Evaluation of meso-scale dispersion modelling for accidental release

A modelling system, PERLE, has been developed at Meteo-France, for the Crisis Meteorological Cell (CMC) of atmospheric accidental release, in case of emergency. Mesoscale meteorological fields are simulated by the nonhydrostatic MESO-NH model (8km and 2km resolution nested grids), that also supply an overview of the plume on a regional scale with an eulerian passive tracer. The dispersion model is a lagrangian particle model, to describe the pollutant cloud in the vicinity of the release during the first critical few hours. Two stochastic dispersion models have been evaluated and compared: DIFPAR and SPRAY. The performance of PERLE is tested against a large range of stability and turbulence conditions. A set of fourteen test cases has been conducted on Kr-85 measurements, released in La Hague plant gaseous waste and measured in the surrounding near field (<5km) by IRSN. Most of the meteorological conditions concern a marine BL, with neutral static stability and a small diurnal variability. In these situations, both models have shown quasi-similar and correct results beyond 800m from the source. Below that, modelling ATC are largely underestimated, partly due to insufficient meteorological horizontal resolution. The system is also applied to the Toulouse Sud campaign (1993), capturing the essential features of the plume. However the 2km meteorological resolution is not sufficient to capture the topographic influence on the flow structure with accuracy. A sensitivity study on the height release has been performed for the neutral and convective boundary layer on Toulouse: the asymmetric structure of turbulence in the CBL appears correctly with SPRAY. PERLE is able to yield reliable simulations of atmospheric dispersion in inhomogeneous turbulence, for emergency purposes, with competitive computation time (< 30min).