UAP AnalysisIndependent · the declassified record

Geographic distribution

173 of 212 records geolocated

Only records with a usable location appear here. Hollow, faded markers are placed at a regional centroid because the document redacted or omitted exact coordinates — treat their position as approximate, never precise.

Dept. of StateDept. of DefenseOtherFBICIADept. of Energy approximate / centroid
Loading map…

Toggle the nuclear-sites layer to see proximity; the panel reports the share of currently-shown incidents within 100 mi of a major nuclear site. Proximity is context for inquiry, not evidence of a link.

Geographic patterns — and why to read them carefully

latitude, hemisphere, proximity to strategic sites

Two patterns are measurable in the geocoded records. Both are real, and both are dominated by the same fact: this is a U.S.-government corpus. It records where the government observes and mandates reporting — not where the phenomenon is.

Latitude distribution · 173 geocoded
-10°–0°1
10°–20°11
20°–30°24
30°–40°83
40°–50°45
50°–60°7
60°–70°2

~74% fall in the 30°–50°N band and 64% are in the Western hemisphere — which is also the latitude of the continental U.S. and its main military theaters. The band coincides with jet-stream latitudes, but we have no wind/altitude data to test that, and it cannot be separated from simply being where the U.S. operates.

Proximity to strategic installations · 99 CONUS records vs random null
within 50 km15.2× vs chance
incidents21%
random1%
within 100 km4.8× vs chance
incidents26%
random5%
within 150 km3× vs chance
incidents35%
random12%

Tested against 4,000 random U.S. land points (so it controls for base density): incidents sit a median 226 km from a major installation vs 396km for random points. The clustering is real — not just “the U.S. is full of bases.”

Read the proximity carefully — what it does and does not show

That installations attract incidents is established. That they attract objects is not — because the same map is predicted by where we are able to see and obligated to report:

  • It is a government dataset. Every record originates with a U.S. agency, so locations are concentrated where the U.S. has jurisdiction, installations, and operations.
  • Reporting is mandated near assets. Military aircrew and base personnel are required to file; a sighting over a base enters a channel that a sighting over open country does not.
  • Monitoring density is wildly uneven. Bases, nuclear sites and coasts carry radar, FLIR and trained observers. You can only detect what you have sensors pointed at — so detections concentrate where the sensors are.
  • ~31% are maritime, but that largely reflects naval operating areas (carrier groups, fleet exercises), not an independent affinity for water.

So the honest statement is the 15.2× clustering within 50 km is consistent with two explanations we cannot separate here: objects drawn to strategic sites, or strategic sites being where observation and reporting are densest. The pattern narrows the question; it does not answer it.

Do they cluster in time, too?

Yes — the 33 near-installation records are not spread evenly. 30 of them fall before 1960, concentrated in 1947–1954, with a tight 1948–50 cluster of 11 incidents around the New Mexico nuclear-weapons corridor (Los Alamos, Sandia/Kirtland) — several logged within a few km of the labs. The most repeatedly-visited sites in the record:

Edwards/Muroc · 9 incidents · 1947–1966Sandia/Kirtland · 7 incidents · 1949–1964Wright-Patterson · 5 incidents · 1948–1954Los Alamos · 5 incidents · 1948–1986Holloman · 2 incidents · 1949–1957Wright Field · 2 incidents · 1954

Same caution, sharper: that 1948–50 window is exactly when the “green fireballs” alarmed the Atomic Energy Commission enough to stand up a dedicated instrumented watch (Project Twinkle) at those nuclear sites — and 1947 was the original nationwide sighting flap. So the space–time cluster reflects a burst of monitoring and reporting at sensitive sites at least as much as any concentration of objects. A genuine spatio-temporal cluster exists in the record; its cause is not separable here.