[MITgcm-devel] more seaice

[Jinlun Zhang]

Hi Martin,

Sorry for the delayed reply for your emails. Cool you have been able to 
run the version with new indexing and other modifications.

As for the calculation of surface stress, as Dimitris mentioned, the one 
implemented right now is a temporary fix. As a temporary fix, I prefer 
the existing treatment. I have been trying to find a permanent solution 
that would get real stress at the ice-ocean interface into the system, 
but it is tough to do.  I don't know how AWI ice models calculate 
stress. I remember one AWI paper published some years back says that 
some spatial average is needed in calc'ing stress, otherwise, the code 
would blow up. If it is true, then it is the same problem we have here, 
but I would not want to use spatial average. By the way, would you mind 
checking AWI ice models to see if they use the LSR solver or use the  
Hibler (1979) solver? I sent out LSR model (rectangular version) to some 
European groups more than 10 years ago, I hope it did not lost in 
translation (-:. And I hope they can use a parallel LSR, if not already 
using some other parallel code.

As for the mask in the LSR solver, better not use AREA=0 as mask. This 
is because it would create velocity discontinuities between areas with 
and without ice.

I haven't got deep into the new LSR routine, since you have run the code 
for many years, it is probably ok except that the metric terms have not 
been included, as you mentioned.

Cheers, Jinlun



Martin Losch wrote:

> Hello Dimitris,
>
> FORCEX/Y at that point are air-ice-stress interactions, the  
> coefficient ist call DAIRN. The ice-ocean stresses are put in later  
> after the computation ETA,ZETA,Eij with the coefficent DWATN. So I  
> think that at this point averaging does not make much sense: It would  
> mean that the ice is driven both by wind over ice and wind over open  
> water.
>
> I think in the c-grid version I'll try to do it the way I find it  
> intuitive: separate wind stress over ocean and ice and compute stress  
> on ocean as the average of ice-ocean stress and wind stress (on the  
> ocean). This is how it is done in the various coupled AWI ice models.
>
> Martin
>
> On Mar 6, 2006, at 7:54 PM, Dimitris Menemenlis wrote:
>
>>> The way I understand this is that the stress on the ice is an  
>>> average over ocean stress and ice stress ( the statement before  
>>> FORCEX(I,J,bi,bj) = ...), whereas the stress over the ocean is  just 
>>> the ocean stress not weighted at all. I find this a little  
>>> inconsistent if not wrong. I would think that these
>>>  terms should be treated separately, with only the ice stress  
>>> driving the
>>> ice, right?
>>
>>
>> Martin, yes it is inconsistent.  As Jinlun mentioned in an earlier  
>> message, the
>> original formulation for ocean stress, the one marked by CPP flag
>> SEAICE_ORIGINAL_BAD_ICE_STRESS in ostres.F caused model  
>> instabilities.  As a
>> temporary fix, the presence of ice is ignored in the computation of  
>> ocean
>> surface stress, variables WINDX and WINDY.
>>
>> Regarding ocean stress at bottom of ice, variables FORCEX and  
>> FORCEY, my understanding is that dynsolver assumes that thin ice  
>> covers the open ocean everywhere, hence the weighted sum of ice- 
>> covered and ice-free components in the computation of FORCEX and  
>> FORCEY.  But I do not know whether and why this thin-ice assumption  
>> is required nor what would be impact of setting the URT/VRT mask to  
>> zero where AREA=0.
>>
>> Dimitris
>