Fifth ILRS AWG Meeting (Toulouse 2001)
Minutes of ILRS Analysis Working
Group Workshop #5
Toulouse, France
September 17-18,
2001
(written by G. Appleby, R. Noomen)
Agenda: see Appendix 1.
Attendees:
Graham Appleby, Franois Barlier, Richard Biancale, Randall Carman, Roberto Devoti, Elisabeth Fragner, Cinzia Luceri, Maria Mareyen, Horst Mller, Ron Noomen (chairman), Konstantin Nurutdinov, Toshimichi Otsubo, Erricos Pavlis, Toshitaka Sasaki.
For a more detailed list: see Appendix 2.
Monday September 17, 2001
1. Opening
The meeting starts with a few moments of silence to commemorate the victims of the terrorist attacks to the United States, on September 11.
Introductory remarks were given by Noomen (Appendix 3). The chairman explained the background of continuing with this workshop, in spite of the international developments and travelling difficulties, in particular for the US colleagues. The alternatives appeared to have more drawbacks, however.
Membership: Noomen had some errors in his list of AWG members (Mareyen and Nurutdinov, members since March 2001, were missing). In addition, the e-mail address of Jrgen Mller has changed, and Horst Mller replaces Detlef Angermann. An updated membership list is included in Appendix 3. New action item Noomen: check the AWG membership list on the ILRS web pages, in the ILRS e-mail exploder and in his own administration.
In addition, the chairman reminded the participants on the background of the current analysis activities, the role of ILRS within the international community, and provided a list of questions and possible discussion items for this workshop (Appendix 3).
2. Minutes and action items AWG Nice
The minutes of the previous ILRS AWG meeting were briefly discussed. Most important elements were the addition of Etalon observations and the extension of the solve-for parameters with EOPdots, the time-derivatives of the standard EOP parameters xpole, ypole and UT1-UTC (both extensions are optional). The action items coming out of that meeting were revisited one by one. About half of them have been completed and the outcome was reported here, or reference was made to other items on the current agenda. Some of the remaining action items are standing ones, that is they deserve our continuous attention; they will be included in the updated list of action items coming out of this workshop.
The action item on various versions of the SINEX format description (Husson) has been resolved: both the reference to the official SINEX format description and that of the ILRS pilot project interpretation now refer to unique web addresses. The action item on providing ocean loading coefficients (Pavlis) triggered a report on a recent development by Scherneck: analysts can obtain ocean loading coefficients from his website at www.oso.chalmers.se/~loading
3. Pilot project positioning + earth orientation
3.1. Individual contributions series A and B
The chairman started this agenda item with a brief overview of the analysis strategies, measurements and contributions (Appendix 4). Basically, the analysts may contribute with a so-called series A solution, which is based on LAGEOS-1/2 data only and basically covers 1999. As an alternative, a number of "series B solutions are also requested, with different choices as for the data to be used and the solve-for parameters. At this moment, the series B solutions (B1-B4) cover the period April-September 2001.
The intensive tracking campaign of the Etalon spacecraft, requested by the ILRS AWG in March 2001 and running for half a year, has been successful in terms of number of passes: the pass statistics (provided by Husson) shows significant improvements during the past months. Also, the observations of both LAGEOS and Etalon satellites are transferred to the appropriate directories on the CDDIS computers in a very timely way (courtesy Dube).
An overview of the pos+eop solutions currently available shows that 5 institutes have re-analysed the 1999 observations (series A), whereas the series B analysis have been performed by 5 research groups, too (albeit with different levels of completeness) (Appendix 4).
Brief presentations are given by the various contributors, when present:
CRL reported about their experience with the series B solutions. Stations are excluded altogether from a particular 28-day period if the number of combined LAGEOS normal points is less than 30. The IERS 1996 Conventions serve as the basis for the computations. As for data weighting, the stations are distributed in 4 different categories, based on their long-term stability. The relative weighting between LAGEOS and Etalon is 1:4, i.e. a single Etalon normal point contributes 4 times as much to the total solution as does an individual LAGEOS normal point. This proved necessary to obtain any effect on the analysis results by the addition of Etalon data. The parameter solutions are computed in an unconstrained way, as requested by the ILRS AWG. Station coordinates are solved for as common parameters, and range biases are estimated for the LAGEOS and Etalon satellites independently (but pair wise). EOPs (and EOPdots) are computed daily, as requested for this project.
