This study attempts to establish the first continental-scale multiobservation calibration and assessment of a land surface model (LSM) over the conterminous United States by using the Colorado State University Unified Land Model (CSU ULM) within the NASA GSFC's Land Information System and the Parameter Estimation (PEST) model. This study aims to calibrate the vegetation and soil optical parameters in different landcover classes by comparing model-predicted surface albedo and those derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) (including black- and white-sky albedo for visible and near-infrared band). The sum of squared deviations (Ψ) between model- and MODIS-derived albedo is iteratively reduced via the Gauss-Marquardt-Levenberg (GML) algorithm. The first calibration process (1) reduced Ψ by about 80% for noncalibrated as well as calibrated seasons and years, (2) revealed the functional biases related to diffuse-radiation upscattering parameters in two-stream canopy radiation scheme (which was fixed before the second calibration), and (3) shows that the parameter related to the leaf angle distribution function could not be tuned. The second calibration was implemented from the lessons learned from the first calibration, and results in the more realistic convergence of the parameters. After calibration, the summertime surface energy budget simulated by offline ULM changed significantly over the less vegetated regions; for example, net shortwave radiation and available energy increased by more than 40 W m−2 and radiative temperature increased by more than 1.6 K in the postcalibrated experiment.