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

Schmitz, Uwe uwe.schmitz at bezreg-koeln.nrw.de
Tue Jan 22 07:55:42 PST 2013


Jan, 

> That's clear, thanks for the information.
> 
> I have two questions:
> 
> - Is there an error factor available for each data point, 
> i.e. how precise are the grid shifts really?

considering the circumstances while creating the grid,
we decided against filling the "accuracy" columns
of ntv2 format.
So how precise is the grid? Well, the grid is documented
as being sub-meter accuracy, but this is much understated.
In 90% of the targeted area precision is as good as
the "best known" transformation method of each
individual state (difference <5cm). Areas around
state borders differ more, dependent on how
good the transformations of the involved neighbours 
match.

> - Are the original measurements available, to try out 
> different interpolations?
> 

At the time of creating BeTA2007 we are in luck that
another project had just collected all the data
needed for calcualting the grid shifts based on
the state transformation methods. To waste no
time we ask them for calculating the partial grids 
for us.
These grids are archived an can be brought back
if necessary.

> Just curious for the German situation, I'm going to ask the 
> same question at the Dutch topographic service. There is no 
> ntv2 file available for our country, and I would like to know 
> what happens if I try to create one with this method, unknow 
> to them :-)
> 

Here in Germany we did it, because the "usual" 
Helmert-Transformation results in errors around 2m 
at the far ends. That is to much if you like
to combine cadastral (buildings) with topographic
(roads) data. The ntv2 solution results in much 
better accuracy and can be tuned to match the 
individual solutions of the federal states.

Uwe




More information about the Proj mailing list