Zachary E. Ross

Abundant off-fault seismicity and orthogonal structures in the San Jacinto fault zone

Distinct on-fault and off-fault seismicity in the trifurcation area of the San Jacinto fault zone. The trifurcation area of the San Jacinto fault zone has produced more than 10% of all earthquakes in southern California since 2000, including the June 2016 Mw (moment magnitude) 5.2 Borrego Springs earthquake. In this area, the fault splits into three subparallel strands and is associated with broad VP/VS anomalies. We synthesize spatiotemporal properties of historical background seismicity and aftershocks of the June 2016 event. A template matching technique is used to detect and locate more than 23,000 aftershocks, which illuminate highly complex active fault structures in conjunction with a high-resolution regional catalog. The hypocenters form dipping seismicity lineations both along strike and nearly orthogonal to the main fault, and are composed of interlaced strike-slip and normal faults. The primary faults change dip with depth and become listric by transitioning to a dip of ~70° near a depth of 10 km. The Mw 5.2 Borrego Springs earthquake and past events with M > 4.0 occurred on the main faults, whereas most of the low-magnitude events are located in a damage zone (several kilometers wide) at seismogenic depths. The lack of significant low-magnitude seismicity on the main fault traces suggests that they do not creep. The very high rate of aftershocks likely reflects the large geometrical fault complexity and perhaps a relatively high stress due to a significant length of time elapsed since the last major event. The results provide important insights into the physics of faulting near the brittle-ductile transition.

Aftershocks driven by afterslip and fluid pressure sweeping through a fault-fracture mesh

A variety of physical mechanisms are thought to be responsible for the triggering and spatiotemporal evolution of aftershocks. Here we analyze a vigorous aftershock sequence and postseismic geodetic strain that occurred in the Yuha Desert following the 2010 M_w 7.2 El Mayor-Cucapah earthquake. About 155,000 detected aftershocks occurred in a network of orthogonal faults and exhibit features of two distinct mechanisms for aftershock triggering. The earliest aftershocks were likely driven by afterslip that spread away from the main shock with the logarithm of time. A later pulse of aftershocks swept again across the Yuha Desert with square root time dependence and swarm-like behavior; together with local geological evidence for hydrothermalism, these features suggest that the events were driven by fluid diffusion. The observations illustrate how multiple driving mechanisms and the underlying fault structure jointly control the evolution of an aftershock sequence.

Anomalously large complete stress drop during the 2016 Mw 5.2 Borrego Springs earthquake inferred by waveform modeling and near‐source aftershock deficit

The 2016 M_w 5.2 Borrego Springs earthquake occurred in the trifurcation area of the San Jacinto Fault Zone and generated more than 23,000 aftershocks. We analyze source properties of this earthquake along with 12,487 precisely located aftershock hypocenters to obtain an unusually detailed view of the rupture process and energy budget for this moderate earthquake. Source time functions are obtained using an empirical Greens function approach and are inverted for a slip distribution on the fault plane. The rupture propagated unilaterally to the northwest over a distance of 1.8 km, resulting in clear directivity signals. Two asperities are identified and the maximum slip is 2.54 m, resulting in a static stress drop of 78.2 MPa. Over 97% of the aftershocks occur more than 1 rupture length from the slip area. We conclude that the Borrego Springs earthquake had a complete stress drop and estimate the seismic efficiency to be 15–26%.

Evolution of seismicity near the southernmost terminus of the San Andreas Fault: Implications of recent earthquake clusters for earthquake risk in southern California

Three earthquake clusters that occurred in the direct vicinity of the southern terminus of the San Andreas Fault (SAF) in 2001, 2009, and 2016 raised significant concern regarding possible triggering of a major earthquake on the southern SAF, which has not ruptured in more than 320 years. These clusters of small and moderate earthquakes with M ≤ 4.8 added to an increase in seismicity rate in the northern Brawley seismic zone that began after the 1979 M_w 6.5 Imperial Valley earthquake, in contrast to the quiet from 1932 to 1979. The clusters so far triggered neither small nor large events on the SAF. The mostly negative Coulomb stress changes they imparted on the SAF may have reduced the likelihood that the events would initiate rupture on the SAF, although large magnitude earthquake triggering is poorly understood. The relatively rapid spatial and temporal migration rates within the clusters imply aseismic creep as a possible driver rather than fluid migration.

The 2016 Kumamoto Mw = 7.0 Earthquake: A Significant Event in a Fault–Volcano System

The 2016 Kumamoto earthquake sequence occurred on the Futagawa–Hinagu fault zone near the Aso volcano on Kyushu island. The sequence was initiated with two major (M_w ≥ 6.0) foreshocks, and the mainshock (M_w = 7.0) occurred 25 h after the second major foreshock. We combine GPS, strong motion, synthetic aperture radar images, and surface offset data in a joint inversion to resolve the kinematic rupture process of the mainshock and coseismic displacement of the foreshocks. The joint inversion results reveal a unilateral rupture process for the mainshock involving sequential rupture of four major asperities. The slip area of the foreshocks and mainshock and the aftershock loci form a detailed complementary pattern. The mainshock rupture terminates near the rim of the caldera, leaving a ~10 km long gap of aftershocks. This area is characterized by high temperature and low shear wave velocity, density, and resistivity, which may be related to the partially melted geothermal condition. Ductile material property near the volcano may act as a “material barrier” to the dynamic rupture. Topographic weight of the caldera increases compressional normal stress on the fault plane, which may behave as a “stress barrier.” Long-term seismic hazard and deformation behaviors related to these two types of barriers are discussed in terms of the associated frictional mechanism. Significant postseismic creeps observed near the volcano area indicates a velocity strengthening frictional behavior near the rupture termination, which confirms that the “material barrier” mechanism is likely the dominant rupture termination mechanism.

