Consistency of Crustal Loading Signals Derived from Models and GPS: Inferences for GPS Positioning Errors

After applying corrections for surface load displacements to a set of station position time series determined using the Global Positioning System (GPS), we are able to infer precise error floors for the determinations of weekly dN, dE, and dU components. The load corrections are a combination of NCEP atmosphere, ECCO non-tidal ocean, and LDAS surface water models, after detrending and averaging to the middle of each GPS week. These load corrections have been applied to the most current station time series from the International GNSS Service (IGS) for a global set of 706 stations, each having more than 100 weekly observations. The stacking of the weekly IGS frame solutions has taken utmost care to minimize aliasing of local load signals into the frame parameters to ensure the most reliable time series of individual station motions. For the first time, dN and dE horizontal components have been considered together with the height (dU) variations. By examining the distributions of annual amplitudes versus WRMS scatters for all 706 stations and all three local components, we find an empirical error floor of about 0.65, 0.7, and 2.2 mm for weekly dN, dE, and dU. Only the very best performing GPS stations approach these floors. Most stations have larger scatters due to other non-load errors. These global error floors have been verified by studying differences for a subset of 119 station pairs located within 25 km of each other. Of these, 19 pairs share a common antenna, which permits an estimate of the fundamental electronic noise in the GPS estimates: 0.4, 0.4, and 1.3 mm for dN, dE, and dU. The remaining 100 close pairs that do not share an antenna include this noise component as well as errors due to multipath, equipment differences, data modeling, etc, but not due to loading or direct orbit effects since those are removed by the differencing. The WRMS dN, dE, and dU differences for these close pairs imply station error floors of 0.8, 0.9, and 2.1 mm, respectively, almost the same as the error floors inferred from the global results where orbit errors should be fully expressed. This match implies that GPS orbit errors are only a minor part of the IGS weekly position uncertainties. A similar comparison for periodic signals implies that about a third of the GPS draconitic harmonics probably arise from sources local to the stations whereas the remaining two-thirds comes from orbital effects.