A secondary pollution modelling system for simulating airborne dispersion and chemical reactions is tested over a regional scale domain located in the North-West of Italy, covering areas characterized by different emission levels and where urban and industrial areas are present. This region is often affected by severe pollution episodes, which are driven by anthropogenic emissions and meteorological conditions. The domain is centred over a main plain area, surrounded by elevated Alpine mountains and the sea. This valley is characterized by unfavourable meteorological conditions, such as weak circulation, frequent low wind and stagnant conditions. Moreover the complexity of the orographic site affects both reconstruction of a proper emission picture and meteorology, in particular at the domain borders where the Alpine mountains are located. The model resolution was 5 km and simulations have been run for the whole year 1999. Emissions are derived from different inventories based on different territorial units. The meteorological input was provided by the meteorological model RAMS, whereas the dispersion and chemical reactions were computed by CAMx photochemical model. The results of the simulations are compared together with measured sulphur dioxide, nitrogen dioxide and particulate matter data. Two winter periods were considered because of the PM accumulation processes that take place during the cold season. The simulated PM concentrations are also compared with the results of the same modelling system, applied over a larger scale and at lower (25 km) resolution. CAMx provided quite satisfactorily performances in reproducing the seasonal evolution of PM10, particularly at 5 km resolution. Afterwards, model performances have been compared against SO2 and NO2, two of the most important particulate matter precursors. As expected, the comparison shows a worsening in the model performance moving from gas species to PM10, confirming that some processes related to particulate matter are still missing in chemical transport models. Moreover, PM10 performances proved to be related more to NO2 than SO2. In fact, while the former gives rise mainly to local scale production, sulfate formation is generally related to large-scale processes, hence less linked to the precursor concentration at the receptor point.