Tidal analysis of GNSS reflectometry applied for coastal sea level sensing in Antarctica and Greenland

• Published in: Remote sensing of environment 2020-10, Vol.248, p.111959, Article 111959
• Main Authors: Tabibi, Sajad, Geremia-Nievinski, Felipe, Francis, Olivier, van Dam, Tonie
• Format: Article
• Language: English
• Subjects: Altimetry; Antarctica; Antennas; Coastal waters; Constituents; Correlation; Diurnal; Diurnal variations; Global navigation satellite system; Global positioning systems; GNSS; GNSS-R; GPS; Greenland; Model accuracy; Ocean models; Ocean surface; Ocean tides; Oceans; Orbits; Polar environments; Polar regions; Reflectometry; Root-mean-square errors; Satellite observation; Sea ice; Sea level; Signal to noise ratio; SNR; Terrestrial environments; Tidal analysis; Tidal harmonics
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2020.111959

We retrieve sea levels in polar regions via GNSS reflectometry (GNSS-R), using signal-to-noise ratio (SNR) observations from eight POLENET GNSS stations. Although geodetic-quality antennas are designed to boost the direct reception from GNSS satellites and to suppress indirect reflections from natural surfaces, the latter can still be used to estimate the sea level in a stable terrestrial reference frame. Here, typical GNSS-R retrieval methodology is improved in two ways, 1) constraining phase-shifts to yield more precise reflector heights and 2) employing an extended dynamic filter to account for the second-order height rate of change (vertical acceleration). We validate retrievals over a 4-year period at Palmer Station (Antarctica), where there is a co-located tide gauge (TG). Because ice contaminates the long-period tidal constituents, we focus on the main tidal species (daily and subdaily), by employing a deseasonalization filter. The difference between sub-hourly GNSS-R retrievals of the ocean surface and TG records has a root-mean-square error (RMSE) of 15.4 cm and a correlation of 0.903, while the tidal prediction has a RMSE of 1.9 cm and a correlation of 0.998. There is excellent millimetric agreement between the two sensors for most eight major tidal constituents, with the exception of luni-solar diurnal (K1), principal solar (S2), and luni-solar semidiurnal (K2) components, which are biased in GNSS-R due to the leakage of the GPS orbital period. We also compare the GNSS-R tidal constituents from seven additional POLENET sites, without co-located TG, to global and local ocean tide models. We find that the root-sum-square-error (RSSE) of eight major constituents varies between 26.0 cm and 56.9 cm for different models. Given that the agreement in tidal constituents between the TG and GNSS-R was better at Palmer Station, we conclude that assimilating the GNSS-R retrievals into tidal models would improve their accuracy in Antarctica and Greenland, provided that care is exercised to avoid the orbital period overtones and also sea ice. •GNSS Reflectometry (GNSS-R) is used to estimate tidal constituents in Antarctica and Greenland.•There are no major systematic biases between GNSS-R and tide gauge (TG) in polar regions.•GNSS-R and TG daily and subdaily constituents agree at the mm level.•GNSS-R sea level time series and tidal constituents will be publicly available.•GNSS-R retrievals could improve accuracy of ocean tide models.