This paper presents a new method for retrieving sea surface heights from Global Navigation Satellite Systems reflectometry (GNSS‐R) data by inverse modeling of SNR observations from a single geodetic receiver. The method relies on a B‐spline representation of the temporal sea level variations in order to account for its continuity. The corresponding B‐spline coefficients are determined through a nonlinear least squares fit to the SNR data, and a consistent choice of model parameters enables the combination of multiple GNSS in a single inversion process. This leads to a clear increase in precision of the sea level retrievals which can be attributed to a better spatial and temporal sampling of the reflecting surface. Tests with data from two different coastal GNSS sites and comparison with colocated tide gauges show a significant increase in precision when compared to previously used methods, reaching standard deviations of 1.4 cm at Onsala, Sweden, and 3.1 cm at Spring Bay, Tasmania.

No abstract

No abstract

Current GNSS-R (GNSS reflectometry) techniques for sea surface measurements require data collection over longer periods, limiting their usability for real-time applications. In this work, we present a new, alternative GNSS-R approach based on the unscented Kalman filter and the so-called inverse modeling approach. The new method makes use of a mathematical description that relates SNR (signal-to-noise ratio) variations to multipath effects and uses a B-spline formalism to obtain time series of reflector height. The presented algorithm can provide results in real time with a precision that is significantly better than spectral inversion methods and almost comparable to results from inverse modeling in post-processing mode. To verify the performance, the method has been tested at station GTGU at the Onsala Space Observatory, Sweden, and at the station SPBY in Spring Bay, Australia. The RMS (root mean square) error with respect to nearby tide gauge data was found to be 2.0 cm at GTGU and 4.8 cm at SPBY when evaluating the output corresponding to real-time analysis. The method can also be applied in post-processing, resulting in RMS errors of 1.5 cm and 3.3 cm for GTGU and SPBY, respectively. Finally, based on SNR data from GTGU, it is also shown that the Kalman filter approach is able to detect the presence of sea ice with a higher temporal resolution than the previous methods and traditional remote sensing techniques which monitor ice in coastal regions.

Global Navigation Satellite System Reflectometry (GNSS-R) tide gauges are a promising alternative to traditional tide gauges. However, the precision of GNSS-R sea level measurements when compared to measurements from a co-located tide gauge is highly variable, with no clear indication of what causes the variability. Here we present a modelling technique to estimate the precision of GNSS-R sea level measurements that relies on creating and analyzing synthetic Signal-to-Noise-Ratio (SNR) data. The modelled value obtained from the synthetic SNR data is compared to observed RMSE between GNSS-R measurements and a co-located tide gauge at five sites and using two retrieval methods: spectral analysis and inverse modelling. We find that the inverse method is more precise than the spectral analysis method by up to 60% for individual measurements but the two methods perform similarly for daily and monthly means. We quantify the contribution of dominant effects to the variations in precision and find that noise is the dominant source of uncertainty for spectral analysis whereas the effect of the dynamic sea surface is the dominant source of uncertainty for the inverse method. Additionally, we test the sensitivity of sea level measurements to the choice of elevation angle interval and find that the spectral analysis method is more sensitive to the choice of elevation angle interval than the inverse method due to the effect of noise, which is greater at larger elevation angle intervals. Conversely, the effect of tropospheric delay increases for lower elevation angle intervals but is generally a minor contribution.