John,
Thanks for the quick response.
1. Sadly, I'm just a recipient of the netCDF files created by the v4.0
API and the upstream source of this data is not planning on upgrading anytime
soon. That said, I expected you guys might say that so I did verified that the
code works this way in 4.3 too.
2. Understood. I've already finished the code that does the netCDF ->
GRIB conversion so this was more of a "why does it work that way?" type of
email.
3. Gotcha. I wasn't even thinking of it from a GIS standpoint. After
spending time writing code that handles GRIB data, I'm used to the first point
in a GRIB record being 0,0 and going from there. The only reason I brought it
up was that it was the first indication that the projection being used to
generate the x and y is not using LA1 and LO1 as the originating point.
4. I'm guessing this is where I'm getting hung up. All this time, I've
been working with the assumption that LA1 and LO1 from the GDS was origin of
the projection. How was it decided that LATIN1 and LOV were going to be used
for the origin instead? I've talked to a two other software devs who work with
GRIB (full disclosure, one is a meteorologist, the other is not), and they
haven't seen it done this way either.
I appreciate the input here. To be fair, now that I understand how it works,
my conversion code spits out great GRIB formatted messages and, as I mentioned
above, this was primarily a follow up to see why it is done this way and
determine if there's a problem.
Bryan
From: netcdf-java-bounces@xxxxxxxxxxxxxxxx
[mailto:netcdf-java-bounces@xxxxxxxxxxxxxxxx] On Behalf Of John Caron
Sent: Monday, June 25, 2012 1:30 PM
To: netcdf-java@xxxxxxxxxxxxxxxx
Subject: Re: [netcdf-java] GRIB-2-netCDF Projection Question
Hi Bryan:
1. The classes you are working with in NetCDF-Java 4.0 are no longer supported.
Please look at ucar.nc2.grib.* classes in NetCDF-Java 4.3.
2. NetCDF-Java does not support writing GRIB files, only reading them. So we
support GRIB -> netcdf/CF, but not the other way around. In principle, if you
understand the CF conventions, you can write the correct GRIB GDS from them.
3. The convention that projection x and y values must be positive is not used.
Thats more of a GIS thing and is rather artificial; one typically has to add
"false_northing" and "false_easting" to make it true.
4. The "origin of the projection" may also be different than what you are used
to. It is essentially the place where x=0, y=0 in the projection defined by the
LambertConformal class; LA1 and LO1 is the position of one of the corners of
the grid. Theres no documentation other than the code for GRIB, but you might
find this helpful:
http://www.unidata.ucar.edu/software/netcdf-java/reference/StandardCoordinateTransforms.html
John
On 6/25/2012 12:14 PM, Rockwood, Bryan wrote:
Greetings,
As part of the process of creating a GRIB1 file from subgridded netCDF data
generated by the NetCDF-Java 4.0 API (long story), I had a difficult time
figuring out how the X and Y values from the netCDF were calculated. Since
there were negative X and Y values, it appeared that the projection used to
calculate X and Y were not grid based. So, I obtained the Java netCDF source
and proceeded to debug through the code. After some exploring, I found a class
called GridHorizCoordSys in the package ucar.nc2.iosp.grid. In said class is a
private method called makeLC and the first thing it does is create a projection
for the GRIB record. Now here's where I'm confused. This is the code in
question:
proj = new LambertConformal(
gds.getDouble(GridDefRecord.LATIN1),
gds.getDouble(GridDefRecord.LOV),
gds.getDouble(GridDefRecord.LATIN1),
gds.getDouble(GridDefRecord.LATIN2));
Looking at the javadoc for the LambertConformal constructor, the first two
parameters should be the latitude and longitude origin of the coordinate
system. In my experience when dealing with Lambert Conformal grids, this is
typically GridDefRecord.LA1 and GridDefRecord.LO1, respectively, from the GRIB
GDS. The code above is using GridDefRecord.LATIN1 and GridDefRecord.LOV
instead. Once I found out how the projection was being created, it was then
possible to get the LA1 and LO1 from the X and Y within the netCDF to create a
valid GDS for the GRIB data I'm generating. My questions are as follows:
Why use the LATIN1 and LOV to set the lat and lon origin of the projection
instead of LA1 and LO1?
Is this documented someplace? I'm curious how someone such as myself would
know to create a projection like this without debugging through the code.
Bryan Rockwood
_______________________________________________
netcdf-java mailing list
netcdf-java@xxxxxxxxxxxxxxxx<mailto:netcdf-java@xxxxxxxxxxxxxxxx>
For list information or to unsubscribe, visit:
http://www.unidata.ucar.edu/mailing_lists/
