New methodology to quantify source contributions to nitrate, sulfate and ammonium using PMF-AI

Area of Science:

• Environmental Science
• Atmospheric Chemistry
• Data Science

Background:

• Inorganic aerosol (IA) is a major component of PM2.5, posing health risks.
• Identifying IA sources is challenging due to diverse origins and lack of specific tracers.
• Effective reduction of IA levels necessitates accurate source apportionment.

Purpose of the Study:

• To identify heavy metal sources contributing to IA in Kaohsiung, Taiwan.
• To quantify individual IA source contributions using a novel PMF-AI technique.
• To analyze IA source contributions using domain knowledge and SHAP.

Main Methods:

• Utilized PM2.5-associated heavy metal measurements.
• Applied the Positive Matrix Factorization (PMF) model for source identification.
• Employed PMF-AI and Shapley Additive Explanation (SHAP) for source contribution analysis.

Main Results:

• The PMF-AI model explained up to 99% of the variance in daily IA values.
• Oil combustion and biomass burning/waste incineration/agrochemical applications were identified as significant IA sources.
• SHAP analysis confirmed the substantial influence of these sources on IA concentrations.

Conclusions:

• The novel PMF-AI approach accurately quantifies IA source contributions.
• This method provides crucial insights for developing effective IA reduction policies.
• The adaptable technique can benefit regions addressing IA-related environmental issues.