[gdal-dev] Bug reading a byn file for Canadian geoid model?

[Greg Troxel]

Javier Jimenez Shaw via gdal-dev <gdal-dev at lists.osgeo.org> writes:

> I am trying to read this file, in theory based in ITRF
> https://webapp.csrs-scrs.nrcan-rncan.gc.ca/geod/process/download-helper.php?file_id=GSD95
> (there is another one based in NAD83xxx)
>
> However in with gdalinfo I see this:
> "Warning 1: CRS EPSG:4140 is deprecated. Its non-deprecated replacement
> EPSG:4617 will be used instead. To use the original CRS, set the
> OSR_USE_NON_DEPRECATED configuration option to NO."
>
> EPSG:4140 https://spatialreference.org/ref/epsg/4140/ is NAD83(CSRS98),
> that is not ITRF, as far as I know.

It is definitely not ITRF.  Canada and the US have both had multiple
realizations of NAD83 after jointly adopting NAD83(1986).  CSRS98 is the
second Canadian realization after the original, and I think is also
called CSRSv2

https://natural-resources.canada.ca/science-data/science-research/geomatics/geodetic-reference-systems/spatial-reference-system-updates

It is defined as a transform from ITRF96, but that's not equal to, but
surely an euler pole rotation.

My guess is that this is just the process of having a datum, then a
revision, then adopting the view that there will be many revisions, and
viewing it as a system, which was perhaps not entirely foreseen from the
start, and then moving from ad hoc names to a vN scheme that emphasizes
and communicates the minor revision nature.

As for deprecation:

https://spatialreference.org/ref/epsg/4140/
https://spatialreference.org/ref/epsg/4617/
https://spatialreference.org/ref/epsg/4955/

this looks like a right mess of ensemble thinking, from taking the CSRS
definitions -- which are surely geodetically rigorous -- and forcing
them into EPSG.

4955 is not showing as an ensemble or as a system in EPSG.

It seems Canada relabeled the base NAD83(CSRS) as a system and then has
realizations of that system.

It is a problem in the NAD83 world that really there is an ensemble, at
last in practice, but the formal language doesn't talk about it that
way, and doesn't really emphasize the system aspect.  It's not so much a
problem in the NAD83 world as much as a problem in talking about the
NAD83 world in the EPSG language.

> However in byndataset.h there is this line:
>
> constexpr int BYN_DATUM_0 = 4140;  /* ITRF2008 (GRS80 based WGS84) */
>
> https://github.com/OSGeo/gdal/blob/master/frmts/raw/byndataset.h#L168

That line is very odd, but I wonder if the comment is just off, and the
"deprecated/replacing" is straight from EPSG and not from that file.
What if you change that to 994140 - do you still get the warning?

> Are we parsing the header wrongly?

It does not seem odd to me that what is likely a geoid model from/for
1995 is expressed in CSRS98.