Confusion over Gauss Laborde projection

[warmerdam]

Before releasing PROJ 4.6.1 I'd like to ensure that the implementation and naming of the Gauss Laborde Sphere Geometric Mean (glabsgm) projection is brought into line between PROJ.4 and libproj4.

In response to:

o Added the glabsgm (Gauss Laborde / Sphere Geometric Mean) projection

Gerald writes:

Excuse me for forgetfulness. This is really the Gauss-Schreiber projection that we went over a few months ago. Aka +proj=stmerc

[dgrichard]

As we have discussed by mails with Gerald few weeks ago (Subject: Regarding projection stmerc and Gause-Laborde - june, 23rd 2008 - 22:18), we finally agreed that Gauss Laborde / Sphere Geometric Mean was not a Gauss-Schreiber projection (citation) :

"astly, I will probably change stmerc to glaborde. I think I may have a new lead on the mysterious Schreiber projection. ;-)"

So, if agreed : I can change : - src/PJ_glabsgm.c to PJ_glaborde.c - src/Makefile.in, src/Makefile.am, src/Makefile (or may only one of them is necessary ?) - nad/IGNF

[gie]

Main question is what, exactly, are we referring to? Is it the projection described in NT/G 73, "Projection Cartographique Gause-Laborde?"

If it is, then it produces the same results as the libproj4 procedure +proj=stmerc, AKA Gauss-Schreiber projection. I believed we discussed that Schreiber probably was the first to describe the projection and thus the name. Snyder, in two references describes a Schreiber projection that matches the technique used here.

PS: "Geometric Mean" seems to be a very incorrect description as the process is the conformal conversion of the elliptical system to a spherical earth and then performing a transverse Mercator projection.

[warmerdam]

I shall give you two some more time to work out whether this is or is not the same thing.

Richard - if it is helpful, I don't mind PROJ.4 supporting both names though we really ought to pick one (in line with libproj4) to document and consider standard.

[gie]

To demonstrate stmerc is the same as the Gause-Laborde I offer the following script:

ARGS="+proj=stmerc +ellps=intl +lon_0=0.96923951127r" ARGS2="+x_0=160000 +y_0=2388648.4517" lproj $ARGS $ARGS2 +lat_0=-0.36855536038r <<EOF -f %.4f 0.96865773483r -0.36651914294r EOF

which produces as a result:

156534.1770 62916.9250

This matches example on p. 3/3 of NT/G 73 "Projection Cartographique Gause-Laborde".

Note: 'r' suffixed values denote radian arguments. I cannot remember how far back into proj history this feature extends.

[dgrichard]

After long talk within my organisation, we came to the following conclusion : stmerc is ok if stmerc means simplified transverse mercator (and not schreiber transverse mercator). Other choices would have been sgltmerc for schreiber-gauss-laborde transverse mercator, or glstmerc.

So, I have made changes to switch to stmerc :

$ proj -l=stmerc

   stmerc : Simplified Transverse Mercator (Schreiber, Gauss-Laborde)
        Cyl, Sph&Ell
        lat_0= lon_0= k_0=

I am going to commit those changes to the trunk.

[gie]

"Simplified" is a poor term: simplified from what? "Simple" is no better. I am afraid I do not understand the problem with using the common practice of naming the projection after the earliest developer. From "Map Projections---A Reference Manual" by Bugayeskiy and Snyder, p. 159-160: "In 1820-30 Gauss developed and published a double conformal projection preserving scale along the central meridian, and it was in practice for calculating the Hannover triangulation. The theory of this projection was also published by Oskar Schreiber in 1866." Kruger went on to do the same thing without the double projection.

Note: lproj demonstrates that the scale is 1 only at the origin of the projection and only maintains a low value along the central meridian. Rapid increase in the scale error occurs with increase in longitude from the central meridian. Lproj error? I doubt it because I have had good results matching other projections with use the Gauss transformation to the sphere. Also, from the American Society of Photogrammetry and Remote Sensing and/or LSU (probably Mugnier) the description: "The second step is a Gauss-Schreiber Transverse projection of the sphere onto a cylinder secant about the central meridian of the projection." This is *not* the same as having constant scale along the central meridian. This comment is from the description of the Laborde Madagascar projection (see  http://www.asprs.org/resources/grids/02-2000-madagascar.pdf). Note also that Laborde's cartographic efforts appear to occur early in the 20th century and thus a much later time line than Schreiber.

Getting back to the bottom line, I can only support the use of the designation of Schreiber for this projection. *Certainly* not "simplified." Laborde already has his claim to fame with the Madagascar projection and it can do nothing but confuse the matter by including his name with this projection. If we really want to be fair then "Gauss-Schreiber" is the correct designation. A footnote can be added that Laborde also used this method for such and such application.

