[IGSMAIL-6475] use of IGS orbits for homogeneous long-term processing

[Jim.Ray]

Author: Jim Ray

Because this question has come from several users, here
is some advice for homogeneous long-term processing of
GPS data using IGS repro1 and operational orbits together.

Some confusion arises because the repro1 products for
weeks 0730-1459 and the operational products for weeks
1460-1631 were based on the IGS05 terrestrial reference
frame plus igs05.atx antenna calibrations (see IGSMAIL
6445).  Meanwhile, since week 1632 the IGS08 terrestrial
frame plus igs08.atx calibrations have been used for all
operational products.

Because the satellite orbit positions in the SP3 product
files are determined from the dynamical physics together
with the rotational orientation of the TRF and because the
IGS05 and IGS08 frames share a common ITRF rotational
datum (see IGSMAIL 6354), the IGS orbits for the entire
period from week 0730 till now are highly self-consistent.
The only material impact of the antenna calibration change
arises from certain station-specific changes, which should
be very accurately compensated by the corresponding position
adjustments applied to create IGS08 (see IGSMAIL 6354).  In
fact, parallel solution tests by the ACs before the week
1632 framework change verified the very close correspondence
of IGS05 and IGS08 orbits, with typical differences within a
few mm (see IGSMAIL 6384).

So a long-term processing using these orbit products in a
differential mode should yield station position time series
that are highly homogeneous *provided* that users rely only
on the IGS08/igs08.atx framework for the entire period.
(The IGS05/igs05.atx framework is obsolete and is no longer
recommended.)

What cannot be used consistently over the full IGS history
are the satellite clocks, for PPP for example.  The clock
offsets are directly impacted by changes in the atx satellite
PCOs so a consistent long-term PPP solution is not possible
in any simple way.  In principle it might be feasible to
"correct" the SV clocks for changes in PCOs for each SV to
approximate consistent long-term clocks, but I know of no
tests of such a method.