Hello,
I'm doing some work on updating the GDAL NetCDF driver to be able to
properly export to and import from CF-1 compliant files encoded with
map projection (Grid Mapping) information (see
http://trac.osgeo.org/gdal/ticket/2893 for background).
The driver as it stands doesn't export properly to NetCDF (eg from
GeoTiff) for many projections: the issue is that GDAL stores
projection information internally in OGC WKT form, and the driver has
to 'map' these to appropriate CF-1 attributes when exporting - and
several of these mappings aren't correct currently (eg Lambert
Conformal Conic and Transverse Mercator were ok, but Albers Equal
Area, Azimuthal Equidistant and others weren't working).
The main issue seems to be that for many projections, where OGC WKT
uses an attribute like 'latitude_of_center', the CF-1 closest
corresponding attribute seems to be 'latitude_of_projection_origin'.
For a projection such as Albers Equal Area, do these mean exactly the
same thing?
Testing the resultant translated file in IDV once I've added the above
attribute mapping suggested the reprojected file is displaying with
the correct coordinates. I'm not a geodsy or projections expert, but
looking at the remotesensing.org and other pages suggested these
parameters may be equivalent.
Also with regard to longitude paramaters: it seems for some CF-1
projections such as Albers Equal Area, the 'longitude_of_center'
attrib in WKT needs to map to 'longitude_of_central_meridian' in CF-1,
whereas in others such as Lambert azimuthal equal area,
'longitude_of_center' needs to map to
'longitude_of_projection_origin'?
(
http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.5/apf.html)
Would appreciate any comments, or even ways to improve testing
strategy used. I'm aware of the CF-1 compliance checkers, and possibly
should try the NetCDF Java API to check that points correspond.
cheers, Patrick.
Received on Tue Sep 27 2011 - 20:27:54 BST