Lastly, this naming debate is too important to be limited to a small group and should be open to discussion by a wider audience and a broader consensus obtained.

[dgrichard]

I am really getting confused. Searching "stmerc proj" on Google leads me to a lot of mails including the ones asking to change from stmerc to glaborde :

 http://lists.maptools.org/pipermail/proj/2006-June/002293.html
 http://www.nabble.com/Changing-the-name-or-stmerc-to-glaborde-td18077847.html
 http://news.gmane.org/group/gmane.comp.gis.proj-4.devel/last=0/force_load=very/?page=12
 http://usergroups.map-s.de/article.php?id=2962&group=gmane.comp.gis.proj-4.devel

I proposed to make such a change : glabsgm to glaborde ...

You replied with Réunion Gauss-Laborde (NT/G 73 within my organisation) is the same as libproj stmerc known as Gauss-Schreiber projection ...

After internal talks within my organisation, I said that Réunion Gauss-Laborde is the well-known name of that projection within France. We accept using stmerc if the s is not for Schreiber because of the well-known usage in France ...

I put in the metadata of the stmerc projection both Schreiber and Gauss-Laborde and we are still stuck in exchanging about the name when the work has already been done ...

My feeling is that I wanted to reach a consensus as fast as possible in using the glaborde first as seen in the above links, then in using stmerc to be compliant with libproj, and proposed new names to have both Schreiber and Gauss-Laborde in. None of the chosen acronyms seem to be good enough to reach that consensus and I don't want to spend the next days in choosing it.

As already stated, I can change again the trunk with glaborde or sgltmerc or glstmerc, but I will do it once. Someone has to make the choice.

[martinoty]

To support what Didier Richard just said, I would like to insist on the fact that this projection is known to be used on one single territory in the whole world: La Réunion island. The French HABIT is to called it "Gauss-Laborde". Maybe it is a pity or maybe not, but that's the habit. Hence, it is an absolute necessity that the phrase "Gauss-Laborde" should appear in the projection's name or in the metadata. That's a point that cannot be bargained on.

Concerning the name: Gerald Evenden himself proposed to rename stmerc to glaborde (see mails referenced above). We agreed on this.

It looks as if he changed his mind, and puts forward the name stmerc. We agreed on this again, provided the metadata are modified.

It seems that's still a problem. I'm afraid I no longer see the point.

I would like to stress that French users are waiting for a fast release of PROJ.4 4.6.1, especially to be able to make conversions between NTF and RGF93. That's what matters.

[gie]

After doing a google on "Gauss-Schreiber", the name is well established as the proper name for the double conversion transvers Mercator. I also found a hit:

 http://www.bl.uk/ukmarc/appd0322az.html

dg Formosa (Taiwan) Gauss-Schreiber Co-ordinate System fs Japan Gauss-Schreiber Grid ge Korea Gauss-Schreiber Co-ordinate System

These need to be verified as to current use.

It also appears that several book level references (I can't check because of costs for copies) also refer to Gauss-Schreiber.

It seems to me that to apply Larborde to this projection only because he used it on a small island in the Indian Ocean (that isn't on any of my atlases) seems to be *very* questionable. Especially in light of the fact that it may have been or is use in other regions and is known by another name. I must also opine that a singular national habit of nomenclature is a rather shaky foundation for naming a projection.

Lastly, I did suggest Laborde at one time but that was before I did more research on the background of the projection. I was definitely in error and apologize.

On another matter, it is recognized that the scale is not constant along the central meridian as implied in the earlier citation and questioned by yours truly.

[dgrichard]

Seems we have reached a consensus on using "Gauss-Schreiber transverse mercator".
I will add for our users the "aka Gauss-Laborde Reunion" in the metadata.
The acronyme will be gstmerc.

I will post the changes to the trunk as soon as I receive a definitive go from Frank and Gerald.

[gie]

I agree to "gstmerc" and I will change my distribution. I will note in documentation of the Laborde factor regarding Reunion.

File proj_stmerc.c modified and renamed to proj_gstmerc.c, modifications to Makefile and proj_list.h.

[gie]

Archive file lbp4.3_20080728.tar.bz2 is now available on

 http://members.verizon.net/~gerald.evenden/proj4/

I am predicting a complaint about +proj=gstmerc +lat_0=, but we will wait and see.

Enjoy.

[dgrichard]

Thanks Gerald, I am moving on.

[warmerdam]

Folks, can we close this ticket now?

[dgrichard]

Replying to warmerdam:

Folks, can we close this ticket now?

Yes, my mistake. I close it directly if you don't mind.