IERS Message No. 12

[central_bureau at iers.org]

***************************************************************************
IERS Message No. 12                                         August 17, 2001
***************************************************************************


IERS Bulletin A changes to AAM assimilation

Predicted values of atmospheric angular momentum (AAM) were introduced
into the IERS Bulletin A combination more than a year ago in an effort
to improve our near-term UT1-UTC predictions.  However, late last year
we observed that the AAM-based values sometimes degraded the UT1
predictions as much as they helped.  We suspended the use of AAM data at
the beginning of 2001 to re-evaluate our entire AAM process.  Starting
with the 14 August 2001 issue of IERS Bulletin A (Rapid Service/Prediction
of Earth orientation) -- Vol. XIV, No. 65 -- the AAM data have been
restored to Bulletin A.  The AAM results continue to come from the
National Centers for Environmental Prediction (NCEP) at
<ftp://ftp.ncep.noaa.gov/pub/cpc/long/aam/nmc/>.

After the additional studies, our approach for assimilating AAM data has
been changed.  Differences include use of both pressure and wind AAM terms,
use of the previously unavailable AAM analysis data at the 24-hour interval
for the latest AAM analysis file, smoothing of the AAM-based LOD before
integrating to UT1R, and removal of a best-fit sinusoid from the AAM
UT1R time series.  The analysis data value at the 24th hour was previously
unavailable and allows for a better merging of the analysis and forecast
periods into a single time series.  Then, the resulting 15-day AAM-based
LODR series (10 days of the most recent analysis + 5 days of forecast) are
smoothed to remove variability with frequencies greater than the Bulletin's
2-day Nyquist frequency.

The AAM-derived LODR is then integrated and the differences between a
UT1 combination of geodetic-only data and the AAM-based UT1R values for
the 10-day AAM analysis period are computed.  A best-fit linear trend
and sinusoid are fit to this difference time series.  These fits are then
weighted and removed from the entire AAM UT1R time series (analysis and
forecast).  The resulting rectified AAM-derived UT1R time series is then
included in the Bulletin A UT1 combination together with the geodetic data.
This procedure reduces the effects of systematic trends (both periodic and
linear) that do not appear to be present in actual UT1 variability.

Our studies indicate that this processing of this AAM-derived UT1R time
series may reduce the UT1 prediction error by ~50% at 5 days into the
future.  The table below shows the UT1 RMS errors for our daily Bulletin A
solutions (without AAM-based UT1R) and for the new AAM-based UT1R series
for a 160-day period earlier this year; in both cases our "final" series
is used retrospectively as truth.

  Assimilation    ------- RMS UT1 Error versus Prediction Interval -------
  Test Method     0 Day     1 Day     2 Days    3 Days    4 Days    5 Days
  wrt finals.data  (us)      (us)      (us)      (us)      (us)      (us)
 -------------------------------------------------------------------------
 BullA (no AAM)   136.8     248.3     405.8     600.5     827.7     1084.3
 AAM UT1R          97.2     150.4     220.4     294.4     381.4      483.1
 -------------------------------------------------------------------------
 [for the 160 days from 06 February through 15 July 2001]

This work has been done by Tom Johnson (tj at CasA.usno.navy.mil) and Brian
Luzum (bjl at maia.usno.navy.mil).

Sincerely yours,
Jim Ray


IERS Bulletin A Web site: http://maia.usno.navy.mil/


***************************************************************************
IERS Messages are edited and ditributed by the IERS Central Bureau.
To subscribe or unsubscribe, please write to <central_bureau at iers.org>.
Archives: http://www.iers.org/iers/publications/messages/
***************************************************************************