Three decades of GNSS-derived geocenter motion: disentangling geophysical signal from systematic errors

This study presents a long-term evaluation of geocenter time series provided by ten International GNSS Service (IGS) Analysis Centers, spanning the period from 1994 to 2025. These series are part of the most recent third IGS reprocessing campaign (IGS Repro3), along with subsequent updates. We analyze geocenter motion across four temporal bands: high-frequency artifacts (2–40 d), draconitic signals (40–150 d), seasonal variations (150–400 d), and interannual changes ( > 400 d). The seasonal annual and semi-annual component, which is central to geocenter studies, accounts for only about 20–35 % of the total variance of GNSS series depending on the component, underscoring the significant influence of GNSS-related artifacts. The combined solutions of individual GNSS series exhibit a substantial reduction in scatter compared to the individual AC outputs: a 40–50 % improvement in the high-frequency band (2–40 d) and a 35–40 % reduction in the draconitic band (40–150 d). Cross-validation with independent datasets, including satellite laser ranging (SLR) to LAGEOS satellites, precise orbit determination of low Earth orbiters, and geophysical loading models, shows strong agreement at seasonal timescales (correlations > 0.8), confirming the geophysical origin of annual geocenter variations in GNSS solutions. However, while the magnitude of amplitude variations of the annual signal is internally consistent across most GNSS contributions, it remains misaligned with external series. Moreover, a sliding window spectral analysis reveals a marked temporal evolution of GLONASS orbital artifacts near ∼ 8-day periods intensified during the constellation’s expansion phase (2005–2012). In contrast, theoretically predicted Galileo signals remain below detectability thresholds. To combine the individual solutions, we apply Förstner’s variance component estimation methodology. Our results demonstrate that an effective combination of geocenter estimates can be achieved at the parameter level, while maintaining a quality level comparable to the full SINEX combination methods used by the IGS.