Revealing the sources of water-soluble PM_2.5 oxidative potential with explainable machine learning

Area of Science:

• Environmental Chemistry
• Atmospheric Science
• Toxicology

Background:

• Quantifying the individual contributions of organic compounds and transition metals to PM2.5 oxidative potential (OP) is challenging due to their complex redox interactions.
• Conventional regression methods are insufficient for accurately assessing the sources and impacts of PM2.5 OP.
• Understanding the drivers of PM2.5 OP is crucial for mitigating associated health risks.

Purpose of the Study:

• To develop and apply a novel Extreme Gradient Boosting coupled with SHapley Additive exPlanations (XGBoost-SHAP) model to quantify the contributions of chemical species and sources to PM2.5 OP.
• To investigate the key redox-active components and sources of water-soluble PM2.5 OP in Beijing using dual-assay data (DTT consumption and ·OH production).
• To establish a robust nonlinear regression approach for better understanding PM2.5 OP sources and informing mitigation strategies.

Main Methods:

• Utilized an XGBoost-SHAP model for nonlinear regression analysis of PM2.5 oxidative potential.
• Employed dual-assay data measuring DTT consumption (DTTv) and ·OH production (OHv).
• Incorporated 20 chemical features, 5 meteorological features, and 9 water-soluble organic compounds (WSOC) source features derived from Positive Matrix Factorization (PMF).

Main Results:

• Moderately hydrophilic water-soluble organic compounds (WSOC) were identified as the most influential factors for both DTTv and OHv.
• Primary and secondary WSOC sources contributed significantly to DTTv (44.9% and 55.1%, respectively).
• Secondary WSOC predominantly drove OHv, accounting for 84.6% of its activity. Simpler WSOC source factors showed better performance in summer simulations.

Conclusions:

• The XGBoost-SHAP model provides a robust method for dissecting the contributions of various factors to PM2.5 oxidative potential.
• Water-soluble organic compounds, particularly secondary WSOC, are critical contributors to PM2.5 oxidative potential in Beijing.
• Findings offer valuable insights for targeted strategies to reduce PM2.5-related health risks.