Multifeature Water Discrimination Method Based on CYGNSS Intermediate Frequency Data

Changes in wetlands, rivers, and other water resources significantly impact the environment and economy, highlighting the need for their real-time, high-accuracy detection. Current GNSS reflectometry methods use reflected signal amplitude for water body detection, ensuring near real-time capability due to the high revisit period of source satellites. However, these methods face challenges in distance resolution due to limited signal features. This article proposes a method utilizing CYGNSS raw intermediate frequency (IF) data for multidimensional analyses, including amplitude, time delay, phase, and frequency, to improve water body discrimination. The high sampling rate of IF raw data enables a minimum interval of 3 m between sampling points, enhancing distance resolution. The preprocessing of raw IF data involves creating complex correlation matrices and applying three compensation methods: direct signal delay, reflected signal frequency, and Doppler drift correction. Features such as amplitude, delay, phase, and frequency are extracted from the matrices using techniques like complex amplitude calculation, peak localization, phase extraction, and wavelet transforms. Various discrimination methods, including threshold, normalized variance, Z-score, variance, and circular correlation (Cir Cor) discriminations, are used based on these features. The fusion of these indicators provides the final water body discrimination results. Case studies in the Amazon River area show that this multifeature method achieves high discrimination accuracy (~90%) and provides high temporal resolution (20 Hz). This approach offers high precision in water body retrieval, particularly for smaller water bodies, reducing the distance resolution from the kilometer level to the meter level.