[MITgcm-support] [EXTERNAL] Banding, Checkerboarding in KPP Viscosities

[Matthew Mazloff]

Hello

I have been through this and don’t really have an answer. But here is what I remember

First thing to consider:
The “surface” velocity used in KPP should not be the surface cell velocity. This doesn’t matter for coarse model runs, however, so the default is to use surface cell velocity as that is much faster than finding the proper “surface” velocity. For high-res this matters! And is likely causing your noise. The code supposedly works in the mitgcm by defining KPP_ESTIMATE_UREF

In kpp_forcing_surf.F we find:
C     |   - velocity shear relative to surface squared (this is  |
C     |     not really a surface affected quantity unless it is  |
C     |     computed with respect to some resolution independent |
C     |     reference level, that is KPP_ESTIMATE_UREF defined ) |


Second thing to consider: 
There is an option to vertically smooth KPP output. To use define:  ALLOW_KPP_VERTICALLY_SMOOTH
This could also help you, however you have to decide how much to smooth it by setting num_v_smooth_Ri in data.kpp
The default is 1 which won’t help, so try making it bigger - perhaps 10 or more, and see if that helps

Final note, we found that KPP output looked better when we horizontally smoothed it. Check your KPP options to see what smoothing you are using. And check you diagnostics to see if you also have noise in the horizontal.

Matt



> On Apr 2, 2020, at 2:24 PM, Dimitris Menemenlis <dmenemenlis at gmail.com> wrote:
> 
> Hi Ed, I was not questioning the need for a convective adjustment scheme but whether a simpler scheme than KPP, e.g., cAdjFreq/ivdc_kappa might suffice.
> 
> Senja, do they instabilities occur within or below KPPhbl?  Are they stable in time and vertical structure or do they oscillate?
> 
> 
>> On Apr 2, 2020, at 2:10 PM, Edward Doddridge <edward.doddridge at utas.edu.au <mailto:edward.doddridge at utas.edu.au>> wrote:
>> 
>> I would have guessed that KPP is still required. With an aspect ratio of 1:200 this simulation should still be in the hydrostatic regime, despite the high resolution.
>>  
>> If you have the computational resources you could try reducing dx and running a non-hydrostatic simulation without KPP.
>>  
>> Cheers,
>> Ed
> 
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20200402/eed9e8ee/attachment-0001.html>