Earthquake waveforms recorded at the surface are strongly affected by near-surface processes, including nonlinear wave propagation, scattering, attenuation, topography effects and local noise. These complexities hinder accurate ground motion prediction at the receiver site and obscure weak earthquake signals in surface recordings, especially at high frequencies. Traditional approaches analyze borehole-surface discrepancies to model shallow structures and apply numerical solvers for near-vertical wave propagation. Yet, these attempts are constrained by simplified physical assumptions, high computational costs, and inherent model uncertainties. To address these challenges, we develop U-shaped Neural Operators for Vertical Continuation (UNO-VC), a fully data-driven machine learning framework designed to map seismic waveforms directly through sedimentary layers. UNO-VC facilitates a two-way vertical continuation strategy: (1) upward continuation to predict surface waveforms given bedrock recordings, and (2) downward continuation to back-propagate surface recordings to a bedrock depth, effectively denoising the waveforms for cleaner signals. We apply and test this method on the Garner Valley Downhole Array dataset, which comprises over two decades of high-quality Southern California earthquake recordings. Quantitative analysis demonstrates that UNO-VC outperforms multiple benchmarks. Upward continuation accurately reproduces observed ground motion intensities and waveform shapes; downward continuation yields cleaner earthquake waveforms than raw surface recordings and more detections from an automated phase picker. This framework benefits ground motion prediction and calibration efforts, and mitigates site-effect uncertainties for refined seismic analysis.

Abstract Moment magnitude Mw is derived from seismic moment and was defined with the intent to be a consistent measure of earthquake size. Here, we demonstrate that reported Mw values depend systematically on the rigidity structure assumed in the moment tensor inversion, which varies between regions and users. We perform regional moment tensor inversions for 70 target earthquakes in the Walker Lane and southern California regions of the western United States, comparing magnitude results across a suite of commonly used 1D velocity models. Inversions performed with slower, lower rigidity velocity models produce lower Mw estimates than inversions performed with faster, higher rigidity velocity models: the epistemic uncertainty from the choice of velocity model averages ±0.08 for the analyzed earthquakes, with larger uncertainties for shallower events. Using magnitudes based on seismic potency (Mp) instead of Mw can reduce this uncertainty by a factor of 2–3. We introduce a potency-based magnitude Mpv that, by correcting for both the rigidity at source depth and characteristic rigidity of the velocity model, produces a robust magnitude scale that is independent of the velocity model. This formulation can be used to generate uniform and consistent regional and global earthquake catalogs necessary for a wide range of applications in earthquake science and engineering.

The Mendocino Triple Junction (MTJ), where the Gorda, North American and Pacific plates meet, is one of the most seismically active regions in California. The tectonic movements along the Mendocino transform fault zone (MTFZ), Gorda slab (GS) and northern San Andreas Fault systems (NSAF) lead to high background seismicity rates but relatively low aftershock productivity. To improve the understanding of earthquake processes in the area, we analyse relations between background seismicity, aftershock productivity and stress parameters. We apply the nearest-neighbour approach to investigate the spatial distributions and properties of background and clustered seismicity, and invert focal mechanisms of events in Voronoi cells for features of the deviatoric stress field. The results indicate that the intensity of background seismicity and aftershock productivity decrease with distance from the MTJ, defined here for simplicity as the hypocentre of the 1992 Mw 7.2 main shock. We also find that the stress regime is the most compressive in the area directly surrounding the MTJ. In the MTFZ and GS, the compressive stress decreases with increasing distance from the MTJ, correlating with the reduced aftershock productivity and background seismicity. In the NSAF, the observed relations between the stress, aftershock productivity and background seismicity are not clear, possibly due to crustal extension related to the slab window and elevated heat flow. Compared to the MTFZ and GS, the NSAF has a higher foreshock proportion, lower aftershock proportion and small-to-medium main shock magnitudes, indicating more swarm-like clusters in this region. The inverted stress regimes in the MTFZ and NSAF are dominated by strike-slip faulting. The GS exhibits mostly strike-slip and normal mechanisms despite the subduction environment, which may reflect slab bending and reactivation of pre-existing normal faults.

