[MITgcm-support] Setting up viscosities for the model

[Martin Losch]

Hi Kunal,

viscAhGrid is a non-dimensional parameter that basically derives from the stability criterion. You can see in the header of pkg/mom_common/mom_calc_visc.F how it is used to compute the viscosity. The main point is that it should be < 1 to ensure stability (independent of grid spacing, because it is scaled by the grid spacing).

viscAh and viscAhGrid can both be specified together, but if you want to replace viscAh by viscAhGrid (which I always do) the relation is like this: viscAh = viscAhGrid * 0.25 * L**2 / deltaT (where L is the grid spacing in meters), or viscAhGrid = 4*deltaT* viscAh/L**2

2. yes, viscAz (or viscAr) serves as the background viscosity for KPP
3. in the monitor output (only if monitorFreq > 0.) in the stdout you’ll find the maximum CFL number (advcfl_uvel_max, etc)

specifying a bottom drag is inconsistent with no_slip_bottom=.True. (either one of the other)
I would use Leigth instead of Smagorinsky viscosity (but you don’t need to have nonzero values for all of the options, you can choose).
replace globalFile by useSingleCpuIO =.TRUE.,
Try to use a delZ without jumps, the rule of thumb is to have dz(k+1) <= 1.4*dz(k)
for constant grid spacing I find it more convenient to replace delX by dxSpacing:
 dxspacing=2891.9791,
 dyspacing=2892.8394,

Hope that helps,
Martin

> On 2. Oct 2020, at 14:31, kunal madkaiker <kunal.madkaiker02 at gmail.com> wrote:
> 
> Dear Team,
> 
> I am having slight confusion about setting the horizontal eddy viscosity in my model. Here, I am only considering Laplacian horizontal viscosity (i.e. not considering Leith, Biharmonic viscosity, etc). I have a 360x780 grid, with 3000m x 3000m spatial resolution, length of mesh ~20 deg, width of mesh ~10 deg and timestep of 120 sec.
> 
> Referring to the manual, I find two criterias to be satisfied i.e., CFL Constraint on Viscosity (Section 2.19.1.7) and Numerical Stability Criteria (Section 4.3.2.1, eq 4.26). Currently, I am specifying viscAh=5000 m2/s. 
> The upper limit for Ah comes to be 18750 > 5000. Also,
> i.e. the stability for horizontal viscosity. For my setup, it gave a value of 0.53 < 0.6 (solving using Adams-Bashforth). Now, when I use viscAh=6000, Sl becomes 0.64 > 0.6, but my model simulation went through without crashing. I am trying to understand how the model still ran.
> 
> Also, I believe using viscAhgrid would be a better option. But the manual doesn't mention how viscAhgrid is being computed/ what's the formula (range is between 0 and 1). I have set viscAhGridMax=0.5 in data
> 
> I am trying to address the following questions:
> 1. How to compute viscAhgrid? (or do I have to use Sl value as viscAhgrid?)
> 2. Can both viscAz and Implicitviscocity=.TRUE. be set at the same time (since KPP needs implicitviscocity=.TRUE). I have kept viscAz=0.001
> 3. Is there any way to check if the run has been stable by using the U and V momentum tendencies from dissipation and advection term in the output? Or any other variable in the output (S,T,U,V,W,eta)? 
> Note: I am specifically interested in using the Laplacian eddy viscosity. 
> I am also attaching my data file. Kindly assist how I can improve the setup further. 
> 
> Regards
> Kunal
> <data>_______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support