Probabilistic assessment of heavy metal risks in sewage sludge using BCR speciation and Bayesian networks

The environmental risk associated with heavy metals in sewage sludge is commonly assessed using total metal concentrations; however, this approach fails to account for differences in metal mobility and bioavailability governed by chemical binding forms. In this study, an integrated probabilistic framework was developed to assess heavy metal mobility in sewage sludge by combining BCR sequential extraction, speciation-based risk indices, and Bayesian network modelling. Sewage sludge samples collected from multiple municipal wastewater treatment plants under different seasonal and operational conditions were characterized in terms of physicochemical properties, total metal content, and chemical speciation. The application of conventional risk indices revealed substantial inconsistencies in risk classification, reflecting their divergent conceptual foundations and limited ability to provide coherent decision support. The Bayesian network model integrated heterogeneous inputs, including metal speciation fractions, risk indices, and operational factors, to generate probability-based assessments of overall metal mobility risk. The results demonstrated that metals associated with labile fractions exhibited the highest probability of elevated mobility risk, while metals predominantly bound to stable fractions were consistently classified as low risk across scenarios. Sensitivity analysis confirmed that chemical speciation was the dominant driver of mobility risk, with physicochemical and operational factors influencing risk indirectly through their effect on metal binding behavior. The proposed framework advances sewage sludge risk assessment beyond deterministic and retrospective approaches by explicitly accounting for uncertainty and enabling scenario-based evaluation. By providing probabilistic, decision-oriented insights, the integrated approach offers a robust tool for supporting sustainable sewage sludge management and can be adapted to other complex environmental systems. This study represents one of the first attempts to integrate chemical speciation and Bayesian inference into a unified probabilistic framework for sewage sludge risk assessment.