[MITgcm-support] numerical noise with CFL<0.02

[Martin Losch]

Hi Antoine,

there are more stability criteria than just the advective CFL number. See the tutorials for some examples, e.g. here: <https://mitgcm.readthedocs.io/en/latest/examples/barotropic_gyre/barotropic_gyre.html#numerical-stability-criteria>

I would always use viscAhGrid instead of vischAh. With values < 1 (I usually use something around 0.1 as a first guess) you’ll make sure that you don’t violate any stability criterion there. (see pkg/mom_common/mom_calc_visc.F or documentation for details)

Also try a smaller ivdc_kappa (order 1-10 maybe) to see if that helps.

What type of noise do you see? Stationary fields? Does it increase with time? In the horizontal velocity fields (on C-grids at coarse resolution you’ll always see some of that, can be suppressed by biharmonic viscosity e.g. with viscA4grid = 0.1)? Or in the tracer fields?

Martin

> On 30. Nov 2018, at 03:28, Antoine Hochet <ahochet at ucsd.edu> wrote:
> 
> Hi all,
> I am running a North Atlantic box with thermal forcing only, at 80km resolution with LLC4320 numerics. I can't understand why I have numerical noise even though CFL < 0.02
> Details of the configuration are below.
> Thanks for your help,
> Antoine
> 
> LLC4320 numerics:
> http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/llc_hires/llc_4320/input/
> http://wwwcvs.mitgcm.org/viewvc/MITgcm/MITgcm_contrib/llc_hires/llc_4320/code/
> except for 
> # rhoConstFresh=999.8,
> # viscC4Leith=2.15,
> # viscC4Leithd=2.15,
> # viscA4GridMax=0.8,
> # useAreaViscLength=.TRUE.,
> # debuglevel=-1,
> # convertFW2Salt=-1.,
> # useRealFreshWaterFlux=.TRUE.,
> # globalFiles=.TRUE.,
> # writeBinaryPrec=32,
> # temp_EvPrRn=0.,
> # implicSurfPress = 0.6,
> # implicDiv2DFlow = 0.6,
> # bottomDragQuadratic = 0.0021,
> and no KPP and instead:
> bottomDragLinear =3.E-4,
> viscAz=1.E-3,
> viscAh=1.E3,
> diffKhT=7.E2,
> diffKzT=1.E-4,
> ivdc_kappa=100.,
> cg2dMaxIters=1000,
> cg2dTargetResidual=1.E-13,
> 
> 
> 
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support