Abstract: The Los Alamos sea ice model CICE5.0 has been coupled to Beijing Climate Center Climate System Model (BCC_CSM2.0). To evaluate the effects of parameter optimization and improve the model performance on Arctic sea ice simulation, three parameters, including the emissivity, the maximum melting snow grain radius and the reference temperature of snow grain radius, are selected for sensitivity analysis. All numerical experiments are carried out with the ice-ocean model based on the BCC_CSM2.0 while the atmospheric circulation model is replaced by data model and forcing data are from Coordinated Ocean-ice Reference Experiments (CORE) data set. Results show that the model performance on Arctic sea ice simulation has been satisfactorily improved after the three parameters are tuned. (1) The sea ice thickness increases remarkably not only in the winter but also in the summer and is more comparable with observations. (2) The spatial distribution of summer ice concentration resembles observations better and the simulated sea ice extent significantly improves. Analysis of sea ice mass budget reveals that parameter optimization has enhanced the physics of albedo, which in turn affects the absorption of solar radiation and finally improves the model capability for simulations of sea ice concentration and thickness. However, some problems still remain especially in the simulation of sea ice area in the winter.