ALBERT

Description: The air-sea gas transfer velocity (K-660) is typically assessed as a function of the 10-m neutral wind speed (U-10n), but there remains substantial uncertainty in this relationship. Here K-660 of CO2 derived with the eddy covariance (EC) technique from eight datasets (11 research cruises) are reevaluated with consistent consideration of solubility and Schmidt number and inclusion of the ocean cool skin effect. K-660 shows an approximately linear dependence with the friction velocity (u*) in moderate winds, with an overall relative standard deviation (relative standard error) of about 20% (7%). The largest relative uncertainty in K-660 occurs at low wind speeds, while the largest absolute uncertainty in K-660 occurs at high wind speeds. There is an apparent regional variation in the steepness of the K-660-u* relationships: North Atlantic 〉= Southern Ocean 〉 other regions (Arctic, Tropics). Accounting for sea state helps to collapse some of this regional variability in K-660 using the wave Reynolds number in very large seas and the mean squared slope of the waves in small to moderate seas. The grand average of EC-derived K-660 ( - 1.47 + 76.67 u * + 20.48 u *(2) o r 0.36 + 1.203 U-10n + 0.167 U (2)(10n) ) is similar at moderate to high winds to widely used dual tracer-based K-660 parametrization, but consistently exceeds the dual tracer estimate in low winds, possibly in part due to the chemical enhancement in air-sea CO2 exchange. Combining the grand average of EC-derived K-660 with the global distribution of wind speed yields a global average transfer velocity that is comparable with the global radiocarbon (C-14) disequilibrium, but is similar to 20% higher than what is implied by dual tracer parametrizations. This analysis suggests that CO2 fluxes computed using a U-10n (2) dependence with zero intercept (e.g., dual tracer) are likely underestimated at relatively low wind speeds.