Detecting and quantifying PM and NO contributions from train and road traffic in the vicinity of a major railway terminal in Dublin, Ireland.

Air pollution from transport hubs is a recognised health concern for local urban inhabitants. Within the domain of transport hubs, significant attention has been given to larger airport and port settings, however concerns have been raised about emissions from urban railway hubs, especially those with diesel trains. This paper presents an approach that adopts low-cost monitoring (LCM) for fixed site monitoring (FSM) to quantify and disaggregate PM and NO contributions from railway station and road traffic on air quality in the vicinity of railway station in Dublin, Ireland. The NO sensor showed larger discrepancies than the PM sensor when compared to the reference monitor. Machine learning models (XGBoost and Random Forest (RF) regression) were applied to calibrate the LCM devices, with the XGBoost model (NO R = 0.8 and RSME = 9.1 μg/m & PM, R = 0.92 and RSME = 2.2 μg/m) deemed more appropriate than the RF model. Local wind conditions, pressure, PM concentrations, and road traffic significantly impacted NO model results, while raw PM sensor readings greatly influenced the PM model output. This highlights that the NO sensor requires more input data for accurate calibration, unlike the PM sensor. The monitoring results from the one-month monitoring campaign from May 25, 2023 to June 25, 2023 presented elevated NO and PM concentrations measured at the railway station, which translated to exceedances of the annual WHO limits (PM = 5 μg/m, NO = 10 μg/m) by 1.6-1.8 and 3.2-5.2 times respectively at the study site. A subsequent data filtering technique based on wind orientation, revealed that the railway station was the main PM source and road traffic was the main NO source when winds come from the railway station. This study highlights the value of LCM devices alongside robust machine learning techniques to capture air quality in urban settings.