The angular velocities of the plates and the velocity of Earth's centre from space geodesy

Using space geodetic observations from four techniques (GPS, VLBI, SLR and DORIS), we simultaneously estimate the angular velocities of 11 major plates and the velocity of Earth's centre. We call this set of relative plate angular velocities GEODVEL (for GEODesy VELocity). Plate angular velocities depend on the estimate of the velocity of Earth's centre and on the assignment of sites to plates. Most geodetic estimates of the angular velocities of the plates are determined assuming that Earth's centre is fixed in an International Terrestrial Reference Frame (ITRF), and are therefore subject to errors in the estimate of the velocity of Earth's centre. In ITRF2005 and ITRF2000, Earth's centre is the centre of mass of Earth, oceans and atmosphere (CM); the velocity of CM is estimated by SLR observation of LAGEOS's orbit. Herein we define Earth's centre to be the centre of mass of solid Earth (CE); we determine the velocity of CE by assuming that the portions of plate interiors not near the late Pleistocene ice sheets move laterally as if they were part of a rigid spherical cap. The GEODVEL estimate of the velocity of CE is likely nearer the true velocity of CM than are the ITRF2005 and ITRF2000 estimates because (1) no phenomena can sustain a significant velocity between CM and CE, (2) the plates are indeed nearly rigid (aside from vertical motion) and (3) the velocity of CM differs between ITRF2005 and ITRF2000 by an unacceptably large speed of 1.8 mm yr(-1). The velocity of Earth's centre in GEODVEL lies between that of ITRF2000 and that of ITRF2005, with the distance from ITRF2005 being about twice that from ITRF2000. Because the GEODVEL estimates of uncertainties in plate angular velocities account for uncertainty in the velocity of Earth's centre, they are more realistic than prior estimates of uncertainties. GEODVEL differs significantly from all prior global sets of relative plate angular velocities determined from space geodesy. For example, the 95 per cent confidence limits for the angular velocities of GEODVEL exclude those of REVEL (Sella et al.) for 34 of the 36 plate pairs that can be formed between any two of the nine plates with the best-constrained motion. The median angular velocity vector difference between GEODVEL and REVEL is 0.028 degrees Myr(-1), which is up to 3.1 mm yr(-1) on Earth's surface. GEODVEL differs the least from the geodetic angular velocities that Altamimi et al. determine from ITRF2005. GEODVEL's 95 per cent confidence limits exclude 11 of 36 angular velocities of Altamimi et al., and the median difference is 0.015 degrees Myr(-1). GEODVEL differs significantly from nearly all relative plate angular velocities averaged over the past few million years, including those of NUVEL-1A. The difference of GEODVEL from updated 3.2 Myr angular velocities is statistically significant for all but two of 36 angular velocities with a median difference of 0.063 degrees Myr(-1). Across spreading centres, eight have slowed down while only two have sped up. We conclude that plate angular velocities over the past few decades differ significantly from the corresponding angular velocity averaged over the past 3.2 Myr.