The potential of GNSS-IR for monitoring daily water level fluctuations up to 15 m

Global Navigation Satellite System Interferometric Reflectometry (GNSS-IR) offers an alternative "remote sensing water level gauge" service. Current GNSS-IR water level monitoring relies on the Lomb-Scargle periodogram (LSP) or Windows LSP (WinLSP) methods, which estimate water levels by extracting one dominent frequency from one Signal-to-Noise Ratio (SNR) arc. These classic GNSS-IR methods mainly focus on quasi-static or metrescale scenarios. However, in cases of rapid and large water level changes, the SNR sequences transform into nonstationary signals, rendering classic GNSS-IR methods ineffective. To address the challenges posed by nonstationary SNR sequences, we propose a new GNSS-IR water level monitoring approach that integrates wavelet analysis and cluster analysis. We validated the performance of this method using simulation and field data from the Xilongchi Pumped Storage Power Station (38 degrees 32'31"N, 113 degrees 16'49"E). The results indicate that the classical WinLSP method is almost ineffective, with only 52 % of the water level retrievals differing from the water level gauge measurements within 1 m. In contrast, the proposed method achieves a precision of 30 cm in tracking daily water level fluctuations of approximately 15 m. Furthermore, the Pearson correlation coefficients between GNSS-IR retrievals and water-level gauge measurements are greater than 0.99. This study pioneers monitoring of daily water level fluctuations in pumped storage power stations via GNSS-IR, filling a gap in scenarios involving daily water level changes up to 15 m.