search

JCSDA-internal / soca

Sea-ice Ocean Coupled Assimilation
Apache License 2.0
13 stars 4 forks source link

running errors for the explicit diffusion operator #1045

Open hyunsookkim-NOAA opened 5 months ago

hyunsookkim-NOAA commented 5 months ago

Description

We tested the explicit diffusion operator for a 1/12-d and 55 z*layer regional MOM6. It failed and generated errors (see below):

FATAL from PE 160: write_energy : Ocean velocity has been truncated too many times. U_velocity_truncations

Time 2022 243 12.03 U-velocity violation at 479: 50 8 ( -7.56 E 44.25 N) Layers 47 to 47. dt = 200.0 Layers: 47 u(m): 1.690E+01 u(3): 1.063E+01 CFL u: 5.299E-01 CFL0 u: 5.299E-01 CAu: 7.225E-02 PFu: 3.106E+01 diffu: -1.498E-02 a: 1.093E-06 8.938E-03 hvel: 1.568E+02 Stress: 6.706E-03 dubt: -1.423E+01 ..... V_velocity_truncations

Time 2022 243 12.03 V-velocity violation at 470: 5 34 ( -45.04 E 45.75 N) Layers 47 to 47. dt = 200.0 Layers: 47 v(m): 1.671E+01 v(3): 1.051E+01 CFL v: 5.372E-01 CFL0 v: 5.377E-01 CAv: -7.782E-04 PFv: 3.187E+01 diffv: -7.862E-03 a: 1.189E-06 1.611E-01 hvel: 1.273E+02 Stress: 9.498E-04 dvbt: -1.486E+01

To Reproduce:

GNU compiler on Orion

travissluka commented 5 months ago

These are errors from the model running, not SOCA. There is something wrong with the restart file used to run your forecasts, check it to see what it looks like

hskang-aoml commented 5 months ago

I checked analysis fields (T,S & SSH) which were imported to the new restart file (after DA). They are closed at the boundaries. Now, they are opened at the boundaries with nco commands and forecast is running OK.

ocn.3dvar.an.*nc in SCRATCH/cycle/run.var/Data is an output file of "soca_var.x". Analysis fields in this file are closed at the boundaries, which is different from the ones in the bump_nicas case where they are opened at the boundaries.

So, with the explicit diffusion operator for the hat10 domain, a post process needs to be done to open the analysis fields at the boundaries.

It looks like that open boundaries are taken care of in "soca_var.x" with the bump_nicas case.