[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [CF-metadata] cf futures (long)

Manuel,

> >   http://www.unidata.ucar.edu/packages/netcdf/examples/files.html
> > 
> I've noticed you have classified ECMWF ERA40 under COARDS convention,
> not CF. If this is intentional, I'd appreciate any comments you may
> have regarding how to make our data more CF compliant.

Sorry, I think it was just a mistake.  The file has the global
attribute

  :Conventions = "CF-1.0" ;

so it was clearly intended to comply with the CF conventions.  I fixed
it by moving the ERA40 sample file to be under the CF convention.
These tables of example files have not been announced or linked to
from our site, so I think no one has gotten the wrong impression except
perhaps for 2 ESRI programmers who requested some sample files.

I did run your file through the CF compliance checker at

  http://titania.badc.rl.ac.uk/cgi-bin/cf-checker.pl

just to see what it would show, and it complained about the units of
the form:

  tcc:units = "(0 - 1)" ;
  lcc:units = "(0 - 1)" ;
  mcc:units = "(0 - 1)" ;
  hcc:units = "(0 - 1)" ;

used for cloud cover.  There are a couple of other possibilities for
such units, but I think can see why you chose "(0 - 1)".  

To stop the CF compliance checker from complaining, you could use
"percent", but then the unpacked values would have to be between 0 or
100, rather than 0 or 1.  So "percent" is not ideal, because it's not
backwards compatible with what you have, unless you changed the
"scale_factor" attributes.

Another possibility would be to use the "flag" attributes described
here:

  http://www.cgd.ucar.edu/cms/eaton/cf-metadata/CF-1.0.html#flags

which would require omitting units and using three other attributes
instead of one, e.g. something like:

    tcc:valid_range = 0, 1 ;
    tcc:flag_values = 0, 1;
    tcc:flag_meanings = "none total" ;

if you really only ever used 0 or 1 as unscaled values.

For cases like this, udunits needs probably needs some enhancements.
By the way, we're planning to merge udunits into netCDF so people
won't have to build the packages separately as of the next release.

Anyway, thanks for pointing out the mistake in the example files, and
please let me know if you see any other problems or have other
exemplary example datasets to add to this list.

--Russ

_____________________________________________________________________

Russ Rew                                         UCAR Unidata Program
address@hidden          http://www.unidata.ucar.edu/staff/russ