[MITgcm-devel] thsice_get_exf

[Jean-Michel Campin]

Hi Martin,

I am in favor of this "#include  ....h" solution, but
would like to make a precision:
- Isn't  the main difference between the 2 calls sitting in
  the i,j loop (inside the S/R in one case, outside in the other) ?
- I don't think there is so much more things to compute for
  thsice_get_exf.F (and we can perfectly don't return the
  unused output in the exf_bulkformulae.F). But it seem
  important to me that the compute the derivatives
  regarding surf.temp (at least for the evap) since it
  might depend on the saturated pressure form (yes,
  I have seen formulae that are different ...).

See you,

Jean-Michel

On Thu, Jun 08, 2006 at 12:00:16PM -0400, Martin Losch wrote:
> Hi Jean-Michel/Patrick,
> 
> I had another look at thsice_get_exf and compared to  
> bulkf_formula_lay it is a mess, even after Jean-Michel has cleaned up  
> all of the obvious mistakes (such as factors of 1000).
> 
> I would like to have a routine that is called both from  
> thsice_solve4temp and exf_bulkformulae (so a pendant to  
> bulkf_formula_lay). That would make it much easier to switch between  
> different bulk formulae, Large and Pond, vs. Large and Yeager, or the  
> NCEP bulkformulae, whatever they are. I think it is probably a fact  
> overlooked by many "uninitiated oceanographers" (such as myself),  
> that the bulkformulae for computing fluxes have to match the ones  
> used in the atmospheric models that produced the atmospheric fields.
> 
> The problem is that exf_bulkformulae does not need all of the stuff  
> that's needed in thsice_get_exf.F
> Also, for the adjoint it is probably problematic to call subroutines  
> for every gridpoint, so maybe it is possible to have code snippet  
> that is included into these routines from a "header"-type file, such  
> as #include "exf_bulkformulae_core.h"
> or
> #include "exf_bulkformulae_ncep.h"
> The extra stuff that is required for thsice_get_exf.F can be computed  
> outside of this snippet.
> 
> What do you think about this?
> 
> Martin
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel