Measurement report: Unraveling PM10 sources and oxidative potential across Chinese regions based on CNN-LSTM data preprocessing and receptor model

Abstract

The oxidative potential (OP) of particulate matter is a key driver of PM10-induced adverse health effects, triggering oxidative stress and inflammatory responses that increase respiratory and cardiovascular disease risks. To evaluate PM10 and its OP characteristics across China, samples were collected from twelve representative monitoring stations from June 2022 to May 2023. A deep learning model combining Convolutional Neural Networks and Long Short-Term Memory networks (CNN-LSTM) was employed to reconstruct anomalous PM10 data, achieving R2 values of 0.8840 for test sets. Significant spatial variations in PM10 were observed, with highest concentrations in the northwestern regions (Xi’An: 98.20 ± 52.92 µg m−3, Dunhuang: 90.36 ± 54.72 µg m−3), the lowest in the northeast (Longfengshan: 40.04 ± 24.04 µg m−3, Dalian: 40.35 ± 15.66 µg m−3), and elevated levels in suburban areas (average: 85.43 ± 46.69 µg m−3). Urban sites showed the highest OP values (0.61 ± 0.21 nmol H2O2 m−3), with significantly higher PM10 concentrations in northern regions compared to southern ones (p < 0.05). Most sites exhibited peak PM10 and OP levels in winter and lowest in summer. Source apportionment using Positive Matrix Factorization (PMF) revealed dust (13.2 %–27.4 %), secondary aerosols (6.9 %–36.2 %), traffic (16.6 %–21.4 %), and biomass burning (22 %–39.3 %) as main contributors to PM10. Mass-normalized OP (OPm) analysis revealed traffic, biomass burning, and coal combustion sources showing consistently high values (0.008–0.022 nmol H2O2 µg−1). These findings highlight the need to control traffic, biomass burning, and coal combustion sources and other major sources to reduce OP and protect public health.