Ambient PM2.5 pollution poses a major risk to public health in China, contributing to significant mortality and morbidity. While overall PM2.5 concentrations have declined in recent years, the changes in PM2.5 chemical constituents remain inadequately understood due to limited ground monitoring networks. We developed a Super Learner model that integrates MISR satellite data, chemistry transport model simulations, and land use information to predict daily OC concentrations across China from 2003 to 2019 at a 10-km spatial resolution. The model achieved high predictive accuracy with a cross-validation R2 of 0.84 and an RMSE of 4.9 μg/m3. Our findings show elevated OC levels in Northern China, driven by industrial activities with concentrations exceeding 30 μg/m3 during the heating season. In contrast, forest fires were the primary contributors in Yunnan, raising OC concentrations to 20-30 μg/m3 during fire seasons. Over the 17-year period, the national OC trend declined by 1.3 % annually. Regionally, the Beijing-Tianjin-Hebei region and the Fenwei Plain experienced faster reductions at annual rates of 1.5 % and 2.0 %, respectively, while Yunnan exhibited no significant trends. To better understand pollution source contributions, we analyzed the OC/EC ratio, which indicated higher ratios in less populated rural areas, suggesting agricultural and biogenic emissions, while lower ratios in urban clusters pointed to primary sources such as traffic and industrial activities. Notably, since 2013, significant decreases in the OC/EC ratio have been observed in the North China Plain, likely reflecting the impact of stringent air pollution control policies on biomass burning. This study provides valuable exposure estimates for epidemiological research on the long-term health effects of OC in China, offering insights for evaluating air quality policies and guiding future management strategies.
Keywords: Elemental carbon; MISR; Organic carbon; PM(2.5) constituents; Super learner.
Copyright © 2024. Published by Elsevier B.V.