Using satellite ice drift and ice concentration data, observed sea surface temperature and sea level pressure data we have developed a statistical sea ice model (ISMO) and computed the vectors of ice velocity for the period 1899–2000 and ice concentration for the period 1966–2000. On the basis of the ISMO results, we estimate the sea ice fluxes through the main straits of the Arctic Ocean and simulate trajectories of the ice drift from the locations of potential sources of contaminants. The ISMO is verified by comparing with the observational results of the sea ice fluxes and extents, and with the historical drift data from the International Arctic Buoy Programme. For example, the relation between the observed (not satellite data) and calculated April ice extent in the Nordic seas for the period 1966–1998 ( R=0.57) has good agreement, and the pathways of simulated and observed trajectories are quite close; origins derived from calculated backward trajectories are generally consistent with actual buoy drifts. We also compare the modelling sea ice flux through Fram Strait with the estimation using ice motion from satellite passive microwave data for the period 1978–1996, finding a significant correlation ( R=0.71) between the calculated and observed time series of the monthly area flux. We estimate the average sea ice export from the Arctic Ocean through Fram Strait and the strait between Svalbard and Franz Josef Land over the 35-year simulated record (1966–2000) as 639,000 and 70,000 km 2 year −1, respectively. Estimation of the ice flux from the Kara Sea to the Barents Sea through the strait between Franz Josef Land and Novaya Zemlya gives 104,000 km 2 year −1 for the same period. The average sea ice import from the Kara, Laptev and East Siberian seas to the Arctic Ocean over the same period is about 332,000 km 2 year −1. The average sea ice export from the Arctic Ocean through the strait between Wrangel Island and Cape Barrow is 7000 km 2. The difference between the annual mean ice fluxes estimated by ISMO and most previous estimations for Fram Strait is 0.1–30%. We investigate the seasonal and interannual variability of the sea ice flux through the main straits of the Arctic Ocean, and show strong seasonal variability and significant variations from year to year (e.g., in Fram Strait from a minimum of 435,000 km 2 year −1 in 1985 to a maximum of 940,000 km 2 year −1 in 1995). Calculation of forward trajectories shows that sea ice from most potential sources of contaminant can reach the open Polar Basin and Fram Strait. Contaminated sea ice from potential sources in the Kara and Laptev seas can reach Fram Strait within 2–4 years, and from the East Siberian, Chukchi and Beaufort seas within 6–11 years. Based on simulated trajectories from different potential sources in the Arctic Ocean, we conclude that the most important regions for monitoring contaminants are Fram Strait and the Barents Sea Opening.