Fourth ILRS AWG Meeting (Nice 2001)
Minutes of ILRS Analysis Working Group Workshop #4
Nice, France, March 22-23, 2001
(written by P.Shelus, R. Noomen)
Agenda: see Appendix 1.
Attendees:
Graham Appleby, Francois Barlier, Richard Biancale, Roberto Devoti, Richard Eanes, Ramesh Govind, Van Husson, Rainer Kelm, Cinzia Luceri, Maria Mareyen, Joelle Nicolas, Ron Noomen (chairman), Konstantyn Nurutdinov, Mike Pearlman, Erricos Pavlis, Pete Shelus, Philippe Yaya. For a more detailed list: see Appendix 2.
Thursday March 22, 2001
1. Opening
Introductionary remarks were given by Noomen. The chairman thanked the analysts for their contributions, and thanked Husson for all the work done for the ILRS. The EGS organization is credited for arranging and paying for the meeting room.
2. Minutes AWG Matera
The minutes of the previous ILRS AWG meeting were made by Luceri, Devoti and Noomen, for which they are thanked. They are briefly discussed.
The main items in the Matera minutes are the extension of the pilot project "pos+eop" to cover 1 year, a further development of the SINEX format (in particular on the contents of some (new) blocks and the scaling of standard deviations and (co)variances), and the constraints handling. The text of the Matera minutes is accepted.
3. Actions since AWG Matera
3.1. Reports, presentations, membership
Presentations: papers representing the ILRS were given at the 1999 Fall Meeting of the AGU (on EOPs; Eanes), and during the EGS General Assembly of 2000 (SLR orbits; Noomen). Noomen encourages the analysts to take a more active stance in meetings and conferences by submitting more ILRS-related papers, to better promote ILRS better. The remark is made that sometimes the invitations for conferences and such are not circulated timely, or are not distributed within the community at all. Action item all: forward such invitations to the ILRS ACs and AACs (use the email exploders for this).
Reports: Both presentations mentioned above have resulted in written papers, and both of them have just passed the reviewing stage.
Membership: Springer has left the geodetic community to work for a commercial company. Weber (University of Wien, but working at AIUB for a significant part of his time) will be invited to replace Springer (action item Noomen). In addition, Mareyen (BKG) and Nurutdinov (NCL) will join the AWG; both already take an active role in ILRS analysis activities and discussions.
Organization: the question is raised what the exact difference is between an ILRS AC and an ILRS AAC. According to the current Terms Of Reference, an AC is required to contribute with both operational and scientific data analyses products. The current grouping, based on the invitations and responses received at the time of the initiation of the ILRS (i.e. more than 2 years ago), may have become obsolete and may be in need for revision. A committee is formed to study the definition of ACs and AACs and work on a description which better reflects the present situation. Members are Shelus (chairman), Govind, Husson, Noomen and Pearlman.
For more details on agenda items 1-3: see Appendix 3.
4. Pilot project "positioning + earth orientation"
4.1. Individual contributions
Brief presentations are given by the various contributors:
ASI computes a combined LAGEOS-1 and LAGEOS-2 solution, using GEODYN and SOLVE. Station coordinates and EOPs (a priori Bulletin B) are common, range biases are estimated per station and per satellite (100 m a priori sigma). A minimal internal constraint is applied. Station coordinates residuals w.r.t. ITRF97 amount to about 15 mm for good stations, with values up to 13 cm for poor stations. For more details: see Appendix 4.
AUSLIG weighs all stations equally (1 m). Range and timing biases are solved for on a pass-per-pass basis (a priori sigmas are 1 m and 0.001 sec). The outcome of the range biases is about 1 cm; the station coordinates show an agreement with ITRF of about 15 mm. Eanes makes the comment that the relation between data weights and a priori sigmas of the biases is such that the latter are overconstrained as a result. For more details: see Appendix 5.
BKG uses a priori sigmas of 1 or 10 m (station coordinates), 1 m (pole) and 10 m (UT). Mareyen questions the estimation of a common range bias for both LAGEOS-1 and LAGEOS-2; sometimes a combination solution introduces conflicts. Various techniques for expressing the internal quality of solutions are presented. As a result of this, it is evident that the quality of the network is not constant (a consequence of variations in data yield). Eigenvalues for all 4-week data periods are shown and discussed. For more details: see Appendix 6.
