[IGSREPORT-2451] Some issues raised by ITRF94
Geosciences Laboratory, NOAA
Jim
Thu Feb 29 08:34:00 PST 1996
******************************************************************************
IGS Electronic Report Thu Feb 29 8:34:00 PST 1996 Message Number 2451
******************************************************************************
Author: Jim Ray (Geosciences Laboratory, NOAA)
Subject: Some issues raised by ITRF94
To: VLBI data modellers and IGS users of ITRF94
(distribution list at bottom)
With the recent release of ITRF94, Claude Boucher, Zuheir Altamimi, and their
colleagues at IGN are to be congratulated on a superb job. This will be an
important resource for a wide range of reference frame applications.
In looking through some of the ITRF94 results, I have noticed a few items
that should probably be addressed by the VLBI data analysts (mostly) but
are relevant to IGS users of ITRF94. (I am not competent to discuss the
SLR or DORIS results, so those are excluded here.)
[1] Tromso "problem"
_______________________________________________________________________________
It has been recognized for some time that the VLBI results give an anomalous
velocity for this site, with a residual East motion of ~12 mm/yr compared to
NUVEL-1. The site was occupied by mobile VLBI only twice, in 1989 and 1992.
The first occupation was seriously compromised by a variety of problems,
including the famous reindeer "encounter", so that only 1 of the 4 sessions
was geometrically robust. The 4 1992 sessions all seem reliable. So, the
velocity estimate really has no redundancy. Any problem with the 31 July 89
session would be completely absorbed into the velocity adjustment. There have
been repeated searches for possible problems with the mobile VLBI system
eccentricity but nothing suspicious has ever been uncovered.
The ITRF94 residuals for Tromso are shown below (epoch 1993.0). It now seems
clear that the VLBI coordinates are skewed in the East direction, almost
certainly due to the suspect velocity. Fortunately, the GPS data are now
strong enough that the TROM coordinates are probably not much affected in the
combination.
I strongly recommend that the Tromso data from the 1989 occupation be deleted
in future VLBI solutions used for such combinations. Meanwhile, the IGS
should consider replacing Tromso as a frame-tie site in their work. Onsala
is a much, much better choice with net residuals of 3 mm horizontal and 4 mm
vertical in the ITRF94 combination of 2 GPS, 5 VLBI, 1 combination, and 3
local tie solutions.
10302 TROMSO
DOMES CDP RX RY RZ NX NY NZ RE RN RU
NUM. NUM. * m m m m m m
---- ---- ----------- ------ ------ ------ ---- ---- ---- ------ ------ ------
M002 7602 1 RG GSFC .002 -.025 -.001 .1 -3.0 .0 -.024 .006 -.003
M002 7602 1 RN NOAA .008 -.022 .015 .3 -1.4 .3 -.023 .005 .014
M002 7602 1 RO USNO .029 -.014 .079 1.2 -1.5 1.2 -.023 .007 .082
M002 7602 1 GA .000 .003 .000 .0 2.5 -.3 .000 .000 .000
M003 TROM 2 PB CODE -.004 .002 -.007 -.8 .3 -1.4 .003 .001 -.008
M003 TROM 2 PE EMR .003 .001 .060 .2 .1 1.2 .000 .018 .057
M003 TROM 2 PJ JPL -.001 .009 .001 -.3 2.2 .1 .009 -.001 .001
M003 TROM 2 GA .000 -.020 .003 -.1 -6.8 1.0 -.015 .005 .013
-------------------- -------------- ------------- ------
.023 2.0 .013 .034
[2] VLBI antenna deformations
_______________________________________________________________________________
Large VLBI antennas deform under the influence of gravity. Among the various
effects is elevation-dependent sagging of the antenna structure. For an
az-el mount, the signature of this effect is indistinguishable from a height
offset which will cause the estimated antenna height to be displaced from
the physical reference point. The effect scales strongly with antenna size.
A 13-mm height offset was documented by Carter et al. [1980] for the 37-m
Haystack antenna and a number of JPL reports have described similar effects.
The NOAA submission for ITRF94 did not model any antenna deformations, and I
suspect that the same is true for GSFC and USNO. JPL does have such models
but I do not know whether they were used for the ITRF94 submissions.
