Machine learning assessment of industrial corrosion and public health risks in data limited aquifers of Northeastern Algeria

Industrializing arid regions face a dual challenge: ensuring water suitability for emerging industries while protecting public health amidst chronic data scarcity. This study confronts this issue in northeastern Algeria by developing and validating an interpretable machine learning (ML) framework to predict industrial water quality and associated health risks from limited groundwater data. Based on 45 samples collected from the Debila and El-Oued regions of the Souf Valley, Artificial Neural Network (ANN), Random Forest (RF), and Decision Tree (DT) models were trained to predict seven corrosion and scaling potential (CSP) indices, namely the Langelier Saturation Index (LSI), Ryznar Stability Index (RSI), Aggressive Index (AI), Puckorius Scaling Index (PSI), Larson-Skold Index (LS), Chloride-Sulphate Mass Ratio (CSMR), and Revelle Index (RI), alongside the Canadian Council of Ministers of the Environment Water Quality Index (CCME-WQI). Physicochemical analysis revealed pervasive water quality deterioration, with 60% of samples exhibiting corrosive tendencies (LSI < 0) and 85% showing high corrosion risk (RSI > 6.8). All sites were uniformly classified as “Poor” by the CCME-WQI, and Monte Carlo simulations (10,000 iterations) confirmed significant non-carcinogenic health risks from nitrate for children (mean Hazard Index > 1.0). The ANN and RF models demonstrated robust predictive power, achieving R² values above 0.90 for six of the eight predicted indices. SHAP (SHapley Additive exPlanations) analysis identified the dominant hydrochemical drivers for each index, providing a transparent and actionable basis for targeted monitoring. This study presents a novel, validated, and transferable ML-based decision support tool that simultaneously integrates industrial process risks with public health assessment, offering a replicable water management paradigm for data-limited, industrializing regions worldwide.