An eddy-covariance flux system was successfully applied over open sea, leads and ice floes during the Arctic Ocean Expedition in July-August 1996. Wind-driven upward aerosol number fluxes were observed over open sea and leads in the pack ice. These particles must originate from droplets ejected into the air at the bursting of small air bubbles at the water surface. The source flux F (in 10(6) m(-2)s(-1)) had a strong dependency on wind speed, log(F) = 0.20 (U) over bar - 1.71 and 0.11 (U) over bar -1.93, over the open sea and leads, respectively (where U is the local wind speed at about 10 in height). Over the open sea the wind-driven aerosol source flux consisted of a film drop mode centered at similar to100 nm diameter and a jet drop mode centered at similar to1 mum diameter. Over the leads in the pack ice, a jet drop mode at similar to2 mum diameter dominated. The jet drop mode consisted of sea-salt, but oxalate indicated an organic contribution, and bacterias and other biogenic particles were identified by single particle analysis. Particles with diameters less than similar to100 nm appear to have contributed to the flux, but their chemical composition is unknown. Whitecaps were probably the bubble source at open sea and on the leads at high wind speed, but a different bubble source is needed in the leads owing to their small fetch. Melting of ice in the leads is probably the best candidate. The flux over the open sea was of such a magnitude that it could give a significant contribution to the condensation nuclei (CCN) population. Although the flux from the leads were roughly an order of magnitude smaller and the leads cover only a small fraction of the pack ice, the local source may till be important for the CCN population in Arctic fogs. The primary marine aerosol source will increase both with increased wind speed and with decreased ice fraction and extent. The local CCN production may therefore increase and influence cloud or fog albedo and lifetime in response to greenhouse warming in the Arctic Ocean region.