Sandstone groundwater salinization modelling using physicochemical variables in Southern Saudi Arabia: Application of novel data intelligent algorithms

Abstract

Reliable modelling and simulation of groundwater management are crucial for sustainable development. Groundwater salinization is considered challenging and has recently led to the development of several emerging advancements and technologies, which grant a feasible solution to integrated water management and desalination processes. For this purpose, Electrical Conductivity (EC) as the early Salinization sign is modelled using various computational techniques, namely, Least Square-boost (LSQ-Boost), Gaussian Process regression (GPR), support vector regression (SVR) and stepwise linear regression (SWLR). The experiment data from sandstone aquifers include parameters from the physical, chemical and hydrogeochemical aspects. Four different input combinations (C1-C4) were developed using linear and ranking nonlinear feature selection and validated modelling results weres assessed by mean square error (MSE), mean absolute error (MAE), root means square error (RMSE), Nash-Sutcliffe efficiency (NSE), and correlation coefficient (R). The analysis also considers the effect of multicollinearity, and the variables affected, such as TDS (mg/L), were not included in the first three combinations. The novel GPR proved superior to other models, with GPR-C1 justified quantitatively (MSE = 0.0255, MAE = 25260.49 and RMSE = 0.1595) in the verification phase. Other intelligent models (SVR, LSQ-Boost) depicted promising for C3 and C4 combinations with more than 88–90% predictive accuracy. The explored novel GPR algorithm offered an excellent and reliable EC prediction tool. The study also suggested using direct correlated positive variables, including hydrochemical and topographic factors, in modelling groundwater salinization. This would lead to more effective water-resources-related planning and decision making.