[IGSMAIL-317] Discussion on Orbit Metrics

[G.]

IGS Electronic Mail       18-AUG-1993 08:46:04       Message Number 317
***********************************************************************
 

>From: G. Blewitt
Subj: Discussion on Orbit Metrics
---------------------------------

  I would like to open the discussion on orbit metrics a little further 
(for a note on the limitations of IGSMAIL for open discussions, see the end of
this message). Clyde Goad responded positively to my message IGSMAIL #311 and 
I would like like to present a few more thoughts on this topic to get 
analysis centers thinking before the Ottawa meeting.
  I agree with Clyde in that I think a some type of standard deviation 
statistic should go with each orbit. Although I believe all centers
currently put such statistics in a table in the summary file, we should 
discuss putting values in the orbit files themselves. 
   I also think that in  addition to this, there should at least be a 
flag .TRUE. or .FALSE. to indicate whether or not the analysis
center recommends using the product for precise regional geodesy. This is
because the standard deviation may not be a homogeneous measure between 
analysis centers.  I think, though, that if the orbit is very bad
(e.g., known maneuver with > 100 meter residuals) then perhaps
it would indeed be better to remove the orbit. Perhaps the criteria for this
should be oriented toward baseline accuracy.  For example: if including
the "bad" orbit would cause > 1 part per million error, then don't include it.

  As an elaboration on the .TRUE./.FALSE. scheme, individual orbits could
be assigned a number as follows:
         Quality Index       Characteristics
         -------------       ---------------
            0                Not recommended for anything
            5                Baseline accuracy 10**-5 to 10**-6
            6                Baseline accuracy 10**-6 to 10**-7
            7                Baseline accuracy 10**-7 to 10**-8
            8                Baseline accuracy 10**-8 to 10**-9
where the "baseline accuracy" is the desired requirement of the surveyor.
Of course, we could try to design some standard tests to more precisely
and uniformly assign these indices (e.g., using the fixed orbit products 
to test the accuracy on known baselines of various lengths). This is tricky,
since at least 4 satellites must be in view, so it might be hard to uniquely
assign an index to a given satellite. On the other hand, it would not be
too difficult, if say almost all satellites were of quality 8 and just a 
couple of satellites with lower quality.  

   To get started, however, a rigorous definition of the index is not 
necessary, since even a reasonable guess on the part of the analysis center 
is better than no information at all.

   I think that orbit standard deviations are a good idea; however, I don't 
think we should expect general users to know how to interpret them.  Standard
deviations better serve the analysts themselves, for purposes of assessing
improvements/degradations in system performance.

   Finally, if other analysis centers do start submitting those orbits which
are known to be problematic rather than simply removing them from the 
file, then we need to be careful with the global RMS metric in the 
pair-by-pair orbit comparisons between analysis centers. The global RMS will 
be dominated by the "bad" satellites, which was the original reason for 
alerting IGS in my mail message IGSMAIL #311.  It would be appropriate to 
remove the satellites flagged "bad" from the RMS calculation, assuming that 
regional geodesists will not use such satellites.  (Another possible scheme, 
using a "weighted RMS", is not as useful, because in practise the end-user 
typically does one of two things with an orbit product: either 
(a) use and fix the orbit, or (b) don't use the orbit).

   I agree with Clyde that this should be a topic of discussion at Ottawa. 

      			Geoff Blewitt
      			JPL IGS Analysis Center
      			geoff at logos.jpl.nasa.gov

P.S. Neither IGSMAIL or IGSREPORT seem to serve the purpose of a forum
for discussion.  I propose a new mail box, IGSANALYSIS for passing around
ideas and relevant progress reports on research between analysis centers. 
The advantages of using this rather than personal e-mail are 
(1) referencable and archived IGSANALYSIS mail, (2) updated distribution 
list automatically used by all, and (3) it would stimulate more discussion,
and I think such discussion has been (understandably) lacking in the scramble 
to produce results for IGSREPORT every week.
  IGSANALYSIS messages should be less formal than the more "official" tone 
of IGSMAIL and IGSREPORT messages, otherwise it will not serve its intended 
purpose.  If you work for an IGS Analysis Center, please send me personal 
e-mail to register your opinion on IGSANALYSIS, and I'll try to summarize
the level of support for this in advance of the meeting at Ottawa.