[MITgcm-devel] depressed Eta and sIceLoad

[Martin Losch]

Dear all,

What should we/I do about pkg/seaice/growth.F
The current status does not seem acceptable to me. I could check in  
my version of growth.F which contains
1. masking evap with (1-area(:,:,2)); I consider that a bug fix and  
that should definitely be included. It will get rid of many problems  
with too much convection under ice.
2. Jinlun's fix for snow, I guess that should be hard coded and not  
put within #ifdef's. This fix gets rid of these enormous snow  
accumulations without upsetting the fresh water flux, but it does not  
get rid of all the snow over water (1e-3m seem to remain a little of  
the ice edge).
3. the little flooding algorithm (also in thsice), within #ifdef's  
and a run time flag, off by default. This  fix does not get rid off  
all the snow either, but it is appealing to convert snow into ice,  
once it is below the water line.

Further I have implemented advection of snow in pkg/seaice/ 
seaice_advdiff.F. I feel that this should also be part of the code,  
but it's probably not required for solving the problems that started  
this discussion thread. Still I would include it hard wired.

All of this will change the lab_sea experiment and all other seaice  
experiments. Since part of the changes (but the part with the most  
impact) are due to the bug-fix with evap, none of the previous  
results can be reproduced anyway, so maybe it's okay to just go ahead  
and add all (or some) of the above things without caring too much  
about backward-compatibility/reproducibility?
Please let me know what you think and I'll go ahead and do it.

Martin

On 25 Oct 2006, at 10:25, Martin Losch wrote:

> Hi Jinlun et al.,
>
> please have a quick look at 10day averages of HEFF and HSNOW in  
> July and January (after 100 years) in
> http://mitgcm.org/~mlosch/ice_iter72360.png (Jul)
> http://mitgcm.org/~mlosch/ice_iter72020.png (Jan)
> run22 is without advection of snow (both runs are with evap*(1- 
> area)), run26 is with advection of snow. Unfortunately, the runs  
> are not exactly comparable because I had to use a 1-order upwind  
> scheme for advection the snow, so that run26 uses the 1-order  
> upwind scheme (for HSNOW, HEFF and AREA) and run22 the traditional  
> 2nd order central differences scheme (but only for HEFF and AREA),  
> no "flooding" in either case. What you see is that the advection  
> reduces the amplitudes of the snow be a factor of 100, but maybe  
> that's just the 1-order upwind scheme? Also the snow is in areas  
> where it shouldn't be.
> Not shown: My (it's not mine, but as opposed to Jinlun's  
> suggestion) simple flooding scheme removes this snow at the  
> "expense" of much increased ice thicknesses, Jinlun's scheme  
> reduces the snow heights even further (I only have 10year yet, but  
> it's already <20cm as opposed to <50cm in "my" runs), and the ice  
> thickness is not increased as much. Also the freshwater flux into  
> the ocean appears to be reduced (higher surface salinities) so that  
> as a first conclusion I would say that Jinlun's fix appears to be  
> quite appropriate (as opposed to mine). More analyses to follow.
>
> Martin
>
>
> On 24 Oct 2006, at 23:19, Jinlun Zhang wrote:
>
>> Hi Martin,
>> Advecting snow is certainly better than not, but I am not sure if  
>> it would solve the problem of getting big numbers for snow.
>> Jinlun
>>
>> mlosch at awi-bremerhaven.de wrote:
>>
>>> Hi Jinlun,
>>> thanks, I'll try this out tomorrow. What about advecting HSNOW?  
>>> Why can that be neglected?
>>>
>>> Martin
>>>
>>> Martin Losch
>>> Alfred Wegener Institute Postfach 120161, 27515 Bremerhaven,  
>>> Germany; Tel./Fax: ++49(0471)4831-1872/1797
>>>
>>>
>>>
>>> ----- Original Message -----
>>> From: Jinlun Zhang <zhang at apl.washington.edu>
>>> Date: Tuesday, October 24, 2006 7:56 pm
>>> Subject: Re: [MITgcm-devel] depressed Eta and sIceLoad
>>>
>>>
>>>> Dimitris and all,
>>>> The ncep precip in Antarctic is way too much, but I have  
>>>> moidfied growth.F  to, hopefully, improve things a little bit.  
>>>> The modification would allow the ocean to melt  the left-over  
>>>> snow when ice is gone (seach jz for the modi's in the code) .  
>>>> See if there is any improvement ot of it.
>>>> Jinlun
>>>>
>>>> Dimitris Menemenlis wrote:
>>>>
>>>>
>>>>> It looks bad:
>>>>> http://ecco2.jpl.nasa.gov/data2/cube/cube38/pickup/HSNOW.jpg
>>>>> 20 to 60 m of snow over open water.  This is after a 12-year
>>>>> integration on the cubed sphere.
>>>>>
>>>>> Dimitris