How to foresee the dispersion of a pollutant plume of badly known source in case of nuclear or industrial accident?

After the Chernobyl accident, the experts of atmospheric dispersion have stressed on the urgent need of developing modeling tools for simulating the transport of pollutants in the atmosphere. Thus, risk forecasting centres have developed numerical models simulating the dispersion of pollutant plumes. More recently, it was shown that in an uncertain context, for instance when the source or the amount of pollutant released is unknown, the use of data assimilation and inverse modeling techniques were necessary. In these techniques, two information sources are coupled in an optimal way: the pollutant concentration measurements, and the numerical dispersion model. The goal is to identify the pollutant source or to establish the best possible forecasting of the future evolution of the pollutant plume. This dossier makes a status of the progress made on this topic by the 'data assimilation' team of the CLIME project (integrated software chains for environment modeling) carried out by the CEREA, a joint research centre of the Ecole des Ponts and EdF R and D. (J.S.)