[CF-metadata] standard names for defined global attribute names

From: Jonathan Gregory <j.m.gregory>
Date: Thu, 29 Mar 2012 15:55:06 +0100

Dear Nan

> It seems that, to indicate various sources for different variables
> within a file,
> one would use institution as a variable attribute instead of a
> global. This is
> the approach that NODC has used in their new NetCDF templates.

That is already legal in CF. The source and institution are allowed as global
or variable attributes.

If source and institution become legal standard names, it means they can be
used as the values of string-valued scalar or auxiliary coordinate variables.
A scalar string coordinate variable is just a string (a 1D char array), so it
would be functionally equivalent to a variable attribute. However, it could
be an array of strings i.e. a 2D char variable, as you say:

> is this meant
> to be a 2D variable? It makes a lot of sense in that context.

You could use this if the data had a dimension which corresponded to a source
or institution e.g.

dimensions:
  ninst=10;
  lat=180;
  lon=360;
  stringlength=30;
variables:
  float ppn(ninst,lat,lon);
    ppn:standard_name="precipitation_flux";
    ppn:units="kg m-2 s-1";
    ppn:coordinates="institution";
  char institution(ninst,stringlength);
    institution:standard_name="institution";
data:
  institution="GFDL", "MPI", "CSIRO", ...;

Here, institution is an auxiliary coordinate variable.

As you say,

> this should
> either be dimension (1) or share a dimension with one of the coordinates?

This is described in sect 6.1 of the convention.

Best wishes

Jonathan
Received on Thu Mar 29 2012 - 08:55:06 BST