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

Martin Losch Martin.Losch at awi.de
Tue Mar 11 05:16:15 EDT 2008


Gus,

useOldFreezing=true resets all temperatures in the entire domain to  
-1.9 (or -1.8, have a look at freeze.F), from top to bottom, if they  
are below -1.9degC (or -1.8?). Effectively you add  heat by doing  
that and this heat is not accounted for anywhere else. So in the  
absense of any heat forcing at the surface, you ocean model will not  
conserve heat with this flag (in fact it will gain heat). So I'd turn  
it off.

I have run many different (uncoupled) configurations where the heat  
flux is prescribed at the surface (or computed from atmospheric  
parameters and bulk formulae and then prescribed) and I have never  
seen this type of behavoir, even in 1000-9000 year runs. My suspicion  
is that your system is loosing heat somewhere, probably in the  
coupling. I suggest that you diagnose your net heat flux at the  
surface and then you'll see how big the imbalance is. Is it possible  
that your atmosphere is loosing the heat and transfering that  
imbalance to the ocean? I know for sure that if the MITgcm (ocean  
model) conserves heat very accurately and global mean temperature  
drifts are only possible, if you have a non-zero net heat input.

I would try to find out whether your cold temperatures come from the  
surface (I bet they do) and are "convected" downwards.

I try to answer your other questions below,

Martin
On 10 Mar 2008, at 21:41, Gus Correa wrote:

> Hello MITgcm experts
>
> We've been working on a coupled ocean-atmosphere setup similar to  
> Dror Mirzayof's,
> although on a lat-lon grid, and using a different mixed layer  
> atmosphere
> (Seager et al., Journal of Climate, 1995, p. 1951-1964).
>
> We have the same problem mentioned by Dror Mirzayof and by Jeff Scott,
> where a few grid cells next to the ocean bottom get cold in the  
> beginning of the run. In our case the problem starts on the  
> Southern Ocean,
> but eventually the whole ocean bottom gets filled with cold water  
> ( ~ -1.8 degrees C).
> The ocean bottom heat loss doesn't stop, and on the long run (~1000  
> years)
> the ocean loses about half of its initial heat content.
> This seems to prevent a correct heat balance at the surface as well.
>
> Our model is on lat-lon coordinates with a resolution
> of 4 degrees by 4 degrees by 24 levels (similar to Jeff Scott's).
>
> Our "data" namelist sets the following parameters
> (a bit fortuitously, they were copied from the verification case  
> global_ocean.90x40x15):
>
> useOldFreezing=.TRUE.
> useRealFreshWaterFlux=.TRUE.
> useCDscheme=.TRUE.
> useNHMTerms=.TRUE.
> hFacMin=.05
> hFacMindr=50.0
>
>
> Most namelist options are not set, so I believe the time step is  
> *not* staggered (i.e. default),
> and the advection scheme is the default (centered 2nd order, right?).
>
> Questions:
>
> 1) Do you have any advice on how to prevent the ocean bottom to  
> lose heat?
> 2) What is the right advection scheme to use to prevent bottom heat  
> loss? (Which namelist options to set?)
using a flux limited scheme is the only way to prevent overshoots,  
but you cannot affect the heat balance by advection schemes. Go with  
the default (2) as long as you have these heat problems, and then you  
can switch to temp/saltAdvScheme=77 or 33 (only with  
staggerTimeStep=true) in order to reduce the over and undershoots.
> 3) Do we need to use staggered time step to prevent the bottom  
> cooling? (Which namelist option to use?)
No
> 4) Should we continue to use shaved bottom cells or not? (And how  
> do we cancel shaved cells in the namelist?)
partial cells do not have any influence on the heat budget
> 5) Does the "CD" scheme have any impact on this "bottom cooling"  
> problem?  Should we continue to use it?
CD does not have any influence on the heat budget, it removes noise  
from the solution, you could replace it by higher viscosity (or  
biharmonic viscosity)
> 6) Should we change "useOldFreezing" to ".FALSE.",
Yes



More information about the MITgcm-support mailing list