Land surface heterogeneity over an area of 71 km × 74 km in the lower Walnut River watershed, Kansas, was investigated using models and measurements from the 1997 Cooperative Atmosphere Surface Exchange Study (CASES-97) field experiment. As an alternative approach for studying heterogeneity, a multiscale atmospheric and surface dataset (1, 5, and 10 km) was developed, which was used to drive three land surface models, in uncoupled 1D mode, to simulate the evolution of surface heat fluxes and soil moisture for approximately a 1-month period (16 April–22 May 1997) during which the natural grassland experienced a rapid greening. Model validation using both surface and aircraft measurements showed that these modeled flux maps have reasonable skill in capturing the observed surface heterogeneity related to land-use cover and soil moisture. The results highlight the significance of rapid greening of grassland in shaping the surface heterogeneity for the area investigated. The treatment of soil hydraulic properties and canopy resistance in these land surface models appears to cause the majority of differences among their results. Several factors contributing to the discrepancy between modeled and aircraft measured heat fluxes in relation to their respective time–space integration were examined. When land surface heterogeneity is pronounced, modeled heat fluxes agree better with those measured by aircraft in terms of spatial variability along flight legs. When compared to Advanced Very High Resolution Radiometer/Normalized Difference Vegetation Index (AVHRR/NDVI) data, it is demonstrated that modeled heat flux maps with different spatial resolutions can be utilized to study their scaling properties at local or regional scales.