[MITgcm-support] Adding constant submerged and surface ice.

[Jean-Michel Campin]

Hi Roger,

Just a comment regarding the 1rst point:
> 1) ... with a time period less than the simulation time step, 

The implementation of relaxation in MITgcm (surface relaxation
without exf: tauThetaClimRelax,tauSaltClimRelax or with pkg/exf:
climsstTauRelax,climsssTauRelax or 3-D relaxation with pkg/rbcs)
is fully explicit (forward in time); the stability criteria
for fully explicit relaxation is something like:
  deltaT/TauRelax < 2 (without Adams-bashforth, i.e., forcing_In_AB=.False.)
  deltaT/TauRelax < 1 (with AB-2) and much less with AB-3
but to ensure a non-oscilatory response (strict-stability), the criteria
is twice more strict:
  deltaT/TauRelax < 1   (without AB)
  deltaT/TauRelax < 1/2 (with AB-2)

Using a relation time-scale (TauRelax) slightly smaller than the time
step might not make the model to blow up (other parts such as
diffusion, and specially implicit vertical diffusion, have a stabilisation
effect) but will generate some noise in the time-evolution of the tracer
(like oscillations).

The implementation of an implicit (backward in time) relaxation
has been on the list of to-do things but will not be available very soon.

Cheers,
Jean-Michel

On Wed, Feb 12, 2014 at 04:55:46AM +0000, Roger Malvin Jr wrote:
> Hi Everyone
> 
> I'm modeling a body of water which has a constant cover of ice, with a known and substantial depth, and one one boundary of the domain there is a glacial wall which we can also assume is constant in temperature and size.
> 
> Having completed the tutorials, I've made a simulation of my own and the package is mostly  working, with topology and initial temperature conditions included. 
> 
> I have two issues:
> 1) I am using a climate relaxation file to just reset the surface temperature to zero with a time period less than the simulation time step, as a brute force method of creating a permanent ice cover and 0 degree surface boundary. However, when I do this (compared to when I just let the surface rest at ambient temperature, usually it settles to about 0.5 C) there is a lot of noise, particularly around the edges of my topology where the depth is small (as little as 2 meters compared to the dz = 5m element size I am using). Even in areas where it is deeper, down to about 20 m, around the edges there is still this issue. When run the simulation without the forced 0 C surface temperature the velocities and temperature profiles at and near the surface are much smoother, and don't provide somewhat un-physical situations like a checkerboard of positive then negative velocities next to the edges seen with the relaxation file forcing the boundary.
> 
> 2) I have not been able to find any information on how I can implement the glacial wall in the simulation. I know how to intialise the temperatures along the wall to be zero, but I need something like the climate relaxation file to set these elements back to zero every time step, but even that seems like it would produce unreliable results. Does anyone know if any of the sea ice packages, or anything else for that matter, allows us to set temperature boundary conditions away from the surface?
> 
> Basically, if someone knows how to produce a box with one Ice side and ice surface in MITGCM that would take me most of the way there.
> 
> Thank you in advance for any assistance,
> Harry
>  		 	   		  

> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support