Physical Heliacal Visibility Phase 4 Jones MYSTIC Pilot Checkpoint

Date: 2026-07-31

Status: corrected v2 550 nm pilot and pre-holdout threshold gate closed; replacement sealed holdouts passed separately; spectral and runtime models not admitted

Candidate identifier: jones_paranal_scattered_moonlight_2013_v1

Pilot identifier: jones_paranal_mystic_550nm_pilot_v2

Outcome

The corrected, independently assembled libRadtran 2.0.6 MYSTIC pilot completed all 15 declared cases. Every threshold frozen after those measurements and before execution of the three new reserved holdouts passed. The result does not admit a spectral grid, runtime model, production data pack, public API, or network dependency.

The generated numerical artifact remains outside the repository. The repository contains the generator and validator contracts, exact source and tool receipts, compact measured checkpoints, and the frozen threshold declaration. No libRadtran or ESO source bytes are redistributed.

V1 Invalidation and V2 Correction

The original v1 pilot, threshold, and holdout evidence is invalid for scientific admission. libRadtran applies each aerosol_file explicit property file from its listed altitude upward to the next boundary; the uppermost line is only a top marker. V1 listed each physical file at its layer's upper boundary. That shifted every layer upward and placed the last physical layer above the intended 20 km profile top.

The v1 builder and validator shared the same incorrect serialization, so receipt, repeatability, convergence, and holdout checks could not detect the source-semantics error. No engine, public API, runtime model, or production data pack used the invalidated evidence. The exact old receipts and source rule are preserved in phase4_jones_mystic_v1_invalidation_checkpoint_2026-07-31.json.

V2 writes a null top marker, an explicit null gap down to the 20 km profile top, and each physical file at its inclusive lower boundary. The pilot builder and independent validator reconstruct and check that ownership separately.

Pilot Boundary

The pilot remains one-wavelength evidence at 550 nm. Its 15 executed cases cover target and Moon altitude, relative azimuth and separation, lunar phase over the admitted 1.55-97 degree ROLO domain, waxing and waning sides, Moon-Earth distance, three photon counts, three independent 300,000-photon seeds, and an exact fixed-seed repeat.

Three fresh geometry combinations were reserved before corrected-v2 execution and were not used to select thresholds. The sealed protocol assigns each holdout 1,000,000 photons and seed 271828183, with an exact repeat of holdout_v2_interior_cross_axis.

The pilot uses checksum-locked ESO solar, ROLO, and mie_m15s1.dat inputs. The aerosol phase function remains source-owned rather than independently reconstructable. The radiative-transfer assembly is independent; the aerosol microphysics is not claimed to be independent.

Explicit Design Limitation

The pilot places the atmosphere and ground at the 2,640 m observer altitude. Jones uses a 2,000 m lower model boundary beneath the observer. This difference must be resolved or explicitly versioned before spectral or production admission.

Generated Artifact Receipt

The corrected-v2 external artifact manifest is locked at:

• bytes: 41237;
• SHA-256: f3a139617abe43ffc1098a6b2df5e90dc10a48a32b1219316a772e6d60245aa7;
• executed cases: 15;
• reserved holdouts: 3;
• directional radiance range: 3.74437e-09 to 8.66392e-08 W m-2 nm-1 sr-1; and
• runtime or network dependency: none.

The source archive, uvspec binary, and 1,300-file data tree remain bound by SHA-256 values 64930cc40b6e4a37aa220520974d330fc1563796f466a649b2238131f2d69840, d4e94259296a65f7700a0911f0dc7fc14aacde89985befac0266fe0a18531b7a, and 191e39b7d1d6554517227a21dfa0feea11ffc308889ffccbe938408ed0fbf207.

Frozen Threshold Results

Check	Observed	Maximum
Relative standard error, all pilot cases	0.0110471732	0.0125
Relative standard error, base 1M case	0.0034758176	0.004
Standard-error times square-root-photons max/min	1.009499576	1.05
Three-seed 300K sample relative standard deviation	0.0039599941	0.01
Sample spread / mean reported relative error	0.623234396	1.0
Aerosol absolute angle reconstruction error	0.2069652263	0.25
Aerosol relative angle reconstruction error	0.0075170261	0.01
Absolute raw k0 - 1	0.0006827257	0.001
Unrepresented AOD fraction above profile top	5.21443e-07	1e-06
Radiance/source-ratio max/min	1.0000041606	1.00001
Absolute MYSTIC-SkyCalc difference	0.0174360371 mag	0.15 mag

All directional outputs were finite and positive. The fixed-seed repeat was byte-identical for mc.rad.spc, mc.rad.std.spc, mc0.rad, mc0.rad.std, and randomseed.

