Quantifying regional transport contributions to PM2.5-bound trace elements in a southeast coastal island of China: Insights from a machine learning approach

Abstract

Identifying and quantifying pollution sources and their associated health risks are essential for formulating effective pollution control policies. This study analyzed PM_2.5-bound trace elements based on one year of sampling data collected from a low-PM_2.5 island in southeastern coastal China. A de-weathered model based on the eXtreme Gradient Boosting (XGBoost) algorithm was applied to remove meteorological influences and estimate local baseline pollutant concentrations. By combining backward air mass trajectories with de-weathered concentrations, we quantified the variation in transport contributions among different trajectory types. Results indicated that meteorological factors reduced PM_2.5 and anthropogenic trace element concentrations by 36.7 %–58.4 % in summer, but increased them by 6.4 %–26.0 % in winter. In contrast, elements related to shipping emissions showed an opposite trend. Positive matrix factorization (PMF) identified industrial and shipping emissions as the two main sources of trace elements, originating from distinct regions. Shipping emissions contributed greatly health risks in summer, while industrial emissions dominated in other seasons. The non-carcinogenic risk (NCR) remained within acceptable levels, whereas carcinogenic risks (CR) exceeded recommended thresholds. Marine airflows (MA), inland airflows (IA), and local airflows (LA) altered trace element concentrations by −3.7 %, +6.4 %, and −5.4 %, respectively. These airflow types changed NCR by −16.4 %, +8.2 %, and −13.5 %, and CR by −4.1 %, +4.7 %, and −28.9 %, respectively. These findings underscore the substantial impact of regional transport on trace elements and the critical need for coordinated regional air quality management, offering new insights into pollutant sources and their associated health risks in relatively less polluted coastal regions.