[IGSREPORT-3105] updated LODR comparison results
NOAA Geosciences Laboratory
Jim
Wed Oct 30 08:58:24 PST 1996
******************************************************************************
IGS Electronic Report Wed Oct 30 8:58:24 PST 1996 Message Number 3105
******************************************************************************
Author: Jim Ray (NOAA Geosciences Laboratory)
Subject: updated LODR comparison results
To: IGS ACs, IERS/CB, USNO, IGS Reports
Summary
-------
This is an update of my previous comparisons of LODR results from the IGS
ACs relative to inferred LODR from VLBI. The methodology is the same as
in J.R. Ray, JGR, 101(B9), 20141-20149, 1996, with the addition of 11 months
of more recent data. Part of the motivation is to assess analysis changes
made by most ACs starting 30 Jun 96.
During the most recent few months, the LODR results of the various ACs have
become much more similar to one another than during earlier periods. This
convergence appears to be explained by improvements in the results of ACs
with previously poorer performance accompanied by increased scatter in the
LODR results of the formerly best performers. Over the 27-month span from
03 Jul 94 to 05 Oct 96, the weighted rms LODR differences compared to VLBI
range from 36.7 to 86.3 usec/day. However, over the most recent 52 days
up to 05 Oct 96, the wrms differences fall between 50.8 and 69.0 usec/day.
It is likely that some of the recent increased scatter is due to larger
VLBI errors, but this is not likely to be the dominant cause.
LODR comparison results
-----------------------
To characterize the variation in LODR determinations from the IGS ACs as a
function of time, weighted mean and weighted rms differences, relative to
VLBI, are tabulated below over 5 time spans ending on 05 Oct 96. The top
line of each pair of entries contains the mean LODR differences while the
second line of each pair gives the wrms differences. The column headed
"VLBI ERR" gives the mean and rms values of the estimated uncertainty in
the VLBI-inferred LODR time series.
----- wtd mean/wtd rms LODR differences wrt VLBI ----
----- (all units usec/day) ----
most
recent EMR VLBI
# days COD EMR UT1 ESA* GFZ JPL NGS SIO ERR
--------------------------------------------------------------------
826 -22.8 -12.5 5.0 -38.6 -26.9 -10.6 -63.3 32.8
39.0 61.6 36.7 65.8 58.4 59.7 86.3 10.4
413 -22.9 -1.7 4.2 -12.3 -21.9 2.9 5.7 -5.0 33.7
39.7 57.9 38.2 53.6 55.2 56.4 55.6 50.1 11.4
207 -28.8 -19.6 0.8 -19.4 -31.6 5.3 -7.5 -8.0 35.7
44.3 60.8 44.6 55.5 54.4 70.9 54.4 57.2 12.1
103 -21.8 -6.7 1.1 -7.8 -18.2 31.0 -12.4 -4.3 35.0
50.0 56.8 45.7 54.1 47.2 76.9 53.0 58.5 11.1
52 -22.4 -7.4 0.3 -23.5 -31.3 -0.6 -34.2 -19.0 35.6
57.8 60.1 53.2 58.8 50.8 69.0 52.8 63.5 11.8
--------------------------------------------------------------------
Note: Considers IGS LOD values from MJD 49536.5 (03 Jul 94), in the
longest instance, to MJD 50361.5 (05 Oct 96).
* After deletion of LOD value from ESA for MJD 50313.5; including
that day would give a wtd. mean difference of -41.2 usec/day and
a wrms of 161.4 usec/day for the most recent 52 days.
