Auto-associative models for pollutants atmospheric dispersion (MADiPA)

(Modèles Auto-associatifs pour la Dispersion de Polluants dans l’Atmosphère)

Duration: 18 months, starting in September 2020.

Partially funded by AMIES through the PEPS program.

Research conducted by Valentin Pibernus under my supervision, in partnership with INRIA :

Atmospheric release of noxious materials requires fast decision making, which increasingly relies on numerical simulation. Atmospheric dispersion models predict pollutant spatial spreads from meteorological forecast and conjectures about the emission dynamics (the “source term”). These prediction are uncertain due to the high complexity of the phenomena at hand, and the uncertainty of input data. For instance, a fire started in the Lubrizol chemical plant near Rouen (France) on 26 September 2019, causing massive release of ill charaterised material in the atmosphere. The divergences between the stances of the stakeholders illustrate well the magnitude and impact of the uncertainty inherent to such accident.

Find more details in the posts-doc (filled) job offer (in French).