[GRASSLIST:7508] Re: "Illegal latitude" problem

[Hamish]

> I've encountered a strange problem in r.in.gdal where I get the
> following error:
> 
> "WARNING: G_set_window(): Illegal latitude for South"
> 
> What I did was:
> 
> "GRASS 6.0.0 (GL1KMWGS84):~/DATA/GL1KMWGS84/PERMANENT > r.in.gdal
> -oe o10g out=o10g"
> 
> This is especially strange because I'm converting GLOBE DEM tiles,
> four of which share the same south latitude, and no others produce
> this error.
..
> WARNING: G_set_window(): Illegal latitude for South
..
> Incidentally, I got the same error for North latitude when trying to
> convert the global GLCF SRTM-GTOPO30 mosaic (GeoTiff), as well as an
> in-house global elevation grid (Arc/Info, not sure of original
> source).
> 
> Has anyone seen this before and come-up with a fix?


Yes, this has been seen before, search the mailing list archives (both
the users and devel list, I can't remember which) for discussion on this
topic from the last month or two. (Specifically the SRTM-GTOPO30
datasets)

Not surprisingly, a GRASS lat/lon location won't let you set the
latitude bound higher than 90 degrees. These datasets will, for example,
have the top half of a 3" cell extend into the >90 degree territory. 

GRASS measures the rasters' region bounds as the outer extent of the
boundary cells, not the value at the cell centers.

In my personal opinion, the datasets are broken in this case, not GRASS.
GRASS should be tolerant of floating point error taking the latitude
bound slightly beyond 90, but not accept half a cell's worth of error.

Work-around: import into a simple XY location, not a lat/lon location.
After cleaning the data (e.g. by chopping off the offending line of
data) you can copy the map into a lat/lon location.

Use gdalinfo to explore what the data bounds are.



Hamish