Heiko,
On 06/12/16 13:55, Heiko Klein wrote:
Hi,
we just extended our meteorology forecast runs with ensemble members.
Data without ensemble/realization dimension can be found here:
http://thredds.met.no/thredds/dodsC/arome25/arome_metcoop_test2_5km_latest.nc
and these work well with the netcdf-java base software like
thredds/idv/ncwms , e.g. precipitation_amount_acc(time,height0,y,x)
As soon as we combine several realizations/ensembles into the same file
http://thredds.met.no/thredds/dodsC/meps25files/meps_allmembers_full_2_5km_latest.nc
all gridded feature stop working for
precipitation_amount_acc(time,height0,ensemble_member,y,x). We can still
use opendap, but ncwms and idv no longer find this variable.
netcdf-tools still detect the variable as a FeatureType GRID
f:D(5)->R(7) (in FeatureType->FeatureScan) but is not able to detect the
coordinateSystem in FeatureType->Grids.
I even tried to remove the height0 dimension (which is of size 1), but
without any change.
Add the _CoordinateAxisType to the ensemble_member coordinate/variable.
I made it in the attached .ncml which is just pointing to your file and
adding the attribute.
Is 'standard_name=realization' an unsupported feature of netcdf-java?
no, the netcdf-java doesn't recognize the realization standard_name as a
hint for the Ensemble Coordinate Axis Type.
But, it would be easy to fix it.
There was a longer discussion about it in 2010:
http://mailman.cgd.ucar.edu/pipermail/cf-metadata/2010/053546.html
Should we use some axis="ensemble" or _CoordinateAxis keyword to help
detect the ensemble dimension?
Yes, .... never ending, the current status is this:
http://cf-trac.llnl.gov/trac/ticket/142
my proposal would be to include the standard_name='realization' as hint
for the Ensemble until the CF governance take a decision.
Or is the problem within thredds/ncwms/idv not able to deal with other
dimensions than x,y,z,t?
The real issue is an exception related with the run date. The
netcdf-java Grid discovery is complaining, but my guess is because you
choose a order on dimensions that is making the Coordinate System
builder to complain.
In fact in your dataset the forecast_reference_time si defined as scalar
coordinate and should be added to the list of coordinates on the
variables (to be CF friendly). But this scalar (no dimensions) is
confusing to the coordinate system builder. If you can, define it as
singleton coordinates (1D)
Best regards
Antonio
Sorry for cross-posting, not really sure where the problem lies. We're
using thredds 4.6 and netcdf-java toolsUI 4.6
Best regards,
Heiko
--
Antonio S. Cofiño
Associate Professor and Researcher
Grupo de Meteorología de Santander
Dep. of Applied Mathematics and Computer Sciences
Universidad de Cantabria (Spain)
Academic Visitor
National Centre for Atmospheric Science
Department of Meteorology
School of Mathematical, Physical and Computational Sciences
University of Reading (UK)
http://antonio.cofino.es
<?xml version='1.0' encoding='UTF-8'?>
<netcdf xmlns='http://www.unidata.ucar.edu/namespaces/netcdf/ncml-2.2'
location='dods://thredds.met.no/thredds/dodsC/meps25files/meps_allmembers_full_2_5km_latest.nc'
>
<variable name='ensemble_member'>
<attribute name='_CoordinateAxisType' value='Ensemble' />
</variable>
</netcdf>
