[MITgcm-devel] some idea for pkg/shelfice

Wed Apr 10 10:15:33 EDT 2013

Hi Jean-Michel,

I am happy with this, maybe also good to get Patrick's opinion? There's also an OpenAD test now, isn't it?

Martin

On Apr 10, 2013, at 3:38 PM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:

> Hi Martin,
> 
>> I have started to do (A), and have 2 issues:
>> 1) it does change results, because of machine truncation
>> (conversion shelficeLoadAnomaly --> shelficeMass --> shelficeLoadAnomaly)
>>  can always put a test " IF ( SHELFICEloadAnomalyFile .EQ. ' ' ) THEN"
>>  but I would prefer not to, and update the isomip results.
> shoud I proceed with these changes and update the isomip output ?
> 
>> 2) I am getting many "TAF recomputation warnings" (10 more than before)
>> and not sure how to fix this.
> and then will leave this to TAF experts
> 
> Cheers,
> Jean-Michel
> 
> On Tue, Apr 09, 2013 at 11:32:49AM +0200, Martin Losch wrote:
>> Dan, you are right about this.
>> 
>> Jean-Michel, great and thanks for fixing this. If that was the only problem "erroneously related to the NLFS", then that sheds (again) a poor light on me. I probably should have found that problem years ago.
>> 
>> Martin
>> 
>> On Apr 8, 2013, at 11:59 PM, Daniel Goldberg <dngoldberg at gmail.com> wrote:
>> 
>>> OK. In general, one cannot use an AB update on the first timestep, right? Because the tendency from the previous timestep (or two timesteps) is needed?
>>> 
>>> Dan
>>> 
>>> 
>>> On Mon, Apr 8, 2013 at 5:46 PM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
>>> Hi Martin,
>>> 
>>> The issue with the CD-scheme is fixed, I think.
>>> I was doing a test with very large gradient of Eta (balanced by
>>> phi0surf) and the 1rst iteration Adams-Bashforth in CD scheme
>>> was not consistent with the rest of the momentum (no AB on 1rst iter).
>>> 
>>> Cheers,
>>> Jean-Michel
>>> 
>>> On Mon, Apr 08, 2013 at 09:49:08AM -0400, Jean-Michel Campin wrote:
>>>> Hi Martin,
>>>> 
>>>> I have started to do (A), and have 2 issues:
>>>> 1) it does change results, because of machine truncation
>>>> (conversion shelficeLoadAnomaly --> shelficeMass --> shelficeLoadAnomaly)
>>>>  can always put a test " IF ( SHELFICEloadAnomalyFile .EQ. ' ' ) THEN"
>>>>  but I would prefer not to, and update the isomip results.
>>>> 2) I am getting many "TAF recomputation warnings" (10 more than before)
>>>> and not sure how to fix this.
>>>> 
>>>> Regarding the other things, need to look into the details of realFreshWater
>>>> code, to see if it has a good chance to works when kSurf <> 1.
>>>> And will check also the CD-code with NonLin-FreeSurf (should not be too
>>>> difficult since it happens at the 1rst iteration), but will be easier once
>>>> I am able to specify the ice-shelf mass.
>>>> 
>>>> Cheers,
>>>> Jean-Michel
>>>> 
>>>> On Thu, Apr 04, 2013 at 04:24:33PM +0200, Martin Losch wrote:
>>>>> Hi Jean-Michel,
>>>>> 
>>>>> some useful suggestions!
>>>>> 
>>>>> I totally agree with A. Do you want to do that?
>>>>> 
>>>>> B: I found this
>>>>>> C     with "real fresh water flux" (affecting ETAN),
>>>>>> C     there is more to modify
>>>>>>      rFac = 1. _d 0
>>>>>>      IF ( SHELFICEconserve .AND. useRealFreshWaterFlux ) rFac = 0. _d 0
>>>>> in the code, but rFac is never used. Not sure what happened here. I think we postponed the real freshwater flux business here. If you know how to do that, I am all for it.
>>>>> 
