In order to improve root parameterization in land surface model, the submodel for root in CERES-Maize was coupled in the SSiB2 after calibrating of maize parameters in SSiB2. The effects of two improved root parameterization schemes on simulated results of land surface flux were analyzed. Results indicated that simulation accuracy of land surface flux was enhanced when the root module provided root depth only with the SSiB2 model (scheme Ⅰ). Correlation coefficients between observed and simulated values of latent flux and sensible flux increased during the whole growing season, and RMSE of linear fitting decreased. Simulation accuracy of CO₂ flux was also enhanced from 121 days after sowing to mature period. On the other hand, simulation accuracy of the flux was enhanced when the root module provided root depth and root length density simultaneously for the SSiB2 model (scheme Ⅱ). Compared with the scheme Ⅰ, the scheme Ⅱ was more comprehensive, while its simulation accuracy of land surface flux decreased. The improved root parameterization in the SSiB2 model was better than the original one, which made simulated accuracy of landatmospheric flux improved. The scheme Ⅱ overestimated root relative growth in the surface layer soil, so its simulated accuracy was lower than that of the scheme Ⅰ.