Problem with periodic boundary condition

[agomer]

Hi,
I'm new to ROMS and I'm using ROMS Agrif right now. I'm running a very idelized simulation of a hurricane passing over a square flat basin on a f-plane. All the fluxes (heat, fresh water, bottom fluxes...) are set equal to zero, a part from the surface momentum flux, which is defined analytically.
On the North and South boundaries there's a wall, while the basin is defined to be periodic in the East-West direction.

The problem is that it seems that there's downwelling along all the four boundaries, starting from the east part of the North and South boundaries and then propagating everywhere (but only in a pretty thin layer close to the boundaries).

Attached there are few frames of an horizontal section of the temperature at 100 m depth.

Does anyone have any ideas where this problem could be coming from? I can provide more info and files if that would help. Any suggestions would be very much appreciated!

[kate]

On the North and South boundaries there's a wall, while the basin is defined to be periodic in the East-West direction.

Clearly, you don't actually have periodic boundaries. I don't know how AGRIF decides these things, but if it is compiled in, you might need a "make clean" and a fresh build. Can you see the code after cpp has gone through? You can check at that phase for the periodic code.

[agomer]

I understand that it seems that there's no periodicity in the EW direction in the pictures that I've attached, by here I post two other pictures (from the same simulation), where it is more evident. The first is the horizontal u-velocity field at the same depth (100 m) and the second is again the temperature field, but at an earlier instant.
Moreover, it seems that in the code the periodicity is defined and in the .out file, the list of cpp flags activated contains the EW_PERIODIC:


Simple Rankine example Big5
506 ntimes Total number of timesteps for 3D equations.
2160.00 dt Timestep [sec] for 3D equations
60 ndtfast Number of 2D timesteps within each 3D step.
1 ninfo Number of timesteps between runtime diagnostics.

5.000E+00 theta_s S-coordinate surface control parameter.
0.000E+00 theta_b S-coordinate bottom control parameter.
1.000E+01 Tcline S-coordinate surface/bottom layer width used in
vertical coordinate stretching, meters.
Restart File: big5_rst.nc nrst = 3600 rec/file: -1
History File: big5_his2.nc Create new: T nwrt = 2 rec/file = 0

Fields to be saved in history file: (T/F)
T write zeta free-surface.
T write UBAR 2D U-momentum component.
T write VBAR 2D V-momentum component.
T write U 3D U-momentum component.
T write V 3D V-momentum component.
T write T( 1) Tracer of index 1.

F write RHO Density anomaly.
F write Omega Omega vertical velocity.
T write W True vertical velocity.
F write Akv Vertical viscosity.
F write Akt Vertical diffusivity for temperature.
F write Hbl Depth of KPP-model boundary layer.
F write Bostr Bottom Stress.
F write Wstress Wind Stress.

T write U-Wstress comp. U-Wind Stress.
T write V-Wstress comp. V-Wind Stress.
F write Shflx [W/m2] Surface net heat flux
F write Shflx_rsw [W/m2] Short-wave surface radiation

1025.0000 rho0 Boussinesq approximation mean density, kg/m3.
3.000E-04 rdrg Linear bottom drag coefficient (m/si).
0.000E+00 rdrg2 Quadratic bottom drag coefficient.
0.000E+00 Zob Bottom roughness for logarithmic law (m).
0.000E+00 Cdb_min Minimum bottom drag coefficient.
0.000E+00 Cdb_max Maximum bottom drag coefficient.

1.00 gamma2 Slipperiness parameter: free-slip +1, or no-slip -1.

WARNING: Unrecognized keyword: lin_EOS_cff --> DISREGARDED.

0.000E+00 visc2 Horizontal Laplacian mixing coefficient [m2/s]
for momentum.
0.000E+00 tnu2( 1) Horizontal Laplacian mixing coefficient (m2/s)
for tracer 1.

Activated C-preprocessing Options:

RANKINE
SOLVE3D
UV_COR
UV_ADV
NONLIN_EOS
UV_VIS2
UV_MIX_S
TS_MIX_S
TS_DIF2
LMD_MIXING
LMD_SKPP
LMD_SKPP2005
LMD_RIMIX
LMD_CONVEC
ANA_BSFLUX
ANA_BTFLUX
ANA_GRID
ANA_INITIAL
ANA_SMFLUX
ANA_STFLUX
EW_PERIODIC
ANA_SRFLUX
ANA_SSFLUX
ANA_SST
VAR_RHO_2D
M2FILTER_POWER
UV_VADV_SPLINES
NTRA_T3DMIX
TS_VADV_AKIMA
LIMIT_BSTRESS
ANA_BRY_WKB
DBLEPREC
Linux
QUAD
QuadZero
GLOBAL_2D_ARRAY
GLOBAL_1D_ARRAYXI
GLOBAL_1D_ARRAYETA
START_2D_ARRAY
START_1D_ARRAYXI
START_1D_ARRAYETA
PRIVATE_1D_SCRATCH_ARRAY
PRIVATE_2D_SCRATCH_ARRAY
PRIVATE_1DXI_SCRATCH_ARRAY
PRIVATE_1DETA_SCRATCH_ARRAY
WESTERN_EDGE
EASTERN_EDGE
SOUTHERN_EDGE
NORTHERN_EDGE
MYID
MPI_master_only
ZEROTH_TILE
SINGLE_TILE_MODE
LF_AM_STEP
FIRST_TIME_STEP
FIRST_2D_STEP
NOT_LAST_2D_STEP
PUT_GRID_INTO_RESTART
PUT_GRID_INTO_HISTORY
PUT_GRID_INTO_AVERAGES
NF_FTYPE
nf_get_att_FTYPE
nf_put_att_FTYPE
nf_get_var1_FTYPE
nf_put_var1_FTYPE
nf_get_vara_FTYPE
nf_put_vara_FTYPE
nf_put_var_FTYPE
NF_FOUT
fast_indx_out


NUMBER OF THREADS: 1 BLOCKING: 1 x 1.

hmin hmax grdmin grdmax Cu_min Cu_max
5000.000000 5000.000000 .200000000E+05 .200000000E+05 0.56377655 0.56377655
volume=8.000000000000000000000E+16 open_cross=0.000000000000000000000E+00

[kate]

You also have EASTERN_EDGE and WESTERN_EDGE. Are they the same as EASTERN_WALL and WESTERN_WALL?

[agomer]

No, they are not the same thing.
EASTERN_EDGE and WESTERN_EDGE should be the regions at the edges of the domain, where variables are exchanged in case of periodic conditions.

But I am not sure, does anyone know more about that?