Observed and modelled chemical weather during ESCOMPTE

The new MOdèle de Chimie Atmosphérique à Grande Echelle (MOCAGE) three-dimensional multiscale chemistry and transport model (CTM) has been applied to study heavy pollution episodes observed during the ESCOMPTE experiment. The model considers the troposphere and lower stratosphere, and allows the possibility of zooming from the planetary scale down to the regional scale over limited area subdomains. Like this, it generates its own time-dependent chemical boundary conditions in the vertical and in the horizontal. This paper focuses on the evaluation and quantification of uncertainties related to chemical and transport modelling during two intensive observing periods, IOP2 and IOP4 (June 20–26 and July 10–14, 2001, respectively). Simulations are compared to the database of four-dimensional observations, which includes ground-based sites and aircraft measurements, radiosoundings, and quasi-continuous measurements of ozone by LIDARs. Thereby, the observed and modelled day-to-day variabilities in air composition both at the surface and in the vertical have been assessed. Then, three sensitivity studies are conducted concerning boundary conditions, accuracy of the emission dataset, and representation of chemistry. Firstly, to go further in the analysis of chemical boundary conditions, results from the standard grid nesting set-up and altered configurations, relying on climatologies, are compared. Along with other recent studies, this work advocates the systematic coupling of limited-area models with global CTMs, even for regional air quality studies or forecasts. Next, we evaluate the benefits of using the detailed high-resolution emissions inventory of ESCOMPTE: improvements are noticeable both on ozone reactivity and on the concentrations of various species of the ozone photochemical cycle especially primary ones. Finally, we provide some insights on the comparison of two simulations differing only by the parameterisation of chemistry and using two state-of-the-art chemical schemes for regional photochemical modelling. Regional air quality modelling is found to be highly sensitive to the emission inventory dataset and also to the vertical and horizontal boundary conditions and detailed representation of chemistry. Interestingly enough, they infer the same range of errors compared to total model errors