Zeta following the bathymetry and velocity profiles

[julian.kuhlmann]

My ROMS application with one-way nesting into a global ocean model and realistic atmospheric forcing runs in a stable way, but the results are a bit unrealistic. In particular, the SSH (variable zeta) and the horizontal velocities closely follow the bathymetry (which I smoothed already to avoid such artifacts). Additionally, and possibly linked, the velocity profiles within the domain are rather homogeneous in the vertical dimension, meaning high velocities (~0.5 m/s) in great depths (~3000-4000 m), which is probably not very realistic. I modified the vertical mixing of momentum (AKV_BAK), the Charnock parameter (CHARNOK_ALPHA), and switched between different GLS mixing schemes (in particular k-epsilon and gen), and switched from mixing along s surfaces to mixing along geopotential surfaces, but it all does not change the results very much. What other parameters or CPP options could I change to make my model behave more realistically? Or has anybody encountered similar problems?

[rduran]

If you are running without forcing and just an arbitrary horizontally-constant stratification then it is the pressure gradient error - not necessarily surprising.

If this is a regular run, you have baroctropic flow following f/h contours - not necessarily surprising.

Granted that 0.5m/s 3km deep is something to look into but I believe we need more info of what you are doing to be able suggest specifically.

Then again, it could be that some smoothing of bathymetry did not solve your pressure-gradient problem:
you might need to smooth some more.
you might need to decrease \delta x
you might need to increase \delta \sigma (i.e. less sigma levels or different stretching).

The only way to see if your bathymetry is good enough (i.e. you can live with the error) is running the typical unforced test, the a-priori estimates (r-factors) won't tell you the whole story (if that is what is going on).

[julian.kuhlmann]

Thanks for the response! I am running with reanalysis forcing. In runs without forcing, there is the same effect but smaller. I doubt that the bathymetry is the problem, since there have been other studies with a similar setup. Maybe this is just a meaningful effect simulated by the model that for some reason is not so apparent in satellite altimetry. I'll keep investigating...