Risk assessment of potentially toxic elements and mapping of groundwater pollution indices using soft computer models in an agricultural area, Northeast Algeria

Groundwater (GW) quality and contamination by potentially toxic elements (PTEs) are major concerns for environmental sustainability, particularly in arid regions. The aim of this study was to assess the human health risks associated with GW contamination by PTEs in the Terminal Complex (TC) aquifer of the Tolga oasis, located in northeastern Algeria. Seventeen GW samples were analyzed using standard methods to determine contamination levels and associated health risks. Results showed that GW was generally contaminated with lead (Pb), which exceeded the WHO permissible limit of 0.01 mg/L in 76.47% of the samples. Although some samples were rich in Cr and Mn, their levels were below WHO guidelines. Pollution indices, including Contamination Factor (CF), Heavy Metal Pollution Index (HMI), and Nemerow Pollution Index (NPI), indicated that over 50% of the samples had medium to high pollution levels. These indices were further estimated using artificial neural network (ANN) and Multiple Linear Regression (MLR) machine learning models, whose performances were validated by Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Nash-Sutcliffe Efficiency Coefficient (NSE). The Taylor diagram analysis showed that MLR models were more accurate than ANN models in estimating GW pollution indices. Mapping these indices using support vector machine (SVM) algorithms and applying chemometric statistical techniques, including principal component analysis (PCA), revealed that alteration of geological formations and anthropogenic activities significantly affected GW contamination by PTEs in the study area. The assessment of health risks associated with heavy metals revealed a significant non-carcinogenic risk, particularly for children, with 41.17% of samples exceeding the hazard index threshold of 1 due to Pb exposure, while carcinogenic risks were low. This study establishes predictive models based on heavy metal pollution indices, providing crucial information on the spatial distribution of GW contamination. The results support the development of targeted mitigation strategies and intervention plans to safeguard GW resources and public health in the region.