proj4 +towgs84 property in EPSG:2039 gives offset in transformation

[shmueldarchi]

I used your proj4 value for transforming coordinates from WGS84 to EPSG:2039 with OL Transform and Proj4js. The transformation had an offset of about 100 meters.
After researching, I found that the value had a transformation from:
+proj=tmerc +lat_0=31.7343936111111 +lon_0=35.2045169444444 +k=1.0000067 +x_0=219529.584 +y_0=626907.39 +ellps=GRS80 +towgs84=23.772,17.49,17.859,0.3132,1.85274,-1.67299,-5.4262 +units=m +no_defs +type=crs
to:
+proj=tmerc +lat_0=31.7343936111111 +lon_0=35.2045169444444 +k=1.0000067 +x_0=219529.584 +y_0=626907.39 +ellps=GRS80 +towgs84=-24.0024,-17.1032,-17.8444,-0.33009,-1.85269,1.66969,5.4248 +units=m +no_defs +type=crs.
After the change, the transform worked correctly.

NOTE
I didn't see any open source that gives this value by default, so if it's not correct, I would like to know.

BR,

[rouault]

The same issue has just been reported to PROJ at OSGeo/PROJ#4170 regarding EPSG:31370 "Belgian lambert 72". It would be nice to have this corrected, as exporting wrong transformation parameters is causing downstream issues.

[rouault]

@shmueldarchi inverting the sign of the 4th, 5th and 6th parameters of towgs84 also do it:

+proj=tmerc +lat_0=31.7343936111111 +lon_0=35.2045169444444 +k=1.0000067 +x_0=219529.584 +y_0=626907.39 +ellps=GRS80 +towgs84=23.772,17.49,17.859,-0.3132,-1.85274,1.67299,-5.4262 +units=m +no_defs +type=crs

[rouault]

Cf OSGeo/PROJ#4170 (comment) for the analysis of the root cause: the parameters of Helmert transformation using the Coordinate Frame rotation convention are used unmodified in a TOWGS84 clause. But a TOWGS84 clause requires them to use the Position Vector convention, which differs from the Coordinate Frame one by the sign of the rotation terms (4th, 5th and 6th terms)

[klokan]

Thanks for spotting the bug! Thank you, @rouault, for your comments here...

This should be fixed together with OSGeo/PROJ#4170 in the upcoming update of the webpage by @TomPohys

Perhaps we could schedule a call to discuss the defined problems during FOSS4GE?
I was just talking to @hobu on LinkedIn... we are keen to help and reduce pains, and make this project better

[rouault]

This issue (and #190 as well) seems to have been fixed, but looking at https://github.com/maptiler/epsg.io/commits/master/ , I don't see any relevant commit but a README update (does https://github.com/maptiler/epsg.io reflect what runs on the epsg.io website... ?) . @jachym ?