[MITgcm-support] run with different time steps

[Abbas Dorostkar]

Dear all,

I am running a Lake model for 11 days (269-280 Julian Days) with high
resolution grid spacing (1380X540X98).  At the beginning of simulation, I
can use deltaT=10 sec but after 2.5 days (when Internal Solitary Waves
initiate), I need to use deltaT=1 or even less otherwise the model crashes.
Therefore, I should use multiple time steps in the model to save run time.

I did some tests with a low resolution model to see the result.  In the
first run, the model was run with deltaT=50 for the whole period. In the
second run, I got the pickupfiles of the first run after 2 days and
restarted the model with deltaT=25 for 271 to 280 Julian Days.  I compared
the result of these two simulations (from 271-280 Julian Days) with field
data and the result of second run is worse quantitatively and qualitatively.
 However, if I restart the second model without changing deltaT, the result
is the same as first run.  I am wondering if I am doing something wrong or
if this problem is common. Is it related to the time-stepping method I use.
I have attached the data file here:

Tref =,
sRef=0.,
viscAz=1.E-3,
viscAh=0.0E-2,
viscC2smag=0.7,
viscAhGridMax=0.5,
 viscC2Leith=0.7,
 no_slip_sides=.FALSE.,
 no_slip_bottom=.TRUE.,
 bottomDragQuadratic=0.001,
 diffKhT=1.E-7,
 diffKzT=1.E-7,
 f0=0.0,
 beta=0.E-11,
 tAlpha=2.E-4,
 sBeta =0.E-4,
 gravity=9.81,
 gBaro=9.81,
 rigidLid=.FALSE.,
 implicitFreeSurface=.TRUE.,
 eosType='MDJWF',
 nonHydrostatic=.FALSE.,
 readBinaryPrec=64,
 writeBinaryPrec=32,
 writeStatePrec=32,
 hFacMin=0.5,
 implicitViscosity=.TRUE.,
 implicitDiffusion=.TRUE.,
 tempAdvScheme=33,
 exactConserv=.TRUE.
staggerTimeStep=.TRUE.,

 &
# Elliptic solver parameters
 &PARM02
 cg2dMaxIters=1000,
 cg2dTargetResidual=1.E-12,
 cg3dMaxIters=300,
 cg3dTargetResidual=1.E-12,
 &

# Time stepping parameters
 &PARM03
 niter0=3456,
  endTime=950400,
#nTimeSteps=12096,
 baseTime=86400,
 deltaTClock=25.,
 abEps=0.1,
 pChkptFreq=43200.0,
 chkptFreq=0.0,
 dumpFreq=1200,
 monitorFreq=1000,


Please let me know your comments.

Thanks very much,

Abbas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20090407/8d5a6bb2/attachment.htm>