On the systematic differences between various GNSS solutions

Space geodesy is the branch of science that deals with determining the shape and dimension of the Earth using quasars and artificial satellites. However, the use of space techniques is not limited to geodesy, many recent geodynamical interpretations are based on the velocities determined using GNSS (Global Navigations Satellite System). In this regard, reference frame differences as well as processing strategies - Precise Point Positioning (PPP) or Network Solution (NS) - may induce systematic differences of significant values. A first comparison between NGL (Nevada Geodetic Laboratory) PPP-produced and IGS (International GNSS Service) NS-produced (repro3) products showed systematic differences between both sets of displacement time series. While the discrepancies in the noise parameters are quite understandable and interpretable, the systematic disagreements in the velocities or amplitudes of annual signals are concerning. The repro3 daily combined solutions are aligned in origin and orientation to the IGSR3 reference frame, which inherits its origin and orientation from ITRF2014. So the daily repro3 station positions are expressed with respect to the ITRF2014 origin, which follows CM on the long-term, but reflects CF at sub-secular time scales. NGL time series, just like the IGS ones, are aligned to a linear reference frame, IGS14 in their case. Since both the IGS repro3 and NGL time series are in fact with respect to the ITRF2014 origin, this cannot explain the annual amplitude differences we observe. A difference in scale may contribute though, because the NGL time series are aligned in scale to IGS14, but the repro3 time series are not aligned in scale to any reference frame. They inherit their scale from the radial satellite phase center offsets (z-PCOs) in igsR3.atx, so in general we may expect: (1) an average ~8 mm difference between the vertical positions in both sets of time series at epoch 2010.0, (2) an average ~0.2 mm/yr rate between the vertical positions in both sets of time series, (3) small differences in the vertical seasonal signals due to the network effect occurring when the NGL solutions are aligned in scale to IGS14. But the fact that the NGL solutions are aligned in scale to the reference frame, while the IGS solutions are not, might explain some other systematic differences. To verify that, we compared the aforementioned series to a set of special IGS repro3 time series produced at IGN and aligned not only in origin and orientation, but also in scale to the IGSR3 reference frame. The comparison between the series was made at 607 stations distributed around the world. The Up component was investigated, and the data have been cut to the same ranges in all three solutions (namely, IGS, NGL and IGN) to avoid misinterpretations and have a minimum length of 5 years. Velocities as well as amplitudes of annual and draconitic oscillations have been analyzed revealing very interesting spatial patterns in the differences.