Groundwater modelling and GIS-based vulnerability mapping coupled with evolutionary metaheuristic optimization in the eastern coast of Saudi Arabia

Developing an efficient and reliable intelligent approach to the available groundwater (GW) resources appears crucial for achieving Saudi Vision 2030 on the availability of freshwater resources, the prosperity of people, and economic development. The present study is based on a real-field investigation and experimental analysis using ion chromatography (IC) and inductively coupled plasma mass spectrometry (ICP-MS). Subsequently, ArcGIS 10.3 software and artificial intelligence (AI)-based metaheuristic optimization (MO) were used to create vulnerability maps and a modelling schema for the potassium (K⁺) and sodium (Na⁺) in the coastal region of eastern Saudi Arabia, respectively. For this purpose, extreme gradient boosting (XG-Boost) was used as a standalone model while differential evolution (DE) and firefly algorithms (FA) as optimization techniques. The results were validated using different statistical indices and graphical visualization. The optimal objective function for each data set through multiple iterations based on the root means square error (RMSE) index and the number of features was done using DE algorithms. The performance results of the optimized XGBoost algorithm (DE-XGBoost and FA-XGBoost) and the XGBoost algorithm indicated that FA algorithms outperformed merit with high accuracy for both K⁺ and Na⁺. The numerical comparison depicted that the mean absolute error (MAE) for K⁺ and Na⁺ FA-XGBoost was 0.0173 and 0.028, respectively. The results showed that the FA-XGBoost method produced more accurate K⁺ and Na⁺ prediction GIS-maps than the other two algorithms. Hence, the current results justified the potential use of the intelligent tool for water resources management.