[MITgcm-support] change deltaT

Jody Klymak jklymak at uvic.ca
Mon Aug 14 13:50:12 EDT 2017


DeltaT needs to be set to satisfy [the CFL 
condition](https://en.wikipedia.org/wiki/Courant–Friedrichs–Lewy_condition), 
ie. dt < dx/c where c is the speed of the fastest signals in your model, 
typically the mode-1 phase speed if surface waves aren’t being used.  
So, the # of grid cells only matters insofar as if your domain size 
doesn’t change, dx will increase/decrease, but it is dx that is the 
fundamental parameter.

In the typical ocean c<4 m/s, so 60s should be good down to something 
like 500-m lateral resolution.

You also need to be cognizant of dz values, particularly in cells with 
hFac<<1.  Thats why there is a hFacMin parameter - if you set this to 
something ridiculously small you can get w > dz/dt.

Its possible that a model run started from rest will have a strong 
initial adjustment period where you need a small deltaT, but then can 
increase it substantially after a few inertial periods (by restarting 
from a pickup file).

Cheers,  Jody



On 14 Aug 2017, at 10:35, rahele shafiee wrote:

> Hello everyone
>
> I'm running a  modell with grid points 510*450*32 for 1  year. 
>  I've chosen DeltaT=60, but the runtime is too long.
>
> I selected a experimental test with grid points 10*12*2 and DeltaT=60 
>  and I run for 1 year and  lasted 11 minutes, but  with 
> DeltaT=108000  it took a second to run. The same test with 
> DeltaT=108000 for 5 years lasted 21 minutes and with 
> DeltaT=5*108000=540000 lasted a second.
>
> How can I reduce the runtime of the model  by modifying  DeltaT? Is 
> it possible to change DeltaT unstable the model after some time?
>
> Thanks for your guidance
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mailman.mitgcm.org/mailman/listinfo/mitgcm-support
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.mitgcm.org/pipermail/mitgcm-support/attachments/20170814/f9978c60/attachment.html>


More information about the MITgcm-support mailing list