[CF-metadata] vector quantities (e.g. velocities) in CF-NetCDF

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 27 Apr 2007 18:22:03 +0100

Dear Bryan

> Surely knowing about vectors
> is something we care about?

Yes, it is. You suggest encoding in the file that quantities are components
of wind, but my argument is that there are *many* relevant relations between
quantities that *could* be encoded in a file, and we cannot possibly do that.
So I suggest not putting this relation in the file, that's all. It is not
metadata because you don't need it to identify the quantity. We know what an
eastward wind means from the standard name. The fact that an eastward wind is
a component of wind is implicit in the definition of eastward wind. The
definition of eastward wind does not reside in the netCDF file, but in the
standard name table and in our heads. That's why I think the standard name
table (or another table) is the right place to store machine-readable
information about what are the components of vectors, not the netCDF file.

A practical reason is that if one relies on the file itself pointing out
the components of vectors, any system such as Jon's will only work with files
that are equipped with this information. If we have an external table which
says this, Jon's system will work with any file. External tables are not a
bad thing per se. For instance, we also rely on the udunits definition file
to be able to interpret and convert units. That table tells us the relation-
ship between seconds and minutes, and the udunits software tells us the
relationship between m and km. We argue against use of external tables only as
a substitute for self-explanatory metadata, but eastward_wind is perfectly
self-explanatory.

Cheers

Jonathan
Received on Fri Apr 27 2007 - 11:22:03 BST