[CF-metadata] non-standard standard_names

From: Jonathan Gregory <j.m.gregory>
Date: Thu, 13 May 2010 08:57:55 +0100

Dear Martin et al

Some quantities are so specific to a particular dataset or model that it would
not be worth the effect of defining a standard name for them, since they will
never be compared with data from another source, the main reason for standard
names being to indicate which quantities should be regarded as comparable. In
those cases, I think having your own private attribute such as htap_standard_
name is the best solution. Other projects have done that before; for instance
the aquaplanet model intercomparion project has APE_names. This is Steve's
solution (2). The attribute is not standard and the vocabulary is private; it
perfectly legal to have non-CF attributes in CF files.

If the project is itself comparing data from different sources, it would be
make sense to use standard names. As Roy says, CF standard names aren't purely
geophysical quantities any more. There are others related to measurements and
technical details. I think it's fine to have standard names for such things
in cases where different data sources are dealing with comparable quantities
and they have to work out how to identify them in a common way.

I would much prefer keeping to a single attribute and namespace. It's much
more convenient to look at one attribute than to search for many, as John
says. Fragmentation of the namespace can in principle be resolved by mappings,
but in practice that would not happen, I am sure and Roy also suggests, just
because it would be a lot of work, and if it's not done, that would seriously
undermine the usefulness of standard names. Part of the reason why agreeing
them takes so long is that we have to understand what they mean, in order to
be sure that the new ones are not duplicating existing ones, and that they are
constructed in a consistent way, which will make it easier for future users to
understand them too.

Therefore, like others, I prefer a fast-track approach, like John's suggestion
and Steve's (1), to deal with new standard names. I think we need a CF central
repository for proposed standard names. That would be useful anyway to monitor
their progress. I know that BADC have been working in this direction, although
I don't know the status. The repository should be kept up to date by modifying
it as proposed standard names are amended, accepted or rejected. We could say
that some reasonable interval after a standard name has been proposed, such as
a small number of weeks, and if it's still under discussion at that time
(i.e. hasn't been rejected or accepted yet), it would be allowed to use it in
the standard_name attribute, prefixed by "proposed" e.g.
  standard_name="proposed: mass_concentration_of_alcohol_in_beer".
That means, as others have said, that the record of proposed names and their
eventual fate would have to be maintained indefinitely, but this should be
easier to manage if there's just one such repository - a central CF one - than
many separate project websites.

With such a system, it worries me that some projects might propose standard
names and start to use them without making a serious effort to agree them as
standard. I think that effort is well worthwhile, in terms of clarifying both
the concepts and the vocabulary.

Cheers

Jonathan
Received on Thu May 13 2010 - 01:57:55 BST