1. Scripps Institution of Oceanography, University of California, San Diego, La Jolla, California 92093–0244, USA
2. Present address: Cooperative Institute for Research in Environmental Sciences, University of Colorado at Boulder, Boulder, Colorado 80309–0216, USA.

Correspondence to: Britton B. Stephens^1,2 Correspondence and requests for materials should be addressed to B.B.S. at the University of Colorado at Boulder (e-mail: Email: britt@cmdl.noaa.gov ).

Ice-core measurements indicate that atmospheric CO₂ concentrations during glacial periods were consistently about 80 parts per million lower than during interglacial periods¹. Previous explanations for this observation^{2, 3, 4, 5, 6, 7, 8, 9} have typically had difficulty accounting for either the estimated glacial O₂ concentrations in the deep sea, ¹³C/¹²C ratios in Antarctic surface waters, or the depth of calcite saturation; also lacking is an explanation for the strong link between atmospheric CO₂ and Antarctic air temperature¹. There is growing evidence that the amount of deep water upwelling at low latitudes is significantly overestimated in most ocean general circulation models^{10, 11} and simpler box models previously used to investigate this problem. Here we use a box model with deep-water upwelling confined to south of 55 °S to investigate the glacial–interglacial linkages between Antarctic air temperature and atmospheric CO₂ variations. We suggest that low glacial atmospheric CO₂ levels might result from reduced deep-water ventilation associated with either year-round Antarctic sea-ice coverage, or wintertime coverage combined with ice-induced stratification during the summer. The model presented here reproduces 67 parts per million of the observed glacial–interglacial CO₂ difference, as a result of reduced air–sea gas exchange in the Antarctic region, and is generally consistent with the additional observational constraints.