Algorithmes d’optimisation pour la surveillance et l’estimation de la pollution de l’air

Abstract

The implementation of an air quality monitoring system requires taking in consideration complex meteorological phenomena, sources of emission and limitations drawn by the costly equipment. The three main contributions made by this thesis regarding monitoring and estimation of air pollution are: a review of techniques for estimating air quality, influence of air pollution’s spatial and temporal variability on precision of interpolation methods, and a suggestion for a possible optimization of Dunkirk air quality monitoring network. Firstly, we did a bibliographic synthesis on Air Quality Modeling techniques, detailing their advantages and limits in studying air pollution. Then, we estimated air pollution in the Hauts-de-France region by means of spatial interpolation methods. We proposed an optimization of Inverse distance Weighting (IDW) interpolation technique that allows improving the coefficient of determination (R²). The influence of these phenomena was filtered by averaging the PM10 data at different time scales. The best R² obtained corresponded to the 24 hours averaging period, similar to the periodicity of some local weather phenomena such as sea breezes occurring in coastal areas. Furthermore, we suggest an approach to optimize the network of measurement stations in Dunkirk that minimizes root-mean-square error of air pollution estimation obtained by IDW interpolation using data of ADMS (Atmospheric Dispersion Modeling System) and the Gaussian plume model. It was shown that the optimized configuration allows obtaining better PM10 concentration estimations compared to the real deployed measuring stations network of ATMO. Actual ATMO stations are located near the emission sources, while for the resulting topology of stations optimization on diffuse sources pollution (ADMS) stations were scattered throughout the region, and for point source pollution (Gaussian plume) stations surrounded the emission source. Finally, a reliable and efficient approach was proposed for improving the accuracy of estimation of air pollution in an area of special interest.