The orbital fit ranges between 10 and 15 mm, and becomes better when the EOPdot parameters are estimated (i.e. an increment of the number of parameters). Otsubo raised questions on the best way to interpolate the EOPs and EOPdots: when EOPs are estimated only, a cubic interpolation using 4 epochs is used; when EOPdots are estimated in combination with EOPs, a cubic interpolation using 2 epochs is used. This issue needs further investigation.
Otsubo was not able to draw a clear conclusion on the improvements (or deterioration) of the quality of the parameters when Etalon data are added and/or when EOPdots are estimated as additional parameters. For more details: see Appendix 5.
ASI has contributed new solutions, both for the series A and for the series B.
In the series A analyses, a single range bias is estimated for each station, for each 4-week data interval, for LAGEOS-1 and LAGEOS-2 independently (i.e. 2 parameters per station). The data weight is 1 meter (larger values for poor stations), and the range bias estimates are unconstrained. In addition to the previous solutions, the orbit parameterization has been extended with weekly empirical cross-track accelerations; this has resulted in an improvement of the rms-of-fit from 4 cm to 2 cm, on average. The formal uncertainty of the Helmert parameters (based on station coordinates only) amounts to 1-2 mm; when EOPs are added for the determination of these parameters, the formal uncertainty of the Helmert parameters for scale and origin are reduced further, but the sigmas for orientation are increased (which is unexplained). The RMS differences of the x and y-pole solutions w.r.t. IERS C04 amount to 0.33 and 0.25 mas, respectively (UT: 0.11 ms).
In the series B analyses, the models and procedures from the series A analyses were used as much as possible. This includes the (equivalent of) 1-meter constraints on a-priori station coordinates and EOPs, which was expanded with a 1 m/day (equivalent) for pole rate and LOD. The addition of the Etalon observations has slightly increased the rms-of-fit. The addition of Etalon data has also slightly increased the rms difference of the xpole (ypole) solutions w.r.t. IERS C04, from 0.40 mas (0.32 mas) for B1 solutions to 0.42 mas (0.34 mas) for B2 solutions. This may indicate that a further tailoring of the analysis model for Etalon is necessary. For more details: see Appendix 6.
AUSLIG: no representative present; no new solution for either series A or series B submitted.
BKG is working on an update of the series A solutions and extension of the B solutions, which necessitates modifications to the Utopia software for orbit determination and parameter estimation.
CSR: no representative was present, but the CDDIS directories include a time-series of 28-day solutions labelled A and A2, starting on Jan 3, 1999 (i.e. the beginning of the series A series) and extending until (1st epoch) August 12, 2001. Upon closer inspection, these A2 solutions appear to be based on 4 satellites, so this series must be considered as a B2 solution; the A solutions (using data from the 2 LAGEOS satellites only) are hence considered as B1 solutions. In that case, the intervals of the time series do not match those of the series B solutions as agreed during the AWG workshop in Nice. Noomen will verify this conclusion with Eanes, and ask him to redo the analysis with a slightly changed measurement window (to allow a proper comparison and combination with other solutions) and resubmit with the appropriate B labels; then, the solutions now labelled A will be deleted from the CDDIS directory.
DGFI has re-analysed the series A data. Range biases are modelled if significant (based on a preliminary analysis); troposphere biases are not modelled or estimated. Depending on the quality of the stations, data weights of 2 or 4 cm are used. In estimating the EOPs, a polynomial fit through epoch values is made. The UT1 solutions based on LAGEOS-2 observations only show a residual signal, which may indicate a tailoring of the computation model to LAGEOS-1.
A solution for the series B component has not been made yet, since this necessitates further developments and studies of the EOP representation in the DOGS software. For more details: see Appendix 7.
GRGS: no representative present (at the time of passing this agenda item); no new solution for either series A or series B submitted.
