations could account for the distribution and pacing of D-O and Bond events. Physical aspects of the system such as sea ice extent and oceanic advective flow rates could determine the characteristic 1500 year time scale of D-O events. The model results with respect to the structure of the water column in the Nordic seas during stadial and interstadial phases are in agreement with paleoproxy observations. Key Points Sea ice and convection can interact to produce millennial-scale oscillations A simple dynamical model is used to demonstrate this hypothesis Oscillations could explain the Dansgaard-Oeschger and Bond events.