The Kumbh moves between four river sites — Prayagraj, Haridwar, Ujjain and Nashik — on a cycle set by the positions of Jupiter, the Sun and the Moon, which works out at roughly twelve years at each. A temporary city is built on the floodplain with roads, power, hospitals and police stations, and dismantled afterwards.
The event is the Amrit Snan, the bathing days, when the akharas — the monastic orders, in a fixed and jealously defended order of precedence — process to the water. The 2025 Prayagraj mela ran 13 January to 26 February with principal bathing days on 14 and 29 January, 3, 12 and 26 February.
The attendance figure is the most instructive number in this atlas, and we do not record it. More than 660 million was announced for 2025. It is a count of dips taken, assembled from manual counts, train arrivals, satellite imagery and camera footage across forty-five days; a pilgrim who bathed on four days is in it four times. The Economist observed that the figure implies half of India attended. There is no published method, no named issuing body in the sources we reached, and no statement of whether it counts people or events. It is filed unusable with that reasoning rather than printed.
The death toll is disputed in the same way. The crush of 29 January was given officially as about 30 dead; a count based on hospital and police records put it at at least 79, and opposition parties said 1,500 remained unaccounted for a week later.
The claim that the 2025 edition was a once-in-144-years alignment is contested and we do not assert it.
PlanningFOUR SITES on a roughly twelve-year rotation. The bathing days — Amrit Snan — are the event; the rest is a city.
CalendarNot computable here, and the rule now says so instead of pretending. The dates come from the positions of Jupiter, the Sun and the Moon, worked out for each site by astrologers and then announced by the state government and the akharas — so this entry is read from published tables rather than derived, and the engine returns that fact rather than a date. The error this replaces is worth naming: the record was classed cyclic, carried a nine-month window and NO cycle, and therefore computed a Kumbh Mela every single year at four cities at once. It was found by a check that compares what an entry SAYS with what its rule DOES — the entry's own date line read 'roughly every twelve years at each of four sites'.
AlongsideArdh Kumbh Mela Q-IN-001 — the two meet in the same place and week, and neither contains the other.
One entry, 3 separate runnings
Four sites, each on its own cycle of about twelve years, and the atlas had them as ONE annual window from January to September — which said, in effect, that there is a Kumbh Mela somewhere every year. The half-Kumbhs at Prayagraj and Haridwar are a separate record in this atlas and are not listed here.