[MITgcm-support] non-hydrostatic pressure and KPP

Fri May 3 08:44:24 EDT 2019

Hi Martin,
We are using MITgcm to reproduce gravity currents induced by differential cooling in lakes. I followed your suggestion, and we are using the non-hydrostatic capabilities of MITgcm + 3D Smagorinsky. But, still our grid resolution is relatively coarse (dx = dy = 5 m, dz = 0.2 m), and limited right now by the available resources, so we cannot fully resolve the convective cells developing in the epilimnion (and of course, not at all, the turbulent features between the gravity current and the surrounding water), so I believe we still need some kind of parameterization for the mixing, right? 

Viscosities are calculated with 3D Smagorinsky, but how is the model dealing with diffusivities in this case?

I have attached an example of model vs. measurements (qualitative) for two types of simulations (M1 is close to the shore and above the thermocline in the lake and M3 is deeper and crossed by the thermocline) in case this helps to clarify:
Hydrostatic + KPP + viscC2Smag = 0.3
Non-hydrostatic + smag3D_coeff = 5e-4 (please, note, for example, how isotherms are tilted in this simulation when compared to the measurements and the hydrostatic simulation, which could be a sign of only resolving partly the mixing induced by the convective cells)

For both types of simulations (hydrostatic and non-hydrostatic),the average RMSE errors between measured and modelled temperatures are similar and ~ 0.1°C

The "data" file for the non-hydrostatic simulation is also attached.

By the way, is there an easy way to output viscosities and diffusivities from the model (3D output and/or diagnostics)?

Thanks in advance for your help
Cheers

Cintia

-----Original Message-----
From: MITgcm-support <mitgcm-support-bounces at mitgcm.org> On Behalf Of Martin Losch
Sent: Thursday, 24 January 2019 17:08
To: MITgcm Support <mitgcm-support at mitgcm.org>
Subject: Re: [MITgcm-support] non-hydrostatic pressure and KPP

Hi Cintia,

KPP is not a “turbulent closure” for LES-like simulations, but a “mixed layer model” that parameterizes near surface vertical mixing processes in climate-scale models with grid spacing of order kilometers to tens of kilometers (have a look at the original publication of Large et al 1994 for more details). For your application you should probably turn it off, because it is trying to parameterize also the processes that non-hydrostatic dynamics are supposed to resolve.

There are some (simpler) turbulent closure schemes avaiable, e.g. Smagorinsky, please have a look at the documenation: <https://mitgcm.readthedocs.io/en/latest/algorithm/algorithm.html#nonlinear-viscosities-for-large-eddy-simulation>

There it says, that 3D smagorinski is not implemented, but I don’t think that this is true anymore. There’s a flag useSmag3D for that, but you may want to check the code for more details.

Martin

> On 24. Jan 2019, at 12:39, Ramón Casañas, Cintia <Cintia.RamonCasanas at eawag.ch> wrote:
> 
> Dear all,
>  
> I got the warning and found these lines about KPP and the use of the non-hydrostatic capabilities of MITgcm (see below).
> I’m new with MITgcm. We are trying to run MITgcm in a small lake (2.5 km in length and 0.5 km wide) and we would like to resolve the convective cells developing during night-time cooling in the lake. The grid size that it’s needed to resolve or at least partially resolve these cells (dx = dy = 5m), taking into account non-hydrostatic effects, is still coarse (compared to the grid size that would be needed for a LES), so, a turbulence closure is still needed to model the mixing. 
> So please, could you explain why the non-hydrostatic capabilities cannot be used with a turbulence closure scheme? Or is this only specifically related to the KPP formulation?
>  
> Thanks in advance
>  
> Cintia
>  
> Hi Sherry,
>  
> thanks to version control, you can easily checkout old versions of the code and try to find, when (in “code-time”) the instabilities occur. 2009 ended with checkpoint62 (get it with "cvs co -r checkpoint62 MITgcm”, but you can also use dates, see cvs documentation, see MITgcm/doc/tag-index for record of what has been done over time). With “divide and conquer” (e.g. if 2009 code works try 2012 code, if that works 2014, if not 2010.5, etc), it doesn’t take that long, but it’s tedious anyway. Once you’ve found the point, where the model starts to have problems, you can use http://mitgcm.org/viewvc/MITgcm/MITgcm/ to look at differences.
>  
> KPP does not make too much sense together with a non-hydrostratic simulation, where you (usually) want to resolve the mixing processes, rather than parameterise them. I would take the warnings seriously.
>  
> Martin
> > On 18 Aug 2016, at 07:59, Sherry <schou at hawaii.edu> wrote:
> > 
> > Hi all,
> > 
> > I have a 2-D nonhydrostatic configuration of an internal wave beam forced through specifying u, v, w, and temperature at the western boundary. At the moment I am trying to reproduce an experiment that was done in 2009 (Grisouard and Staquet, 2010), but for some reason I am getting instabilities after about 10 wave cycles, even though I am using the same input parameters. Could there be some changes in the model since 2009 that would affect default values? Any advice about how to track these down?
> > 
> > To control the instabilities I am trying higher diffusion values, as well as turning on KPP. However, it seems there is some internal inconsistency with the configuration being nonhydrostatic, and I get the following warnings (from config_check.F):
> > 
> > WARNING  CONFIG_CHECK: Implicit viscosity applies to provisional 
> > u,vVel WARNING  => not consistent withfinal vertical shear (after 
> > appling 3-D solver solution WARNING  CONFIG_CHECK: Implicit 
> > viscosity not implemented in CALC_GW WARNING  CONFIG_CHECK: Explicit 
> > viscosity might become unstable if too large
> > 
> > Should I be concerned about these warnings? Is there any way around them since nonhydrostatic requires 3-D solver and KPP requires implicit viscosity?
> > 
> > Thanks for reading!
> > 
> > Sherry Chou
> > University of Hawaii
> > _______________________________________________
> > 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://mailman.mitgcm.org/mailman/listinfo/mitgcm-support

_______________________________________________
MITgcm-support mailing list
MITgcm-support at mitgcm.org
http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
-------------- next part --------------
A non-text attachment was scrubbed...
Name: HydrostaticLeft_MeasuredRight.png
Type: image/png
Size: 335657 bytes
Desc: HydrostaticLeft_MeasuredRight.png
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20190503/ba1346fe/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: NonHydrostaticLeft_MeasuredRight.png
Type: image/png
Size: 461655 bytes
Desc: NonHydrostaticLeft_MeasuredRight.png
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20190503/ba1346fe/attachment-0003.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dataNH
Type: application/octet-stream
Size: 6823 bytes
Desc: dataNH
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20190503/ba1346fe/attachment-0001.obj>