[Proj] *SOLVED*: Creating a nadgrid shift file with GDAL

[Andre Joost]

Am 21.01.2013 15:12, schrieb Jan Hartmann:

> 5) create two point files, one for the lons and one for the lats. The
> coordinates should be the wgs84 values, and the z-value the lon or lat.

I think you made a mistake there, or not stated it clearly.
According to the Beta2007 specification, the grid is related to the 
*source* CRS ellipsoid, that is EPSG:4289 Amersfoort latlon for 
EPSG:28991 and 28992, and EPSG:4314 DHDN for German Gauss-Krüger CRS.

The shift values are from bessel ellipsoid (or Airy, Clarke or whatever) 
to the WGS84 ellipsoid. For reprojection, an iterative solution is 
mentioned.

As far as the accuracy is concerned, ntv2 is by design accurate than 
7-parm Helmert. Simple reason is that Helmert takes the height into 
account. The height vecors of two ellipsoids are not in parallel, so 
another height gives you another lat/lon coordinate in the target 
system. ntv2 can not care about that. As approximation, the natural 
surface height is taken when shift values are made of Helmert 
transformations. And think of continental tectonic moves...

Greetings,
André Joost