Files
aimsight-labeling/auto/analyses/NEBRASKA_2026-10-22T0000_2026-11-01T0000/analysis.md
T
maximeandClaude Opus 5.5 ddb17c2189 auto: labels requalified under the 5-noise rules (only what stands out from the noise): 32 analyses refined, the 6 paused periods verified, the long labels requalified with the fixes the period analysts found; 1153 -> 961 labels
- 32 analyses: each label kept, rewritten or removed by its analyst (section Requalification of the report; mm magnitudes where the targets have them)
- SNCF01 x3, WILDSTRUBEL 2026-09-01 -> 2026-10-06 x3: verified from their existing labels under the new rules
- long labels (no analysis): auto/analyses/long-labels/<STATION>/; NORDEND-169 and -195 run to 07/11, so NORDEND-283 and -284 (their continuation) are removed
- reading-rules.md, stations.md: the citations of removed or rewritten labels point to their successor, or to [expert]

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-17 23:58:00 +01:00

11 KiB
Raw Blame History

Labels: NEBRASKA, 2026-10-22T00:00Z → 2026-11-01T00:00Z (UTC)

Requalification, 2026-11-17, of the labels of this analysis under the new rule (labeling/labeling-rules.md, section 1): only what stands out from the noise of the corrected curves, at least one target beyond 5 of its own noises, is labelled. Input: work/requalify_input.jsonl (7 labels, identical to the previous labels.jsonl, kept in work/labels.before-requalify.jsonl); previous analysis kept in work/analysis.before-requalify.md. The reading was not redone. The station measures every 4 h (61 instants, none missing). No weather/ folder for NEBRASKA: no weather available. No image-quality columns (older camera). Spanning labels NEBRASKA-2 (REFCOLH22 lost, a Target issue, reference rejected at every instant of the range) and NEBRASKA-6 (long drift, Displacement) are outside this analysis and not requalified here.

Result: 4 kept as they are (NEBRASKA-8, -9, -10, -11, all reviewed 2026-11-17), 0 rewritten, 3 removed (NEBRASKA-15, -16, -17).

1. My labels

The four labels kept are unchanged (start, end, family, description, review). Their original evidence (previous analysis):

Id Verdict Evidence
NEBRASKA-8 (26T00→26T20, Thermal) kept Housing temperature a little lower 26T02→26T14 (30–30.5 °C vs 31–31.5). Pose: tilt 0.0362 (00) → 0.0375 → 0.0399 → 0.0410 (12) → 0.0383 (16) → 0.0394 (20), rotation −0.0229 → −0.0283 (08) → −0.0243 (20); pan last at rest at 00:00. corr_y COLJ20 −0.090 (00) → −0.009 → +0.029 (08) → −0.029 (20); POUJ19 −0.016 → +0.093 (08) → +0.009 (20); POUJ20, COLJ19 the same; corr_x nearly flat. 26T16 still touched (POUJ19 +0.046), 26T20 back. Bounds, family, description hold.
NEBRASKA-9 (27T20→28T16, Thermal) kept Temperature 31.5 → 30.25–30.75 from 28T01 to 28T10. Pan 0.0009 (27T20) → 0.0066 (04), rotation −0.0256 → −0.0194 (04), tilt 0.0440 and scale 0.999456 at 08, all back at 16 (tilt 0.0387, scale 0.999577). corr_y down from 28T00: COLH19 +0.065 → −0.079 (04), POUH19 +0.068 → −0.018, POUJ20 +0.015 → −0.062, COLJ20 −0.016 → −0.125, COLH20 0.109 → 0.035; corr_x peak at 08 (COLJ20 +0.064, POUJ19 +0.058, POUH21 +0.057, COLJ19, POUH20 about +0.04). Back at 28T16. Holds.
NEBRASKA-10 (28T20→29T12, Thermal) kept Coldest night: 27.25 °C at 29T07. Tilt 0.0387 (28T16) → 0.0422 (20) → 0.0480 (04); pan last at rest at 28T20 (0.0026) then 0.0075 (00), 0.0081 (08): the later of the two axes rests until 28T20, so the start is right (labeling-rules 2, camera cycle start; one step of difference stays within the uncertainty anyway). corr_x up 29T00–08 (COLJ20 0.773/0.837/0.806, POUJ20 0.717 at 04); corr_y jumps up at 04 (COLJ20 −0.117 → +0.065, POUJ19 −0.061 → +0.113, POUH19, COLH19, COLH20), lower at 00 and 08. error_value at 04: REFH22 0.155 → 0.194, COLH19 0.163, COLH20 0.184, far below 0.7. 29T12 back at the new level of 29T16 (corr_x COLJ20 0.778/0.773, tilt 0.0421/0.0406). Holds.
NEBRASKA-11 (29T20→30T16, Thermal) kept Temperature 31.75 (29T19) → 27.75 (30T06). Pan 0.0037 → 0.0057 at 29T20, tilt last at rest at 29T20 (0.0416) then 0.0449, 0.0481 (04); scale 0.999397 at 04; back at 16. corr_y jumps up at 04 (COLJ20 −0.027 → +0.077, POUJ19 +0.027 → +0.106, POUH20 −0.013 → +0.065, POUJ20 −0.024 → +0.048), down at 08; corr_x higher 00–12. error_value at 04: REFH22 0.207, COLH20 0.216, COLH19 0.163. Back at 16. Holds.

