Through their influence on the exchange of energy, mass, and momentum between the land surface and atmosphere, surface characteristics, such as vegetation type and density, exert a critical control on a broad range of meteorological, hydrological, and ecological processes. As a result, accurately representing the impacts of surface heterogeneity on land-atmosphere interactions is a key prerequisite to correctly modeling weather and climate, particularly on local and regional scales. Concentrating on the domain of the 2002 International H2O Project (IHOP_2002), more specifically, the Walnut River Watershed in southeastern Kansas, this work combined in-situ, airborne, remotely-sensed, and modeled data to (1) characterize the relationship between surface heterogeneity and spatial variability in surface and moisture fluxes and (2) assess the ability of the High Resolution Land Data Assimilation System (HRLDAS) and Noah land surface model (Noah) to reproduce the observed fluxes. The model was run at a temporal resolution of one hour and a spatial resolution of 30 m. In order to compare the observed fluxes and the model output, estimates of the observed fluxes were calculated as the product of the model output and a footprint weighting function derived from the observational data. Based on an analysis of the airborne and tower measurements of the surface fluxes, the variability along the 50-km flight path for the sensible (10 to 20 W m-2) and latent (40 to 60 W m-2) heat fluxes was tied primarily to variations in vegetation density with soil moisture content playing a secondary role. A comparison of the model-based estimated and observed fluxes showed that HRLDAS failed to predict both the magnitude and variability of the airborne flux measurements. For example, using the sensible heat flux data from 22 June, not only were the modeled and observed fluxes poorly correlated (r=0.63), the model overestimated the flux by more than a factor of two. The discrepancy between the model and observations was traced to both the coarse resolution of the model inputs and the limitations of the model’s physics.