Spatial patterns of conditions leading to peak O3 concentrations revealed by clustering analysis of modeled data

• Published in: Air quality, atmosphere and health Vol. 12; no. 6; pp. 743 - 754
• Main Authors: Pineda Rojas, Andrea L., Leloup, Julie A., Kropff, Emilio
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.06.2019; Springer Nature B.V; Springer
• Subjects: Air quality; Air quality models; Atmospheric Protection/Air Quality Control/Air Pollution; Cluster analysis; Clustering; Computer simulation; Diurnal; Earth and Environmental Science; Emission measurements; Environment; Environmental Health; Geophysics; Health Promotion and Disease Prevention; Metropolitan areas; Model testing; Ozone; Ozone measurements; Physics; Urban areas; Clustering analysis; Ozone; Air quality modeling; Monte Carlo simulations; Buenos Aires
• ISSN: 1873-9318; 1873-9326
• DOI: 10.1007/s11869-019-00694-9

Air quality models are currently the best available tool to estimate ozone (O 3 ) concentrations in the Metropolitan Area of Buenos Aires (MABA). While the DAUMOD-GRS has been satisfactorily evaluated against observations in the urban area, a Monte Carlo (MC) analysis showed that it is the region around the MABA, where the lack of observations impedes model testing, that concentrates not only the greatest estimated O 3 peak levels but also the largest model uncertainty. In this work, we apply clustering analysis to these MC outcomes in order to study the spatial patterns of conditions leading to peak ozone hourly concentrations. Results show that families of conditions distribute, as emissions, radially around the city. A cluster exhibiting an O 3 morning peak dominates in low-emission areas, a behavior that can be explained both from theory and from the few monitoring campaigns carried out in the city. Its distinct dynamics compared with the typical O 3 diurnal profile occurring in the urban area suggests the need of new ozone measurements in the surroundings of the MABA which could contribute to improve our understanding of O 3 formation drivers in this region. The results illustrate the potential of applying clustering analysis on large ensembles of modeled data to better understand the variability in model solutions.