Unstructured Grid Support in Oceananigans.jl

[trinefrisbaek]

Hi,
I’m very impressed by the Climate Modeling Alliance’s development of Oceananigans and its related components. I am thinking of using the model for simulations in Danish coastal regions. Given the many fjords and complex bathymetry, most existing models for these areas rely on unstructured grids. I was wondering whether there are any plans—either in the near or distant future—to support unstructured grids in Oceananigans.jl. Or if the model’s current structure is such that implementing unstructured grids would be fundamentally infeasible?

Any insights would be really appreciated.
Thank you.

[glwagner]

It's not infeasible of course, since many unstructured models have been developed. But the effort required would essentially be a rewriting of the entire core of the code (most of the user interface could probably still be used, but I'm actually not 100% sure about that).

Note that we already omit computations in land regions when using the active_cells_map. See section 3.2 of Silvestri et al.

There is the additional possibility of saving the memory allocated in land regions (in addition to the computation). Basically this means extending the active_cell_map concept to the underlying arrays in Field.

This still does not have the benefit of aligning cell boundaries with the geometry, though.

Of course, before doing any of that, it would be advisable to try it out first!