[PROJ] Some questions about UK CRS

[Even Rouault]

On jeudi 27 février 2020 11:28:45 CET Peter Gagliardi wrote:
> Hi everyone,
> 
> I have a separate question about some CRS in the UK that a client wanted us
> to support.
> 
> 1. For EPSG:7405 OSGB36, it looks like it tries to open
> `OSTN15_OSGM15_GB.txt` and fails. From
> https://www.ordnancesurvey.co.uk/business-government/tools-support/os-net/fo
> r-developers, the developer pack includes a `OSTN15_OSGM15_DataFile.txt`, is
> this related, or are they two different files for two different purposes?

I had looked a bit at that some months ago. OS delivers grids in a specific 
format that would require further processing to be usable by PROJ. So 
typically proj.db has records for the geoid conversion OSTN15_OSGM15_GB.txt, 
but has no mapping to a file it would know how to handle (a .gtx typically)

> 
> 2. I see that EPSG:7952 and EPSG:7953 are not listed in the database.

They are in EPSG, but not imported in proj.db, as they use a METHOD["Ordnance 
Survey National Transformation",ID["EPSG",9633]] that PROJ doesn't know how to 
use. If I remember well, the issue is that the grid is referenced in projected 
coordinates, and not geographic one. Processing would be needed.

> From
> the developer pack, it looks like this isn't a projection, but rather a
> horizontal shift + a vertical shift. Our client provided .gsb and
> .gtx files for us to use, but now... how do I specify both horizontal and
> vertical shift at once? I tried using the following:
> 
> cs2cs +proj=pipeline +step +proj=hgridshift +grids=OSTN15_NTv2.gsb
> +step +proj=vgridshift +grids=OSGM15.gtx +to
> EPSG:4979
> 
> However, I get the error:

Yes, "+proj=pipeline +step +proj=hgridshift +grids=OSTN15_NTv2.gsb +step 
+proj=vgridshift +grids=OSGM15.gtx" expresses a transformation not a CRS, so 
you cannot use that as the source CRS of cs2cs.
You may want to use that pipeline directly in the 'cct' utility
Also be careful about the direction of +proj=vgridshift. By default, it 
subtracts the value found in the grid (that is it goes from ellipsoidal height 
to orthometric height)


-- 
Spatialys - Geospatial professional services
http://www.spatialys.com