Spatial Characterization of Tailpipe and Non-Tailpipe Emissions in Toronto Using Land-Use Regression Modelling

CHRISTI JOSE, Nicole Trieu, Yee Ka Wong, Sophie Roussy, Cheol H. Jeong, Greg J. Evans, SOCAAR, University of Toronto

     Abstract Number: 389
     Working Group: Urban Aerosols

Abstract
Traffic-related air pollution (TRAP) remains a critical environmental health challenge, with a growing concern over the contribution of non-tailpipe (NTP) versus tailpipe (TP) emissions. This study applied land-use regression (LUR) modelling to explore the spatial variation of TP and NTP emissions across Toronto, Canada. Weekly saturation measurements were conducted at approximately 40 monitoring sites across the city during three seasonal campaigns - Fall (October-November 2023), Winter (January-March 2024), and Summer (July-August 2024), yielding over 900 PM₂.₅ and PM₁₀ filter samples for trace element analysis, and more than 480 Ogawa badge samples for nitrogen oxides (NO, NO₂, NOx).

Using a supervised forward stepwise regression framework, guided by prior expectations of variable directionality, LUR models were developed based on GIS-derived traffic and land-use characteristics calculated within multiple buffer distances (25-1000 m). The NOx model, representing TP emission, performed well (adjusted R² = 0.84), with highway length and traffic signals identified as key predictors. Positive Matrix Factorization from a near-road field campaign resolved Ba, Cu, and Fe as NTP brake-wear markers, supporting the strong intercorrelations (~0.80) observed in the saturation measurements. An integrated brake-wear metric derived from principal component analysis captured greater spatial variance (adjusted R² = 0.85), outperforming individual metal models for Ba, Cu and Fe (adjusted R² = 0.70-0.79). The joint model relying on similar predictor variables such as roadway density and traffic signals, provided a more robust representation of brake-wear emissions. The study also assessed local emission contributions through background subtraction method, wherein concentrations at background sites were subtracted from those at monitoring sites. As the project advances, we will compare LUR models for additional NTP tracers such as Ca and Zn to better understand their distinct spatial behaviours and refine joint modelling approaches for NTP emissions to enhance their accuracy and interpretability.