Signal Coherence and Water Detection Algorithms for the ESA HydroGNSS Mission
Jilun Peng, Weiqiang Li, Estel Cardellach, Gabrielle Marigold, Maria-Paola Clarizia
Abstract
The algorithms to detect the presence of water on land surfaces using the Global Navigation Satellite System (GNSS) reflected signals collected aboard of the future ESA Scout 2-satellite mission HydroGNSS are presented. HydroGNSS will be ready for launch in H2/2024, into polar orbits, and it will operate at dual frequency, dual-polarization, and in two simultaneous acquisition modes (low-rate power and high-rate complex signal modes). The overall strategy to generate level-2, point-by-point along track water classification using all these signals is introduced, yet currently available datasets only permit the validation of the algorithms applied to one single frequency and polarization. Two signal coherence indicators are selected per each acquisition mode, and together with geolocation, large-scale surface roughness, and land cover type are the inputs of the random forest classifier, which is an ensemble learning method, with the monthly Global Surface Water (GSW) as a reference target. The algorithms are tested with NASA/CYGNSS low-rate power delay-Doppler maps (DDMs) and with CYGNSS and U.K./TechDemoSat-1 (TDS-1) raw intermediate frequency (IF) sampled signals. The results of the validation are analyzed at different scales, with results achieving the mission’s 90% accuracy requirement. Final retraining and validation using actual dual-polarization and dual-frequency HydroGNSS data will be conducted once the satellites are in-orbit.