Evaluation of a mesoscale dispersion modelling tool during the CAPITOUL experiment

Atmospheric transport and dispersion were investigated during the CAPITOUL campaign using measurements of sulphur hexafluoride (SF6) tracer. Six releases of SF6 tracer were performed (March 9-11 and July 1-3, 2004) in the same suburban area of Toulouse conurbation, during the Intensive Observing Periods (IOP) of CAPITOUL. Concentration data were collected both at ground-level along axes perpendicular to the wind direction (at distances ranging between 280 m and 5000 m from the release point), and above the ground at 100 m and 200 m height using aircraft flights. Meteorological conditions were all associated with daytime anticyclonic conditions with weak winds and convective clear and cloudy boundary layers. A meso-scale dispersion modelling system, PERLE, developed at Meteo-France for environmental emergencies in case of atmospheric accidental release, was evaluated in terms of meteorology and dispersion, for the different tracer experiments, in its operational configuration. PERLE is based on the combination of the non-hydrostatic meso-scale MESO-NH model, running at 2 km horizontal resolution, and the Lagrangian particle model SPRAY. The statistical meteorological evaluation includes two sets of simulations with initialisation from ECMWF or ALADIN. The meteorological day-to-day error statistics show fairly good Meso-NH predictions, in terms of wind speed, wind direction and near-surface temperature. A strong sensitivity to initial fields concerns the surface fluxes, crucial for dispersion, with an excessive drying of the convective boundary layer with ALADIN initial fields, leading to an overprediction of surface sensible heat fluxes. A parameterization of dry and shallow convection according to the Eddy-Diffusivity-Mass-Flux (EDMF) approach (Pergaud et al. 2008) allows an efficient mixing in the Convective Boundary Layer (CBL) and improves significantly the wind fields. A statistical evaluation of the dispersion prediction was then performed and shows a realistic behaviour of the system, with a good location of the concentration maxima. But the lateral spread of the plumes is quasi-systematically underestimated, mainly in July, even when meteorological conditions are well reproduced. In the same way, higher integrated concentration values are slightly overestimated. The remove of the EDMF scheme in Meso-NH artificially improves the horizontal dispersion, underlying compensating errors. Sensitivity tests performed on the Lagrangian time scales in the coupling Meso-NH-SPRAY have been conducted. But they don't solve the shortcoming and lead to the conclusion that SPRAY could have some difficulties to correctly reproduce the mixing for daytime thermal convection. © Springer-Verlag 2008