[MITgcm-devel] Re: thsice+exf test

[Martin Losch]

Jean-Michel,

I found the culprits that cause differences between useSeaice=T and F.

The first is the usual one: EXF. There are two routines called  
exf_filter_rl.F and exf_filter_rs.F which basically apply a the  
landmask, but only if useSEAICE=F (because of the B-Grid for the  
traditional seaice). Since I use the C-Grid version of seaice I can  
comment the if-statements out.

The second problem is within seaice_dynamics.F/seaice_ocean_stress.F

I will fix the second problem as agreed on the phone and introduce a  
stress reduction factor that is 1 by default (1=reduction,0=no  
reduction). I'll leave the first one as it is, so that you'll have to  
fix it for yourself locally. Eventually, when we will have abandoned  
the B-grid in pkg/seaice in the future we can think about fixing  
exf_filter_rl/rs.F

Martin

On Jun 13, 2006, at 5:41 PM, Jean-Michel Campin wrote:

> Hi Martin,
>
> Sorry, I arrived too late to call you, what about trying
> tomorrow (morning for me ?) ?
>
> It's not a very good practice to mixed the e-mails answer,
> as I am going now, but the thing that I did not find very
> elegant in the construction of seaice_calc_rhs.F,
> is that it's has some arguments (xA,yA) that are not passed to
> SEAICE_ADVECTION ; and the only real thing that is done is
> to add the diffusion tendencies to the advective one.
> And I thought it would be clearer (and also more flexible,
> specially at this stage of testing => see below) to replace
> 1 call to SEAICE_calc_rhs by 2 calls,
> one to SEAICE_ADVECTION and the other to SEAICE_DIFFUSION.
> So that those 2 S/R would be at the same level.
>
> An example of "flexibility": right now, we advect the product
> hEff*Qice1,2 ; I don't think we would like to diffuse the
> same thing, but rather Qice1,2 . It's not yet completely
> clear to me, but this is something that I might be interested
> to test.
>
> Now I go back to those tests.
> I tried also with SEAICEuseDynamics=T and
> 1rst.O upwind advection scheme. I run into this precision problem
> and change the advection routines gad_dst2u1_adv_?.F.
> But it's still blowing up at iteration 19 with no "early sign"
> of something getting wrong.
> It would help if we can identify what the seaice is doing
> when SEAICEuseDynamics=F (sorry to insist).
>
> Cheers,
> Jean-Michel
>
> On Tue, Jun 13, 2006 at 03:13:05PM +0200, Martin Losch wrote:
>> Hi Jean-Michel,
>>
>> that sounds great.
>> However, "BBerr - Bug: IceT(1) > Tmlt" is an error, isn't it. Do you
>> see where it comes from?
>>
>> Do the results look similar at all?
>>
>> Martin
>>
>> PS. I don't know if you get my emails at all. I have been sending
>> mails to many people outside AWI today, but no replys so far. Could
>> you send me short email to tell me that you got this, thanks.
>>
>> On Jun 11, 2006, at 12:06 AM, Jean-Michel Campin wrote:
>>
>>> Hi Martin,
>>>
>>> I run your c32 set-up (with thsice+exf), for 1 year,
>>> a) with seaice & SEAICEuseDynamics=F
>>> b) without seaice (useSEAICE=F)
>>> and get 9 "BB" warnings (all of this type:  BBerr - Bug: IceT(1) >
>>> Tmlt)
>>> in the 1rst case (a), and only 2 (with Tice1 not above Tmlt1 
>>> +8.e-3 oC,
>>> and this might not be too bad) in the second case (b).
>>>
>>> I would be tempted to conclude that, for now,  without  
>>> seaiceDynamics,
>>> it's working (apart from those differences between useSEAICE=T/=F,
>>> that
>>> I did not investigate yet).
>>> What do you think ?
>>>
>>> Cheers,
>>> Jean-Michel
>>