CRL (Appleby reports on behalf of Otsubo) has estimated the parameters that have been required: 1-day UT and pole, and station positions at 28-day intervals. In addition, range biases (common for LAGEOS-1 and LAGEOS-2) have been estimated. As for data weighting, the stations have been divided into 4 different categories and processed accordingly. For more details: see Appendix 7.
CSR has provided 3 solutions: (1) solving for all parameters requested plus a 5x5 gravity field model (to improve on the standard linear model for C2,1 and S2,1), (2) solving for station coordinates and EOP only, and (3) a solution constrained to "hidden" parameters. The stations have been divided into 4 categories, based on their quality, with various values for the data variances (2.5 cm2, 5 cm2, 10 cm2 and 20 cm2, respectively). One bias per station has been estimated in all scenarios (common for both satellites). The second solution overconstrains the UT parameters, because acceleration solutions (constant AT, 1cpr R and N) from solution 1 are used and constrained.
DGFI has provided independent solutions for LAGEOS-1 and LAGEOS-2, plus a combination solution. The data weights are either 2 cm or 4 cm, depending on the quality of the station. Biases (either range or time) are estimated in a preliminary analysis first, and kept fixed at this value in the final analysis if exceeding 3 cm or 0.03 ms, respectively; if smaller, they are kept fixed at zero. Biases are handled on a pass-by-pass basis. The analysis includes nutation corrections, but corrections for the troposphere model (such as a tropo bias) are not included. For more details: see Appendix 8.
GRGS does the analysis in a 2-step approach: first, orbits are computed using GINS, and next EOPs and other parameters are computed using DYNAMO. The overall fit is about 4 cm, and about 3 cm for the best stations. The dynamic model includes a constant and 1cpr AT acceleration, plus accelerations in the BX and BY directions. Range biases are not estimated or modeled. The a priori sigmas are 3 mas for pole, 0.3 ms for UT and 1.7 mm, 1.7 m or 17 m for station coordinates. It appears as if some of the parameters are overconstrained. Differences of the EOP solutions w.r.t. IERS C04 are presented. For more details: see Appendix 9.
JCET discusses the computation model that has been used in the analysis. It includes a new model for ocean tides (GOT99.2), cross-track 1-cpr accelerations (estimated for each 2 days), and range biases for LAGEOS-1 and LAGEOS-2 independently, once per week. For more details: see Appendix 10.
NERC applies a data screening before the actual analysis: only stations with more than 30 normal points are accepted. The data weights are either 10 mm, 30 mm or 100 mm, depending on the quality of the station. For more details: see Appendix 11.
4.2. Comparisons and combinations
ASI has addressed various aspects of the solutions. First of all, they have studied the statistical quality of the solutions by looking at the formal uncertainties of the Helmert parameters, for the cases of using (i) station coordinates only and (ii) both station coordinates and EOPs, to estimate the Helmert parameters. It turns out that the addition of the EOPs results in a better statistical quality of the rotation parameters in particular. The differences of the individual solutions w.r.t. ITRF97 is about 1-2 cm, with AUSLIG at slightly larger values. ASI has also used the results of a Helmert transformation based on coordinates only and a similar transformation based on coordinates and EOPs to investigate the internal consistency of individual solutions. This appears to be a problem for all contributions. In addition, ASI has made a time-series of combination solutions, based on all individual contributions. The input solutions were scaled and edited such that the weighted rms-of-fit of the individual solutions amounts to about 1. The combined solutions show a difference w.r.t. ITRF97 of about 1.5 cm for the first 9 intervals; after that the difference increases to about 3 cm. The critical points are the proper choice of (or process of obtaining) the weighting factors, errors of the combined solution and internal constraints. For more details: see Appendix 12.
