||The net air-sea exchange of CO2 in the ocean area enclosed by the Fram Strait in the north and by the Greenland-Scotland ridge in the south is estimated on the basis of oceanic pCO(2) data from three different expeditions. The data comprise observed pCO(2) as well as pCO(2) calculated from data on dissolved inorganic carbon (C-t), total alkalinity (TA) and pH. Data on atmospheric content of CO2 are available from two of the expeditions and are supplemented with monthly mean values from ocean station ''M'' (66 degrees N, 2 degrees E). Wind speed drawn from a data set with a spatial resolution of 75 km and a temporal resolution of six hours are used in the calculation. The pCO(2) data are interpolated to the same spatial resolution as that of the wind speed data, before calculating the net flux. The objective of this work is to supplement earlier estimates of the net CO2 flux between atmosphere and ocean in the Nordic Seas, and to find a plausible shape of the seasonal variation of the net CO2 air-sea exchange. The calculations, per formed with the parameterization of the gas exchange coefficient according to Liss and Merlivat (1986), give a net uptake by the Nordic Seas of atmospheric CO2 amounting to approximately 2.4 moles C m(-2) year(-1). This is equivalent to 0.05 Gtonnes C year(-1), if the ice-free area is assumed to be 1.7*10(12) m(2). The estimate is in the lower range of the earlier estimates. Sensitivity tests of the calculated fluxes are performed as to the influence on the results by the choice of parameterization of the gas exchange coefficient, errors of observation in the pCO(2) data, choice of time scale of wind speed data and choice of interpolation method. The sensitivity tests indicate that the largest variations of the individual net fluxes arise from the uncertainty of the parameterization of the gas transfer and in the case of pCO(2) calculated from C-t, TA and pH, also from errors of observation. The results from three different equations of gas transfer (Liss and Merlivat, 1986; Tans et al., 1990; Wanninkhof, 1992) are compared. The parameterizations by Tans et al. and by Wanninkhof result in an increase of the estimated net fluxes of approximately 75 % compared to the net fluxes obtained by using the equations by Liss and Merlivat. Regarding the situation examined here, this uncertainty corresponds to an error of observation in oceanic pCO(2) of 3-11 %. The choice of time scale of the wind speed data and the choice of interpolation method do not seem to be of the same significance as far as the uncertainty of the result is concerned. Another uncertainty of the estimated yearly net flux arises from the fact that the pCO(2) data are very sparse in time, but the magnitude of this uncertainty is difficult to estimate.