Large earthquakes involve processes that occur primarily at seismogenic depths of 5-10 km or more and that are thus difficult to observe in detail with traditional seismic networks. The FaultScan project establishes the first long-term, dense nodal array experiment by acquiring continuous data at 300 stations along the San Jacinto Fault (SJF, South California) at the Piñon Flat Observatory for 2.5 years (April 2022-Nov. 2024).We report on the experiment logistics, data quality and describe the characteristics of the High-Frequency (HF, >1 Hz) back-ground noise with a focus on incoherent wind generated noise, train traffic tremors and far distant coherent urban noise. We illustrate how these HF coherent noise sources recorded mostly as body-waves can be used to track velocity changes across the SJF by applying seismic interferometry between the array and nearby permanent seismic stations.We further describe how much slant-stacking increases the level of signal to noise ratio for the detection of small earthquakes along the San Jacinto Fault and show how newly detected events improve our description of foreshock/aftershock patterns. Our search for tectonic tremors along the San Jacinto Fault turns up empty but we observe tremor-like signals, mostly T-phases coming exclusively from the Tonga subduction and one intriguing T-phase like sequence originating from the October 2023, offshore Japan volcanic crisis.

Quantifying earthquake size across scales requires linking physical source parameters to empirical magnitude measures, especially for microearthquakes where attenuation, scattering, and instrumental limitations complicate source characterization. We combine regional results with near-source observations from the Tony Creek dual Microseismic Experiment, dense nodal data from Ridgecrest, and regional broadband data from San Jacinto to extend potency–magnitude scaling to local-magnitude ML ≈ −1.8. Bayesian spectral inference yields robust potency estimates across seven potency magnitude MP units and nine ML units. The combined datasets follow a continuous quadratic relation MP = 0.040M2 L + 0.556ML + 1.133. Synthetic modeling shows that source scaling, attenuation, and the finite Wood–Anderson passband can produce similar nonlinearity, although rupturescaling differences may also contribute. Despite sensitivity to observational and regional factors, the independent datasets are broadly consistent with the established quadratic relation, supporting it as a general empirical scaling over the investigated size range. Plain Language Summary Local magnitude is a common measure of earthquake size based on recorded ground motion, whereas seismic potency measures permanent deformation at the source. Estimating potency for very small earthquakes is difficult because the relevant signals are weak and affected by attenuation and instrumental bandwidth. We combine close-range recordings of induced microearthquakes in western Canada with independent tectonic earthquake observations from Ridgecrest and the San Jacinto Fault in California, together with a regional catalog of larger earthquakes. The combined observations reveal a smooth, curved relation between local magnitude and potency across a broad range of earthquake sizes. Tests with synthetic waveforms show that much of this curvature can arise from wave attenuation and the limited frequency range used to measure local magnitude, even when the earthquake sources scale similarly. The derived relation provides a practical empirical conversion between commonly reported magnitude and physical source size, while the exact coefficients may vary among regions and observing systems.

ABSTRACT The Statewide California Earthquake Center has developed the CyberShake platform to perform 3D simulation-based probabilistic seismic hazard analysis. Using earthquake event rates, kinematic ruptures, and a 3D seismic velocity model, all constrained by observations, CyberShake utilizes wave propagation simulations and a reciprocity-based approach to generate hundreds of thousands of seismograms per site of interest. These simulation results are used to produce hazard data products including uniform hazard spectra, significant durations, site-specific hazard curves, and regional hazard maps. Previous CyberShake modeling efforts have computed seismic hazard models in southern California and Central California. In this article, we present CyberShake Study 24.8, which implements the CyberShake workflow for 315 sites in the greater San Francisco Bay area, extending physics-based hazard estimates to the entire San Andreas fault system. We construct the velocity model by tiling together three 3D and 1D models and applying a near-surface VS30-based merged taper to the top 700 m. We used the AWP—ODC wave propagation code to perform the ground-motion simulations, and the Graves–Pitarka kinematic rupture generator to create slip time histories for individual events. CyberShake Study 24.8 produces overall slightly reduced hazard compared to the NGA-West2 ground-motion models (GMMs). Specifically, CyberShake produces lower hazard immediately surrounding the San Francisco Bay likely due to use of a minimum VS of 400 m/s, lower hazard in near-fault regions, and higher hazard at long periods near sedimentary basins in the velocity model. We find larger directivity effects for CyberShake in northern California than in southern California, likely due to increased uniformity in fault characteristics and orientation in northern California. Validation results using the 1989 M 6.9 Loma Prieta event show that CyberShake reproduces observed ground motions with reduced bias compared with GMMs and 1D simulations.

