VISITATIONS

Hatch's database catalogs all UFO sightings, not just entity encounters. We filter down to only the records where Hatch noted some kind of occupant or figure — a humanoid, a pseudo-human, a monster, a robot, a man-in-black, or similar — narrowing the 18,123-record catalog to the 1,508 entity encounters that matter for Visitations. Hatch's labels are translated onto our entity taxonomy, a Hynek classification is inferred from his activity tags (abduction becomes CE-IV, conversation becomes CE-V, otherwise CE-III), and dates are normalized while preserving his approximate-date markers. Coordinates, source references, credibility and strangeness scores, and the original title and description all travel through unchanged.

When one of our written cases is built from a specific Hatch record, the pin for that record is suppressed and the full case pin takes its place — no duplicate dots, no data discarded. And when a Hatch record describes the same real-world event as a Magonia record, the two collapse into a single pin on the map; the popup then shows a lineage strip of catalog pills, so you can click across to read each catalog's own account of the same event. Pairs at zero day-delta and within five kilometres auto-confirm; ambiguous cases are reviewed by hand.

Magonia has no machine-readable "occupant present" flag — unlike Hatch, which tags every record with one of eight entity codes, Vallée's catalog gives us only a short prose description per case. Keyword filters that approximate "does this record mention a being?" turn out to be both lossy (missing cases like "two men wearing furs" or "shapes like potato bags moving about the object") and false-positive-prone (snagging maritime language like "witnesses aboard the ship Vultur"), so we read every record individually and decide include or exclude per entry. From the 923 records in the catalog we keep the 329 that describe a non-witness being, each assigned to an entity type bucket matching our taxonomy with a Hynek classification inferred from the narrative. Dates are normalized across Vallée's many formats and locations are split into locality and country with US-state recognition so "Springfield, Illinois" resolves correctly.

Magonia records don't carry coordinates natively, so each stub's locality is geocoded automatically against a global gazetteer (with an on-disk cache and hand-placed overrides for the stragglers). The vast majority resolve cleanly; a handful of garbled OCR spellings — "Bajeola" for Bajada, "Pasnembi" for Panambi, "Aria" for Arica — had to be recovered by cross-referencing other catalogs and period reports before the true locations could be pinned.

As with Hatch, when one of our written cases is built from a specific Magonia entry the stub is suppressed and the written case's pin takes its place. Cross-catalog duplicates between Magonia and Hatch are collapsed into a single pin — same mechanism as above — so the popup lists every catalog that indexed the event as a chronological strip of pills.