[MITgcm-devel] seaice

[Martin Losch]

The story continues:

I ran my configuration with (always with exf and exf_bulkformulae)
1. full seaice model
2. without seaice model
3. with seaice model but unmodified stresses
4. with seaice model but with unmodified buoyancy fluxes (store the  
fluxes at the beginning of seaice_model and restore them at the end  
of seaice_model)
    a. all unmodified
    b. only EmPmR unmodified
    c. only qnet and qsw unmodified (blows up!!!)

Whenever EmPmR is unmodified the circulation is "bad". That is, the  
ACC breaks into the Weddell Sea. In particular runs 2, 4a and  4b  
look reasonable (of course the ice is terrible in a few cases but I  
don't care about that now), but run 1 and 3 are bad. Unfortunately  
the run 4c explodes very quickly and I don't know why.

I compared the sfluxtave between run 1 and run 4b (empmr unmodified  
by ice model). The icemodel adds a freshwater flux to the model that  
can be 10 to 20 larger than the atmospheric freshwater flux (in the  
yearly average!!), especially along the boundary of the Weddell Sea.  
This makes the surface waters along the boundaries much saltier in  
run1 (with full seaice) than in run 4b (everything but EmPmR  
modified). As a consquence run4b shows a strong salinity gradient  
from the coast (fresh) to the center of the Weddell Sea, and a  
Weddell Gyre develops, whereas in run1 (with full seaice) this  
gradient is very weak. Comparison with levitus surface salinity does  
not help, because in the mean there the structure is a different (and  
I don't know to what extend I can believe that).

Is it reasonable to have yearly average salt fluxes (sflux) on the  
order of -0.002 to 0.008 (mainly due to the ice model, instead of  
atmospheric fluxes of -0.002 to 0.0006), divide by s*rhoFresh =  
(35*1000) to get freshwater fluxes? May there be a problem in the  
seaice?

Martin