Separation of diurnal and seasonal variation from urban ozone monitoring by Bayesian statistics

Abstract

The variation pattern of potential ozone (PO) in four urban sites, Osaka, Kobe, Tokyo, and Kawasaki, were separated into the diurnal and seasonal components using Bayesian statistical analysis to evaluate the individual contributions from in situ photochemical production and the background concentration to total ozone (O₃) concentrations. The Bayesian model involved a diurnal component expressed as the product of the time-of-day dependent diurnal variation pattern (D_h), its month-dependent amplitude (B_m), and a month-dependent seasonal component (A_m). The estimated D_h for the four urban sites showed good agreement with each other, having the diurnal cycle resembling those of the hydroperoxy radicals observed at an urban site. B_m showed a seasonal pattern with a significant increase in the summer season. The sudden increase in B_m in August was characteristic of two sites in the Kanto area. The variation in A_m was similar among the four sites, however, a slight difference in the variation of A_m during the summer season was detected between the Kansai and the Kanto regions. When the diurnal and seasonal components were regarded as the local photochemical production and the background contributions, respectively, the local contribution was the highest in August. The extremely large photochemical production in the Kanto area seemed to cause the vague autumn maximum in the seasonal variations of PO in the area.