Re: representing multiple nests

Ramesh_S_Mantri wrote:

> Dear All,
>   as some might be aware, quite often models produce data over a 
> geographical region in multiple nests, typically with resolution
> increasing from outer to inner nests.
>
> COAMPS represents nests and their geometry in a data header file.
> I was told that GRIB-1 uses standard tables and pointers to represent
> nests and that GRIB-2 has an improved scheme (I'll need to look into
> it later).
>
> I have 2 questions:
> 1) Is there any standarad (or accepted) mechanism for representing
>    nests in a NetCDF file? I have looked at CF conventions and it appears
>    to me that it is possible to represent multiple fields (as separate
>    variables), but there doesn't appear a way to represent multiple
>    nests. I would appreciate if you could suggest a solution.
>
> 2) Does your GRIB-1 decoder write out nesting information in NetCDF
>     in any manner?
>
> This issue is important to us since consumers of data sometimes are 
> interested in nesting information. Thanks in advance.
> sincerely,
> Ramesh
>  
Im afraid i dont know any details of GRIB nesting.

We are planning to implement Balaji's CF proposal when its ready. You 
might want monitor the CF email list to track progress there.

here is the reply from the CF list:

Dear John

There is no explicitly stated way to represent nested grids. Of course, any
grid, however constructed, can be recorded by regarding its coordinates as
1D lists, with no knowledge of the structure of the grid. This should always
be done anyway. The issue is how to record the "higher-level" information
about its construction. It's similar to the issue of relating u and T grids.

Perhaps Balaji's proposal at the GO-ESSP meeting would help with this too. I
believe he intends to propose it to CF.

Best wishes

Jonathan