[Proj] UTM32N-GaussBoaga datum shift problems

[Oscar van Vlijmen]

> From: "Giorgio Agugiaro"
> Date: Mon, 8 Jan 2007 15:58:48 +0100
> Subject: [Proj] UTM32N-GaussBoaga datum shift problems
...
> I have some troubles with datum-shift results using cs2cs.
...
> I calculated (and checked the results with other programs) the 7 parameters
> from the following test points:
> WGS84-UTM32N
> Northing    |   Easting | h_ell| Point_id
> -----------------------------------------
> 5022649.211  718871.227  52.537  #1000
...
> Roma40, Zone 1 (GaussBoaga)
> Northing    |    Easting  | h_orto | Point_id
> -------------------------------------------------------
> 5022670.5000  1718900.5000  12.00000 #1000
...
> In order to check the transformation I do:
> cs2cs -rsf %.4f +datum=WGS84 +proj=utm +zone=32 +north +unit=m +to
> +ellps=intl
> +proj=tmerc +lat_0=0.0 +lon_0=9.0 +k=0.9996 +x_0=1500000 +y=0.0 +unit=m
> +towgs84=-927.3934,-426.678,166.8859,-129.10138,-5.11282,-118.30878,65.
> 7674 mg84grid.txt
...
> Output is:
> 5022672.0592    1718901.5497 12.1290
...

Disturbing. I got x=1718900.9018, y=5022671.7300, h=12.0459 and there
shouldn't be any significant difference between my procedures and
proj/cs2cs.
Looking at the "other programs" values of 5022670.5000  1718900.5000
12.00000: I can't reproduce them either with the given information, not even
with the Italian Gauss-Boaga equations from Hirvonen instead of PROJ tmerc.

...
> On a side note, I used successfully cs2cs in a similar datum-shift with
> these test points and parameters:
> 
> Points
> AA   WGS84-UTM32N                 45 24 05.7680 N   11 52 02.5800 E   59.57
> Gauss Boaga Roma40           45 24 03.3760 N   11 52 03.3920 E
...
> calculated 7 parameters were (units as above):
> -135.9406,-28.3596,-10.6462,-0.24989,-2.48329,-1.57131,-15.6918

I got 45d 24m 03.3772s, 11d 52m 03.3812s, 59.5709 m
That differs by 0.238 m on the globe. Disturbing....

Returning to:
> I calculated (and checked the results with other programs) the 7 parameters
> from the following test points:
It is possible that the error in the closest fit of the calculated
transformation parameters is a bit larger than you hoped for.
But that is no explanation for the differences mentioned above.