Variability of air pollution (PM1) analyzed using explainable Machine Learning

Air pollution and in particular high concentrations of particulate matter (PM) are known to be harmful to human health. However, the quantification of factors leading to high levels of PM remains challenging, as both anthropogenic and meteorological factors contribute to high pollution events. Here, a novel approach using a machine learning algorithm is proposed to identify and quantify drivers of concentrations of speciated particles with a diameter below 1µm (PM1) using meteorological data from the Site Instrumental de Recherche par Télédétection Atmosphérique (SIRTA) observatory located southwest of Paris. PM1 concentrations were modelled and effects of meteorological conditions on modelled PM1 concentrations were analyzed. Mixing layer, wind direction and temperatures showed to have high explanatory power to the model.