The sensitivity of a thermodynamic sea ice model to changes in surface energy fluxes in the Arctic is investigated. The main emphasis of the paper is on the testing of the model sensitivity to changes in surface albedo parametrization. Climatologies of turbulent and long-wave fluxes in the Arctic are scarce, and those that exist are shown to generate significant differences in the predicted ice thickness. There is considerable disagreement in the literature on albdeo values, and in particular, proposed albedos of bare, puddled ice range from 0.4 to 0.66. The differences among published model simulations are shown to be potentially explicable in terms of this range in bare ice albedo. A new ice albedo parameterization is proposed, and its sensitivity is tested. It is shown that the increase in surface albedo with cloud cover can cause a doubling of the ice thickness, and the need to include melting snow as an albedo class distinct from dry snow is demonstrated. The value of bare-ice albedo is shown to be important in determining whether the ice is in a multiyear or a seasonal ice zone, and the need for more observational data on the extent and role of melt puddles is emphasized.