Abstract High ozone levels are regularly reached during summer period in South-European urban areas, calling for careful design of primary pollutants emission reduction strategies. In this perspective the CALGRID modelling system has been applied to Milan metropolitan area, located in the Po Valley, the most industrialised and populated area in Italy. For the first modelling exercise, a simulation domain of 100×100 km 2 has been considered and a summer period, characterised by high photochemical activity, has been selected. Hourly emissions have been derived by spatially and temporally disaggregating national inventories data, while standard upper-air and ground-based meteorological data have been used as input to the CALMET pre-processor. A careful analysis of simulation results versus local network monitoring data has revealed some critical points, related to both modelling assumptions and practical data availability. A satisfactory reproduction of daytime ozone behaviour has been, in fact, accomplished, both in urban and suburban sites, while nighttime primary pollutants accumulations and consequent ozone consumption simulated by the model have not found correspondence in the measurements. Nitrogen dioxide has been also successfully modelled, mostly in city surroundings, whereas higher discrepancies have been found in some urban stations. Possible explanations of these facts are discussed in the paper, giving an insight for further work.