In the absence of a standard, I have also made some choices about how
to handle ensemble metadata with GrADS and the GDS. Below is a
shortened example of the ncdump output from a GFS ensemble data set
behind GDS. I put the metadata that GrADS needs in customized
attributes (e.g. grads_name, grads_dim, etc.) I steered clear of using
data or coordinate variables of type 'char', because I feel attributes
are the better place for this kind of metadata, and GrADS and the GDS
have no interface for handing data variables that are not numeric. I
have always assumed it would be necessary to add more metadata to make
this data set CF compliant, but in the meanwhile, the ensemble axis is
a plain old linear dimension. --Jennifer
--
Jennifer M. Adams
IGES/COLA
4041 Powder Mill Road, Suite 302
Calverton, MD 20705
j...@cola.iges.org
# ncdump -c http://monsoondata.org:9090/dods/gfsens/gfsens.2010031612
netcdf gfsens {
dimensions:
ens = 22 ;
lat = 181 ;
lev = 7 ;
lon = 360 ;
time = 31 ;
variables:
double ens(ens) ;
ens:grads_name =
"avg
,c00
,p01
,p02
,p03
,p04,p05,p06,p07,p08,p09,p10,p11,p12,p13,p14,p15,p16,p17,p18,p19,p20" ;
ens:grads_length =
"31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31,31" ;
ens:grads_tinit =
"1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1" ;
ens:avg = "unweighted mean of all members" ;
ens:c00 = "unperturbed control forecast" ;
ens:p01 = "positively perturbed forecast" ;
ens:p02 = "positively perturbed forecast" ;
<additional attributes snipped>
ens:p19= "positively perturbed forecast" ;
ens:p20 = "positively perturbed forecast" ;
ens:grads_dim = "e" ;
ens:grads_size = "22" ;
ens:grads_mapping = "null" ;
ens:units = "count" ;
ens:long_name = "ensemble member" ;
double time(time) ;
time:grads_dim = "t" ;
time:grads_mapping = "linear" ;
time:grads_size = "31" ;
time:grads_min = "12z16mar2010" ;
time:grads_step = "6hr" ;
time:units = "days since 1-1-1 00:00:0.0" ;
time:long_name = "time" ;
time:minimum = "12z16mar2010" ;
time:maximum = "00z24mar2010" ;
time:resolution = 0.25f ;
<other coordinate variables snipped>
float t(ens, time, lev, lat, lon) ;
t:_FillValue = -9.99e+33f ;
t:missing_value = -9.99e+33f ;
t:long_name = "temperature [k] " ;
<other variables snipped>
data:
ens = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18,
19, 20,
21, 22 ;
time = 733848.5, 733848.75, 733849, 733849.25, 733849.5, 733849.75,
733850,
733850.25, 733850.5, 733850.75, 733851, 733851.25, 733851.5,
733851.75,
733852, 733852.25, 733852.5, 733852.75, 733853, 733853.25,
733853.5,
733853.75, 733854, 733854.25, 733854.5, 733854.75, 733855,
733855.25,
733855.5, 733855.75, 733856 ;
}
On Mar 16, 2010, at 1:22 PM, Jonathan Gregory wrote:
Dear John
Im not sure if we ever converged on how to write ensemble files.
No, we didn't. If I remember correctly, we couldn't agree on the
relationship
between attributes and coordinates, and that was a sticking-point. I
thought
that we ought to allow multivalued auxiliary coord variables for
ensemble
axes, corresponding to attributes that might be used to describe a
single
model. I would give these standard_names to identify them. Paco
started the
discussion, and in the end he made his own decisions for what to do
for the
ENSEMBLES project (because CF had not concluded), so he might
comment on it.
Particularly, how does software recognize the ensemble dimension?
We didn't agree. But at the end of the discussion, Steve also
suggested that
it could be identified by an axis attribute, just as you have. I
like that
idea too. But it could also be given a standard name, like other
coordinate
variables e.g. ensemble_member_identifier. I suppose the values of
this
coord var might be either numeric or strings, though it your example
it is
integer. I haven't looked at the trac ticket to see if we discussed
these
points before.
Perhaps we should add :axis = "Ensemble" to the model coordinate
variable?
I'd suggest lower case, like most CF keywords. The other axis
attributes are
all single letters (X Y Z T) but I don't think it would be a problem
to use
a word.
Best wishes
Jonathan
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
CF-metadata@cgd.ucar.edu
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata