Comprehensive assessment of Per- and polyfluoroalkyl substances (PFAS) pollution characteristics and environmental health risks in the Northeast Black Soil Region of China.

Abstract

Per- and polyfluoroalkyl substances (PFAS) pose long-term risks to ecosystems and human health because of their persistence, bioaccumulation, long-range transport, and toxicity. Focusing on the Northeast Black Soil Region of China, a major grain production base with intensive agricultural activities and complex industrial inputs, this study developed a comprehensive analytical framework integrating multi-source data (2005-2025), including literature, environmental monitoring, soil properties, and toxicity databases. The framework combined meta-analysis, principal component analysis (PCA), machine learning, and entropy-weight-based risk assessment to systematically characterize PFAS pollution patterns, spatial heterogeneity, and priority risks. Random-effects meta-analysis revealed moderate-to-strong positive correlations among PFAS congeners (pooled effect size = 0.336) with significant heterogeneity (I² = 96.80%). PCA showed that the first two principal components explained over 48% of the total variance, whereas six components accounted for more than 85%. Among eleven machine-learning models, Logistic Regression achieved the best performance and stable generalization when the sample size exceeded 60 samples. SHapley Additive exPlanations (SHAP) identified latitude, longitude, and soil organic carbon as the dominant predictors. Risk assessment indicated that PFOS and PFOA exhibited the highest priority levels (ToxPi: 0.872 and 0.499; EHPi: 0.839 and 0.679), whereas approximately 60% of PFAS were classified as low risk and 20% as medium-to-high risk. This framework provides quantitative support for targeted PFAS management and offers a transferable strategy for regional-scale assessment of emerging contaminants.