[CF-metadata] standard name proposals

From: Brian Eaton <eaton>
Date: Thu, 7 Aug 2003 10:01:49 -0600

Hi Jonathan and John,

On Wed, Aug 06, 2003 at 05:28:06PM +0100, Jonathan Gregory wrote:
> > If the packed data was in the 1024-1028
> > range, then having an add_offset of -1000 would make the unpacked data
> > technically not agree with the units?
>
> No, I don't think so. Most units don't have an implicit "origin". The units
> of kg m-3 are unchanged by subtracting 1000 kg m-3.
>
> > Maybe this would just be a good place
> > for the "comment" attribute?
>
> I think we do need to record the 1000 in a standardised attribute. add_offset
> seems the best choice to me. I wonder if Brian has a view. It is intended
> really for packing, but it is functionally what you want to do.

By convention the units associated with a variable are the units of the
unpacked data. Using an add_offset of -1000 with packed data values of
1024-1028 implies that the true values of density and potential density are
in the range 24-28 kg/m3, which is not correct. I think the best solution
is Jonathan's original suggestion that the "sigma" quantities require their
own standard names since they are actually derived from the true density
and potential density values.

Brian
Received on Thu Aug 07 2003 - 10:01:49 BST