Advancing estuarine box modeling: A novel hybrid machine learning and physics-based approach

Abstract

Estuaries play a crucial role in the maintenance of the ecological balance of coastal ecosystems. Salinity intrusion can disrupt these fragile ecosystems, impacting aquatic life and human activities in coastal regions. An accurate prediction of salinity intrusion is essential for managing water resources and preserving ecosystems. This paper introduces a novel hybrid tool, called Hybrid-EBM model, designed to predict the salt-wedge intrusion length and the salinity at river mouth of an estuary. Combining the state-of-the-art Estuary Box Model (EBM) with machine learning algorithms, the new Hybrid-EBM model provides an accurate forecast of the salinity intrusion events. Experimental results highlight the effectiveness of Hybrid-EBM in salinity prediction with an RMSE of 3.41 psu against the 4.22 obtained by EBM. The outputs of this paper represent a significant advancement in the understanding of the impacts of salinity intrusion along the estuarine ecosystems, contributing to the sustainability of the coastal regions worldwide. • Salt-wedge intrusion into estuaries can seriously damage these ecosystems. • A new hybrid-model is proposed to predict salt-wedge intrusion events into estuaries. • The presented hybrid-model combines machine learning and physics-based models. • This hybrid model outperforms the state-of-the-art models. • The study provides valuable insight into salt-wedge intrusion impact on ecosystems.