[PROJ] Differences in coordinate transforms from proj 9.1 to 9.2

[Greg Troxel]

Erixen Cruz via PROJ <proj at lists.osgeo.org> writes:

> I think https://github.com/OSGeo/PROJ/pull/3621. Before that PR, PROJ
> did a shift with a geotiff between GDA2020<->GDA94 as an intermediate
> step for transforming to WGS84. Now, WGS84 to either GDA is null (no
> operation). I remember reading some bulletin published by an
> Australian government or agency that confirms this behaviour is
> "correct"? That geotiff grid shift is intended only for conversion
> between 94 and 2020 and not as a more accurate way to get to WGS84.

WGS84 is an ensemble with 2m of fuzz so there is really no such thing as
accurate, when one means 10cm or sub-meter, transforms to/from or
coordinates in WGS84.

My memory is fuzzy, but I think GDA94 might be close to WGS84(TRANSIT)
and GDA2020 close to a more recent realization of WGS84 or ITRF, as of
its reference epoch.

If you could about accuracy you should stop using WGS84(ensemble) and
use specific realizations or specific ITRF realizations.  However,
geojson and maybe web standards refer to WGS84(ensemble).  Everywhere,
but especially AU, if you care about sub 50 cm, you need to be clear on
station velocities in ITRF, whereas regional datums like GDA2020 and
NAD83(2011) are much closer to plate-fixed.