NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.

Re: [netcdf-java] Rotated latitude/longitude GRIB2 coordinates

FYI I have submitted a pull request (and Christian found some test failures in the Unidata suite): 
https://github.com/Unidata/thredds/pull/531

Kind regards,
Ben.

On 14/03/16 14:23, Ben Caradoc-Davies wrote:

John,

I have partly implemented NetCDF-Java support for rotated
latitude/longitude GRIB2 files with the unusual NCEP GDS template 32769:
http://www.nco.ncep.noaa.gov/pmb/docs/grib2/grib2_temp3-32769.shtml

Although this is an unusual GDS, my investigations indicate that it is
the same projection as GDS template 1 but defined by different GDS
parameters. I am thus keen to reuse the existing implementation
Grib2Gds.RotatedLatLon to reduce code duplication.

During implementation, I encountered some behaviour of GribIosp and
Grib2Gds.RotatedLatLon that I do not understand:

(1) Grib2Gds.RotatedLatLon appears to make no transformation to the
rotated coordinate system. The generated lat/lon will extend from
la1/lo1 to la2/lo2 (the unrotated corners). What happened to the new
rotated origin? I expected the rotated grid to centred on (0,0) if the
southern pole parameters set the origin as the centre of the grid (not
mandatory but a popular choice):
https://github.com/Unidata/thredds/blob/master/grib/src/main/java/ucar/nc2/grib/grib2/Grib2Gds.java#L535


(2) The coordinate variables created in GribIosp for rotated
longitude/latitude GRIB2 grids do not appear to conform to the CF
Conventions as they lack rlat/rlon and standards_name
grid_latitude/grid_longitude and do not contain multidimensional lat/lon
as required. Instead they appear to contain a simple lat/lon as for an
unrotated grid:
http://cfconventions.org/cf-conventions/cf-conventions.html#grid-mappings-and-projections


(3) Rather than using GdsHorizCoordSys.isLatLon as for GRIB1, for GRIB2
files GribIosp uses GDS template range checks to decide if a grid is
lat/lon. This seems to lack extensibility. What is the reason for this
implementation choice, and can I change it?
https://github.com/Unidata/thredds/blob/master/grib/src/main/java/ucar/nc2/grib/collection/GribIosp.java#L240

https://github.com/Unidata/thredds/blob/master/grib/src/main/java/ucar/nc2/grib/grib2/Grib2Utils.java#L116


Is anyone using NetCDF-Java support for GDS template 1 GRIB2 files? Does
anyone have any sample files for GDS template 1 and known good NetCDF CF
Conventions equivalents?

I suspect the NetCDF-Java implementation of GDS template 1 may be
incorrect; if I am right, I intend to fix it as a side-effect of
implementing support for GDS 32769. If I am wrong, please stop me now!

Kind regards,



--
Ben Caradoc-Davies <ben@xxxxxxxxxxxx>
Director
Transient Software Limited <http://transient.nz/>
New Zealand