Seismicity rate modeling for prospective stochastic forecasting: the case of 2014 Kefalonia, Greece, seismic excitation

Abstract

We examined the January-February 2014 earthquake doublet (M w = 6.1 and M w = 6.0) and the associated aftershocks which form a seismic excitation adequately well recorded by a dense local seismological network. It started on January 26 with the main shock, causing a lot of panic and followed by numerous aftershocks. The second main shock with M w = 6.0 occurred 7 days later on an along-strike adjacent fault segment. The close proximity of the two main shocks, in both space and time and the intense aftershock sequence, triggered the investigation of the occurrence probability evolution for the stronger aftershocks and possibly a third main shock in the seismic excitation. This purpose was further motivated by the potential of the area for hosting a stronger (M w C 6.0) earthquake based upon both historical information and instrumental data. Aftershock rate modeling was done on subsequent data samples by the restricted epidemic-type aftershock sequence stochastic model, and probabilities for the occurrence of strong (M w C 5.0) earthquakes were calculated during the progress of the aftershock sequence. We executed daily model simulations and probability forecasts for 30 days focusing in more detail on the impact of some model parameters on the prospective forecasting. Trying to be near to a real-time case, all forecasts were done on data up to the moment of forecasting.

Precursory Activity Before Larger Events in Greece Revealed by Aggregated Seismicity Data

“…Additionally, it is well known that aftershock sequences mean that seismicity rates are often far from homogeneous (Felzer and Brodsky 2005;Marsan 2003). One approach is to try to identify and remove aftershocks, to produce a possibly near-homogeneous mainshock sequence (Gomberg et al 2001;Kilb et al 2000;Matthews and Reasenberg 1988;Wyss and Wiemer 2000), although results may be dependent upon the choice of methodology, such as declustering algorithms used.…”

“…Another way to describe the temporal patterns of seismicity is by means of cascade models (Felzer et al 2015;. Testing data for consistency with an ETAS-type model can be complex, partly because of issues related to, e.g.…”

“…Console et al (2006) worked with data from Greece and applied statistical models on short and long time scales, while combining temporal and spatial modelling to investigate the activity prior to major events. Gospodinov et al (2015) applied the Restricted Epidemic Type Aftershock Sequence (RETAS) model aiming to identify precursors and periods of relative quiescence prior to strong aftershocks and estimate the occurrence probabilities of new strong events in a sequence.…”

Evidence of Precursory Patterns in Aggregated Time Series

“…Papadopoulos et al (2000) identified foreshock activity during less than 4 months prior to 12 out of 17 strong earthquakes (M>5) around the Corinth Gulf. Gospodinov et al (2015) recently presented a day by day forecast of seismicity working with data in the Kefalonia area. Here, we use interevent times between earthquakes within a specified area as a timespecific proxy for seismicity rate.…”

Στοχαστική Μοντελοποίηση Της Σεισμογένεσης Στην Ελλάδα

“…Η εκδοχή του μοντέλου που εφαρμόζεται στην παρούσα διατριβή είναι αυτή που προτάθηκε από τους Console & Murru (2001) και Console et al (2003). Με την εφαρμογή του έγιναν προγνώσεις σε πραγματικό χρόνο από τους Console et al (2007) (Gospodinov et al, 2007(Gospodinov et al, , 2015Papadimitriou et al, 2013;Karakostas et al, 2014) αλλά και σε σεισμικές εξάρσεις (Mesimeri et al, 2018) , όπου Εi είναι η σεισμική ενέργεια του i-οστού σεισμού και N(t) ο αριθμός των σεισμών έως το χρόνο t. Σε αυτή την περίπτωση η=0.75 και η σχέση (2.3) γίνεται 𝑆 𝑖 = 10 0.75(𝑀 𝑖 −𝑀 𝑡ℎ ) .…”