GOCE – Last days’ orbits H. Bock, A. Jäggi, U. Meyer Astronomical Institute, University of Bern, Switzerland

PSD.1

40th COSPAR Scientific Assembly 2014 2 -10 August 2014 Moscow, Russia

Astronomical Institute University of Bern

Background and Motivation • The

first ESA Earth Explorer core mission GOCE ended officially on 21 October 2013, because the satellite ran out of fuel.

• Three

weeks later, on 11 November 2013, the satellite re-entered the Earth’s atmosphere near the Falkland Islands in the South Atlantic.

• GPS-based

orbit determination was possible until few hours before re-entry.

• Data

from both GPS receivers are available during the last days.

Copyright: Bill Chater

Slide 2

Astronomical Institute University of Bern

Background and Motivation GOCE orbit height derived from GPS

21 October 2013

10 November 2013



Last available GPS measurements: 10 November, 17:15:20 UTC

Slide 3

Astronomical Institute University of Bern

Background and Motivation 



 

In the frame of the European GOCE Gravity Consortium (EGG-C) AIUB was responsible for the generation of the GOCE Precise Science Orbit (PSO) product => reduced-dynamic and kinematic orbit. Internal validation: Orbit overlap analysis and differences between reduced-dynamic and kinematic orbits for consistency checks. External validation: Satellite Laser Ranging (SLR) measurements. Reduced-dynamic orbits were generated with the same orbit parameterization for the entire mission.

Two main questions for this study: 



How can the orbits be validated, because SLR measurements are no longer available (only three passes)? Is the orbit parameterization of the reduced-dynamic orbit still reasonable for the last three weeks of GOCE?

Slide 4

Astronomical Institute University of Bern

What have we done?

We look at the following possibilities for validation: 

Orbit differences between reduced-dynamic and kinematic orbit.



Comparison of orbit solutions from the two GPS receivers.

Parametrization of the reduced-dynamic orbit is adapted by  

changing the constraints of the empirical parameters replacing the background models (e.g., gravity field model) by more recent models

Slide 5

Astronomical Institute University of Bern

GOCE internal orbit validation

RMS (cm)

RMS of differences between red.-dyn. and kinematic orbits during official mission time

18 15 12 9 6 3 0

radial +9cm

along−track +6cm

out−of−plane +3cm

3−D

Bock et al. (2014)

Jul

Jan

29 December 2009

Jul

Jan Jul Jan Date in 2009−2013

Jul

Jan

Jul

28 December 2012

Differences between reduced-dynamic and kinematic orbits

Slide 6



show consistency between the two orbit types and



reveal data problems and gaps in the kinematic orbit

Astronomical Institute University of Bern

Differences red.-dyn.  kinematic orbits radial

m

0.1 0 −0.1

At the beginning of the mission the differences between reduced-dynamic along-track and kinematic orbits

m

0.1 0 −0.1



show only few outliers and



only small systematics are present out-of-plane

m

0.1 0 −0.1 0

29 December 2009 Slide 7

6

12 Hours

18

24

Astronomical Institute University of Bern

Differences red.-dyn.  kinematic orbits m

0.5

radial

0 −0.5 

m

0.5



0



−0.5

m

0.5

12 Hours 28 December 2012 and 29 December 2009 Slide 8

Kinematic orbit shows more “outliers” and systematic effects But: Kinematic orbit is independent from physical models and therefore

itout-of-plane should be possible to validate the reduced-dynamic orbit modeling using the differences between the reduced-dynamic and the kinematic orbits 18 24



0 −0.5

End of 2012 the data quality is worse than end of 2009 along-track

0

6

Astronomical Institute University of Bern

GOCE internal orbit validation RMS of differences between red.-dyn. and kinematic orbits for official mission

RMS (cm)



radial +9cm along−track +6cm out−of−plane +3cm 3−D 18 15 12 9 Larger 6 RMS values for the last three weeks reveal that the 3 Bock et al. (2014) parameterization of the 0 reduced-dynamic orbit isJul not Jan Jul Jan Jul Jan Jul Jan Jul ideal at all Date in 2009−2013

200

Last three weeks

RMS (cm)

radial

SLR validation (3 passes) 2.64 ± 5.52 cm Slide 9

along−track

out−of−plane

3−D

160 120

31 October 2013 3D-RMS: 21.7 cm

80 40 0

295

300

305 Day in 2013

310

Astronomical Institute University of Bern

315

Differences red.-dyn.  kinematic orbits m

0.5

radial

0 −0.5

m

0.5

along-track

0 

−0.5

m

0.5



0 −0.5



12 Hours Original solution; 31 October 2013 Slide 10

0

6

Large once-per-revolution signal in radial and along-track component Empirical orbit parameters are not out-of-plane able to catch the full signal Constraints are obviously too tight 18

24

Astronomical Institute University of Bern

Reduced-dynamic orbit determination

Orbit models and parameterization:

    

2.5 x 2*10-8 m/s2 5x 10 x 25 x 50 x

2

Test solutions with weaker constraints:

0 −2

μm/s2



EIGEN5S 120x120, FES2004 50x50 (fixed by GOCE Standards) Six initial orbital elements Three constant accelerations in radial, along-track, out-of-plane 6-min piece-wise constant accelerations in radial, along-track, out-of-plane (2*10-8 m/s2) 2 μm/s

   

0 −5 −10 −15 2

2



30 h processing batches (not for the last 10 days), 10 s sampling, undifferenced processing, ionosphere-free linear combination, CODE Final GNSS orbits and clocks (5 s) and Earth Rotation Parameters

μm/s



0 −2

Slide 11

0

6

12 18 Hours of day 294/2013

Astronomical Institute University of Bern

24

Solutions with weaker constraints 3D RMS of differences between red.-dyn. and kinematic orbits

3D RMS (cm)

120 100







2.5x

5x

10x

25x

50x

80 60 40 20 0



orig

295

300

305 Day in 2013

Test solutions with weaker constraints show better consistency with kinematic orbits.

