Deep learning for river water quality monitoring: a CNN-BiLSTM approach along the Kaveri River

• Published in: Sustainable water resources management Vol. 10; no. 3; p. 125
• Main Authors: Geetha, T. S., Chellaswamy, C., Raja, E., Venkatachalam, K.
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.06.2024; Springer Nature B.V
• Subjects: Artificial neural networks; Cloud computing; Contamination; Data collection; Data storage; Data transfer (computers); Decision making; Deep learning; Development Economics; Earth and Environmental Science; Earth Sciences; Environmental monitoring; Field cameras; Hydrogeology; Hydrology/Water Resources; Information storage; Long short-term memory; Machine learning; Neural networks; Original Article; Resource management; River water; Rivers; Sustainability management; Sustainable Development; Sustainable practices; Transfer functions; Water levels; Water monitoring; Water Policy/Water Governance/Water Management; Water quality; Water quality management; Water resources; Water resources management; Water supply; Kaveri River; Deep learning; BiLSTM network; Water quality monitoring; Convolutional neural network
• ISSN: 2363-5037; 2363-5045
• DOI: 10.1007/s40899-024-01102-6

River water plays a crucial role in the development of urban civilizations, necessitating the need for its continuous monitoring, despite the challenges involved. This paper focuses on the application of deep learning techniques to monitor the water quality in rivers. Such monitoring is essential to determine both the quantity and quality of available water, enabling sustainable management practices and informed decision-making. To accomplish this, various water testing stations have been established along the Kaveri River, collecting data from each location. By employing a hybrid approach that combines Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM), the level of contamination introduced into the water supply can be assessed. The data is collected using field cameras placed at the test sites to measure the water levels and a range of sensors to quantify the degree of contamination. A cloud-based web interface facilitates real-time monitoring, data storage, and data transfer functions. The proposed CNN-BiLSTM model was evaluated, resulting in an improved detection accuracy of 4.62%. This approach proves valuable in precisely assessing the quality and quantity of water in rivers, making it a useful tool for water resource management.