Each anomaly layer plotted independently. Brighter zones indicate multi-layer convergence — the primary signal of a site warranting ground investigation. Toggle layers to isolate specific signal types.
Sites ordered chronologically. Clustering in certain epochs may indicate bombardment cycles or active periods of unknown causation.
Confidence weights data quality, source independence, publication volume, and recency. A site can have high convergence but lower confidence if it relies on few primary sources or outdated surveys.
Live API feeds are fetched client-side when site cards are opened. Cached data supplements live feeds for sites outside observatory coverage zones. All anomaly descriptions are cross-referenced against peer-reviewed publications.
| Database | Data type | Coverage | Status | Updated |
|---|
| Layer | Type | Weight | What we look for |
|---|
Historical and mythological accounts georeferenced and cross-matched against known geological anomaly zones. Independent cultural convergence on the same region is treated as a supporting data layer.
Witnessed events with official investigations, physical evidence, and geographic coordinates. These cases are distinct from the geological anomaly sites — they represent documented incidents with known or estimated impact locations, investigated by military, government, or scientific bodies.
Primary source declassified government documents relevant to UAP research, crash site investigation, and aerial anomaly analysis. All documents are publicly available through official government archives. Links open the original source document.
An original quantitative analysis of 137,009 UFO reports from the NUFORC database (1930–2022). We scored each report on observed behavior — silent flight, hovering, instant acceleration, electromagnetic effects — drawn from the free-text summaries, then tested whether the highest-anomaly reports cluster geographically.
Seven binary dimensions were scored via regex on cleaned summary text. Each captures a physically anomalous behavior that is hard to fake or misremember — and, unlike object shape, is not something media coverage can prime a witness to report:
BAI = sum of dimensions present. BAI≥2 is treated as a strong signal (1.0% of reports, n=1,170 in the US subset); BAI≥3 as very strong (43 reports, 0.04%).
An event study tested report characteristics against major media shocks. Witness-reported shape spikes hard after media events — the Phoenix Lights produced a +3.87pp jump in triangle reports (p=0.0002). The BAI showed zero statistically significant response to any media event (all p>0.09). This is the core validation: BAI measures something the culture isn't putting there.
The silent + hover + instant triplet appears in every decade from the 1960s through the 2010s, with no cross-witness coordination mechanism. The apparent rate decline post-1995 is a denominator effect — the internet era floods the dataset with low-quality misidentification reports, while the absolute count of BAI≥2 reports continues to grow.
State-level OLS regression of BAI rate on geographic features. Low mean elevation is the strongest single predictor. Continental shelf width and submarine canyon access are positive and significant. The full model explains ~22% of cross-state variance (R²=0.222).
Grouping geocoded reports by nearest canyon region, New England canyons show the highest BAI rate and the Pacific the lowest. Two independent quality measures — the algorithmic BAI and NUFORC's human-expert Tier 1 grading — point the same direction: Tier 1 reports are more canyon-proximate than the baseline (56% within 500km vs 42% overall).
If you have identified a location with anomalous geological, geochemical, or historical characteristics that may warrant investigation, submit it here. All submissions are reviewed against our nine-layer scoring methodology before being added to the database.