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

[Schmitz, Uwe]

Jan, 

> I forward this mail from Hermann Peifer, about the 
> differences between the official German grid shift file, and 
> the one computed by gdal utilities based on 31 test points. 
> It's a matter of different interpolation methods. Can anyone 
> comment on this pretended accuracy of the official grid shift 
> files, and about the interpolation methods used, in Germany 
> or elsewhere?
> 
I was involved in the process of creating the
BeTA2007 grid as a member of the (AdV-)working group.

So let me try to shed some light on this.

The resulting BeTA2007 grid is (as Hermann mentioned)
*not* created using one standalone interpolation/transformation 
method.
It's an amalgamation of the most suitable sub-grids
used by the individual federal states (Bundeslaender).
Each subgrid has been calculated by using the method
which was preferred by the indiviual state.
The subgrids were then "composed" to a country 
wide grid, using a special algorithm to average
the differences at state boundaries.

Not enough, there have been additional constraints to
preserve. So it's not unusual that a grid, computed
based on some arbitrary chosen test data, has nothing
to do with something obtained through the above
outlined process.

The test data is only meant for verifying 
implementations and to ensure, that an unknown
software is doing a good job ;-)

Hope that helps a little bit in understanding
BeTA2007.

Uwe