[GRASSLIST:2201] Re: Another problem with r.proj

[Eric G. Miller]

On Sat, Jul 28, 2001 at 11:29:22AM -0700, Rich Shepard wrote:
> On Sat, 28 Jul 2001, Eric G. Miller wrote:
> 
> > > 	PROJ_UNITS:
> > > unit: unknown
> > > units: unknowns
> > > meter: 1
> >
> > Does it help if you change the PROJ_UNITS to?
> >
> > unit: meter
> > units: meters
> > meter: 1.0
> 
> Eric,
> 
>   I didn't pick this up until I sent the response. I just changed it and,
> no, it makes no difference.
> 
> > Also, it looks like you need to specify the input "mapset" for the
> > location (i.e. "spc").
> 
>   Tried that, too. I posted to the list as well as to Len when he suggested
> this.

Hmm. I don't know why it is failing to get the projection info.  Are you
sure your arguments for location, mapset and dbase are correct?  r.proj
uses standard library functions to retrieve this information...

> > Actually, the datum transformation needs to be done via geodetic
> > coordinates following the inverse projection from UTM.  Unfortunately,
> > the projection routines in GRASS don't do this at the present.
> 
>   Would it be a Good Thing at this point for me to use r.out.ascii and
> v.out.ascii on these data sets and then use the USGS 'proj' program? I don't
> know if that does datum convertions as well as projection conversions.
> 
> > It may or may not be easier to use i.rectify to transform the DEM --
> > provided you can find suitable control coordinates. (hint: select 6 or 8
> > points spread across the input source; transform them to geodetic
> > coordinates; use m.datum.shift to get the corresponding datum shifted
> > geodetic coordinates; project the coordinates to your target projection;
> > voila, you have control points in both coordinate systems).
> 
>   I can't see this as being easier. :-)

Well.  Neither option seems particularly convenient.  Rewriting the
whole coordinate system management framework and associated projection
routines has been discussed for 5.1.  It's sort of a hodgepodge of two
different approaches right now, neither of which address datum
transformations.

>   So, does it look like a bug in r.proj to you?

I dunno.  I still would ask that you double check your arguments to the
command.  It seems it is not looking in the right place for projection
info.

-- 
Eric G. Miller <egm2 at jps.net>