[CF-metadata] scalar coordinate variables

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 24 Dec 2010 21:26:29 +0000

Dear Steve

I'm reply to your comment on scalar coord vars on the email list rather than
the trac site because I don't want to confuse the discussion in the ticket
about clarifying the standards document. Unlike you, I think they do offer
significant advantages.

* Creating an extra dimension (of size 1) feels like an overload of machinery
if the intention is simply to attach a single coordinate datum to a data
variable. This is quite a frequent need. The commonest examples are surface
air temperature, at a height of 1.5 m or 2.0 m, or windspeed at 10 m. We quite
often recommend, for instance in connection with particular standard names,
that the value of a certain other parameter could be specified, e.g. a
radiation_wavelength for radiative quantities. Scalar coord vars are a neat
way to do this. They are something between multivalued coord vars and
attributes in terms of function: easier than coord vars, and more powerful
than attributes because they can themselves have attributes.

* Apart from not having to create a dimension, using a scalar coord var means
avoiding having to decide where to insert this dimension in the order of the
dimensions of the data variable. Obviously it makes no difference to the data
at all, so this is often a meaningless and arbitrary decision which it is
useful not to have to make.

Best wishes

Jonathan
Received on Fri Dec 24 2010 - 14:26:29 GMT