기계학습 기반 농도 재구성을 통한 산업도시에서의 다환방향족탄화수소와 중금속의 장기 흡입 위해성 평가

Abstract

Long-term exposure to hazardous air pollutants (HAPs) is difficult to evaluate in industrial cities due to limited spatio-temporal coverage of routine monitoring. Annual concentrations of 13 polycyclic aromatic hydrocarbons (PAHs) and 8 metals were reconstructed for 19 air quality monitoring stations in Ulsan, South Korea, from 2001 to 2023 using Random Forest models based on criteria air pollutants and meteorological variables. Validation yielded R values of 0.85 for Σ7 PAHs and 0.66-0.80 for individual metals, while independent passive air sampling supported the spatial reproducibility of the model. Traffic emissions, petrochemical activities, metal processing, and heavy manufacturing were identified as major sources. Compared with a constant-concentration assumption based on the 2023 annual mean, year-specific concentration reconstruction showed that PAH exposure was underestimated by up to 14% in residential areas, whereas metal exposure was overestimated by up to 18% near industrial areas. Residence-history scenarios indicated that temporal changes in pollutant levels and residential location can alter time-integrated exposure. Across 11 long-term monitoring stations, year-specific inhalation cancer risks ranged from 2.92 × 10-8 to 4.52 × 10-8 for PAHs and 4.48 × 10-6 to 5.73 × 10-6 for metals. Long-term concentration reconstruction can provide a more reliable basis for time-integrated exposure and risk assessment in industrial cities.