[Proj] Call for review of derived NTv2 for Swiss projection CH1903

[Mikael Rittri]

Hello Sandro,
you wrote:

> ETRS (seems to be the same as WGS84?)

I would say that ETRS was the same as WGS84 on January 1, 1989. But ETRS is
attached to the Eurasian tectonic plate. Since 1989, the continental drift
has moved Europe and ETRS toward northeast by about 2.5 cm per year,
compared to WGS84. 

> anyway it'd be interesting to see if your 0.008 mm error gets even lower
> with this.

Actually, it got slightly worse:

Test point	in CH1903		Diff Sandro	Diff Swisstopo
7.630605°E 46.590702°N	0.008 mm		0.124 mm
8.980228°E 47.020130°N	0.004 mm		0.125 mm
9.857142°E 47.571428°N	0.005 mm		0.112 mm

The first difference column gives the difference between the concatenated
shift (using CHENYX06a.gsb followed by the 3-par shift) and the direct
shift using your file CHENYX06a_wgs84.gsb.

The second difference column gives the difference between the concatenated
shift and the direct shift using the Swisstopo file CHENYX06_etrs.gsb.

One reason for the differences may be that I have assumed that the height
above the Bessel ellipsoid is zero for all test points. This height
shouldn't affect the direct shifts, but it might affect the result of the
3-parameter shift between CH1903+ and WGS84.  It might be better to look up
the ground elevation at each test point, and convert it from orthometric
height to height above Bessel ellipsoid (or just hope that the height type
isn't so important for this test). 
    But maybe it is something else that causes the differences.

> We've asked for redistribution permissions to get the file included
> in the proj-data package.

That would be nice.

Best regards,

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

-----Original Message-----
From: Sandro Santilli [mailto:sandro.santilli at gmail.com] On Behalf Of Sandro Santilli
Sent: Monday, February 20, 2012 10:20 AM
To: Mikael Rittri
Cc: PROJ.4 and general Projections Discussions
Subject: Re: [Proj] Call for review of derived NTv2 for Swiss projection CH1903

On Thu, Feb 02, 2012 at 01:55:14PM +0000, Mikael Rittri wrote:
> Sandro wrote:
> 
> > I got a max of 9mm drift.
> >
> > This mail is to ask others to try the NTv2 file with real data and let
> > me know if it's working as expected.
> 
> Looks excellent to me.
> 
> I have used Carmenta Engine to construct two implementations of CH1903.
> One uses a concatenated datum shift, where the original NTv2 file
> transforms to CH1903+, and the Helmert datum shift continues to WGS84.
> The other uses a direct datum shift to WGS84 based on your new NTv2 file.
> 
> Then I selected three arbitrary points expressed in CH1903 LongLat and
> transformed them to WGS84 LongLat with both implementations. The largest 
> difference between the two implementations was 0.008 millimeters. 
> 
> So it seems that your drift of 9 millimeters is caused only by the
> difference between the official FINELTRA algorithm and the NTv2 file.

Thanks a lot for the review Mikael.

Eventually I received an NTv2 file directly converting from CH1903 to
ETRS (seems to be the same as WGS84?) from Swisstopo.

My tests don't show any major difference between using this "official"
grid and the one I made, anyway it'd be interesting to see if your
0.008 mm error gets even lower with this.

Could you try it ? It's attached.

We've asked for redistribution permissions to get the file included
in the proj-data package. Still waiting for a response...

--strk; 

  ,------o-. 
  |   __/  |    Delivering high quality PostGIS 2.0 !
  |  / 2.0 |    http://strk.keybit.net
  `-o------'