Prediction of heavy metal and PM2.5 concentrations in atmospheric particulate matter using key magnetic parameters

Abstract

Heavy metal pollution is harmful to the human health and the environment, and it is of great significance to accurately predict the concentration of heavy metals in atmospheric particulate matter. However, the prediction of heavy metals in atmospheric particulate matter has not yet been reported, and traditional geochemical methods are inefficient and time-consuming. In this study, environmental magnetic parameters were introduced as independent variables of machine learning to predict the concentration of heavy metals in atmospheric particulate matter and classify PM_2.5 concentrations. Four popular models were constructed to predict heavy metal concentrations. Moreover, using magnetic parameters and PM_2.5 concentrations as feature values, the correlation between magnetic parameters, PM_2.5 and heavy metal concentrations were explored. The results show that all heavy metals are positively correlated with χ_lf, SIRM, HIRM and χ_ARM, and the GA-SVM model has the best prediction performance. Additionally, the optimal GA-SVM model was used to perform sensitivity analysis on Fe heavy metal concentration and to conduct PM_2.5 concentration classification prediction, it was found that SIRM, HIRM and χ_ARM have a significant effect on the prediction results, and the prediction results are highly accurate. The research results have reference significance for the prediction of pollutant concentrations in the future.