[GRASS5] [bug #4122] (grass) r.in.gdal won't import some data into lat long projections

[Hamish]

> this bug's URL: http://intevation.de/rt/webrt?serial_num=4122
> ---------------------------------------------------------------------
> 
> Subject: r.in.gdal won't import some data into lat long projections
> 
..
> GRASS Version: 6.0.1
> 
> I have an image file (short-int png) that is a projected raster map in
> the EPSG:4326 (lat/long, wgs84) projection. It's got a nice bounding
> box that goes around the outside of the pixels. The bounding box is
> -119,41.8,-116.8,44.5. The georeferencing is in a text file with the
> coverage that states the northern, southern, eastern, and western
> borders. This data does not need to be rectified or reprojected; it is
> already in the projection of the grass database, so importing the data
> should be simple.
> 
> In other projections, when I have encountered this sort of data, I
> have imported it "overriding" the projection of the file with the
> projection of the current database. I then used r.region to set the
> bounding box of the raster and am done. So this ought to be easy. I
> do:
> 
> r.in.gdal input=coverage.png output=coverage -o
> 
> However this returns:
> 
> Over-riding projection check.
> Proceeding with import...
> WARNING: G_set_window(): Illegal latitude for South
> 
> This is because, in the png image, the coordinates of the data run
> along the lines of -2000 to 0 in the y direction and -2000 to 0 in the
> x direction. -2000 isn't a valid southern latitude. Thanks for the
> warning. It's even a good thing for the program to stop here to make
> sure the importer knows that something's amiss. However I really do
> mean to do this as the numbering of the pixels in the PNG image is
> uninteresting. This says it is a warning, not an error, yet there is
> no way to continue ignoring this warning. What matters to me, and to
> grass, is the region I am going to assign it after this step.
> 
> Solutions:
> 
> 1. (best in my opinion) Provide a method to ignore this warning. Make
> sure that our bases are covered in that similar export warnings and
> reprojection warnings are in place.
> 
> 2. Change the text from "WARNING: G_set_window(): Illegal latitude for
> South" to "ERROR: G_set_window(): Illegal latitude for South". Provide
> another easy avenue for importing data that doesn't include its
> georeferencing but is already projected. Note the idiosynchracies of
> the latlong projection with respect to this behaviour in the
> documentation.
> 
> 3. Change the r.in stuff to have options for the window be provided
> with the command.
> 
> 4. Add an import command that wraps up the r.in infrastructure so that
> it supports shifting data (providing georeferencing from outside the
> file). If this is done (I think it'd have to use i. commands) we might
> as well go all the way and handle reprojection too.


Unreferenced images are imported as pixels, once you get more than 90
rows it gets upset.

import into a new XY location, use r.region to set correct bounds, check
resolution is correct with r.info, and either copy mapset into a 
projected location or change your XY location with g.setproj.

gdal import often support "world files" containing geo-ref info.


Hamish