2026-10-26T00:00Z → 2026-10-26T20:00Z (NEBRASKA-8, kept)

  • Phenomenon of the catalogue: the camera drifts when the temperature of its housing drops; a compensation on two references.
  • Precedents: none needed (requalification; the label is reviewed).
  • Family: Thermal displacement, unchanged: the corrected curves keep the camera movement of the cool morning and return with it.
  • Measured: POUJ19 corr_y +5.2 noises from its level of the 24 h before (+0.098 px, +0.45 mm at 11.7 m) at 26T08; ranges.py --k 5 --rolling 1d: POUJ19 corr_y +5.3 at 26T08 (also beyond 5 at --rolling 2d); COLJ20, POUJ20, COLJ19 corr_y +3.4 to +3.5 noises at 26T08. Just above the threshold.
  • Images: none new.
  • Confidence: medium: a single target just beyond 5 noises, at the peak measurement; the label holds.

2026-10-27T20:00Z → 2026-10-28T16:00Z (NEBRASKA-9, kept)

  • Phenomenon of the catalogue: same.
  • Precedents: none needed.
  • Family: Thermal displacement, unchanged.
  • Measured: at 28T08, POUH21 corr_x +5.8 noises (+0.056 px, +0.83 mm at 37.8 m), POUH20 +5.3 (+0.60 mm at 30.8 m), COLH20 +5.0; COLJ24 corr_y −5.0 at 28T04 (−1.44 mm at 68.8 m). --rolling 1d, 12h and 2d: POUH20 cx +5, POUH21 cx +6 at 28T08.
  • Images: none new.
  • Confidence: high that it crosses the threshold (two targets, three rolling windows).

2026-10-28T20:00Z → 2026-10-29T12:00Z (NEBRASKA-10, kept)

  • Phenomenon of the catalogue: same.
  • Precedents: none needed.
  • Family: Thermal displacement, unchanged.
  • Measured: at 29T04, corr_x COLJ20 +10.3 noises (+0.46 mm), POUJ20 +10.0 (+0.51 mm), COLJ19 +9.9 (+0.56 mm), POUJ19 +9.5 from the level before; --rolling 1d: COLJ19 cx +7.5, POUJ20 cx +7.3, POUJ19 cy +7.4, COLJ20 cy +5.6 at 29T04; COLJ24 corr_y −6.7, COLJ23 −5.1 at 29T08.
  • Images: none new.
  • Confidence: high: the strongest episode of the range.

2026-10-29T20:00Z → 2026-10-30T16:00Z (NEBRASKA-11, kept)

  • Phenomenon of the catalogue: same.
  • Precedents: none needed.
  • Family: Thermal displacement, unchanged.
  • Measured: at 30T04, POUJ19 corr_y +7.0 noises (+0.61 mm), COLJ20 corr_y +6.0 (+0.53 mm); --rolling 1d: COLJ20 cy +5.7, POUJ19 cy +6.9, COLJ24 cy −5.1 at 30T04 (same at 12h and 2d).
  • Images: none new.
  • Confidence: high.

2. What I left without a label

Removed by the requalification (no target beyond 5 noises from its level on the corrected curves, whatever the reference: the 24 h before the label, or ranges.py --rolling at 12h, 1d, 2d):

  • NEBRASKA-15, 2026-10-23T20:00Z → 2026-10-25T04:00Z, Thermal displacement: largest deviation from the level before, COLJ20 corr_x +4.8 noises at 24T20 (+0.06 px, +0.22 mm), COLJ24 corr_y −4.5 noises at 24T00 (−0.05 px, −1.30 mm: a far target, 68.8 m, many mm per px), COLH21 corr_y +4.0; none at --k 5 with --rolling 12h, 1d or 2d. With --outside the label (window ±1 day) COLJ24 reaches −6.9 at 24T00, but that noise is computed on the 3 days around only, which hold the other cold nights' curves partly removed; on its noise over the range (as ranges.py without --outside) COLJ24 stays at −4.5. The camera movement is real (pose and temperature, previous analysis) but the compensation leaves on the curves less than the threshold: compensated thermal, removed.
  • NEBRASKA-16, 2026-10-31T00:00Z → 2026-10-31T16:00Z, Thermal displacement: largest deviation COLH19 corr_x −3.2 noises at 31T00, POUJ19 corr_y −2.9 at 31T04 (−0.26 mm); none at --rolling (not even at --k 3 with --rolling 2d, where only COLH20 cx −3 shows at 31T16). Compensated thermal, removed.
  • NEBRASKA-17, 2026-10-31T20:00Z → 2026-11-01T00:00Z, Thermal displacement: at most 1.7 noises (POUH21 corr_y); the 01T00 measurement (and the following up to 11-03) stays under 3 noises at --rolling 2d. Compensated thermal, removed; nothing to extend into the next range.

