Subject: Re: Problem with CCM 3.6.6 (T31) blowing up within 2 years ...
From: David Noone (dcn@gps.caltech.edu)
Date: Fri Mar 01 2002 - 12:40:33 MST
Hi Dave,
Sounds like a CLF violation alright! In the last output line, NSTEP=
28660, it is greater than 2! If you are running with the Eulerian
dynamic core, this MUST stay below 1. The problem turns up in the
semi-Lagrangian code due to the latitude scanning algorithm, but is
symptomatic of bigger problems. Although I haven't had problem running
at T31 with 1800 second, the easiest fix is to make the timestep smaller
(try going back to 1200?). You could also try bumping up the diffusion
coefficients dif4 _and_ dif2. Also, if you change the timestep, you can
consider changing the time filter coefficient (EPS) to have the same
amount of temporal smoothing - but I doubt this is the problem as the
CCM3 default is quite strong. Notice in your output that the courant
number becomes problematic at k=1 (the model top). The del^2 diffusion
is only applied in the top (pressure only) layers of the model, and
worth playing with. Perhaps double it? An early modelling paper
(Smagorinski? Phillips? circa 1950/60s) suggested it scales in grid
point models as the cell spcing to the power of 4/3 from "eddy
viscosity" arguments. Keep and eye on the courant number in the output
line when you restart. It should drop and stay below 1 on the order of a
day or so. This answer is a numerical solution, and doesn't address the
more important problem of working out if it is something in the dynamics
that has gone haywire.
Hope this helps,
David
On Fri, 1 Mar 2002, Dave Newman wrote:
> Dear CCM users --
>
> I am having a problem with CCM 3.6.6 blowing up within 2 years, and
> wanted to see if anyone was experiencing anything similar, or if anyone
> had any suggestions on what may be causing the problem. Here is some
> relevant information:
>
> Version: CCM 3.6.6
> Res: T31
> DIME: 1800 (=30 mins)
> Mode: DST_FRC (i.e. no feedbacks)
> DIF4: 0.2E17
> Machine: SGI
>
> The run stops at NSTEP=28660 (approx day=597, year=1.6) with the message:
> CUBYDR: Departure point out of bounds: jcen,icount,pgls= 9, 1726, 1728
>
> I noticed that the Courant limit seemed to be creeping up as well.
> Also, I reran exactly the same model/case at T42, and it ran perfectly
> fine for several years (well beyond day=597)
>
> Does anyone have any suggestions or comments? Any help would be greatly
> appreciated!
>
> Thanks,
> Dave Newman
> newman@uci.edu
>
> --------------------------------------------------------------
>
> Below is some more detailed info on this run ...
>
> Input data
> ==========
>
> dstccm50: CCM 3.6.6 dst-1.0.15: dcr mbl newman test
> User name to build pathnames = newman
> Initial dataset is: JAN1.T31.0198.nc
> Time-variant boundary dataset (sst) is: T31M5079.nc
> Time-variant boundary dataset (ozone) is: ozn.0596.r8.nc
>
> History file 1 will be a monthly average
> Fields on history file 2 are instantaneous
> Base day,seconds of day = 0 0
> Base date,seconds of date = 101 0
> Elapsed time to end run (NELAPSE) 87600
> Time step in seconds (DTIME) 1800.
> Time filter coefficient (EPS) 0.060
> DEL2 Horizontal diffusion coefficient (DIF2) 0.250E+06
> DEL4 Horizontal diffusion coefficient (DIF4) 0.200E+17
> Number of levels Courant limiter applied 1
> Lowest level for dry adiabatic adjust (NLVDRY) 3
> Frequency of Shortwave Radiation calc. (IRADSW) 2
> Frequency of Longwave Radiation calc. (IRADLW) 2
> Frequency of Absorptivity/Emissivity calc. (IRADAE) 24
> Frequency of SST Initialization calc. (ITSST) 1
> SST dataset will be reused for each model year
> OZONE dataset will be reused for each model year
> Output files will be disposed ASYNCHRONOUSLY
> divergence damper NOT invoked
> global co2 volume mixing ratio = 3.5500000000000001E-4
> number of user advected tracers= 4
> All advected user tracers will be initialized to zero
> Visible optical depth (tauvis) = 0.14000000000000001
> Orbital parameters calculated for given year: 1950 AD
>
> Final output
> ============
>
> NSTEP = 28658 8.779757585849342E-05 6.328269167202861E-06 252.587
> 9.84572E+04 2.387539556912451E+01 1.61 0.32
> COURLIM: *** Courant limit exceeded at k,lat= 1 37 (estimate = 1.391),
> solution has been truncated to wavenumber 22 ***
> *** Original Courant limit exceeded at k,lat= 1 37 (estimate =
> 1.391) ***
> NSTEP = 28659 8.781644339713775E-05 6.556245519381330E-06 252.583
> 9.84572E+04 2.387403670644274E+01 1.86 0.48
> COURLIM: *** Courant limit exceeded at k,lat= 1 37 (estimate = 1.405),
> solution has been truncated to wavenumber 22 ***
> *** Original Courant limit exceeded at k,lat= 1 37 (estimate =
> 1.405) ***
> NSTEP = 28660 8.785450835741513E-05 7.065918497627775E-06 252.578
> 9.84572E+04 2.387370473511656E+01 2.09 0.70
> CUBYDR: Departure point out of bounds: jcen,icount,pgls= 9, 1726, 1728
> IOT Trap
> Error: thread 1 (pid 2480255) has returned unexpectedly
> Received signal 6: Abort or IOT
>
>
>
---------------------------------------------------------------------- California Institute of Technology, Pasadena, CA, 91125, USA Division of Geological and Planetary Sciences Mail Stop 100-23, Phone (626) 395-6982 http://www.gps.caltech.edu/~dcn
This archive was generated by hypermail 2b27 : Fri Mar 01 2002 - 12:52:55 MST