[PROJ] Newbie cs2cs question

[Even Rouault]

Hi,

1) In the genral case, there is no transitivity when doing A -> B, and A -> C and then C -> B. You 
will generally not get the same result, if A, B and C are not in the same datum.

e.g:
GDA94 -> WGS 84 : null transformation (they were coincident in Jan 1st 1994)
WGS 84 -> GDA2020: null transformation (they were coincident in Jan 1st 2020)

but
GDA94 -> GDA2020: ~ 1.8m shift !

2) However, in your case, I'm not completely clear why you don't get transitivity in using 
EPSG:32756 since it uses the same datum as EPSG:4326... This seems related to using the 
deprecated +init=epsg:XXXX syntax instead of the EPSG:XXXX one (if you the EPSG:XXXX 
one, you'll need to specify the geographic coordinates in latitude, longitude order).

3) The result you get with "cs2cs -d 6 +init=epsg:4326 +to +init=epsg:7856+5711" is the 
optimal one I believe. Setting PROJ_DEBUG=3 as environment variable, I can see "
Using coordinate operation axis order change (geographic3D horizontal) + Inverse of 
GDA2020 to WGS 84 (2) + GDA2020 to AHD height (1) + Map Grid of Australia zone 56", which 
is the most reasonable option to me, using AUSGeoid2020 for the vertical adjustment.

4) But if you used the non deprecated syntax, "cs2cs -d 6 EPSG:4326 EPSG:7856+5711" you'd 
get a suboptimal result involving AGD66, GDA94 and AUSGeoid09 internal steps. I've traced 
that issue in
https://github.com/OSGeo/PROJ/issues/2348

5) If you are interested in centimer-level accuracy, and have a coordinate epoch associate 
with your WGS 84 coordinate (since WGS 84 is a dynamic datum), you'd better using EPSG:
9057 / "WGS 84 (G1762)" to get a non-null transformation between WGS 84 and GDA2020

e.g:

echo -33 151 5 2020.6 | cs2cs -d 8 "WGS 84 (G1762)" GDA2020

returns

-33.00000030	150.99999988 5.00013120 2020.6


Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20200905/94848c5e/attachment.html>