Utilizing Normal Time-Frequency Transform-Assisted GNSS-R to Retrieve Sea Surface Height

Global navigation satellite system reflectometry (GNSS-R), a new remote sensing technology, has been used to monitor sea surface height (SSH) in recent years. In the process of SSH retrieval by GNSS-R, estimating the oscillation frequency of signal-to-noise ratio (SNR) data is particularly critical. However, the sampling time of the commonly used Lomb-Scargle periodogram (LSP) is too low, and the wavelet transform (WT) is sensitive to noise and has a frequency shift in the non-L1 norm. To solve these problems, we propose a new method using the normal time-frequency transform (NTFT) to assist GNSS-R in retrieving SSH. NTFT corrects the frequency offset in the WT and has better noise resistance. We applied it to the SSH retrieval of three GPS sites (SC02, AT01, and BRST) and compared the results with those of LSP and WT. The results show that except for the SC02 site, where the retrieval accuracy of NTFT is slightly lower than that of LSP, the accuracy and quantity of NTFT retrieval are better than those of LSP and WT at the other two sites. At the AT01 site, the accuracy of NTFT is 38.5% and 35.2% higher than that of LSP and WT, respectively, which verifies the accuracy of NTFT in determining the instantaneous frequency and anti-noise. We also performed NTFT analysis on SSH retrieved via GNSS-R to obtain the time-varying harmonic constants of the main tidal components. Compared with LSP and WT, the NTFT retrieval of high-frequency tidal components is significantly more accurate than that of low-frequency components.