The ITRF94 residuals are shown below for Algonquin. It can be seen that
all the VLBI solutions for the 46-m antenna are biased in their heights
while the GPS solutions are biased in the opposite sense, with a systematic
height difference of roughly 2 cm. A similar discrepancy has been apparent
in the weekly IGS TRF combinations. This could be caused by a local tie
error, as well as by antenna deformation.
40104 ALGONQUIN (7282 = 46-m dish)
DOMES CDP RX RY RZ NX NY NZ RE RN RU
NUM. NUM. * m m m m m m
---- ---- ----------- ------ ------ ------ ---- ---- ---- ------ ------ ------
M002 ALGO 1 PB CODE .015 -.011 .011 1.5 -1.8 1.2 .012 -.002 .018
M002 ALGO 1 PE EMR .003 -.023 .055 .1 -1.8 1.1 -.002 .022 .055
M002 ALGO 1 PJ JPL .000 -.002 .001 .1 -.4 .3 .000 .000 .002
M002 ALGO 1 GA -.001 .001 -.001 -.5 .9 -1.0 .000 .000 .000
S001 7282 3 RG GSFC -.003 .007 -.008 -.5 1.4 -1.4 -.001 .000 -.010
S001 7282 3 RN NOAA -.003 .007 -.004 -.3 .6 -.3 -.001 .003 -.008
S001 7282 3 RO USNO -.002 .012 -.012 -.4 2.8 -2.5 .000 .000 -.017
S001 7282 3 GA .004 -.007 .011 1.2 -2.4 3.6 .008 .011 .000
-------------------- -------------- ------------- ------
.012 1.5 .006 .019
The ITRF94 residuals for the Westford site do not show any obvious indication
of a 13-mm height error but this is a small effect compared with the data
scatter and I do not know whether the offset might have been included in
one of the local site ties.
40440 WESTFORD (7205 = Haystack 37-m dish; 7209 = Westford 18-m dish)
S002 7205 2 RG GSFC -.001 -.006 .005 -.2 -1.0 .8 -.003 .001 .007
S002 7205 2 RN NOAA -.002 .005 .000 -.3 .4 .0 -.001 .004 -.004
S002 7205 2 RO USNO .000 -.004 .004 .0 -1.0 .9 -.002 .000 .006
S002 7205 2 GA .001 .002 .000 .5 .9 -.2 -.002 .000 .000
S003 7209 3 RG GSFC .000 -.002 .002 -.1 -.6 .5 -.001 .000 .003
S003 7209 3 RN NOAA -.002 .002 .002 -.2 .2 .2 -.001 .003 -.001
S003 7209 3 RO USNO -.001 .003 -.002 -.3 1.3 -.8 -.001 .000 -.004
S003 7209 3 GA -.005 -.007 .002 -1.6 -1.9 .4 -.002 .007 .005
S003 7209 3 GB .000 .000 .000 .0 .0 .0 .000 .000 .000
S020 WES2 4 PB CODE .012 -.005 .002 1.2 -.8 .3 .010 -.004 .008
S020 WES2 4 PJ JPL -.004 .015 -.018 -.6 1.3 -1.8 .001 -.003 -.024
S020 WES2 4 GB -.022 .053 -.054 -.2 .5 -.5 .048 -.063 -.001
-------------------- -------------- ------------- ------
.017 .9 .015 .021
If antenna deformations are important, they could show up at all 3 DSN sites.
However, the ITRF94 residuals for these sites are complex and not easily
interpreted. Nevertheless, it is perhaps noteworthy that the JPL solution
gives systematically different height residuals for the largest antennas.
These sites deserve close study to determine the effects, if any, on the
combined GPS coordinates since these are all used for the IGS frame tie.
In any event, the IGS should consider removing Algonquin from the set of
frame tie sites, whether the 2-cm height difference is due to antenna
deformation, local tie errors, or other causes.
[3] Constraints on VLBI velocity estimates
_______________________________________________________________________________
The NOAA VLBI solution adjusted the coordinates and velocities of all sites
freely, regardless of the data spans. This is possible because the adjustment
was done at the mean epoch for each site. Upon exporting the results, a
common reference epoch is used for all sites but those with short observing
histories have been projected using NUVEL velocities rather than estimated
values. These can be distinguished in the export files by having zero
velocity errors and zeros in the associated parts of the covariance matrix.
