NUCLEAR
PROXIMITY
Is the UAP phenomenon reported closer to nuclear-weapons sites than chance would predict? We tested our strongest theme spatially — honestly, controlling for the obvious confound that reports come from where people are.
Does the reporting cluster near nuclear sites — or just near people?
Sightings happen where people are. Any raw “UFOs near reactors” claim has to clear that bar first. So we test each nuclear layer against a population-weighted null — random points drawn from real cities, matched by country and era — and ask whether the observed cases fall closer than that null predicts.
Enrichment = observed ÷ population-null. Above 1.0× means closer than chance; below means further.
This measures where cases are reported, not where craft were. The reporting-feedback confound — nuclear cases amplified because they’re nuclear — cannot be removed.
The fossil-fuel row is the placebo: same population and industrial footprint, nothing nuclear. A nuclear layer only counts as signal if it beats the placebo.
Enrichment vs population-matched chance
Each bar diverges from the 1.0× chance line — right means cases fall closer to that site type than population predicts, left means further. Significant results (p < 0.05) glow and carry a ✓. Click any layer to inspect it.
Nuclear weapons
SIGNAL ✓The headline result. Cases fall dramatically closer to warhead and weapons-storage sites than population predicts — over 4× at the tightest radius — and the effect is strongest as you close in, exactly the shape a real spatial signal should have.
Is it nuclear — or just “secret military”?
The strongest objection: reports cluster near secret bases, and weapons sites are secret bases. So we pit nuclear weapons against non-nuclear military bases — Nellis, Edwards, Holloman, Fallon: secret, but not nuclear. Both measured at 50 km.
Weapons enrichment (2.17×) far exceeds generic secret-military bases (1.20×, only marginally significant). The nuclear dimension does real work beyond “reports cluster near secret activity.”
Does the signal survive on corroborated cases?
The corpus grew mostly with single-source archival sightings, so the honest worry is that a pile of uncorroborated reports manufactures the effect. We re-run the weapons test separately on L3+ cases (corroborated across ≥2 sources or a primary document) and L0 single-source cases — each with its own matched null.
The weapons signal holds on corroborated-only cases — L3+ enrichment at 25 km is 4.68× (p<0.001, n=492), if anything stronger than the single-source stratum. It is not an artifact of the archival influx.
How the test works
Geocode & date the cases
1,116 cases carry a location and a date, and 244 nuclear sites are filtered to those active in each case’s era — you can’t be near a site that didn’t exist yet.
Build a population-weighted null
Random comparison points are drawn from real GeoNames cities, matched by country and era, so the null carries the same “sightings happen where people are” bias as the real data.
Permute 500 times
For each of five site layers and three radii, we compare the observed share of cases within range against 500 reshuffles of the null — giving an enrichment ratio and a p-value.
Guard with a placebo & a confound
Fossil-fuel plants (same footprint, nothing nuclear) set the floor; non-nuclear military bases test whether the effect is really about “secret bases.” Weapons beats both.
Within 100 km, USA vs rest-of-world
The weapons signal is far stronger in the USA — where the weapons complex and the reporting culture both concentrate. Bars show share within 100 km; the figure at right is the median distance to the nearest site.
Nuclear assets that move
A fixed-site distance test buries encounters at sea as random ocean points. But modern US carriers (since 1961) and most submarines are nuclear-powered — a mobile nuclear layer. These are logged categorically, not distance-tested.
