[IGSMAIL-8026] Preliminary IGS repro3 SINEX solutions available

Wed Feb 10 09:48:45 UTC 2021

Dear colleagues,

As today was the original IERS deadline for solution submissions to ITRF2020,
a first set of daily and weekly combined terrestrial frame solutions from the
3rd IGS reprocessing campaign (repro3) has been made available as preliminary
IGS contribution to ITRF2020. The final IGS repro3 terrestrial frame solutions
will be released by the extended IERS deadline (10 April 2021). It is expected
that by then, all participating Analysis Centers (ACs) will have completed their
reprocessing, and some minor issues identified in current AC contributions will
be resolved.

For now, the preliminary IGS repro3 terrestrial frame solutions are combinations
of the following AC contributions:
* COD (Center for Orbit Determination in Europe): 1994-01-02 to 2019-12-31
   - GLONASS included starting from 2002-01-01
   - Galileo included starting from 2013-01-01
* ESA (European Space Agency):                    1995-01-01 to 2020-12-31
   - GLONASS included starting from 2009-01-01
   - Galileo included starting from 2015-01-01
* GFZ (GeoForschungsZentrum):                     1994-01-02 to 2020-12-31
   - GLONASS included starting from 2012-01-01
   - Galileo included starting from 2013-12-21
* GRG (Groupe de Recherche en Géodésie Spatiale): 2000-05-03 to 2020-12-31
   - GLONASS included starting from 2008-11-04
   - Galileo included starting from 2016-12-31
* JPL (Jet Propulsion Laboratory):                1994-01-02 to 2019-12-28
   - GPS-only
* MIT (Massachusetts Institute of Technology):    2007-01-07 to 2019-12-28
   - Galileo included starting from 2017-01-01
* NGS (National Geodetic Survey):                 1994-01-02 to 2020-12-31
   - GPS-only
* TUG (Graz University of Technology):            1994-01-02 to 2020-12-31
   - GLONASS included starting from 2009-01-01
   - Galileo included starting from 2013-01-01
* ULR (Université de la Rochelle):                2008-01-01 to 2020-12-31
   - GPS-only
* WHU (Wuhan University):                         2008-01-01 to 2019-12-31
   - GLONASS included starting from 2010-09-28

Details about the available products, the modeling updates since the repro2
campaign and the combination strategy can be found below. Please feel free to
contact igs-rf at ign.fr for any further information.

With deep thanks to the contributing Analysis Centers for their huge efforts and
high-quality products,

Paul Rebischung

AVAILABLE PRODUCTS
------------------

The preliminary IGS repro3 SINEX products follow the new IGS long file name
convention. They are prefixed "IGS0R03SNX", while the final IGS repro3 SINEX
products will be prefixed "IGS1R03SNX". The available products include:
* daily combined SINEX solutions:   IGS0R03SNX_${yyyy}${doy}0000_01D_01D_SOL.SNX
* --- without covariance matrices:  IGS0R03SNX_${yyyy}${doy}0000_01D_01D_CRD.SNX
* daily combination residuals:      IGS0R03SNX_${yyyy}${doy}0000_01D_01D_RES.RES
* --- in YAML format:               IGS0R03SNX_${yyyy}${doy}0000_01D_01D_RES.YML
* weekly combined SINEX solutions:  IGS0R03SNX_${yyyy}${doy}0000_07D_07D_SOL.SNX
* --- without covariance matrices:  IGS0R03SNX_${yyyy}${doy}0000_07D_07D_CRD.SNX
* weekly ERP files:                 IGS0R03SNX_${yyyy}${doy}0000_07D_01D_ERP.ERP
* weekly combination summaries:     IGS0R03SNX_${yyyy}${doy}0000_07D_07D_SUM.SUM
* --- in YAML format:               IGS0R03SNX_${yyyy}${doy}0000_07D_07D_SUM.YML
where ${yyyy} stands for the 4-character year and ${doy} for the 3-character
day-of-year (first day of the week in case of weekly files).

Note that the opportunity of repro3 was taken to change the style of the
traditional SINEX combination residual and summary files and to make them
additionally available in machine-readable YAML format.

These products are available in weekly sub-directories at:
ftp://igs-rf.ign.fr/pub/repro3

They are currently being uploaded to CDDIS, where they will soon be available
at: https://cddis.nasa.gov/archive/gnss/products/repro3

They will eventually be available at other IGS global data centers as well.

