Cross-media dynamics and prioritized risks of PFAS in textile-impacted environments: using geospatial machine...

Area of Science:

• Environmental Chemistry
• Environmental Science
• Toxicology

Background:

• Per- and polyfluoroalkyl substances (PFAS) pose environmental risks, particularly from textile industries.
• Understanding PFAS cross-media dynamics between soil and tree bark is limited by traditional models.

Purpose of the Study:

• To elucidate PFAS interactions between soils and tree barks in Chinese textile hubs using a machine learning framework.
• To assess environmental risks and identify high-priority PFAS for mitigation strategies.

Main Methods:

• Geospatially-informed machine learning (ML) framework applied across 48 sites.
• Target analysis of 31 PFAS in soil and tree bark samples.
• Development of a novel multi-criteria risk assessment framework (ToxPi, Risk Index, EHPi).

Main Results:

• Higher PFAS concentrations were found in tree barks than in soils.
• Ultra-short-chain and emerging PFAS dominated profiles, with TFA and 8:2 FTS identified as high-risk.
• LightGBM ML model demonstrated superior predictive performance (R² = 0.962-0.974).
• Tree bark PFAS significantly predicted soil contamination, highlighting cross-compartment influence.

Conclusions:

• Machine learning effectively models PFAS dynamics in textile-impacted environments.
• Textile industry emissions contribute significantly to soil and bark PFAS burdens.
• Integrated risk assessment framework aids in prioritizing pollution mitigation strategies for PFAS.