Predicting saltwater intrusion in Sebou estuary (Morocco) using Savenije model enhanced by Grey Wolf Optimization metaheuristic algorithm

In estuarine regions, saltwater intrusion is a major global challenge affecting water quality and freshwater resources. This study develops a hybrid framework coupling the Savenije analytical model with the Grey Wolf Optimizer (GWO) metaheuristic to improve salinity prediction in river estuaries under low-water slack (LWS) and high-water slack (HWS) tidal conditions. The key model parameters (dispersion and Van der Burgh coefficients) were optimized using GWO implemented in Python. The Savenije–GWO model achieved strong predictive performance, with a determination coefficient (R2) of 0.99, a Nash–Sutcliffe efficiency (NSE) of 0.99, a Normalized Objective Function (NOF) of 0.055 (LWS) and 0.047 (HWS), and a Root Mean Square Error (RMSE) of 0.22 g/L (LWS) and 0.59 g/L (HWS). Percent Bias (PBIAS) values of 3.28% (LWS) and −0.47% (HWS) confirm minimal systematic bias, demonstrating the model's robustness and outperforming previously reported analytical models based on gradient-based optimizers (Levenberg-Marquardt). These results confirm strong agreement between observed and predicted salinity values and highlight GWO's effectiveness in enhancing prediction accuracy and stability. The framework supports saltwater intrusion prevention and informed decision-making in estuarine water management and is transferable to other, particularly data-limited, estuarine systems.