From Pascal.R.Willis@jpl.nasa.gov Fri Feb 28 12:36:29 2003
Date: Wed, 26 Feb 2003 15:50:01 -0800
From: Pascal Willis
To: Daniel Gambis ,
Richard Gross ,
Zuheir Altamimi ,
Meise Barbara
Cc: Martine Feissel ,
Gilles Tavernier ,
Jean-Paul Berthias ,
Carey Noll ,
Jean-Jacques Valette ,
Laurent Soudarin ,
Serguei Kuzin , CRETAUX@sc2000.cst.CNES.FR,
"John C. Ries"
Subject: DORIS/EOP precision vs number of satellites + epoch of minimum variance
dear colleagues,
here are a number of ideas and proposals related to current DORIS/EOP
analysis. Some of them could be useful in the preparation of becoming
a new IAG Service in Sapporro.
1) number of DORIS satellites actually used in computation
during the recent IDS Analysis Workshop in Paris, I was surprised to
see that all groups when showing DORIS accuracy for EOP did not
distinguish between solutions using 2, 3 or more DORIS satellites.
When doing point positioning solutions, the precision and the
accuracy of the results can be easily linked with the number of DORIS
satellites used in the estimation. In my opinion, the EOP parameters
precision should also improve with this parameter : A 5-6 satellite
solution should provide far better results than a TOPEX-only solution.
In the past submission of EOP, I have put the statistics of the
number of measurements per satellites in the individual sinex files.
Unfortunately, nobody seems to use it. There is nobody to blame as I
must admit that it is not very easy to look though all these
individual sinex solutions to derive the number of DORIS satellites
actually used in the computation.
I have then prepared for you a synthetic file called mjd_sat.txt
containing the following basic information :
First column is mjd. The second column gives you the number of
satellites. The 3rd and eventually other columns give the names of
the satellites (for more extensive studies), using alphabetical order.
48990.5 2 SPOT2 TOP
48991.5 2 SPOT2 TOP
48992.5 2 SPOT2 TOP
48993.5 2 SPOT2 TOP
48994.5 2 SPOT2 TOP
48995.5 2 SPOT2 TOP
48996.5 1 TOP
48997.5 2 SPOT2 TOP
48998.5 2 SPOT2 TOP
48999.5 2 SPOT2 TOP
.....
52574.5 4 ENVIS1 SPOT4 SPOT5 TOP
52575.5 4 ENVIS1 SPOT4 SPOT5 TOP
52576.5 4 ENVIS1 SPOT4 SPOT5 TOP
52577.5 4 ENVIS1 SPOT4 SPOT5 TOP
52578.5 4 ENVIS1 SPOT4 SPOT5 TOP
52579.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52580.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52581.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52582.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52583.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52584.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52585.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52586.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
52587.5 5 ENVIS1 SPOT2 SPOT4 SPOT5 TOP
You can get it at IGN by anonymous ftp :
ftp lareg.ensg.ign.fr
cd incoming
get mjd_sat.txt
I would be very interested if one group could provide DORIS/EOP
statistics by number of satellites (1,2,3,4 and 5) and not as global
result of all types of observations.
If you find out that such an approach makes a difference in the
statistics, it would then make sense to define a format for such a
file and ask all DORIS Analysis groups to provide this information in
liaison with their EOP series.
2) epoch of comparison of EOP
On the other hand, I would like to raise another technical point
concerning the availability of estimated velocities for the polar
motion in the EOP sinex file. In my opinion, the comparisons of EOP
should not be done at the Modified Julian Days given in the EOP file
but a pre-processing should be done at this step in order to define
the epoch of minimum variance of X and Y polar motion.
In the case of DORIS, like SLR and VLBI, the observations are not
continuous all around the day (by opposition to GPS). For example, if
only the second half of the daily data are available (between 0.5 and
1.0), it is best to compare the polar motion parameters at 0.75.
Comparing at 0.5 will give worse result and will not be a good
estimator of the DORIS accuracy.
The proper mathematical way to do that is to look for the epoch of
minimum variance (minimum of a very simple quadratic form) that will
determine the mean observation value. By doing so, you ensure that
the uncertainty in the polar motion velocity does not affect the
accuracy of the polar motion parameter. It is exactly the equivalent
of comparing the stations coordinates at a given reference epoch
(like 1997.0) instead of looking for the mean observed period (using
the minimum variance) that could be far away in 2001. When doing the
comparison when DORIS observations are available in 2001 (just an
example), results are better because they are not extrapolated.
If EOP analysis groups do not use such a strategy, it then makes
sense for me to do an internal post-processing to provide EOP at
non-fixed epoch or to fixed certain EOP parameters (like polar motion
velocities and UT1-UTC related parameters)
Would such a proposed preprocessing look easy enough to implement or
should I submit different types of solutions as proposed above
(irregular epochs or estimating less parameters)?
3) outliers
I would be also very interested to have access to the list of
outliers that you found in your different analysis. That would help
me a lot in understanding problems in the current DORIS data
processing and it should also help me submit more reliable solutions
in the future. My first guess is they are related to single-satellite
solutions but I may be wrong and there may be some more.
4) different DORIS results from Analysis Centers
It was also very surprising for me to see in the view-graphs
presented by Daniel Gambis the different types of results of DORIS
determination of polar motion. Some seemed to be noisier but more
stable on the very long-term while others seem very precise on
short-term but but noisier on longer-term showing some very
systematic errors changing with time.
I wonder if someone should not spend some time in understanding the
different strategies used by all DORIS groups in order to be able to
define better analysis strategies for the future. It might be
presently one of the limiting factor of the accuracy of DORIS/EOP
solutions.
Looking forward for your comments and suggestions
Pascal