Couple tightly ocean and sea ice

[simone-silvestri]

This PR aims to couple tightly sea ice and ocean in stages rather than full timesteps. It is still a work in progress (I have just checked it works very nicely for the simple example shown in this PR which I will probably remove) because I still need to modify the interface to make sure that the coupling works with all the rest of the infrastructure (for example callbacks in the two specific components)

A video from the example, showing the difference between AB2 + FE with 5 minutes timestep and RK3 + RK3 with per-stage coupling but a 15 minutes timestep (so coupling expenses are equal to the AB2 - FE)

coupling_comparison_sst_sss.mp4

[codecov]

Codecov Report

❌ Patch coverage is 7.40741% with 50 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/EarthSystemModels/rk_coupled_ocean_sea_ice.jl	0.00%	41 Missing ⚠️
src/SeaIces/sea_ice_simulation.jl	0.00%	8 Missing ⚠️
.../EarthSystemModels/time_step_earth_system_model.jl	80.00%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[glwagner]

nice

[simone-silvestri]

We can use the same strategy also for breeze - ocean coupling when we use an RK3 for both

[glwagner]

We can use the same strategy also for breeze - ocean coupling when we use an RK3 for both

that's already the default, although one wrinkle is we have ssprk3 for the atmosphere. maybe the flavor of rk3 doesn't matter though

[simone-silvestri]

Because of a bug, I haven't realized this was just RK3+RK3 vs AB2+FE but the RK3 were not coupled at each stage, just at the end of the timestep (the method was not dispatched upon because EarthSystemModel holds simulations, not models).
I will fix it and try again the test case

[glwagner]

Because of a bug, I haven't realized this was just RK3+RK3 vs AB2+FE but the RK3 were not coupled at each stage, just at the end of the timestep (the method was not dispatched upon because EarthSystemModel holds simulations, not models). I will fix it and try again the test case

interesting, so just using RK3 fixes the noise issue. Maybe another test case will reveal the advantage of stage vs step coupling.