# Physical methane and meter source qualification

Checked 2026-10-04. Original OGI weak-mask experiments, budgets and publication
receipts remain unchanged. These are different tasks and modalities.

## Executed chemical-source clock preflight

The separate [chemical UAS source preflight](../chemical-source-preflight/REPORT.md)
now preserves 16 full published file digests and an executed clock-only audit of
35,659 rows across 12 streams. Under the LI-7810 header's Europe/Amsterdam
timezone, **446 flight-1 analysis rows fall inside flight 2's collection window**.
Matching overlapping timestamps can therefore pair the wrong physical experiment.
Later laboratory analysis needs the original flight/session and AirCore coil/sample
mapping; a candidate-assisted alignment is not an independent timing reference.

No chemical, thermal, humidity, pressure or position measurement values were
decoded or fitted. The original MATLAB metadata errors and quoted-REMARK clock
parser failure remain recorded, alongside the repaired header/projection methods
and exact replay. LI-7810 units are explicit; Axetris logger scales, temperature
channel mapping and portable MATLAB object values remain unresolved. Five flights
share one sensor/site/day, including the known adverse flight 3. No chemical model
or industrial deployment is admitted. The table below preserves the earlier
source-qualification snapshot before this clock-only access.

## Earlier source-qualification snapshot

| Source | Verified rights and version | Practical use | Remaining boundary |
|---|---|---|---|
| [TADI controlled releases](https://zenodo.org/records/18381032) / [February 2026 paper](https://amt.copernicus.org/articles/19/923/2026/amt-19-923-2026.html) | CC BY 4.0, selected archive members match pinned Git objects | Executed cleaned rate-report denominator/quantification audit | No raw gas image/concentration model; published historical data, common facility and unequal schedules |
| [Groningen UAS methane sensor](https://zenodo.org/records/19882114) / [May 2026 paper](https://amt.copernicus.org/articles/19/3333/2026/amt-19-3333-2026.html) | CC BY 4.0 in record API; README matches source MD5 | Candidate thermal-stability and paired chemical-sensor study | Five flights at one farm/day; flight 3 has failed temperature control. Analysis clock offset, AirCore sample mapping and reference independence must be resolved; no numeric rows decoded |
| [2017 OPLS trajectories released in 2025](https://datadryad.org/dataset/doi:10.5061/dryad.h44j0zpxh) | Explicit CC0; version 404636 and complete source file digests pinned | Candidate drone trajectory/quality audit | Molar concentration versus path-integrated units and controlled-release/session joins unresolved; no raw CSV decoded |
| [Glasgow drone methane imaging](https://researchdata.gla.ac.uk/1697/) | CC BY 4.0 on primary record | Gas-sensitive phase-paired imagery candidate | 9 GB distribution requires contacting owner; no outreach or image access. Phase direction and release/no-release references must be established |
| [NRC-GAMMA physical gas meter](https://nrc-digital-repository.canada.ca/eng/view/object/?id=ba1fc493-e65f-4c0a-ab31-ecbcdf00bfa4) | CC BY 4.0; code/README tree pinned `8d8184c5ee10ad850a1c5d8e5c6395b47deca191` | Analog-dial reference candidate | One meter/day, crowd reading labels, 3.8 GiB TGZ; no matched physical image/reference smoke acquired |
| [2026 physical water counters](https://datadryad.org/dataset/doi:10.5061/dryad.7d7wm3860) | Explicit CC0 API; version 422392, file 4579939 | Frozen five-crop existing-OCR smoke proposal | API 401 and published download route 403. Zero image bytes/inference; physical asset/session mapping absent |

Groningen's AirCore-to-flight time assignment partly uses the candidate Axetris
signal, so a future paired benchmark cannot simply call that time mapping an
independent reference. Its README also marks onboard wind channels uncalibrated
and describes offline wind averaging around flight times. Reference/clock/wind
assumptions must be frozen before paired evaluation or any causal-online claim.
The scientific paper's own source/adverse-condition results are not our rerun.

Meter acquisition retains both failed requests, without a full-archive fallback,
credentials or contact forms. The new boundary does not amend the older ETH-only
boundary. No sample was chosen by pixels or successful model output. Published
human digit labels would support string agreement, not independently calibrated
volume or unseen-meter accuracy. Entire exploratory assets/sessions and unknown
derivatives must be excluded from later confirmation; unavailable grouping blocks
that claim.

The new [TADI audit](REPORT.md) is executable with retained
source objects, failures and arithmetic replay. Chemical sensing, physical-meter
accuracy, target-device qualification, site acceptance, field adoption and
qualified expert growth are separate unfinished milestones.