A Detailed Earthquake Catalog for Banda Arc–Australian Plate Collision Zone Using Machine-Learning Phase Picker and an Automated Workflow

The tectonic setting of Timor–Leste and Eastern Indonesia comprises of a complex transition from oceanic lithosphere subduction to arc-continental collision. To better understand the deformation and convergent-zone structure of the region, we derive a new catalog of earthquake hypocenters and magnitudes from a temporary deployment of five years of continuous seismic data using an automated processing procedure. This includes a machine-learning phase picker, EQTransformer, and a sequential earthquake association and location workflow. We detect and locate ∼19,000 events during 2014–2018, which demonstrates that it is possible to characterize earthquake sequences from raw seismic data using a well-trained machine-learning picker for a complex convergent plate setting. This study provides the most complete catalog available for the region for the duration of the temporary deployment, which includes a complex pattern of crustal events across the collision zone and into the back-arc, as well as abundant deep slab seismicity.

Precise Locating of the Great 1897 Shillong Plateau Earthquake Using Teleseismic and Regional Seismic Phase Data

Pinched between the Eastern Himalaya and the Indo-Burman ranges, the Shillong Plateau represents a zone of distributed deformation with numerous visible and buried active faults. In 1897, a great (magnitude 8+) earthquake occurred in the area, and although a subsurface rupture plane has been proposed geodetically, its epicenter remained uncertain. We gathered original arrival time data of seismic waves from this early-instrumental era and combined them with modern, 3D velocity models to constrain the origin time and epicenter of this event, including uncertainties. Our results show that the earthquake has taken place in the northwest part of the plateau, at the junction of the short, surface-rupturing Chedrang fault and the buried Oldham fault (26.0°N, 90.7°E). This latter fault has been proposed earlier based on geodetic data and is long enough to host a great earthquake. Rupture has most likely propagated eastward. Stress change from the 1897 earthquake may have ultimately triggered the 1930 M 7.1 Dhubri earthquake, along a fault connecting the Shillong Plateau with the Himalaya.

Bolide Energetics and Infrasound Propagation: Exploring the 18 December 2018 Bering Sea Event to Identify Limitations of Empirical and Numerical Models

Infrasound observations are an important tool in assessing the energetics of bolides and can help quantify the flux of meteoroids through Earth’s atmosphere. Bolides are also important atmospheric sources for assessing long-range infrasound propagation models and can be used as benchmark events for validating the International Monitoring System (IMS) infrasound network, which is designed to detect nuclear tests in the atmosphere. This article exploits unique infrasound observations from a large bolide recorded on IMS infrasound arrays and high-density infrasound deployments in the United States to assess limitations in infrasound source scaling relationships. The observations provide an unprecedented sampling of infrasound propagation along a transect at an azimuth of 60° from the source to a distance of ∼8000 km. Widely used empirical laws for assessing bolide energetics and state-of-the-art numerical models for simulating infrasound propagation are assessed to quantify important discrepancies with the observations. In particular, empirical laws for equivalent yield, which are based on signal period and are assumed to be relatively unaffected by propagation effects, can be heavily contaminated by site noise. In addition, by modeling infrasound propagation over a range of ∼8000 km, we show that state-of-the-art models do not reproduce the observed amplitude decay over this long range (which decays by a rate of at least 2 higher than can be modeled).

Seismic Wave Propagation and Inversion with Neural Operators

Seismic wave propagation forms the basis for most aspects of seismological research, yet solving the wave equation is a major computational burden that inhibits the progress of research. This is exacerbated by the fact that new simulations must be performed whenever the velocity structure or source location is perturbed. Here, we explore a prototype framework for learning general solutions using a recently developed machine learning paradigm called neural operator. A trained neural operator can compute a solution in negligible time for any velocity structure or source location. We develop a scheme to train neural operators on an ensemble of simulations performed with random velocity models and source locations. As neural operators are grid free, it is possible to evaluate solutions on higher resolution velocity models than trained on, providing additional computational efficiency. We illustrate the method with the 2D acoustic wave equation and demonstrate the method’s applicability to seismic tomography, using reverse-mode automatic differentiation to compute gradients of the wavefield with respect to the velocity structure. The developed procedure is nearly an order of magnitude faster than using conventional numerical methods for full waveform inversion.

