[GRASS5] Re: [GRASSLIST:1874] Re: Importing USGS SDTS DLG maps fromdifferingdatums? I'm confused. [very long]

[Markus Neteler]

Hi Roger

(cc to grass5 as it will be of interest)

On Sat, May 26, 2001 at 01:32:53PM -0600, Roger S. Miller wrote:
> Markus Neteler wrote:
> 
> > >From GRASS 5.1 onwards GRASS will use the latest PROJ 4.x software available
> > from http://www.remotesensing.org/proj/ maintained by Frank Warmerdam
> > including datum transform.
> 
> Markus,
> 
> Thanks for the information.  It good to know that datum shifts will
> eventually be added.  Right now we have a very unfortunately situation. 
> I think it would be a mistake to release GRASS 5.0 without changing it's
> documentation; probably the projection programs need some kind of trap
> added to prevent people from using them for datum shifts.

yes, I agree 100%.
 
> Everyone I know of who uses geographic data deals with conversions
> between NAD27 and NAD83.  Probably anyone in the US who has used GRASS
> in the last few months to do more than run tutorials has used GRASS to
> convert between NAD83 and NAD27 data.  They have every reason to expect
> that the conversion works as described.  But it doesn't.  Their data are
> now corrupt.

.. which is a very unfortunate situation!
 
> My experience... Ground water rights in this area are administered by
> the State of New Mexico, using a ground water model that was built using
> ArcInfo on a lambert conformal conic base, using NAD27.  Like everyone
> else in this area who deals with ground water planning, I have to
> maintain a working version of that model. My GRASS database includes a
> location for lambert conformal conic, NAD27 datum to support that model.
> 
> I spent a day in the field back in March with a consultant who
> represented "the other side" in a ground water permit hearing.  We
> surveyed (with their GPS) the location of each of my clients' wells, and
> the location of 
> their sewage outfall on the left bank of the Rio Grande.  They forwarded
> the results to me after we got back to the office, using UTM, NAD83 in
> an ESRI format.  I built a location for UTM zone 13, NAD83 datum,
> converted their surveyed points to GRASS format and then used v.proj to
> put them in my lambert conformal conic, NAD27 location.  I (and every
> other GRASS user) had every reason to believe that the reprojection was
> correct.  There were no errors.  There were no warnings.
> 
> The surveyed locations didn't come out to where they were supposed to
> be, in fact two of the wells fell within the right-of-way for an
> Interstate Highway.  However, the ground water model has a fairly coarse
> grid in that area and all of the wells fell within the same model cell
> where I already simulated them, so the location error wasn't a problem. 
> It was more of a problem that the sewage outfall fell on the right bank
> of the Rio Grande rather than on the left bank.
> 
> I was lucky in this instance that the model grid was coarse and that the
> exact location of the sewage outfall wasn't very important.  I think you
> can see that much more serious consequences are possible.
> 
> 
> Roger Miller
> Lee Wilson and Associates

in my opinion a full migration to latest PROJ4 *including* the datum shift
should not be a big task. GRASS is using a few wrapper functions to PROJ4
(src/libes/proj). As *.proj use these wrapper functions, a datum shift
supported by PROJ4 should be available then to these GRASS functions as
well through the wrapper (or not?).

We should consider to upgrade PROJ4 already in 5.0 and not postpone it
to 5.1. Say, at least *.proj should take care of the datum shift.
Nowadays it will be somewhat hard to explain why GRASS 5.0.0stable still
won't do datum transforms. As not too much is left for the stable release,
but many, many changes awaiting for 5.1, I suggest to upgrade PROJ4 now.

If I am totally wrong, please don't hesiate to tell me :-)

Regards

 Markus Neteler