A dynamic-thermodynamic sea ice model is developed and coupled with the ''21 wave 9 level'' Melbourne University general circulation model to simulate the seasonal cycle of the global sea ice distribution. We have run the coupled system and obtained a creditable seasonal simulation of global sea ice. When monthly averaged atmospheric data (taken from the mean of the coupled run) are used to force the sea ice model, the seasonal cycle of sea ice extent (to the outer ice edge) is quite similar to that simulated in the interactive run. However, the actual sea ice covered area (i.e., excluding leads) differs considerably between the two simulations. Sea ice is more compact in the monthly averaged forced run than in the interactive run throughout the year in both hemispheres. The sea ice thickness distribution also differs between the two runs. In general, the sea ice is more open and thicker in the seasonal ice zone of the two polar regions for the interactive coupled case than for the mean forcing. We have also run the model forced with daily atmospheric data and the simulated sea ice distribution differs significantly from both the interactive model and the monthly averaged forcing results. These differences highlight the dangers of undertaking studies with sea ice models forced with prescribed atmospheric conditions rather than using a fully interactive atmosphere-sea ice system. ¿ American Geophysical Union 1996