When doing comparisons and making combinations, BKG has struck upon various numerical problems. Most of these appear to be caused by stations with a small number of observations. In theory the problem should not occur for stations with 3 or more observations (the situation of just 1 or 2 observations does occur). However, also in case of a single pass with e.g. 8 data points, numerical problems may occur. As a result, an inversion of the matrices using a standard Gauss technique is not always possible, and one may have to use a Singular Value Decomposition instead. This may pose a problem to the people involved in making combinations, in particular. Solutions to overcome this problem may be (i) to require a minimum number of observations or some measure of geometric strength, and to reject a station altogether if this is not met, or (ii) to put more emphasis on a subnetwork of "core" stations, which have a high quality and a reliable data production. A discussion follows, with no clear outcome. For more details: see Appendix 6.
CSR has done a comparison of individual solutions by inspecting solutions for the baselines between stations. In this way, singularities caused by network (orientation) deficiencies may be avoided. Typically, baselines covering short distances show good consistencies. As an example of this, the baseline between Graz and RGO shows a scatter of 2 mm only. It is recommended not to use ITRF97 for such purposes, because it has an error in absolute scale of 0.7 ppb. ITRF200 is a much better choice when making comparisons.
DGFI has identified a number of different levels along the way towards a combined product. In level 1, the data files are tested for errors, such as in a priori coordinates or eccentricities. This was done for 8 different contributions (GRGS and CSR were not included because of their late submissions). This step yielded problems with the solutions of IAA and JCET. Level 2 concerns a loose constraint solution analysis. Here, a direct comparison is made between the a priori and the a posteriori coordinates solutions. This yields (and should yield) adjustments in the order of 10-20 cm. Level 3 consists of a recovered unconstrained normals analysis. Here, the normals are recovered and the constraints are removed. This should yield the normal euqation based on observations only. When converted to eigenvalues, one should expect 4 small values (reflecting the inherent singularities in the datasets: 2 station latitudes, 1 station longitude and 1 UT). The remainder of the eigenvalues should be large. In any case, the eigenvalues should always be positive. In level 4, a minimal constraint is applied (such as constraining 2 latitudes, 1 longitude and 1 UT), and the results are inspected. This is not the optimum way for a combination product, but it is useful for testing since all solutions are treated in a standard way. Level 5, finally, consists of a combination of the unconstrained normals equations. For more details: see Appendix 13.
HTSI has studied the relation between range biases and station coordinates, in particular baseline lengths and station heights. The 1999 dataset had some bias problems. If biases are not modeled properly, they could adversely affect site coordinates. There is a strong correlation between range bias estimates and station height variations. For the best performing systems, the coordinate solutions would agree significantly better if no biases were estimated. This was demonstrated by adjusting station heights by a value of 1.3 times the range bias. This implies data treatment and/or modeling is inducing errors in the station positions. Seasonal variations in atmospheric pressure (i.e. pressure loading) are known to exist in Australia and China at the 10 and 20 mbar level, respectively. Pressure loading will induce systematic mm level monthly station height variations. For more details: see Appendices 14 and 15.
After this presentation, it was commented that the troposphere model also plays an important role in station height determination. Pavlis will provide a new mapping function.
JCET has made a comparison of 4 different solutions (because of time constraints): ASI, AUSLIG, CSR and JCET. Out of the available data, 23 stations have been selected. A comparison with ITRF97, using variances only (i.e. disregarding the correlations between parameters) shows that some solutions agree almost perfectly with the a priori values; this hints at overconstraining. More general, all solutions agree with ITRF97 at the level of their error estimates. A plea is made to include the EOP rates (time derivates, including LOD) in the next step of the pilot project. For more details: see Appendix 16.
NCL uses the Tanya software to compare and combine network solutions. In doing so, each individual network solution is read and tested against a catalog file (which checks issues like eccentricity, standard deviations etcetera); SINEX format errors are reported when detected. In the next step, the input solutions are deconstrained. Then, an interactive process is started, where all network solutions are subject to a Helmert transformation and the (co)variances are scaled in such a way that the post-fit residuals become more or less identical. In this process, only stations with at least 3 solutions are used. If a certain (station) contribution has a residual value which is too large, the contribution is eliminated automatically. This process is repeated until convergence. The final solution is labeled "G-snx", and it is next mapped onto an ITRF solution. A discussion arises whether a relatively small set of core stations should be used for this purpose (such as 7090, 7110, 7403 and 7840), or the full network; a similar thing is in effect for GPS (51 core stations), although it is realised that the magnitude of the ILRS station network is much smaller. A final remark: station 7837 has been excluded manually, because of a problem with Tanya. For more details: see Appendix 17.
