[MITgcm-support] changes in seaice default behaviour?

Mon Feb 11 11:44:12 EST 2019

Hi Paul,

I don’t understand why the SEAICEscaleSurfStress parameter makes such a big difference. It should only make a difference with ice concenrations < 1 (really <<1). For AREA < 1 it basically reduces the interfacial stress between ice and ocean and ice and atmosphere. Why should that lead to more pile up of ice along the coast? Do you have AREA<1 there (i.e. the effect of offshore katabatic winds being reduced by that?)

SEAICEpressreplfac = 0. is different, because removing the replacement may not be undesirable physically. Replacement pressure was introduced to avoid unforced ice motion by a pressure gradient grad(P) in the absense of forcing. Fair enough, that avoid unphysical behavoir, but what does that also do? You replace P (in grad(P)) by P*Delta/(Delta+DeltaMin) (or P*Delta/max(Delta,DeltaMin), but that makes no difference). When ice is pushed against a coastline or into a “bay” (represented by a single grid cell, maybe), there will be no motion and Delta -> 0, but DeltaMin is constant at O(1e-9). So P->0 and there will be no resistance against compression, so that ice can pile up indefinitely. That’s also unphysical. So you have choice between two unphysical behaviors, which both stem from a regularisation/parameterization. You can probably decrease DeltaMin further to avoid either problem, but then it will get harder and harder so solve, … it’s a dilemma.

