[MITgcm-devel] new package?

[Martin Losch]

Hi again,

sure enough I found a fundamental problem with my new (not-yet-public) 
package "shelfice":
In principle, I tried to avoid pressure gradient errors by staying 
stricltly in z-coordinates (no rstar either). I added shelf-ice 
topography in the same way as it is done for bathymetry (partial 
cells), so that I suddenly have "land" (dry) cells above "ocean" (wet) 
cells. The critical point (I thought) was to introduce the 
geo-potential anomaly phiHyd at the bottom of the shelf-ice so that 
there are no horizontal gradients of phiHyd introduced. That's 
straightforward in the same way we did this for the oceanic pressure 
coordinates, and as then I read this constant field (in time) from a 
file (pLoadFile) (with a linear EOS I can choose my tRef and sRef so 
that the phiHyd contribution is alway zero). I arrive at a model that 
will not move, if initially there are no horizontal density gradients 
(and no forcing). So far so good and we have a big advantage over sigma 
models (or the rstar-option which I tried a year ago without much 
success because of the pressure gradient errors).

Now the problem: I added shelf-ice ocean interactions (drag and 
thermodynamics). For a start, I use simple thermodynamics as in ISOMIP 
(http://fish.cims.nyu.edu/project_oisi/isomip/overview.html), more 
complicated stuff may follow. The idea is that the thermodynamics 
change the density *below* the ice-shelf. Consider an ocean of constant 
temperature and salinity. If the shelf-ice topography intersects a 
layer, say goes from k-1 to k, then you'll have in the first time step 
a change in temperature/density in the grid cells (i,j,k-1) and 
(i,j+1,k), for example. This will lead to a gradient in phiHyd: 
phiHyd(i,j+1,k)-phiHyd(i,j,k) .ne. 0 (phiHyd(i,j+1,k-1)-phiHyd(i,j,k-1) 
is non-zero also, but that's masked properly.). This gradient is 
"singular" and will lead to a singular velocity v(i,j+1,k) before the 
solve_for_pressure call. Next, cg2d_b is computed with this velocity to 
give a field with the corresponding spikes and consequently, eta(n+1) 
will have some spikes like these as well. In the correction step, the 
signal is then added as a barotropic signal, so that the v-velocity 
looks "noisy" (basically there are jumps whereever topography 
intersects layers). In the next timestep phiHyd will also have these 
structures and so on. However, phiHydtot will aways be smooth as the 
spikes in phiHyd and gravity*eta cancel exactly. What I am describing 
is the effect of a combination of lack of resolution (although this 
would only reduce the problem, not get rid off it), and the pressure 
correction method, as far as I can see (I hope I described it well 
enough for you to understand). I have integrations of 3months so far, 
but these noisy structures do not go away, because they are constantly 
forced by the shelfice-ocean thermodynamics.

In summary: no horizontal pressure gradient error (very good!), but 
noise introduced at topography-layer intersections by limited 
resolution and the pressure correction method. Any idea, what could fix 
this?

Martin

On Jan 23, 2006, at 8:40 AM, Martin Losch wrote:

> Hi Jean-Michel,
> the code with simple thermodynamics ("shelf-ice-ocean interaction") is 
> ready, which doesn't mean, that it is free of problems (e.g, I have 
> not had a close look at the effects of shelfice on vertical mixing, 
> but I assume that I need to do something about that, too). But I would 
> like to t check it in anyway, since it requires more than one hook, 
> where there routines are called from, and that would be difficult to 
> maintain up-to date, if it's not in the main code (also I would like 
> you to give me some advice on how to do things better (o:).
> Do you still agree to add the package or should I rather put it on the 
> development branch?
>
> Martin
>
> On Jan 20, 2006, at 7:38 PM, Jean-Michel Campin wrote:
>
>> Hi Martin,
>>
>> I think it's a good time (from my point of view) to
>> add this new package.
>> And good luck with the Ice-Shelf !
>> Jean-Michel
>>
>> On Fri, Jan 20, 2006 at 09:24:32AM +0100, Martin Losch wrote:
>>> Hi there,
>>>
>>> I have started implementing shelf-ice caverns (remember that I work 
>>> at
>>> a polar research institute :o). The main issue lie in the treatment 
>>> of
>>> pressure at the bottom of the shelf-ice, but I think I have sorted 
>>> that
>>> out. I have to admit that so far, I have no idea how well it will 
>>> work
>>> in a large scale or even global configuration with coarse resolution,
>>> but my litte idealized test look promising.
>>> If everything goes well, I would like to add the code as a new 
>>> package,
>>> called "shelfice", to the repository, if nobody objects.
>>>
>>> Martin
>>>
>>> _______________________________________________
>>> MITgcm-devel mailing list
>>> MITgcm-devel at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel