[CF-metadata] bounds

From: Brian Eaton <eaton>
Date: Wed, 25 Jun 2003 23:19:58 -0600

Hi Jonathan,

On Fri, Jun 20, 2003 at 04:40:36PM +0100, Jonathan Gregory wrote:
> Dear Brian
>
> Well, I think we do almost agree! Not quite, though.
>
> > trying to classify grids by an "underlying x-y
> > arrangement". We do not resort to such language in the current CF document
> > when describing the use of 2-D lat/lon arrays for grid coordinates.
>
> Yes, I agree. However, what I meant by that phrase is implied what you have
> said at the start of (3) i.e. there are 2D lat and lon variables and the cells
> have four sides.
>
> I agree with what you have said except the requirement in (3) for them to
> be contiguous. The test you give indicates whether they are contiguous. It's
> well defined how to test it in this case. There might still be 4-sided cells
> in 2D arrays and not be contiguous if there was a row or column "missing".
> This is just an extension of the 1D case and it makes the contiguousness test
> useful, because *some* of the rows and columns may be contiguous, others not.
>
> Consider this situation: suppose you did have a 2D grid completely covering the
> plane and then you removed one row from the middle of the rows. Your statement
> would imply you'd have to rearrange all the bounds from (3) to (4) just because
> of that. I think it is simpler and better to leave them as (3) and just drop the
> single row from the bounds. The test will detect where a row is missing.

My main concern about allowing non-contiguous cells is with describing how
to order the vertices. In the contiguous case, the order is completely
described by the equalities outlined previously. If we allow this
representation for non-contiguous cells, then we must allow for the most
general case in which none of the cells are contiguous. In that case, how
do you propose to specify how the vertices are indexed?

>
> > (3) In the case where the horizontal grid is described by two-dimensional
> > coordinate variables lat(N,M) and lon(N,M), and the associated cells
> > are 4 sided, then the boundary variables are given in
> > the form latbnd(N,M,2,2) and lonbnd(N,M,2,2) rather than in variables
> > that use only a single dimension for the cell vertices. The contiguity
> > of the cells implies the following equalities:
>
> > For 0 < j < N and 0 < i < M
>
> I would state the tests as follows, which seems a bit simpler to me
>
> If cells (j,i) and (j,i+1) are contiguous, then
>
> bnd(j,i,0,1)=bnd(j,i+1,0,0) and bnd(j,i,1,1)=bnd(j,i+1,1,0)
>
> If cells (j,i) and (j+1,i) are contiguous, then
>
> bnd(j,i,1,0)=bnd(j+1,i,0,0) and bnd(j,i,1,1)=bnd(j+1,i,0,1)
>
> > where bnd stands for either latbnd or lonbnd.

This is fine.

Brian
Received on Wed Jun 25 2003 - 23:19:58 BST