[CF-metadata] Question about Coordinate System

[Jonathan Gregory]

Dear Phil

> non-geodesist interpretation of a geodetic datum is that it defines the
> mapping of a Coordinate System (such as latitude-longitude) onto a
> particular figure of the earth within a given reference frame, thus
> providing a Coordinate Reference System.

That sounds similar to my interpretation (that it defines an ellipsoid).

> Although a spherical earth approximation may be appropriate for many
> climate applications, I guess there may be plenty of situations where
> meteorological datasets are actually referenced to a non-spherical
> earth, e.g. met observations referenced against the popular WGS  1984
> datum/ellipsoid. In such cases I would have thought that the geodetic
> datum information ought indeed to be captured.

Yes, if the dataset had one, but I'm sure that most lat-lon meteorological
datasets (the ones I've come across, anyway) do not specify their datum.
They may have been gridded from obs, or they may provide obs on whatever
lat-lon system the country concerned uses. That implies a kind of mixture, I
suppose, which is OK because the differences don't matter for the purposes for
which the dataset is intended. But I really don't know about this. I'm sure
someone on the conventions committee must be an expert! For climate model
output the datum is genuinely undefined since the world is idealised.

> As regards the content and structure of a potential 'geodetic_datum'
> grid mapping attribute, would a solution be to adopt the well-known text
> format devised by the Petrotechnical Open Standards Consortium (POSC)
> and since adopted by the Open Geospatial Consortium (OGC). In Backus-
> Naur Form this looks thus:
> 
> <datum> ::=
>     DATUM [ "<name>", <spheroid>
>     {, <shift-x>, <shift-y>, <shift-z>, <rot-x>, <rot-y>, <scale-
> adjust> }
>     ]
> 
> <spheroid> ::=
>     SPHEROID ["<name>", <semi-major-axis>, <inverse-flattening>]
> 
> And so the WGS 1984 datum example might then look something like:
> 
> DATUM [ "World Geodetic System 1984" SPHEROID ["WGS 1984", 6378137.00
> 298.257223563 ] ]

This looks promising. To make it more self-describing and easier to use I
suggest we might put those elements of the definition in separate attributes.
Is it really that simple though? I would have expected more numbers: perhaps
that's what the <shift>s and <rot>s are.

My ignorance of this is all too obvious. To proceed we need expertise and
use-cases.

Cheers

Jonathan