Filtering methods to extract the tide height from Global Navigation Satellite Systems (GNSS) signals for Hydrographic applications

Abstract

Hydrographic surveys have traditionally relied on the availability of tide information for the reduction of sounding observations to a common (chart) datum usually related to a specific tide level. In most cases, tide information is obtained from tide gauge observations and/or tide predictions using local, regional or global tide models. An emerging method that is increasingly being used is based on Global Navigation Satellite System (GNSS) positioning of the sea surface. This study assesses the suitability of two types of filtering methods to extract the tide signal from GNSS height observations of the water level. The Savitzky-Golay and Gaussian low-pass filters were implemented to eliminate high-frequency components due to waves, dynamic draft and measurement uncertainties. Over a 30-day period, GNSS heights were estimated from GPS and GLONASS carrier phase data collected by a Fugro Starpack GNSS receiver installed on a floating pontoon at Hillarys Boat Harbor, located in Western Australia. Sea level heights recorded by a traditional tide gauge at Hillarys were used as a reference to evaluate the effectiveness of using GNSS water level heights in extracting the tide signal. To assess the filters performance for a range of window lengths, amplitudes and phases of the four major tidal harmonic constituents (M2, S2, K1, and O1) were determined and compared for the unfiltered and filtered water level signals. The study found a high degree of agreement between the unfiltered tide information obtained by GNSS and the tide gauge documented by almost identical amplitudes of the tidal harmonic constituents. There was a correlation coefficient of up to 0.98 and an RMS value of 0.0325 m for the differences satisfying specifications of the International Hydrographic Organization (IHO) for Special Order hydrographic surveys. As a main finding regarding the filter performance, this study shows that the Savitzky-Golay filter suffers considerably less from signal loss (damping of the tidal harmonic constituents) when compared to the Gaussian filter.