MAIN MODELING UPDATES SINCE REPRO2
----------------------------------

Compared to the previous IGS reprocessing campaign and to current IGS
operational products, a number of new models and conventions have been adopted
in repro3. The main updates are detailed below together with their (expected)
impact on terrestrial frame solutions.

* Secular pole: The new IERS secular pole model was adopted in repro3. It
replaces the previous mean pole model of the original IERS 2010 Conventions,
which was used in repro2 and is still used in the IGS operational products. The
repro3 terrestrial frame solutions thus show a degree-2, order-1 deformation
pattern with respect to repro2/operational solutions, but also ITRF2014/IGS14,
whose amplitude is relatively small until ~2010, but then increases linearly. It
reaches ~3 mm in vertical and ~1 mm in horizontal in 2020.

* Satellite z-PCOs: Compared to igs14.atx, the radial phase center offsets
(z-PCOs) of all GPS and GLONASS satellites have been updated in the ANTEX file
used in repro3, igsR3.atx. While the GPS and GLONASS satellite z-PCOs in
igs14.atx had been estimated based on the ITRF2014 terrestrial scale, they were
corrected in igsR3.atx to become consistent with the Galileo satellite z-PCOs
measured and published by the European GNSS Agency. Details about these
corrections can be found in the 2019 technical reports of the IGS Antenna and
Reference Frame working groups (https://doi.org/10.7892/boris.144003) and in
igsR3.atx (latest version: ftp://igs-rf.ign.fr/pub/IGSR3/igsR3_2135.atx). As a
consequence, the terrestrial scale of the IGS repro3 solutions is independent of
ITRF2014, and is instead based on the Galileo satellite antenna calibrations. It
shows an offset of ~1.2 ppb at epoch 2010.0 and a drift of ~0.03 ppb/yr with
respect to the ITRF2014 scale.

* Ground antenna calibrations: In order to enable the processing of Galileo data
in repro3, the L1/L2-only calibrations of several ground antenna types in
igs14.atx were replaced by new multi-GNSS calibrations in igsR3.atx (see details
in the 2019 technical report of the IGS Antenna working group and in the header
of igsR3.atx). As usual, such ground antenna calibration updates induce antenna
type- and station-dependent position offsets of a few mm in horizontal / up to
~1 cm in vertical. Those offsets were estimated for all affected stations that
are part of the IGSR3 reference frame (see below). The results can be found at:
ftp://igs-rf.ign.fr/pub/IGSR3/igs14_to_igsR3_2077.txt

* IGSR3 reference frame: Because of the updates mentioned above, the repro3
solutions could not have been aligned to the IGS14 reference frame without
systematic errors. A particular reference frame called IGSR3, consistent with
the new IERS secular pole model, the igsR3.atx satellite PCOs and the igsR3.atx
ground antenna calibrations, was thus used in repro3. It is available at:
   - ftp://igs-rf.ign.fr/pub/IGSR3/IGSR3_2077.snx
   - ftp://igs-rf.ign.fr/pub/IGSR3/IGSR3_2077.ssc (without covariance matrix)
Details about its computation can be found in the comment section of the SINEX
file. Note that the daily and weekly repro3 solutions are aligned in orientation
and origin to IGSR3, but not in scale. Their scale is a weighted average of
those of the contributing AC solutions, and is thus based on the igsR3.atx
satellite z-PCOs.

* PCO/PCV rotations for antennas not oriented North: Although certain ground
GNSS antennas are not oriented North, the IGS practice until repro3 had been to
ignore these mis-orientations and use the phase center corrections from the IGS
ANTEX files as if all antennas were oriented North. In repro3 however, the phase
center corrections from igsR3.atx have been rotated when necessary to match the
actual antenna orientations reported in the site logs. These PCO/PCV rotations
have impacts of up to a few cm in horizontal / a few mm in vertical on the
estimated positions of the affected stations. Antenna orientations, extracted
from the site logs, are reported in an extra column in the SITE/ANTENNA blocks
of the repro3 SINEX files.

* Inclusion of GLONASS and Galileo: While only two ACs had processed GLONASS
data in repro2, a majority of ACs have processed GLONASS and also Galileo data
in repro3. This may have a positive impact, which remains to be quantified, on
the background noise in the repro3 station position time series. On the other
hand, this leads to additional spurious signals at periods around 8 and 10 days.

