Comparison of Single-Trace and Multiple-Trace Polarity Determination for Surface Microseismic Data Using Deep Learning

Abstract

For surface microseismic monitoring, determination of the P-wave first-motion polarity is important because (1) it has been widely used to determine focal mechanisms and (2) the location accuracy of the diffraction-stack-based method is improved greatly using polarization correction. The convolutional neural network (CNN) is a form of deep learning algorithm that can be applied to predict the polarity of a seismogram automatically. However, the existing network designed for polarity detection utilizes only individual trace information. In this study, we design a multitrace-based CNN (MT-CNN) architecture using several neighbor traces combined as training samples, which could utilize the polarity information of neighbor sensors in the surface microseismic array. We use 17,227 field seismograms with labeled polarities to train two different neural networks that predict the polarities by a single trace or by multiple traces. The performance of the test set and field example of two CNN architectures shows that the MT-CNN significantly produces fewer polarity prediction errors and leads to more accurate focal mechanism solutions for microseismic events.

Recent advances in earthquake seismology using machine learning

“…Naoi et al (2022) automatically read the timing of the initial maximum amplitude using DL; the obtained amplitudes were used to solve moment tensors for laboratory AEs. In these analyses, waveforms from a single station were typically used as input data to a DL network; additionally, multi-station approaches have been explored (Tian et al 2020). Although it is not as active as the problem of arrival time reading, the generalizability of these models has also been studied (Hara et al 2019).…”

EdgePhase: A Deep Learning Model for Multi‐Station Seismic Phase Picking

“…The recent expansion of seismic data and computing resources enables flourishing applications of deep learning (DL) in seismology (Bergen et al., 2019; Bianco et al., 2019; Civilini et al., 2021; Kong, Inbal, et al., 2019; Kong, Trugman, et al., 2019; Kuang et al., 2021; LeCun et al., 2015; Li et al., 2018; Meier et al., 2019; Mousavi & Beroza, 2019, 2020; Nakano et al., 2019; Oord et al., 2016; Perol et al., 2018; Reynen & Audet, 2017; Ross et al., 2019; Saad & Chen, 2020; Saad et al., 2021; Schmidhuber, 2015; Smith et al., 2022; Tian et al., 2020; Valentine & Trampert, 2012; T. Wang et al., 2021). Many of these studies aim to automatically pick P and S arrivals, especially the signals of microseismicity buried under noises.…”

Deep Learning for Efficient Microseismic Location Using Source Migration‐Based Imaging

“…The reversed polarity can also be corrected by source moment tensor inversion (Anikiev et al., 2014; Zhebel & Eisner, 2015), moment tensor imaging (Chambers et al., 2014), or focal mechanism search (Liang et al., 2016). The polarity corrections based on amplitude trend least‐squares fitting (Xu et al., 2020) and convolutional neural network determination (Tian et al., 2020) were also applied to surface microseismic data. In the latter case, the imaging conditions for TRI mainly contained integral imaging, PS cross‐correlation imaging (Artman et al., 2010), interferometric imaging (Li et al., 2014; Wang et al., 2013; Zhang & Zhang, 2022), the PS interferometric cross‐correlation imaging (Zhou & Zhang, 2017; Zhou et al., 2022), energy imaging (Oren & Shragge, 2019; Rocha et al., 2019), and the geometric‐mean imaging (Lyu & Nakata, 2020; Nakata & Beroza, 2015).…”