Buoyancy-driven flow over complex topography?

[raphaelouillon]

Hello,

A few months ago I set up an Oceananigans simulation of a gravity current interacting with a background current (with much help from the dev team) and was very encouraged by the results. One thing that I think I remember not being possible at the time was setting up a non-flat bottom boundary, which was the next important step in the physical problem I was interested in investigating. From looking at the recent Issues and Discussions, it seems like it may now be possible to run simulations with an immersed boundary. Is it feasible to setup a non-hydrostatic simulation of a buoyancy-driven flow (gravity current) over complex topography? By complex topography, I mean gentle hills, but with a topography that is imported from some existing bathymetry file.

Thanks very much for sharing thoughts!

[glwagner]

@raphaelouillon it's possible to set up a nonhydrostatic simulation with an immersed boundary, however because we don't have an exact pressure solver for complex domains, the resulting simulation makes small errors in the impenetrable boundary condition (there is some leakage). In some situations this leakage is small, but probably the errors would need to be evaluated. Our support for non-insulating / free-slip boundary conditions is also limited. All of these things can be fixed with a little development if you'd like to take it up! I'd be happy to help setup an experimental simulation, too.

Here's one such "experimental" simulation:

flow_over_wedge_Nz512_high_Re.mp4

Boundary conditions can be achieved two ways: either (i) refactoring TurbulenceClosures to permit fluxes across immersed boundaries (a lot of work, but probably necessary eventually), or (ii) adding a forcing term or Callback that adds the necessary boundary fluxes (still difficult, but less work).

Developing a new efficient pressure solver is an interesting problem. One idea is to use the existing FFTBasedPressureSolver as a preconditioner in a conjugate gradient iteration. Since we have both a conjugate gradient solver and an fft-based pressure solver, solving that is mostly a matter of assembling tools and writing tests.

[raphaelouillon]

@glwagner in that case please go ahead, it'll take me a while to re-familiarize with the code so thanks a bunch for the help.

[glwagner]

@raphaelouillon do you have Slack / are you on the Julia slack? If so we can chat there!

[raphaelouillon]

@glwagner I do. Is that the MIT JuliaLab Slack, or a public workspace?

[iuryt]

@glwagner
Can you share the code for this experimental simulation?

[glwagner]

@iuryt it's here: https://github.com/CliMA/Oceananigans.jl/blob/glw/diffusion-driven-flow/validation/immersed_boundaries/flow_over_wedge.jl