[PROJ] Meaning of (x) numbers in proj transformations in QGIS

[Even Rouault]

Hi Richard,

> According to Nyall, he uses the info from proj to label the 'best fit'
> green, but I also see 'higher order' numbers in the descriptions of the
> options (numbers between brackets). My main reason to come here:
> apparently (3) is by proj/Nyall seen as current, but then I wonder why
> there is a (4) option...

Tough topic. The numbers in brackets (3), (4) come directly from the name of 
the transformation from the EPSG dataset. I don't think they have a particular 
meaning, except to distinguish transformations which go from the same source 
to the same target. The latest number is probably the most recent, but PROJ 
doesn't make any such hypothesis.
So when PROJ collects all possible transformations and tries to filter and 
sort them from the most relevant to the less, it takes into account different 
parameters: accuracy advertized, area of use, availability of grids, 
supersession information, ... and when all those criteria are equal, it just 
sorts by name... Here EPSG:4833 (Amersfoort to WGS 84 (4)) and EPSG:15934 
(Amersfoort to WGS 84 (3)) have the same accuracy, area of use, don't require 
grids and ... are reported as being both current ! The text remark that EPSG:
4833 replaces EPSG:15934 isn't translated in a record in the 
'epsg_supersession' table (whereas EPSG:11125 and EPSG:1672 are clearly marked 
as superseded by other codes).
So for that particular case, PROJ probably makes the "wrong" (see final 
comment at end of email) choice based on incomplete information from the EPSG 
dataset. You should probably file a change request to IOGP so that they add an 
entry in the supersession table. Unless this is intended by them to keep only 
the last 2 valid entries (perhaps "replaced by" doesn't equate to "is 
superseded by"...), and just tag as supersed the previous ones.

If one wanted to improve the situation from the available information, there 
could be two choices:
- analyze text remarks to guess extra supersession information. That is look 
for pattern like "Replaces FOO (code BAR)"
- take into account the " (number)" trailing information, to put first 
transformation with higher number, but I'm wondering if that wouldn't be 
fragile in other situations.
Not sure which one I dislike the less :-)

Another thing is that currently the import process of the EPSG dataset to the 
PROJ dataset doesn't import remarks, so they are lost to the end user. I 
didn't want to do that for now to limit the size of the database, but that 
might be discussed.

> But my main question: is "which proper information an average QGIS user
> can use to determine what transformation to pick.." if presented with
> one of those dialogs (as I had....)... and not being a projection guru :-)

Ah ah, anyone knowning the answer would be multi-millionaire ;-)

But to come back to your particular transformation, the thing is that which 
one of "Amersfoort to WGS 84 (3)" or "Amersfoort to WGS 84 (4)" should be 
picked up depends probably more on which one was used by other data producers 
to do similar transformations. Given the accuracy of definition of both 
Amersfoort and WGS 84, both are OK. This is mostly an issue of using 
consistently the same one to avoid alignment problems.

Even

-- 
Spatialys - Geospatial professional services
http://www.spatialys.com