1. Geocenter motion estimates from the IGS Analysis Center solutions P. Rebischung, X. Collilieux, Z. Altamimi IGN/LAREG & GRGS 1 EGU General Assembly, Vienna, 26 April 2012

2. Background Global GNSS solutions are sensitive to geocenter motion in two different ways: 2 Through orbit dynamicsThrough loading deformations

3. Background Main limitation of « orbit dynamics »: The non-gravitational forces acting on GNSS satellites are not modeled accurately enough. →ACs have to estimate empirical accelerations which correlate with the CM location (origin). 3 Example of accelerations that would be felt by a satellite if CM was shifted by 1 mm in the Z direction. Accelerations are shown in the « DYB » frame: ― D: Satellite-Sun axis ― Y: Rotation axis of solar panels ― B: Third axis Correlations with some parameters of the CODE model are obvious (constant along D; once-per-rev along B).

4. Methodology Data: Weekly solutions from 7 ACs (COD, EMR, ESA, GFZ, JPL, MIT, NGS) 1998.0 – 2008.0 : reprocessed solutions 2008.0 – 2011.3 : operational solutions Stacking: For each AC, stack weekly solutions into a long-term piecewise linear frame. Geocenter motion estimation: Pseudo-Observations = weekly minus regularized position differences Three possible models 4

5. Methodology 5 Network shift approach CF approach aka degree-1 deformation approach (Blewitt et al., 2001) CM approach (Lavallée et al., 2006) with degree-1 Love numbers in CF framewith degree-1 Love numbers in CM frame Observation equations Estimates of r CM-CF from orbit dynamics from loading deformations In the CM approach, both information contribute to the same estimate (because degree-1 deformations have a translational part in the CM frame). In the following, use: well-distributed sub-network identity weight matrix In the following, use n max =5

6. Sub-annual frequencies corrupted by (odd) draconitic harmonics Z: network shift approach 6 All ACs affected

7. Z: network shift approach Low frequencies well explained by annual + 1 st draconitic: →Progressive phase shift wrt SLR Similar patterns for ACs using the CODE model Different pattern for JPL (and EMR?) Underlying annual signal unreliable: 7

8. Z: CF approach 8 Draconitics smaller than in the network shift approach:

9. Z: CF approach 9 Annual signal in phase with SLR for all ACs, over the whole time period But amplitude over-estimated: Also found with simulations (see Collilieux et al., JoG 2012) Aliasing of >5-degree deformations?

10. Note on the CM approach CM approach ≈ weighted mean of orbit dynamics and loading deformations 10 ≈ 0.65 for X ≈ 0.60 for Y ≈ 0.45 for Z ≈ 0.35 for X ≈ 0.40 for Y ≈ 0.55 for Z with n max =5

11. Z: CM approach 11 Some draconitics averaged; other cancelled (depending on their relative phases in Z shift and Z CF )

12. Z CM alternately: In good agreement with SLR; ≈ 0; Out-of-phase (recently, except JPL). →Is it really reasonable to make this weighted mean? Z: CM approach 12

13. X Network shift approach: Draconitics up to 2 mm Annual signal partly detected CF approach: Draconitics as large as in net. shift Annual signal in phase with SLR Amplitude over-estimated CM approach: Sometimes in good agreement with SLR But not always 13

14. Y Network shift approach: Draconitics up to 2 mm Rather good annual signal CF approach: Draconitics as large as in net. shift Rather good annual signal But slight phase shift for some ACs CM approach: Strikingly good agreement with SLR Net. shift & CF errors cancel out. 14

15. Conclusions (1/2) Network shift (orbit dynamics): All ACs affected by draconitics as large as « true » annual signal Effect of draconitics different for JPL (and EMR?) than for other ACs in Z (JPL’s first draconitic not in phase with other ACs) Underlying annual signals: Unreliable in Z Partly detected in X Agrees well with SLR in Y CF approach (loading deformations): Also corrupted by draconitics As large as in network shift in X & Y But ~twice smaller in Z Annual signals in phase with SLR But amplitudes over-estimated in X & Z 15

16. Conclusions (2/2) CM approach (≈ weighted average): In X & Z, network shift and CF errors cancel sometimes out, but not always. →Isn’t the unification of orbit dynamics and loading deformations questionable? Strikingly good results in Y: network shift and CF errors cancel out. →Why? 16

17. Additional slides 17

18. Note on the network shift approach Using raw cov. matrices gives unrealistic results: Shift estimates are perturbed by correlations with degree-1 deformations. 18 ≈ 0.5 for X & Y ≈ 0.8 for Z

19. X 19

20. Y 20

21. Z 21

22. X: draconitic harmonics Radius = 2 mm 22 Network shift CF CM 1st 2nd 3rd 4th 5th 6th 7th

23. 23 Network shift CF CM Y: draconitic harmonics Radius = 2 mm 1st 2nd 3rd 4th 5th 6th 7th

24. 24 Network shift CF CM Z: draconitic harmonics Radius = 5 mm Radius = 10 mm 1st 2nd 3rd 4th 5th 6th 7th