Physical Heliacal Visibility Phase 1 Altitude/Pressure Checkpoint

Date: 2026-07-30

Status: altitude/pressure interpolation gate passed; Phase 1 remains in progress

Runtime boundary: research evidence only, not a runtime table or data pack

Outcome

The direct-transmission laboratory now has frozen observer-altitude and pressure-ratio nodes, an admitted interpolation coordinate, and independently reserved holdouts. The admitted method is bilinear interpolation in extinction magnitude. It does not extrapolate and it rejects a query unless all four corners of its training cell satisfy both the absolute-pressure and pressure-ratio bounds.

The final v5 artifact contains 5,037 deterministic libRadtran runs and 15,111 spectral values across all six named molecular atmospheres. Independent validation evaluated 12,636 spectral values that were never used as training nodes. Another 396 values were explicitly excluded because one or more training-cell corners failed the already-frozen pressure-domain law.

Admitted Grid

Observer-altitude training nodes are 0, 500, 1,000, 1,500, 2,250, 3,000, 4,000, and 5,000 metres. Fourteen separate altitude holdouts lie between those nodes.

Pressure is represented as a ratio to the selected named atmosphere's surface pressure at the observer altitude. Training ratios are 0.85, 0.925, 1.0, 1.04, and 1.08. Eight separate pressure-ratio holdouts lie between those nodes. The 500-1,100 hPa absolute bound and the 0.85-1.08 ratio bound both remain mandatory.

The validation cross-product uses target true altitudes 0.25, 5, and 20 degrees and wavelengths 400, 550, and 780 nm.

Error Result

Quantity	Measured	Fixed ceiling
Maximum absolute extinction error	0.0124664 mag	0.025 mag
95th-percentile absolute extinction error	0.00404537 mag	0.01 mag
Maximum relative transmission error	0.0114163	0.025

No acceptance threshold was relaxed after observing a holdout.

The worst case is the subarctic-winter profile at 3,250 m, pressure ratio 1.08, target altitude 0.25 degrees, and 400 nm. Its predicted extinction is 18.1641139 mag versus 18.1516475 mag from the withheld libRadtran run.

Failed Designs Preserved

Four compact failure receipts remain source controlled.

1. The initial altitude holdouts failed the unchanged error ceiling.
2. A denser altitude-node design still failed because the native AFGL vertical grid changes surface-relative Shettle aerosol discretization when the bottom boundary crosses native levels.
3. The scientifically corrected site-relative grid passed numerically, but its validator incorrectly required arbitrary double values in a binary32-staged atmosphere profile to round-trip as doubles.
4. The next artifact passed on Linux but was rejected on Windows because the validator required byte equality for a recomputed logarithm that differed by approximately 1e-15.

The admitted design keeps the source-bound, site-relative 290-level near-horizon grid. The validator requires byte and checksum identity for raw libRadtran outputs and stored canonical receipts, then compares independently recomputed transcendental values with a fixed cross-platform numerical tolerance. It does not weaken the scientific acceptance limits.

Immutable Receipt

Field	Value
Artifact directory	phase1-altitude-pressure-interpolation-2026-07-30-v5
Generation fingerprint	ef95bba5a00667ce3bd1d983f9b9de93b989bd315da0e702f3342153d07bf165
Manifest SHA-256	2264727cf4d1a74bb747aa51cc44e4ba9e703e09c132ab57eb2c0afef863c727
Summary SHA-256	68648160b2d2b0f172fcc37715f1148e614f65f4f0a8e20a1bbfb1b8a4a04ab5
Files	50,768
Bytes	69,923,910
Native Linux validation	Passed
Windows validation of the same immutable bytes	Passed

The Windows validation used an immutable tar transport to local NTFS because walking more than 50,000 individual WSL files through the UNC bridge is prohibitively slow. The manifest and every per-file checksum were revalidated after extraction.

Boundary and Next Gate

No engine module, native code, public API, package dependency, or runtime data changed. The artifact remains external research evidence.

The next Phase 1 gate freezes the solar-altitude, target-altitude, and relative-azimuth adaptive radiance design; establishes a fail-closed deep-twilight law; and runs untouched directional-radiance holdouts.