Horizontal regridding and stabilizing CATKE features #2881
Conversation
glwagner
added
experimental feature 🧪
|
the current regriding tests fail... is this alarming? |
|
Haha, that's not good. |
…A/Oceananigans.jl into glw/more-general-regridding
|
merge main? |
…A/Oceananigans.jl into glw/more-general-regridding
|
@simone-silvestri im getting a lot of errors after merging your suggestions… can you take a look? |
Co-authored-by: Simone Silvestri <33547697+simone-silvestri@users.noreply.github.com>
|
I'll have a look tomorrow! |
| vi_diffusivity_fields = Tuple(diffusivity_fields[n] for n = 1:N if is_vertically_implicit(closure[n])) | ||
|
|
||
| # length(vi_closure) > 1 && error("More than one closure with VerticallyImplicitTimeDiscretization is not supported") |
There was a problem hiding this comment.
remove or is it still the case?
There was a problem hiding this comment.
Hmm, it's not tested and I suspect CATKE may not work with a tuple yet. But I think we should save for a future PR, this PR does enough already. I'll remove the comment
There was a problem hiding this comment.
Sorry let me clarify -- we test tupled closures with implicit vertical diffusion and they pass (I also tested myself and found that it works). CATKE is special though, I think there is a bug with the linear component (only CATKE has this). But I'd like to debug and implement a test in another PR.
This PR adds regridding in x (and there was previously untested regridding in y). Regridding is only supported for tracer fields (though it's not a huge amount of work to extend the schemes for any field).
I believe the regridding is conservative for
RectilinearGrid, and also conservative forLatitudeLongitudeGridwhen the grids are aligned (ie when all grid lines of the coarser of the two grids is shared with the finer grid).In the case of partial overlaps on curvilinear grids, the volumes of the overlaps are approximated and thus the regridding is not perfectly conservative. It could make sense to emit a warning in this case.
In principle, this should allow regridding in 3D by forming intermediate regridding in x, y, and z. We leave multi-dimensional regridding to the user for now.
This PR also adds a
minimum_turbulent_kinetic_energyandnegative_turbulent_kinetic_energy_damping_timescaletoCATKEVerticalDiffusivity, either of which can help stabilize 2D or 3D simulations with CATKE (where due to advection TKE may go negative).