From ries@csr.utexas.edu Fri Feb 28 12:31:29 2003
Date: Fri, 28 Feb 2003 01:45:35 -0600
From: "John C. Ries" <ries@csr.utexas.edu>
To: Jean-Paul.Berthias@cnes.fr, Gilles.Tavernier@cnes.fr,
    pascal.willis@ensg.ign.fr, Feissel@ensg.ign.fr
Subject: orbit combinations

Hi Everyone,

I thought I would try an 'IGS style' orbit combination where I 
averaged 4 Jason-1 orbits to see if the result is better than the 
individual orbits. Not surprisingly, one ends up with an average that 
is just about the average of what went in. In a combination of 2 very 
good orbits (Pascal's and Bruce's), my reasonably good orbit, and a 
somewhat poorer DEOS orbit. I found that the combination of all 4 is 
just about in the middle. The combination of Pascal's and Bruce's was 
about as good as either one, since they were roughly equal, and 
better than the combination of 4.

It would seem that the 'best' orbit should be selected. It probably 
cannot be created through averaging multiple orbits. On the other 
hand, if the IGS is merely trying to create a 'reliable' product, 
rather than the best product, averaging may dilute errors in any 
particular orbit. However, analysis of IGS orbits for Jason-1 
processing by Scott Luthcke indicates that some IGS orbits can be 
quite bad, so even this goal is not always obtained through simple 
averaging.

If we ever decided to produce an orbit product, I would recommend a 
single 'best' product, verified against alternative orbits for 
blunders, than any sort of combination procedure.

Best regards,
JR