Near-surface water balance modeling is often used to evaluate land-atmosphere interactions, deep drainage, and groundwater recharge. The purpose of this study was to compare water balance simulation results from seven different codes, HELP, HYDRUS-1D, SHAW, SoilCover, SWIM, UNSAT-H, and VS2DTI, using 1--3 year water balance monitoring data from nonvegetated engineered covers (3 m deep) in warm (Texas) and cold (Idaho) desert regions. Simulation results from most codes were similar and reasonably approximated measured water balance components. Simulation of infiltration-excess runoff was a problem for all codes. Annual drainage was estimated to within ¿64% by most codes. Outliers result from the modeling approach (storage routing versus Richards' equation), upper boundary condition during precipitation, lower boundary condition (seepage face versus unit gradient), and water retention function (van Genuchten versus Brooks and Corey). A unique aspect of the code comparison study was the ability to explain the outliers by incorporating the simulation approaches (boundary conditions or hydraulic parameters) used in the outlying codes in a single code and comparing the results of the modified and unmodified code. This approach overcomes the criticism that valid code comparisons are infeasible because of large numbers of differences among codes. The code comparison study identified important factors for simulating the near-surface water balance.