A novel method for establishing typical daily profile of PM concentrations in underground railway stations

Abstract

The air quality in underground railway stations (URS) poses a significant public health concern due to extremely high concentrations of particulate matter: PM₁₀ and PM_2.5. Indeed, PM sources are strong and numerous, such as train braking and tunnel effect and URS are often confined spaces with low air change rates. Despite multiple PM measurements within URS, the variability of those concentrations from stations to stations is still poorly understood. We present here a methodology for establishing a daily profile of particle mass concentrations, based on a 5-year long measurement series in a Parisian URS. This approach incorporates an extensive data cleaning process based on the identification of URS operation periods and physically inconsistent or mathematically aberrant data, together with a linear regression model. This methodology delivers three usable outcomes: a typical profile for weekdays, a typical profile for weekends, and a PM concentration Daily Amplitude Coefficient (DAC) for the considered period. The DAC is a daily metric of the pollution levels, that enables the analysis of temporal trends and facilitates the comparison with other data with other acquisition frequency. The methodology developed here in a specific URS for PM₁₀ measurements can be easily applied to different particle size fractions or to other measured parameters exhibiting a daily profile. Weekdays PM₁₀ concentrations exhibit two distinct peaks corresponding to morning and evening rush hours, with an average daytime concentration of 193 µg/m³. These peaks are delayed by ∼1 hour compared to the train traffic. Weekends show consistently lower PM levels with no observable peaks, averaging 157 µg/m³ during the day. Our analysis reveals the long-term temporal evolution of PM concentration within the URS, highlighting seasonal patterns with higher PM₁₀ concentrations observed in summer (up to 400 µg/m³) and lower values in winter (down to 250 µg/m³). This indoor seasonal evolution is not correlated with the outdoor temporal evolution, showing higher concentrations during the winter. Furthermore, our results show that the optimal period (DAC∼1) for conducting experiments to obtain reliable profiles is during the spring months (April, May, June).