Machine Learning Models for Monitoring Salinity in River Estuaries: A Case Study of the Po River

Saltwater intrusion into river estuaries is a pressing environmental and socio-economic concern, posing a threat to freshwater ecosystems, agriculture, and coastal communities' water resources. Despite this, estuaries remain inadequately monitored today. The present study propose a multi-branch machine learning approach to predict the estuaries' salinity. A comprehensive learning dataset was constructed using an unstructred grid model, named SHYFEM, focusing on the Po River branches and spanning the year 2018. Machine learning algorithms including Support Vector Machine (SVM) and Random Forest (RF) were chosen as the primary models for analysis. SVM emerged has top-performing model, with an RMSE of 0.976 psu, MAE of 0.576 psu and an R2 score of 0.925. The proposed methodology provides valuable insights for monitoring and managing salinity intrusion in Po River region. However, its applicability extends beyond the Po River to other areas facing similar natural and anthropogenic conditions.