[MITgcm-support] Re : Water at the bottom of NA below freezing point

[Jeff Scott]

Hi Dror,

I've had this problem (i.e. cold, deep temperatures below the seaice 
freezing point), to varying degrees (no pun intended), in several 
different global coarse-resolution configurations. Jean-Michel and I have 
numerous discussions about it, and we are fairly sure it is simply the 
usual suspect: the advective overshoot-undershoot problem. That being 
said, it is not easy to come up with a good working solution.

The most recent global configurations I've worked with were 4x4x15 layers, 
4x4x22, and 2.5x2x22 (the middle being more of a "test" setup). All were 
lat-lon, so I can't say how this would be affected by the cubed sphere 
(but I think the problem would still exist). All three configurations have 
the problem, but it seems the cold water occurs in only a few places or 
even just one place, which differ between setups. In the 4x4x15 and 
2.5x2x22, the runs are still "usable" in that the cold water remains 
fairly localized. In the 4x4x22, I gave up because the cold water was 
occurring near strong flow in the Drake Passage and filling the whole deep 
ocean. In the other configurations, the SH was all fine but the problem 
was in "overflow" region of the NA, near the bottom of bumpy topography. 
In one test I actually tried changing the topography to "fill in" the cold 
gridpoint in one of these runs, but the cold point simply moved higher in 
the water column, thwarting my efforts. (incidently, using partial cells 
makes the problem considerably worse).

I've also tried different advection schemes (besides 2). 33 and especially 
77 are not usable, as they radically affect the model's temperature 
structure, being rather diffuse. I've also tried the Prather/SOM scheme; 
indeed, this pretty much fixes the cold temp problem, but is at high CPU 
cost -- and even this scheme is noticeably diffusive. Since my work 
involves explicitly trying to minimize overall (numerical and 
parameterized) diffusion, I gave up on this solution too.

Jean-Michel and I have discussed some other avenues to perhaps address 
this problem, but I haven't had time to pursue it... Thus I've learned to 
live with local temps even below -4C.

(I think the scary thing however is that this problem is not limited to 
cold water; such advective undershoot/overshoot is occuring in other 
places, you just don't notice it so easily. But for example, look to see 
if you are getting convection occurring below the surface in the eastern 
tropical Pacific - this seems suspect...)

Anyone else have some insight into this problem and/or possible solutions?

Jeff


On Thu, 6 Mar 2008, Dror Mirzayof wrote:

> Hi Dimitris,
> and thanks for your response
>
> The answers for your questions are :
>
> 1) At the surface we do not find any temperature below freezing point (~-1.9 
> deg) in our output.
> The output is of 100 years, in time resolution of 1 year. The forcing is not 
> seasonal but constant with time.
>
> 2) The experiments we mentioned in our questions, were initiated from a 
> spinup run,
> in which we corrected all temperatures to be above the freezing point.
>
> Dror
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support
>