A method of determining the large-scale sea ice drift using 85.5 GHz Special Sensor Microwave Imager data are presented. A cross-correlation method is applied to sequential images of gridded data covering the entire Arctic. Individual correlation results are validated with ice velocities derived from buoy data. The satellite-derived mean drift values and the variabilities of the ice drift correspond closely with the buoy data. Similarly, time series of buoy data and associated satellite data are in good agreement even over large time periods. An example of a satellite-retrieved 3 day mean drift field demonstrates the potential of the method for providing large-scale ice circulation patterns. Mean drift fields of the winter periods 1987--1988 and 1992--1993 indicate a considerable interannual variability of the sea ice drift pattern in the Arctic Ocean. The Arctic region is divided into seven larger areas, and the area flux between these regions has been derived. The Kara Sea and the Laptev Sea show the largest area ice export with 0.02 and 0.015 km2 s-1, respectively. The central Arctic export through Fram Strait amounts to 0.12 Sv during the winter of 1992--1993 with a maximum of 0.15 Sv in January. ¿ 2000 American Geophysical Union