Martin

> On 11. Feb 2019, at 17:32, Holland, Paul R. <pahol at bas.ac.uk> wrote:
> 
> Hi again everyone
> 
> OK, I am pleased to report that with both SEAICEscaleSurfStress=.false. and SEAICEpressreplfac = 0., I am able to significantly reduce my problems with thick sea ice in coastline embayments.  Furthermore, the simulations appear to be much more stable than with diffusion turned on, and I am able to increase the timestep by a factor of 2.5.  So I am generally very happy with this! However, there are two issues.
> 
> Firstly, I am still getting ice that is gradually thickening in coastline bays, and so the problem has not gone away, it has just been reduced to the level where I can get away with it.  If I do longer runs or have a 'more sensitive' geometry or free surface formulation, I think I will still be in trouble.  I might have to then resort to using a small amount of ice thickness diffusion, unless a better solution arises.
> 
> Secondly, both the choices above seem undesirable in that they are disabling developments made for good physical reasons. In using them to remove my thick ice, this is really a 'trick' for which they were not intended. One could easily argue that using diffusion is also undesirable.  So what is the solution?
> 
> My areas of thick ice are in the vicinity of ice shelves but are not being caused by supercooled water.  So I think they are fundamentally a dynamical issue - the ice is continuing to pile up even though it is >10m thick.  I can't help thinking the rheology should be preventing this from happening....
> 
> All thoughts welcome, as ever,
> 
> Cheers
> 
> Paul
> 
>> Hi Folks
>> 
>> I can report back as well.  I did some test cases where I turn off ice diffusion
>> and turn on SEAICEpressreplfac = 0., with a few different choices for
>> SEAICE_deltaMin.  In the diffusive runs I get a maximum ice thickness of
>> about 6m and a nice seasonal cycle, but in all other cases I got ice thicknesses
>> steadily growing in coastline corners (e.g. reaching 15m after 10 years).
>> 
>> This is all with SEAICEscaleSurfStress=.true. so it seems that turning that off is
>> the next thing to try for me.  There is very little multi-year ice in my area of
>> interest.  Perhaps when ice concentration is low in the summer, having
>> SEAICEscaleSurfStress=.true. means there is not enough stress to flush the
>> thick ice out?  Let's find out....
>> 
>> I'll report back if I get a simulation to work without diffusion!  Thanks for your
>> help so far, any more comments gratefully received...
>> 
>> Cheers
>> 
>> Paul
>> 
>>> Date: Mon, 4 Feb 2019 13:46:20 +0100
>>> From: Martin Losch <Martin.Losch at awi.de>
>>> To: MITgcm Support <mitgcm-support at mitgcm.org>
>>> Subject: Re: [MITgcm-support] changes in seaice default behaviour?
>>> Message-ID: <FFD5CAE6-48FA-4C5F-8212-569BC0DAF743 at awi.de>
>>> Content-Type: text/plain; charset="utf-8"
>>> 
>>> Hallo Christoph,
>>> 
>>> thanks for the useful feedback.
>>> 
>>> I understand the drag parameter (more surface stress to move ice out of
>>> unpleasant corners). The scaleSurfStress parameter is strange, when True,
>> the
>>> surface stress terms are multiplied by ?AREA?, which they should according
>> to
>>> Connolley et al (2004), so this will reduce the surface stress only where
>>> AREA<1, which is probably not in the areas where you have your ice
>> thickness
>>> problems, oh well ?
>>> 
>>> BTW, there?s a SEAICE_drag_south parameter, which lets you set a
>> different
>>> value for the drag coefficient in the southern hemisphere than for the
>> northern
>>> hemisphere.
>>> 
>>> Martin
>>> 
>>>> On 4. Feb 2019, at 13:05, Christoph Voelker <christoph.voelker at awi.de>
>>> wrote:
>>>> 
>>>> Hi Martin and those interested in the sea-ice behaviour, I just wanted
>>>> to report back that playing with the defaults that Martin suggested
>> indeed
>>> solved my problem.
>>>> I did one run, where I switched back the four new defaults,
>>>> 
>>>> SEAICEscaleSurfStress to .FALSE.
>>>> SEAICEaddSnowMass to .FALSE.
>>>> SEAICE_useMultDimSnow to .FALSE.
>>>> SEAICE_drag to 0.002
>>>> 
>>>> But I also tried to be less drastic and did a few runs changing only one of
>> them
>>> at a time.
>>>> The two parameters that had the strongest effect are
>>>> SEAICEscaleSurfStress and SEAICE_drag, the others also lowered the
>>> maximum seaice_heff (in monitor_seaice*.nc) somewhat.
>>>> 
>>>> Cheers, Christoph
>>>> 
>>>> Am 01.02.19 um 10:22 schrieb Martin Losch:
>>>>> Hi Paul,
>>>>> 
>>>>> it?s a shame that you need to use these diffusivities. For the original
>>> advection scheme (2nd order central difference), this was necessary, but
>>> shouldn?t not be required when you use one of the flux limited schemes.
>>>>> 
>>>>> I found that in long runs turning of the replacement pressure (as
>> Christoph
>>> does with SEAICEpressReplFac = 0.) usually avoids having too tick ice in
>>> individual corners along the coastline (See Kimmritz et al 2017 for an
>>> explanation). I am not saying that this is more physical than extra diffusion,
>> but
>>> is probably affects the solution less away from these critical points.
>>>>> 
>>>>> The diffusion has changed long ago. I don?t think that that is Christoph?s
>>> problem.
>>>>> 
>>>>> Martin
>>>>> 
>>>>> 
>>>>>> On 31. Jan 2019, at 20:02, Holland, Paul R. <pahol at bas.ac.uk>
>>>>>> wrote:
>>>>>> 
>>>>>> Hi Christoph
>>>>>> 
>>>>>> I have had trouble with excessively thick sea ice around coastlines also.
>>> The only thing I found that would touch it is ice diffusion, e.g.
>>>>>> 
>>>>>> SEAICEdiffKhHeff = 10.0,
>>>>>> SEAICEdiffKhSnow = 10.0,
>>>>>> SEAICEdiffKhArea = 10.0,
>>>>>> SEAICEdiffKhSalt = 10.0,
>>>>>> 
>>>>>> I don't think this is particularly physical, but it works for me!  I tried
>>> changing all the ice rheology parameters and none of them solved this
>> problem
>>> for me.  Be careful with the values of diffusion though - high values caused
>>> blow-ups and undesirable results for me.
>>>>>> 
>>>>>> If I understand it correctly, in the old original version of the sea ice
>> code
>>> there was a diffusivity somewhere deep in the code (parameter DIFF1),
>> which
>>> went away at some point, so I think this is just compensating for that
>> change.
>>> But I think this change may have happened before checkpoint 66f.
>>>>>> 
>>>>>> Others have found that ice shelf meltwater leads to very thick ice
>> forming
>>> as well, but that seems to be a much more localised problem.
>>>>>> 
>>>>>> Perhaps Martin can comment on the above?
>>>>>> 
>>>>>> Cheers
>>>>>> 
>>>>>> Paul
> 
> 
> 
> This email and any attachments are intended solely for the use of the named recipients. If you are not the intended recipient you must not use, disclose, copy or distribute this email or any of its attachments and should notify the sender immediately and delete this email from your system.
> UK Research and Innovation has taken every reasonable precaution to minimise risk of this email or any attachments containing viruses or malware but the recipient should carry out its own virus and malware checks before opening the attachments. UK Research and Innovation does not accept any liability for any losses or damages which the recipient may sustain due to presence of any viruses.
> Opinions, conclusions or other information in this message and attachments that are not related directly to UK Research and Innovation business are solely those of the author and do not represent the views of UK Research and Innovation.
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support