[GRASS-user] reprojection from/to gnomonic projection two years later

[Paul Kelly]

On Mon, 18 Jun 2007, Milan Salek wrote:

> Hello,
>
> I am referring to the reprojection problem of GRASS6.x described here:
> http://www.intevation.de/rt/webrt?display=History&serial_num=3630
>
> To sum up, the GRASS version 5.3+ (i.e., the current version 6.2, too; I'v not
> tested CVS version) incorrectly reprojects the maps between coordinate system
> using gnomonic projection (using only sphere as a reference body) and any other
> coordinate system with projection using ellipsoid. There is obvious erroneous
> north-south shift of approx 0.19 degree (at 50 deg latitude), very likely due to
> the different reference body (and mayby because of  the mix of of
> geodetic/geometric latitude - ?).
>
> The system proj (http://proj.maptools.org/) works OK, there is not a problem.
>
> The old [vr].proj in GRASS 5.0.3 is reprojecting perfectly, without any shift,
> obviously handling the reference body correctly. I think that there could be

The difference is, GRASS < 5.3 uses the PROJ.4 functions pj_fwd() and 
pj_inv() internally, while GRASS >= 5.3 uses pj_transform() internally. So 
to get to the bottom of the difference you need to investigate why 
pj_transform() handles things differently. I know pj_transform() does some 
kind of ellipsoid conversion that I don't quite understand.

BUT, as I said in the bug report it is not valid to use the PROJ.4 
Gnomonic projection on an ellipsoid; it is only valid for a sphere:
"According to the output of the "proj -lP" command:
gnom : Gnomonic
         Azi, Sph.
gnomonic projection can only be used with a spherical earth model. So you
are going to get undefined behaviour / errors if re-projecting between an
ellipsoid and a sphere."

The fact that you get the results you want with GRASS < 5.3 does NOT 
necessarily mean that is "working properly". It could be two bugs 
cancelling each other out or some such. The issue is beyond GRASS. If you 
really want to get to the bottom of it you're going to have to take it up 
on the PROJ mailing list. It would be helpful before doing that, if you 
could re-produce the behaviour with a point on the command line using the 
proj command (which uses pj_fwd()/pj_inv(), and should give you what you 
want) and the cs2cs command (which uses pj_transform()) and should 
"hopefully" give the "wrong" results.

Paul

> some possibility to fix the problem if this old version works OK. Maybe the
> problem could have been connected with the introduction of the datum
> transformation function in grass5.3 (http://grass.itc.it/devel/roadmap.php).
>
> I would like to ask if there is a possibility to fix the shortcoming.
> As a workaround, I still have to maintain the old version of GRASS 5.0.3 at one
> computer just for the reprojections using gnomonic projection. For some
> reprojection tasks I have available external programs, but the correction of the
> above mentioned error would be very helpful. If the correction of the error (or
> feature?) :o) is not planned in the future, let me know, too, please.
>
> BTW, is anywhere a repository of old GRASS 5.0.3 for cygwin? I'v found only
> GRASS 5.4.
>
> Thanks in advance for any response.
>
> Milan Salek
>
> -- 
>    ___________
>    \   o o   /    Milan Salek
>     \   ^   /     Regionalni predpovedni pracoviste
>      \  ~  /      Regional Forecasting Office Brno
>      ( ( ) )      Czech Hydrometeorological Institute
>    ( (  .  ) )    Kroftova 43, 616 67 Brno, Czech Republic
>  ( (    .   ) )
> (____________) )  tel.: +420-541 421 072
>      ////         fax : +420-541 421 018
>     ////          E-mail: salek at chmi.cz
>                   cell phone: +420-724 185 618
> http://www.chmi.cz
>
> Ceska meteor. spolecnost / Czech Meteorological Society:
> http://www.chmi.cz/poboc/BR/metspol/metspol.html
> ------------------------------------------------------------
>
> _______________________________________________
> grassuser mailing list
> grassuser at grass.itc.it
> http://grass.itc.it/mailman/listinfo/grassuser
>
>