[Proj] epsg parameters

[Paul Kelly]

On Wed, 2 Apr 2008, Frank Warmerdam wrote:

> The problem is that if EPSG offers more than one transformation to
> WGS84 for a datum, the automated translation gives up and offers none
> of them preferring for the user to make the decision themselves.
>
> I suspect, given sufficient investigation, you would find that EPSG
> offers more than one transformation for the problem datums.
>
> I'm not saying this is a good decision in the automated translation.

I tend to think it's the best you can do given the situation. We have had 
a similar/related issue in GRASS, and the current kludgy solution is to 
store our own database of datum transformation parameters along with extra 
information on where geographically the transformation is valid and the 
accuracy of it. When importing a co-ordinate system with ambiguous datum 
information the user is then presented with this choice and forced to make 
it.

I think it's in general a bad idea to include datum transformation 
information without any user interaction, because many users have the 
misconception that there is somehow a "correct" set of parameters which 
will always give them accurate results. When of course, as we've seen 
discussed on this list many times, accurate datum transformation is very 
much a localised thing: a transformation that gives accurate results in 
one region of a country may give very inaccurate results in another 
region, and a transformation that covers the whole country is in general 
not very accurate anywhere.

One minor complication that I don't think I have ever seen given much 
discussion (and the main reason I'm sending this mail) is that some 
classical datums have recently been redefined by the relevant national 
mapping agencies in terms of their relationship to WGS84 (see Note 1) In 
this situation then (in my understanding) there is one and only one 
"correct" set of datum transformation parameters, and it *would* be valid 
to include this without any user interaction (I'm not necessarily talking 
about generation of the PROJ epsg file, just in general).
I was wondering though did anybody know if there was any standard way of 
specifying (for example in the WKT format, or in the EPSG database) that a 
set of datum transformation parameters is "exact" in this way?

Paul

Note 1: E.g. OSGB36 in the UK 
http://www.ordnancesurvey.co.uk/oswebsite/gps/information/coordinatesystemsinfo/guidecontents/guide5.html
Although this uses a grid-based datum shift rather than a set of 
parameters, the prinicple of the datum having an exact relationship with 
WGS84 is still the same.