We show that simulated oceanic absorption of an atmospheric gas is very sensitive to the representation of a process that occurs beneath sea ice. As sea ice forms, salt is rejected, locally increasing surface sea-water density. This dense water can sink to the pycnocline at the base of the mixed-layer. Previous studies have not considered the impact of this subgrid-scale process on transient tracers in the ocean. To assess the potential importance of this process to the oceanic absorption of atmospheric gases, we performed two idealized simulations: a Control simulation in which salt rejected during sea-ice formation is placed in the model's 25 m thick surface layer; and a Test simulation in which salt rejected during sea-ice formation is distributed uniformly through the upper 160 m beneath the forming sea ice. Our treatment of rejected salt is highly idealized, and is intended to demonstrate the need for a physically-based parameterization of subgrid-scale convection for use in ocean general circulation models that takes into account the subgrid-scale heterogeneity of surface buoyancy forcing. Distributing rejected salt more deeply during periods of ice formation helps to maintain vertical density gradients, inhibiting grid-scale convection, especially in the Southern Ocean. This greatly diminishes simulated ocean uptake of CFC-11, and generally improves simulated CFC-11 and salinity fields. The modeled global ocean inventory of CFC-11 for year 1990 is about 30% lower, and modeled column inventories in the Southern Ocean are up to 90% lower, in our Test simulation relative to our Control simulation. We infer that a more detailed treatment of subgrid-scale processes occurring beneath sea ice may diminish simulated oceanic absorption of anthropogenic CO2, especially in the Southern Ocean. ¿ 1998 American Geophysical Union