Using a novel high-performance computing implementation of a nonlinear continuum damage breakage model, we explore interactions between 3D co-seismic off-fault damage, seismic radiation, and rupture dynamics. Our simulations demonstrate that off-fault damage enhances high-frequency wave radiation above 1 Hz, reduces rupture speed and alters the total kinetic energy. We identify distinct damage regimes separated by solid-granular transition, with smooth distributions under low damage conditions transitioning to localized, mesh-independent shear bands upon reaching brittle failure. The shear band orientations depend systematically on the background stress and agree with analytical predictions. The brittle damage inhibits transitions to supershear rupture propagation and the rupture front strain field results in locally reduced damage accumulation during supershear transition. The dynamically generated damage yields uniform and isotropic ratios of fault-normal to fault-parallel high-frequency ground motions. Co-seismic damage zones exhibit depth-dependent width variations, becoming broader near the Earth's surface consistent with field observations, even under uniform stress conditions. We discover a new delayed dynamic triggering mechanism in multi-fault systems, driven by reductions in elastic moduli and the ensuing stress heterogeneity in 3D tensile fault step-overs. This mechanism affects the static and dynamic stress fields and includes the formation of high shear-traction fronts around localized damage zones. The brittle damage facilitates rupture cascading across faults, linking delay times directly to damage rheology and fault zone evolution. Our results help explain enhanced high-frequency seismic radiation and delayed rupture triggering, improving our understanding of earthquake processes, seismic radiation and fault system interactions.

Following the 2019 Mw 7.1 Ridgecrest, California, earthquake, 15 dense 1D and 2D arrays (461 sites) were deployed around the main ruptures, including four 1D arrays across the surface ruptures of the mainshock. The dense arrays captured numerous aftershocks and provided an unprecedented dataset for studying near-fault ground motions. We combine the near-fault ground motions measured from the dense arrays with a regional groundmotion dataset and develop a near-fault partially nonergodic ground-motion model (GMM). The partially nonergodic GMM results are used to investigate several aspects of the near-fault ground motions that are related to source and fault zone properties. We find that shallow events tend to produce weaker short-period energy due to smaller stress drop or stronger attenuation but stronger long-period energy from surface-wave excitation. Significant ground-motion variations across the fault zone are observed, which include amplifications within the damage zones due to both reduced seismic velocities and trapped waves at certain periods. We also observe double-couple radiation pattern effects only in long-period ground motions, agreeing with a period-dependent earthquake rupture process. Between 3 and 15 days following the mainshock, the ground motions of aftershocks decrease with time, likely reflecting decreasing stress drops. No temporal evolution of site response and wave propagation effects within the fault zone are observed between 10 and 35 days following the mainshock. Our study demonstrates the capability of dense array data to improve our understanding of earthquake processes, fault zone properties, and near-fault seismic hazard.

We perform comparative analyses of seismic signals and fault-zone structures at the Banning Fault (BF) and Mission Creek Fault (MCF) strands of the Southern San Andreas Fault (SoSAF) in the Coachella Valley, California. The research utilizes data recorded by densely spaced linear and two-dimensional (2D) seismic arrays deployed across two faults. Continuous recordings are first segmented and classified into earthquakes, rail and road traffic events, and non-traffic noise using various detection algorithms. We then employ beamforming on traffic signals and noise to resolve apparent Rayleigh-wave velocities in the 5-10 Hz frequency band, imaging the upper 100 m of each fault damage zone. Exploiting the dual-2D array coverage, we further extend beamforming analyses to regional earthquakes to reconstruct body-wave propagation directions and quantify their azimuthal rotations induced by fault-zone structures at depths of 0.5-4 km. Our results reveal strongly asymmetric shallow damage zones on both strands, with velocity reductions of about 14.6% on the BF and 35.6% on the MCF toward the northeast. At greater depths, both faults are bimaterial interfaces that separate bounding rocks with higher seismic velocities on the northeast side, in agreement with previous large-scale P-wave tomography. Integrated observations document a depth-dependent reversal in velocity contrast across these bimaterial faults, consistent with rupture dynamics that promote northwestward propagation along the SoSAF and enhanced shallow damage accumulation on the stiffer block. The MCF, in particular, exhibits a pronounced asymmetric damage zone and increased wavefield scattering, affirming its role as the principal strand of the SoSAF in this region.

