Research Projects

Water resource management and hydrological modeling are crucial for addressing water scarcity, flood management, and climate adaptation challenges. This is especially true in countries like India, where large regions face significant challenges due to data scarcity, limited hydrological monitoring infrastructure, and complex climatic and topographic conditions. Traditional hydrological models, reliant on in-situ observations, often struggle in areas with sparse/obsolete monitoring networks. The recent launch of NASA's Surface Water and Ocean Topography (SWOT) satellite in December 2022 provides high-resolution, global surface water data, offering new insights into surface water dynamics. However, the accuracy of SWOT data could be limited by spatiotemporal biases due to factors like river morphology, topography, and weather, which reduce its reliability. This project primarily addresses these gaps by evaluating SWOT data for estimating surface water information, focusing on quantifying and correcting these biases with respect to ground-based observations. This would significantly be useful for hydrological applications, including flood mapping and water resource management. While SWOT satellite data can provide valuable insights into surface water dynamics, its predictive capacity is limited due to its long revisit cycle (~21 days) and uncertainties associated with data retrieval. Therefore by developing a Distributed Physics-Informed Deep Learning (DPIDL) hydrologic model by integrating SWOT and ground observation, we aim to address this research gap. The DPIDL model combines process-based hydrological knowledge with machine learning techniques, offering more accurate predictions of river discharge. Its scalable and interpretable framework is adaptable to different regions across India and potentially to other developing countries. Catchment classification and optimal stream gauge network design are crucial for improving watershed monitoring and water resource management. Existing networks often face limitations due to obsolete infrastructure, inadequate coverage, and inefficiencies in data collection. By leveraging geomorphological, climatic, and hydrological attributes (derived from bias corrected SWOT measurements) along with simulated discharge information from the DPIDL model, this project aims to develop a robust catchment classification framework. This framework, combined with advanced statistical techniques such as Entropy, Bayesian concepts, and Multi-objective optimization, will enable the efficient expansion and redesign of stream gauge networks. The approach will enhance network coverage, reduce redundancy, and optimize data collection, ultimately improving the accuracy and reliability of hydrological monitoring. The proposed methodologies will be demonstrated in the Cauvery River Basin, chosen for its significance in water resource management, seasonal streamflow variability, and representative upstream-downstream interactions.