[GRASSLIST:4337] Re: GIS visitor requests directions

[Victor Wren]

On Tue, Aug 20, 2002 at 01:36:37AM +0100, Glynn Clements wrote:

> The projection utilities currently ignore the datum, except for the
> special case where one of the locations uses NAD27 and the other
> NAD83. It wouldn't necessarily be that hard to add generalised datum
> shifts; GRASS already has the necessary code, it just isn't used by
> *.proj.
> 
> I don't necessarily agree with Victor; the facts that is reporting a
> problem with converting between locations which both use NAD83, and
> with errors of around 200 miles both suggest that something else is at
> play here.

Actually, it did not report any error.  It simply imported the map in
the wrong location without indicating that anything at all was awry.

The location shift was consistent with importing State Plane raster into
a UTM map without actually translating coordinates (I'd have to experiment
to be sure -- it SHOULD have fallen outside the boundarids of the datum,
but I've noticed things like that aren't always checked)

> AFAIK, differences between datums are typically less than 1km

If the translation is done correctly, that's true, but my assertion is that
the software is getting confused or misidentifying the projection, perhaps 
because both projections are NAD83 datum.  The differences in raw numbers
between STPL and UTM are potentially in the millions of meters.  Without
looking at the code, which I'm not competent to do, it would
be pure speculation to guess how it might be handled if a state plane 
projection had a northing in the millions of meters (clearly impossible)

>, and
> differences between international and US survey feet are around 2ppm
> (an error of ~40 metres between antipodes). An error of 200 miles
> (i.e. ~2.9 degrees) would have to be due to something major, e.g. 
> using the wrong projection parameters.

I agree.  The question is only whether I am the one making the 
mistake, or whether it is r.proj.  I have my own favorite answer, of
course, but I'm biased.  I'm still gamely learning my way around this
beast, though, so my room for making errors is still large.  I just
gave up after the first attempt, because I was optimistically attempting
to use it in a way which it was clearly labelled not to work.  If you 
say that it SHOULD have worked under G5.0P5, I'll give it another try.

Victor Wren