Strike-slip faults-where tectonic plates grind past each other horizontally-are a defining feature of many densely populated continental seismic zones worldwide, including the San Andreas fault system in California, the North and East Anatolian faults in T & uuml;rkiye, and the Sagaing fault in Myanmar (Burma). Although their lateral motion has long been recognized, a growing body of global evidence is now highlighting a more hazardous aspect of these systems: supershear earthquakes-fast propagating ruptures that exceed the speed of shear waves and can cause disproportionately intense shaking and destruction. Four of the last six Mw 7.0+ earthquakes on strike-slip faults have been recognized as supershear events, including the damaging Mw 7.7 Myanmar and the Mw 7.8 Pazarcik earthquakes, highlighting the need to confront the potential implications of such future events.

Seismic faults are surrounded by damaged rocks with reduced rigidity and enhanced attenuation. These damaged fault zone structures can amplify seismic waves and affect earthquake dynamics, yet they are typically omitted in physics‐based regional ground motion simulations. We report on the significant effects of a shallow, flower‐shaped fault zone in foreshock‐mainshock 3D dynamic rupture models of the 2019 Ridgecrest earthquake sequence. We find that the fault zone structure both amplifies and reduces ground motions not only locally but at distances exceeding 100 km. This impact on ground motions is frequency‐ and magnitude‐dependent, particularly affecting higher frequency ground motions from the foreshock because its corner frequency is closer to the fault zone's fundamental eigenfrequency. Within the fault zone, the shallow transition to a velocity‐strengthening frictional regime leads to a depth‐dependent peak slip rate increase of up to 70% and confines fault zone‐induced supershear transitions mostly to the fault zone's velocity‐weakening roots. However, the interplay of fault zone waves, free surface reflections, and rupture directivity can generate localized supershear rupture, even in narrow velocity‐strengthening regions, which are typically thought to inhibit supershear rupture. This study demonstrates that shallow fault zone structures may significantly affect intermediate‐ and far‐field ground motions and cause localized supershear rupture penetrating into velocity‐strengthening regions, with important implications for seismic hazard assessment.

The Main Marmara fault (MMF) in northwestern Türkiye poses the highest seismic risk in broader Europe. The 2025 moment magnitude (MW) 6.2 event was the largest earthquake along the MMF in >60 years. We integrated observations from multiple temporal scales including the decade-long evolution of M > 5 earthquakes, their rupture dynamics, and aftershock patterns. We show a series of eastward-propagating M > 5 events and a gradual eastward partial rupture of the MMF over the past ~15 years. The seismically active portion of the fault includes creeping and transitional segments with some of the most recent seismicity located near the presumably locked Princes' Islands segment south of Istanbul that has the potential to generate a M ~7 earthquake. Our analysis highlights the necessity of real-time monitoring of this part of the MMF.

Earthquake swarms may be driven by slow processes within fault zones, such as fluid migration, or silent slip events. A prominent example is the 2024 Mw 7.5 Noto earthquake in Japan, which was preceded by a significant seismic swarm, likely triggered by upward fluid migration (Wang et al. 2024). These swarms provide insights into dynamic fault processes, and there is thus a need to detect smaller earthquakes for a clearer understanding of the temporal evolution of seismicity and the mechanisms driving fault activity (Shearer et al., 2022).To address these challenges, we used a dense array of 300 seismic nodes deployed for three years at the Piñon Flat Observatory (South.Cal.) along the San Jacinto Fault, one of the most seismically active areas in California. Using advanced slant stacking techniques tuned to the fault geometry at crustal P- and S-wave velocities, we significantly enhanced our detection capabilities down to approximately magnitude -2. By detecting four times as many events as the standard USGS catalog, this allowed us to highlight three distinct swarm episodes that were not identified before. These episodes exhibit a characteristic progression: an initial activation phase, a steady state culminating in a peak, followed by a final decay.We further investigate the magnitude distribution and spatial migration to propose a possible driving mechanism. This approach can be extended to other fault zones to unveil hidden fault activity.

