update surface fluxes and update dependencies

[szy21]

Purpose

Use the new interface in SurfaceFluxes v0.15 to calculate surface fluxes. For the bucket simulation, call the turbulent_fluxes! in ClimaLand directly. Also updates dependencies.

To-do

Content

---

• I have read and checked the items on the review checklist.

[simone-silvestri]

Is this PR going to allow using a skin temperature?

[szy21]

Is this PR going to allow using a skin temperature?

I think we will mimimal changes in this PR to recover the current behavior of the coupler, as we want to merge this relatively fast. But we can allow using a skin temperature now. I want to do it in a separate PR.

[szy21]

working amip: https://buildkite.com/clima/climacoupler-coarse-nightly-amip/builds/763

[juliasloan25]

The src changes look reasonable to me! One thought is we might want to merge this after we resolve the nightly AMIP compat troubles, and figure out why it started failing with the larger coupling timestep

[juliasloan25]

should we rename these all to use the same variable names as SurfaceFluxes? Especially for F_turb_moisture -> evaporation

[szy21]

I'm fine with that. I guess it depends on whether the coupler wants some convention of variable names - e.g. we have F_turb_energy and F_turb_moisture. It will change many files though - maybe we can do it in a separate PR?

[szy21]

The src changes look reasonable to me! One thought is we might want to merge this after we resolve the nightly AMIP compat troubles, and figure out why it started failing with the larger coupling timestep

I was thinking about merging this: CliMA/ClimaAtmos.jl#4198 soon, as we want to merge it before working on another breaking change on thermodynamics in ClimaAtmos (and ClimaLand and ClimaCoupler). But it will break nightly AMIP. I can make a patch release of ClimaAtmos now, and a breaking release with the ClimaAtmos PR, so we can resolve the nightly AMIP compat with the patch release of ClimaAtmos?

I'm not too worried about the larger coupling timestep, but am happy to wait for a few days. Kat and I were not sure it's correct for integrated land - e.g. if the land timestep is 8 minutes and coupling timestep is 32 minutes, does the atmosphere get the flux at the end of the four land timesteps? This will cause some significant errors in energy conservation, although I don't know whether it will make things more unstable.

[szy21]

Actually, amip breaks this week. I'm leaning towards there might be some issue with ClimaLand v1.3. Let's talk next week about how to move forward?

[juliasloan25]

Dependency updates in this PR:

  [13f3f980] ↑ CairoMakie v0.12.18 ⇒ v0.15.8
  [b2c96348] ~ ClimaAtmos v0.33.2 ⇒ v0.34.0
  [d414da3d] ↑ ClimaCore v0.14.45 ⇒ v0.14.46
  [908f55d8] ↑ ClimaCoreMakie v0.4.6 ⇒ v0.4.7
  [08f4d4ce] ~ ClimaLand v1.3.0 ⇒ v1.4.0
  [5c42b081] ↑ ClimaParams v1.0.9 ⇒ v1.0.10
⌃ [6ba0ff68] ↑ ClimaSeaIce v0.3.10 ⇒ v0.3.12
  [aa8a2aa5] ↑ EnsembleKalmanProcesses v2.6.0 ⇒ v2.6.1
  [db073c08] ↑ GeoMakie v0.7.12 ⇒ v0.7.16
  [ee78f7c6] ↑ Makie v0.21.18 ⇒ v0.24.8
⌃ [9e8cae18] ↑ Oceananigans v0.100.7 ⇒ v0.103.1

[juliasloan25]

CMIP runs out of GPU parameter memory on the P100s with this Oceananigans update. It still runs fine on A100s. I'll make the CI runs soft fail so we can at least continue working on AMIP while we figure out a solution for CMIP.

See CliMA/Oceananigans.jl#5191