[MITgcm-support] model crashing, cfl parameter

[Dimitris Menemenlis]

if it crashes early on, i.e., within first few hours or days, my guess is that perturbation
to initial conditions could be the problem. try a lower time step initially,
set monitorfreq=deltat, until advcfl_w decreases consistently below .5 then gradually
increase time step, while continuing to monitor advcfl_w until model has gotten rid of
initial shock.

On Jul 10, 2014, at 3:49 PM, Neil Patel <nigellius at gmail.com> wrote:

> I'm currently running the mitgcm at a 1/6° resolution globally. I was able to get a stable run using a modern day forcing and input files, using the exf package with era 40 data for the atmospheric forcing and the WGHC climatology as the starting input file. I've added a perturbation to both the exf forcings and the wghc climtologies. Using either both the input and the forcing perturbations, or just the forcing perturbation, the model crashes. Just before the crash, the advcfl_vel_max parameters diverge to a large number. Looking at the output, the cell that crashes is a cell right near a land / sea boundary, in more of a sea rather than open, fully deep ocean. The time steps and viscosity is the same as the previous runs that didn't crash, only thing changed was the input files.
> 
> Does anyone here know what might be the cause of this? I suppose I could try a very low time step (or maybe just lower the momentum time step) and then raise the time step after a few model days? The current time step I have 300 seconds (for both the momentum and tracer equations) and I am using the Leith viscosity method.
> 
> Anyhow, any suggestions on what to look for / debug would be appreciated.
> 
> Thanks,
> 
> Neil
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support