Hydrogeochemical characteristics prediction using multiple linear regression: case study on unconfined aquifer in northeastern Tunisia

Abstract

The main factors and mechanisms controlling the groundwater chemistry and mineralization are recognized through hydrochemical data. However, water quality prediction remains a key parameter for groundwater resources management and planning. The geochemical study of groundwater of a multilayered aquifer system in Tunisia is recognized by measurements of the pH, EC, total dissolved solids (TDS), major ion concentration and nitrates of 36 samples from pumping wells covering the aquifer extension and analyzed using standard laboratory and field methods. The calcite precipitation, gypsum, anhydrite and halite dissolution, and direct and reverse ion exchange are the principal process of chemical evolution in the Nadhour-Saouaf aquifer system. Using stepwise regression, the concentration groups of (Ca, Cl, and NO3), (Cl, SO4, and Mg), and (Ca and Na) exhibit significant prediction of TDS in Plio-Quaternary, Miocene, and Oligocene aquifer levels, respectively. The highest values of R 2 and adjusted R 2 close to 1 revealed the accuracy of the developed models which is confirmed by the weak difference between the measured and estimated values varying between −12 and 8%. The important uncertainty parameters that affected the estimated TDS are assessed by the sensitivity analysis method. The concentration of (Cl), (Ca and Cl), and (Na) are the major parameters affecting the TDS sensitivity of the Plio-Quaternary, Miocene, and Oligocene aquifer levels, respectively. Hence, the developed TDS models provide a more simple and easy alternative to other methods used for groundwater quality estimation and prediction as proven from external validation on groundwater samples unconsidered in the model construction.