[MITgcm-devel] seaice growth

[Martin Losch]

Hi Matt,

I am glad, that LSOR was not the culprit. I was so happy, when it finally became stable after so many attempts and a new instability would have worried me a lot.

I usually use advection scheme 33, because it has less numerical diffusion than 77.

Here, we discussed some of the issues <http://mitgcm.org/pipermail/mitgcm-devel/2011-May/004769.html>
I agree that this default requires the user to do something, which is a bit unfortunate. Maybe the warning should also go to STDERR, so that it is not overlooked?

Martin


On Aug 23, 2011, at 5:07 AM, Matthew Mazloff wrote:

> Hello,
> 
> Thanks for the help.  Turns out large HEFFs were largely just a red herring.  The problem seems to be primarily that I was using default DIFF1, and this changed from 0.004 to 0 along the line.  I finally payed attention to the STDERR warning, and added
>      DIFF1             = .004,
> in data.seaice.  My stability issues are now resolved!  But I have questions:
> 
> As you can see I am using the default advection scheme,  SEAICEadvScheme    = 2, is this not recommended?  Perhaps I should just change my advection scheme to 33 or 77 -- I am not using seaice dynamics in the adjoint run.
> 
> Regardless, I'm not sure it is the best idea to have default parameters (SEAICEadvScheme    = 2, with DIFF1             = 0.0) that are not stable....perhaps the warning should be switched to a stop.
> 
> Thanks
> Matt
> 
> 
> 
> 
> 
> On Aug 18, 2011, at 12:52 AM, Martin Losch wrote:
> 
>> Matt,
>> 
>> what is "extreme"? I have runs where LSOR "handles" 30m of ice. Theoretically, the ice strength must not become "nearly infinite", because then the convergence of the iterative solver is bad. This is achieved by limiting the ice strength/viscosity (not HEFF) and this is described even in the documentation (o:
>> The runtime variable for \Delta^*=SEAICE_zetaMaxFac=2.5e8, however, that your can modify, is not described in the documentation, probably, because I forgot it or I did not want it to be changed?
>> 
>> I agree with Ian: Hack the code in seaice_dynsolver.F and leave seaice_growth.F alone.
>> 
>> I am a little surprised that the LSOR solver crashes. Since revision 1.39 it has been remarkably stable.
>> 
>> Martin
>> 
>> On Jul 28, 2011, at 1:50 AM, Matthew Mazloff wrote:
>> 
>>> Hi Jean-Michel,
>>> 
>>> Ooops, sorry, missed it at the top there!
>>> 
>>> OK, so I see now that #define SEAICE_CAP_HEFF in SEAICE_OPTIONS did nothing because capping HEFF is only allowed in LEGACY code.
>>> 
>>> Are there any "hacks" that I can use in the new sea ice code to keep the model well conditioned and running.  I would like to diagnose how the ice is piling up, and not having it crash as soon as it starts to pile up would be nice
>>> 
>>> Thanks,
>>> Matt
>>> 
>>> 
>>> 
>>> 
>>> On Jul 27, 2011, at 9:23 AM, Jean-Michel Campin wrote:
>>> 
>>>> Hi Matt,
>>>> 
>>>> I don't know which version of seaice_growth.F you are looking at,
>>>> since:
>>>>> #include "SEAICE_OPTIONS.h" ?
>>>> has been there from the beginning (seaice_growth.F version 1.1)
>>>> 
>>>> Cheers,
>>>> Jean-Michel
>>>> 
>>>> On Wed, Jul 27, 2011 at 09:16:27AM -0700, Matthew Mazloff wrote:
>>>>> Hello,
>>>>> 
>>>>> Shouldn't seaice_growth.F include the header file:
>>>>> 
>>>>> #include "SEAICE_OPTIONS.h" ?
>>>>> 
>>>>> Thanks
>>>>> Matt
>>>>> 
>>>>> 
>>>>> ps> This comes up because I am having issues of too much ice
>>>>> accumulation in the Southern Ocean with the latest seaice package.
>>>>> It may very well be due to forcing or ocean state, not a seaice
>>>>> model problem.  When the thickness gets too extreme, however, the
>>>>> LSR solver crashes.  So I can make a little progress diagnosing
>>>>> this, I want to set
>>>>> #define SEAICE_CAP_HEFF
>>>>> and felt it should go in
>>>>> SEAICE_OPTIONS.h
>>>>> but realized that it cannot go there...thus the question
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> 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
>> 
>> 
>> _______________________________________________
>> 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