310

315

SLR validation RD orbits 2.64 ± 5.52 cm

Differences between 5x and 50x weaker constraints are marginal.

7.25 ± 7.55 cm

Except the very last days, these solutions are acceptable.

3.78 ± 4.07 cm

SLR validation is not very meaningful because of the very small number of passes

Slide 12

4.76 ± 5.03 cm 3.43 ± 3.73 cm 3.40 ± 3.73 cm

Astronomical Institute University of Bern

Differences red.-dyn.  kinematic orbits m

0.5

radial

0 −0.5

m

0.5

along-track

0 −0.5 

m

0.5

Large once-per-revolution signal is very much reduced out-of-plane

0 −0.5

12 18 24 Hours Original solution and 10x weaker constraints; 31 October 2013 Slide 13

0

6

Astronomical Institute University of Bern

m

4 2 0 −2 −4

m

4 2 0 −2 −4

m

Differences red.-dyn.  kinematic orbits

4 2 0 −2 −4

radial

along-track



Orbit differences are significantly larger for the very last hours out-of-plane (different scale!!)

The GPS data quality at this stage of the mission (150 – 130 km altitude) is still surprisingly 18 good!!! 24



12 Hours Original solutions and 10x weaker constraints; 10 November 2013 Slide 14

0

6

Astronomical Institute University of Bern

Comparison with second GPS receiver RMS of differences between red.-dyn. and kinematic orbits: 1 Aug – 20 Oct 2013 SSTI-A Mean 3D-RMS: 5.86 cm

RMS (cm)

15 radial

along−track

cross−track

3−D

radial

along−track

cross−track

3−D

10 5 0

SSTI-B Mean 3D-RMS: 4.43 cm

RMS (cm)

15 10 5 0

 

Slide 15

215

225

235

245 255 265 Day of Year in 2013

275

285

Since 1 August 2013 both GPS receivers were running SSTI-B was operated with an updated firmware version, which reduced the number of data losses on L2 but led to a slight increase of the carrier phase noise. Astronomical Institute University of Bern

Solutions with weaker constraints – second GPS 3D RMS of differences between red.-dyn. and kinematic orbits

80

3D RMS (cm)

AA orig



Slide 16

BB orig

BB 10x

300

305 Day in 2013

60 40 20 0



AA 10x

295

Orbit differences from SSTI-B show in average slightly better performance SLR validation is only a snap-shot from the three passes

310

315

SLR validation RD orbits (3 passes) SSTI-A

SSTI-B

2.64 ± 5.52 cm

10.54 ± 11.87 cm

3.78 ± 4.07 cm

2.94 ± 4.28 cm

Astronomical Institute University of Bern

Solutions with weaker constraints – second GPS 3D RMS of differences between red.-dyn. and kinematic orbits

80

3D RMS (cm)

AA orig





BA BB orig orig

BA BB 10x

60 40 20 0



AA AA 10x 10x

295

300

305 310 315 Day in 2013 If we look at the differences between the reduced-dynamic orbits from SSTI-B and the kinematic orbits from SSTI-A, the differences are very similar Reason for this is the quality of the kinematic orbit, which is slightly better for SSTI-B because of less data gaps The differences in the quality of the kinematic orbit are not critical for the validation of the reduced-dynamic orbit

Slide 17

Astronomical Institute University of Bern

Improved background modeling 



In order to improve the background models the gravity field model EIGEN5S 120x120 is replaced by GOCO03S 200x200 for the first 11 days of the decay phase. Test solutions with original and weaker constraints are repeated. Old solutions

30

3D RMS (cm)

orig



5x

10x

25x

50x

20 10 0



2.5x

294

296

298 300 Day in 2013

302

304

No improvements with respect to the old solutions can be noticed with the better gravity field model. Other perturbations, mainly the atmospheric drag, are dominating.

Slide 18

Astronomical Institute University of Bern

Summary 







How can the orbits of the last days of GOCE be validated ?=> The differences between kinematic and reduced-dynamic orbits may be used for validation, because the quality of the kinematic orbit is still very good. Is the orbit parameterization of the reduced-dynamic orbit still reasonable for the last three weeks of GOCE? => No, the constraints are too tight; 10x weaker constraints are reasonable. Orbits from both GPS receivers are as expected very similar and comparison confirms the results from the main GPS receiver. Updates in the background modeling of the reduced-dynamic orbit determination did not improve the results of the reduced-dynamic orbits, because other perturbations, in particular atmospheric drag, are dominating.

Slide 19

Astronomical Institute University of Bern