Proposal for estimating ground-level ozone concentrations at urban areas based on multivariate statistical methods

Abstract

This study focuses on describing ozone patterns and estimating ozone concentrations in urban settings through the classification of an urban area into homogeneous typologies, according to hourly ozone concentrations, and the development of accurate estimation models for each typology. For these proposals, a hierarchical cluster analysis was conducted in order to define homogeneous subareas, and multiple linear regressions were subsequently applied with the aim of obtaining ozone predictions, employing chemical and meteorological variables as predictors. Seville metropolitan area (Spain) is a densely populated area of the Mediterranean Basin that exhibits environmental problems related to ozone pollution episodes. Ozone exceedances are a consequence of the combination of road traffic and industry emissions with hot temperatures and high solar radiation, mainly during anticyclonic events. Cluster analysis evince that this area can be divided into 3 categories according to hourly ozone concentration in summer. Cluster 1 is comprised of monitoring stations located in the outskirts of the city of Seville; Cluster 2 corresponds to monitoring stations located within the city of Seville; and Cluster 3 is comprised of a monitoring station specialized in traffic emissions. Multiple linear regression shows that the relative weight of meteorological variables decreases when moving from the urban periphery towards the urban center, whereas the weight of chemical variables increases. Coefficients of determination (R2) values were 0.885, 0.890 and 0.830 and root mean squared error (RMSE) were 11.226, 11.874 and 11.260 μg m−3 for Cluster 1, 2 and 3, respectively.