Volcanoes
11,178 eruptions. A case study in why era-splitting matters.
Decision Chain: Volcanoes
Ingest Smithsonian Global Volcanism Program data. 11,178 eruption events.
11,178 events ingested.
Fit self-exciting point process to eruption timestamps.
Full-dataset branching ratio 0.912. Appears highly self-exciting.
Investigate the high ratio. Volcanic monitoring networks expanded dramatically over the 20th century — more stations means more detected eruptions.
Monitoring expansion creates artificial clustering in the historical record. Era-corrected branching ratio: 0.112.
After removing the monitoring expansion artifact, assess whether real volcanic triggering exists.
Corrected ratio 0.112. Some genuine physical triggering exists (eruptions can trigger nearby eruptions), but the effect is modest.
Monitoring expansion inflated the ratio 8x
Without era correction, volcanoes appear to be 91.2% self-exciting — nearly as contagious as social media posts. After correcting for the expansion of monitoring networks worldwide, the true branching ratio drops to 0.112. Some genuine physical triggering exists (stress transfer, shared magma systems), but the uncorrected value is almost entirely an artifact of improved detection. This dataset, along with fireballs, proves that era-split analysis is mandatory for any long-running geophysical dataset.