Richardson-number stability weight on the geometric SGS variance term - #4837
Merged
Merged
Conversation
Contributor
|
📖 Docs preview for this PR: https://clima.github.io/ClimaAtmos.jl/previews/PR4837/ |
szy21
force-pushed
the
zs/sgs_ri_weight
branch
from
September 18, 2026 02:42
fe0b7e9 to
20cebce
Compare
szy21
marked this pull request as ready for review
September 18, 2026 02:42
szy21
force-pushed
the
zs/sgs_ri_weight
branch
2 times, most recently
from
September 18, 2026 02:53
ba1c3f2 to
e03c7b5
Compare
szy21
force-pushed
the
zs/sgs_ri_weight
branch
from
September 18, 2026 03:52
e03c7b5 to
a1f2abc
Compare
costachris
approved these changes
Sep 18, 2026
szy21
force-pushed
the
zs/sgs_ri_weight
branch
from
September 18, 2026 05:53
8cffba5 to
ab42703
Compare
Multiply the horizontal resolved-gradient (geometric) SGS variance term of Ri₀ = k Ri_crit, with N² the saturated moist buoyancy gradient in every cell (the cached buoyancy-gradient coefficients evaluated with cloud fraction 1) and S² the cached strain-rate norm. The term then acts only in air that is absolutely stable to a saturated displacement and not turbulent, where the turbulence closure cannot carry the variance; in mixed and conditionally unstable layers it is faded out instead of being counted on top of the closure. - new provisional parameter `sgs_variance_geometric_Ri_factor` (k; default 0 → weight exactly 1, default path bitwise identical to main) - the applied weight is cached in `p.precomputed.ᶜgeo_weight` (1 = term on and unweighted, 0 = term off) and exposed as the `sgs_geo_weight` diagnostic - evaluated after `materialized_mixing_length!`, which fills `ᶜstrain_rate_norm` on the non-EDMF path - pointwise tests of the weight, parameter and diagnostic test entries, NEWS, closures docs table Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
szy21
force-pushed
the
zs/sgs_ri_weight
branch
from
September 18, 2026 18:29
ab42703 to
a9d3a66
Compare
szy21
enabled auto-merge
September 18, 2026 18:30
szy21
added a commit
that referenced
this pull request
Sep 19, 2026
…tion for the geometric SGS variance term
Options on the horizontal resolved-gradient (geometric) SGS variance term, all
defaulting to the merged behaviour (default and unweighted `tq` paths unchanged):
- `sgs_variance_horizontal_form: isentropic` (IsentropicHorizontalVariance): q′q′
gains geo_h |∇_h q_tot − r ∇_h θ_li|², the squared gradient of q_tot along the
θ_li surface, with r = clamp((∂q/∂z)(∂θ/∂z)/((∂θ/∂z)² + θz_min²), ±r_cap) the
regularised, capped vertical q/θ_li ratio (`sgs_variance_isentropic_min_dtheta_dz`
= 1e-3 K/m, `sgs_variance_isentropic_slope_cap` = 5e-4 kg/kg/K). No θ′θ′
geometric term. The term carries the validity weight 1[N² > 0] on the saturated
moist N² of the Richardson weight, so it vanishes where the layer is
moist-neutral or unstable. The three DSS'd gradient invariants come from one
evaluation of each gradient (`hgrad_invariants_pair!`, second C12 scratch buffer
allocated for this form only); θz_min > 0 is checked at configuration time.
- `sgs_variance_element_filter: linear` (ElementLinearFilter): the gradient
invariants are replaced by their element-linear (lumped GLL{2}) restriction
(`element_linear!`), removing the element-boundary enhancement of the nodal
spectral-element gradient while conserving each element's WJ-weighted mean.
- `sgs_variance_geometric_tke_scale` (tke₀, default 0 = off): the term is
multiplied by tke₀/(tke₀ + max(tke, 0)) on the prognostic EDMF TKE
(`sgs_geometric_tke_weight`), fading it where the closure reports an active
turbulent layer. Requires prognostic TKE (checked in `get_atmos` and in
`set_covariance_cache!`); multiplies the Richardson weight when both are set.
- `tq_correlation_model: diagnosed` (DiagnosedTqCorrelation): the T–q correlation
sampled by the quadrature is diagnosed per cell from the gradient covariance
T′q′ (the vertical closure's cross term plus the geometric cross term
∇_h θ_li · ∇_h q_tot with the same weight and element filter as the variances),
clamped to ±`sgs_correlation_max` (default 1) and rescaled with the σ_q bound so
the correlation is invariant under it; the prescribed value is the fallback
where the variances vanish. The correlation the quadrature samples now lives
in `p.precomputed.ᶜcorr_Tq` (constant model: filled with the prescribed value),
and the `env_q_tot_temperature_{covariance,correlation}` diagnostics report it.
Requires the `tq` form and a nonzero `sgs_variance_horizontal_scale_factor`
(configuration-time checks; the vertical closure alone gives exactly ±1).
The Richardson and TKE weights (and the isentropic validity step) are now
materialized once per call into the persistent `p.precomputed.ᶜgeo_weight`
(1 = term on and unweighted, 0 = inactive), replacing the lazy per-use
`ᶜsgs_geo_weight` broadcast, and exported as the `sgs_geo_weight` diagnostic.
Squashed rebase of zs/sgs_variance_rh onto main after #4837 merged its
Richardson weight; the RH form of the original branch is dropped.
Validation: `ClimaAtmos` precompiles; `test/parameter_tests.jl` and
`test/parameterized_tendencies/microphysics/cloud_fraction.jl` pass; coarse
prognostic-EDMF aquaplanet smoke (h_elem 4, 3 steps) of the `tq` + TKE-weight
path and the isentropic + element-filter + TKE-weight path solves with finite
variances and ᶜgeo_weight in [0, 1]; the TKE weight without prognostic TKE is
rejected at configuration time. The diagnosed-correlation path solves with
|ᶜcorr_Tq| ≤ sgs_correlation_max, and its two invalid combinations (with the
isentropic form; without the geometric term) are rejected at configuration time.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Purpose
Multiply the horizontal resolved-gradient (geometric) SGS variance term of #4802 by w = Ri₊² / (Ri₊² + Ri₀²), Ri₊ = max(N², 0) / max(2 S², ε), Ri₀ = k Ri_crit, with N² the saturated moist buoyancy gradient in every cell and S² the cached strain-rate norm. The term then acts only in air that is absolutely stable to a saturated displacement and not turbulent, where the turbulence closure cannot carry the variance.
Note: This is a temporary solution. It helps with the boundary layer and tropical RH, but the physical explanation is not entirely clear. A more physical solution may be some maximum length scale for the horizontal mixing length that depends on stability.
To-do
Content