P-wave arrival picking and first-motion polarity determination with deep learning

Determining earthquake hypocenters and focal mechanisms requires precisely measured P-wave arrival times and first-motion polarities. Automated algorithms for estimating these quantities have been less accurate than estimates by human experts, which is problematic for processing large data volumes. Here, we train convolutional neural networks to measure both quantities, which learn directly from seismograms without the need for feature extraction. The networks are trained on 18.2 million manually picked seismograms for the southern California region. Through cross-validation on 1.2 million independent seismograms, the differences between the automated and manual picks have a standard deviation of 0.023 seconds. The polarities determined by the classifier have a precision of 95% when compared with analyst-determined polarities. We show that the classifier picks more polarities overall than the analysts, without sacrificing quality, resulting in almost double the number of focal mechanisms. The remarkable precision of the trained networks indicates that they can perform as well, or better, than expert seismologists.

Diverse Volumetric Faulting Patterns in the San Jacinto Fault Zone

We examine locations, magnitudes, and faulting types of post‐2000 earthquakes in the trifurcation area of San Jacinto fault zone to clarify basic aspects of failure processes in the area. Most M ≥ 3.5 events have strike‐slip mechanisms, occur within 1 km of the main faults (Clark, Buck Ridge, and Coyote Creek), and have hypocenter depths of 10–13 km. In contrast, many smaller events have normal source mechanisms and hypocenters in intrafault areas deeper than 13 km. Additional small events with hypocenter depth <13 km occur in off‐fault regions and have complex geometries including lineations normal to the main faults. Five moderate earthquakes with M 4.7–5.4 have high aftershock rates (~150 M ≥ 1.5 events within 1 day from the mainshock). To obtain more details on aftershock sequences of these earthquakes, we detect and locate additional events with the matched filter method. There are almost no aftershocks within 1 km from the mainshocks, consistent with large mainshock stress drops and low residual stress. The five aftershock sequences have almost no spatial overlap. While the mainshocks are on the main faults, most aftershocks are located in intrafault and off‐fault regions. Their locations and spatial distribution reflect the mainshock rupture directions, and many also follow structures normal to the main faults. The significant diversity of observed features highlights the essential volumetric character of failure patterns in the area. The increasing rate of moderate events, productive aftershock sequences, and large inferred stress drops may reflect processes near the end of a large earthquake cycle.

Dissipative Intraplate Faulting During the 2016 Mw 6.2 Tottori, Japan Earthquake

The 2016 M_w 6.2 Tottori earthquake occurred on 21 October 2016 and produced thousands of aftershocks. Here we analyze high-resolution-relocated seismicity together with source properties of the mainshock to better understand the rupture process and energy budget. We use a matched-filter algorithm to detect and precisely locate >10,000 previously unidentified aftershocks, which delineate a network of sharp subparallel lineations exhibiting significant branching and segmentation. Seismicity below 8 km depth forms highly localized fault structures subparallel to the mainshock strike. Shallow seismicity near the main rupture plane forms more diffuse clusters and lineations that often are at a high angle (in map view) to the mainshock strike. An empirical Greens function technique is used to derive apparent source time functions for the mainshock, which show a large amplitude pulse 2–4 s long. We invert the apparent source time functions for a slip distribution and observe a ~16 km^2 patch with average slip ~3.2 m. 93% of the seismic moment is below 8 km depth, which is approximately the depth below which the seismicity becomes very localized. These observations suggest that the mainshock rupture area was entirely within the lower half of the seismogenic zone. The radiated seismic energy is estimated to be 5.7 × 10^(13) J, while the static stress drop is estimated to be 18–27 MPa. These values yield a radiation efficiency of 5–7%, which indicates that the Tottori mainshock was extremely dissipative. We conclude that this inefficiency in energy radiation is likely a product of the immature intraplate environment and the underlying geometric complexity.

Generalized Seismic Phase Detection with Deep LearningShort Note

To optimally monitor earthquake‐generating processes, seismologists have sought to lower detection sensitivities ever since instrumental seismic networks were started about a century ago. Recently, it has become possible to search continuous waveform archives for replicas of previously recorded events (i.e., template matching), which has led to at least an order of magnitude increase in the number of detected earthquakes and greatly sharpened our view of geological structures. Earthquake catalogs produced in this fashion, however, are heavily biased in that they are completely blind to events for which no templates are available, such as in previously quiet regions or for very large‐magnitude events. Here, we show that with deep learning, we can overcome such biases without sacrificing detection sensitivity. We trained a convolutional neural network (ConvNet) on the vast hand‐labeled data archives of the Southern California Seismic Network to detect seismic body‐wave phases. We show that the ConvNet is extremely sensitive and robust in detecting phases even when masked by high background noise and when the ConvNet is applied to new data that are not represented in the training set (in particular, very large‐magnitude events). This generalized phase detection framework will significantly improve earthquake monitoring and catalogs, which form the underlying basis for a wide range of basic and applied seismological research.