[IGSMAIL-6356] Advice for coping with the change to IGS08+igs08.atx

Jake Griffiths Jake.Griffiths at noaa.gov
Mon Mar 7 06:24:21 PST 2011

As announced in IGS Mails #6354 and #6355, the IGS will switch from the (IGS05 + 
igs05.atx) framework to (IGS08 + igs08.atx) in the coming weeks. Many users have 
accumulated results expressed in the old framework, either based on the regular 
IGS operational products since November 2006 or in concert with the IGS 1st 
reprocessing campaign (acc.igs.org/reprocess.html). This note provides advice to 
those users on transforming their old results to be consistent with the new 
frame, to high accuracy.

The most direct method to establish consistency with the (IGS08 + igs08.atx) 
system is to reprocess all old data in the new framework. The IGS will do that 
over the next few years, in its 2nd reprocessing campaign. But many users may 
not find this option to be timely or cost-efficient for the near term. An 
alternative strategy is outlined below that does not require data reprocessing.

Jake Griffiths, Jim Ray, Paul Rebischung, Ralf Schmid

We assume that a user has a long time series of data reduced using igs05.atx 
antenna calibrations and that the results have been stacked into a long-term 
linear frame aligned to the IGS05 reference frame (RF), or possibly some other 
well-defined RF. To generate an equivalent frame solution aligned to the IGS08 
RF and consistent with igs08.atx antenna calibrations involves two main 
steps--application of position corrections to account for antenna calibration 
changes and a Helmert alignment to IGS08.SNX.


A simple, direct Helmert transformation from the user igs05.atx-calibrated frame 
to a set of the included IGS08 stations is *not* recommended. For one, the IGS08 
positions have been derived from ITRF2008 values by directly applying 
station-dependent corrections to account for the impact of changing from 
igs05.atx antenna calibrations (which were used to realize IGS05 and ITRF2008) 
to igs08.atx. Equivalent position corrections are needed for the user frame 
before attempting any Helmert alignment.

1.1 Station-dependent corrections

For the IGS08 RF stations contained in the user frame, the same corrections as 
used to create IGS08 can be applied, which are available at:


For each antenna type with an antenna calibration update, sets of station 
positions were estimated using the Bernese GPS Software 5.0. A static PPP 
processing strategy was used with 24h of data every 30 days during the span for 
which the particular antenna (plus radome) was deployed at the station. For each 
span, one set of positions was estimated using the igs05.atx calibration of the 
antenna, the other using the igs08.atx calibration. Then, the two sets of 
positions were differenced (igs08 minus igs05) and the weighted-average taken 
over the differences to provide the corrections found in the above table.

These station-dependent corrections are best to use. However, since additional 
user corrections may also be needed for the non-IGS08 stations to ensure overall 
self-consistency, the procedure in the next section can be applied for non-IGS08 
stations, and, if more convenient, also for IGS08 stations.

1.2 Latitude-dependent correction models by antenna type

The station-dependent corrections above have been calculated for each individual 
station separately, whereas antenna type correction models are fits over 
latitude. Here, for each antenna type with an updated calibration, two sets of 
station positions were estimated using the Bernese PPP engine and 24h of data 
for 82 well-distributed global stations. For each updated calibration, it was 
assumed that all 82 stations were equipped with the antenna of interest, which, 
of course, is not true. However, both sets of positions were then differenced, 
minimizing common errors, and latitude-dependent functions fitted to the global 
ensemble of East, North and Up position differences.

Discrepancies (de, dn, du) derived by subtracting the latitude-modeled from the 
directly calculated position corrections have been computed. Most de and dn 
discrepancies are smaller than 0.5 mm in magnitude, and most in du are smaller 
than 1 mm in magnitude. Though there are several that are larger (units below 
are mm):

  CONZ  A    2  TPSCR3_GGD      CONE    0.52   0.17  -0.75
  CONZ  A    3  TPSCR3_GGD      CONE    0.52   0.17  -0.75
  JAB1  A    6  LEIAT504        NONE   -0.65  -0.04  -1.73
  EISL  A    2  ASH701945C_M    NONE   -0.76   0.21  -1.57
  NKLG  A    1  TRM29659.00     NONE    0.03  -0.02   1.09
  KERG  A    3  ASH701945E_M    SNOW   -0.03  -0.25  -1.04
  PIMO  A    4  ASH701945C_M    NONE    0.28   0.16  -1.53
  EISL  A    2  ASH701945C_M    NONE   -0.76   0.21  -1.57
  JAB1  A    6  LEIAT504        NONE   -0.65  -0.04  -1.73
  HOFN  A    4  TPSCR3_GGD      CONE    0.19   0.01  -2.65
  SCOR  A    1  ASH701941.B     SCIS   -0.16  -0.38  -3.38

The overall average ENU discrepancies and standard deviations are:

  avg. de = -0.02 (+/- 0.19) mm
  avg. dn = -0.02 (+/- 0.13) mm
  avg. du = -0.13 (+/- 0.67) mm

Thus, on aggregate, the deficiencies of the modeled corrections are negligible 
compared to the sigmas in IGS08. This is confirmed by rigorously comparing the 
IGS08.SNX file and a pseudo SINEX file derived from applying the modeled 
corrections (see Table 1).

Table 1. Helmert parameters from a comparison of IGS08.snx and a pseudo IGS08. 
The parameters were derived using the 91 primary IGS08 core sites. Note that the 
sigmas were *not* rescaled to force chi2=1, though, they were reduced by a 
factor of sqrt(2).

