[PROJ] Krovak S-JTSK centimeter accuracy

[Erixen Cruz]

Hello,
We want to use Proj to transform from EPSG:4979 to EPSG:5513+5705 (Krovak
S-JTSK + Baltic height). The CZ government has published how to go from
ETRF2000 (which I believe Proj approximates with WGS84) to S-JTSK with
centimeter accuracy here
<https://www.cuzk.cz/Zememerictvi/Geodeticke-zaklady-na-uzemi-CR/GNSS/Nova-realizace-systemu-ETRS89-v-CR.aspx>
 (English
<https://www-cuzk-cz.translate.goog/Zememerictvi/Geodeticke-zaklady-na-uzemi-CR/GNSS/Nova-realizace-systemu-ETRS89-v-CR.aspx?_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp>).
We can observe that Proj does not apply any grid shift files and relies on
a helmert transform and Krovak projection only:
$ OSGeo/PROJ/build/bin/projinfo -s EPSG:4979 -t EPSG:5513+5705 -o PROJ
...
PROJ string:
+proj=pipeline
 +step +proj=axisswap +order=2,1
 +step +proj=unitconvert +xy_in=deg +xy_out=rad
 +step +proj=push +v_3
 +step +proj=cart +ellps=WGS84
 +step +inv +proj=helmert +x=589 +y=76 +z=480
 +step +inv +proj=cart +ellps=bessel
 +step +proj=pop +v_3
 +step +proj=krovak +axis=swu +lat_0=49.5 +lon_0=24.8333333333333
       +alpha=30.2881397527778 +k=0.9999 +x_0=0 +y_0=0 +ellps=bessel

The PDF from the CZ gov't suggests that after the Krovak coordinate
operation, the horizontal grid shift from the provided file be applied in
this projected frame.

First question:
Does Proj support this local coordinate grid shift? Looking through the
documentation, the shift you can get with `xyzgridshift` expects geocentric
coordinates, so it isn't appropriate here I think. `hgridshift` expects
geodetic coordinates (lat/lon), so it isn't appropriate either.

Second question:
The Czech Technical University in Prague has a page about using Proj to
transform to/from S-JTSK here
<http://freegis.fsv.cvut.cz/gwiki/S-JTSK_/_Grid> (English
<https://freegis-fsv-cvut-cz.translate.goog/gwiki/S-JTSK_/_Grid?_x_tr_sch=http&_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp>).
It suggests using grid shift files from some CZ researchers with the old
`nadgrids` and `geoidgrids` paradigm. We can confirm that this leads to
sub-10cm accuracy. Is there anything wrong with going with this method, and
would Proj accept a PR for adding these grid shift files to get the better
accuracy?

Thanks for reading, and thanks for any advice you can offer.
Sincerely,
Erixen Cruz
3D Software Developer, Cesium <https://cesium.com/>
erixen at cesium.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20220316/8a93afb8/attachment.html>