Open anton-seaice opened 3 weeks ago
I reported the shortage of info in the stack trace here: https://github.com/ACCESS-NRI/ACCESS-OM3/issues/13
I will set the DT_THERM back to 1350s and try again :)
With DT_THERM set to the default, the Bgrid configuration is still running (almost 2 years).
With the Cgrid (https://github.com/ACCESS-NRI/access-om3-configs/compare/34e2136cab181fd38068d9e35bbc85f3f9de9db9...anton-seaice:MOM6-CICE6:dev-025deg_jra55do_ryf-cgrid?expand=1), I get the errors with eta dropped below bathyT
The errors are :
WARNING from PE 1036: btstep: eta has dropped below bathyT: -1.1825361824810603E+01 vs. -1.1805749893188477E+01 at -1.9538E+02 5.8943E+01 340 793
and
WARNING from PE 1036: Extreme surface sfc_state detected: i= 332 j= 788 lon=-197.125 lat= 58.423 x=-197.125 y= 58.423 D= 6.5867E+01 SSH=-6.2522E+02 SST=-1.7967E+00 SSS= 3.3376E+01 U-= 4.8522E+00 U+= 4.8522E+00 V-= 4.8793E+00 V+=-4.8252E+00
It doesn't say which extreme sfc_state detected is causing the issue, but presumably SSH.
There is a bit of a "shelf?" in the crash location (a nameless bay, western bering sea):
But nothing else obvious going on - I looked at daily mean of SSH, and sea-ice concentration.
I can't see how turning on C-grid in CICE has impacted this ? Should I turn on time-step output for SSH and re-run (or is eta
saved some other way ? )
SSH=-6.2522E+02
is pretty extreme! This indicates numerical instability. Have you tried a shorter barotropic or baroclinic timestep?
From the messages it looks like the error might originate in the barotropic solver, so you could try reducing the magnitude of DTBT
, and/or try setting DTBT_RESET_PERIOD=0.0
.
Following recent discussions , I ran the 025deg ryf with
Diabatic_first=True Dt_therm = 5400s (1.5 hours)
and otherwise the default configuration (the om2 grid / bathymetry / initial conditions / cice b-grid)
It ran until the 18th March in the second year
And crashed with too many truncations :
Is this consistent with your work @minghangli-uni - should I increase the number of allowable truncations and see if they stabilise?
Config is here: https://github.com/ACCESS-NRI/access-om3-configs/compare/34e2136...51fbffa1d152d7773093c9d26924fd0d08fa5880 Output is in:
/g/data/tm70/as2285/payu/dev-025deg_jra55do_ryf-bgrid/work