             TX(mm)  TY(mm)  TZ(mm)  S(ppb)  RX(mas)  RY(mas)  RZ(mas)
   offsets    0.00    0.00    0.00   0.000    0.000    0.000    0.000
       ±      0.06    0.06    0.06   0.010    0.002    0.002    0.002

   rates      0.00    0.00    0.00   0.000    0.000    0.000    0.000
       ±      0.00    0.00    0.00   0.001    0.000    0.000    0.000

Although the comparisons made for IGS08.SNX are promising, actual changes due to 
the igs05.atx->igs08.atx calibration updates may differ from those obtained by 
the latitude-dependent models, especially for non-IGS08 stations, for reasons 
such as:
  •  the models were obtained from simulations using data from stations not
     actually equipped with the antennas of interest and only 24h of data
  •  the precise impact of antenna calibration changes on station positions is a
     function of latitude *and* perhaps other parameters, such as station
     environment (observation masks and multipath) and elevation cut-off angle

1.3 Utilities

A few Perl utilities, designed for applying position corrections with some level 
of modularity in mind, are available at:



  snx2ant.pl      - builds antenna information table from igs05.atx-calibrated
  comp_corr.pl    - evaluates latitude-dependent models for each
                    antenna(+radome) in antenna information table to obtain
                    position corrections
  cmp2disc.pl     - matches position corrections to solution numbers given in a
                    table of discontinuities for the user's igs05.atx-calibrated
  apply_shifts.pl - applies position corrections to solutions in igs05.atx-
                    calibrated SINEX
  date_conv.pm    - Perl module called by the above scripts

Running the scripts without command-line arguments will provide usage 
information and other information about the input files.

The scripts for computing, matching and applying the modeled position 
corrections assume that three files are provided by the user. Those files are:
  •  a table of antenna information for sites in the igs05.atx-calibrated
     solution in a particular format (see usage info in comp_corr.pl)--if
     the SITE/ID and SITE/ANTENNA blocks in the SINEX file of the stacked
     frame are accurate, then snx2ant.pl can be used to create the ant.info table
  •  a list of sites and solution numbers appearing in the stacked frame, also in
     a particular format (see usage info in cmp2disc.pl)
  •  a definitive, SINEX-formatted discontinuity table (named soln.snx below)
     containing a record for each site in the igs05.atx-calibrated SINEX
     file--if a site has no discontinuities, we recommend inserting a record
     that corresponds to the solution number and data span

A fourth file containing a table of coefficients for the latitude-dependent 
models is also required. As stated in IGS Mail #6354, such a table is available 
containing models for each antenna type with *.atx calibrations. The file can be 
found in two places:


All antenna types appearing in both igs05.atx and igs08.atx have 
latitude-dependent models in the lat_models.txt table. The models for 
antenna/radome combinations with unchanged calibration values have zeros for 
their coefficients, so a zero position correction is applied when the model is 
evaluated. In the special cases where the user’s stacked frame solution contains 
sites with uncalibrated equipment, no correction is available to make its 
position consistent with IGS05 or IGS08. Users should take care in how such 
stations are used in order to avoid introducing frame inconsistencies in 
downstream processing.

The sequence of Korn shell commands using the above scripts and tables could 
look something like:

  snx2ant.pl igs08.atx igs05-atx.SNX > ant.info 2> log

  comp_corr.pl ant.info lat_models.txt > ant.info_corr 2>> log

  cmp2disc.pl pts.dat ant.info_corr soln.snx > shifts.dat 2>> log

  apply_shifts.pl shifts.dat igs05-atx.SNX > igs08-atx.SNX 2>> log

where the igs05-atx.SNX and igs08-atx.SNX frames are consistent with usage of 
igs05.atx and igs08.atx antenna calibrations, respectively. The user should also 
note that the scripts provide diagnostic information to the log file that we 
hope can be helpful in computing, matching and applying position corrections to 
the data intervals in the user’s regional solution. Of course, one can imagine 
the case where a discontinuity due to an antenna change was omitted because no 
position jump was detected. Such cases leave the user to decide whether 
additional discontinuities are to be inserted, and the matching process be 
repeated, or to simply omit that particular station/solution from the final 
SINEX file.

It is important to note that the SOLUTION/EPOCHS block from the user's SINEX 
file is not used to match latitude-dependent models to SINEX solution numbers. 
By definition, the SOLUTION/EPOCHS block contains information about the span of 
data used in the estimation of the position and velocity. The span of data may 
not coincide with the discontinuity interval. Thus, the definitive beginning and 
ending epochs used for matching are assumed to be in the user's soln.snx 
discontinuity file.


Once station-dependent corrections are computed and applied, then a Helmert
alignment to IGS08 should be performed. However, one must take care to match 
discontinuity intervals between sites in IGS08 and the user’s solution.

Note that within the entire IGS contribution to ITRF2008 of about 558 station 
time series, which span up to about 12.5 yr, there is an overall average of 0.9 
position discontinuities per station. The continuous span of data without any 
break averages about 8.7 yr with a standard deviation of 3.2 yr. Consequently, 
it is only practical to realize long-term GNSS RFs these days by including 
position discontinuities, which is the case with the IGS08 RF. Aligning a user 
frame to IGS08 requires that the same set of discontinuities applied in IGS08 be 
used in the user frame for the stations used in the alignment. The set of IGS08 
discontinuities is available at:


We recommend that, once all station-specific corrections have been applied to 
the user frame coordinates, the solution be aligned to IGS08. The set of sites 
used for the Helmert alignment should be based on the IGS08 core network (e.g., 
ftp://igs-rf.ensg.eu/pub/IGS08/IGS08_core.txt), while also taking care to match 
IGS08 discontinuities. Of course, the user can adopt their own set of 
discontinuities, but it may be most convenient to use the same as those used by 
the IGS for any common IGS08 stations.

More information about the IGSMail mailing list