[CF-metadata] curvilinear cartesian coordinates case

From: Jonathan Gregory <j.m.gregory>
Date: Fri, 1 Dec 2006 18:35:31 +0000

Dear John and Brian

Oh dear, I am going to disagree again. Yes, I can see there is an idea of a
grid which could exist separately from a data variable. However, I don't think
it really makes sense to point to the projection coordinates from the
grid_mapping variable, because there could be several sets of projection
coordinates (e.g. velocity and temperature grids) that used the same
grid_mapping variable, because they shared the same map projection. I think
that the ideas of "grid" (a particular set of coordinates) and "mapping" (a
transformation between one space and another) are distinct. The grid_mapping
variable presently deals with the mapping. The grid is a property of the
data variable, because it is the only place the particular combination of
coordinate variables is specified that constitute the grid.

Brian wrote
> I routinely create files that contain only grid data. A typical use of
> these files is as input to an application that does regridding. Such an
> application may, for example, take an input datafile at some resolution, an
> input grid file at a different resolution, and produce an output datafile
> at the resolution contained in the grid file.

I can see that is useful, and that you want a way to specify a grid, which
means that particular combination of coordinate variables. A CF-compliant
way to do it would be to have a file with a single data variable in it,
but presumably you want to avoid that because the data variable would be a
waste of space as it is of no interest. Is that right?

Cheers

Jonathan
Received on Fri Dec 01 2006 - 11:35:31 GMT