Atmospheric boundary layer (ABL) simulations over remotely sensed boundary conditions using a large eddy simulation (LES) code are employed here to explore the dynamical coupling of heterogeneous land surfaces and the ABL. The LES was recently extended to incorporate remotely observed surface states, and the ability to account for the soil and vegetation (i.e., two sources) contributions to the mass and energy exchanges <Albertson et al., 2001>. In this effort we explore the impacts of changes in the magnitude of surface variability (i.e., spatial contrasts). We focus our simulations and analysis with rescaled surface temperature fields to explore a wider range of contrasts (i.e., spatial variance). We demonstrate that the increase in temperature contrast has negligible effect on regionally averaged fluxes. However, the strength of coupling (or feedback) between spatial fields of land surface and surface layer temperature (z ~ 10 m) increases with increasing temperature contrast. This dampens increases in the spatial variance in the sensible heat flux relative to increases in the spatial variance in surface temperature, suggesting the feedbacks act to limit the spatial variability in the flux. We also use the LES to explore the errors induced in spatially distributed heat flux predictions from using spatially uniform atmospheric variables in a related two-source energy-balance radiometric surface temperature scheme. The use of spatially uniform atmospheric variables is commonly employed when computing surface fluxes from remotely sensed land surface data. This leads to significant differences in the spatial distribution of land surface fluxes when compared to LES derived fluxes. This was particularly evident in the overestimated Bowen ratio (¿), primarily for locations with relatively low vegetation cover.