[Proj] Is egm08_25.gtx misplaced half a pixel?

[Mikael Rittri]

Hello,

I have tried to use the geoid height file 

   http://download.osgeo.org/proj/vdatum/egm08_25/egm08_25.gtx,

which is discussed here:

   http://trac.osgeo.org/proj/wiki/VerticalDatums

But I am beginning to think that the file is misaligned
(well, at least if ones uses GDAL 1.8 or later to read it).

>From gdalinfo (version 1.8.1), I get

  Origin = (-180.041666666666660,90.041666666666686)
  Pixel Size = (0.041666666666667,-0.041666666666667)

But I think the correct origin of the data is really

           (-180.020833333333333,90.020833333333333)

that is, half a pixel farther east and south. 

I suspect this may be caused by the nasty AREA_OR_POINT issue.
The file is dated 3 Sep 2010, so I guess it was created before
GDAL 1.8, when the semantics of AREA_OR_POINT = POINT changed.

http://trac.osgeo.org/gdal/wiki/rfc33_gtiff_pixelispoint

Or maybe the cause is something else.

Some evidence: I have used the gtx file to look up the geoid 
height (using a Carmenta Engine version that uses GDAL 1.8.1), 
at a position where the geoid slopes towards southeast, 

    145dE 13dN

and I found the value 44.925678 m.  

But according to Charles Karney's online geoid calculator 
http://geographiclib.sourceforge.net/cgi-bin/GeoidEval
the value should be 45.7529 m (with a few millimeters 
uncertainty, perhaps).  So the difference is 0.8272 m. On the 
other hand, if I move the position half a pixel = 1.25 arc 
minutes east and south, that is

    145d1.25'E 12d58.75'N

then Charles's calculator gives me 44.9287 m, so the difference
is now just 3 mm from the gtx result for the original position.

More evidence: gdalinfo on Charles's file egm2008-2_5.pgm 
(from http://geographiclib.sourceforge.net/html/geoid.html) says

Origin = (-0.020833333333333,90.020833333333329)
Pixel Size = (0.041666666666667,-0.041666666666667)

so its origin is, indeed, half a pixel farther east and south
(well, there is a also 180 degree longitude difference due to 
 the longitude range being 0 to 360 in this file). 

I guess I could file a ticket, but I am not quite convinced 
that I am right.  Has anyone else noticed any problems with
the gtx file? 

Another issue with the gtx file: I would have preferred if it had
one more column; that is, if the column for 180dW was repeated 
for 180dE. That would make it simpler to interpolate results for
a position slightly west of 180dE.  

Regards,

Mikael Rittri
Carmenta
Sweden
http://www.carmenta.com