IAA: no representative present; no new solution for either series A or series B submitted.
IAAK: no representative present, but new solutions for both series A and series B have been submitted.
JCET has a large experience in estimating EOPdots, but has been unable to provide a formal series B contribution to the project at this moment (either in the form of network/EOP solutions or in the form of a full presentation) because of a computer breakdown. Pavlis reports that the orbital fits for the B1 and B3 type of solutions (i.e. LAGEOS data only) typically amount about 10 mm (EOPs only) and 7 mm (EOPs + EOPdots), respectively.
NERC reports on their general experiences concerning the series A and B analyses. Typically, data weights of 10, 30 and 100 mm are used for the observations, depending on the quality of the station. The values used for the LAGEOS and Etalon spacecraft are identical. The computation model includes constant along-track accelerations (estimated at 5-day intervals) and 1-cpr accelerations in along-track and cross-track directions (estimated at 10-day intervals). The centre-of-mass value for the Etalon satellites is modelled differently for the different types of SLR systems (single-photon systems need smaller CoM corrections than high-energy systems); this difference can amount to more than 30 mm. As for the LAGEOS satellites, a similar procedure should be followed, but at this moment one unique value of 251 mm is used. The estimation scheme includes a range bias per station, which is estimated as a common parameter for all 4 satellites; this should also be changed in new computations. For more details: see Appendix 8.
3.2. Comparisons and combinations
ASI: Devoti reports on the ASI experiences on the comparison and combination of series A and series B solutions. As a general remark, he mentions that the error to be associated to the combination product(s) cannot be obtained from a direct propagation of the uncertainties of the input solutions (these are correlated obviously, by virtue of using the same observations), and that an alternative will have to be developed. This remains something to be investigated in the future.
ASI has been working on techniques to reduce the effect of estimating Helmert parameters, which (should) eliminate the systematic differences between various network solutions. One option is for a rigorous solution of these (7) parameters, based on both station coordinates and EOPs simultaneously; such an approach will also yield whether a specific network/EOP solution is self-consistent or not, and thus may be used as an additional quality assessment tool. Activities and results on this issue have been reported during previous AWG workshops. A drawback of any procedure where Helmert parameters are estimated is that the (statistical) quality of the combination product may be lost w.r.t. the original input solutions, so that the gain from combining solutions is not maximal.
As an alternative, ASI has studied a technique in which the systematic effects can be considered as insignificant, and therefore need not to be estimated at all. In such a case, the statistical quality of the individual solutions is retained in the combination product. To this aim, ASI has developed a quality assessment parameter which includes the absolute value of the Helmert parameters in relation to their the formal uncertainty; if these match, the systematic effects are considered insignificant and not contributing to the quality of the final results, and a combination or comparison can equally well be done without the estimation of Helmert parameters. This so-called looseness index should ideally have a value of 1; a significantly larger value indicates that the particular network/EOP solution deviates systematically from the reference or other solution(s), and should be handled differently: either with the estimation of Helmert parameters, or without, but with a scaling of the standard deviations.
As for the series A solutions that have been submitted so far, those computed by ASI, CRL, DGFI and NERC appear all reasonably consistent (i.e. estimation of Helmert parameters is not necessary, and will only lead to a weakening of the combination solution). As for IAAK, significant differences were found, and it proved necessary to apply scaling factors for the original standard deviations of up to 40,000 for about half of the monthly solutions, which indicates problems with the analysis (possibly related to stations with a poor data yield).
Also, the EOPs have been combined without elimination of (relative) systematic differences, but with scaling of the original standard deviations. Next, the results have been projected upon the IERS C04 series to provide the absolute reference, which includes the estimation of a bias and rate. Here again, the IAAK solutions appear to have a larger offset than the other solutions. Further details are given in Appendix 9. This technique has been in use by IGS for the combination of GPS orbits (and other parameters?) already for a large number of years.
