An Earthquake-Clustering Model in North Aegean Area (Greece)

The investigation of short-term earthquake-clustering features is made feasible through the
application of a purely stochastic Epidemic-Type Aftershock Sequence (ETAS) model. The learning
period that is used for the estimation of the parameters is composed by earthquakes with M>= 2.6 that
occurred between January 2008 and May 2017. The model predictability is retrospectively examined
for the 12 June 2017 Lesvos earthquake (Mw6.4) and the subsequent events. The construction of timedependent
seismicity maps and comparison between the observed and expected earthquake number
are performed in order to temporally and spatially test the evolution of the sequence, respectively.
The generation of 127 target events with M >= 3.0 in the period June–July 2017, just before the main
shock occurrence, is examined in a quantitative evaluation. The statistical criteria used for assessing
the model performance are the Relative Operating Characteristic Diagram, the R-score, and the
probability gain. Reliable forecasts are provided through the epidemic model testifying its superiority
towards a time-invariant Poisson model.