Summary Geocenter motion, defined as the displacement of Earth’s center of mass relative to its center of figure, is crucial for maintaining the International Terrestrial Reference Frame origin and quantifying large-scale mass redistribution. However, whether observing geocenter motion by tracking satellite orbits or inferring it using geophysical models, accurately acquiring such subtle motions imposes stringent requirements on the consistency and precision of both tracking data and geophysical models. This study improves geocenter motion estimates derived from the combination of GRACE/GRACE-FO time-variable gravity (TVG) and Ocean Bottom Pressure (OBP) models (the GRACE-OBP method) in two ways. First, we apply a forward modelling technique to mitigate land–ocean leakage in GRACE/GRACE-FO TVG fields, which demonstrably outperforms empirical coastline buffer-zone corrections in controlled simulation experiments. Second, we introduce the Bayesian Three-Cornered Hat (BTCH) method to optimally combine geocenter series derived from multiple GRACE solutions and two independent OBP models (ECCO2 and MPIOM), producing an improved geocenter product without requiring a ground-truth reference. Uncertainty analysis shows that the noise level is governed primarily by the GRACE solution, and that BTCH provides a clearer advantage over equal-weighted averaging when the number of input series is limited, reducing the noise level by about 30 per cent. After restoring atmospheric and oceanic contributions, our improved geocenter series shows good agreement with the CSR SLR-derived geocenter product. Although uncertainty levels vary among individual solutions, the estimated annual and secular trend signals are broadly consistent and show limited sensitivity to the choice of GRACE TVG solution and OBP model. Using the improved geocenter series, we revisit the annual geocenter oscillation and its drivers; the results indicate that cryospheric mass variability and land-ocean mass exchange (i.e. sea-level fingerprints) provide non-negligible contributions to the annual geocenter cycle and improve consistency with observations. Finally, the improved geocenter series yields the lowest uncertainty in degree-1 mass variations, with a global RMS of 0.55 mm. Incorporating these degree-1 terms into mass budget assessments yields secular trends of 38.8 Gt/yr for the Antarctic Ice Sheet and 0.57 mm/yr for global mean ocean mass, highlighting the need for accurate geocenter corrections to support reliable long-term climate monitoring.
Yu et al. (Fri,) studied this question.