Measurements of the dry deposition velocity of ozone have been made by the eddy correlation method during ESCOMPTE (Etude sur Site pour COntraindre les Modèles de Pollution atmosphérique et de Transport d'Emissions). The strong local variability of natural ecosystems was sampled over several weeks in May, June and July 2001 for four sites with varying surface characteristics. The sites included a maize field, a Mediterranean forest, a Mediterranean shrub-land, and an almost bare soil. Measurements of nitrogen oxide deposition fluxes by the relaxed eddy correlation method have also been carried out at the same bare soil site. An evaluation of the deposition velocities computed by the surface module of the multi-scale Chemistry and Transport Model MOCAGE is presented. This module relies on a resistance approach, with a detailed treatment of the stomatal contribution to the surface resistance. Simulations at the finest model horizontal resolution (around 10 km) are compared to observations. If the seasonal variations are in agreement with the literature, comparisons between raw model outputs and observations, at the different measurement sites and for the specific observing periods, are contrasted. As the simulated meteorology at the scale of 10 km nicely captures the observed situations, the default set of surface characteristics (averaged at the resolution of a grid cell) appears to be one of the main reasons for the discrepancies found with observations. For each case, sensitivity studies have been performed in order to see the impact of adjusting the surface characteristics to the observed ones, when available. Generally, a correct agreement with the observations of deposition velocities is obtained. This advocates for a sub-grid scale representation of surface characteristics for the simulation of dry deposition velocities over such a complex area. Two other aspects appear in the discussion. Firstly, the strong influence of the soil water content to the plant response, specifically in conditions of stress, is confirmed. Second, we point out the difficulty in interpreting measurements of nitrogen oxide deposition velocities: a synergetic approach combining measurements and modeling is practical.