GNGTS 2023 - Atti del 41° Convegno Nazionale

Session 1.1 GNGTS 2023 better suitable for automated processing. Figure shows the comparison between the coherence matrices of a true seismic event and a false one. When applied to continuous seismic data, waveform staking methods produce coherence matrices with clear patterns that can be used to distinguish true events from false ones (i.e. noise). In fact the coherence matrices for a seismic event generally show a single and well focused maximum while pure noise waveforms produce blurred images with low coherence values or many poorly focused maxima. Deep Learning algorithms are the perfect tool to classify these kinds of images and improve the detection capability of waveform based techniques. The aim of this work consists in the development of a framework that simultaneously detect and locate seismic events through a Convolutional Neural Network (CNN) that classify earthquake signals from noise; more specifically the CNN classify the coherence matrix, which is the output of waveform based method. In order to produce the synthetic coherence matrices needed to train the CNN, we first generate synthetic waveforms for a set of seismic events, then we add noise records with the same spectral properties of the observed one and generated through stochastic modeling. For each synthetic events or pure noise recording we use the waveform stacking to generate a coherence matrices that will be used to train the CNN. One important feature of the workflow here exposed is that the training is performed entirely on synthetics without the need of large labeled data, often missing when new microseismic networks are deployed. To test the workflow we apply it to the recently released dataset collected within the COSEISMIQ project (Grigoli et al 2022). This dataset consists of 2 years of continuous seismic waveforms acquired by a temporary microseismic network deployed in the Hengill geothermal region (Iceland). The released dataset also contains seismicity catalogues that have been used both as source of information for the synthetics generation and as a benchmark for the performance of the detector. References Mousavi, S.M. and Beroza, G.C., 2022. Deep-learning seismology.  Science , 377 (6607), p.eabm4470. Li, L., Tan, J., Schwarz, B., Stan ě k, F., Poiata, N., Shi, P., Diekmann, L., Eisner, L. and Gajewski, D., 2020. Recent advances and challenges of waveform‐based seismic location methods at multiple scales. Reviews of Geophysics , 58 (1), p.e2019RG000667. Grigoli, F., Clinton, J.F., Diehl, T., Kaestli, P., Scarabello, L., et Al. 2022. Monitoring microseismicity of the Hengill Geothermal Field in Iceland.  Scientific Data , 9 (1), pp.1-11.