wave-like disturbances in velocity forming near edges

[sonaljit.m]

Hello.

In the previous thread, I discussed about my runs blowing up due to unrealistic velocities forming along the bathymetry.

Following this, I tried to make my runs more simpler by removing any variability in the bathymetry, and initializing with constant quantities everywhere in the domain. This is just an idealistic run with no spatial variabilities. Details are below:

- Temperature (and SST) is constant at 21.0 everywhere.
- Salinity (and SSS) is constant at 34.0 everywhere.
- velocities are constant at 0.001 m/s everywhere (including u,v,ubar and vbar).
- Bottom bathymetry is constant to 3000 m everywhere in the domain.
- Removed any land masking (since simulation is idealized).
- Fluxes are 0.0 everywhere.

I tried two boundary conditions - (i) closed on all walls, and (ii) periodic on east/west walls.

My preprocessor flags are as below:

#define UV_ADV
#define UV_C4ADVECTION
#define UV_COR
#define UV_QDRAG
#define LIMIT_BSTRESS
#define UV_VIS4
#define MIX_GEO_UV

#define CURVGRID
#define NONLIN_EOS
#define PJ_GRADPQ4

#define TS_C4HADVECTION
#define TS_C4VADVECTION
#define MIX_GEO_TS
#define TS_DIF4

#define SALINITY
#define SOLVE3D
/*#define MASKING*/

#define VISC_GRID
#define DIFF_GRID

#define GLS_MIXING
#define CANUTO_A
#define CHARNOK

It turns out that these idealized runs are blowing up as well. The error messages are:

Found Error: 01 Line: 302 Source: ROMS/Nonlinear/main3d.F

ROMS/TOMS - Output NetCDF summary for Grid 01:
Found Error: 01 Line: 350 Source: ROMS/Utility/close_io.F

ROMS/TOMS - Blows up ................ exit_flag: 1


After a few time steps, very large unrealistic velocities are forming near the boundaries at the bottom, as shown in plot below:

The regions where the unrealistic values are happening, are marked and zoomed. There is a wave-like pattern in these disturbances, occurring along the adjacent grid points.

Below, is a vertical transect plot.

As you can see, these disturbances are forming near the bottom only.

Now, for the bottom drag I am using quadratic drag, and there are no bottom fluxes (just like there are 0 surface fluxes).

I previously thought that these were due to some problem with the bathymetry. But it turns out that bathymetry is not the issue. Has anyone come across these problems before? Any suggestions?

[kate]

And your timesteps are? Both DT and NDTFAST? How about the CFL report from ROMS?

[sonaljit.m]

DT 30 seconds.
NDTFAST 60.
horizontal resolution is 1000 m.

This is a plot of the u velocity along the longitude, at the bottom.

These zig-zag patterns are forming next to the boundary.

With shorter time steps, the simulation is running a bit further, but eventually blowing up again.