ESOC Navigation Support OfficeIGS Workshop 2008

MiamiESOC IGS Reprocessing

T.A. Springer, F. Dilssner, E. Schoenemann, I. Romero, J. Tegedor, F. Pereira, J. Dow

AbstractEarly in 2008 ESOC has replaced its old GNSS analysis software with its new software, called Napeos. One of the key design criteria of this new ESOC GNSS analysis software has been short processing times! The now fully operation software allows to generate an IGS final solution using 100 stations within 30 minutes on a Linux PC with the Intel Fortran compiler. A network of 150 stations takes around 60 minutes to process. On a “quad-core” CPU four jobs can be run simultaneously without any performance loss. Thus on a single quad-core PC we are able to process 96 days (24 hours x 4 Cores) of IGS data per day! With such a performance rapid reprocessing decades of IGS data becomes feasible. On a single PC reprocessing a full year of IGS data takes less than 4 days. So reprocessing 10 years of data will take about 40 days on a single PC and even less when using multiple PC nodes. This opens up a great potential for scientific research. New models can be incorporated into the software, tested, and validated. After that an N-year reprocessing can be started to study the effect of the new model and compare to previous reprocessed results. The scientific value of such an efficient tool is tremendous!

This poster shows some selected results from our first reprocessing efforts using the year 2007. We have reprocessed this year twice to study the effect of the simple Earth Albedo and infra-red radiation model in our software. The reprocessed results are also compared to the original IGS products for 2007. Using the obtained orbit, clock, and ERP products we will look at internal and external consistencies of the results. We also look at a “fast Fourier transformation” (FFT) of the obtained time series and will show some interesting results. Last but not least we will use the SLR observations of the two GPS satellites carrying SLR reflectors to study the quality of the different IGS orbits.

Conclusions• The IGS reprocessing effort is a very important and valuable undertaking because:

• It will provide a homogeneous set of products from the contributing ACs.

• Even in 2007 and 2008 significant processing changes were made at different ACs causing jumps in different parameters, orbit scale, geocenter, etc. etc.

• All changes make analysing long time series extremely difficult if not impossible.

• Reprocessing will allow us to identify AC dependent modelling differences (and errors) because signals are no longer “obscured” by modelling changes

• Reprocessing results should have a much better accuracy, even already for 2007!

• The ESA independent orbit overlap study has revealed:

• Cross-track differences in orbit overlaps are larger than along-track differences!?

• Significant 14-day period in difference in the right ascension of the ascending node!

• Similar signals were observed in the Z-rotation of the IGS orbit comparisons for most ACs!

• Possible causes may be imperfections in the sub-daily UT1 and high rate Nutation models.

• Of course there are several other (model) error sources that may lead to 14-day periods

• Earth Albedo and Infra-Red Radiation pressure (ON and OFF) results from ESOC have revealed:

• Earth Albedo and Infra-Red Radiation may explain the observed bias in the SLR O-C residuals.

• The simple ESOC model does, however, not remove the correlation of the SLR residuals with the Sun-Satellite geometry (U0 and B0 angles).

• A more elaborate model is needed (e.g. the model(s) developed at UCL).

• The SLR validation of IGS Orbits has demonstrated:

• Independent measurements are extremely useful.

• Good agreement between SLR residual mean and orbit scale

• Good agreement between SLR residual sigma and orbit combination statistics

• ESA reprocessed solution (despite CMC error) very good!

Quality of New ESOC SoftwareThe quality of our new software is demonstrated in the figure below where we show the RMS of different orbits compared to the IGS final orbits. The comparisons includes in particular:

• ESA: the original ESA IGS orbit submissions using the old ESA software.

• ES1: the new reprocessed ESA IGS orbits using the new ESA software, Napeos.

The figure clearly shows that the ES1 orbits are superior to the old ESA orbits, 16mm vs 22mm.

Earth Albedo and Infra-Red Radiation PressureESOC has reprocessed the year 2007 twice. Once ignoring the Earth Albedo and Infra-Red radiation pressure (solution label “OFF”). The second time taking these effects into account using a simple model as available in the Napeos software (solution label “ON”). This full year reprocessing with and without taking the effect into account allows us to study the differences in the obtained results. We have done this in two ways, firstly comparing the orbits directly. Secondly by looking at the SLR O-C residuals.

Orbit Comparison

For the orbit comparison we have compared the orbits to each other including a 7 parameter Helmert transformation. The only significant difference was observed for the orbital scale. The figure below shows the difference in orbital scale for each of the 6 orbital planes, A – F, from top to bottom. For clarity each plane was offset by 0.1 ppb. The mean scale difference between the two orbit solution series over the full year is 0.455 ppb. The average RMS difference between the two orbit solutions over the full year is at the few millimeter level.

We may conclude that our simple Earth Albedo and Infra-Red radiation pressure model primarily affects the scale of the orbits with a small variation depending on the orientation of the orbital plane.

SLR Validation of IGS OrbitsThe figure below shows the statistics of the two-way SLR O-C residuals of the IGS orbits. For this analysis the final orbits, not the reprocessed orbits, of all the ACs contributing to the IGS final orbits were used. In addition the two series of reprocessed ESA orbits (labeled ON and OFF) are included as well.

A couple of interesting observations may be made.

