Source-specific control leverage for reducing oxidative potential of PM_2.5 revealed by explainable machine learning

Area of Science:

• Environmental Chemistry
• Atmospheric Science
• Toxicology

Background:

• Oxidative potential (OP) of fine particulate matter (PM2.5) is a critical health metric beyond mass concentration.
• Understanding OP's complex chemical drivers and source contributions is essential for effective emission control.
• Nonlinear interactions among PM2.5 components and emission sources complicate OP variability analysis.

Purpose of the Study:

• To investigate the nonlinear drivers of PM2.5 OP variability using year-long observational data from central Taiwan.
• To characterize OP responses to chemical species and source contributions using explainable machine learning.
• To develop a source-resolved framework for evaluating OP mitigation strategies.

Main Methods:

• Analysis of year-long PM2.5 chemical composition and OP data.
• Application of explainable machine learning and regime-level response mapping.
• Integration of positive matrix factorization for source identification and interaction analysis.

Main Results:

• Nonlinear dependencies and interactions involving transition metals, water-soluble organic carbon, and aerosol liquid water significantly influence OP.
• Source combinations can amplify or suppress OP depending on pollution regimes.
• Coal/heavy oil combustion, industrial emissions, and secondary pollutants show high OP mitigation leverage.

Conclusions:

• A system-level framework linking OP response dynamics to source-resolved mitigation leverage was established.
• Source-specific control strategies are crucial for effectively reducing PM2.5 OP.
• This study provides quantitative insights for toxicity-oriented air quality management beyond traditional mass-based controls.