The only processes which could have decreased atmospheric CO2 during glacial climates without appreciably changing the carbon isotope distribution in the sea appear to be (1) dissolved calcium carbonate input to the ocean, e.g., coral reef buildup and erosion, (2) the ocean solubility pump, due to changes in surface temperature or air-sea exchange, or (3) decreased biological production of calcium carbonate. It is assumed here that one of these mechanisms caused part of the atmospheric CO2 changes recorded in the 200-kyr-long Vostok ice core. Two residual CO2 records are generated by scaling the Δ13C difference between planktonic and benthic foraminifera in marine sediment cores to ~82 ppm CO2 per 1% increase in ΔΔ13C and subtracting from this the measured CO2 concentration in the ice core. Both residual CO2 records exhibit two broad maximums between about 20-50 ka and 140-200 ka, indicating that during these times, about 40 ppm of the CO2 decrease from interglacial levels cannot be explained by the interaction of the ocean's biological and vertical mixing cycles.

The shape of the residual CO2 curve is similar qualtitatively to the variation of calcium carbonate in central equatorial Pacific sediments during this time period, which would imply that changes in dissolved carbonate input to the ocean contributed the added component of CO2 change. However, recent models of atmospheric CO2 change in response to changing alkaline inptu to the ocean exhibit about a 25 to 35 ppm decrease per 1013 mol yr-1 increase in dissolved CaCO3 input. If compensation for changing input is occurring mostly within an area of about 40¿106 km2 below the lysocline in the Indo-Pacific, the change in carbonate accumulation rate corresponding to a -40 ppm CO2 change would be a minimum of 3 mg cm-2 yr-1. This can be compared to glacial increases of 0.5 to 1.0 cm-2 yr-1 during the last 200 kyr in central equatorial Pacific sediments. Thus, the added glacial accumulation of carbonate does not seem to match quantitatively with the 40 ppm amplitude of the residual CO2 signal, leading one to suspect that solubility plays a greater role than expected on the basis of a 2¿C cooler surface ocean. ¿ American Geophysical Union 1995.