The location of past earthquakes is used to make areal sources of future earthquakes. The principal being applied is this: Future earthquakes will most likely happen in the vicinity of past earthquakes. And if a large earthquake has happened somewhere in the past, future large earthquakes are more to occur where past smaller earthquakes have occurred rather than where no past earthquake have occurred.
Earlier methods used areal source zones to delineate where future earthquakes would occur. A geographically continuous area of geologically-related earthquakes would be enclosed in a geometric zone. Inside the zone earthquakes would occur, having the same rate of occurrence and range of magnitudes as the historical earthquakes. The earthquakes would be equally likely at all locations within the zone.
Later methods assume that the vicinity of past earthquakes is instead described by a decreasing probability, the further a site is from a past earthquake. Now the earthquake rates for future earthquakes depend on the spatial distribution of past earthquakes, rather than being equally likely inside a zone. However the relative magnitude distribution is assumed the same everywhere.