Ozone pollution regimes modeled for a summer season in California’s San Joaquin Valley: A cluster analysis

This study demonstrates an application of cluster analysis to model simulation data for California’s San Joaquin Valley (SJV) for the purpose of identifying meteorologically representative pollution regimes. Principal component analysis is employed to facilitate exploring and visualizing temporal variations in highly resolved gridded model data. Six regimes are clustered according to the spatial distribution of SJV 8 h ozone maxima. Meteorological effects (temperature and winds) are shown to explain the observed ozone spatial distributions in the SJV, and their relationship to those in upwind San Francisco Bay Area air basin (SFB) under certain prevailing wind flow patterns. In general, average ozone levels in the SJV increase with temperature, while their spatial distributions depend on flow regimes, especially the strength of sea breezes and upslope flows. More ventilated flow regimes, associated with stronger sea breeze and upslope flows, cause eastward transport of pollutants, increasing ozone in the southeastern SJV and decreasing it in the northwest SJV. The opposite occurs during the most stagnant conditions associated with the weakest sea breeze and upslope flows. The two most prominent relationships between the SFB and SJV were found to be associated with the most ventilated and the most stagnant conditions, respectively, indicating a strong inter-basin transport (or the lack thereof) event. Spatial representativeness of existing measurement sites and the confounding influences of emission changes on clustering results are also investigated. Existing measurement sites are able to capture ozone spatial patterns in the SFB and Sacramento Valley (SV), whereas those along the western side of the SJV are under-represented. Differences in day-of-week emissions produce minor effects on spatial ozone distributions and the clusters are largely stable under these changes.

► Cluster analysis and principal component analysis applied to modeled results.
► Pollution regimes characterized by spatial ozone and underlying meteorology.
► Visualized variability in air quality model results with high-dimensionality.
► Determined spatial representativeness of measurement locations.
► Determined influences of emission changes on clustering stability.