zeta gradually increasing over the entire surface

[sonaljit.m]

Hello,
I am running an idealized simulation in ROMS using an uneven terrain grid that represents the VI-Puerto Rico bathymetry. I smoothed the bathymetry from the original ETOPO 1km version, such that rx0 = 0.16 and rx1 (Haney number) = 3.92. The bathymetry plot is shown below. (This simulation does not have any land-sea masking)

bathy.png (64.44 KiB) Viewed 2334 times

The initial condition consists of a vertical temperature gradient of 0.0015 degrees, and the surface temperature is 30 C (in which case, the temperature at 4000 m deep would be 24.4 C). There is no horizontal gradient. Salinity is constant everywhere at 33 PSU. The velocities are initialized to 0.0 everywhere as well. The sea-surface height zeta is also initialized to 0. Also, the heat flux and wind stresses are 0.

Boundary condition used here is Radiation (rad) on all walls. I am not using any nudging. The climatology file provided is also idealized with exactly the same temperature, velocities, zeta, etc. In other words, none of the properties change over time.

Below, I am showing timelapse videos of zeta, pressure gradient (u_prsgrd), and the u velocity at the surface:

zeta: https://youtu.be/IoZnvpsz7tg
u_prsgrd: https://youtu.be/A1JJJBs82D4
u: https://youtu.be/InZVe2HAo50

Notice that zeta is gradually increasing over the entire area; it’s magnitude rising from 0 to as high as 6 m. Also, the u velocity (as well as v velocity) rises to near 2 m/s at the end of 100 time steps, which is obviously not realistic.

My question is – Why would zeta rise over time, given that there is no influx from the boundaries? Is this an artifact of the pressure gradient anomaly due to the sigma grid? Can this anomaly be removed, or at least minimized by some means? Is more smoothing necessary?

Thanks,
Sonaljit.

[dcherian]

Open boundaries are not well posed. I found that I had to nudge at the open boundaries to conserve volume in the domain. I don't think this is a pressure gradient problem, but an open boundaries problem.

The same simulation with closed boundaries should not have this trouble and you can judge the magnitude of pressure gradient error.