Estimating contributions of vehicular emissions to PM in a roadside environment: A multiple approach study.

Vehicular emissions (VE) are among the major sources of airborne fine particulate matter (PM) in urban atmospheres, which adversely impact the environment and public health. Receptor models are widely used for estimating PM source contributions from VE (PM), but often give inconsistent results due to different modelling principles and assumptions. During December 2015-May 2017, we collected nine-months of hourly organic carbon (OC) and elemental carbon (EC) data, as well as 24-h PM speciation data including major species and organic tracers on select days from an ad hoc roadside site in Hong Kong. The weekday vs. holiday and diurnal variations of EC tracked closely with those of traffic flow volume, indicating EC as a reliable tracer for PM in this area. We applied multiple approaches to estimate the PM, including the EC-tracer method with the hourly OC-EC data, and chemical mass balance (CMB) and positive matrix factorization (PMF) analyses with the filter-based speciation data. Considering source profile variability, CMB gave the lowest PM estimate among the three approaches, possibly due to the degradation of organic markers (i.e., hopanes). The PM derived from the EC-tracer method and PMF were comparable, accounting for ~12% (3.4-4.0 μg/m) of PM averaged across 20 samples in both approaches, but a larger sample size is needed for a more robust PMF solution. The monthly PM derived from the EC-tracer method was in the range of 3.2-6.6 μg/m. The continuous measurement reveals a decreasing trend in PM throughout the entire sampling period, indicating the effectiveness of a recent vehicle control measures implemented by the Government in phasing out pre-Euro IV diesel commercial vehicles. This work implies that hourly OC-EC monitoring at strategically located spots is an effective way of monitoring vehicle control measures. It provides reasonable estimate of PM through comparing with other more sophisticated receptor models.