Seismic Constraint on Heterogeneous Deformation and Stress State in the Forearc of the Hikurangi Subduction Zone, New Zealand

The Hikurangi subduction zone (HSZ) is the collisional boundary between the Pacific and Australian tectonic plates along the eastern coast of the North Island of New Zealand. The region is believed to be capable of hosting large megathrust earthquakes and associated tsunamis. Recent studies observe a range of slip behavior along the plate interface, with a sharp contrast between locked and creeping parts of the megathrust along the margin. This work uses teleseismic scattering data (receiver functions [RFs]) recorded at 53 long-running seismograph stations on the North Island of New Zealand to constrain the structure and mechanical properties of the forearc in the HSZ. We observe directional variations in RF phases at P–S converted delay times (i.e., depths) associated with the overlying forearc crust and note a general correlation with spatial variations in plate coupling as well as other geophysical properties. Our results suggest differences in the nature of crustal deformation (and stress state) along the Hikurangi margin, with evidence of clockwise rotation and/or extension in the northern HSZ, where the overriding forearc crust is uncoupled from the subducting Pacific slab.

Paleotsunamis on the Southern Hikurangi Subduction Zone, New Zealand, Show Regular Recurrence of Large Subduction Earthquakes

Geological records of subduction earthquakes, essential for seismic and tsunami hazard assessment, are difficult to obtain at transitional plate boundaries, because upper-plate fault earthquake deformation can mask the subduction zone signal. Here, we examine unusual shell layers within a paleolagoon at Lake Grassmere, at the transition zone between the Hikurangi subduction zone and the Marlborough fault system. Based on biostratigraphic and sedimentological analyses, we interpret the shell layers as tsunami deposits. These are dated at 2145–1837 and 1505–1283 yr B.P., and the most likely source of these tsunamis was ruptures of the southern Hikurangi subduction interface. Identification of these two large earthquakes brings the total record of southern Hikurangi subduction earthquakes to four in the past 2000 yr. For the first time, it is possible to obtain a geologically constrained recurrence interval for the southern Hikurangi subduction zone. We calculate a recurrence interval of 500 yr (335–655 yr, 95% confidence interval) and a coefficient of variation of 0.27 (0.0–0.47, 95% confidence interval). The probability of a large subduction earthquake on the southern Hikurangi subduction zone is 26% within the next 50 yr. We find no consistent temporal relationship between subduction earthquakes and large earthquakes on upper-plate faults.

Stability of the Fault Systems That Host-Induced Earthquakes in the Delaware Basin of West Texas and Southeast New Mexico

The Delaware basin of west Texas and southeast New Mexico has experienced elevated earthquake rates linked spatiotemporally to unconventional petroleum operations. Limited knowledge of subsurface faults, the in situ geomechanical state, and the exact way in which petroleum operations have affected pore pressure (Pp) and stress state at depth makes causative assessment difficult, and the actions required for mitigation uncertain. To advance both goals, we integrate comprehensive regional fault interpretations, deterministic fault-slip potential (DFSP), and multiple earthquake catalogs to assess specifically how faults of two systems—deeper basement-rooted (BR) and shallow normal (SN)—can be made to slip as Pp is elevated. In their natural state, the overall population faults in both the systems have relatively stable DFSP, which explains the low earthquake rate prior to human inducement. BR faults with naturally unstable DFSP and associated earthquake sequences are few but include the Culberson–Mentone earthquake zone, which is near areas of wastewater injection into strata above basement. As a system, the SN faults in the southcentral Delaware basin are uniformly susceptible to slip with small increases in Pp. Many earthquakes sequences have occurred along these shallow faults in association with elevated Pp from shallow wastewater injection and hydraulic fracturing. Our new maps and methods can be used to better plan and regulate petroleum operations to avoid fault rupture.

