Machine Learning in Contaminated Soil Remediation: From Monitoring, Source Apportionment, and Risk Assessment to Practical Applications

Abstract

Severe soil contamination over recent years has sparked massive research focus on soil remediation. Traditional remediation approaches are categorized as physical, chemical, and biological methods, alongside their composite combined remediation modes. Matching remediation strategies to site-specific pollution features facilitates enhanced remediation efficiency. Nonetheless, complicated soil matrices and heterogeneous pollutants in terms of types, physicochemical traits, and concentrations render the screening and optimization of remediation technologies highly challenging. Recent advances have facilitated the widespread application of artificial intelligence (AI) in soil remediation research. Machine learning (ML), as a core branch of AI, substantially enhances the management and analytical performance of soil monitoring data. By collating and analyzing literature related to soil remediation, machine learning, and artificial intelligence, this review outlines state-of-the-art advances in ML-driven data processing and remediation optimization and further compares the performance and characteristic properties of various machine learning models for soil remediation across different pollutant categories. Machine learning leverages historical data analysis, scenario simulation and nature-inspired optimization strategies to greatly reduce soil sampling requirements and improve data utilization efficiency. This review summarizes the key strengths and inherent drawbacks of machine learning techniques in contaminated soil remediation and further indicates the promising application prospects of machine learning in this field.