Hi Roland-
On 11/5/13 8:45 AM, Roland Schweitzer - NOAA Affiliate wrote:
I see hints in the Java netCDF API that it can understand a grid that
has an ensemble axis. What's not clear to me is how one constructs a
netCDF data source (either in the file or via some sort of TDS
aggregation) that represents the ensemble.
It's documented if you read the code. ;-) In the CF1Convention parser:
if (sname.equalsIgnoreCase("ensemble")) { // experimental
v.addAttribute(new Attribute(_Coordinate.AxisType,
AxisType.Ensemble.toString()));
continue;
}
where sname is the standard name. So, if you add a standard name of
"ensemble" to your coordinate axis with a CF-1.X Conventions, that will
be picked up as an ensemble dimension by the netCDF-Java library (and
hence TDS).
I generally set the following attributes:
long_name: "ensemble";
standard_name: "ensemble";
axis: "E";
(I guess with the above code the latter should bet axis: "Ensemble", but
it's not really used as far as I know).
You can also have this as a char variable using the names of the
perturbations (e.g. E1,E2,E3,E4). I mostly use ints.
The CF conventions document
doesn't seem say anything about ensembles.
There was a discussion a while back about having a CF convention for the
ensemble dimension, but there was never any consensus. I think people
got bogged down in the difficult cases instead of implementing something
that worked most of the time (like what John did) and iterate from
there. I've been using John's stuff for several years either directly
in a netCDF file or through NcML aggregation.
Google is failing me, can anybody offer some examples or documentation
links for data sources with an ensemble axis?
I figured this out from private conversations with John instead of going
through support so those didn't end up in the Google-verse.
Hope this helps.
Don
--
Don Murray
NOAA/ESRL/PSD and CU-CIRES
303-497-3596
http://www.esrl.noaa.gov/psd/people/don.murray/
