Applying Multivariate Analysis and Machine Learning Approaches to Evaluating Groundwater Quality on the Kairouan Plain, Tunisia

Groundwater on the Kairouan Plain, Tunisia, exhibits diverse chemical characteristics primarily influenced by water–rock interaction, dolomite dissolution, evaporation, and ion exchange. Analysis revealed three distinct water types: Ca-Mg-SO4, Na-Cl, and mixed Ca-Mg-Cl/SO4, with ions following the order Na+ > Ca2+ > Mg2+ > K+ and SO42− > HCO3− > Cl−. While most groundwater is suitable for agriculture, irrigation water quality indices (IWQIs) indicate high-to-severe restrictions in some areas. Machine learning models, particularly ANN-HyC-9 and XGBoost regression, demonstrated high accuracy in predicting IWQIs, with R² values exceeding 0.823 for testing datasets. These findings provide critical insights for sustainable water resource management.

Salem, S., Gaagai, A., Ben Slimene, I., Moussa, A., Zouari, K., Yadav, K., Eid, M., Abukhadra, M., El-Sherbeeny, A., Gad, M., Farouk, M., Elsherbiny, O., Elsayed, S., Bellucci, S., & Ibrahim, H. (2023). Applying Multivariate Analysis and Machine Learning Approaches to Evaluating Groundwater Quality on the Kairouan Plain, Tunisia. Water, 15(19), 3495. https://doi.org/10.3390/w15193495