Employing Machine Learning Approaches and Multivariate Analysis to Assess Groundwater Quality for Irrigation in the Mornag Plain, Tunisia

Abstract

The crucial assessment of water quality in the Mornag Plain, Tunisia, is fundamental for reservoir management, ensuring suitability for consumption, and preserving environmental integrity. Employing a diverse range of methodologies, such as water quality indices (WQIs), statistical analyses, geographic information systems (GIS), and decision tree (DT) model, provided a nuanced understanding of the compositional attributes of groundwater designated for irrigation. Detected categories of water, for instance Na-Cl, Ca-Mg-SO4, and a combination of Ca-Mg-Cl/SO4, displayed unique chemical signatures. These patterns were shaped by diverse processes including interactions between water and rock, the breakdown of dolomite, the concentration of minerals through evaporation, the swapping of ions, and human impact. Evaluating groundwater's suitability for irrigation purposes by employing measures like Na%, SAR, SSP, and MH, demonstrated that a significant portion of the samples conformed to approved norms. However, the analysis revealing 65.6% of the IWQI and every instance of PS falling within the spectrum of high to severe constraints, as well as ranging from detrimental to unsatisfactory classes, underscores the obstacles in sustaining superior irrigation water standards. The predictive model, DT, demonstrated robust results in forecasting all water quality indices. The DT-HyC-9 model stood out as the top performer in prediction accuracy, demonstrating a robust correlation with prime factors affecting IWQI, as shown by substantial R2 metrics in both the training and evaluation phases. Likewise, the DT-HyC-3 approach showcased remarkable precision in forecasting IWQI, highlighting the critical role of three essential factors. These results provide crucial guidance for policymakers engaged in managing water quality, laying the groundwork for the sustainable management of water resources within the Mornag Plain. The integration of advanced methodologies and predictive models enhances the understanding of groundwater dynamics, facilitating informed decision-making for the region's water sustainability.