As for the series B solutions, the technique described above has been applied to assess the quality of the various types of solutions. This has been done for the first three 4-week periods only. In summary, the UT results show a 2-weekly signal, which is believed to be of tidal origin. The looseness index and the overall quality have been established for the various types of solutions (with and without Etalon data; with and without estimation of EOPdots), for each analysis institute individually. Details are given in Appendix 9. The general conclusion is that there is no evidence for real improvements by adding Etalon data and/or extending the parameter base at this stage; further development, both of the comparison/combination technique and of the individual analysis strategies and computation models, is necessary. For more details: see Appendix 3.
BKG has studied the effect of a number of orbit parameterisation issues on analysis results. In particular, the influence of the arc length and the frequency of solving for empirical accelerations (radial, cross-track) have been investigated (the length of the individual data periods being fixed at 28 days). In general, longer time spans for arcs and parameters are to be preferred from the point of view of physical stability and quality, but are poor in terms of orbital fit; the reverse holds for the short time spans. Based on parameter quality, fit statistics and correlations between parameters, an arc length of 14 days and intervals of 14 (7) days for the constant (1-cpr) accelerations, respectively, appear to be optimal. Noomen remarks that the results of this interesting study need to be put in a broader perspective by also studying the effect of uncertainties of other elements of the computation model. It is recommended that this issue will be further developed in the pilot project benchmarking; Noomen will contact the coordinator of that project (Husson) and ask him to invite Mareyen as co-coordinator. For more details: see Appendix 10.
Tuesday September 18, 2001
The workshop continues on the second day with agenda item 3.2.
DGFI has worked on the comparison and combination of the series A solutions. The objective of their work is (1) to make a combination product, (2) to optimise this combination product, and (3) to develop a technique to arrive at a reliable and realistic error estimate of this combination product. For the quality control of the input solutions, 2 different criteria are used: (1) the rms of the differences between coordinates of a subset of SLR stations, according to the individual network solutions and the combination or reference network solution; and (2) the values for the eigenvalues of the covariance matrix; the 4 smallest values should be near 0 (reflecting the 4 degrees of freedom of the SLR network+EOP solution). In summary, it turns out that effectively none of the solutions satisfy the latter criterion, which may be attributed to over-constraining by the analysts or problems with the scaling and interpretation of the covariance matrices by DGFI. Mareyen suggests that the problem may also be related to the uneven station distribution. This issue is to be studied further. For more details: see Appendix 11.
3.3. Future of positioning + earth orientation
After some discussion, and having reflected on the results that have been presented during the 2 days of the workshop, it is concluded that all elements of this pilot project need some form of further development. As for the series A solutions, analysts will be given until December 1, 2001 to further refine their solutions, when deemed necessary. From then on, the series A solutions will be fixed and be available for reference purposes, possibly as a test bed for comparison and combination studies.
The series B solutions clearly need further refinement. Specific issues are the EOP treatment (in particular concerning the EOPdots), the modelling and parameterisation of the Etalon orbits, and the relative data weighting between LAGEOS and Etalon; other elements of the analyses may also need further development. It is not yet clear whether the full potential of the four-satellite solution has been realized.
A similar conclusion holds for the comparison and combination of solutions, both for series A and for series B products.
To extend the dataset of LAGEOS and Etalon observations that is currently available, Noomen will ask the ILRS MWG and GB for a continuation of the Etalon intensive tracking campaign by another 6 months, starting at the end of the current campaign (October 1, 2001) until April 1, 2002.
Also, it is decided that it still is too early for the development and release of requests for proposals (RFP) for contributions to an official ILRS analysis product and for an official ILRS combination centre. At this moment, the ILRS analyses and comparison/combinations are considered to include too many uncertain factors.
4. SINEX format
The SINEX format (its use and interpretation, plus extensions for the ILRS pilot project positioning and earth orientation) has been discussed intensively during all previous AWG workshops.
In spite of that, a number of issues were raised during this workshop. First of all, no decision was made in Nice on the Parameter Type (SOLUTION/ESTIMATE, SOLUTION/APRIORI, SOLUTION/MATRIX_ESTIMATE, SOLUTION/MATRIX_APRIORI blocks) when referring to biases which are common for more than 1 satellite. In Nice, it was decided to use the codes L1 and L2 for LAGEOS-1 and LAGEOS-2 biases, respectively. Pavlis suggested using the coding LC (LAGEOS, combined) for a common LAGEOS range bias, and equivalently E1, E2 and EC for range biases in the Etalon data. This was accepted.
