[MITgcm-support] Small timestep - rapid overadjustment

[Martin Losch]

Hi Chris,

I don't know enough about your configuration, but the only time dependent stability parameter that I can think of is for the horizontal viscosity: viscAh*deltaT/deltaX^2. Maybe this has some effect (not on the stability, as you decrease you time step, but maybe on the solution). Did you try using viscAhgrid (value of <1, see mom_calc_visc.F or the online documentation for more details) instead of viscAh?

Martin

On Jan 4, 2013, at 8:33 PM, Chris Horvat <horvat at fas.harvard.edu> wrote:

> Hi, 
> 
> I'm running a reentrant ocean channel configuration (120km by 40km), and OBCS is setting temperatures on the north and sound boundaries in order to force a temperature gradient. Initial conditions interpolate between the boundaries. It is an Arctic simulation, so the deep water is warmer than the top water. (Temp. profile at http://imgur.com/4RZ0X)
> 
> 
> Here's the problem. If I set my timestep to be too large (~3600 s), numerical instability kills me in a few steps. If I set smaller (~600 s), I can get much further before I run into a similar numerical error. (maybe 2 mo.s), and the first thing that happens, as expected, is the interior begins to relax to match the imposed gradient. (seen at http://imgur.com/Ofo5Q)
> 
> So my thought is, setting deltaT = 30s would help... but instead what I see is an incredibly unrealistic change in the temperature profile. The gradient is immediately wiped out in favor of extreme gradients on either side of the channel with weak interior gradient. (Seen at http://imgur.com/Od1lM)
> 
> Could someone give me some guidance as to what might be causing this?
> 
> Thanks!
> 
> Chris
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support