[CF-metadata] Conventions for Unstructured Grid Data

From: Jonathan Gregory <j.m.gregory>
Date: Wed, 10 Nov 2004 14:14:58 +0000

Dear Rich cc Steve, Balaji

CF has rudimentary support for unstructured grids in section 5 of the standard.
First, if a horizontal grid is not latitude and longitude, auxiliary coordinate
variables of lat and lon should be provided. That is so that any program can
at least know how to locate the data, although it will not understand how the
data points are arranged in space. Second, there is a mechanism in section 5.6
for describing the grid when it is a projection of some kind. If the software
implements this section, it allows 1D coordinates to be related to lat and lon.
This could be extended for arrangements of points that aren't familiar map
projections, where formulae or algorithms can be given for computing points
in lat and lon from coordinates or indices of the grid.

Balaji (at GFDL) is concerned with this issue in connection with ESMF. They
intend to make proposals. We discussed it recently, in particular the issue of
whether other "metrics" of the grid, such as information about how points are
connected, could be described by CF conventions. There is a method in section
7.2, which could be extended, for providing some further info of this kind.
Another possibility is that metrics of the grid could be supplied as data
variables, identified by particular standard names.

Best wishes

Jonathan
Received on Wed Nov 10 2004 - 07:14:58 GMT