Results for each AC are discussed individually below. The following
postscript plots are available at ftp://ray.grdl.noaa.gov/pub/
lodr.ps 24463 bytes -- LODR values & errors from VLBI
lodr.ps.Z 8403 bytes -- " compressed
dlod1.ps 97443 bytes -- delta LODR for COD, EMR, EMR(UT1), ESA
dlod1.ps.Z 29655 bytes -- " compressed
dlod2.ps 87733 bytes -- delta LODR for GFZ, JPL, NGS, SIO
dlod2.ps.Z 27279 bytes -- " compressed
VLBI
----
The n9609 series of UT1 and LOD values determined from VLBI data by USNO
were used here. The VLBI analysis procedures were changed since the series
n9504 used for the previous comparisons. The primary differences are: use
of improved CALC models; estimation of polar motion rates; estimation of
atmospheric gradients. The latter 2 changes apply only to the 24-hr sessions
and cause an increase in the UT1 and LOD formal errors. Over the 489-day
previous study period the average LODR error increased by 7.8%. Comparing
LODR values inferred from the n9609 and n9504 time series, the differences
have a mean value of -1.5 usec/day and a wtd. rms of 29.2 usec/day. As
seen in the table above (and the ps plot), the VLBI-based LODR errors
increase somewhat in more recent data, presumably due to changes in the
selection of radio sources available for the brief, quasi-daily UT1 sessions.
However, this increase does not appear large enough to explain the observed
increase in wrms LODR differences for COD and EMR(UT1).
COD
---
The average LODR bias for COD has remained very stable over the 5 periods
tabulated above, about -23 usec/day. However, the wrms scatter compared to
VLBI has increase steadily in the most recent half year or so. Inspection
of the ps plot indicates that the wrms increase is caused by larger day-to-
day scatter as well as larger, faster drifts in the LODR bias beginning
apparently around late Nov 95 but becoming more prominent after the changes
made on 30 Jun 96. To some extent, this behavior was expected based on
recent orbit model changes made at CODE (reported by Tim Springer).
EMR
---
The EMR(UT1) results have shifted slightly in the mean and are now basically
unbiased relative to VLBI. However, there also seems to be a distinct
increase in the day-to-day LODR scatter relative to VLBI starting around late
Feb 96. Considering that the VLBI-based LODR errors also increase about that
same time (see ps plots), the increased LODR scatter could possibly be due to
increased VLBI errors rather than to changes at EMR although the mean and
rms VLBI errors do not seem large enough to permit this.
The directly estimated LODR values from EMR no longer show the large,
fortnightly modulation previously apparent, starting with the analysis
changes made on 30 Jun 96. It was established by Gerd Gendt, Jan Kouba,
and me that the LODR periodicities (also apparent in the GFZ results) were
caused by neglect of diurnal and semi-diurnal UT1 variations and have been
eliminated by including the IERS-recommended model in recent data analyses.
ESA
---
The ESA results have been comparatively stable since about mid 95, although
drifts in LODR bias, sometime large, are apparent (see ps plot). A faulty
LOD value was reported for MJD 50313.5 (18 Aug 96) which was omitted from the
tabulated results above because it otherwise dominates the recent statistics.
GFZ
---
As noted above, the incorporation of a model for diurnal/semi-diurnal UT1
variations on 30 Jun 96 has eliminated the previous large, fortnightly
periodicities in LODR. While it is clear that this change has improved the
LODR scatter, a longer data span is required to fully characterize the new
GFZ performance.
JPL
---
The JPL data can be broken into 3 spans. The early data, until mid-Oct 94,
showed a steady drift in LODR bias. The middle span, until late Feb 96,
has relatively small scatter with occasional excursions in bias. In the
most recent data, the LOD bias has drifted markedly and the day-to-day
scatter apprears to have increased, sometimes dramatically.
NGS
---
NGS, which did not report LOD results before 06 Aug 95, shows similar
performance throughout, with distinct drifts in LODR bias together with
moderate day-to-day scatter.
SIO
---
The SIO results improved significantly beginning around Jul 95. Since then,
the day-to-day scatter has often been quite small, although drifts in the
LODR bias are apparent. In the most recent data, the scatter appears to
have increased although, as mentioned before, this could possibly be
caused by an increase in VLBI error. SIO has several data gaps including
several week-long periods, with a total of 31 days having no LOD values.
[Mailed From: Jim Ray (NOAA 301-713-2850) <jimr at ray.grdl.noaa.gov>]
More information about the IGSREPORT
mailing list