Automated Generation of Lakes and Reservoirs Water Elevation Changes From Satellite Radar Altimetry

Limited access to in-situ water level data for lakes and reservoirs have been a major setback for regional and global studies of reservoirs, surface water storage changes, and monitoring the hydrologic cycle. Processing satellite radar altimetry data over inland water bodies on a large scale has been a cumbersome task primarily due to the removal of contaminated measurements as a result of surrounding land. In this study, we proposed a new algorithm to automatically generate time series from raw satellite radar altimetry data without user intervention. With this method, users with a little knowledge on the field can now independently process radar altimetry for diverse applications. The method is based on K-means clustering, interquartile range, and statistical analysis of the dataset for outlier detection. Jason-2 and Envisat radar altimetry data were used to demonstrate the capability of this algorithm. A total of 37 satellite crossings over 30 lakes and reservoirs located in the U.S., Brazil, and Nigeria were used based on the availability of in-situ data. We compared the results against in-situ data and root-mean-square error values ranged from 0.09 to 1.20 m. We also confirmed the potential of this algorithm over rivers and wetlands using the southern Congo River and Everglades wetlands in Florida, respectively. Finally, the different retracking algorithms in Envisat; Ice-1, Ice-2, Ocean, and Sea-Ice were compared using the proposed algorithm. Ice-1 performed best for generating water level time series for in-land water bodies and the result is consistent with previous studies.