[MITgcm-devel] obcs_seaice_sponge.F

[Matthew Mazloff]

Hello

While we are changing obcs_sponge.F I would request an CPP flag that  
allows one to turn off normal flow sponge, but leave on tangential  
flow sponge.  Only having a choice between U and V is limiting for the  
case of several OBCS.

The reason for this is if one does not have flat topography in the  
sponge layer one has to be very careful about restoring to a velocity  
in that it can introduce large W.  Coastlines can be incredibly  
problematic if they just out into the domain.

So in my code, I have

#ifdef SPONGE_NORMAL_FLOW

around U sponging on the east and west boundaries, and around V  
sponging in the north and south boundaries.  Undefining it says only  
prescribe normal flow, but you can sponge tangential flow.

Upwelling in the sponge layers has been killing me now that I BGC  
model -- and this does help a bit.  Though ideally one would just have  
flat topo in the sponge layers...

anyway, my 2 cents

-Matt







On Sep 20, 2012, at 12:23 PM, Menemenlis, Dimitris (3248) wrote:

> Jean-Michel, I just checked in obcs_seaice_sponge.F and verification  
> experiment.
> I did not on turn on sponge for oceanic fields, as you recommend  
> below.
>
> Regarding obcs_sponge.F
> I just realize that Ian and Michael also run into problem described  
> below and fixed in a local copy:
> http://ecco2.jpl.nasa.gov/data3/MPS_output/pub/patrick/code/obcs_sponge.F
>
> Would it be OK if I check in their fix, with ALLOW_OBCS_LINEAR  
> undefed by default,
> so as not to affect existing set-ups, and with proper comments and  
> explanation in OBCS_OPTIONS.h
> That way the next person that may look at this routine does not have  
> to find the bug anew?
>
> Let me know.
>
> Dimitris Menemenlis
>
> On Sep 20, 2012, at 11:46 AM, Jean-Michel Campin wrote:
>
>> Hi Dimitris,
>>
>> Yes, this is part of the long list of issues with OBCS SPONGE.
>> And by the way, I think it would be better not to turn on the sponge
>> for oceanic fields when you add this to exp seice_obcs, because  
>> regarding
>> velocity (as you mentionned) there are other issues.
>>
>> Jean-Michel
>>
>> On Thu, Sep 20, 2012 at 05:45:58PM +0000, Menemenlis, Dimitris  
>> (3248) wrote:
>>> Jean-Michel and Patrick, I know that it is no longer actively  
>>> supported but, for the record, there
>>> seems to be a problem with obcs_sponge.F, and it dates back to  
>>> very first check in 10+ years ago.
>>>
>>> According to comments, the relaxation time scale is supposed to be  
>>> linear
>>> and controlled by U/Vrelaxobcsinner/bound:
>>>
>>> C U/Vrelaxobcsinner/bound :: relaxation time scale (in seconds) on  
>>> the boundary
>>> C                            (bound) and at the innermost grid  
>>> point of the sponge
>>> C                            layer (inner); relaxation time scales  
>>> in-between
>>> C                            are linearly interpolated from these  
>>> values
>>>
>>> But as coded, the relaxation is quadratic because there is an  
>>> additional attenuation term due
>>> to linearly decreasing u/v/t/srelax as it goes away from the  
>>> boundary.
>>>
>>> Additionally, as coded, relaxation will be zero if spongeThickness  
>>> = 1.
>>>
>>> I will fix these two problems in obcs_seaice_sponge.F before  
>>> checking in,
>>>
>>> Let me know if you want me to do anything about these two problem  
>>> in obcs_sponge.F
>>>
>>> Cheers
>>>
>>> Dimitris Menemenlis
>
>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel