Society Of Marine Science and Technology International Journal Of Coastal, Offshore And Environmental Engineering(ijcoe) 2980-8731 Articles in Press 2026 08 15 Hybrid SUTRA–Machine Learning Modeling of Seawater Intrusion and Salinization Dynamics in the Esteghlal Minab Dam–Coastal Aquifer System 250742 10.22034/ijcoe.2026.591349.1261 EN Mehrdad Jafari Department of Civil Engineering, Marvdasht Branch, Islamic Azad University, Marvdasht, Iran 0009-0005-0186-0170 Saeed Khorram Department of Civil Engineering, Marvdasht Branch, Islamic Azad University, Marvdasht, Iran 0000-0001-9127-9649 Hossein Khorshidi Assistant professor, Department of Civil Engineering, Marvdasht Branch, Islamic Azad University, Marvdasht, Iran Abolghasem Sayadi Assistant professor, Department of Civil Engineering, Marvdasht Branch, Islamic Azad University, Marvdasht, Iran Journal Article 2026 07 13 Seawater intrusion and salinization of coastal aquifers are critical threats to freshwater security in arid and semi-arid coastal regions. The Minab plain in southern Iran, hydraulically affected by the Esteghlal Minab Dam, intensive groundwater abstraction, reduced inland recharge, and proximity to the Strait of Hormuz, represents a vulnerable hydrogeological system in which reservoir operation and coastal groundwater dynamics interact. This study develops a hybrid physically based and data-driven framework for simulating seawater intrusion and salinization dynamics in the Esteghlal Minab Dam–coastal aquifer system. The numerical core is based on SUTRA, a variable-density groundwater flow and solute transport model, adapted from a validated coastal reservoir–aquifer conceptual framework. The model incorporates inland groundwater input, reservoir–aquifer exchange, tidal forcing, density-driven flow, pumping stress, recharge variability, and sea-level fluctuation. Climatic and management scenarios are designed to evaluate salinity distribution, salt mass, exchange fluxes, and seawater intrusion length under drought, increased abstraction, reservoir-level variation, sea-level rise, and combined stress conditions. To reduce computational cost and support uncertainty analysis, XGBoost and CatBoost are trained as surrogate models using SUTRA-generated outputs. Their performance is assessed using RMSE, MAE, NSE, and R2. The proposed framework also employs Monte Carlo-based uncertainty analysis to identify dominant controls on aquifer salinization. The results indicate that reduced inland recharge and increased abstraction intensify seawater intrusion, whereas maintaining adequate reservoir levels and controlled pumping can limit salinity migration. The hybrid SUTRA–machine learning approach provides a rapid, interpretable, and decision-oriented tool for sustainable management of coastal groundwater systems affected by dam operation and marine forcing.