[Proj] Trouble georeferencing data stored in a spherical projection.

Mikael Rittri Mikael.Rittri at carmenta.com
Tue Mar 6 05:38:03 PST 2012


Hello, (I hope this is approximately the right forum).


I have had some trouble with meteorological GeoTIFF data, stored in stereographic
based on a GEOGCS called "NinJo sphere", without any EPSG code, using a sphere
with radius 6378137 m.

The main problem wasn't the lack of an ordinary geodetic datum, since the most detailed
file had the resolution 4.6 km, much more than the typical difference between datums.

What bothers me more, with spherical data, is that it seems impossible to know if the
spherical latitude should be interpreted as identical to WGS84 latitude (geodetic latitude),
or if it should be interpreted as geocentric latitude that needs conversion to WGS84.
The difference between the two interpretations can be up to 21 km, which can be
significant even when the datum choice isn't.

I was hoping to find that one of the alternatives was a de facto standard, but
instead I found evidence that both of them are in use.

Spherical latitude = geodetic latitude is used by:

      Data in Web Mercator, like Open Street Map,

      Meteorological data in any of the models WRF, MM5 and NinJo.

Spherical latitude = geocentric latitude is used by:

      MISR Regional UAE2 Imagery (from NASA Earth Data).



In an old post, Melita Kennedy has written that it is "probably more correct" to use the
second convention; maybe she has experience with data like MISR.

The Proj.4 FAQ discusses Web Mercator and mentions "many other datasets ... such as
many NASA datasets" as using the same convention, so maybe NASA is not consistent
in its choice of convention.



Anyway, what I would like is some standard way to tag the metadata, in GeoTIFF in
my case but more generally in WKT, to specify which convention is used.
Of course, the problem is already solved in Proj.4, as the FAQ says, since we can use

     +nadgrids=@null

for the first convention, and

     +towgs84=0,0,0

for the second one.

In GeoTiff, specifying TOWGS84[0,0,0] in the same way will be possible as soon
as the proposal http://trac.osgeo.org/geotiff/wiki/TOWGS84GeoKey is supported,
but there will still be no way to specify the first convention.

In WKT, too, only the second convention can be specified.

So I wonder, is there already some discussion going on about how the first convention
could be specified? (Perhaps within OGC or EPSG.)  If not, can we start it?

Best regards

Mikael Rittri
Carmenta
Sweden
http://www.carmenta.com<http://www.carmenta.com/>



REFERENCES:

The way to interpret WRF and MM5 data: http://www.ie.unc.edu/cempd/projects/mims/spatial/grids_ellipsoids_map_proj.html

The way to interpret NinJo data: personal trial-and-error.

The way to interpret MISR UAE2 data: http://eosweb.larc.nasa.gov/PRODOCS/misr/uae/imagery/regional/map_projection.html

Melita Kennedy's post: http://lists.osgeo.org/pipermail/metacrs/2008-August/000144.html

Proj.4 FAQ: http://trac.osgeo.org/proj/wiki/FAQ#ChangingEllipsoidWhycantIconvertfromWGS84toGoogleEarthVirtualGlobeMercator










-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/proj/attachments/20120306/515963ad/attachment.html>


More information about the Proj mailing list