From tah at chandler.mit.edu Wed Jul 2 23:14:18 2003
From: tah at chandler.mit.edu (Tom Herring)
Date: Thu, 3 Jul 2003 02:14:18 -0400 (EDT)
Subject: [IGSREPORT-10387] Wk 1223 MIT T2 Analysis Report
Message-ID: <200307030614.CAA25131@mtglas.mit.edu>
******************************************************************************
IGS Electronic Report 02 Jul 23:15:10 PDT 2003 Message Number 10387
******************************************************************************
Author: Thomas Herring
********************************************************************************
MIT ANALYSIS REPORT GPS-WEEK 1223 Dates: 03jun15 to 03jun21
********************************************************************************
MIT (Massachusetts Institute of Technology)
Room 54-618
77 Massachusetts Avenue,
Cambridge, MA 02139
e-mail: tah at mit.edu
tel: (617)253-5941 (Tom Herring)
fax: (617)253-1699
PRODUCTS GENERATED:
The following products are generated and uploaded into CDDIS @ NASA/GSFC
MIT1223g.SNX Combined sinex file from all available analysis centers
MIT1223g.SUM Name of this summary file
MIT1223g.RES File of the individual AC position estimates residuals
to the combined solution for the week.
0.0 Starting Week 1203 we changed the method used to compute the variance
scaling for each sinex file. The old method is described in item 2 of
the analysis procedures below. In the new method, we determine the scaling
factor by the ratio of the adjustments to the IGS00 frame from a loose,
rotated and translated solution, to the sigmas determined for the adjustment.
The new method is robust when some of the reference sites have changed
position (because of outlier detection during the rotation/translation
step) and is not senstive to any aproiri rescaling factors used in the
analysis.
0.0 Starting Week 1202 adopted IGS01P37_RS54 for reference frame. Dropped
KOKB and FAIR from frame definition.
0.1 Starting Week 1125, we removed AREQ from the sites used in determining
the variance re-scaling of the center SINEX files. This site
appears to have moved by 0.5 meters at around week 1120.
0.2 Starting Week 1110, we stopped applied a pole-tide correction to
SIO sinex files because this correction is already made in their
analysis.
0.3 Starting Week 1066, we adopted the IGS realization of the ITRF97
(referred to as IGS97). We also replaced our constrained solution
a loose solution with translation (+-1 mm) and orientation (+-0.1 mas)
constraints applied.
0.4 Starting Week 1020, we adopted the ITRF97 reference frame
0.5 Starting Week 1002, we re-ordered the summary to give the residual
summary and the SINEX file errors first.
CENTER SINEX ERRORS: (corrected during processing).
CNT Site Type of Error Used values IGS.SNX values
HISTORY and PROCEDURES
----------------------
0.0 Starting Week 998 we adopted full checking of the center sinex
headers and we correct the solutions for differences between the
headers and a modified version of IGS.SNX. (The modifications are
to put full precision antenna L1/L2 offsets and to correct the WUHN
entry. The station log for this site shows height to BCR not ARP).
A new error reporting block has been added below. We also have
removed some of the routine documention in this summary.
0.1 Starting week 993 we started to add corrections for the pole-tide to
the SIO, NGS, and ESA sinex. According to the analysis descriptions
none of these centers apply the pole tide correction. Starting with
the re-submission of week 992, we corrected the LOD values in JPL sinex
files from regularized LOD (should be called LODR in the SINEX file)
to standard LOD. (Since starting multiday polar motion combinations
we have been correcting the sign error in the JPL LOD values.
0.2 Starting week 989 we include as Section 5.0 of this summary, the
Earth rotation parameter estimates from the constrained combination
solution in the IERS format.
0.3 Starting Week 980 we include multiday polar motion/UT1 estimates into
our analysis. We also no longer need to deconstrain the AC sinex
files although to add back into the covariance matrices at rotational
uncertainity which is coupled between station positions and EOPs.
0.4 Starting Week 0943 we base the analysis on the ITRF96 system and the
reference site list for the ITRF96 system. The increase in the number
of sites used to obtain analysis variances increases the variance
factors for all centers by factors between 2 and 4. The Chi**2 for
the center residuals is now much closer to 1.0
0.5 Starting Week 0931 we added a new residual file to our submissions.
A description of the contents of the file is given in the file itself.
We also changed the treatment of the translation of the individual
AC solutions. The generation of the weighting factors and the constrained
combined solution remains unchanged. However, in the unconstrained
analyses (used for AC comparisons) we now allow explicit translation
and scale parameters in the solution. The net effect is to generate
more realisitic sigmas for the center of mass position estimates and to
make the coordinate estimates immune to real center of mass translations
and scale changes.
0.6 Starting Week 0901 we started using a new apriori file not Non-ITRF94 coordinates
based on the analysis of week 0822-0900 sinex files.
1.0 Analysis Procedures:
(1) Each centers sinex solutions are de-constrained by adding back
in a rotational uncertainty.
(2) Covariance matrices are re-scaled by computing the chi**2/freedom of
the position residuals to the reference sites in IGS00 from an loosely
constrained solution that is rotated and translated on to the IGS00 system.
During the rotation/translation step, outliers (4-sigma based on fit) are
removed and this new scheme is more robust when some of the reference sites
have moved (due earthquakes and/or equipment changes).
(2o) This procedure was replaced Week 1203 with the one described above.
Covariance matrices from each center are re-scaled to generate
unit increment in chi square per degree in the tight constraints solution.
The values are are based in the chi**2 per degree of freedom increment
when the loosely constrained solution is constrained to the ITRF1996
coordinates of the 55 Reference sites. Orientation changes are estimated
in this step, but translations and scale are NOT estimated.
(3) The tight solution is generated with the following constraints on
the core sites. Starting Week 0858, ITRF94 is used for the coordinate
system. Starting Week 0878, we changed the site selection here to use
WTZR and GOL2. Starting Week 0897 we stopped using MADR as a core site.
We continue not use MADR in the reference site list.
Earth rotation parameters are estimated for each center (even when not given
in the original SINEX files. In this case the partial derivatives are
computed from the station coordinates).
(4) When the combined and center solutions are compared to ITRF97, the
heights are given a variance 10 times larger than the horizontal coordinates.
The same variance ratio is used when the RMS to ITRF97 core sites is
computed.
(5) Starting Week 0826, we show the differences in the coordinates for
five non-core sites, in remote areas. Four of these sites are processed
by all centers. These sites are not used in the RMS calculations or
the transformation estimation to the 13 core sites. The change in procedure
was because FORT is no longer being processed by all centers. AREQ has
been added to the list for this reason. After Week 0943, when the reference
site list increased we no long summarize additional sites.
(6) Starting Week 0826, we start to show a comparison of each centers
results with the combined solution for the week. When the constraints
are applied to loose combined solution to translate, rotate and scale it
to the ITRF97, the constraints are actually applied using the full
covariance matrix of the combined solution (rather than simply directly
applying the transformations to the coordinates). The approach we use
constrains the center of mass to have zero variance relative to ITRF97.
The result of this is that the combined solution has the same center
of mass position and scale as ITRF97, and therefore when each of the
centers are compared to the combined solution, they generate very similar
translations and scales as they do the ITRF97. Had we