Detecting Floods with GNSS Reflectometry: A CYGNSS-Based Deep Learning Approach for Surface Water Monitoring

This study presents a novel deep learning framework for mapping global surface water using GNSS-R data from CYGNSS constellation. The method combines Delay Doppler Maps (DDMs) with ancillary geophysical parameters to train separated models against the Landsat-based Global Surface Water (GSW) product. These separated models are studied for each land-cover in the Global Lakes and Wetlands Database (GLWD v2). Global water/land classifications at 3km monthly and daily maps are produced and evaluated against external datasets. Three major flood events are studied: the 2022 Pakistan floods, the 2022 eastern Australia floods, and the 2024 Rio Grande do Sul floods in Brazil.