[MITgcm-support] funky ice dynamics in doubly periodic domain

[Martin Losch]

Hi Jean,

if you use a spatially homogeneous forcing fields (the same value for all grid points) that just varies in time, I am not surprised, especially if the wind blows only along the channel (is that the case?). In summer the ice is mobile because AREA<1. Already at AREA<.90 there is hardly any resistence in the ice with default paramerers because of the  exp-function term in the strength parameterisation. As AREA->1 ice strength increases and it will be very hard to deform ice by convergent motion. With a solid ice cover and no divergence in the wind field I don’t see any reason, why the ice should move at all, unless you use free slip boundary conditions and there is some mean drift effect (btw, 10cm/s or even 5cm/s is not very small in that case). 
If you want to see something “interesting” in winter you need to introduce some sort of divergence into the ice motion, e.g. by introducing divergent wind forcing, or add some obstacle in the channel (see verification/offline_exf_seaice for an example). But still, with 100% ice cover and fairly thick ice there is going to be very little ice motion. That’s the whole point of the set of parameterisations, that we call a sea ice model.

Martin
PS. here’s a master thesis about sea ice in an idealised domain and some idealised forcing. Nils had to work a little to get the ice into motion.
http://epic.awi.de/38870/

> On 23 Feb 2016, at 07:58, Jean Mensa <jean.mensa at yale.edu> wrote:
> 
> Hi Jean, how big is your domain, and which ECMWF wind product are you using?
> 
> domain is 200x400km and I use the era-interim product averaged in space at each given time. This is to avoid discontinuity at the boundaries (does that answer your second question?)
>  
> And what does your effective sea ice thickness look like spatially as a function of time,
> especially at boundaries of your domain where you are bound to have some surface
> stress divergence or convergence?  Do you get any cracks?
> 
> ice thickness looks smooth. My impressions is that the lack of lateral constrains prevents the ice from cracking.
>  
> The reason I ask is because maybe what you are seeing is exactly what you should expect
> to see.  If there is not sufficient divergent or convergent wind stress on the ice
> 
> there is none!
>  
> it could move as a monolithic slab as it thickens.  That is the sea ice cover is mostly driven by internal
> rather than surface stress, there is sufficient mass to smooth out wind speed fluctuations,
> and it integrates surface stress over space time.
> 
> glad to see that there might be a physical explanation... I am running a simulation from thinner ice to see if it keeps better correlation with the wind forcing,
>  
> One way to test this hypothesis is to reduce ice strength, say:
>       SEAICE_strength    = 1  _d +04
> to see if the onset of the “steady ice speed” period happens a little later.
> 
> I will try that. Thank you for the suggestion!
>  
> Another would be to close boundaries at edges of your domain then your ice
> should be immobilized (maybe?).
> 
> I would like the sea ice to be moving. I was thinking of prescribing ice velocity at the boundaries as if ice was driven only by wind stresses (mDu/Dt = tau_air) but I am afraid that the model will respond generating artificial ice stresses...
> 
> thank you for your comments,
> Jean
> 
> 
> Dimitris Menemenlis
> 
> > On Feb 22, 2016, at 10:22 PM, Jean Mensa <jean.mensa at yale.edu> wrote:
> >
> > Hi all,
> > I have a doubly periodic high resolution simulation forced with ECMWF atmospheric forcing (forcings.png) running the seaice pkg with VP rheology. Thermodynamics variables look good (ice.png) but ice velocity does something funky and I am not sure why. During summer season ice thickness is minimum (H_I) and concentration low (A_I) and ice dynamics seems to be driven by wind stress. This is all good and fine except that during winter ice velocity (S_I) seems to converge to some constant value decoupling from wind stress. This does not seem right to me. Is that an artifact of the doubly periodic configuration? Notice that the simulation runs fine and gives no failed convergence warnings.
> > Thanks,
> > Jean
> 
> 
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> https://urldefense.proofpoint.com/v2/url?u=http-3A__mitgcm.org_mailman_listinfo_mitgcm-2Dsupport&d=AwIGaQ&c=-dg2m7zWuuDZ0MUcV7Sdqw&r=5oFkuOCa8D2MEaB0wBLw8HmmtUZ0uoNntNFgQ9b_kVE&m=VVWlWZ2WJkeKLM7h2oVXrDmOL11dniPP3-KdLDSKhvM&s=tjnNlJ3365Gfwo-iy-c43P1M29gbQiiwWkUwbiGx4ZY&e=
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support