SCALODEEP: A Highly Generalized Deep Learning Framework for Real‐Time Earthquake Detection

Abstract

The detection of earthquake signals is a fundamental yet challenging task in observational seismology. A robust automatic earthquake detection algorithm is strongly demanded in view of the ever‐growing global seismic dataset. Here, we develop an automatic earthquake detection framework based on a deep learning approach (SCALODEEP). It extracts high‐order features embedded in three‐component seismograms by encoding a time‐frequency representation of the data (scalogram) into a deep network with skip connections. The SCALODEEP is trained and validated on an open‐source dataset from North California, and then employed to seismicity detection in four areas, including Arkansas, Japan, Texas, and Egypt. Despite vastly varying characteristics of regional earthquakes (e.g., focal mechanism, duration, and noise level), SCALODEEP successfully detects seismic signals over a broad range of local magnitudes (as low as −1.3normalML) and outperforms conventional algorithms such as STA/LTA, FAST, and template matching. Compared to recently proposed deep learning based frameworks (e.g., CRED and Earthquake transformer), SCALODEEP achieves a superior generalization ability via a sophisticated network architecture. In summary, our study offers a promising new tool to improve existing earthquake detection systems and, as importantly, sheds light on designing an effective deep learning network for generalized earthquake detection.

Augmenting Seismic Data Using Generative Adversarial Network for Low-Cost MEMS Sensors

“…In this section, we first present the basic theoretical framework of standard GAN in Part A, the characteristics of real seismic sequences are analyzed in Part B according to the evolution process of the seismic event. Finally, we also design the EQGAN model based on appropriate algorithms in Part C. Earthquake/Phase detection ConvNet incomplete/low SNR waveform Using transformation method [25] Eaethquake detection Unsupervised technique Microseismic dataset Using transformation method [26] Earthquake detection SCALODEEP Small training dataset Using generalizad deep learning based on a small daraset [27] Earthquake detection CPIC Small-sized dataset Using generalizad deep learning based on a small daraset [28] Earthquake detection CapsNet Small training dataset Using generalizad deep learning based on a small daraset [31] Synthesize speech HMM Speech dataset Developing a data augmentation approach [32] Text generation LSTM Text dataset Developing a data augmentation approach [33] Text summarization Seq2seq Text dataset Developing a data augmentation approach [29] Seismic data augmentation Conditional GAN Seismic dataset Developing a data augmentation approach [30] Short seismic waveform generation EarthuqakeGen Seismic dataset Developing a data augmentation approach A. THEORETICAL BASIS In 2014, Goodfellow [36] proposed the concept of GAN, an epoch-making unsupervised learning algorithm framework (Fig.…”

Real‐Time Earthquake Detection and Magnitude Estimation Using Vision Transformer

“…Recently, deep learning is widely used in the field of seismology, including earthquake detection and picking (Mousavi et al., 2020; Saad & Chen, 2021; Saad, Huang, et al., 2021; Zhu & Beroza, 2018), seismic data denoising (Saad & Chen, 2020; Saad, Oboué, et al., 2021; Yang et al., 2021; Zhang et al., 2019), estimation of earthquake location (Mousavi & Beroza, 2019; Münchmeyer et al., 2021), and seismic data inversion (Li et al., 2019; Liu et al., 2021; Ren et al., 2020; Zhang et al., 2021). The convolutional neural network (CNN) is a commonly used architecture due to its ability to extract significant features from the input data.…”

3D Microseismic Monitoring Using Machine Learning

“…Recently, machine learning (ML) methods for locating passive seismic sources are emerging. While most ML methods focus on intelligent data processing, for example, arrival picking (Mousavi et al., 2016; Mousavi & Langston, 2016; Qu et al., 2020; Saad et al., 2021; Saad & Chen, 2021; Song et al., 2010; Stork et al., 2020) or waveform detection (Chen, 2018; Saad & Chen, 2020), much less attention is given toward using ML to directly predict the source locations from the recorded seismic data. For the few ML‐based source‐location prediction methods (Mousavi & Beroza, 2019; Perol et al., 2018; Shen & Shen, 2021; van den Ende & Ampuero, 2020; Zhang et al., 2020), they are not generalizable due to the lack of a sufficiently large and diverse training database.…”