I am facing a problem with my CROCO setup, I was hoping if anyone familiar with this problem might have a possible solution. I am trying to simulate an idealized problem of an island within the open ocean. The island is elliptical shaped, as shown here in green landmask:
The min depth is 10 m, and steeps down to 1500 m towards the open ocean. The grid details are given below.
- Horizontal resolution: 500 m
- Min depth 10 m, max depth 1500 m
- Haidvogel number: 0.203, Haney number: 2.96
- theta_s = 5, theta_b = 0.0, Tcline = 10.0
Based on what I’ve read, the Haidvogel no should be < 0.25 and Haney no should be < 3.0, in order to give stable results.
I initialized it with climatology from HYCOM 2018 and forcing from ECMWF.
I started with a time step dt of 10 sec and NDTFAST 120, but it is blowing up within 5 steps. Then I reduced the time step to 2 sec keeping NDTFAST the same, but it is still blowing up after 66 sec. Here is a timelapse video showing the zonal velocity transect and the SSH at each timestep till it blows up:
video timelapse attached
I noticed that around the 54th time step, the velocity forms zig-zag patterns near the landmask. These patterns amplify to unrealistically large values. Also the SSH forms such patterns in the same area, and blows up eventually.
I further reduced the time step to 1 sec and it seems to run fine without blowup (1 sec timelapse). But running the model at such small time steps is quite unrealistic, so I am wondering what am I doing wrong here?
Here is a list of preprocessors I am using for my setup: