A reasonable initial state of ice concentration is essential for accurate short-term forecasts of sea ice using ice-ocean coupled models.In this study,sea ice concentration data are assimilated into an operational ice forecast system based on a combined optimal interpolation and nudging scheme.The scheme produces a modeled sea ice concentration at every time step,based on the difference between observational and forecast data and on the ratio of observational error to modeled error.The impact and the effectiveness of data assimilation are investigated.Significant improvements to predictions of sea ice extent were obtained through the assimilation of ice concentration,and minor improvements through the adjustment of the upper ocean properties.The assimilation of ice thickness data did not significantly improve predictions.Forecast experiments show that the forecast accuracy is higher in summer,and that the errors on five-day forecasts occur mainly around the marginal ice zone. A reasonable initial state of ice concentration is essential for accurate short-term forecasts of sea ice using ice-ocean coupled models. In this study, sea ice concentration data are assimilated into an operational ice forecast system based on a combined optimal interpolation and nudging scheme The scheme produces a modeled sea ice concentration at every time step, based on the difference between observational and forecast data and on the ratio of observational error to modeled error. The impact and the effectiveness of data assimilation are investigated. Significant improvements to predictions of sea ​​ice extent were obtained through the assimilation of ice concentration, and minor improvements through the adjustment of the upper ocean properties. the assimilation of ice thickness data did not significantly improve predictions. Forecast experiments show that the forecast accuracy is higher in summer, and that the errors on five-day forecasts occur mainly around the marginal ice zone.