[CF-metadata] How to include EPSG codes or WKT information in a CF file [SEC=UNCLASSIFIED]

From: Jonathan Gregory <j.m.gregory>
Date: Wed, 1 Aug 2012 09:42:35 +0100

Dear Paul, Etienne et al.

> From my perspective I treat an EPSG code as a unit definition (in this
> case a coordinate) as I would treat 'metres' as a unit definition for
> length. CF conventions do not require the description of what a metre
> implies; that is assumed knowledge, and I believe the EPSG code should
> be treated likewise. If we make it easy to define the CRS, we will see
> far more users actually specify one!

I see the analogy, but it's not a perfect one, because units are a much
simpler business. There is only a handful of basic units, and the translations
of all the derived units are easily accessible, well-known and completely
dependable. In particular, there is readily available software for parsing
and converting units.

So, while I agree with your order of preference

> Option 1: Long form : full parameter attribution (as per your ticket)
> Option 2: Short form : EPSG code
> Option 3: Compatibility form : wkt representation

I think that Option 1 is much better than Option 2. Option 2 would be worth
considering if there was a online service, or a machine-readable file available
from a public repository, that CF software could use to translate EPSG codes
into and out of parameters that could be equivalenced to CF grid_mapping
attributes. As far as I understand (but I'm ready to be corrected) this is not
the case. If it were the case, it would be possible to check automatically that
the EPSG code and the grid_mapping in a CF file were consistent. Without such
a check, I think it is dangerous to allow codes, because it is inevitable that
some people will put codes in files without making sure that the grid_mapping
is consistent with the code, and then data-analysts will get different results
according to what information their software uses. This would obviously be a
failure of the purpose of CF to "define metadata that provides a definitive
description of ... the spatial and temporal properties of the data."

Best wishes

Jonathan
Received on Wed Aug 01 2012 - 02:42:35 BST