[CF-metadata] Help needed with area_type and "surface type classification" datasets

From: Jonathan Gregory <j.m.gregory>
Date: Tue, 27 Sep 2011 15:05:15 +0100

Dear Seth

> Attached below are the 93 unique values for surface types from
> these 3 models. If it looks like we could hash out the details
> and extend the existing area_types list could be extended to
> cover this list in fairly short order, then I think we can say
> that the current system will be sufficient. If it's a bigger
> job than that, then I don't think it's adequate.

I don't suppose we could do this quickly, as I expect you anticipate. The
reason why it would take time is because we would have to establish whether
any of these terms actually meant the same thing. So this raises the question
of what the intended use of the metadata is. There could be two aims:

* describe the contents of each dataset by itself.

* indicate which quantities in one dataset are to be regarded as comparable
with quantities in another dataset.

The first aim is easier. It can be met by defining an attribute which is
allowed to contain some values from a vocabulary, which could be a vocabulary
external to CF. Provided the values in the vocabulary are self-explanatory,
the data is self-describing. However, if there are several datasets using
different vocabularies, as you describe, this approach doesn't address the
second aim, which is the one that is usually more difficult. It's the second
aim which is particularly important to CF; that's what standard names are for.
It can't be done without understanding what all the terms in each vocabulary
mean, and that takes effort!

Of course, we could allow area_type to have non-standardised contents, or use
a different attribute for this purpose. That postpones the effort until the
time when you want to do it, so it serves the purpose of data archival more
than data intercomparison, I suppose.

Best wishes

Jonathan
Received on Tue Sep 27 2011 - 08:05:15 BST