Machine learning approaches for monitoring environmental metal pollutants: Recent advances in source apportionment, detection, quantification, and risk assessment

Abstract

Metal pollutants pose significant and enduring threats to human health and the environment, mainly due to their non-biodegradable nature. Traditional monitoring of these pollutants involves costly and time-consuming analytical methods. However, this conventional approach struggles to grasp the intricate and nonlinear connections between metals and the factors affecting pollution dynamics. Machine learning (ML) has emerged as a promising solution, capable of handling large datasets, diverse inputs, and complex patterns. This review highlights insights from developing innovative ML models to monitor metal pollutants. Beginning with an overview of commonly used ML algorithms for this purpose, the review comprehensively explores recent advancements in ML applications across various environmental matrices. It covers source apportionment, detection, quantification, and ecological risk assessment in solids, water, and air. The review also identifies research gaps and proposes future directions for maximizing ML's potential in preventing and managing environmental metal pollution.