Friday March 23, 2001
4.2. Comparisons and combinations (continued)
Questions/issues:
- station coordinates? EOPs?
- L1+2 vs L1 vs L2
- between institutes - editing
- weighting
- is L1+2 better?
- is the combination product better?
- RFP for official ILRS combination center(s)
4.3. Issues
procedures
SINEX format
analysis standard
4.4. EOP: UT vs LOD
4.5. Future of "positioning + earth orientation"
multi-sat?
time-series?
combinations?
operational analyses?
official ILRS product(s)?
time line?
The remaining part of agenda item 4 is discussed. It is recognized that in particular the 3rd component of the EOPs is a very weakly observed by SLR, because of its correlation with the station longitudes and with the ascending node of the satellite orbit(s).
In the current situation, EOP solutions provided by the ILRS Acs or AACs receive very little weight in combination solutions, and the UT/LOD component is sometimes ignored at all.
In addition, a debate has been going on within the community for some time about the best representation of this third component: should it be UT1-UTC or should it be its time-derivative, LOD? During the previous workshop in Matera more and more doubts on UT as the best representative parameter were expressed, and it was decided then to discuss this topic again in Nice and to consider the use of LOD instead.
Several analysts mention that an interesting alternative to improve the LLR/SLR EOP determination exists: the inclusion of the satellites Etalon-1 and Etalon-2. These satellites are flying at such a high altitude that they are hardly sensitive to variations in the gravity field and tidal forces, so that (the precession of) their ascending nodes can be modeled very accurately. After a lengthy debate, it is concluded that the addition of the Etalon satellites to the current satallites LAGEOS satellites can be expected to result in significant improvements of our EOP product.
To substantiate this, the ILRS will be asked officially to organize an intensive tracking campaign for Etalon-1 and -1, starting on April 1 and taking 6 months. The data will be analyzed together with the LAGEOS observations, as part of the extension of the pilot project "pos+eop".
In addition to UT/LOD, the estimation of GM and assessment of station characteristics (range biases, frequency biases) and possibly other parameters will also benefit from the addition of these high-flying satellites.
Apart from adding the Etalon data, it is also decided to add the time derivatives of the EOPs (i.e. dXPOLE/dt, dYPOLE/dt and LOD) as additional parameters in the pilot project (still on a daily basis).
The SINEX format has provisions for this. Analysts will be invited to contribute to 4 different cases, covering half a year period starting on April 1, 2001: (1) LAGEOS-1/2, EOP, (2) LAGEOS-1/2 + Etalon-1/2, EOP, (3) LAGEOS-1/2, EOP + EOPdot, and (4) LAGEOS-1/2 + Etalon-1/2, EOP + EOPdot. The datasets will have to be analyzed in batches of 28 days, and results are to be submitted to CDDIS as soon as possible. The first evaluation will take place in September, in the next AWG workshop (held in conjunction with the SPIE conference in Toulouse, France).
In addition, all analysts will be invited to reanalyze the 1999 LAGEOS-1/2 dataset, following the remarks and comments given during the presentations and discussions during this workshop (such as the transition to ITRF2000 as a priori for station coordinates).
For more details: see Appendix 18.
5. SINEX format
The SINEX format as it is now in use for the pilot project "pos+eop" is discussed. This includes a.o. mandatory elements in the "COMMENTS" block (intended at better describing the analysis technique and products), plus a better definition of the scaling of the statistical uncertainties (both a priori and a posteriori).
Eanes has a request for more specific parameters to be added to the list of possible entries in the SINEX format, to accomodate future needs of our community and to complete the full description of the CSR contribution for this project in particular. To cut the discussion short, analysts will be invited to provide such feedback by means of email correspondence when back home (action item Noomen). For the moment, the SINEX format suits the direct purposes of the "pos+eop" project.
