Long term simulations with Chemical Transport Models (CTM) are becoming very important for the assessment of air quality and for the evaluation of the impact of emission reduction policies. The meteorological input is very important in these simulations, since atmospheric dispersion strongly affects near-surface concentrations of pollutants: this is particularly true in regions of low wind and complex terrain, such as the Po Valley. Six-months simulations were performed over Milan region, using the same CTM and different meteorological inputs, and the resulting concentrations of ozone and other pollutants were compared with observations. Hourly values of wind field and turbulence parameters were produced using the direct output of a LAM and a meteorological pre-processor (Calmet) based on a dense network of local observations; all other CTM input were left unchanged, and were made available in the framework of City-Delta project. Prior to CTM simulations, the reconstructed wind fields were also compared with a set of independent observations: the LAM was able to reproduce the most significant features of the Po valley circulation, including mountain breeze, but generally overestimated wind speed and underestimated the temporal variability of near-ground wind; on the other hand, winds reconstructed from observations were sometimes incoherent and probably lacked organised circulations in the upper boundary layer. The results of CTM simulations showed a very high sensitivity to meteorological input, with differences in 6 months average ozone concentration exceeding 10 ppb on most of the domain. When applying CTM to the peculiar situation of the Po Valley, a great care must be put in the meteorological input; the use of a combination of the output of a LAM and local observations could significantly improve the evaluation of pollutant concentrations..