The SkyCalc 2.0.9 comparison is source-owned operational evidence, not an independent oracle. Its 550 nm flux_sml value converts to 1.5224540982e-08 W m-2 nm-1 sr-1; the matching corrected MYSTIC case produced 1.4982e-08 W m-2 nm-1 sr-1.

Independent Validation

The validator independently reconstructs the atmosphere, O4 companion, lunar source, corrected aerosol column and 512 Legendre moments, explicit layer ownership, geometry, case set, repeat files, and convergence diagnostics. It also verifies the exact generator and external-source receipts.

The threshold evaluator binds that artifact to the frozen v2 threshold specification and exact SkyCalc capture. Its checkpoint status is corrected_v2_pilot_passes_threshold_gate_holdouts_not_executed.

Machine-Checkable Files

• scripts/visibility_reference_lab/phase4_jones_mystic_pilot_spec.json
• scripts/build_visibility_phase4_jones_mystic_pilot.py
• scripts/validate_visibility_phase4_jones_mystic_pilot.py
• tests/artifacts/visibility_reference_lab/phase4_jones_mystic_pilot_checkpoint_2026-07-31.json
• scripts/visibility_reference_lab/phase4_jones_mystic_admission_thresholds.json
• scripts/audit_visibility_phase4_jones_mystic_thresholds.py
• tests/artifacts/visibility_reference_lab/phase4_jones_mystic_threshold_checkpoint_2026-07-31.json
• tests/artifacts/visibility_reference_lab/phase4_jones_mystic_v1_invalidation_checkpoint_2026-07-31.json
• tests/unit/test_visibility_phase4_jones_mystic_pilot.py
• tests/unit/test_visibility_phase4_jones_mystic_thresholds.py

Reproduction

python3 scripts/build_visibility_phase4_jones_mystic_pilot.py \
  --uvspec /home/nilad/.cache/moira/visibility-reference-lab/work/libRadtran-2.0.6/bin/uvspec \
  --lib-radtran-archive /home/nilad/.cache/moira/visibility-reference-lab/source/libRadtran-2.0.6.tar.gz \
  --data-root /home/nilad/.cache/moira/visibility-reference-lab/work/libRadtran-2.0.6/data \
  --eso-archive /mnt/c/Users/nilad/AppData/Local/Temp/moira-jones-20260731/SM-01.tar.gz \
  --output-root /home/nilad/.cache/moira/visibility-reference-lab/work/jones-mystic-pilot-v2-20260731-2

python3 scripts/validate_visibility_phase4_jones_mystic_pilot.py \
  --artifact-root /home/nilad/.cache/moira/visibility-reference-lab/work/jones-mystic-pilot-v2-20260731-2 \
  --uvspec /home/nilad/.cache/moira/visibility-reference-lab/work/libRadtran-2.0.6/bin/uvspec \
  --lib-radtran-archive /home/nilad/.cache/moira/visibility-reference-lab/source/libRadtran-2.0.6.tar.gz \
  --data-root /home/nilad/.cache/moira/visibility-reference-lab/work/libRadtran-2.0.6/data \
  --eso-archive /mnt/c/Users/nilad/AppData/Local/Temp/moira-jones-20260731/SM-01.tar.gz

Gate State

Closed:

• corrected, checksum-bound 550 nm pilot matrix and generator;
• independent artifact validation and explicit aerosol-layer ownership;
• repeatability, convergence, aerosol representation, and lunar-source checks;
• corrected-v2 pre-holdout numerical threshold freeze; and
• fresh sealed holdouts in the companion holdout checkpoint.

Still open:

• an independently generated spectral grid and interpolation thresholds;
• untouched spectral holdouts and response-integrated products;
• release/distribution disposition for any derived data artifact;
• engine implementation and typed component receipt;
• public-surface parity; and
• Phase 4 closure.

The lower-boundary disposition listed by the original checkpoint was closed on 2026-08-02 by the Jones MYSTIC lower-boundary checkpoint. This pilot remains the exact corrected-v2 observer-bottom control.