Modelling urban roads particulate matter and airborne traffic-related metals exposure: AERMOD-Support vector machine hybrid framework for Monte Carlo health risk assessment

Abstract

Urban roads air pollution from fine particulate matter (PM_2.5) is a major environmental concern. Toxic metals associated with PM_2.5 pose health risks but remain poorly studied in Pakistan. In present investigation, the dust and PM_2.5 bound metals pollution with associated health risks were evaluated by cumulative field measurements, dispersion modelling, and machine learning. For this purpose, dust was collected from solar panels, and airborne PM_2.5 was gathered using a low volume sampler for dual seasons (June and December). Along urban roads, the PM_2.5 dispersion generated by vehicular activities was simulated using the AERMOD Gaussian dispersion model. Then, utilizing AERMOD outputs and Wheater data for Support Vector Machine (SVM) model were trained to forecast metals (As, Cd, Cr, Ni, Pb, and Zn) in both dust and PM_2.5. Elevated concentrations of Zn, Cu, and Cd were observed in PM_2.5 and deposited dust, and AERMOD modeled PM_2.5 mass concentrations peaked at 115.7 μg/m³, exceeding applicable regulatory limits. The SVM approach attained significant predictive accuracy for metal concentrations (R² = 0.75, 0.81, 0.85 for As, Cd, and Pb). Health risk perspective, hexavalent chromium (Cr (VI)) posed the substantial noncarcinogenic and carcinogenic risk in both seasons. The observed and SVM predicted metals total cancer risk ranged for PM_2.5 (TCR ≤10⁻⁴) and dust (TCR <10⁻⁶), and hazard index varied for PM_2.5 (HI > 1) and dust (HI < 1) in both seasons, respectively. The study was limited by minor roads data and lack of local emission standards; future research should include deeper local modelling and extensive traffic emissions databases.