Abstract:
We analyze two versions of the CSSM (CSSM3-4) with widely different temporal evolution of the minimum sea ice extent in the Arctic Ocean. In CCSM3, abrupt reductions in summer Arctic sea ice are present in all ensemble members leading to a seasonally ice-free Arctic as early as 2040 (Holland et al, 2006). In all ensemble members, the abrupt sea ice declines are preceded by a pulse of heat of Atlantic origin into the Arctic Ocean, causing an increase in open water area, and absorbed solar radiation at the surface triggering the positive ice-albedo feedback. In CCSM4, no abrupt reductions in summer Arctic sea ice are simulated; instead, the model simulates a slow decline in sea ice leading a seasonally ice-free Arctic by the end of the century. A key difference between the two versions of the CCSM is the partitioning of ocean heat of Atlantic origin between Fram Strait and the Barents Sea gate. In CCSM4, ocean heat fluxes increases slowly but steadily in all gates of the Arctic. When North Atlantic Drift waters enter the Arctic through the Fram Strait, they sink at depth and interact little with the sea ice cover while circulating within the Arctic Ocean. In CCSM3, the ocean heat flux through Fram Strait decreases slightly with time in the 21^st century and the pulses of ocean heat enter the Arctic entirely through the Barents Sea gate.  This ocean heat interacts with the sea ice cover on the northern end of the Barents Sea before merging with the Fram Strait branch at depth and leads to a negative ice thickness anomaly that is advected in part in the Fram Strait and in part in the Beaufort Gyre. The thinner (weaker) central Arctic pack ice in turns allows for more sea ice divergence along the Eurasian coastline and for a rapid transition to a seasonally ice free Arctic. The creation of new open water also triggers the ice-albedo and cloud-ice albedo feedback, with an increase in liquid water cloud and associated increase in downwelling longwave radiation.