Artificial Intelligence (AI) and Monte Carlo Simulation-Based Modeling for Predicting Groundwater Pollution Indices and Nitrate-Linked Health Risks in Coastal Areas Facing Agricultural Intensification

Abstract

This study assesses groundwater quality and nitrate-related health risks in the Skhirat coastal aquifer (Morocco) using a multidisciplinary approach. A total of thirty groundwater wells were sampled and analyzed for physico-chemical properties, including major ions and nutrients. Multivariate statistical analyses were employed to explore contamination sources. Pollution indices such as the Groundwater Pollution Index (GPI) and Nitrate Pollution Index (NPI) were computed, and Monte Carlo simulations (MCSs) were conducted to assess nitrate-related health risks through ingestion and dermal exposure. Furthermore, Random Forest (RF), Gradient Boosting Regression (GBR), Support Vector Regression (SVR) with radial basis function kernel, and Artificial Neural Networks (ANN) models were tested for predicting groundwater pollution indices. Results of hydrochemical facies revealed Na+-Cl− dominance in 47% of the samples, suggesting strong marine influence, while nitrate concentrations reached up to 89.3 mg/L, exceeding World Health Organization (WHO) limits in 26.7% of the sites. Pollution indices indicated that 33.3% of samples exhibited moderate to high GPI values, with 36.7% of the samples exceeding the threshold for NPI. The MCS for nitrate health risk revealed that 43% of the samples posed non-carcinogenic health risks to children (Hazard Index (HI) > 1). RF outperformed other models in predicting GPI (R2 = 0.76) and NPI (R2 = 0.95). Spatial prediction maps visualized contamination hotspots aligned with intensive horticultural activity. This integrated methodology offers a robust framework to diagnose groundwater pollution sources and predict future risks.