[CF-metadata] size-one axes

From: Jonathan Gregory <j.m.gregory>
Date: Tue, 22 Jul 2003 23:25:21 +0100

Dear Brian and Ag

I had not thought of allowing the size-one coordinate variables to be scalars.
I agree, this is even simpler. We may need a new term for them as they would
no longer be coordinate variables in the Unidata sense. We could call them
"scalar coordinate variables".

If COARDS compliance is not thought to be an issue, I am quite happy to do
this for xyzt as well. I agree, this is the most common case, and we want to
encourage data-writers to supply the information. Variables are much better
than attributes, I think, because they can themselves have essential
attributes such as units and bounds.

I agree with Brian that the existing coordinates attribute is OK to use for
this; I don't think we need a special one.

The example becomes:

dimensions:
  lon=96;
  lat=73;
  forecast_time=3;
variables:
  float LWflux(forecast_time,lat,lon);
    LWflux:coordinates="wavelength analysis_time";
  float lon(lon);
  float lat(lat);
  double forecast_time(forecast_time);
  double analysis_time;
  float wavelength;

Does that look good?

Jonathan
Received on Tue Jul 22 2003 - 16:25:21 BST