A deep learning approach for the development of an Early Earthquake Warning system
Marco Carratù, Vincenzo Gallo, Vincenzo Paciello, Antonio Pietrosanto
Abstract
In the recent period, machine learning approaches have been widely used in many different fields. For example, in such applications where high immunities to noisy conditions are required. This is the case of an Early Earthquake Warning (EEW) system, a common technology used today to issue an alert in case of incoming seismic events. However, since most seismometers are installed in different locations of the Earth's surface, and different mechanical properties characterize them, each interpretation of a seismic earthquake could result in a highly complex task to be done in real-time using traditional approaches. Therefore, the proposed research has investigated the development of an innovative EEW system based on a novel deep learning system using both Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). The novel approach has been trained on about 5000 events retrieved from the IRIS University consortium. The achieved results have shown the excellent architecture capability in fully discovering the arrival of seismic events and good performance in the scoring of event intensity.