[CF-metadata] Getting back to ensembles

From: Pamment, JA <J.A.Pamment>
Date: Wed, 13 Dec 2006 03:25:37 -0000

Hello Jamie, Paco, Jonathan and Roy,

Thank you all for your recent comments. Firstly I would like to make a
general point about the purpose of the standard name status reports. They
are intended to achieve two things: (1) to keep everyone up to date with the
state of play on the various standard names threads without having to
follow every discussion in detail; (2) to identify the unresolved issues
and prompt further discussion. Writing the summaries helps me to keep things
straight in my own mind and I hope they are useful to others, if only to
reassure people that their proposals have not been forgotten, especially when
progress is slow. Preparing the summaries is, of course, a subjective process
and sometimes I will get things wrong - I am always happy for people to point
out when I have made a mistake or misrepresented something.

My summary of the ensembles discussion concluded with:
> Given the direction that the discussion has taken since the initial
> proposals were made on 15th October 2006 I will now close "ensembles" as a
> standard names issue. The names will _not_ be added to the table.
> However, ensembles will most definitely remain open as a CF1.0 conventions
> issue.
This seems to be causing some concern. Before writing this I reread the whole
of the ensembles discussion and I believed this to be the point that had been
reached in the conversation. Since Bryan's posting of 27th October there has
been no further discussion of the specific standard name proposals. Rather,
the question has been whether or not standard names is the correct mechanism
for labelling metadata.

My overall impression was that people felt that being able to specify an
external vocabulary or standard metadata would be a useful addition to the CF
standard. There seemed to be general agreement that it could
accommodate ensemble metadata and also other types of metadata that may be
needed in the future. Rereading again today I realise that there are a couple
of points raised by Jonathan that should have been included in my summary but
weren't. They are:
(1) if we adopt Bryan's proposal then we need a way of deciding whether new
metadata added within CF should be standard names or standard metadata;
(2) when using external vocabularies we would need to be sure of the format and
content - these may need to be agreed with the vocabulary owner.
Regarding (1), Bryan has suggested that standard names should be used to
describe the content of variables that relate to the physical world while
standard_metadata would describe how the data were produced (i.e., details of a
model or observing system). Jonathan (23rd October) has suggested that it is
not always straightforward to decide whether something should be a standard name
or standard metadata, so it would be better not introduce the standard_metadata
vocabulary.

The upshot of this is that there are actually three alternative approaches on
the table, rather than the two detailed in my summary:
(1) use standard names for ensemble metadata;
(2) use standard names to describe physical variables, for metadata use
external vocabularies where possible and introduce a new CF vocabulary called
standard_metadata for situations where a suitable external vocabulary does not
exist;
(3) use external vocabularies for metadata where possible and add new standard
names where a suitable external vocabulary does not exist.

Clearly there needs to be a decision on which approach to take before any
specific proposals can be taken forward. On balance, it seems I was premature
in declaring the issue completely closed as far as standard names are
concerned. However, I would certainly say that the standard name proposals are
"on hold" pending a decision on strategy.

It remains my impression that external vocabularies are generally agreed to be
a useful tool, albeit one that should be used with care. For ensembles in
particular, my understanding is that no suitable external vocabulary exists, so
in this instance we will need to add the metadata within CF. All
of which leaves one question: do we want to introduce a new attribute called
standard_metadata, with an associated vocabulary, or do we want to add metadata
names as standard names? I would still argue that this is really a conventions
issue at the moment, rather than a standard names issue, and that it deserves
as wide a discussion as possible.

Best wishes,
Alison

------
Alison Pamment Tel: +44 1235 778065
NCAS/British Atmospheric Data Centre Fax: +44 1235 445858
Rutherford Appleton Laboratory Email: J.A.Pamment at rl.ac.uk
Chilton, Didcot, OX11 0QX, U.K.
Received on Tue Dec 12 2006 - 20:25:37 GMT