A land surface model (a modified version of the Simple Biosphere Model, Version 2; SiB2) was parameterized and tested against two years of eddy covariance flux measurements made over un-grazed tallgrass prairie and a winter wheat field in Oklahoma, USA. The land surface model computed 30-min estimates of sensible and latent heat flux and carbon dioxide flux that agree well with the patterns observed in the field, simulating in particular the contrasting seasonal timing of fluxes in the wheat, where leaf area and physiological activity peak in the spring, and prairie, where leaf area and physiological activity peak in summer. However, systematic errors in flux estimates were also identified for particular times of day and season. These systematic errors are sometimes related to difficulty in correct definition of vegetation structure (LAI) and physiological activity. This was observed particularly in the wheat site towards the end of the growing season when senescence, which reduces both the amount and the physiological activity of leaves, is difficult to parameterize. Simulation errors are also attributed to problems in the mathematical description of water stress, soil respiration, and the leaf-to-canopy scaling methodology. SiB2 tends to predict too much photosynthetic activity at low solar angles, while developing soil moisture stress before it is seen in the field. Systematic errors in energy balance terms (heat and water fluxes) occur for bare soil and dormant vegetation, related to simulation of soil heat flux. Daytime errors in sensible and latent heat fluxes average 20 W m−2, and can be more than 100 W m−2 at certain times. In regional and global climate models, the effect of land surface sub-model errors on atmospheric dynamics will depend in part on the magnitude of the systematic error, but also on the spatial extent and temporal duration over which the systematic error persists.
