Comparing Traditional and Deep Learning Signal Features for Event Detection in the Utah Region

Abstract

Advances in deep learning in the past decade have recently been applied to various algorithms in the seismic event monitoring data processing pipeline. In this article, we apply PhaseNet (Zhu and Beroza, 2018)—a deep learning model for seismic signal detection, to backprojection event detection in the Utah region using the Waveform Correlation Event Detection System (WCEDS). We compare PhaseNet-WCEDS with the original short-term average/long-term average (STA/LTA) version of WCEDS from Arrowsmith et al. (2016, 2018). Using the Unconstrained Utah Event Bulletin (Linville et al., 2019) as the “ground truth,” we present the precision and recall for each method for a variety of tuning parameters, with PhaseNet-WCEDS recall being approximately 86%, whereas STA/LTA-WCEDS recall was 66% across a range of detection thresholds. Furthermore, we show that the PhaseNet-WCEDS recall advantage holds across various subregions and event source types in the Utah region. We also introduce a local to near-regional event criteria test that reduces false positives by 55% whereas only reducing true positives by 7% for PhaseNet-WCEDS (60% and 17%, respectively, for STA/LTA-WCEDS). Using the event commonality score (ECS, Draelos et al., 2015), we explore the ECS-based event categories for PhaseNet-WCEDS and STA/LTA-WCEDS for two important subsets of our Utah data set—the Circleville aftershock sequence and events in the central mining region.

Customization of a deep neural network using local data for seismic phase picking

“…Figure 7 shows the prediction results related to the position shifts of P arrivals in the input data. As reported by Heck et al (2022), the performance of PhaseNet varied with position changes of P arrivals in the input window. However, KFpicker exhibited excellent performance and was consistent for P arrivals, regardless of variations in P arrival positions in the input window.…”

“…PhaseNet (Zhu and Beroza, 2019) was used to generate candidate picks, which were then used to identify ground-truth picks. To address the inconsistency of PhaseNet caused by variations in the input window position (Heck et al, 2022) and maximize the number of candidates, we shifted the input window by a 1 s stride and selected the picks exceeding a 0.1 threshold based on the point-by-point maximum of overlapping prediction outputs. Through careful visual inspection, a total of 1,354 ground truth picks were identified, consisting of 489 P picks and 865 S picks, primarily due to the low SNR of the P phases.…”

“…Szegedy et al (2013) and Goodfellow et al (2014) demonstrated that DL models are vulnerable to adversarial attacks due to imperceptible perturbations added to the original input, which cause the model to commit mistakes. This phenomenon is also observed in DL pickers, where a position change or a slight perturbation in the input waveform can cause differences in the results (Heck et al, 2022;Park et al, 2022). The performance of DL pickers is critical, particularly for near-real time monitoring of seismicity, as continuous waveforms are processed with a controlled level of false-positive rate (Munchmeyer et al, 2022).…”

“…DL pickers can be generalized to other regions or source types that were not used for their training (Mousavi et al, 2020;Liao et al, 2021;Tan et al, 2021;Walter et al, 2021;García et al, 2022;Harsuko and Alkhalifah, 2022;Heck et al, 2022;Liao et al, 2022;Münchmeyer et al, 2022;Wang et al, 2023). However, the performance of the picker is sometimes degraded when applied to those regions or source types (Jiang et al, 2021;Heck et al, 2022;Han et al, 2023). This issue can be resolved by retraining the neural network (Chai et al, 2020;Johnson et al, 2020;Lapins et al, 2021).…”

Recent advances in earthquake seismology using machine learning

“…Incidentally, in the initial process of earthquake catalog development, the analyst typically identifies a window involving a seismic waveform and then measures arrival time in a cutout waveform. However, this event detection phase can be skipped, resulting in the arrival time measurements being performed directly on the entire continuous waveforms (Park et al 2020;Heck et al 2022;Retailleau et al 2023). The fraction of false detections of arrival time measurements and the efficiency of the subsequent phase association calculations likely determine which of the two approaches is effective.…”