# Sky Truth — Plano, Texas — Friday October 23, 2026

**The satellite detector has not passed its accuracy test, so nothing below that depends on it is used to judge the sky. Predictions are still made and locked as normal.**

## The test that matters (all days so far)

- Flights whose air was dry by every measurement, yet whose own trail lasted: 0. The number expected by coincidence is not measured yet (needs a day of flight history).

## Today, hour by hour

Flight outcomes: H = predicted lasting trail and its own trail was seen; M = predicted, not seen; FA = predicted no lasting trail but its trail lasted; CN = predicted none, none seen.

| Hour | Air over the area | Flights predicted to leave lasting trails | Flight outcomes | Unexplained lines (max per scan) | Candidate |
|---|---|---|---|---|---|
| 6 PM | uncertain | 0 | 0H 0M 0FA 0CN | 0 |  |
| 7 PM | uncertain | 0 | 0H 0M 0FA 17CN | 0 |  |
| 8 PM | uncertain | 0 | 0H 0M 0FA 40CN | 0 |  |
| 9 PM | uncertain | 0 | 0H 0M 0FA 23CN | 0 |  |

Candidate hours today: 0. A candidate is NOT an anomaly: candidates include coincidences. Only the test at the top of this report says whether there are more than chance explains.

## Running totals (all days)

**Primary result (raw humidity)** (0 scored flights)
- Flights predicted to leave lasting trails whose trails the satellite saw: n/a
- Flights predicted NOT to leave lasting trails whose trails lasted anyway: n/a
- Skill score (hit rate minus false-alarm rate; 0 = no better than always predicting trails, +1 = perfect): n/a

**Pre-registered sensitivity test (humidity divided by 0.97 for known model dryness; constant correction of Teoh et al. 2020, derived for ERA5)** (0 scored flights)
- Flights predicted to leave lasting trails whose trails the satellite saw: n/a
- Flights predicted NOT to leave lasting trails whose trails lasted anyway: n/a
- Skill score (hit rate minus false-alarm rate; 0 = no better than always predicting trails, +1 = perfect): n/a

Satellite detector, measured on 223 expert-labelled contrail images (OpenContrails): flags 54% of images that contain contrails, and 47% of images that contain none; 12% of its lines lie on real contrails.

Note: the 2 km satellite cannot see narrow trails, so some misses are expected.

**Coverage (all days)**
- Flight verdicts that were firm (all models agreed) and could be scored: 138 of 138 (100%); disputed or missing data: 0
- Hours where the air over the whole area was clearly moist or clearly dry: 0 of 4 (0%)

## Weather sources compared (mean of max ice-humidity at cruise levels)

- ecmwf: 93% over 29 windows
- gfs: 81% over 29 windows
- hrrr: 75% over 25 windows
- raob: 123% over 8 windows
## Satellite lines traced to aircraft

- Plano area: 0.00 lines per scan on average (36 scans)
- Low-traffic control area (West Texas), same detector: 0.03 lines per scan — this is roughly how often natural cloud alone produces a line
- Lines matched to a transponder track after wind drift: 0 of 0 (n/a)
- Control area lines with no aircraft behind them (natural-cloud rate): 1 of 1, 0.03 per scan

## High-resolution passes (VIIRS, 375 m)

- 2:35 PM NOAA-21 (78.9° up): ok, 0 lines; that hour's prediction: not scored

## Satellite-measured humidity

- Calibration pairs (balloon vs satellite): 1 — not yet in use (needs 20 pairs and a real fit)

Aircraft with known type (engine-specific physics): 735 of 945 (78%)


## Air chemistry on lasting-trail days vs no-trail days (EPA, Dallas Hinton)

- No chemistry compared yet (plan confirmed in Bitcoin: not yet).

## Data source health

- aircraft_db: 1 of 1 runs succeeded
- ecmwf: 19 of 19 runs succeeded
- gfs: 19 of 19 runs succeeded
- goes: 35 of 35 runs succeeded
- goes_wv: 31 of 31 runs succeeded
- hrrr: 19 of 19 runs succeeded
- metar: 19 of 19 runs succeeded
- opensky: 10 of 10 runs succeeded
- opensky_control: 25 of 25 runs succeeded
- opensky_main: 59 of 59 runs succeeded
- raob: 19 of 19 runs succeeded
- viirs: 36 of 36 runs succeeded

## Prediction ledger

- Integrity: INTACT — 42 predictions, chain intact
- Latest hash: `eedc8b5787892b5fcc80a7d86327eb3222d0d5527219ade2c5336af58370439a`
- Bitcoin timestamp (OpenTimestamps): stamped every hour (anchors folder) and daily (stamped). Proof files end in `.ots`; anyone can check them with `ots verify`.