* Others:
   - ACs have used "modern" ocean tidal loading models in repro3 (FES2014 or
     equivalent). This can be expected to reduce spurious aliased tidal signals
     in the repro3 station position time series to some extent.
   - ACs have refined their solar radiation pressure models since repro2. This
     may be expected to reduce draconitic signals in the repro3 station position
     time series to some extent.
   - The sub-daily EOP tide model from Desai & Sibois (2016;
     https://doi.org/10.1002/2016JB013125) was adopted in repro3 in replacement
     of the current IERS model. This has been shown to have practically no
     impact on station position estimates, but should reduce spurious aliased
     tidal signals in the repro3 polar motion [rate] time series to some extent
     (Sibois, 2019; http://acc.igs.org/workshop2019.html).
   - Contrary to repro2, all ACs have used time-variable gravity field models in
     repro3. The expected impact on terrestrial frame solutions is sub-
     millimetric (Amiri et al., 2016; http://acc.igs.org/workshop2016.html).

COMBINATION STRATEGY
--------------------

The daily combinations of the AC repro3 SINEX solutions follow essentially the
same strategy as the one used in repro2 and described in Rebischung et al.
(2016; https://doi.org/10.1007/s00190-016-0897-6). Some minor changes are
detailed below.

* Terrestrial scale: In the repro2 daily combinations, no scale factors were
estimated between the AC solutions and the combined solutions. In repro3, the
estimation of such scale factors was restored to study possible subtle
differences between the terrestrial scales realized by the different AC
solutions. The terrestrial scale is in practice handled in a similar way as
geocenter coordinates during the repro3 daily combinations. Scale factors with
respect to the IGSR3 reference frame are first made explicit in the 
unconstrained AC normal equations, by a similar re-parameterization as for
geocenter coordinates. The preprocessed AC normal equations are then inverted
using no-net-rotation, translation or scale (NNR+NNT+NNS) constraints with
respect to IGSR3 (instead of NNR+NNT only in repro2). The AC scale factors are
then combined together with station positions, ERPs and geocenter coordinates,
and scale residuals are in particular provided in the combination residual
files. The combined solution is at this step obtained using NNR+NNT+NNS
constraints with respect to IGSR3 and includes combined geocenter coordinates as
well as a combined scale factor with respect to IGSR3. In a last post-processing
step, the combined normal equation is re-inverted with NNR+NNT constraints only
while fixing the combined scale factor to 0. By doing so, the combined scale
information (based on the igsR3.atx satellite z-PCOs) is brought back into
station coordinates.

* Station metadata errors: Like in repro2, care has been taken to exclude from
the repro3 combinations AC station position estimates for which incorrect
metadata were reported in the AC SINEX files. The policy adopted in repro3 was
to confront metadata in the AC SINEX files with site logs collected from
different sources and systematically reject stations with either:
   - an incorrect antenna type,
   - an eccentricity error larger than 1 mm in either East, North or Up,
   - an antenna orientation error larger than 10°.
This lead in particular to reject all stations with antenna mis-orientations
larger than 10° from the solutions of COD, NGS and WHU, as these ACs did not
apply PCO/PCV rotations and reported 0 or no mis-orientations in their SINEX
files. An exception was made for antenna orientation errors in case of GFZ, as
they applied PCO/PCV rotations but did not report antenna mis-orientations in
their SINEX files. All detected metadata errors are reported in Section 6 of the
combination summary files, while rejected stations are listed in Section 4.

* AC weighting and outliers: The weighting of AC solutions in the combinations
was refined compared to repro2. The same unbiased variance component estimator
is used, but it is now iteratively applied until all AC variance factors have
converged to better than 1e-3 in a relative sense. (This explains why the global
variance factors reported in the daily combined SINEX files are very close to
1.) The criterion for rejecting outliers from the AC solutions was also changed
compared to repro2. Now, any AC station position estimate with a *normalized*
residual larger than 5 in either East, North or Up is considered as an outlier.
The overall iterative repro3 combination procedure can be described as follows:
   - Assign a priori AC variance factors in the same way as in repro2.
   - Perform an initial combination and re-weight AC solutions using unbiased
     variance component estimates. No outliers are rejected yet.
   - Iteratively combine AC solutions. After each iteration, reject outliers and
     re-weight AC solutions again.
   - Iterate until no outlier remains and AC variance factors have converged.

* Weekly combined solutions: The weekly combined repro3 solutions are obtained
by iteratively stacking the 7 daily combined solutions of the week. The daily
solutions are given a uniform a priori variance factor of 1, and a single common
corrective variance factor is estimated. After each iteration, station position 
estimates with normalized residuals larger than 5 are removed from the
corresponding daily solutions. Iterations are performed until no such outlier
remains.