Improving water level monitoring using CYGNSS DDMs and machine learning techniques

Water level monitoring plays a vital role in hydrology, as it not only facilitates the optimization of water resource management but also supports flood disaster early warning. Traditional methods mostly rely on in situ water level and empirical curves, which are unsuitable for complex water bodies. The CYGNSS (Cyclone Global Navigation Satellite System) satellite system offers unique sensitivity to surface water, showing significant advantages in water level monitoring. This study leverages the statistical features of CYGNSS Delay-Doppler Maps (DDMs) to develop a water level estimation model based on machine learning (ML) techniques. Evaluations against in situ data show the model accurately estimates water level variations, with the Root Mean Square Error (RMSE) improving from 0.735 m to 0.470 m (~36% decrease) after incorporating DDM features. To further evaluate model performance, additional tests were conducted on independent spatiotemporal stations, which confirmed its robustness and generalizability. This study improves real-time water level monitoring accuracy and offers novel approaches for regional and global hydrological monitoring and disaster early warning.