[CF-metadata] Question from NODC about interplay of standard name modifiers, cell_methods, etc.

From: Jonathan Gregory <j.m.gregory>
Date: Tue, 26 Mar 2013 08:12:59 +0000

Dear Steve

> What you have just described is the state of CF standard_name today.
> But elements of it remain problematic. What do you see as the
> *purpose* of CF standard_names? What are the use cases by which
> standard_names will lead to better interoperability?

These remarks sound as though standard names were an innovation! In fact, as
you know, the distinction between standard names, cell methods and coordinates
is as old as CF itself. We have been following this approach, and expanding
the standard name table, for the last 13 years. There surely must be many cases
where the adoption of standard names has led to better interoperability, or
else people wouldn't continue to request them. One example is the hundreds of
analysis projects based on CMIP3 and CMIP5 AOGCMs, leading to hundreds of
papers cited in the IPCC reports.

Specifically, the reasons to separate cell methods from standard names were
(a) to make them more precise, so that it is easy to say which dimension the
method applies to and to allow combinations e.g. the temporal standard
deviation of a zonal mean, (b) to limit profileration of standard names.

> One use case that we are already seeing is that standard_name
> attributes are harvested from CF files and used in search engines.
> If a variable named as sea_water_temperature in a file is really
> time_variance_of_sea_water_temperature, because of its cell_methods,
> then we need to be pro-active in making sure that the search engines
> harvest the information found in cell_methods, too. This level of
> sophistication has not yet emerged into the ISO metadata world. I'm
> concerned that if we invoke sophisticated ISO machinery(*) to solve
> the problem, we may have a solution in name only -- will this level
> of complexity have a hope of becoming widely enough used to bring
> interoperability?

I don't think the burden is entirely on "us", the definers of the CF standard,
nor on the providers of data. People developing software which seeks to index
data should interest themselves in how the metadata is defined, shouldn't they.
Maybe we could clarify the statements in the preamble to the CF convention
which point out the essential metadata, in case they look there.

The need for a unified summary of the essential metadata, for discovery and
related purposes, is what motivated the common concepts trac ticket, which
was never finalised because it is a complicated issue. The more recent ticket
94 on CF string syntax looks like it could fulfil this role. That doesn't aim
to replace standard name etc. but to provide an additional facility.

Cheers

Jonathan
Received on Tue Mar 26 2013 - 02:12:59 GMT