Statistical analysis of daily seismic event rate as a precursor to volcanic eruptions

We analyse time series of daily seismic event rate for the Kilauea, Hawaii, volcano between 1959 and 2000. Individual eruptions are not always preceded by an increase in event rate, and many increases in event rate do not lead to eruption. However, a mean field accelerating behaviour does emerge 10–15 days before eruption in data stacked in phase with the eruption time. In phase space the pre‐eruptive dynamics is well defined by Voight's [1988] equation, but so is that of the seismicity in the period between eruptions. We conclude that the underlying dynamics of the 'background' seismicity is similar to that of magma eruption. We use Bayesian methods to compare different time‐to‐failure models that have been suggested for precursors. Only a short‐term forecast can be achieved, using a linear fit to inverse rate.