Single Layer Recurrent Neural Network for detection of local swarm-like earthquakes-the application

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F67985530%3A_____%2F19%3A00508961" target="_blank" >RIV/67985530:_____/19:00508961 - isvavai.cz</a>

Result on the web

<a href="https://academic.oup.com/gji/article-abstract/219/1/672/5532359?redirectedFrom=fulltext" target="_blank" >https://academic.oup.com/gji/article-abstract/219/1/672/5532359?redirectedFrom=fulltext</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1093/gji/ggz321" target="_blank" >10.1093/gji/ggz321</a>

Result language

angličtina

Original language name

Single Layer Recurrent Neural Network for detection of local swarm-like earthquakes-the application

Original language description

We present results of applying a local event detector based on artificial neural networks (ANNs) to two seismically active regions. The concept of ANNs enables us to recognize earthquake-like signals in seismograms because well-trained neural networks are characterized by the ability to generalize to unseen examples. This means that once the ANN is trained, in our case by few tens to hundreds of examples of local event seismograms, the algorithm can then recognize similar features in unknown records. The detailed description of the single-station detection, design and training of the ANN has been described in our previous paper. Here we show the practical application of our ANN to the same seismoactive region we used for its training, West Bohemia/Vogtland (border area Czechia-Saxony, local seismic network WEBNET), and to different seismogenic area, Reykjanes Peninsula (South-West Iceland, local seismic network REYKJANET). The training process requires carefully prepared data set which is preferably achieved by manual processing. Such data were available for the West Bohemia/Vogtland earthquake-swarm region, so we used them to train the ANN and test its performance. Due to the absence of completely manually processed activity for the Reykjanes Peninsula, we use the trained ANN for swarm-like activity in such a different tectonic setting. The application of a coincidence of the single-station detections helps to reduce significantly the number of undetected events as well as the number of false alarms. Setting up the minimum number of stations which are required to confirm an event detection enables us to choose the balance between minimum magnitude threshold and a number of false alarms.

Czech name

—

Czech description

—

Type

J_imp - Article in a specialist periodical, which is included in the Web of Science database

CEP classification

—

OECD FORD branch

10507 - Volcanology

Project

Result was created during the realization of more than one project. More information in the Projects tab.

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year

2019

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Name of the periodical

Geophysical Journal International

ISSN

0956-540X

e-ISSN

—

Volume of the periodical

219

Issue of the periodical within the volume

1

Country of publishing house

GB - UNITED KINGDOM

Number of pages

18

Pages from-to

672-689

UT code for WoS article

000484124800041

EID of the result in the Scopus database

—