Joint Regional Waveform, First-Motion Polarity, and Surface Displacement Moment Tensor Inversion of the 3 September 2017 North Korean Nuclear Test

The 3 September 2017 Mw 5.2 North Korean underground nuclear test (DPRK2017) is the largest man-made explosion with surface displacements observed by Synthetic Aperture Radar (SAR) and showed as much as 3.5 m of horizontal permanent deformation. Although regional distance waveform-based seismic moment tensor (MT) inversion methods successfully identify this event as an explosion, the inverted solutions do not fit the SAR displacement field well. To better constrain the source, we developed an MT source-type inversion method that incorporates surface ground deformation (accounting for free-surface topography), regional seismic waveforms, and first-motion polarities. We applied the source-type inversion over a grid of possible source locations to find the best-fitting location, depth, and point-source MT for the event. Our best-fitting MT solution achieves ∼70% horizontal geodetic fit, ∼80% waveform fit, and 100% fit in the first-motion polarities. The joint inversion narrows the range of acceptable source types improving discrimination, and reduces the uncertainty in scalar moment and estimated yield. The method is transportable and can be applied to other types of events that may have measurable geodetic signals such as underground mine collapses and volcanic events.

Laboratory Evidence of Transient Pressure Surge in a Fluid-Filled Fracture as a Potential Driver of Remote Dynamic Earthquake Triggering

Seismic waves carrying tiny perturbing stresses can trigger earthquakes in geothermal and volcanic regions. The underlying cause of this dynamic triggering is still not well understood. One leading hypothesis is that a sudden increase in the fluid-pore pressure in the fault zone is involved, but the exact physical mechanism is unclear. Here, we report experimental evidence in which a fluid-filled fracture was shown to be able to amplify the pressure of an incoming seismic wave. We built miniature pressure sensors and directly placed them inside a thin fluid-filled fracture to measure the fluid pressure during wave propagation. By varying the fracture aperture from 0.2 to 9.2 mm and sweeping the frequency from 12 to 70 Hz, we observed in the lab that the fluid pressure in the fracture could be amplified up to 25.2 times compared with the incident-wave amplitude. Because an increase of the fluid pressure in a fault can reduce the effective normal stress to allow the fault to slide, our observed transient pressure surge phenomenon may provide the mechanism for earthquake dynamic triggering.

Machine-Learning-Based High-Resolution Earthquake Catalog Reveals How Complex Fault Structures Were Activated during the 2016–2017 Central Italy Sequence

The 2016–2017 central Italy seismic sequence occurred on an 80 km long normal-fault system. The sequence initiated with the Mw 6.0 Amatrice event on 24 August 2016, followed by the Mw 5.9 Visso event on 26 October and the Mw 6.5 Norcia event on 30 October. We analyze continuous data from a dense network of 139 seismic stations to build a high-precision catalog of ∼900,000 earthquakes spanning a 1 yr period, based on arrival times derived using a deep-neural-network-based picker. Our catalog contains an order of magnitude more events than the catalog routinely produced by the local earthquake monitoring agency. Aftershock activity reveals the geometry of complex fault structures activated during the earthquake sequence and provides additional insights into the potential factors controlling the development of the largest events. Activated fault structures in the northern and southern regions appear complementary to faults activated during the 1997 Colfiorito and 2009 L’Aquila sequences, suggesting that earthquake triggering primarily occurs on critically stressed faults. Delineated major fault zones are relatively thick compared to estimated earthquake location uncertainties, and a large number of kilometer-long faults and diffuse seismicity were activated during the sequence. These properties might be related to fault age, roughness, and the complexity of inherited structures. The rich details resolvable in this catalog will facilitate continued investigation of this energetic and well-recorded earthquake sequence.