RII Track-4:@NASA: Combining Physics and Deep Learning for Accurate River Discharge and Bathymetry Estimation from the Surface Water and Ocean Topography Mission

点击翻译按钮获取中文摘要

Estimation of river bottom elevation and discharge plays an essential role in many practical applications such as safe and efficient maritime transportation, flood risk management, and water management planning. However, in-situ measurement of river depth and discharge is logistically challenging, expensive, and time-consuming. Recently launched Surface Water and Ocean Topography (SWOT) satellite mission opens a new possibility to estimate river discharge and depth measurements by providing the river water elevation above sea level, river width, and slope over the watershed to continental scales. This proposed research will enable the development and application of large-scale accurate river depth measurements and discharge estimation using the recently acquired SWOT data sets. The research will also focus on the effect of surface-groundwater interaction on river discharge estimation. The fellowship program will train the PI and a graduate student from the University of Hawaii at Manoa in both technical aspects of SWOT data processing and analysis as well as computational approaches for real-time water depth estimation. The collaborative research will ultimately improve the research capacity of the home institution and benefit the State of Hawaii, which is the state that resorts to groundwater resources and is subject to both water management and balance.This EPSCoR Research Infrastructure Improvement (RII) Track-4: EPSCoR Research Fellows (RII Track-4:@NASA) will provide a fellowship to an Associate Professor and training for a graduate student at the University of Hawaii at Manoa. This work would be conducted in collaboration with researchers at NASA Jet Propulsion Laboratory (JPL). The PI and one graduate student will visit the Terrestrial Hydrology group at NASA JPL to learn the SWOT mission data acquisition and associated river discharge estimation from experts in the field. The PI has actively developed river dynamics-based machine-learning techniques to estimate river bathymetry and discharge and their corresponding uncertainties in a computationally efficient unified framework. The main objectives of this research study are to advance the currently used methods to account for the surface-groundwater interaction, improve the accuracy of the discharge estimation, and develop a new data assimilation method to estimate spatiotemporal river bathymetry and discharge using the SWOT data. For close-to-real-time bathymetry estimation, the shallow water equations will be approximated through physics-informed neural networks with a controlled accuracy for real-time simulation. The proposed methods will be packaged in a Python library, with the intent of releasing them in a public repository once properly tested and quality checked. Expertise gained during this fellowship will enable the PI to train and educate students at the University of Hawaii at Manoa in cutting-edge remote sensing data analysis and machine learning-based assimilation techniques.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.