An Improved Height Rate Correction Method Based on Robust Regression for Sea Level Estimation in GNSS Interferometry Reflectometry

Abstract

Global Navigation Satellite System-Interferometry Reflectometry (GNSS-IR) sea level altimetry, which is based on analysis of GNSS signals reflected from the sea surface, has demonstrated unique advantages for sea level monitoring. The signal-to-noise ratio (SNR) of a geodetic GNSS receiver can be used to estimate the distance between antenna and sea surface called reflecting height (RH), and subsequently to retrieve sea level. The classical SNR analysis method uses the multipath frequency of the SNR to estimate RH by assuming the sea is static within a limited period. Then, the bias caused by the moving sea surface is corrected by the height variation rate \( \dot{h} \) from the dynamic surface and by the elevation angle variation rate \( \dot{e} \) from the dynamic satellite. However, this method cannot correct this height rate error completely because \( \dot{h} \) cannot be calculated accurately from the raw RH series. Recently, a dynamic SNR method has been developed that can estimate \( h \) and its variation rate \( \dot{h} \) concurrently from the variational multipath frequency based on the least square method. However, this method has no ability to avoid error, leading to many outliers in retrieval series. So, we introduced the robust regression method to improve this method to avoid errors. The different performance of the classical SNR method and the improved dynamic SNR method is firstly analyzed and compared, using the SNR data from the Kachemak Bay GPS site. The results showed the retrievals of improved dynamic method based on robust regression achieves a higher accuracy than that of classical method. And it has the potential for short-term monitoring (waves and individual days when an event such as a large storm surge might occur).