[MITgcm-support] online calculation of energy flux

[Jody Klymak]

Hi Ryan,


On Jan 31, 2014, at  12:46 PM, Ryan Abernathey <ryan.abernathey at gmail.com> wrote:

> Wow, great! I didn't know this existed. Thanks Jody.

Well, it only exists as an attempt.  Its not numerically robust, and it is really very tailored for internal wave baroclinic/barotropic problems.  

I should have also said that for the calculation of APE it assumes that the derivative of N^2 is small, and that it assumes a constant background N^2 in time that density perturbations are working against.  This is good for internal waves for short process study runs.  Its unlikely to be good for long-term studies where N^2 should be allowed to evolve freely.  i.e. it doesn't close a APE/background PE  in any intelligent way.  

Having background PE in the energy equations is a real nuisance.....


Cheers,   Jody


> 
> On Fri, Jan 31, 2014 at 3:06 PM, Jody Klymak <jklymak at uvic.ca> wrote:
> Hi Robert and Sonya,
> 
> My online energy diagnostic is at:
> 
> https://github.com/jklymak/MITgcmcode
> 
> It is after Kang and Fringer, and Kang's thesis, and does a barotropic/baroclinic decomposition for the linear terms. The non-linear terms are not formally separable, and so I don't separate them! I think I have most of the terms correct, but bugs wouldnt surprise me.  Its not well documented, but hopefully largely self-explanatory.  
> 
> It certainly does not attempt to deal w/ model discretization issues or calculate explicit dissipation from general dissipative terms.  I would suggest that folks who write viscosity submodules should consider calculating dissipation for their users!
> 
> If anyone uses this, or in particular finds bugs, I'd love to hear about it.
> 
> Thanks,  Jody
> 
> 
> 
> On Jan 31, 2014, at  10:19 AM, Ryan Abernathey <ryan.abernathey at gmail.com> wrote:
> 
>> Robert and Sonya,
>> 
>> This is an issue that came up at the recent ECCO meeting. Many people would probably like to have such kinetic energy diagnostics from MITgcm. Although the energy equation is easy to write down, diagnosing all of the terms in a way that is consistent with the model discretization can be very difficult. Many people have probably tackled it in the past and come up with a "good enough" solution, no standard solution has emerged.
>> 
>> Peng's suggestion is correct, but the problem is that mean(up) is not currently an available diagnostic. That would be fairly trivial to add, especially if you already know how to diagnose it from offline output. The current pressure diagnostics are filled in model/src/dynamics.F, which could be modified to include the pressure flux term.
>> 
>> More generally, if you want a complete energy budget, you will also want the advective transport of kinetic energy, which involves tracking "triple correlation" terms such as mean(uv^2). This is similar to the problem encountered in the diagnosis of tracer variance budgets. Several people (including Jean Michel, myself, and Liam Brannigan) have recently become interested in adding such diagnostics to the model. Nonlinear advection schemes can make this difficult for tracers, but perhaps the momentum advection is actually simpler.
>> 
>> The biggest issue I see in closing the kinetic energy budget is the dissipation term (epsilon). If you are doing energy budgets for internal waves, presumably this term is a leading order one for you. A robust method for diagnosing dissipation (consistent with the numerics and LES options such as Smagorinsky) would be a great contribution.
>> 
>> I personally wish the MITgcm community could come together to produce a definitive answer to this problem. But of course, it is not a high priority for most people, and it is a very difficult numerical problem. Plus it is possible that the necessarily diagnostics would significantly slow down the code execution. Whatever solution you come up with, you should definitely share it with the community, perhaps through the MITgcm_contrib repository, for the benefit of the next grad student who is faced with this simple sounding but actually very deep task.
>> 
>> Best,
>> Ryan
>> 
>> 
>> On Fri, Jan 31, 2014 at 11:45 AM, Sonya Legg <sonya.legg at noaa.gov> wrote:
>> Hi Robert, 
>> 
>> You should state that it's the product u'p' that you want, and you want to do it online so that you can calculate time-averages. Otherwise you might get a response that u is already output, as is p, so why do you need to add anything? 
>> 
>> Sonya
>> 
>> 
>> On 1/31/2014 11:39 AM, Robert Nazarian wrote:
>>> Hi,
>>> 
>>> I’m trying to add an online evaluation of u’ and p’ to the source code to calculate the energy flux. Is there a particular diagnostic that would be best to do this in? If not, is there a particular subroutine that’s ideal to write such a diagnostic? Previously, I did these calculations offline but am hoping to incorporate it into the code itself.
>>> 
>>> Thanks,
>>> Rob
>>> 
>>> Robert Nazarian
>>> 
>>> Program in Atmospheric and Oceanic Sciences
>>> 
>>> Princeton University
>>> 
>>> rn2 at princeton.edu
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> 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
> 
> 
> _______________________________________________
> 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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20140131/f8942abe/attachment.htm>