[Proj] towgs84 - problem with datum rotation parameters

[Paul Kelly]

On Fri, 6 Aug 2004, VAN DAELE, Toon wrote:

> Hi all,
>
> For the transformation of WGS84 to BD72 (belgian datum), I'm using
> towgs84 with the 7 parameter transformation. The published parameters
> are (national geographical institute, NGI):
>
> dX=-99.059, dY=53.322, dZ=-112.486, rX=-.419, rY=.83, rZ=-1.885,
> ds=0.99999
>
> With these parameters the result are about 70 meters away from the
> locations I got with the transformation tool from the NGI. After a few
> hours of trying I found out the calculation is perfect (differences less
> than a centimeter) when only the sings of the rotation parameters are
> changed, thus:
>
> dX=-99.059, dY=53.322, dZ=-112.486, rX=.419, rY=-.83, rZ=1.885,
> ds=0.99999
>
> I've checked this with several widely spread control points, and it's
> perfect. Is there an explanation why the rotation parameters could be
> defined in the opposite direction? Have other people experienced the
> same problem with the 7 parameter transformation?

Yes the two different rotation direction conventions is a common and 
well-documented problem. The CRS website is a good reliable source:
http://crs.bkg.bund.de/crseu/crs/descr/eu-countrysel.php?country=BE
gives the parameters as in your second example except that the sign of the 
scale (the last parameter) is reversed, i.e. -1. Does this make any 
difference?

The parameters used in GRASS
http://freegis.org/cgi-bin/viewcvs.cgi/*checkout*/grass/src/libes/gis/datumtransform.table
for the bel72 datum are the same as those on the CRS website, but if you 
can confirm that it still works with the scale factor sign change then I 
will update the parameters in the GRASS datum list to the more precise 
values you have given.

Paul Kelly