>>>>> C: I use the CD scheme, because I needed to get all the smoothing I needed (have a look at section 3 of Losch (2008)). I am sure the CD-scheme can be replaced by biharmonic viscosity. I never got the shelfice package to work with the non-linear free surface. even without r*, but I cannot remember if I tried it only with CD turned on. It's probably better to try without.
>>>>> 
>>>>> The noise that "called for" the CD scheme is due to the step-wise topography (and partial cells). I always planned to try r* to get a (numerically) "smooth" surface, by depressing the surface layer according the SHELFICEloadAnomaly (or better your SHELFICE_Mass). That might solve many numerical problems, e.g. noise, dynamic coupling, and introduce new ones (e.g., layers that get too thin, pressure gradient errors). I just never got around to trying that out.
>>>>> 
>>>>> Martin
>>>>> 
>>>>> On Apr 3, 2013, at 3:47 AM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
>>>>> 
>>>>>> Hi Martin,
>>>>>> 
>>>>>> Some here have been using pkg/shelfice and, in order to make
>>>>>> more progress, there are few things that could be changed:
>>>>>> 
>>>>>> A) add a new variable for the mass of ice (per unit area):
>>>>>> I propose to do this, and to allow to specify (through a file)
>>>>>> the ice-shelf mass.
>>>>>> To start with, fully backward compatible with no result changes,
>>>>>> will allow either to specify:
>>>>>>  a) R_shelfice & SHELFICEloadAnomaly (same as now)
>>>>>> or b) R_shelfice & SHELFICE_Mass
>>>>>> and with the relation between the 2:
>>>>>> SHELFICE_Mass = SHELFICEloadAnomaly/gravity - Ro_surf*rhoConst
>>>>>> (since Ro_surf is negative).
>>>>>> At the end of shelfice_thermodynamics.F, will compute SHELFICEloadAnomaly,
>>>>>> and in case of (a), SHELFICEloadAnomaly from the file will be converted
>>>>>> (in shelfice_init_varia.F) to SHELFICE_Mass using the same relation.
>>>>>> Advantages:
>>>>>> 1) mass of ice is (much) more natural than SHELFICEloadAnomaly;
>>>>>> + it does not dependent on hFacMin truncation.
>>>>>> 2) should allow in the future to step forward SHELFICE_Mass.
>>>>>> 3) they are few places where SHELFICE_Mass can be used
>>>>>> (e.g., pressure @ the base of the ice-shelf = SHELFICE_Mass*gravity,
>>>>>> instead of R_shelfice as it is now,
>>>>>> also total thickness of the ice = SHELFICE_Mass/rho_ice ...)
>>>>>> but will let you decide what should be modified or not.
>>>>>> 
>>>>>> B) implement real-fresh-water flux with pkg/shelfice:
>>>>>> I think it would not be too hard to allow the surface forcing
>>>>>> to be applied at the surface level (kSurf) even if kSurf <> 1
>>>>>> This way, if we put the SHELFICE melting into PmE, we will get
>>>>>> the real-fresh-water formulation with little more modif.
>>>>>> 
>>>>>> C) current isomip experiment: found a problem when turning on
>>>>>> the non-lin free-surf (without z*), got some strange (and wrong)
>>>>>> currents because of the CD-scheme. Would need confirmation,
>>>>>> Likely due to a Pb in CD-scheme with non-lin free-surf and
>>>>>> large variations of hFac (+ kSurf <> 1 ?), will need to be confirmed.
>>>>>> Can you remember us why CD-scheme is used (since the resolution
>>>>>> is high enough, ~10.km, to run without) ?
>>>>>> 
>>>>>> Cheers,
>>>>>> 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
>>> 
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>> 
>>> 
>>> 
>>> -- 
>>> 
>>> Daniel Goldberg, PhD
>>> NSF Postdoctoral Fellow
>>> Earth, Atmosphere and Planetary Sciences, MIT
>>> Cambridge, MA 02139
>>> 
>>> em: dgoldber at mit.edu
>>> web: http://ocean.mit.edu/~dgoldberg
>>> _______________________________________________
>>> 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