In addition, there is some debate on the existence of various versions of the SINEX format. Husson is asked to make an inventory of all possible SINEX descriptions, both offical versions and ILRS specific ones.
For more details: see Appendix 19.
6. Pilot project "harmonization"
Husson gives a status report on the pilot project "QC harmonization". The results obtained by the various analysis centers appear to show quite some difference, but is is evident that the range bias trends for a particular station, as observed by the analysts, follows a similar pattern. Differences in station positions are responsible for this: the range bias is correlated with station height, and the time bias relates to the horizontal components of the station position. Currently, only range bias stabilities from CSR are used in the Global Performance Report Card, but future report cards will incorporate bias stability metrics from each participating LAGEOS analyst center.
In addition, Husson discusses the "colocation" technique, which is capable of comparing relative biases for networks at scales of about 5000 km. This technique can be used to identify relative bias (changes) at the level of 2-4 mm. A typical value for absolute range bias stability is 5 mm.
For more details: see Appendix 20.
7. Pilot project "benchmarking"
Husson gives a brief status report on the pilot project "benchmarking". He will make a draft proposal for the development of this project and circulate that. This will be discussed in the next AWG workshop.
For more details: see Appendix 21.
8. Pilot project "orbits"
Eanes gives a brief status report on the pilot project "orbits". Issues are customers, comparison of results, satellites to be included, product format and others. Eanes will make a draft proposal for the development of this project and circulate that. This will be discussed in the next AWG workshop.
9. Miscellaneous
9.0. ILRS core stations
Pearlman discusses the qualification of the network of SLR/LLR stations. Currently, a station is considered an ILRS core station if it obtains 400 passes per year on the LAGEOS satellites, 1000 passes per year on LEO satellites, and if it shows a certain level of system stability. There appears to be a need for another, more subtle categorization of stations, which provides room for "newcomers", but also gives full credits to the "best of best" stations and stations that contribute significantly to one category but that cannot contribute to another because of technical limitations.
A new advisory subgroup will be tasked with the assessment of the initial quality of new stations that come on line. This subgroup comprises Appleby, Husson and Pearlman.
9.1. Report of the IVS/IGS/ILRS working group
Appleby reports on the status of this working group. The main activities appear to be with IVS and IGS, although there is some general doubt on the feasability of the ideas to use a combination of techniques to solve the problem of GPS phase center location. The overall coordinator (Corey) appears to be unable to include a possible contribution from ILRS at this moment. ILRS will continue its "wait and see" attitude.
For more details: see Appendix 22.
9.2. Handling of bias problems
Husson raises the issue of how to handle biases (or other data problems, for that matter) that are detected by a station (or someone else) at a certain moment. After a brief discussion, it is concluded that (1) the corrections are to be applied by the stations and the data is to be resubmitted, and (2) the corrections (and actions) are to be reported officially, i.e. through the CDDIS Bulletin, SLRmail and/or other appropriate channels which are used by the analysts.
9.3. Calibration Colloquium (September 2001): topics for discussion
Husson raised the question for items which could be discussed during the SPIE meeting in Toulouse (September 2001). As for the AWG, 2 things were mentioned: calibration issues and GM values.
9.4. Next meeting
In principle there are a number of candidates for a next venue of the ILRS AWG: the EOS/SPIE Symposium in Toulouse (September 17-21), the SALRO workshop in Riyadh (2nd half of September) and the AGU Fall Meeting in San Francisco (December 10-14). Since the next ILRS General Assembly (and related meetings) will be held in conjunction with the SPIE symposium, the Monday and Tuesday of the week in which this symposium takes place are chosen for our next workshop: September 17 and 18, 2001.
10. Action items
Noomen summarized the action items coming out of this workshop. See Appendix 23.
11. Closure
Noomen thanked the attendees for their participation, their input to the discussions and the contributions that they have made to the ILRS pilot projects.
