Nuclear Facilities and UAP: The Statistical Case for a Pattern

One of the most persistent claims in UAP research is that sightings cluster near nuclear facilities — weapons storage sites, ICBM fields, nuclear power plants, and strategic military installations. The claim dates back to the late 1940s, when sightings near Los Alamos, Oak Ridge, and Hanford — the three pillars of the Manhattan Project — were frequent enough to generate classified Air Force concern.

But is this pattern real, or an artifact of reporting bias?

AUSPEX applies formal statistical testing to answer this question.

UAP incidents occur near nuclear and strategic military facilities at a rate higher than chance would predict. Proponents point to dozens of well-documented cases:

• ◆Malmstrom AFB (1967) — ICBMs taken offline during UAP overflight
• ◆Minot AFB (1968) — B-52 crew and ground personnel observed UAP near missile silos
• ◆Loring AFB (1975) — unknown craft penetrated the weapons storage area
• ◆Rendlesham Forest (1980) — adjacent to RAF Woodbridge, a NATO nuclear weapons storage site
• ◆Chernobyl (1986) — multiple witnesses reported UAP during the reactor disaster
• ◆Fukushima (2011) — UAP reported over the reactor complex during the meltdown

The pattern is suggestive. But anecdote is not evidence. We need a null model.

The fundamental question: if UAP incidents were distributed randomly across the United States, how many would we expect to fall within 100km of a nuclear/strategic facility by pure chance?

To build a null model, we need to know what percentage of U.S. land area falls within the "proximity zone" of nuclear sites. AUSPEX defines 15 nuclear and strategic installations:

| Site | Type | |------|------| | Malmstrom AFB, MT | ICBM Field | | Minot AFB, ND | ICBM Field | | F.E. Warren AFB, WY | ICBM Field | | Whiteman AFB, MO | B-2 Stealth Bomber | | Nellis AFB / NTS, NV | Nuclear Test Site | | Hanford Site, WA | Plutonium Production | | Oak Ridge, TN | Uranium Enrichment | | Los Alamos, NM | Weapons Laboratory | | Savannah River, SC | Tritium Production | | Area 51 / Groom Lake, NV | Classified Test Facility | | Edwards AFB, CA | Flight Test Center | | Vandenberg SFB, CA | ICBM Test Launch | | Wright-Patterson AFB, OH | Foreign Technology Division | | Langley AFB, VA | Air Combat Command | | Cheyenne Mountain, CO | NORAD |

Each site gets a 100km radius proximity zone. The combined area of these 15 circles (accounting for some overlap) covers approximately 4.8% of the contiguous United States land area.

If incidents are randomly distributed, we would expect approximately 4.8% of U.S. incidents to fall within 100km of one of these sites. This is our expected frequency under the null hypothesis.

We then count the observed frequency — how many incidents in the AUSPEX corpus actually fall within 100km of a nuclear site — and compare the two using a chi-squared goodness-of-fit test with 1 degree of freedom.

The formula:

chi-squared = sum of (observed - expected)^2 / expected

For two categories (near nuclear / not near nuclear):

chi-squared = (O_near - E_near)^2 / E_near + (O_far - E_far)^2 / E_far

At the p < 0.05 significance level, the critical chi-squared value with 1 degree of freedom is 3.84. If our computed chi-squared exceeds this threshold, we reject the null hypothesis — the clustering is statistically significant.

Running this analysis on the AUSPEX incident corpus produces the following results, which you can verify in our Analysis module (MODULE 12 > MILITARY PROXIMITY):

The analysis computes observed vs. expected incident counts near our 15 defined sites, performs the chi-squared test, and reports both the test statistic and the significance determination.

Interpretation caveats:

Our current null model uses a naive area-fraction approach (4.8% of U.S. area). This has known limitations:

1. ◆

   Population density confound — nuclear facilities tend to be in less populated areas, which should mean fewer reports nearby, not more. If anything, the area-fraction model is conservative (it overestimates the expected near-nuclear count).

2. ◆

   Reporting bias — military personnel near these facilities may be more likely to report sightings due to security awareness. This could inflate the near-nuclear count.

3. ◆

   Observer density — NUFORC and MUFON reports correlate with population density. A proper null model should use population-weighted expected values, not raw area.

4. ◆

   Small sample size — with 75 verified incidents in our core corpus, statistical power is limited. The test becomes more meaningful at scale (140K+ incidents from NUFORC/MUFON).

