[IGSMAIL-1439] GNAAC NCL now producing P-Sinex

[Phil]

******************************************************************************
IGS Electronic Mail      Thu Sep 26  9:52:19 PDT 1996      Message Number 1439
******************************************************************************

Author: Phil Davies
Subject: GNAAC NCL now producing P-Sinex

Beginning this week 0872, weekly P-Sinex files are submitted to IGS
Data Centres by Associate Analysis Centre NCL (Newcastle, England),
starting with the week 0867 Polyhedron solution. This is in addition
to the weekly G-Sinex we already deposit. Because the GNAAC
definition is open to various interpretations, I briefly summarise
below the procedure we have initially adopted to create these files.


(i) Stage one - Global network estimation

Weekly A-Sinex files from COD, EMR, ESA, GFZ, JPL and SIO are used.
Significant station constraints are removed where present to give
unbiased A-networks. The remaining orientation constraints of each
are augmented to ensure loose (several m) SDs of orientation. A
variance scaling factor is applied to each A-network. A-networks are
combined by stacking normal equation blocks (assuming no correlation
between A-networks) to estimate the NCL G-network.  No frame
parameters are estimated. Stations estimated by fewer than three ACs
are excluded from the G-network.

A posteriori statistical testing on the G-network includes (a) MINQE
variance component estimation to test the A-network scale factors,
which may be updated after a hypothesis test; (b) three-dimensional
data-snooping on the A-network station observations, primarily to
trap AC antenna height blunders; (c) Overall chi-square test. The
G-network estimation is iterated as necessary following these
procedures.

An ITRF Core constraints block is added to the loose (i.e.
approximately free) G-network to give the constrained G-network which
is written out in a G-Sinex file NCLwwwwG.SNX.  The accompanying SUM
file shows Helmert transformations between A and G networks, residual
sizes, excluded observations, and station information discrepancies.


(ii) Stage two - Polyhedron assembly

Currently R-Sinexes from EUR, GSI and PGC are used (SIR and ASI
having no unique stations). Significant station constraints are
removed where present to give unbiased R-networks. The covariance of
each is augmented to give large SDs of all seven Helmert parameters.
An ad hoc variance scaling factor is applied to each R-network.
Also, an 'extra' R-network is formed from a combination of the
non-Global A-network station estimates using the Core station set as
Anchor stations; any stations which also appear in a 'real' R-network
are deleted from this extra block.

R-networks are adjusted to the G-network by backsubstitution of
G-network coordinates and covariance for the R-network Anchor station
parameters. The Polyhedron is a concatenation of the G-network and the
adjusted R-networks. The full Polyhedron covariance matrix is
computed. The loose and ITRF-constrained Polyhedra are obtained in
this way from the loose and constrained G-networks respectively.
Note that using this approach Global station coordinates and
(co)variances in the Polyhedron are unchanged from their G-network
values, and that Polyhedron coordinates do not depend on R-network
variance scaling. Adjusted R-networks take their reference frame
definition from the G-network Anchor station positions.

The ITRF-constrained Polyhedron estimate and full matrix is written
out in a P-Sinex file NCLwwwwP.SNX. This includes the same a priori
constraints block as the G-Sinex, which can be removed by users if
required to give the loose Polyhedron; this has the same reference
frame definition as the loose G-network. The P-Sinex includes almost
all the stations in the input A and R Sinexes; those also included in
the G-Sinex are considered 'first order' stations. At the moment the
accompanying SUM file does not contain any comparison statistics - I
intend to include these 'gradually' over the next weeks.





[Mailed From: "Philip Davies" <P.B.H.Davies at newcastle.ac.uk>]