Time-series output from the Eta forecast model of the National Weather Service's National Centers for Environmental Prediction was evaluated by comparison with measured values for a location in Oklahoma. The measured data were drawn from the archives of the Department of Energy's Atmospheric Radiation Measurement program for the southern Great Plains site. Surface energy budget components and cloud indicators were examined for the first half of 1997. Overall, the Eta surface energy budget was found to be nearly balanced, as intended by the model physics, except for one instance when light snowfall occurred in the spring. Despite this balance, an average 50 W m-2 excess in incoming solar radiation was found. Approximately half of this excess was attributed to deficient extinction of shortwave radiation by water vapor and aerosols in the model, while the remainder appeared to be due to cloud treatment errors. The excess shortwave flux was offset by a smaller negative bias in the downwelling infrared flux and the use of slightly high albedos, in addition to errors of lesser magnitude in the latent and sensible heat fluxes. The upwelling infrared flux and ground heat flux were closer to measured values. Ambiguity in the definition of a cloud as well as measurement limitations hindered the analysis of cloud prediction. However, a rough cloud comparison indicated that the Eta model has more difficulty predicting convection than the movement of large storm systems. In addition, cirrus clouds were predicted too frequently in the winter, while low and midlevel clouds were often missed in the spring. This study demonstrates that single-point time-series data can be used effectively both to ascertain the quality of model output and to investigate the treatment of individual physical processes within models. ¿ 1999 American Geophysical Union