[MITgcm-devel] bounds in ctrl_map_init_ecco.F not cover Arctic

[Martin Losch]

Hi An,

if these values need frequent adjustments, they should become runtime parameters, shouldn’t they?

Alternatively, your ctrl_map_ini_ecco.F could be moved into your code directory (or into the one of the tests that are failing).

I personally stumbled over these bounds in the past, and I think that by default, there shouldn’t be any bounds (i.e. boundsVec=0) (and individual experiiments can have individual ecco_map_ini_ecco.F’s, or introduce runtime parameters, I am repeating myself), but then I wonder, what *are* good defaults for ECCO? BTW, there are BFGS optimization algorithms with bounds (http://www.ece.northwestern.edu/~nocedal/lbfgsb.html), that would make hard-coding them here unnecessary (also cleaner from an optimization point of view)

Martin

On Apr 30, 2014, at 2:39 AM, An T Nguyen <antnguyen13 at gmail.com> wrote:

> hello,
> 
> The current lower bounds for 2 control fields, salt0 and background diffusivity, are too high for the Arctic region.  I'd like to change them as follows:
> 
> old:
> #ifdef ALLOW_SALT0_CONTROL
> c--   Salinity field.
>       boundsVec(1)=29.
>       boundsVec(2)=29.5
> 
> #ifdef ALLOW_DIFFKR_CONTROL
>       boundsVec(1)=1. _d -6
>       boundsVec(2)=2. _d -6
> 
> 
> new:
> #ifdef ALLOW_SALT0_CONTROL
> c--   Salinity field.
>       boundsVec(1)=18.
>       boundsVec(2)=20.0
> 
> #ifdef ALLOW_DIFFKR_CONTROL
>       boundsVec(1)=1. _d -7
>       boundsVec(2)=3. _d -7
> 
> However, I think the changes will affect some results on current tests... So I'm putting it here for discussion and advise on how to proceed.
> Thanks, 
> An
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel