Improve constructors for various types by nefrathenrici · Pull Request #4236 · CliMA/ClimaAtmos.jl

nefrathenrici · 2026-01-26T20:13:45Z

This PR adds defaults and improves constructors for various types to facilitate runscripts for case setups:

Diagnostic/PrognosticEDMFX
RRTMGPI Modes
Hyperdiffusion (renamed from ClimaHyperdiffusion), new function cam_se_hyperdiffusion for CAM_SE hyperdiff
SmagorinskyLilly
Viscous/RayleighSponge
AtmosNumerics
EDMFXModel
tracer nonnegativity methods (+updated markdown docs on the topic)

The PR also:

enables passing z_mesh to grids
removed call_cloud_diagnostics_per_stage from default_model_callbacks(water::AtmosWater,...) since it's configured from the atmos model
added show-methods for various structs

The second commit of this PR adds:

a new runscript for the RCEMIP-II box CRM (runscripts/rcemipii_box_CRM.jl), providing a configurable setup for running the model with various grid resolutions and diagnostics.

szy21 · 2026-02-02T21:18:17Z

Yes, renaming to HyperDiffusion is good.

haakon-e · 2026-02-02T22:08:18Z

Yes, renaming to HyperDiffusion is good.

@szy21 should the yaml name be updated to "hyperdiffusion", too?

edit: It looks a bit strange to set hyperdiff: hyperdiffusion though, since (to my understanding) this option requires you to specify the coefficients yourself, too. But no strong opinions on what's best here.

edit 2: How about updating the setting to:

hyperdiffusion:
  help: |
    Fourth-order hyperdiffusion. Options:
    - `custom_coefficients` (default): Use custom coefficients for hyperdiffusion: 
      set these with the `vorticity_hyperdiffusion_coefficient`, `hyperdiffusion_prandtl_number`, 
      `divergence_damping_factor` config options.
    - `CAM_SE` (Lauritzen et al. 2018): Use CAM_SE coefficients for hyperdiffusion
    - `~` (no hyperdiffusion): Disable hyperdiffusion
  value: "custom_coefficients"

szy21 · 2026-02-02T22:20:08Z

Yes, renaming to HyperDiffusion is good.

@szy21 should the yaml name be updated to "hyperdiffusion", too?

It looks a bit strange to set hyperdiff: hyperdiffusion though, since (to my understanding) this option requires you to specify the coefficients yourself, too. But no strong opinions on what's best here.

I think it's fine to keep hyperdiff (or it's consistent with others, e.g. we have vert_diff: VerticalDiffusion, even though it looks a bit strange). You don't need to specify the coefficients - they have default values.

haakon-e · 2026-02-02T23:20:17Z

src/parameterized_tendencies/radiation/RRTMGPInterface.jl

+                   AbstractRRTMGPMode
+    idealized_h2o::Bool = false
+    idealized_clouds::Bool = false
+    cloud::ACR = InteractiveCloudInRadiation()


@nefrathenrici I set this default to InteractiveCloudInRadiation() instead of nothing as you had done originally, which is consistent with the default_config.yml. Is that fine?

szy21

Looks good, but maybe @nefrathenrici wants to take another look. What's the difference between Base.@kwdef and @kwdef? We are using both at different places in the codebase now.

src/simulation/grids.jl

szy21 · 2026-02-02T23:22:02Z

src/simulation/AtmosSimulations.jl

                "diagnostics" => diagnostics,
                "netcdf_interpolation_num_points" => nothing,
                "netcdf_output_at_levels" => false,
+                "output_default_diagnostics" => false,


why do we want this to be false here?

There's some duplicate code for default diagnostics which we'll clean up later. If you look at the signature of this function, the argument use_default_diagnostics is a bool that determines if we add default diagnostics, which are then set in L45 onwards.

We set it to false on L89 to avoid adding these twice.

src/solver/types.jl

haakon-e

Adding some comments throughout the changed files to (hopefully) make review easier.

haakon-e · 2026-02-02T23:22:22Z

src/prognostic_equations/remaining_tendency.jl

just formatting changes here

haakon-e · 2026-02-02T23:23:52Z

src/parameterized_tendencies/sponge/rayleigh_sponge.jl

+αₘ(s::RayleighSponge, z, α) = ifelse(z > s.zd, α, zero(α))
+ζ_rayleigh(s::RayleighSponge, z, zmax) = sinpi((z - s.zd) / (zmax - s.zd) / 2)^2
+β_rayleigh_uₕ(s::RayleighSponge, z, zmax) = αₘ(s, z, s.α_uₕ) * ζ_rayleigh(s, z, zmax)
+β_rayleigh_u₃(s::RayleighSponge, z, zmax) = αₘ(s, z, s.α_w) * ζ_rayleigh(s, z, zmax)


renamed β_rayleigh_w to β_rayleigh_u₃, which seemed more consistent with β_rayleigh_uₕ, but did not update the struct field from α_w.

haakon-e · 2026-02-02T23:24:43Z

src/parameterized_tendencies/sponge/viscous_sponge.jl

reformatting to remove FTs and improve readability

haakon-e · 2026-02-02T23:25:07Z

src/prognostic_equations/implicit/implicit_tendency.jl

making use of rayleigh_sponge_tendency_u₃

haakon-e · 2026-02-02T23:25:32Z

src/simulation/AtmosSimulations.jl

    integrator::OD
 end

-Base.summary(::AtmosSimulation) = "AtmosSimulation"


Base.summary and Base.show is moved to show.jl file

haakon-e · 2026-02-02T23:41:16Z

src/solver/types.jl

+    κ₂::FT
 end

+ViscousSponge(params) = ViscousSponge(;


convenience constructor using ClimaParams parameter set.

haakon-e · 2026-02-02T23:42:56Z

src/solver/types.jl

 end
-Base.broadcastable(x::AtmosNumerics) = tuple(x)
-
-function Base.summary(io::IO, numerics::AtmosNumerics)


moved this function to show.jl

haakon-e · 2026-02-02T23:43:19Z

src/solver/types.jl


 Base.broadcastable(x::AtmosModel) = tuple(x)

-function Base.summary(io::IO, atmos::AtmosModel)


moved this to show.jl

haakon-e · 2026-02-02T23:46:50Z

src/utils/show.jl

this file contains two generic methods that attempts to print more helpful information about model structs. These generic methods are then (L100 onwards) applied to the Base.show methods for various model structs. This is by far not complete, and is something we update as needed.

Note that these are largely relevant for interactive sessions, not non-interactive runs.

haakon-e · 2026-02-02T23:47:10Z

src/ClimaAtmos.jl

updating some formatting and importing the show.jl file.

haakon-e · 2026-02-02T23:55:56Z

Base.@kwdef and @kwdef? We are using both at different places in the codebase now.

No difference. Previous versions of Julia didn't export @kwdef from Base. It is now exported, so there's no need to prefix @kwdef with Base., so I'm just removing where I see it.

- add user-friendly model constructors for: - tracer nonnegativity methods - prescribed vertical diffusion models - eddy viscosity models - sponge models (+ documentation & defaults) - radiation models (+ defaults) - hyperdiffusion - atmosnumerics - some formatting updates - enable passing z_mesh to commongrids - dont pass call_cloud_diagnostics_per_stage via callback_kwargs Co-authored-by: Haakon Ludvig Langeland Ervik <45243236+haakon-e@users.noreply.github.com>

nefrathenrici force-pushed the ne/refactor branch 5 times, most recently from d9b62bb to f86415a Compare January 30, 2026 18:19

nefrathenrici marked this pull request as ready for review January 30, 2026 18:21

nefrathenrici requested a review from szy21 January 30, 2026 18:22

nefrathenrici linked an issue Jan 30, 2026 that may be closed by this pull request

Types need to be cleaned up #4218

Closed

haakon-e self-requested a review January 30, 2026 22:28

haakon-e force-pushed the ne/refactor branch from f86415a to 1fb9b6b Compare February 2, 2026 21:49

haakon-e force-pushed the ne/refactor branch from 1fb9b6b to 36e3c2f Compare February 2, 2026 22:53

haakon-e mentioned this pull request Feb 2, 2026

rft: improve model interfaces, add runscript example #4254

Closed

haakon-e reviewed Feb 2, 2026

View reviewed changes

szy21 approved these changes Feb 2, 2026

View reviewed changes

haakon-e reviewed Feb 2, 2026

View reviewed changes

haakon-e force-pushed the ne/refactor branch 3 times, most recently from 6225cbf to bcda57f Compare February 3, 2026 01:33

haakon-e mentioned this pull request Feb 3, 2026

Updated the yaml output to reference the toml file in the output #4264

Open

1 task

nefrathenrici force-pushed the ne/refactor branch 2 times, most recently from fdb4d62 to 36ac9a9 Compare February 5, 2026 00:41

haakon-e force-pushed the ne/refactor branch from 36ac9a9 to 7575ce5 Compare February 5, 2026 03:11

nefrathenrici force-pushed the ne/refactor branch from 7575ce5 to 89c2432 Compare February 5, 2026 21:26

nefrathenrici force-pushed the ne/refactor branch 2 times, most recently from c4e9815 to 99da0eb Compare February 5, 2026 23:45

nefrathenrici enabled auto-merge February 5, 2026 23:46

nefrathenrici disabled auto-merge February 6, 2026 00:44

feat: add rcemip-ii runscript

2084b15

nefrathenrici force-pushed the ne/refactor branch from 99da0eb to 2084b15 Compare February 6, 2026 00:45

nefrathenrici enabled auto-merge February 6, 2026 00:45

nefrathenrici added this pull request to the merge queue Feb 6, 2026

Merged via the queue into main with commit 61344cd Feb 6, 2026
45 checks passed

nefrathenrici deleted the ne/refactor branch February 6, 2026 02:57


		Base.broadcastable(x::AtmosModel) = tuple(x)

		function Base.summary(io::IO, atmos::AtmosModel)

Conversation

nefrathenrici commented Jan 26, 2026 • edited by haakon-e Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szy21 commented Feb 2, 2026

Uh oh!

haakon-e commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

szy21 commented Feb 2, 2026

Uh oh!

haakon-e Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

szy21 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

haakon-e left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

haakon-e commented Feb 2, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

nefrathenrici commented Jan 26, 2026 •

edited by haakon-e

Loading

haakon-e commented Feb 2, 2026 •

edited

Loading

haakon-e Feb 2, 2026 •

edited

Loading