[MITgcm-support] Re: nonHydrostatic

Dimitris Menemenlis menemenlis at sbcglobal.net
Thu Aug 9 03:53:59 EDT 2007


Geof, maybe the explanation for your puzzle lies in truncation error and a 
marginally stable configuration.  Out of curiosity I ran a similar test to yours 
in the lab_sea configuration, after turning off all packages and surface forcing 
  and starting with salinity 35 and a stable but uniform vertical temperature 
profile.  First experiment is with diffKzS=0

> [dimitri at skylla build]$ grep dynstat_salt_ output.diffKzS0 | tail -5
> (PID.TID 0000.0001) %MON dynstat_salt_max             =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_min             =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_mean            =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_sd              =   1.4210854715202E-14
> (PID.TID 0000.0001) %MON dynstat_salt_del2            =   0.0000000000000E+00

and second experiment is with diffKzS=1e-5.

> [dimitri at skylla build]$ grep dynstat_salt_ output.diffKzS1 | tail -5
> (PID.TID 0000.0001) %MON dynstat_salt_max             =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_min             =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_mean            =   3.5000000000000E+01
> (PID.TID 0000.0001) %MON dynstat_salt_sd              =   6.5060044744523E-14
> (PID.TID 0000.0001) %MON dynstat_salt_del2            =   1.4898476717550E-15

So although salinity remains at 35 in both experiments, in the case with 
diffKzS=1e-5, its standard deviation is ever so slightly larger.  maybe this 
truncation error is sufficient to trigger the instabilities that you observe, 
particularly given the extremely fine resolution that you are using.

Dimitris


> Hi Dimitris,
> 
> Thank you for your help so far. By looking at someones elses data file, I tried to put the nonHydrostatic=.TRUE., and now, even if I have still some small velocities (10^-7m.s-1, compared to 1m.s-1 after 50000s), the results seems more reasonable. So it seems much better, even if I'm still not satisfied because there are some velocities. I will launch a simulation to see whether it still blows up after or not.
> You can find some pictures there :
> http://perso.crans.org/gaubry/dossier/stage/hydro.html
> Does it inspire you anything?
> Thank you,
> -- 
> Geoffroy



More information about the MITgcm-support mailing list