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

[Martin Losch]

Hallo Dimitris et al.

this is kind of a bummer, but I just found (and corrected) a bug in  
exf_wind.F, somehow the exponents where evaluated as (1/3) (=0 in  
Fortran) instead of (1./3. = 0.33333333), so that the wind fields the  
exf_wind backed out from the wind stress where completely stupid for  
ustar .ge. ustofu11.

This bug (or rather the stupid wind fields) is the reason, why I  
changed the seaice model to use the stress fields from file, instead  
of the ones that are computed from wind fields that have been  
computed from stress fields. What should I do? Should I revert the  
seaice model back to its original state? Or should we stick to the  
current version? Both versions have their advantages/disadvantages.  
Further, this option of reading stress and not wind is a bit esoteric  
to my mind, but I need this for a model intercomparison project here  
at AWI and why not have the seaice model be able to use both kind of  
fields. Opinions?

Unfortunately the whole thing does NOT fix my problem with holes in  
the domain.
In my previous emails about this I talked about being able to  
reproduce the problem in a cs32 configuration, but that turned out to  
be connected to the way ustar was computed in exf_bulkformulae after  
Jean-Michels modifications. After fixing that I am now unable to  
reproduce the problem with holes in the domain. However I would  
expect that the new ecco-godae llc configuration, which has a hole in  
the south pole would have the problem, too, but apparently not, since  
nobody has complained since. So I guess I have to go back to the  
500CPU configuation and try to figure out what is really the problem.  
Oh well ...

Martin

On 8 Mar 2007, at 15:30, Dimitris Menemenlis wrote:

> Martin, thanks for explanation.  The main difference comes from  
> lower wind speed cutoff, which is
>
>>       parameter ( umin           = 0.5 _d 0
>
> in exf_constants.h versus
>
>>       SEAICE_EPS        = 1.      _d -10
>>       SEAICE_EPS_SQ     = -99999.
>>       IF ( SEAICE_EPS_SQ .EQ. -99999. )
>>      &     SEAICE_EPS_SQ = SEAICE_EPS * SEAICE_EPS
>
> in seaice_readparms.F
>
> I am not sure which is more appropriate or robust from adjont  
> perspective ofr the SEAICE_BUDGET computations?
>
>> In addition I would like exf to prepare all input files (including  
>> uwind an
>> vwind), in case they are need somewhere.
>
> OK.
>
> Do your latest changes fix the "hole" problem?  Or is that still  
> outstanding?
>
> D.
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel