Hi Heiko,
> One of my fimex-users complaint about fimex not being able to read
> CF files, although they looked very similar to the netcdf CF example
> file on the unidata webpage:
> http://www.unidata.ucar.edu/software/netcdf/examples/RUC.cdl At the
> end, my user was missing the grid_name attribute, but was using
> instead only the _CoordinateSystems = "projectionCoordSys"
> attribute.
Yes, some of those examples are relatively old (the RUC one was from
late 2004) and we haven't been maintaining or replacing them as
needed.
> According to the netcdf user guide, names starting with '_' should
> only be used for internal purposes. I think above file makes too
> much usage of such attributes, and is therefore a bad example. It
> confuses more than it helps. In my opinion, it would be nice to see
> this example removed from the web-page of CF-examples at
> http://www.unidata.ucar.edu/software/netcdf/examples/files.html
I have removed it, and would like to have a better replacement that
would exemplify best practices in the use of the CF Conventions. We
could create such a file, but it would be better if someone would
contribute real data that would also serve as a good example for how
CF can be used to capture meaning in model output or observational
data.
If someone has an example they would like to contribute or nominate,
please let us know at support-netcdf.unidata.ucar.edu. This isn't a
competition, so we may decide that multiple examples are worthy of
emulation and add them all to the published file examples.
Thanks for pointing out the problems with RUC.cdl and the associated
RUC.nc. We look forward to getting some good replacement candidates.
--Russ
"I never metadata I didn't like"
