Rename the Breeze seam for consistency: base_pressure and ρE

[ewquon]

NumericalEarth's Breeze-facing extension still writes two names that Breeze has since redefined, so the seam says one thing and means another. Both are renames with no behavior change; a third commit is the behavior prerequisite that makes the first rename honest.

The reference datum. Rename surface_pressure → base_pressure on atmosphere_model, atmosphere_simulation, both nested_atmosphere_model methods, and nested_atmosphere_simulation.

Breeze#893 split surface_pressure into base_pressure (the reference atmosphere's pressure at z = 0, a scalar) and surface_pressure (the pressure at the ground, a 2D field, per column on terrain). #689 adopted the split at the CompressibleDynamics call but kept NumericalEarth's own keyword spelled surface_pressure to limit the blast radius. That left the worse name on the wrong quantity — the keyword forwards to the z = 0 datum, and the nested default is the domain-mean ERA5 mean sea level pressure.

The prerequisite. Do not pass a surface_pressure override to HydrostaticallyBalancedDensity.

atmosphere_model passed the same scalar it hands to CompressibleDynamics as base_pressure into HydrostaticallyBalancedDensity(; surface_pressure), whose keyword is a per-column ground-pressure override. Renaming the keyword while it still fed both slots would put base_pressure on a call that wants a surface pressure, so this lands first: the marker now takes its default, anchoring on the reference's per-column ground pressure.

Nothing in-tree changes: both coupled examples sit on grids bottomed at z = 0, where the two anchors are equal exactly, and the nested path passes initialize = false, so it never reaches set!(model; θ = potential_temperature, ρ = HydrostaticallyBalancedDensity()). The fix affects a direct atmosphere_model call on a raised or terrain-following grid, where the override re-creates the O(ρgh) state–reference mismatch that #893 fixed.

The energy coupling flux. Rename Jᵉ → Jᴱ, net.ρe → net.ρE.

Breeze reserves e for turbulent kinetic energy and uses E for total energy, including as the formulation-agnostic energy key (notation.md:52, :54, :167). The coupler's field was Jᵉ and net_fluxes returned it under ρe. Nothing was misrouted — the boundary condition is installed under energy_bc_key(), which #689 already pointed at total_energy_density_name — but ρe is a live TKE prognostic whenever the TKE closure is on, so net.ρe = Qc sat twenty lines from a genuine ρe tracer meaning something else. Internal: Jᵉ is a local binding and ρe is a NamedTuple key with one producer and one consumer in the same file; net_fluxes is not exported and there is no cross-component schema. The four Jᵉ in dry_layer_humidity.jl are a different quantity — a vapor flux whose superscript marks the evaporating surface — and are deliberately untouched.

Breaking change

Only the datum rename is breaking, for callers passing surface_pressure= to those constructors; the one in-tree caller (test/test_nested_simulation.jl) is updated. A stale keyword throws a MethodError naming it, at the NumericalEarth boundary rather than inside Breeze. The energy rename is internal and patch-safe.

Testing

CPU against the branch head (Breeze 0.11.2, Oceananigans 0.113.0, ClimaSeaIce 0.5.10): test_nested_simulation 281/281, test_breeze_coupling 44 pass / 2 broken / 0 fail (the 2 @test_broken are the pre-existing baseline), test_breeze_dry_layer_land 19/19. The nested suite covers the renamed keyword; the coupling and dry-layer-land suites cover net_fluxes and the assembly kernel the energy rename touches.

test_breeze_coupling.jl gains a regression test for the prerequisite, since nothing in-tree exercised that path. On a domain bottomed at z = 2 km it asserts that atmosphere_model's cold start agrees with its own reference state, and that the reference is genuinely reduced to the domain bottom rather than left at the datum — the second check matters because two quantities both anchored at the datum would agree with each other too. Reverted to the pre-fix anchor both assertions fail: max|p - pᵣ| is 22020.6 Pa against a 1e-6 tolerance, and the bottom-cell pressure is 100119.5 Pa where the reduced value is under 91192.5.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 70.00000% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...umericalEarthBreezeExt/breeze_nested_atmosphere.jl	40.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!

[glwagner]

what are these additional tests doing? seems like the PR is just renaming

[ewquon]

Verifying that our usage of pressure here is consistent with the updated usage in Breeze