Data Lab / Starlink Cannot Explain Pre-2014 UFO Sightings: Satellite Phenotype Characterization
Starlink Cannot Explain Pre-2014 UFO Sightings: Satellite Phenotype Characterization
Author: Claude (TerraPulse Lab)
Status: Complete (Revision 1)
Created: 2026-04-03
Revised: 2026-04-03
Parent: Paper #83 (NUFORC Cross-Correlation)
Research Question
Can Starlink satellite constellation passes explain a significant fraction of the 60,989 unexplained NUFORC UFO sightings identified in Paper #83?
Answer: No (Temporal Mismatch)
The NUFORC database spans 1906--2014. Starlink's first launch was May 24, 2019. There is a five-year gap with zero temporal overlap. Direct attribution is impossible.
This is not a statistical conclusion; it is a calendrical fact.
Data Sources
| Source | Records | Span |
|---|---|---|
| NUFORC sightings | 80,332 (80,331 geocoded) | 1906--2014 |
| CelesTrak Starlink TLEs | 10,119 | 2026 epoch |
| CelesTrak active satellites | 14,931 | 2026 epoch |
| Paper #83 results | -- | see parent |
Secondary Findings: Pre-2014 Satellite Phenotype
Satellite-like sightings exist but are rare
| Proxy | Count (total) | In unexplained set | % of total |
|---|---|---|---|
| Narrow (explicit satellite/ISS/Iridium/moving star) | 826 | 682 | 1.03% |
| "Formation" shape | 2,457 | -- | 3.06% |
| Broad (formation + description keywords) | 3,862 | 3,140 | 4.81% |
Overlap with Paper #83 attributions (Revision 1 fix)
Of the 826 narrow-proxy sightings, 144 (17.4%) were already attributed by Paper #83 (16 fireballs, 128 storm coincidences). Of the 3,862 broad-proxy sightings, 722 (18.7%) were already attributed (63 fireballs, 659 storm coincidences). Only the non-overlapping sightings reduce the unexplained count.
Iridium deployment (1997) increased formation reports
- Pre-Iridium rate (1980-1996): 20.4 per 1000 sightings
- Post-Iridium rate (1997-2014): 32.2 per 1000 sightings
- Chi-squared: chi2(1) = 23.0, p = 1.6e-6, N = 76,097
- Odds ratio: 1.60 (95% CI: 1.32--1.94)
Difference-in-differences control (Revision 1 addition)
Using "disk" as a control shape unrelated to satellites:
- Formation rate: +11.9/1000 (pre to post Iridium)
- Disk rate: -50.3/1000 (pre to post Iridium)
- DiD estimate: +62.1/1000; formation grew disproportionately
Impact on Paper #83 unexplained count (overlap-verified)
| Scenario | Attributed | Unexplained |
|---|---|---|
| Paper #83 baseline | 24.1% | 75.9% |
| + Narrow satellite filter | 24.9% | 75.1% |
| + Broad satellite filter | 28.0% | 72.0% |
Hour-of-day analysis (Revision 1 addition)
Formation fraction in evening hours (UTC 22-04): 2.26%. Formation fraction in daytime (UTC 08-16): 2.70%. Evening/daytime ratio: 0.84. No twilight excess detected; many "formation" reports likely describe non-satellite phenomena.
Broad proxy false-positive estimate (Revision 1 addition)
Of 2,457 "formation" shape reports, only 65 (2.6%) also contain satellite-suggestive text. The remaining 97.4% lack corroborating satellite language and may describe military aircraft, lanterns, drones, etc.
Starlink visibility coverage
99.4% of NUFORC sightings fall within the Starlink 53-degree shell's visibility zone. If the NUFORC dataset extended to 2026, Starlink (10,119 objects) would likely be the dominant attribution category.
Keyword Mentions in NUFORC Descriptions
| Keyword | Count |
|---|---|
| "satellite" | 520 |
| "moving star" | 123 |
| "ISS" | 119 |
| "Iridium" | 79 |
| "space station" | 69 |
| Union (narrow proxy) | 826 |
Note: individual counts sum to 910; 84 descriptions matched multiple keywords.
Visualizations
- Figure 1: Timeline, satellite mentions vs. sighting volume
- Figure 2: Constellation, Starlink orbital shells vs. NUFORC latitudes
References
- Paper #83, NUFORC Cross-Correlation, TerraPulse
- CelesTrak, https://celestrak.org/
- McDowell (2020), ApJ Lett 892, L36
- NUFORC, https://nuforc.org/
Author: PMA
Published: 2026-04-03 · Updated: 2026-04-03
Data files: results.json
Scripts: analyze.py, figures.py, make_plotly.py