# H10S fixed-control gross-confound audit v1

## Decision

The public environmental branch is **source incomplete**, and both fixed traffic-context comparisons are **indeterminate**. This audit therefore neither identifies an obvious environmental confound nor clears one. It also does not establish that either `+24 h` scene is traffic matched, environmentally matched, quiet, or low exposure.

The immutable MAERSK DETROIT and CRETANSEA H10S event/control energy results remain descriptive two-node comparisons only. Nothing here promotes a vessel, creates a ship detection or contribution result, or changes the old frozen artifacts.

## Prospective freeze and review

The unit was fixed as two scene pairs, not nodes or bands. Exact 600-second event and `+24 h` half-open UTC comparator windows, public variables, gross calipers, ARA snapshots, parser rules, 35/75/150 km rings, row-age rule, output classes, and claim ceiling were frozen before target public values or ARA headers/rows were opened.

Two prevalue reviews rejected implicit ARA parser discretion and a missing pin to the inherited exact-window table. Both were corrected: the exact historical schema contract is frozen in `ara-schema-mapping.json`, undocumented SOG/length/type semantics were removed from traffic gates, and the inherited membership is hash-pinned. A later prevalue review caught an under-specified CMEMS request box. The exact strict-inside box and required `1 x 1 x 12` structure were then bound into the protocol and both manifests. Review 005 approved the resulting one-shot freeze.

## Public environment stop

The first staged execution completed and structurally validated all five exact NASA POWER responses, then the exact frozen Copernicus request returned `12` three-hour timestamps, `2` latitude cells (`-9.0`, `-8.916666666666671`), and `1` longitude cell (`-14.666666666666686`) rather than the required `1 x 1 x 12` structure. The runner stopped before USGS acquisition or any scene-pair environmental evaluation. Its temporary POWER bytes were not published.

One identical-request evidence capture preserved the CMEMS response at `raw/cmems/cmems-wave.nc`, 42,856 bytes, SHA-256 `2e26e7a407100f2dc4a504b0f8cd4981508895e8d32b7a08a231ec5080e4e323`. It reproduced the same structural condition. No date, box, cell, source, version, variable, or selection rule was changed. Under the frozen fail-closed rule, both pair environment classes are `source_incomplete`; no weather, wave, precipitation, or seismic mismatch value is reported.

## Fixed local ARA context

All four pinned CSVs passed the exact frozen parser. Every delivered row had a valid position and explicit UTC `timeOfFix`; every row was within six hours of its file's latest valid fix. The within-pair snapshot-reference separations were 24.1642 hours for MAERSK and 23.0753 hours for CRETANSEA, so both traffic comparisons passed the structural eligibility gate.

Fresh unique MMSI counts were:

| Pair | Scene | within 35 km | within 75 km | within 150 km |
|---|---:|---:|---:|---:|
| MAERSK DETROIT | event | 1 | 1 | 4 |
| MAERSK DETROIT | control | 0 | 1 | 3 |
| CRETANSEA | event | 1 | 1 | 3 |
| CRETANSEA | control | 0 | 0 | 2 |

Neither pair reached the frozen gross gate: an absolute difference of at least 10 and at least a factor of two in the same direction at both 75 and 150 km. Both outputs are therefore `traffic_context_indeterminate`. That label is deliberate: these are irregular latest-known-position snapshots with undocumented provider coverage, not simultaneous or complete traffic censuses, so small count differences cannot establish “no mismatch” or vessel absence.

## Boundaries

No waveform, count, PSD, band value, or unpublished acoustic outcome was opened or recomputed. No ARA contact or acquisition occurred. No alternate control, rematching, interpolation, track/CPA reconstruction, RAM/propagation, upload, or canonical-document edit occurred. The exact source bytes, checksums, QC tables, ring summaries, stop record, and machine-readable decision are retained locally in this versioned package.