[MITgcm-devel] other problem in recent check-in

[Martin Losch]

Hi,
I have been playing with the option, and found this:
1. the global_with_exf results have been produced with fldstartdate 
(s) = 19920101. This makes it impractical to check the transfer from  
one to the next year, because the first record used is the first of  
1992. If fldstartdate was 19920115, you'd need the last record of the  
previous year (1991) to compute the forcing at Jan01 and this would  
be the nice test that I had in mind.
2. The model calendar is gregorian, requiring a longer fldperiod than  
2592000: 365*86400/12 = 2628000 for normal years and  
366*86400/12=2635200 for leap years.

So I see three options:
1. change the model calendar to "model", but then the "gregorian"  
calendar will not be used. Is it tested elsewhere? Yes, in lab_sea,  
for example
2. increase the fldperiods to 2628000 or 2635200 to be on the safe side.
3. forget about testing this yearly fields stuff

option 1 and 2 require updating the test output, option 1 would  
global_with_exf exactly the same as global_ocean.90x40x15 except for  
the capping of wind stress in pkg/exf (which I could also remove).  
Option 3 saves me a lot of work (o:

What do you think.

Martin

On 28 Jan 2008, at 10:40, Martin Losch wrote:

> We could use an experiment, that checks both EXF_INTERPOLATION (as  
> in global_with_exf) and the use of yearly fields. Would it be  
> possible to modify global_with_exf to do so?
> e.g. have a script link/cp the climatology fields
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002 lev_sss.bin
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002 lev_sst.bin
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002 ncep_emp.bin
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002  
> ncep_qnet.bin
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002  
> trenberth_taux.bin
>   -rw-r--r--    1 mlosch   staff      172800 Nov 12  2002  
> trenberth_tauy.bin
> to trenberth_taux.bin_1991 and trenberth_taux.bin_1992, etc, modify  
> data.exf to have the forcing records start in 1991 instead of 1992.  
> That shouldn't change the results of the test and still test the  
> yearlyfields flag for exf (not for obcs).
>
> In case you agree with this, how do I do this linking/copying  
> without increasing the number of files (MBs) in the repository?
>
> Martin
>
>
> On 28 Jan 2008, at 09:20, Martin Losch wrote:
>
>> Again, sorry for that. I really wonder, why it compiled for me  
>> (and still does), although myiter etc were clearly not defined.  
>> (SUPER-computers are fun and reeeaally reliable). D, thanks for  
>> cleaning up after me.
>>
>> Martin
>>
>> On 27 Jan 2008, at 18:15, Jean-Michel Campin wrote:
>>
>>> Hi,
>>>
>>> Seems that global_with_exf does not compile any-more since
>>> Friday pm. Compilation error in exf_set_uv.F
>>>
>>> Jean-Michel
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel