Ten years (1997–2006) of weekly GNSS solutions of 205 globally distributed stations have been used to investigate the impact of the reference frame definition on the estimated station velocities. For that purpose, weekly regional solutions (covering the European region) and global solutions have been, respectively, stacked to obtain regional and global velocity fields. In both cases, the estimated long-term solutions (station positions and velocities) were tied to the ITRF2005 under minimal constraints using a selected set of reference stations. Several sets of global and regional reference stations were tested to evaluate first the impact of the reference frame definition on the global and regional velocity fields and later the impact on the derived geodynamic interpretations. Results confirm that the regional velocity fields show systematic effects with respect to the global velocity field with differences reaching up to 1.3 mm/year in the horizontal and 2.9 mm/year in the vertical depending on the geographical extent of the network and the chosen set of regional reference stations. In addition, the estimations of the Euler pole for Western Europe differ significantly when considering a global or a regional strategy. After removing the rigid block rotation, the residual velocity fields show differences which can reach up to 0.8 mm/year in horizontal component. In Northern Europe, the vertical ground motion is dominated by the Glacial Isostatic Adjustment (GIA). A proper modeling of this effect requires sub-mm/year precision for the vertical velocities for latitudes below 56°. We demonstrate that a profile of vertical velocities shows significant discrepancies according to the reference frame definition strategy. In the case of regional solutions, the vertical modeling does not predict any subsidence around 52° as predicted by the global solution and previous studies. In summary, we evidence the limitation of regional networks to reconstruct absolute velocity fields and conclude that when geodynamics require the highest precisions for the GNSS-based velocities, a global reference frame definition is more reliable.