• The sigma of the IGS Final orbit is a bit better than IGS Rapid (IGR) orbits, the mean is identical..

•Good correlation between SLR bias and the scale differences observed in the IGS orbit combination

• Only JPL mean bias does not correspond to the scale observed for the JPL orbits in the IGS combination. However, after the JPL analysis change in 2007 this seems to have improved

• Good correlation between SLR sigma and the RMS of the IGS orbit combination

• The ESA Reprocessed (ON and OFF) orbits are significantly better than the original ESA orbits.

• The Albedo and Infra-Red modeling (ON) changes the scale of the orbits significantly.

• The ESA Reprocessed solution performs slightly better than all other solutions.

The results underline the importance of having a completely independent technique to monitor the GNSS orbits. At ESOC we will routinely perform this analysis for our reprocessed solutions and are planning to include the other reprocessing AC solutions as well.

Independent Orbit OverlapThe ESOC IGS analysis strategy generates completely independent 24 hour solutions. Therefore, the orbit differences at the arc boundaries, i.e., the day boundaries, give a completely independent measure of the orbit quality. The figure below gives the RMS of the orbit overlaps over the full year 2007 of our reprocessed orbits. The RMS is given in both the position and velocity space using the Radial, Along-, and Cross-track directions. It is surprising to see that the cross-track errors are dominating, because typically one would expect the along-track errors to be dominant!

Two-way SLR Residuals

Over recent years it has been shown that the SLR residuals of the GPS orbits have a dependency on the location of the Sun with respect to the orbital plane. The Earth Albedo effect is of course also depending on the location of the Sun. The observed scale differences, shown above, have the correct sign to explain the bias observed between the SLR observations and the GPS orbits. An analysis of the SLR observations of the two GPS satellites should show if the bias indeed changes and if the sigma of the residuals is reduced. I.e. does the scale variation, as plotted above, reduce the observed dependency of the SLR observations on the Sun – Orbital Plane geometry. A 0.1 ppb scale change corresponds to 2.65 mm radial change (5.3 mm in two-way SLR residuals).

The bar plot (shown on the right) includes the statistics of the “ON” and “OFF” solutions. It shows that the mean of the two-way SLR residuals changes from 46 to 25 mm, i.e., a change of 21 mm. The mean orbit scale change of 0.455 ppb corresponds to a radial orbit change of 12 mm and thus should affect the two-way SLR residuals by 24 mm. Thus the orbit scale change agrees very well with the change in the mean of the SLR residuals. The SLR sigma of the two orbit series does not differ. This means that the 0.1ppb variation observed in the orbital scale does not improve the SLR O-C statistics. Note that GPS-35 is in plane B and GPS-36 in plane C which both show a significant scale variation in the plot shown above.

The figure below shows the SLR O-C residuals of the “ON” solution. The dependency of the SLR residuals on the argument of latitude of the Sun is still very significant. We assume that our Earth Albedo and Infra-red model is to simple to fully account for these effects. So more work will be needed in this area.

Value of ReprocessingThe figure below shows the time series of X- and Y-orbit translations (Y-translations artificially offset by 10 mm below the X-translations) when comparing our reprocessed orbits to the IGS final orbits. A clear periodic signal may be observed which is also present in our Z-translations and the orbit scale. After some investigations a bug in our CMC correction implementation was detected. This bug affected only our SP3 orbits and has been corrected on May 19, 2008. The detection and correction of this bug was only possible thanks to our involvement in the IGS reprocessing efforts. This demonstrates how valuable participation in the reprocessing is!

0

5

10

15

20

25

30

35

Position VelocityRM

S o

f Orb

it O

verla

p in

pos

tion

(mm

) and

Vel

ocity

(mm

/s)

Radial Along Cross

A Closer Look at Orbit Overlaps

Due to the surprising result of cross-track errors dominating in the orbit overlap statistics we took a closer look at the overlaps. Besides looking at the results in radial, along-, and cross-track we also studied the overlaps in Earth Fixed Cartesian (X, Y, Z) and in the Kepler elements. Besides looking at the times series we also made a Fourier analysis of all the 18 time series. We found a surprising result that in the Fourier analysis of the time series of the overlaps of the Right Ascension of the Ascending Node (RAAN) two very significant peaks around the 14 day interval showed up. The two figures below demonstrate this; the top one showing the full Fourier series, the bottom one a “zoom in” around the 14 day period. Based on these results we looked for 14 day periods in results of the orbit comparisons. Here we found a striking agreement for the Z-orbit rotations! The same two peaks around the 14 day period were found for practically all ACs. Only for EMR and JPL the peaks were not very clear. The reason for these peaks in RAAN and Z-orbit ration is unclear at present. We believe that a possible explanation may be found in imperfections in the sub-daily UT1 model and the high frequency Nutation model. Important to notice is that the signal has a significant strength! The two peaks in the RAAN together amount up to ~20 mm and thus contribute significantly to the RMS of the orbit overlaps! It also explains why cross-track the differences are larger than along-track.

0

10

20

30

40

50

60

Mean Sigma

Mea

n an

d Si

gma

of S

LR O

-C R

esid

uals

(mm

)

IGS IGR COD EMR ESA GFZ JPL MIT NGS SIO ON OFF

ESOC will reprocess all the IGS data since 1994

before the end of 2008!