Jeff Freymueller Julie Elliott Geophysical Institute, University of Alaska Fairbanks

1. How well do we know the North American frame today?:Differences between recent realizations and implications Jeff Freymueller Julie Elliott Geophysical Institute, University of Alaska Fairbanks

2. Topics • You can fit velocities to a fraction of a mm/yr, but what is the real uncertainty in the NOAM frame? • Uncertainty in underlying frame (ITRF) • How precise is ITRF, really? • Differences between NOAM angular velocities • Sella et al. (2007) • Initial SNARF pole

3. Uncertainty in ITRF • Uncertainty in ITRF commonly ignored. • The TZ rate difference (1.8 mm/yr) between ITRF2005 and ITRF2000 has gotten a lot of attention. • There may actually be a similar (or larger) difference between ITRF2000 and ITRF97 • If so, uncertainty in frame (geocenter origin) may be much larger than precision of GPS baseline rates.

4. ITRF2000 Velocities – Sella pole

5. ITRF2000 Velocities – other poles Black – Sella 2007 White – REVEL Yellow – SNARF Note systematic residual in REVEL, 2-3 mm/yr

6. NOAM Poles • With past studies, it is common that NOAM poles do not lie within 95% confidence ellipses of other studies • Systematic errors or missing uncertainty • Difference between SNARF and Sella is a rotation about a pole in the SE United States.

7. This is Expected • The least certain component of the plate’s angular velocity vector is a rotation about an axis through the centroid of the network. • Consider the angular velocity vector of the plate expressed in the local east-north-up coordinates at a particular site:

8. 0 The site’s velocity is • Two components of the plate angular velocity are directly determined by the site’s velocity, while the third (local vertical component) is completely undetermined. • When sites span a small area, their local vertical directions will be similar, and this component of the angular velocity will be the least well determined.

9. More About Angular Velocity • We could resolve the undetermined component by taking a minimum norm solution: • In this case the pole is located 90° away from the site. • The pole could also be located anywhere on the great circle that lies between this minimum-norm solution and the site itself. • The component of the angular velocity in the average radial direction will naturally be the least constrained.

10. Strategy for Augmented Covariance • Sella and SNARF differ by almost 1 mm/yr in Alaska, significant relative to CGPS site velocities, and we really can’t tell which is “right” • We thus augment the covariance in two ways: • Add an uncertainty corresponding to the difference in angular velocity between Sella and SNARF • Add an uncertainty in Zdot of 1.8 mm/yr as a conservative uncertainty in the ITRF.

11. Additional Uncertainty Rotation Only

12. Additional Uncertainty Rotation + Zdot

13. Augmented Covariance

14. Conclusions • May be ~2 mm/yr frame differences between all past ITRFs, not just ITRF2000 to ITRF2005. • 1.8 mm/yr in Tzdot between ITRF2000, ITRF2005 • Probably 2-3 mm/yr between ITRF97, ITRF2000 • Latest NOAM poles in ITRF2000 agree to with 0.3 mm/yr over the actual stable part of NOAM • Difference is a rotation about a pole located on Gulf Coast • Difference is ~ 0.5 mm/yr on west coast, almost 1 mm/yr in Alaska • We propose an augmented covariance matrix for the NOAM angular velocity that we think is more realistic than the published covariance.