[MITgcm-support] Energy diagnostics?

[Klymak Jody]

Hi Jeff et al.

On Feb 17, 2009, at 8:51 AM, Jeffery Scott wrote:

>
> Several years ago Alistair and I worked out how to calculate this,  
> i.e. ,  given complications due to Adams-Bashforth time stepping,  
> and we able to diagnose under what conditions/run parameters the  
> MITGCM conserves KE (albeit fairly limited testing of MITGCM  
> options). PE is a bit more messy and I gave up at some point,  
> perhaps others have had more patience.

OK, I'm ready for a hint at this point.  I have the diagnostic  
working, and it improves the standard deviation of the error compared  
to using snapshots by an order of magnitude (Comparing dE/dt to u(P+E)  
- including all the PE terms).  So far so good, and its probably fine.

However, I do have the slight oddity that if I compute dE/dt as  
diff(E)/dt_diagnostic the error is not as small as when I compute  
diff(E)/(dt_diagnostic+dt_model/2).  So, is this the time-stepping  
scheme telling me that it computes the large terms in E at time  
bracketting the flux terms?  Or am I just tilting at windmills here?   
Any thoughts you had about the "complications due to the Adams- 
Bamforth time stepping" - even schematic ones - would be appreciated.

I've written a small document, and will post the code when I've run a  
few more tests.

Thanks,  Jody

>
> However, we did this all offline, so I'm not sure how much help it  
> would be, but happy to share, have some notes written up...
>
> Jeff Scott
>
>
> On Feb 16, 2009, at 3:51 AM, Martin Losch wrote:
>
>> Hi Jody,
>>
>> I am not aware of energy flux diagnostics [(E+P)u] in the code. I  
>> agree with you that for closing the budget they need to be computed  
>> correctly (at each time step and then averaged), horizontal  
>> averaging will also be an issue.
>>
>> Martin
>>
>> On Feb 15, 2009, at 10:25 PM, Klymak Jody wrote:
>>
>>>
>>> Hi all,
>>>
>>> Before I spend a couple of days doing this, has anyone gone  
>>> through the effort of calculating energy diagnostics?
>>>
>>> I'm running barotropic flow over an obstacle (all in 2-D).  If I  
>>> take every timestep I can get a decent energy budget:
>>>
>>> (d/dt)\int E dV = \oint (E+P) u \cdot dA + \int \epsilon dV
>>>
>>> where \epsilon is specified and numerical dissipation.  However (E 
>>> +P)u changes so much over my output timesteps the residual when  
>>> compared w/ dE/dt is unrealistically large.      Averaging the  
>>> linear and non-linear flux over the dump interval should give a  
>>> much better estimate.
>>>
>>> Has anyone done this?  If not, I'll share my code.
>>>
>>> Thanks,  Jody
>>>
>>> --
>>> Jody Klymak
>>> http://web.uvic.ca/~jklymak/
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

--
Jody Klymak
http://web.uvic.ca/~jklymak/