Tomás Chor
Tomás Chor
I ran many tests today, and I found that the only way that I could somewhat reliably run simulations was to switch the preconditioner to what it was before https://github.com/CliMA/Oceananigans.jl/pull/4041....
> Oh wait maybe it works with the latest commit but I definitely do not think it is correct and it will be wrong in some scenarios e.g. when the...
I think this PR is ready for review. This makes simulations with `PerturbationAdvection` OBC run as well as they did before https://github.com/CliMA/Oceananigans.jl/pull/4041. The key to make that happen is to...
> Also I ran your example as is and I can not reproduce the problem, I get zero mass flux. > > Update: I checked and I had the code...
> The only interestying thing I've noticed is that if you set the in and outflow timescales to be the same then the error goes down to ~10^-12 and if...
> You can start by allowing the coefficient to be an array as in AMD: > > https://github.com/CliMA/Oceananigans.jl/blob/7f2d512ceaa2ce3fec3132cf0b08ab39344270d9/src/TurbulenceClosures/turbulence_closure_implementations/anisotropic_minimum_dissipation.jl#L143 > > Then with a callback you can implement any model, including...
> Here's a first attempt at Lagrangian-averaged dynamic Smagorinsky: https://github.com/CliMA/Oceananigans.jl/tree/ss-xk/dynamic-smagorinsky It doesn't run due to `StackOverflorError`, likely due to the way we are taking the averages. I think if we...
> Why don't you implement `DynamicSmagorinsky` and abstract the notion of the `Averaging` to `LagrangianAveraging` and `DirectionalAveraging`. > > Note, you should expect to spend your time validating the closures...
> Nice work! I'm curious about the criteria. Should it be something like > > ```julia > dt = 10 * eps(dt) * sim.dt > ``` > > ? Or...