[MITgcm-support] Problem using partial cells

[Martin Losch]

Hi Isa,

as you might have found out yourself, running the model (in fact any model) at this high resolution requires careful tuning, because there are many stability criteria that one has to observe. In your case it seems that with setting the hFacMin to 0.1 you make a few cell too thin (hFacMin*delR), so that the vertical cfl number is violated. You can check by setting your monitorFreq to your deltaT or less to get monitor output every time step and then look at the variables advcfl_*
As a solution, you can specify hFacMinDr top of hFacMin to control the thickness of the cells = max(hFacMinDr, hFacMin*delR)

Also for debugging, I'd change the debugLevel to at least 1, just to see how long it takes before the model explodes.

Martin

PS. a few considerations about your parameter list:
- you use a lot of viscosity/diffusion parameters. I am not sure if using Smag and Leith simultaneously is physically justified, check with the documenation: <http://mitgcm.org/public/r2_manual/latest/online_documents/node85.html>
- I am not sure how useful diffK4T/S are, because they are grid cell dependent. If your resolution varies over your domain, you'll likely run into problems because the same diffK4 might be to large for the high resolution and at the same time too low for the low resolution. your advection should be stable without extra diffusion.
- multiDimAdvection is set automatically and can be omitted (I'd do so)
- useRealFreshWaterFlux = .TRUE., set  convertFW2Salt automatically to -1, so you can omit that parameter
- with useSingleCPUio, you don't need globalFiles=.TRUE., I think globalFiles is not a very safe option sometimes



On Sep 3, 2012, at 8:58 AM, Isa Rosso wrote:

> Hi all,
> 
> I'm trying to use partial cells in my high-res model (1/80 deg resolution, open boundaries, 150 vertical layers, realistic topography). The model was running so far without this capability, with a time step up to 30 seconds. I tried to run it setting the hFacMin = 0.1, reducing the time step to 5 seconds, but the model crashes. 
> 
> Is there some particular settings I should consider?
> 
> Below the PARAM01 of the data file.
> 
> 
> Thanks,
> isa
> 
> 
> &PARM01
>  tRef               	= 150*5.0,
>  sRef               = 150*34.5,
>  viscAr 		= 5.6614e-04,
>  diffKrT 		= 1.e-5,
>  diffKrS 		= 1.e-5,
>  no_slip_sides 	= .FALSE.,
>  no_slip_bottom 	= .TRUE.,
>  rhonil 		= 1027.5,
>  rhoConstFresh  	= 999.8,
>  eosType 	= 'JMD95Z',
>  hFacMin 	= 0.1,
>  hFacInf 	 	= 0.01,
>  hFacSup 	= 2.,
>  select_rStar 	= 2,
>  nonlinFreeSurf 	= 4,
>  implicitDiffusion 	= .TRUE.,
>  implicitViscosity 	= .TRUE.,
>  viscC4Leith 	 	= 2.0,
>  viscC4Leithd	= 2.0,
>  viscC2Leith 		= 1.5,
>  viscC2Leithd 	= 1.5,
>  viscA4GridMax 	= 0.5,
>  viscAhGridMax 	= 0.5,
>  viscC4Smag 	= 2.2,
>  diffK4T 	= 1.E8,
>  diffK4S 	= 1.E8,
>  useAreaViscLength = .TRUE.,
>  sideDragFactor 	= 0.,
>  useAbsVorticity 	= .TRUE.,
>  selectVortScheme 	= 2,
>  selectKEscheme 	= 3,
>  bottomDragQuadratic = 0.0034,
>  tempAdvScheme 	= 7,
>  saltAdvScheme 		= 7,
>  StaggerTimeStep 	= .TRUE.,
>  multiDimAdvection 	= .TRUE.,
>  vectorInvariantMomentum = .TRUE.,
>  implicitFreeSurface 	= .TRUE.,
>  exactConserv 	= .TRUE.,
>  debuglevel 		= -1,
>  convertFW2Salt 	= -1
>  useRealFreshWaterFlux = .TRUE.,
>  useSingleCPUio 	= .TRUE.,
>  globalFiles 		= .TRUE.,
>  allowFreezing 	= .TRUE.,
>  &
> 
> ***************
> Isa Rosso
> PhD student
> 
> Research School of Earth Sciences
> The Australian National University
> Canberra ACT 0200
> Australia
> 
> ph: +61 2 6125 9958
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support