A Hybrid Modeling Framework for Designing Subsurface Barriers to Mitigate Seawater Intrusion in Coastal Aquifers

Seawater intrusion (SWI) poses a significant threat to coastal aquifers, leading to the deterioration of groundwater quality and undermining long-term water resource sustainability. Physical barriers, such as cutoff walls and subsurface dams, represent a viable mitigation strategy; however, their performance is highly dependent on site-specific design, which necessitates robust scenario-based simulation. In this study, a coupled Python-FEFLOW modeling framework was developed to dynamically assess the hydraulic and salinity responses of various physical barrier configurations, including both top-down and bottom-up implementations. The model quantifies key system responses, namely total dissolved salts (S) within the domain and seawater influx (Qin) across boundaries. Application of the model to the classical Henry problem, under uniform domain properties, revealed that cutoff walls achieved up to a 98% reduction in S and a 76.5% decrease in Qin. Subsurface dams were similarly effective, yielding reductions of approximately 92% in S and 81% in Qin. Notably, barrier placement proximal to the saline boundary led to stagnation zones and distortion of salinity contours, highlighting the importance of strategic positioning. To complement the simulation-based analysis, a hybrid machine learning approach, integrating Random Forest regression with SHAP analysis, was employed to perform sensitivity analysis and identify the most influential design parameters in a computationally efficient manner. This integrated modeling and machine learning framework offers a novel and scalable tool for optimizing physical barrier design This approach enabled a comprehensive and automated evaluation of physical barriers and their effectiveness in mitigating SWI. It also assessed the relative influence of physical barriers design parameters on SWI dynamics. The proposed methodology demonstrates significant potential for enhancing field-scale SWI mitigation strategies.