[MITgcm-support] Simulation Breaking with r*

[Martin Losch]

Hi Senja,

In general, this error message can also just mean that your model explodes, but you are right, one of the reasons may be one or more grid cells running dry. Did you have a look at the horizontal eta field just before the model stops? Maybe there’s some horizontal noise or instability. The eta-values (-0.155m) does not seem extreme compared your dz[0]=0.375m.

I don’t know anything about your configuration except your vertical grid spacing, which is very fine and not monotonically increasing with depth.
I don’t think that this should stop you from simulating large-amplitude seiches (this is what the r*-coordinates are made for), but I suggest that your vertical grid spacing varies smoothly with depth for numerical stability.

Naturally there are limits to how thin (or thick) your cells can get. These limits can be adjusted with the runtime parameters hFacInf, hFacSup. You probably ran into a situation where your hFac becomes smaller than hFacInf. Keep in mind that the thinner the cells, the stronger the vertical cfl-constraint. 
Again I don’t know your namelist files, but maybe there’s something to smooth the simulations? 

Martin

> On 20. Mar 2019, at 15:30, Senja Walberg <senja.w at gmail.com> wrote:
> 
> Hi, 
> I'm running a lake simulation with the nonlinear free surface and the r* coordinate, an eventual aim is to include seiche behaviour in the model. The model is breaking consistently as I mentioned in an email before, and this is the error we have been getting:
> 
>  fail at i,j=  14  11 ; rStarFacC,H,eta =  0.199897  1.939551E-01 -1.551841E-01
> WARNING: r*FacC < hFacInf at       1 pts : bi,bj,Thid,Iter=   1   1   1     63844
> STOP in CALC_R_STAR : too SMALL rStarFac[C,W,S] !
> 
> My understanding is that this is due to near-drainage of the cell. What can we do to solve this problem? Should we change the bathymetry to have a minimum depth, or change the grid spacing? Is it possible with the MITgcm to simulate large-amplitude seiches or is that pointless with the current model? 
> 
> Regards, 
> Senja
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support