[MITgcm-support] Problem with KPP

Martin Losch Martin.Losch at awi.de
Mon Jun 9 03:21:06 EDT 2008


Abbas,

in fact, the model is blowing up, your compiler just continues to  
work with NaN's. Some compilers can do that, which is often annoying.

You need to have a closer look at what's going on by setting you  
moniterFreq to a lower value (it's only in this package, where the  
model checks for a run-away theta), and by producing output to see,  
where stuff blows up first (location and variables). I suspect a cfl- 
problem, but based on your information it could be anything ...

Martin

On 5 Jun 2008, at 20:41, Abbas Dorostkar wrote:

> Hi all,
> I am simulating internal waves in a lake. The model without using KPP
> scheme works well. However, when I implement KPP, it does not blow up
> but I get NAN in output.txt. I wonder what is going wrong. Your  
> help is
> much appreciated.
> I have attached some part of output.txt, data, data.kpp
>
>  (PID.TID 0000.0001)  exf_GetFFieldsRec: first, changed, fac:   F   F
> 0.80000000D+00
> (PID.TID 0000.0001)  exf_GetFFieldsRec: myiter, count0, count1:     24
> 1      2
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
> (PID.TID 0000.0001)  exf_GetFFieldsRec: first, changed, fac:   F   F
> 0.80000000D+00
> (PID.TID 0000.0001)  exf_GetFFieldsRec: myiter, count0, count1:     24
> 1      2
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
>  cg2d: Sum(rhs),rhsMax =                    NaN  4.05115865207719+303
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
> (PID.TID 0000.0001)  exf_GetFFieldsRec: first, changed, fac:   F   F
> 0.79166667D+00
> (PID.TID 0000.0001)  exf_GetFFieldsRec: myiter, count0, count1:     25
> 1      2
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
> (PID.TID 0000.0001)  exf_GetFFieldsRec: first, changed, fac:   F   F
> 0.79166667D+00
> (PID.TID 0000.0001)  exf_GetFFieldsRec: myiter, count0, count1:     25
> 1      2
> (PID.TID 0000.0001)  exf_GetFFieldsRec:
>  cg2d: Sum(rhs),rhsMax =                    NaN                   NaN
>
> -------------------------------------------------
> # Continuous equation parameters
>  &PARM01
>  Tref =,
>  sRef=0.,
>  viscAz=1.E-3,
>  viscAh=1.0E-2,
>  viscA4=0.E12,
>  no_slip_sides=.FALSE.,
>  no_slip_bottom=.FALSE.,
>  diffKhT=1.E-2,
>  diffKzT=1.E-3,
>  f0=0.0,
>  beta=0.E-11,
>  tAlpha=2.E-4,
>  sBeta =0.E-4,
>  rigidLid=.FALSE.,
>  implicitFreeSurface=.TRUE.,
>  eosType='LINEAR',
>  nonHydrostatic=.FALSE.,
>  readBinaryPrec=32,
>  writeBinaryPrec=32,
>  writeStatePrec=32,
>  saltStepping=.FALSE.,
>  hFacMin=0.2,
>  implicitViscosity=.TRUE.,
>  implicitDiffusion=.TRUE.,
> &
> # Elliptic solver parameters
>  &PARM02
>  cg2dMaxIters=1000,
>  cg2dTargetResidual=1.E-12,
>  cg3dMaxIters=400,
>  cg3dTargetResidual=1.E-12,
>  &
> # Time stepping parameters
>  &PARM03
>  niter0=0,
>  startTime=0.0,
>  endTime=604800,
>  deltaT=5.,
>  abEps=0.1,
>  pChkptFreq=0.0,
>  chkptFreq=0.0,
>  dumpFreq=200,
>  monitorFreq=1000,
>  &
> # Gridding parameters
>  &PARM04
>  usingCartesianGrid=.TRUE.,
>  usingSphericalPolarGrid=.FALSE.,
>  delX=546*99.880952381,
>  delY=3000.,
>  delRFile='delz.grid',
> #delZ=132*1.000649793,
>  &
> ----------------------
> # KPP parameters
>  &KPP_PARM01
>  KPPmixingMaps   = .FALSE.,
>  KPPwriteState   = .TRUE.,
>  KPP_ghatUseTotalDiffus=.FALSE.,
>  &
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list