On Mar 16, 2010, at 5:30 PM, John Caron wrote:
Hi Jennifer, in this instance, how do you know its an ensemble
dimension, as opposed to say, a wavelength ? is it
ens:long_name = "ensemble member"
Grads knows it's an ensemble by the attribute:
ens:grads_dim = "e" ;
I could just as easily look for the axis="E" attribute, since I also
look at the axis attribute to find X, Y, Z, or T.
If the E dimension were used for wavelength, then it would have a
different long name but the same grads_dim='e' attribute.
--Jennifer
??
On 3/16/2010 12:36 PM, Jennifer Adams wrote:
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
[email protected]
# 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
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
--
Jennifer M. Adams
IGES/COLA
4041 Powder Mill Road, Suite 302
Calverton, MD 20705
[email protected]
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