In addition, Eanes has used two entries in the SOLUTION/STATISTICS block to provide the sum of the weights (units should be m**-2) and the rms of the residuals of the accepted observations (units: m). However, he used entries that are specific for GPS applications, notably the phase measurements sigma and the code measurements sigma, respectively. It is believed that this may lead to confusion. As a solution, Noomen will contact the coordinator of the SINEX format (Herring) and request the addition of 2 new (optional) entries in this statistics block, to be named SUM OF WEIGHTS and RMS OF RESIDUALS, respectively.
Eanes also commented on the value which is used for the sampling interval entry, also in the SOLUTION/STATISTICS block. So far, the ILRS AWG has recommended specifying a value of 120 sec when using LAGEOS data, but nothing has been foreseen when analysing Etalon observations or a working with a mixture of satellites. It is advised to use 300 sec when generating an Etalon-only solution, and to stick to the value of 120 sec when developing a combined LAGEOS/Etalon solution.
Finally, it has been suggested during previous workshops (and repeated here) that it might be very useful to supply a real normal equation directly, including matrix and right-hand side. This avoids the (sometimes still) cumbersome task of adding loose constraints and, at a later stage, removing them again. The current format description does have an option for providing the matrix component of the normal equation (the INFO option in the SOLUTION/MATRIX_xxx blocks), but not for the provision of the right-hand side. Noomen will discuss this issue with Herring.
The Parameter Type and Sampling Interval modifications will be included in the ILRS description of the SINEX format as soon as possible. The extension s in the STATISTICS block and the inclusion of a full normal equation will have to await coordination with Herring.
5. Pilot project harmonization
The coordinator of this pilot project, Husson, could not attend the workshop, and no new developments or plans could be reported.
6. Pilot project benchmarking
The coordinator of this pilot project, Husson, could not attend the workshop, and no new developments or plans could be reported. Noomen will contact Husson to ask whether Mareyen can serve as co-coordinator.
7. Pilot project orbits
The coordinator of this pilot project, Eanes, could not attend the workshop, and no new developments or plans could be reported.
8. Miscellaneous
8.1. Report of the IVS/IGS/ILRS working group
Appleby reported that he understood that the chairman of the working group (Corey) is working on a report on their investigation of the use of the VLBI technique to determine the phase centre of the GPS satellites. The SLR technique has not yet played any role in this question. When the final report is released, Appleby will send a copy to the ILRS AWG chairman.
8.2. Qualification of stations
The coordinator of this issue, Pearlman, could not attend the workshop, but he informed the chairman that this item would be handled by e-mail.
8.3. Next meeting
The following venues were considered as candidates for having an ILRS AWG workshop in close connection: the AGU Fall Meeting in San Francisco (December 10-14, 2001), the EGS General Assembly in Nice (April 22-26, 2002) and the AGU Spring Meeting in Washington (May 28-June 1, 2002). It was decided to have the next AWG workshop directly before the EGS General Assembly, i.e. on April 18 and 19, 2002.
However, it turned out that the slot of April 18-20 has already been claimed by the IERS, for a 3-day workshop on the ICRS and the Research Combination Centers (with overlap with the ILRS analysis activities). After having polled the attendants of the AWG workshop through e-mail, and having received no serious objections, it was decided to schedule the next AWG workshop for April 29 and 30, i.e. the Monday and Tuesday after the EGS General Assembly, also in Nice.
9. Action items
Noomen summarized the action items remaining from the previous workshop in Nice, and new ones coming out of this workshop (Appendix 12).
10. Closure
Noomen expressed his satisfaction with having had this workshop as planned, albeit with less participation that originally expected. He thanked the attendees for their participation, their input to the discussions and the contributions they have made to the ILRS pilot projects, and wished them a safe trip home.
Appendices:
1. Agenda
2. List of participants
3. Introduction, status report pilot project pos+eop (Noomen)
4. Overview pos+eop series A and B (Noomen)
5. EOP+network solution CRL (Otsubo)
6. EOP+network solution ASI (Luceri)
7. EOP+network solution DGFI (Mller)
8. EOP+network solution NERC (Appleby)
9. EOP+network comparison ASI (Devoti)
10. EOP+network comparison BKG (Mareyen)
11. EOP+network comparison DGFI (Mller)
12. ILRS AWG action items
Appendix 1: Agenda
1. opening
2. minutes and action items AWG Nice
3. pilot project positioning + earth orientation
3.1. individual contributions series A and B
= ASI
= AUSLIG
= BKG
= CRL
= CSR
= DGFI
= GRGS
= IAA
= IAAK
= JCET
= NERC
= questions/issues
- is data yield and quality Etalon good?
- experiences with analysis LAGEOS?
- experiences with analysis Etalon?
3.2. comparisons and combinations
= ASI
= BKG
= DGFI
= questions/issues:
- re-analysis 1999 LAGEOS data better?
- quality of EOPs?
- do EOP+EOPdot yield better products
than EOP only?
- do Etalons improve EOP/EOPdots
significantly?
- quality station coordinates?
= RFP for contribution to official ILRS combination product(s)
= RFP for official ILRS combination centre(s)
3.3. future of positioning + earth orientation
4. SINEX format
5. pilot project harmonization
5.1. status report
5.2. future
6. pilot project benchmarking
6.1. status report
6.2. future
7. pilot project orbits
7.1. status report
7.2. future
8. miscellaneous
8.1. report of the IVS/IGS/ILRS working group
8.2. qualification of stations
8.3. next meeting
? AGU Fall Meeting, San Francisco, Dec 10-14, 2001
? EGS General Assembly, Nice, April 22-26, 2002
? AGU Spring Meeting, Washington, May 28-June 1, 2002
9. action items
10. closure
Appendix 2: List of participants
Graham Appleby, NERC SGF ([email protected])
Franois Barlier, CERGA/GRGS ([email protected]
Richard Biancale, CNES ([email protected]
Randall Carman, BAE Systems/Moblas-5 ([email protected])
Roberto Devoti, Telespazio S.p.A. - ASI/CGS ([email protected])
Elisabeth Fragner, TU Vienna ([email protected])
Cinzia Luceri, Telespazio S.p.A. - ASI/CGS ([email protected])
Maria Mareyen, BKG ([email protected])
Horst Mller, DGFI Munich ([email protected])
Ron Noomen, DEOS ([email protected])
Konstantin Nurutdinov, Univ. of Newcastle, UK ([email protected])
Toshimichi Otsubo, CRL ([email protected])
Erricos C. Pavlis, JCET/GSFC ([email protected])
Toshitaka Sasaki, NASDA ([email protected])
Appendix 12: ILRS AWG action items
1. all submit copy of presentations
2. all review website
3. all update solutions series A (deadline Dec 1, 2001)
4. all update and complete solutions series B (pref. continuously; deadline Feb 1, 2002)
5. Appleby, Noomen prepare minutes of workshop
6. Devoti check parameter for assessing statistical significance Helmert transformation
7. Devoti check consistency IAAK scale factors
8. Dube prepare datasets LAGEOS-1/2, Etalon-1/2 (2001-2002)
9. Eanes develop strategy for orbits
10. Eanes re-analysis series with 1-week shift; correct label A -> B
11. Husson develop strategy for benchmarking
12. Mareyen co-organize benchmarking (Noomen: contact Husson)
13. Mller, Kelm, Mareyen investigate internal criteria sol+pos solutions
(4th eigenvalue)
14. Noll rename series A solutions GRGS
15. Noll move series B solutions to new series B directory
16. Noomen check inclusion of 3 new AWG members (3 locations)
17. Noomen, all forward invitations for conferences to ACs and AACs
18. Noomen SINEX issues
19. Noomen ask for extension of Etalon campaign
20. Nurutdinov e-mail SINEX problems -> Noomen
21. Pavlis provide new mapping function on ILRS web pages
21. Shelus (+4) revision of definition of ACs/AACs