Enhanced remote sensing and deep learning aided water quality detection in the Ganges River, India supporting monitoring of aquatic environments

Abstract

Water pollution is a pressing global concern, impacting numerous communities across the world. Existing water quality monitoring systems rely on static or periodically collected data, presenting limitations in their ability to provide real-time dynamic insights. This research introduces an innovative approach to address this gap—a dynamic data intake system capable of identifying contamination sources, employing remote sensing techniques to track temporal changes, and issuing timely alerts for safeguarding crucial water resources. The proposed system adopts a hybrid methodology, integrating the QAA-v5 algorithm to derive essential parameters. These parameters serve as input for a pre-trained CatBoost model, which facilitates real-time calculations of chlorophyll-a concentrations at specified geographical coordinates. For future forecasting, the system leverages two distinct models: NBeats and CatBoost Time-Series. Notably, the CatBoost model achieves a commendable regression score of 0.985. For a comprehensive assessment and validation of the system's performance, the research draws upon the dataset provided by the International Ocean-Color Coordinating Group (IOCCG). The innovative framework introduced in this study exhibits considerable promise in advancing water quality protection and monitoring, making a significant contribution to the field of environmental research and management.