Engineering topics

Shelly, David R.

Publications and source records attributed to Shelly, David R..

6 records

DOE OSTI2024

The Value of Hyperparameter Optimization in Phase-Picking Neural Networks

The effectiveness of neural networks for picking seismic phase arrival times has been demonstrated through several case studies, and seismic monitoring programs are starting to adopt the technology into their workflows. However, published models were designed and trained using rather arbitrary choices of hyperparameters, limiting their performance. In this study, we use phase picks from both routine and template-matching analyses from multiple regions (Ridgecrest, California; Kilauea, Hawaii; Yellowstone, Wyoming–Montana–Idaho) to test a hyperparameter optimization scheme for phase-picking neural networks and to evaluate their performance. We show that a published model, namely PhaseNet (Zhu and Beroza, 2019), can be simplified and improved with reasonable effort and there are preferred choices of hyperparameters that increase the performance. We also show that models optimized based on the arrival times reported in routine event catalogs consistently perform well when picking arrival times of smaller events, which is crucial for many tasks from microseismicity to explosion monitoring.

58 GEOSCIENCES↗

DOE OSTI2024

Making Phase-Picking Neural Networks More Consistent and Interpretable

Improving the interpretability of phase-picking neural networks remains an important task to facilitate their deployment to routine, real-time seismic monitoring. The popular phase-picking neural networks published in the literature lack interpretability because their output prediction scores do not necessarily correspond with the reliability of phase picks and can even be highly inconsistent depending on how we window the waveform data. Here, we show that systematically shifting the waveforms during training and using an antialiasing filter within the neural network architecture can substantially improve the consistency of the output prediction scores and can even make them scale with the signal-to-noise ratios of the waveforms. We demonstrate the improvements by applying these approaches to a commonly used phase-picking neural network architecture and using waveform data from the 2019 Ridgecrest earthquake sequence.

58 GEOSCIENCES↗

DOE OSTI2021

A Big Problem for Small Earthquakes: Benchmarking Routine Magnitudes and Conversion Relationships with Coda Envelope-Derived Mw in Southern Kansas and Northern Oklahoma

Earthquake magnitudes are widely relied upon measures of earthquake size. Although moment magnitude (M w ) has become the established standard for moderate and large earthquakes, difficulty in reliably measuring seismic moments for small (generally M w <4) earthquakes has meant that magnitudes for these events remain plagued by a patchwork of inconsistent measurement scales. Because of this, magnitudes of small earthquakes and statistics derived from them can be biased. Furthermore, because small earthquakes are much more numerous than large ones, many applications, such as seismic hazard modeling, depend critically on analysis of events characterized by magnitudes other than M w . Therefore, to assess this problem, we apply coda envelope analysis to reliably determine moment magnitudes for a case study of small earthquakes from northern Oklahoma and southern Kansas. Not surprisingly, we find significant differences among M L , m bLg , and M w for M ~2–4 earthquakes examined here. More troublingly, we find that relations designed to convert other magnitudes to M w , which are relied upon for important applications such as seismic hazard analysis, often increase rather than decrease this bias for our dataset. In our case study, we find that converted magnitudes can result in a systematic bias sometimes exceeding 0.5 magnitude units, a difference that typically corresponds to a factor of ~3 in seismicity rate. Moreover, we find a correspondingly large bias in Gutenberg–Richter b-values, controlled primarily by inaccurate magnitude scaling in the conversion relationships. Although this study focuses on a relatively small geographic area, we can expect that similar issues exist with varying severity in other regions. Therefore, magnitudes of small earthquakes and their associated statistics, including seismicity rates and b-values, should be treated with caution.

58 GEOSCIENCES↗

DOE OSTI2020

Geodetic Measurements of Slow-Slip Events Southeast of Parkfield, CA

Tremor and low-frequency earthquakes are presumed to be indicative of surrounding slow, aseismic slip that is often below geodetic detection thresholds. This study uses data from borehole seismometers and long-baseline laser strainmeters to observe both the seismic and geodetic signatures of episodic tremor and slip on the Parkfield region of the San Andreas Fault near Cholame, CA. We note the observed occurrence rates of both the tremors and co-located families of low-frequency earthquakes are not steady but instead exhibit quasiperiodic bursts of increased activity.We show that these periods of elevated seismic activity correlate with statistically significant stacked strain signals consisting of 44 slow-slip events. Modeled individual slow-slip events and their total summed moment, which are constrained by seismic signals and stacked strain, respectively, indicate that the individual moment magnitudes of these events range from $M_w$ 4.6–5.2.We find that the measured geodetic signal likely precedes the seismic signal by several hours, consistent with the aseismic slip preceding and driving the observed seismic tremor activity.We confirm that strike-slip faults, in addition to subduction zones, are capable of producing episodic tremor and slip.

58 GEOSCIENCES↗

DOE OSTI2018

Kinematics of the 2015 San Ramon, California earthquake swarm: Implications for fault zone structure and driving mechanisms

Earthquake swarms represent a sudden increase in seismicity that may indicate a heterogeneous fault-zone, the involvement of crustal fluids and/or slow fault slip. Swarms sometimes precede major earthquake ruptures. An earthquake swarm occurred in October 2015 near San Ramon, California in an extensional right step-over region between the northern Calaveras Fault and the Concord–Mt. Diablo fault zone, which has hosted ten major swarms since 1970. The 2015 San Ramon swarm is examined here from 11 October through 18 November using template matching analysis. The relocated seismicity catalog contains ~4000 events with magnitudes between -0.2 < M d < 3.6. The swarm illuminated three sub-parallel, southwest striking and northwest dipping fault segments of km-scale dimension and thickness of up to 200 m. The segments contain coexisting populations of different focal-mechanisms, suggesting a complex fault zone structure with several sets of en échelon fault orientations. The migration of events along the three planar structures indicates a complex fluid and faulting interaction processes. We searched for correlations between seismic activity and tidal stresses and found some suggestive features, but nothing that we can be confident is statistically significant.

58 GEOSCIENCES↗

DOE OSTI2016

Fortnightly modulation of San Andreas tremor and low-frequency earthquakes

Earth tides modulate tremor and low-frequency earthquakes (LFEs) on faults in the vicinity of the brittle-ductile (seismic-aseismic) transition. Our response to the tidal stress carries otherwise inaccessible information about fault strength and rheology. We analyze the LFE response to the fortnightly tide, which modulates the amplitude of the daily tidal stress over a 14-d cycle. LFE rate is highest during the waxing fortnightly tide, with LFEs most strongly promoted when the daily stress exceeds the previous peak stress by the widest margin. This pattern implies a threshold failure process, with slip initiated when stress exceeds the local fault strength. Furthermore, variations in sensitivity to the fortnightly modulation may reflect the degree of stress concentration on LFE-producing brittle asperities embedded within an otherwise aseismic fault.

58 GEOSCIENCES↗