A statistically significant result does not prove that UAP are attracted to nuclear facilities. It proves that the observed clustering is unlikely to occur by chance under our null model. There are several possible explanations:

• ◆Active surveillance: If UAP represent an intelligence-gathering technology, nuclear weapons infrastructure would be a logical surveillance target.
• ◆Sensor density: Military bases have more radar, more cameras, and more trained observers. They may simply detect more UAP.
• ◆Geomagnetic factors: Some researchers hypothesize that nuclear sites correlate with geomagnetic anomalies that may attract or generate the phenomenon.
• ◆Reporting culture: Military personnel at sensitive installations may report anomalies more rigorously than civilians.

The statistical test cannot distinguish between these explanations. What it can do is establish that the pattern is real — not an artifact of cherry-picked anecdotes.

The AUSPEX proximity analysis is implemented in src/lib/analysis.ts as the proximityAnalysis() function. It:

1. ◆Defines 15 nuclear/strategic sites with lat/lng coordinates
2. ◆For each incident, computes haversine distance to all 15 sites
3. ◆Counts incidents within the specified radius (default: 100km)
4. ◆Computes expected count using the area-fraction null model
5. ◆Runs the chi-squared test (1 degree of freedom)
6. ◆Reports significance at p < 0.05

The analysis is fully transparent and reproducible. All code is available for inspection. The incident corpus is exportable as CSV from the Analysis module.

Our proximity analysis will improve significantly when we integrate:

• ◆U.S. Census population density rasters — for a population-weighted null model that accounts for observer density
• ◆Kulldorff scan statistics — a more rigorous spatial-temporal clustering method used in epidemiology
• ◆Incident-level confidence weighting — weighting confirmed/radar-backed incidents higher than unverified reports
• ◆International nuclear sites — expanding beyond U.S. installations to test the pattern globally

The question of whether UAP cluster near nuclear facilities is one of the most consequential in the field. If the pattern holds under rigorous statistical scrutiny at scale, it has profound implications for understanding what these phenomena are — and what they might want.

The nuclear-proximity case in book form:

• ◆Robert Hastings, UFOs and Nukes: Extraordinary Encounters at Nuclear Weapons Sites (CreateSpace, 2017 ed.) — the foundational book-length treatment, decades of FOIA work and witness interviews — Bookshop · Amazon
• ◆Robert Salas and James Klotz, Faded Giant: The 1967 Malmstrom AFB UFO Incident & Missile Shutdown (BookSurge, 2007) — primary-source account of the Echo Flight / Oscar Flight ICBM-shutdown incident — Amazon

Statistical and methodological context:

• ◆Mick West, Escaping the Rabbit Hole: How to Debunk Conspiracy Theories Using Facts, Logic, and Respect (Skyhorse, 2018) — base-rate thinking and the dataset hygiene that this analysis tries to apply — Bookshop · Amazon
• ◆Edward Tufte, The Visual Display of Quantitative Information (Graphics Press, 2nd ed.) — methodological foundation for the kind of clustering visualization AUSPEX does in the Analysis module — Bookshop · Amazon

Disclosure-era context:

• ◆Leslie Kean, UFOs: Generals, Pilots, and Government Officials Go on the Record (Harmony, 2010) — multiple chapters on nuclear-site encounters from named military personnel — Bookshop · Amazon
• ◆Luis Elizondo, Imminent: Inside the Pentagon's Hunt for UFOs (William Morrow, 2024) — discusses the nuclear-proximity pattern from inside AATIP — Bookshop · Amazon

Primary sources (free):

• ◆Robert Hastings's compiled FOIA documents — ICBM-site incident reports
• ◆DoD AARO — Historical Record Report Vol I — covers nuclear-site reporting in section 5
• ◆The 2010 National Press Club press conference on UFO/nuclear incidents featuring multiple ex-military witnesses

Affiliate disclosure. Some of the book links in this post are affiliate links — primarily through Bookshop.org (which supports independent bookstores), with Amazon as a secondary fallback. As an Amazon Associate, AUSPEX earns from qualifying purchases. If you make a qualifying purchase via these links, AUSPEX may earn a small commission at no additional cost to you. The site does not host or reproduce any copyrighted text from any of the works mentioned; quotations above are brief and used for commentary purposes under fair use.

All analysis referenced in this article is computed live in the AUSPEX Analysis module (MODULE 12 > MILITARY PROXIMITY). Data and methodology are fully transparent.