In the ITRF94 combination, sites with fixed velocity values have been used
only under restricted conditions. All other sites have values freely
determined by the VLBI data only. Those sites which have long enough data
spans for geophysically interesting horizontal rates sometimes give poor
vertical rates due to system limitations. This is particularly true for
most of the mobile VLBI sites in the western U.S. However, the estimated
vertical rate uncertainties are also large in these cases.
The GSFC VLBI solution did not fix any site velocities but did apply a
weighted constraint for sites with short observing histories. (It is not
clear whether the constraint is uniform in all dimensions are sometimes
heavier in the vertical.) The USNO procedures are not clear but it does
not appear from the ITRF94 residuals that heavy constraints have been
applied.
The result of the different methodologies is quite evident in the pattern of
ITRF94 vertical residuals for the two sites below.
40105 PENTICTON
M001 7283 1 RG GSFC -.015 -.018 .022 -.5 -.4 .4 -.004 -.003 .032
M001 7283 1 RN NOAA .038 .079 -.103 .4 .6 -.5 -.006 -.001 -.135
M001 7283 1 RO USNO .027 .051 -.074 .7 .8 -.9 -.002 -.004 -.093
40410 POINT REYES
M001 7251 1 RG GSFC -.003 .000 .010 -.2 .0 .6 -.002 .007 .008
M001 7251 1 RN NOAA .014 .028 -.015 .3 .4 -.2 -.003 .007 -.034
M001 7251 1 RO USNO .053 .075 -.078 1.1 1.1 -1.2 .003 -.005 -.120
In addition, there are a number of sites where only the GSFC constrained
velocity solution satisfied the IGN criteria (because velocity fixing was
not apparent) and therefore totally determines the ITRF94 value.
While I can not propose how best to address the problem of sites with brief
observing histories, it is very clear that combining these inconsistent
solutions is inappropriate. In principle, the SINEX formalism allows
analysis constraints to be removed prior to the combination (as is done
routinely in the IGS weekly combinations), but I do not believe that was
done in this case. The result is a serious under-estimate of the true
coordinate errors and over-weighting of the GSFC solution relative to the
others. I strongly recommend that the sites in ITRF94 affected in this way
be specially designated somehow so that users will be aware.
[4] Biased coordinate residuals
_______________________________________________________________________________
There are several sites in ITRF94 with only a single technique represented
but where the coordinate residuals are biased in one dimension. For example,
the Maspalomas residuals below show a definite East component bias. I do
not understand how this is possible.
31303 MASPALOMAS
DOMES CDP RX RY RZ NX NY NZ RE RN RU
NUM. NUM. * m m m m m m
---- ---- ----------- ------ ------ ------ ---- ---- ---- ------ ------ ------
M001 MASP 1 PB CODE .001 -.013 .001 .1 -1.3 .1 -.012 -.001 .004
M001 MASP 1 PJ JPL -.006 .001 -.002 -.7 .2 -.4 -.001 .001 -.006
M001 MASP 1 CU DUT -.005 -.006 -.003 -.4 -.5 -.2 -.007 -.001 -.004
M002 MASP 2 PB CODE -.004 -.008 -.002 -.4 -.4 -.2 -.009 -.001 -.002
-------------------- -------------- ------------- ------
.006 .6 .006 .004
Concluding remarks
_______________________________________________________________________________
The new ITRF94 is rich in detail and will require considerable study. In
addition to the issues I have noted above, there is a real need to identify
local site ties that may be inaccurate. So, the suggestion that ITRF94 be
maintained as a reference solution for more than 1 year seems reasonable.
However, it is clear that substantial improvements should be possible in the
near future. I would recommend, therefore, that no new realization be
issued for 1995 but that ITRF96 be considered for late 1997.
Regards,
--Jim
distribution:
IERS - Z Altamimi, C Boucher, M Feissel
VLBI - B Archinal, J Bosworth, W Cannon, T Clark, M Eubanks, C Jacobs, C Ma,
D McCarthy, A Niell, V Nelson, A Nothnagel, J Popelar, S Rekkedal,
G Resch, A Rogers, J Ryan, O Sovers
IGS - IGS Reports, ACs
others - D Argus, W Carter, R Eanes, K Larson, B Schutz
[Mailed From: Jim Ray (NOAA 301-713-2850) <jimr at ray.grdl.noaa.gov>]
More information about the IGSREPORT
mailing list