[GRASS-dev] Re: Question concerning g.proj and epsg:2100

[Paul Kelly]

On Sun, 30 Mar 2008, Nikos Alexandris wrote:

[...]
> ( http://wald.intevation.org/tracker/?func=detail&aid=603&group_id=21&atid=204 )
>
> I am struggling to understand why g.proj picks the wrong ellipsoid when
> using epsg:2100 to create a new location in GRASS.

Hi Nikos,
Thanks for reporting this. It looks to be a bug as a result of my change 
to lib/proj/convert.c on 2006-11-05 to add the ellipsoid name into the 
PROJ_INFO files generated from WKT sources. Previously the ellipsoid name 
was missing when a location was generated with g.proj.

The code tries to match the numerical ellipsoid parameters against those 
in GRASS's ellipse.table. As you may know, GRS80 and WGS84 are almost 
identical, close enough for them to make no difference to datum transform 
calculations, for example. Because they are almost the same it is 
obviously very difficult to tell them apart. I thought I'd tested it 
reasonably well, but as you've found it fails in some circumstances. It 
shouldn't affect the accuracy of your calculations though, but is 
obviously less than ideal.

When I originally wrote the ellipsoid comparing code I made a compromise 
between handling the case of a sphere correctly and keeping the best 
precision. I've changed it now so it handles the sphere as a special case 
and it now seems also to work for your example. I'd appreciate some more 
testing though.

This bug will have been present in 6.2.3 but not earlier versions.

Best regards,

Paul