Abstract: A variational data assimilation algorithm for assimilating land surface temperature (LST) in the Common Land Model (CoLM) is implemented using the land surface and canopy energy balance as the adjoint physical constraint. In this data assimilation algorithm, the evaporative fraction (EF) of soil and canopy is adjusted according to the surface temperature observations. The analysis results from CoLM with the LST assimilation algorithm are well compared with the field observations from AmeriFlux data at Bondville site. This algorithm is also tested at the regional scale located in East China. These results indicate that the surface temperature assimilation method is efficient and effective when only one-time observational data per day are available. From the histogram of the LST error, it can be concluded that the LST after assimilation is improved much compared with the MODIS observations especially in the daytime. The evapotranspiration patterns also change much after the assimilation. All these can prove the robustness of the algorithm which this paper proposes in assimilation applications to the CoLM. The variational assimilation method this paper develops can improve the simulation results of land surface model, and has an important meaning to the land surface processes such as land surface hydrology and vegetation ecology etc. Meanwhile, the land surface model also could be coupled with the numerical prediction model, and then the variational algorithm also can improve the results of numerical forecast.