The newly developed nonhydrostatic model MESO-NH, in which the surface scheme Interactions Soil–Biosphere–Atmosphere has been incorporated, is used in this study to assess the impact of increasing the horizontal resolution from 10 km to 1 km on the simulation of surface and turbulent fluxes for the 16 June 1986 case of HAPEX-MOBILHY, a field experiment that took place in southwestern France. Except for a slight deterioration over the cultivated areas surrounding the Landes forest (caused by an inconsistency between the soil texture fields at 10 and 1 km), the simulation of the surface fluxes of sensible and latent heat is generally improved by the increase of horizontal resolution. The contrast of the sensible heat fluxes between the Landes forest and the surrounding cultures is well captured in both 10-km and 1-km runs, but the spatial variability of these fluxes is better represented in the high-resolution results. An oasis-type effect over the larger clearings of the Landes forest is even produced by the model, as was observed. For the 1-km simulation, the comparison of the turbulent fluxes against observations has to include both the grid-scale fluxes resulting from resolved larger eddies within the well-mixed layer, as well as subgrid-scale (i.e., parameterized) fluxes. (At 10-km resolution, all turbulent fluxes are parameterized.) The greater contributions of the grid-scale component are found over the forest, where the larger eddies are more vigorous due to stronger sensible heat fluxes at the surface. For sensible and latent heat fluxes, the grid-scale component is particularly important in the middle of the mixed layer, whereas for turbulent kinetic energy this component is greater near the bottom and top of the mixed layer. In general, the increase of horizontal resolution does not improve significantly the simulation of the turbulent fluxes. Thus, the use of such an intermediate horizontal resolution (i.e., 1 km), lying between that typically used in large-eddy simulation models (<200 m) and that of mesoscale models (>few kilometers), is questionable, even though this resolution is probably optimal for simulating surface fluxes, since it is roughly the same as the resolution of the soil and vegetation databases.