Using 3 months of wind measurements made by the Seasat scatterometer and the relationship between the wind speed and the transfer velocity measured by Liss and Nerlivat <1986>, we deduce the exchange coefficient between air and sea for carbon dioxide on a worldwide scale. Global maps of the exchange coefficient, with a 10¿¿10¿ resolution, are made, neglecting the temperature dependence due to the Schmidt number and solubility temperature variations, as we show that the variations of these two parameters nearly compensate in the range of variation of sea surface temperature in the global ocean. The influence of the ice-covered area of the ocean, where no gas exchange occurs, is taken into account. A mean value of the global average exchange coefficient is derived, equal to 3.63¿10-2 mol m-2 yr-1 μatm-1. Notwithstanding that this number is based on a 3-month period of wind measurements, it is consistent with previous estimates, such as the determination made from radioactive tracers (i.e., carbon 14, radon), the values computed using climatological charts, and estimates derived from large-scale experimental wind data. The influence of our method of averaging on the results is studied using an example in the Gulf of Guinea. Using in situ wind and carbon dioxide measurements, we must stress the importance of seasonal variations in the determination of the net flux on an annual basis. Our results, combined with the global map of the partial pressure gradient of carbon dioxide established by Broecker et al. <1986>, give an estimation of the net flux between air and sea of 1.17¿1015g C yr-1, lower than usually expected (about 2 to 2.5¿1015g C yr-1) from the global budget of the carbon dioxide cycle . Considering the limitations of the data (only 3 months of wind speed measurements combined with low spatial resolution annual mean ΔPCO2 maps), this discrepancy is not surprising, especially as the result is very dependent on the ΔPCO2 used.

Moreover, we study the consequences of assuming a spatially constant exchange coefficient for estimating a net flux. Such an assumption can result in a very large error. ¿ American Geophysical Union 1988