Abstract
We seek to quantify and understand the residual signal in GPS and GLONASS estimates of ocean tide loading displacements (OTLDs) after removing state-of-the-art model estimates. To consider contributions over a broad spatial scale, we estimate OTLD over the Australian continent using $$\sim $$ ∼ 5.5 years of continuous GPS and GLONASS data from 360 sites. We compare these with modelled estimates, with a focus on the lunar semidiurnal M $$_2$$ 2 and diurnal O $$_1$$ 1 constituents. We observe spatially coherent patterns of residual OTLD in each of the east, north, and up coordinate components after the removal of tidal loading using elastic models. We subsequently assess the impact of including anelastic dispersion in the model and show a 0.2 mm reduction in range of the up component residuals at coastal sites. A similar reduction at all sites is observed in the east and north components. Of the seven ocean tide models used, we find that three recent models, FES2014b, GOT4.10c and TPXO9.v1, perform similarly, noting these comparisons are made in the CE frame. However, we show that the latter contains centre of mass (CoM) biases in amplitude up to 0.2 mm and 0.5 mm for M $$_2$$ 2 and O $$_1$$ 1 , respectively, due to the assimilated altimetry data having not been corrected for geocentre motion. We find OTLD estimates are sensitive to the chosen orbit and clock products used in our analysis, with differences of up to 0.5 mm in the east component between solutions using the JPL and either of ESA or CODE products (GPS-only). Our analysis shows that current GNSS estimates of OTLD over Australia are typically accurate to $$\sim $$ ∼ 0.2 mm at which point we are unable to explain spatially coherent residuals when compared to modelled quantities. These may depend on the appropriate treatment of the CoM variation, anelasticity and/or three-dimensional Earth structure. As such, we recommend great care is taken when interpreting OTLD at the level of 0.1 or 0.2 mm, even if it is regionally coherent.