Specifying velocities within your domain

[rwdell]

DDF8_proposedgrid.pdf

(27 KiB) Downloaded 583 times

Dear ROMS friends,

After Dr. Arango detailed some of the difficulties associated with altering the periodic boundary conditions in ROMS (see post viewtopic.php?f=3&t=1877), I turned to thinking about another possible domain configuration that might allow me to study boundary layers over sloping bottoms without the interference of boundaries at either end of the slope. Right now, I'm thinking about a grid like the one in the plot I've attached. The idea would be to have closed boundaries, with a test section (on the left) and a narrow channel (on the right) to allow a recirculation of the boundary layer flow, separated by a narrow island. This way, I can have a net flux in the boundary layer in the text section, but allow that flux to recirculate without disrupting the boundary layer dynamics. I've tried running ROMS using this grid, but I'm having some problems:

(1) My island is currently a seamount, with whose top is at h=5m below the surface. I didn't make it break the surface because I wasn't exactly sure what I would need to do with masking if I did. Is this a bad idea? I've never used a coastal mask (I've only used ROMS in idealized domains to study bottom boundary layers), and I couldn't find any information on the online documentation about how it should be applied. So the question is: should I use a coast here?

(2) I would like to force fluid from the test section into the area behind the island through the opening indicated with an arrow. Is there a way I can specify a velocity within the domain? I only know how to specify the velocity on the boundaries, using u3dbc and related functions, but this would be specifying a barotropic velocity across several internal grid points.

(3) For some reason, when I run the model with the attached grid it integrates in a reasonable way for an hour or so (of model time), then blows up. This is almost 1000 timesteps, so is there any reason why ROMS might integrate happily for 1000 timesteps and only then blow up?

Thanks so much for all your help!

Cheers,
Rebecca

[nganju]

Hi Rebecca-
(1) If you break the surface (i.e. negative h), mask_rho needs to have a value of 0 instead of 1. If you are using ana_grid, you can specify the masked areas easily enough using ana_mask.h. For this domain, you should probably have the seamount break the surface.

(2) I don't know if you can do this using UV_PSOURCE, basically specifying a point source of momentum in the middle of the domain. Alternatively, could you provide a flow bc on the northeastern side of your domain, forcing the flow south in the deep basin?

(3) The best way to diagnose a blow-up is to save the fields in the restart file often enough, so you can see what is blowing up where. I would suspect you have a problem near the seamount, which may be fixed by masking...