[MITgcm-devel] pkg/icefront

[Martin Losch]

Hi Dimitris,

this sounds all very fancy for a "coarse" grid (dx = O(10km)), but  
obviously I have not given this much thought. Go ahead ...

M.

On Dec 14, 2009, at 10:59 PM, Dimitris Menemenlis wrote:

> Hi Martin, the vertical faces are needed for Greenland glaciers,  
> which often terminate and calve at the grounding line, i.e., that  
> have no ice shelf cavity beneath them.
>
> My preference is for a separate package because
>
> (i) rather than search 2D space for ice shelf cavities, the icefront  
> package needs to search a 3D space for the ice fronts,
>
> (ii) probably best to save position of the ice fronts in vectors  
> during initialization in the same way as is done in pkg/downslope,
>
> (iii)  each grid box can have up to three separate contributions  
> from ice fronts while ice shelf cavities only have one contribution  
> from the top,
>
> (iv) shelfice_thermodynamics.F can be rewritten using the downslope  
> formulation (and also shelfice_u_drag.F and shelfice_v_drag.F I  
> think) but they would be radically different and harder to read than  
> they are right now,
>
> (v) eventually the physics of the two packages may diverge to  
> include different processes, for example, a fancier boundary layer  
> scheme for shelfice and a calving scheme for icefront, and
>
> (vi) it's less coding effort to separate the two packages.
>
> I will proceed with a separate package.  OK?
>
> D.
>
> On Dec 9, 2009, at 11:42 PM, Martin Losch wrote:
>
>> Without having thought about this too much I think you should extend
>> pkg/shelfice.
>> But: I only neglected the horizontal faces, because it was simpler  
>> and
>> nobody so far cared about them: dx = 5km and larger, dz = 10m-500m ->
>> area of ice front/area of ice base = 1/10 at maximum (and even this
>> only hypothetically at very large depths, because normally the ice
>> shelves extend to 1500m, where the vertical resolution is better than
>> 500m in ECCO/ECCO2 and any other configuration that I can think of),
>> more like 1/100 - 1/500. Is it really worth it? The entire
>> parameterization is so shaky that add another 1% of uncertainty
>> doesn't really seem to matter, does it?
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel