******************************************************************************
IGS-ACS Mail      11 Jul 12:18:26 PDT 2008      Message Number 25
******************************************************************************

For IGU & IGR ACs,

To follow-up a point from the AC splinter meeting
in Miami Beach, Yves Mireault has kindly interrupted
his summer holiday to provide some information about
the EOP prediction procedure he uses for his IGU
solutions.  See below.  Basically, he computes EOP
rates-of-change from IERS Bulletin A, then propagates
his own most recent (IGA) values with those rates.

Experience with the IGUs shows that the EMR rotations
tend to be among the least scattered.  Usually, the
Rz rotations are most problematic due to the direct
relationship with predicted UT1, which is naturally
more highly variable than are polar motion changes.

An alternative source of EOP predictions that you can
consider is from the JPL operational series, which is
available at:
  ftp://euler.jpl.nasa.gov/keof/predictions/
I spoke to Richard Gross who explained that these files
are updated daily, for both midnight and noon epochs.
Like Bulletin A, the JPL series assimilates predicted
values of AAM excitation to improve the UT1/LOD prediction
performance.  The JPL series is available for the public,
but primarily generated for deep space navigation.

Concerning the relative UT1 prediction performances,
according to the IERS-sponsored EOP prediction campaign,
which posts results at:

http://www.cbk.waw.pl/EOP_PCC/?menu=results

for "ultra-short" prediction periods (<=10 d) the best
series studied for UT1 came from JPL, with an RMS UT1
error of about 65 us over the 1st day and around
250 us after 5 d.  (The IERS series were not studied.)

On the other hand, in the IERS Annual Report the Paris
Observatory claims to predict UT1-UTC to 180 us over the
1st day and 600 us over 5 days, while comparable values
for the USNO Bulletin A are 147 us over the 1st day and
518 us over 5 days.  These are much larger than the
near-term JPL prediction errors.

So it would seem to be useful for the IGS to also
consider using the JPL operational series as an apriori
source, at least for UT1.  Better EOP predictions are
most important for the real-time IGUs but also have
some impact on even the IGR products.

My thanks to Yves and Richard.
Best regards,
--Jim


-------- Original Message --------
Date: Fri, 11 Jul 2008 14:23:14 -0400
From: Yves Mireault
Organization: NRCan - ESS - GSD
To: Jim Ray

Hi Jim,

The estimated part comes directly from my Bernese estimation.
Just in case you don't know/remember, I use the Bernese software
and not Gypsy for the Ultra Rapid. Regarding the EOP predictions,
I did several tests a few years ago using different staregies. I
tried using my last rate to extrapolate, I tried also different
fitting scenarios (using previous EOP values) but none of those
gave better results than using the Bulletin A predicted rates.
To summarize, I use my own values for the estimated part which I
then extrapolate for the next day using Bulletin A EOP rates.
That's as simple as date! And this way, there's no jumps between
the estimated/predicted EOPs which was the case when I tried
using predicted Bulletin A values (i.e.xp, yp, UT) directly.
Based on my own experience, the Bulletin A rates (i.e. simply the
difference between 2 consecutive days) are quite reliable
nowadays. I just have to watch for times when Bulletin A is not
available for several days which doesn't happen anymore. Well
I'll cross my fingers now that I said that!

And when I thought about it, it made sense to use it since who's
better than USNO to predict EOPs anyway? I certainly couldn't
even get close to what they are achieving without putting a lot
of effort into it and frankly, I did not feel like it. Let me
know if this is not clear and/or not acceptable.

Regards,

Yves Mireault
Geodetic Engineer