当前位置：
X-MOL 学术
›
Seismol. Res. Lett.
›
论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
DynTriPy: A Python Package for Detecting Dynamic Earthquake Triggering Signals
Seismological Research Letters ( IF 3.3 ) Pub Date : 2020-10-21 , DOI: 10.1785/0220200216 Naidan Yun 1 , Hongfeng Yang 2 , Shiyong Zhou 1
Seismological Research Letters ( IF 3.3 ) Pub Date : 2020-10-21 , DOI: 10.1785/0220200216 Naidan Yun 1 , Hongfeng Yang 2 , Shiyong Zhou 1
Affiliation
Cite this article as Yun, N., H. Yang, and S. Zhou (2020). DynTriPy: A Python Package for Detecting Dynamic Earthquake Triggering Signals, Seismol. Res. Lett. XX, 1–12, doi: 10.1785/ 0220200216. Supplemental Material Long-term and large-scale observations of dynamic earthquake triggering are urgently needed to understand the mechanism of earthquake interaction and assess seismic hazards. We developed a robust Python package termed DynTriPy to automatically detect dynamic triggering signals by distinguishing anomalous seismicity after the arrival of remote earthquakes. This package is an efficient implementation of the high-frequency power integral ratio algorithm, which is suitable for processing big data independent of earthquake catalogs or subjective judgments and can suppress the influence of noise and variations in the background seismicity. Finally, a confidence level of dynamic triggering (0–1) is statistically yielded. DynTriPy is designed to process data from multiple stations in parallel, taking advantage of rapidly expanding seismic arrays to monitor triggering on a global scale. Various data formats are supported, such as Seismic Analysis Code, mini Standard for Exchange of Earthquake Data (miniSEED), and SEED. To tune parameters more conveniently, we build a function to generate a database that stores power integrals in different time and frequency segments. All calculation functions possess a high-level parallel architecture, thoroughly capitalizing on available computational resources. We output and store the results of each function for continuous operation in the event of an unexpected interruption. The deployment of DynTriPy to data centers for real-time monitoring and investigating the sudden activation of any signal within a certain frequency scope has broad application prospects. Introduction It has been widely reported that earthquakes can be triggered by transient stress changes related to the passage of seismic waves; this phenomenon is known as dynamic triggering (e.g., Hill et al., 1993; Gomberg and Johnson, 2005; Velasco et al., 2008; Peng et al., 2010, 2012; Pollitz et al., 2012; van der Elst et al., 2013; Aiken et al., 2018). Systematic investigations of dynamic triggering can help us explore the responses of faults to stress disturbances, and these responses can be leveraged to monitor temporal variations in stress and thus assess earthquake hazards. Moreover, the mechanisms of earthquake interactions are critical for advancing our understanding of earthquake physics (Hill and Prejean, 2007; Brodsky and van der Elst, 2014). In the Coulomb failure model, brittle failure occurs when a dynamic stress perturbation elevates the stress state on a fault plane, exceeding the frictional strength. Therefore, remote earthquakes with larger dynamic stress perturbations should have stronger triggering abilities (Gomberg and Davis, 1996; Gomberg et al., 2001; Wu et al., 2011; Aiken and Peng, 2014; Wang et al., 2015; Miyazawa, 2016). However, a large number of observations have demonstrated that other characteristics of seismic wavefields, such as the back azimuth, dominant frequency, and types of surface and body waves, also affect the dynamic triggering response (e.g., Prejean et al., 2004; Brodsky and Prejean, 2005; West et al., 2005; van der Elst and Brodsky, 2010; Hill, 2012; De Barros et al., 2017). Moreover, the source attribute, such as rupture direction, can influence the wavefield and hence the triggering intensity, although this effect is still debated (e.g., Gomberg et al., 2001; Jiang et al., 2010). At present, it remains difficult to determine a unified mechanism of dynamic triggering or to exactly describe the dominant mechanism in different regions (e.g., Gomberg et al., 1997; Cocco and Rice, 2002; Brodsky et al., 2003; Perfettini et al., 2003; Brodsky and Prejean, 2005; Johnson and Jia, 2005; Gonzalez-Huizar and Velasco, 2011; Shelly et al., 2011; 1. School of Earth and Space Sciences, Peking University, Beijing, China; 2. Earth System Science Programme, Faculty of Science, The Chinese University of Hong Kong, Shatin, Hong Kong, China *Corresponding author: zsy@pku.edu.cn © Seismological Society of America Volume XX • Number XX • – 2020 • www.srl-online.org Seismological Research Letters 1 Electronic Seismologist Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/doi/10.1785/0220200216/5169098/srl-2020216.1.pdf by Chinese Univ Hong Kong user on 22 October 2020 Hill, 2012; Delorey et al., 2015). To address these questions, many investigations of dynamic triggering utilizing rapidly expanding seismic arrays are needed in a variety of regions. Because of their inception (Hill et al., 1993), many methods have been developed to detect dynamic triggering by identifying triggered earthquakes and then estimating the significance of variations in the rate of seismicity, such as the β statistic (e.g., Matthews and Reasenberg, 1988; Reasenberg and Simpson, 1992). Traditional detection algorithms strongly depend on the accuracy and completeness of earthquake catalogs, which correspond to the network coverage (e.g., Hill et al., 1993). Local earthquakes during teleseismic waveforms are often manually picked, but this process is time-consuming and sometimes subjective (e.g., Peng et al., 2010). Automatic methods for identifying microearthquakes have been applied to observe dynamic triggering (e.g., Miyazawa and Mori, 2005; Delorey et al., 2015; Li et al., 2017; Miyazawa, 2019; Tang et al., 2020); among these methods, the matched filter technique is the most popular (Gibbons and Ringdal, 2006; Peng and Zhao, 2009; Yang et al., 2009), for which an even distribution of template events in the research area is critically important. However, sufficient templates are difficult to obtain in lowseismicity areas, and a large number of templates demand extensive computational times and resources. The convolutional neural network (CNN) has also been widely employed for phase picking and earthquake detection (Perol et al., 2018; Kong et al., 2019; Zhou et al., 2019). The basic idea of the CNN is to optimize the parameters of a neural network model based on a training dataset of earthquakes and then identify events in other waveforms using the trained model. The current development of CNNs is limited by the requirement for a substantial amount of training data, and normally, the network model needs to be retrained when changing the study region (Zhu et al., 2019). To overcome the dependence on the number of local earthquakes in seismicity rate evaluations, Yun et al. (2019) proposed the high-frequency power integral ratio (HiFi) algorithm using the difference in high-frequency energy before and during teleseismic waves (RE) to detect anomalous seismicity. For example, many studies reported the dynamic triggering of small earthquakes by the 4 April 2010 M 7.2 Baja California earthquake in the Geysers geothermal area, California (Aiken and Peng, 2014; Yun et al., 2019). If 5 hr before the P-wave arrival is selected as the background time window Tb, the manual detection of local events can become time-consuming. As an alternative, the HiFi method calculates the highfrequency energy directly. The high-frequency (25–35 Hz) energy in the time window Te aligned with teleseismic waves is richer than that in the time window Tb (Fig. 1a,b). Moreover, to reduce interference due to noise and address disturbances related to background seismicity, high-frequency energy changes among the same time windows on other days (RB) are also evaluated. For example, the high-frequency energy (a)
更新日期：2020-10-21
京公网安备 11010802027423号