Precise point positioning (PPP) requires careful modeling of several error sources affecting GNSS observations. Additionally, consistency between the network and user software is essential to obtain the upmost accuracy. Unfortunately, an inconsistency affecting NRCan's CSRS-PPP software has been causing, for a few years now, a height bias of several millimeters.
With about 1,000 submissions daily, we can say that continuous testing of the CSRS-PPP software is taking place. Optimizations are often implemented in the software to accommodate the impressive variety of data sets collected by users. However, it looks like we omitted to continuously monitor the consistency of the 24-hour static solutions with respect to the IGS weekly solutions. Had we done so, we would have noticed that an additional correction needed to be implemented following the adoption of the IERS 2010 conventions by the IGS.
This correction is the Shapiro delay, a secondary relativistic effect (IERS conventions 2010, equation 11.17). This signal propagation effect causes a range error typically between 1-2 cm and affects the height estimate of the station. When testing NRCan’s new SPARK engine against the actual engine running online (PACE), we noticed a height bias of about 4-5 mm for stations in Canada. The following figure shows the convergence of both software for the latitude, longitude and vertical components, based on 1000 RINEX files (20 stations, 1 day/week for 50 weeks).
Fig 1 Convergence evaluation of NRCan’s PPP engines: SPARK (new) vs PACE (current)
Since a height bias of a few millimeters is typically within the position uncertainty, no clients ever complained about this bias. However, a recent publication comparing online PPP software stated that (Jamieson and Gillins, 2018):
“For the GPS+GLONASS solutions, the VRMS values from TrimbleRTX are 10–20% smaller than those from CSRS-PPP. Based on this finding, it seems that TrimbleRTX combines GLONASS with GPS more effectively than CSRS-PPP does.”
From the explications above, it becomes clear that it is not the integration of GPS and GLONASS that is problematic, but rather the incompatibility between the IGS products and the NRCan PPP software regarding the Shapiro correction.
In summary, if you are a CSRS-PPP user, the upcoming change from PACE to SPARK will create a height discontinuity of a few millimeters in your time series analysis. Don’t worry about it and note that it is an improvement in your results!
p.s. Thanks To Elyes Hassen from NRCan for the above figure!
Jamieson M, Gillins DT (2018) Comparative Analysis of Online Static GNSS Postprocessing Services. J Surv Eng, 144(4) doi: 10.1061/(ASCE)SU.1943-5428.0000256