Beyond the removed labels, the previous analysis already left without a label: 22T12–16 and 23T12–16 (corr_y of COLJ20, POUJ19, POUJ20 at 2–3 noises), 27T12 (several corr_y at 2–2.5 noises for one measurement). Other instants beyond 5 noises at --rolling all fall inside the kept labels (28T08, 29T04, 29T08, 30T04, 26T08); 28T20 (COLH19 cx −6, POUH20 cx −5 at --rolling 12h; COLJ24 cy −5 at 2d) is the bound of NEBRASKA-9/-10, its 12h deviation coming from the 29T04 peak pulling the local level, not a new event (no new event is looked for in this requalification).

3. The documentation

  • labeling-rules.md section 1 (5-noise threshold, thermal kept only if the corrected curves keep it) decided every verdict. The previous analysis had noted that at a 4-hour cadence days.py never reaches 5 noises on the pose; the new rule reads the corrected curves instead, and with them the three weak episodes (24, 31 October, start of 1 November) added by the previous analysis fall out, while the four reviewed ones hold.
  • --rolling choice at a 4 h cadence: 1d is only ±6 measurements; a cool day of 24 h (NEBRASKA-15) would be partly absorbed by it, so 2d and the level of the 24 h before were also read: they agree.
  • --outside on a short window gives a local noise that inflates the deviations (NEBRASKA-15, COLJ24): to compare across labels, the noise over the whole range is safer.

Difficulties

none

Requalification

Commands (outputs in work/requal/):

  • python3 tools/scripts/ranges.py data/NEBRASKA --columns corr_x,corr_y --k 5 --rolling 1d --start 2026-10-22T00:00Z --end 2026-11-01T04:00Z --instants 2 → ranges_rolling1d.txt; same with --rolling 12h and --rolling 2d, --instants 1, the 12 targets → ranges_rolling12h.txt, ranges_rolling2d.txt.
  • per label, ranges.py --k 5 --start <start−1d> --end <end+1d> --outside <start>/<end> → ranges_outside.txt.
  • python3 labeling/auto/analyses/NEBRASKA_2026-10-22T0000_2026-11-01T0000/scripts/night_level.py → night_level.txt: per label, each target's largest deviation from its median of the 24 h before the label, in its noise over the range (median absolute successive difference / 0.6745 / √2), px and mm.
  • ranges.py --k 3 --rolling 2d --start 2026-10-29T00:00Z --end 2026-11-03T00:00Z --instants 1 for the end of the range.
Id Bounds Largest deviation (level before) rolling 1d / 12h / 2d at k 5 Verdict
NEBRASKA-15 23T20 → 25T04 COLJ20 cx +4.8 (24T20), COLJ24 cy −4.5 none / none / none removed
NEBRASKA-8 26T00 → 26T20 POUJ19 cy +5.2 (26T08) POUJ19 cy +5.3 / – / +5 kept
NEBRASKA-9 27T20 → 28T16 POUH21 cx +5.8, POUH20 cx +5.3 (28T08) +6, +5 at all three kept
NEBRASKA-10 28T20 → 29T12 COLJ20 cx +10.3 (29T04) up to +8 at all three kept
NEBRASKA-11 29T20 → 30T16 POUJ19 cy +7.0 (30T04) +7 at all three kept
NEBRASKA-16 31T00 → 31T16 COLH19 cx −3.2 (31T00) none / none / none removed
NEBRASKA-17 31T20 → 01T00 POUH21 cy +1.7 none / none / none removed

Sources read

instructions/analyst.md, work/requalify_input.jsonl, labels.jsonl, analysis.md and report.md of this analysis, work/spanning.jsonl, tools/GUIDE.md (ranges), labeling/labeling-rules.md (sections 1–4), tools/scripts/data.py (the reader, for my script). Tools: ranges.py, inventory.py. The previous analysis's sources: agents/agents/analyst.md, tools/DATA.md, labeling/families.md, labeling/reading-rules.md, labeling/stations.md, labeling/auto/NEBRASKA.jsonl, the BREITHORN and MICHIGAN labels of the range, precedents.py, overview.py, metrics.py, summary.py, days.py, stations.py, thumbnails.py, plot.py, images.