Forecasting Strong Subsequent Earthquakes in Greece with the Machine Learning Algorithm NESTORE

Eleni-Apostolia Anyfadi; Stefania Gentili; Piero Brondi; Filippos Vallianatos

doi:10.3390/e25050797

Forecasting Strong Subsequent Earthquakes in Greece with the Machine Learning Algorithm NESTORE

Entropy (Basel). 2023 May 13;25(5):797. doi: 10.3390/e25050797.

Authors

Eleni-Apostolia Anyfadi^{1

2}, Stefania Gentili³, Piero Brondi³, Filippos Vallianatos^{1

2}

Affiliations

¹ Section of Geophysics-Geothermics, Department of Geology and Geoenvironment, National and Kapodistrian University of Athens, 15784 Athens, Greece.
² Institute of Physics of Earth's Interior and Geohazards, UNESCO Chair on Solid Earth Physics and Geohazards Risk Reduction, Hellenic Mediterranean University Research & Innovation Center, 73133 Chania, Greece.
³ National Institute of Oceanography and Applied Geophysics-OGS, 33100 Udine, Italy.

Abstract

Aftershocks of earthquakes can destroy many urban infrastructures and exacerbate the damage already inflicted upon weak structures. Therefore, it is important to have a method to forecast the probability of occurrence of stronger earthquakes in order to mitigate their effects. In this work, we applied the NESTORE machine learning approach to Greek seismicity from 1995 to 2022 to forecast the probability of a strong aftershock. Depending on the magnitude difference between the mainshock and the strongest aftershock, NESTORE classifies clusters into two types, Type A and Type B. Type A clusters are the most dangerous clusters, characterized by a smaller difference. The algorithm requires region-dependent training as input and evaluates performance on an independent test set. In our tests, we obtained the best results 6 h after the mainshock, as we correctly forecasted 92% of clusters corresponding to 100% of Type A clusters and more than 90% of Type B clusters. These results were also obtained thanks to an accurate analysis of cluster detection in a large part of Greece. The successful overall results show that the algorithm can be applied in this area. The approach is particularly attractive for seismic risk mitigation due to the short time required for forecasting.

Keywords: Greek seismicity; NESTORE; aftershocks; clusters; features; forecasting; machine learning algorithm; training procedure.

Grants and funding

Funded by a grant from the Italian Ministry of Foreign Affairs and International Cooperation and Co-funded by the Erasmus+ programme of the European Union (EU).