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

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 27 Apr 2007 17:56:59 +0100

Dear Jon

I think this was the kind of instance that was suggested before. To deal with
it, I tend to think that it would be better to have extra information *outside*
the netCDF file. This may be an unpopular view, but my reason is that the file
is already adequately self-describing. The relationship between vector
components is a "scientific" relation, which is of the same kind (so it
seems to me) as the relationship between (for instance) TOA net longwave
and net shortwave radiation, or between rainfall flux and rainfall amount.
For particular purposes, programs need to know "scientific" relations of this
kind, but we can't possibly encode all such relationships in the netCDF file.
That is not metadata, but instructions on what you can do with the data.

My proposal therefore would be that we maintain another table in a machineable
form listing the components of vectors (for your case). We can compile this
table now from the standard name table, and add to it as we add to the standard
name table. Possibly some ingenious system of flags in the standard name table
itself might avoid the need for a separate table.

Some relationships can be deduced, given the systematic construction
of standard names. If you open your file, read in a list of all the standard
names which occur in it, and do something like this to the list:
s/northward/NE/
s/eastward/NE/
then identical results will occur in pairs for vectors expressed as northward
and eastward components. But other cases can't be dealt with like this, so
a table is needed in general.

Best wishes

Jonathan
Received on Fri Apr 27 2007 - 10:56:59 BST