An eddy-covariance flux measurement system was applied successfully to measure aerosol number dry deposition over open sea, ice floes, and over the leads in between the ice floes during the Arctic Ocean Expedition 1996. The aerosol number dry deposition velocity geometric mean was 1.9x10(-3) In s(-1) for the open sea. The dry deposition velocity to the ice floes averaged approximately 3 x 10(-4) ms(-1) with no significant difference between the melting summer ice and frozen, partly snow-covered ice in the early freeze-up. Over the leads the dry deposition velocity was 3.4 x 10(-4) and 9.1 x 10(-4) m s(-1) for the summer and the freeze-up period, respectively. For neutral and stable stratification the geometric averages were 6.5 x 10(-4) and 2.9 x 10(-4) m s(-1), respectively. The dry deposition changed mainly with friction velocity and particle size, as expected from theory, and with static stability, and we are able to recommend an empirical parameterization that works reasonable well over the Arctic Ocean pack ice. The resistance through the quasi-laminar sublayer dominated over the aerodynamic resistance and the ultrafine and Aitken mode particles dominated the dry deposition flux. The average turnover time for the submicometer aerosol in the boundary layer was only 4.6 days, but for cases when the ultrafine and Aitken modes dominated the aerosol size spectra, the turnover times were 1.3 and 3.5 days, respectively. The small turnover times were largely caused by the shallow boundary layer over the Arctic Ocean pack ice.