Machine learning-enhanced GALDIT modeling for the Nile Delta aquifer vulnerability assessment in the Mediterranean region

Abstract

Mega-delta aquifers face increasing salinization risks from overexploitation and erratic climate change globally. This study integrates the GALDIT framework with machine learning (ML) models, namely Support Vector Machine (SVM), Generalized Linear Model (GLM), and eXtreme Gradient Boosting (XGBoost), to enhance delta aquifer vulnerability (DAV) assessment to seawater intrusion (SWI). The Nile Delta, the largest freshwater mega-delta aquifer, serves as a case study. Grid search hyperparameter optimization was applied to refine these models using the GALDIT factors (groundwater occurrence, aquifer hydraulic conductivity, groundwater height above sea level, distance from the shoreline, impact of existing seawater intrusion, and aquifer thickness) and adjust conditioned vulnerability index (CVI) based on Total Dissolved Salts (TDS) as input variables. Statistical metrics, including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Mean Squared Error (MSE), Coefficient of Determination (R²), Pearson Correlation Coefficient (r), Nash–Sutcliffe Efficiency (NSE), Root Mean Square Error to Standard Deviation of Observations (RSR), and Index of Scatter (IOS), show that the XGBoost model significantly outperforms SVM and GLM, with exceptional results: R² = 0.9622, RMSE = 0.0430, r = 0.9815, MAE = 0.0206, MSE = 0.0018, NSE = 0.9618, RSR = 0.0005, and IOS = 0.2935. The GALDIT_XGBoost map identified previously undetected high-vulnerability areas west of Alexandria and localized pockets within southern Port Said along the Mediterranean coast. The moderate vulnerability zone expanded, especially in northern Ismailia, compared to the basic GALDIT. Piper diagrams confirmed SWI risks, with dominant Na-Cl and Ca-Mg-Cl facies indicating elevated Cl⁻ and SO₄²⁻ levels. A shift from HCO₃⁻ to Cl⁻ further validated salinization, while Ca-HCO₃ facies represented freshwater. The optimized XGBoost model offers a robust tool for managing mega-delta groundwater and assessing global delta vulnerabilities.