Message-Id:
Date: Thu, 6 Dec 2001 08:58:22 -0800
To: igsleo@igscb.jpl.nasa.gov
From: Thomas P Yunck
Subject: [IGSLEO-32] SLR Comparisons
Sender: owner-igsleo
Precedence: bulk
******************************************************************************
IGS LEO Mail 06 Dec 09:23:09 PST 2001 Message Number 32
******************************************************************************
Author: T. P. Yunck
To: IGS LEO List
Here are some comments I sent yesterday to Henno regarding the new
SLR comparisons with GPS CHAMP orbits (before I had seen Da Kuang's
comments). These deal principally with the anomalously low JPL-NCL
ratio discussed at the end of Henno's web page. --Tom
---------------
Dear Henno,
I've been examining your lovely new website with the SLR comparisons
and I must say it's remarkable. You've done a tremendous amount in a
very short time, and done it very rigorously. Your ratios of direct
agreement to combined SLR agreement are especially interesting. I
suspect there's a fairly simple reason for the anomalously low
JPL-NCL value -- the same reason, in fact, for the anomalously close
JPL-NCL agreement in Fig 2 of your original comparison.
No, no-one is using SLR data in their orbit solutions; but not only
do JPL and NCL use the same software, I suspect they both (I don't
actually know what NCL does) adopt a more highly kinematic strategy
than the others. That means their solutions at each step depend more
strongly on the momentary observing geometry and measurement error,
which are common to all solutions, and are less sensitive to arc
lengths, epoch times, initial conditions, etc, that can strongly
affect the details (though not necessarily the overall performance)
of dynamic solutions.
In the kinematic limit, the JPL and NCL solutions would become
virtually identical, even while their absolute error (departure from
SLR) would increase, and your ratio would approach zero (just as the
diagonal elements of your ratio matrix must be zero). That's just a
longwinded way of saying that kinematic solutions from the same set
of data will tend to be more highly correlated than dynamic
solutions, while not necessarily being more accurate, and thus will
yield lower ratios. And that of course is why the SLR comparisons
are so valuable as an independent, absolute check. Nothing more is
really needed.
Also, I notice that SLR site 7 is rather problematic -- everybody
disagrees with it, all by about the same rather large amount -- so it
should probably be omitted. Eyeballing your charts, I estimate that
without site 7 the revised 1-way RMS values would be roughly 10.4,
15.2, 17.5, 18.8, and 32.3 cm. Even allotting 3 or 4 cm to SLR
error, the GPS-only error would not come down much (e.g., 9.6 cm for
JPL assuming a 4 cm independent SLR error). To estimate the 3D
error, I might divide your 2.33 multiplier by root-2 (i.e., remove
the denominator 2 from your combining equation), giving about 1.65
(which is not too far off a plausible a priori guess of root-3 in
going from 1D to 3D). That would then agree well with our own
estimates of the JPL 3D CHAMP orbit error, derived by other means --
mainly overlap comparisons.
--
-----------------------------------------------------------
T. P. Yunck Ph: 818-354-3369
JPL, M/S 238-540 Fax: 818-393-6686
4800 Oak Grove Drive e-mail: tom.yunck@jpl.nasa.gov
Pasadena, CA 91109
-----------------------------------------------------------