[Proj] Incorrect translations to the Gauss-Boaga system (resend)

[Gino Lucrezi]

A few days ago I sent the following message to this mailing list, but there were no subscribers yet (except for Frank and me) so I am sending it again.

--------8<----cut-here----8<------
I am currently using version 4.4.8 of the proj4 libraries. I mostly use them from within PostGIS, but I get the same results calling them on a stand-alone basis.

One widely used coordinate system here in Italy is Gauss-Boaga,
corresponding to EPSG codes 26591 and 26592.

So, I often have to convert between Gauss Boaga and UTM 33 (epsg:23033)

Here are the results for two known points:
cs2cs +init=epsg:26592 +to +init=epsg:23033 -v <<EOF
2413000.0000 4669000.0
2389238.48 4690935.27
EOF

# ---- From Coordinate System ----
#Transverse Mercator
#       Cyl, Sph&Ell
# +init=epsg:26592 +proj=tmerc +lat_0=0 +lon_0=2.54766666666667 
+k=0.999600
# +x_0=2520000 +y_0=0 +ellps=intl +pm=rome +units=m +no_defs
# ---- To Coordinate System ----
#Universal Transverse Mercator (UTM)
#       Cyl, Sph
#       zone= south
# +init=epsg:23033 +proj=utm +zone=33 +ellps=intl +units=m +no_defs
393000.00       4669000.00 0.00
369238.48       4690935.27 0.00

The first point I could check on a map, and it should have been
393070 4669190 I used a good ruler, and I am sure I got this point with an error smaller than 10 metres.

The second point was converted using another (interactive) program, which is usually regarded as authoritative, and it resulted in 369301 4691118

In either case, the resulting point is about 200 metres off from the correct result.

I didn't modify the epsg file in any way.

Did I do anything wrong?

Gino Lucrezi
Penta Consulting Services Srl
--------8<----cut-here----8<------

Frank replied (among other things):

> It seems that there is no datum shifting information available for either
> coordinate system.  The Guass Boaga is using the Monte Mario datum while the
> UTM coordinate system is using ED50.

> I *think* the ED50 is relatively close to WGS84, but you may need to find
> some approximation for the datum shift needed for ED50.  You might even be
> able to derive a reasonable approximation from your observed errors.

So, if I understand correctly, I need one more parameter like +towgs84= for both systems.

Does anyone have correct values for it?
Should it go in the epsg file?


Gino Lucrezi
Penta Consulting Services Srl