A discussion ensued on how to proceed with this benchmarking project. Since a number of analysis groups have a problem with the implementation of the along-track, piecewise continuous acceleration model, it was decided to reformulate the “A” element and have the orbit integration done without those accelerations. This needs to be included in the description available on the ILRS web pages, as well as brought to the attention of the analysts. Other descriptions that need to be made more explicit refer to the data weighting relative to the a priori standard deviations of parameters, the epoch of station coordinates, and the UT vs. LOD issue (action item Husson, analysts).
Since the project is to result in a pass/fail grade for individual solutions, specific test criteria were discussed. First of all, the range corrections (center-of-mass, troposphere and relativity) will be inspected; the rms of the difference w.r.t. a reference standard may be no more than 0.1 mm. This reference will be generated from the average of the “D” solutions, but these already are known to be consistent at a very high level.
As for station coordinates and EOP solutions, a reference solution will be developed from the “D” solutions, first mapped onto ITRF2000 (using the ILRS AWG Core Stations only). In this process, editing and weighting will have to be applied to a certain extent, depending on the actual consistency of the solutions. The final pass/fail verdict will also be based on “D” solutions, and the provisional criterion is for the rms difference w.r.t. the average to be within 2 times the rms of the position/EOP residuals coming out of the reference determination.
The orbits element of the pass/fail grade will consider both “A” and “D” results. The reference for the “A” comparison will be a direct average (leaving room for editing at this moment), whereas the standard for the “D” solutions will be developed from the solutions propagated into the ITRF2000 frame (the orbit solutions are typically given in an earth-fixed reference frame); again, the editing issue is left open for now. The pass/fail result of the final contributions is again a provisional factor 2 times the residuals of the computation of the average. Importantly, the radial, cross-track and along-track elements of the orbit (differences) will be judged individually.
The residuals will not be used in the pass/fail assessment, since it is recognized that they do not provide new, independent information.
The analysts will still have to provide “B” and “C” solutions, but they may be used to help identify possible problems with rejected “A” and/or “D” solutions by comparing them with similar results computed by other analysis groups. No reference solutions will be made for these “B” and “C” solutions. Analysts are requested to generate new solutions before June 1 (action item analysts), and subsequently reference solutions will be generated before June 15 (action item Husson/Torrence) and pass/fail assessments on individual solutions will be given on June 30 at latest (action item Husson).