The parameterization of the atmospheric boundary layer is essential for accurate numerical weather predictions. The near-surface values of air temperature or wind speed for instance are highly dependent on the complex land – atmosphere interactions over heterogeneous terrain. Over such surfaces, several open challenges remain regarding the growth of internal boundary layers, the determination of mixing layer heights and the spatial distribution of heat and momentum fluxes. The large-eddy simulation (LES) code we apply is based on the robust Lagrangian scale-dependent dynamic subgrid-scale model (Bou-Zeid et al., 2005, Physics of Fluids). The flow is driven by a mean pressure gradient expressed in terms of horizontal geostrophic winds. The code is pseudo-spectral with spectral decomposition in the horizontal dimensions. Land surface heterogeneities take the form of spatially distributed patches of surface temperature and roughness derived from satellite imagery and land use analysis. The simulation of the complete diurnal cycle is possible but we restrict our discussion to fully convective conditions. Data from LITFASS – 2003 (Lindenberg Inhomogeneous Terrain Fluxes between Atmosphere and Surface: a long-term Study) are used to validate the simulation results. The LES domain covers a 99-m meteorological tower with turbulent measurements and five surface micrometeorological stations. We investigate the presence of large-scale turbulent structures that are typical for daytime conditions with a particular interest in the blending height. Overall, this study illustrates how important is the accurate parameterization of heterogeneous terrain in studies of the atmospheric boundary layer.