6533b7d8fe1ef96bd126a0ac

RESEARCH PRODUCT

Prévisibilité à 24h de la pollution en ozone à Dijon et Chalon-sur-Saône

subject

description

International audience; The ATMOSF'air network monitoring air quality would like to be able to predict ozone concentrations 24 hours in advance in accordance with the air quality law of 1996. The University of Burgundy is collaborating with the networks on this project. Primary studies show that tropospheric ozone is the main pollutant in Burgundy and that the meteorological conditions are very important when considering ozone concentrations (Houzé, 1999 ; Richard et al., 2000). We predict the ozone hourly maximal value (predictand) using three types of predictors. First are ozone level from Laboratory of Meteorology Dynamic (LMD) calculated from air mass back trajectories and precursor integration (Vautard et al., 2000), second next day meteorological data measured in Dijon and Chalon-sur-Saône Météo-France stations, and third the pollution variables (precursors): CO, NO, NO2. Predicting models use two linear methods: Linear Multiple Regression and Discriminant Analysis. We obtained one model for each ATMOSF'air station. We used five predictors. Once models have been established, improvement can be made. First results show that this predicting is possible in operational use. Regression models explain about 65% variance and Discriminant Analysis models reach about 68% good classification.