Sea ice is an integral component of the global climate and weather system. The extent of sea ice is mainly influenced by, and has a significant effect on, the surface energy budget and ocean-atmosphere energy exchange. Regional and global changes in sea ice fraction and extent influence oceanic and atmospheric conditions, which in turn influence the evolution of sea ice itself. In this work we investigate the effect of sea ice on numerical weather prediction. A thermodynamic sea ice model is coupled to the NCEP Global Forecast System (GFS) atmospheric model. The sea ice model is based on Winton’s (2000) three-layer thermodynamic process. In each model grid box, the heat and moisture fluxes and albedo are treated separately for ice and open water (Wu et al. 1997). Preliminary results from cycled data assimilation/forecast experiments for a January 2004 case study show satisfactory performance of the new forecast model. While good agreement in the anomaly correlation between the new and old models is observed, the low-temperature bias in the lower troposphere in the high latitudes during winter has been greatly reduced in the new model, especially when a new data assimilation scheme is used (Okamoto et al. 2004). Further assessments of the forecasts are currently under investigation. Full results will be presented at the conference.