Source characterization of trace elements and impacts of salinization using geochemical modelling with machine learning approach and Monte-Carlo based health risk assessment in coastal aquifer zones.

This work investigated the trace element contamination, spatial distribution and their source of origin along with health risk assessment in the coastal region of eastern India. Geochemical modelling, Principal Component Analysis-Multiple Linear Regression (PCA-MLR) and self-organizing map (SOM) has been employed to understand the effect of salinization and the prevailing Eh-pH condition on the trace element behavious. The results show ∼41% and 27% groundwater belong to the heavy-Heavy Metal Pollution Index (HPI) and high-Heavy Metal Evaluation Index (HEI) categories, respectively. Uncertainty analysis from Monte-Carlo simulation predicted (37%, 29%) chance of non-carcinogenic risk (NCR) through oral pathways, similarly (9%, 0.25%) chances through dermal contact, for children and adults, respectively, with higher sensitivity loading by Fe and Mn. Simulation also suggests a higher chance of carcinogenic risk (CR) for children than adults. Geochemical modelling indicates mineral phases (FeCO3, FeSO4·7H2O, Mn3O4, Mn (OH)2) are the major contributors of Fe and Mn in the coastal aquifer. The major factors are extracted by PCA and supported by SOM in which (TDS, B, Sr) showing higher loading in PC1 suggested seawater mixing mechanism, where as positive loading on (As, Ba, Fe, Mn) in PC2 indicates geogenic source with enhanced dissolution due to redox condition, while PC3 and PC4 having loading on (Al, Ba, U), and (Zn), derived from geogenic sources. The study integrates PCA-MLR, Monte-Carlo simulation, and geochemical modelling with SOM, for the comprehensive evaluation of pollution levels, and which will be useful for the sustainable management of water resources in the coastal regions.