In 1983 the project "Travaux d'Oceanographie Spatiale: Capteurs Actifs dans l'Atlantique Nord-Est" (TOSCANE) was initiated in view of calibration, validation, and use of satelliteborne scatterometer and altimeter measurements in terms of wind and wave data, with special emphasis on the ERS-1 satellite to be launched in 1991 by the European Space Agency. The wind measurements from an array of buoys and ship and land stations, deployed during the last experiment within the program, TOSCANE-2, are described, and a method is given to evaluate their accuracy. When comparing wind data from two measuring stations at sea, the averaging time has to be selected according to the separation distance between the two locations to filter out the space-induced wind variability. For buoy and ship comparisons, for distances less than 3 km, a 10-min averaging time was found to be a lower limit. For pairs of buoys separated by 25 or 35 km, the standard deviation of differences decreases with averaging time, in the same manner for speed and direction: by about 40% when passing from 10 min to 6 h averaging (from 1.3 to 0.8 m s-1 for speed, and from 14-degrees to 9-degrees for direction) with a plateau value (50%) after 12 h. Data from the two independent wind systems on board N/O Le Noroit show a high consistency: mean values of differences for direction and speed are 5.9-degrees and 0.29 m s-1, with standard deviations of 4.5-degrees and 0.40 m s-1, the standard deviation of speed differences increasing (0.10 to 0.49 m s-1) over the 1-30 m s-1 speed range; 99.9% of data are within +/- 2 m s-1. When calibrating anemometers with coefficients from wind tunnel tests, differences between buoy speeds are large, as much as 1.4 m s-1 +/- 0.8 m s-1, resulting from, among other causes, the inability of sensors to correctly filter high-frequency fluctuations of wind speed. The buoys were then tuned independently, relative to ship data from dedicated runs, which allowed a reduction of differences to maximum values of 0.5 m s-1 +/- 0.8 m s-1 and 4-degrees +/- 9-degrees. These results are compatible with an assessment of the foreseen 2 m s-1 and 20-degrees accuracy of the ERS-1 scatterometer wind data. In contrast, wind data from a 10-m mast on a flat island are shown not to be accurately representative of wind at sea, even for onshore wind.