Note on the CO 2 air–sea gas exchange at high temperatures

Abstract

The temperature dependency of ocean–atmosphere gas transfer velocities is commonly estimated in terms of Schmidt numbers, i.e. the ratio of kinematic viscosity to diffusivity. In numerical models least square regressions are used to fit the limited number of experimentally derived Schmidt numbers to a function of temperature. For CO 2 a well established fit can be found in the literature. This fit constitutes an integral part in standardized carbon cycle simulation projects (e.g. C4MIP, OC4MIP, Friedlingstein et al., 2006). However, the fit is valid only in the range where diffusivity measurements exist, i.e., from 0 to about 30 °C. In many climate warming simulations like e.g. the MPI contribution to the fourth Intergovernmental Panel on Climate Change Assessment Report (IPCC AR 4), sea surface temperatures largely exceed the validated range and approach or even reach the range, where the standard fits leave the physically meaningful range. Thus, this paper underlines the demand for new measurements of seawater diffusivities for CO 2 and other trace gases especially for the temperature range >30 °C. In this paper we provide improved fits for the temperature dependence of the Schmidt number. For carbon dioxide our fit is compared to the established fit under identical climate change simulations carried out with the 3D-carbon cycle model HAMOCC. We find that in many tropical and subtropical high temperature regions the established fit leads to unrealistically short adaption times of the surface water pCO 2 to altered atmospheric pCO 2. In regions where the local oceanic pCO 2 is not primarily controlled by the atmospheric boundary pCO 2 but by other processes such as biological activity, the atmosphere ocean pCO 2 gradient is clearly underestimated when using the established fit. The effect on global oceanic carbon uptake in a greenhouse world is rather small and the potential climate feedback introduced by this bias seems to be negligible. However, the bias will clearly gain in significance the more regions warm up to over 30 °C. On a regional scale, especially in coastal regions at low latitudes, the effect is not negligible and a different steady state is approached.