[MITgcm-devel] Re: [MITgcm-cvs] MITgcm/pkg/gchem CVS Commit

[Martin Losch]

Hi Steph,

On Dec 12, 2005, at 8:08 PM, Stephanie Dutkiewicz wrote:

> Update of /u/gcmpack/MITgcm/pkg/gchem
> In directory forge:/tmp/cvs-serv24517/pkg/gchem
>
> Modified Files:
> 	gchem_forcing_sep.F
> Log Message:
> o set imin,imax,jmin,jmax to include overlap points
why is it necessary to include the overlap regions? I think we had this 
discussion long ago, and I could convince you that they weren't 
necessary.
As far as I can see there are no horizontal averages or gradients 
involved in any of the bgc-models that I know of (I don't know too 
many, I have to admit), and right after gchem_forcing_sep, we call 
do_blocking_exchanges, so the calculating the bgc-fluxes/tendencies on 
the overlaps is just a waste of computer time (not a big waste, but 
unecessary, and these bgc-models can become expensive, at least mine 
is). I would prefer a solution, in which the overlaps, should they 
really be needed, are hardwired in the corresponding subroutines (as 
you did, e.g. in dic/carbon_chem.F and dic/fe_chem.F, although I don't 
see why: I couldn't find any instances of i-1,i+1,j-1,j+1 or the like 
in the directory dic), or included in model-specfic ifdef's.

What have I overlooked?

Martin