Prediction of river water salinity using water quality indices

Abstract

Salinity intrusion within freshwater systems poses serious threats to agriculture, ecosystems, and even the socio-economic stability of places near coasts, where these changes have occurred due to climate change. Salinity monitoring has been made possible through different approaches, which are usually resource-intensive and time-consuming; therefore, applying such techniques does not allow for their common use during large-scale, real-time assessments. This research investigates the use of remote sensing-based water quality indices to forecast the salinity of river water as a more cost-effective and scalable alternative. Using Sentinel-2 satellite images and Google Earth Engine, four vegetation indices—Normalized Difference Water Index (NDWI), Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Normalized Difference Red Edge (NDRE)—were analyzed to test their relationship with in-situ salinity measurements collected in Giao Thuy, Vietnam. Among the indices tested, EVI showed the highest but weak correlation with salinity levels and was employed to develop a predictive model. The salinity distribution map generated by the model indicates that the salinity level in-creased further inland. However, issues such as cloud interference, inconsistencies in data recorded, and omitting other environmental factors reduced the model's accuracy. Future research should employ machine learning techniques, higher-resolution satellite images, and many influencing factors to accomplish this goal.