Two-week real-time monitoring of personal mass concentration of four PM size fractions in the Paris subway

PM exposure triggers adverse health effects. In subways, PM exposure levels, components and determinants are not well documented. This is in part due to the difficulty of modelling PM time series harboring non-stationary autocorrelation. Henceforth, we developed a Bayesian spline model for the real-time mass concentrations of PM10, PM2.5, PM1, and PM0.3 measured simultaneously in personal breathing zone of Parisian subway workers. The measurements were performed by GRIMM, gravimetric method and DiSCmini during the workers’ work-shifts over two consecutive weeks. Overall, PM10 was predominant in all jobs, closely followed by PM2.5. The PM0.3 concentrations were more than an order of magnitude lower compared to the other PM and showed the highest temporal variation, followed by PM1. For modelling, exposure levels of the study room at Porte de la Villette station (line 7) (0.31μg/m3 for PM0.3, 2.71μg/m3 for PM1, 3.32μg/m3 for PM2.5, and. 3.91μg/m3 for PM10) were used as reference. The PM2.5 raised the highest exposure concern: 15 stations out of 37 on line 7 had higher mass concentrations compared with reference. Twelve of these stations had also significantly higher PM1 mass concentration. Regarding PM10, only seven stations out of 37 had significantly higher mass concentration. Station PM levels were not correlated with the annual number of passengers entering the station, year of station opening or renovation, or the number of platforms and tracks. The correlation with the number of station entrances was consistently negative for all PM sizes. The number of correspondence concourses was negatively correlated with PM0.3 and PM10 and positively correlated with PM1 and PM2.5. Almost all studied environments had higher PM exposure compared to the study room, although the highest levels of PM were measured outdoors. The highest PM10 exposure was observed at the station platform, followed by the subway cabin and train, while ticket counters had the highest PM0.3, PM1, and PM2.5 mass concentrations. Compared with gravimetric and DiSCmini measurements, GRIMM results showed some discrepancies, with an underestimation of exposure levels. Therefore, we suggest to use GRIMM, calibrated by gravimetric methods, for PM sizes above 1μm, and a DiSCmini, calibrated by SEM, for sizes below 700nm.