Groundwater quality assessment for agricultural utilizing indexical and machine learning techniques in Ouled Djellal Aquifer, Southern Algeria

Abstract

Groundwater represents the main water resource for irrigation in the Ouled Djellal region (southeast of Algeria). Despite the importance of groundwater in this area, its quality and sustainability remain insufficiently studied. Therefore, this study aimed to introduce an integrated analytical framework by combining multivariate statistical techniques i.e., Principal Component Analysis (PCA) and Hierarchical Ascending Classification (HAC), irrigation indices (IWQI, SAR, Na%, SSP, PS, and RSC), and machine learning (ML) models such as Artificial Neural Network (ANN), Support Vector Machine (SVM), and Multiple Linear Regression (MLR) to assess and predict groundwater quality for irrigation. The main difference with previous studies is the fact that this work applied Empirical Bayesian Kriging Regression Prediction (EBKRP) to spatialize irrigation indices derived from ML with higher precision. The approach enables cross-validation of model performance and captures complex nonlinear interactions among hydrochemical parameters. The attained results revealed that groundwater quality was varied from moderate to poor for irrigation, driven mainly by salinity and sodicity effects. In addition, the ANN model achieved the highest predictive accuracy (R² = 0.97, RMSE = 1.50), confirming its superiority in modelling complex hydrochemical behavior. The proposed modelling framework represents a methodological advancement for data-scarce arid regions, serving as a practical tool adaptable to groundwater monitoring and irrigation planning in similar regions.