Appendices:
1. Agenda
2. List of participants
3. Introduction, status report pilot project "pos+eop" (Noomen)
4. EOP+network solution ASI (Luceri)
5. EOP+network solution AUSLIG (Govind)
6. EOP+network solution and comparison BKG (Mareyen)
7. EOP+network solution CRL (Appleby)
8. EOP+network solution DGFI (Kelm)
9. EOP+network solution GRGS (Yaya)
10. EOP+network solution JCET (Pavlis)
11. EOP+network solution NERC (Appleby)
12. EOP+network comparison ASI (Devoti)
13. EOP+network comparison DGFI (Kelm)
14. Station coordinates comparison HTSI (Husson)
15. SLR bias detection capabilities using collocation HTSI (Husson)
16. EOP+network comparison JCET (Pavlis)
17. EOP+network comparison NCL (Nurutdinov)
18. "pos+eop" future (Noomen)
19. SINEX (Noomen)
20. "harmonization" (Husson)
21. "benchmarking" (Husson)
22. IVS/IGS/ILRS working group (Appleby)
23. ILRS AWG action items
R. Noomen, Delft, June 24, 2001
1. opening
2. minutes AWG Matera
3. actions since AWG Matera
3.1. reports, presentations, membership
4. pilot project "positioning + earth orientation"
4.1. individual contributions
ASI
AUSLIG
BKG
CRL
CSR
DGFI
GRGS
IAA
JCET
NERC
4.2. comparisons and combinations
ASI
BKG
CSR
DGFI
HTSI
JCET
NCL
questions/issues:
- station coordinates? EOPs?
- L1+2 vs L1 vs L2
- between institutes
- editing
- weighting
- is L1+2 better?
- is the combination product better?
RFP for official ILRS combination center(s)
4.3. issues
procedures
SINEX format
analysis standard
4.4. EOP: UT vs LOD
4.5. future of "positioning + earth orientation"
multi-sat?
time-series?
combinations?
operational analyses?
official ILRS product(s)?
time line?
5. SINEX format
parameters
6. pilot project "harmonization"
6.1. status report
6.2. future
7. pilot project "benchmarking"
7.1. status report
7.2. future
8. pilot project "orbits"
8.1. definition
8.2. products format
8.3. future
9. miscellaneous
9.1. ILRS core stations
9.2. report of the IVS/IGS/ILRS working group
9.3. handling of bias problems
9.4. Calibr / Colloquium (September 2001): topics for discussion
9.5. next meeting
10. action items
11. closure
Appendix 2: List of participants
Graham Appleby, NERC SGF ([email protected])
Francois Barlier, OCA/CERGA/GRGS ([email protected])
Richard Biancale, CNES ([email protected])
Roberto Devoti, Telespazio/ASI-CGS ([email protected])
Richard Eanes, CSR ([email protected])
Ramesh Govind, AUSLIG ([email protected])
Van Husson, HTSI ([email protected])
Rainer Kelm, DGFI ([email protected])
Cinzia Luceri, Telespazio/ASI-CGS ([email protected])
Maria Mareyen, BKG ([email protected])
Joelle Nicolas, OCA/CERGA/GRGS ([email protected])
Ron Noomen, DEOS ([email protected])
Konstantyn Nurutdinov, University of Newcastle ([email protected])
Mike Pearlman, CfA ([email protected])
Erricos Pavlis, JCET/GSFC ([email protected])
Pete Shelus, McDonald Observatory ([email protected])
Philippe Yaya, GRGS ([email protected])
Appendix 23: ILRS AWG action items
All - submit copy of presentation(s)
All - review website
Dube - prepare datasets Lageos-1/2, Etalon-1/2 (2001)
Eanes - develop strategy for "orbits"
Husson - inventory of SINEX descriptions
Husson - develop strategy for "benchmarking"
Noomen - add 3 new AWG members; invite Weber; eliminate Springer
Noomen, all - forward invitations for conferences to ACs and AACs
Noomen - write and submit proposal for tracking campaign Etalon-1/2
Noomen - write and distribute invitation for "pos+eop"
Noomen - invitation for feedback on SINEX contents
Noomen+Shelus - minutes AWG meeting
Pavlis - provide new mapping function
Pavlis - contact Scherneck for complete list of ocean loading coefficients
Shelus (+4) - revision of definition of ACs/AACs (deadline April 10 and April 30)