Analysis of earthquake rupture directivity provides key information for seismic hazard and risk assessment, particularly for faults near urban areas. We analyze directivity patterns for 31 well-constrained ML >= ${M}_{L}\mathit{\ge }$ 3.5 earthquakes along the Main Marmara Fault, in direct proximity to Istanbul. We calculate source mechanisms with a waveform modeling approach and analyze earthquake directivity from apparent source-time functions using empirical Green's functions. Most of the strike-slip earthquakes to the west of the Princes Islands segment display a predominantly asymmetric rupture toward the east with the median directivity trending 85 degrees, consistent with the Main Marmara Fault strike. Consequently, earthquake ground shaking may be more pronounced toward Istanbul. This holds potentially for a large earthquake on the Main Marmara Fault which is late in its seismic cycle. Our results motivate the importance of evaluating the impact of eastward asymmetric ruptures on the probabilistic seismic hazard and risk assessment around Istanbul.

Earthquakes are assumed to be unpredictable, but their forecasting may improve if signatures of preparatory processes can be reliably identified through continuous monitoring. Recent results suggest that years to months before large earthquakes, progressive rock weakening can lead to seismicity localization and coalescence, facilitating fault rupture. If this holds generally, comprehensive, years-long analyses of seismicity may help detect transients signaling proximity to large failure. We test this hypothesis by considering the 2023, MW 7.8 Kahramanmaras,, T & uuml;rkiye, earthquake as a case study. A previous study identified an 8-month long activation of seismicity clusters in a complex fault network within 50 km of the future epicenter. To track earthquake evolution at higher resolution, we developed an enhanced seismic catalog combining deep-learning and classic techniques for the six years preceding the mainshock. Recurrent seismicity on the Narl & imath; Fault, a secondary fault where the mainshock nucleated before propagating onto the major East Anatolian Fault Zone intensified months before the event, exhibiting increased localization and interaction. Moreover, in the weeks preceding the mainshock, seismicity surged on a previously quiescent branch aligning with the future rupture plane, yet no immediate foreshocks were observed in the final hours. We propose that persistent damage-induced weakening near the nucleation region primed the system for failure, ultimately enabling rupture propagation toward the main fault. Our findings underscore the importance of long-term, and consistent high-resolution seismic monitoring and analysis for tracking spatiotemporal seismicity transients that may serve as indicators of proximity to rupture.

We develop an earthquake simulator to study the partitioning of seismic/aseismic slip and dynamics of Earthquakes on a Heterogeneous strike-slip Fault (HFQsim) using a generalized model of a discrete fault governed by static/dynamic friction and creep in an elastic half-space. Previous versions of the simulator were shown to produce various realistic seismicity patterns (e.g., frequency-magnitude event statistics, hypocenter and slip distributions, temporal occurrence) using friction levels and creep properties that vary in space but are fixed in time. The new simulator incorporates frictional heat generation by earthquake slip leading to temperature rises, subsequent diffusion cooling into the half space, and time-dependent creep on the fault. The model assumes a power law dependence of creep velocity on the local shear stress, with temperature-dependent coefficients based on the Arrhenius equation. Temperature rises due to seismic slip produce increased aseismic slip, which can lead to further stress concentrations, aftershocks, and heat generation in a feedback loop. The partitioning of seismic/aseismic slip and space-time evolution of seismicity are strongly affected by the temperature changes on the fault. The results are also affected significantly by the difference between the static and kinetic friction levels. The model produces realistic spatio-temporal distribution of seismicity, transient aseismic slip patterns, mainshock-aftershock sequences, and a bimodal distribution of earthquakes with background and clustered events similar to observations. The HFQsim may be used to clarify relations between fault properties and different features of seismicity and aseismic slip, and to improve the understanding of failure patterns preceding large earthquakes.

Using a novel high-performance computing implementation of a nonlinear continuum damage-breakage model, we explore interactions between 3D co-seismic off-fault damage, seismic radiation, and rupture dynamics. Our simulations demonstrate that off-fault damage enhances high-frequency wave radiation above 1 Hz, reduces rupture speed and alters the total kinetic energy. We identify distinct damage regimes separated by solid-granular transition, with smooth distributions under low damage conditions transitioning to localized, mesh-independent shear bands upon reaching brittle failure. The shear band orientations depend systematically on the background stress and agree with analytical predictions. The brittle damage inhibits transitions to supershear rupture propagation and the rupture front strain field results in locally reduced damage accumulation during supershear transition. The dynamically generated damage yields uniform and isotropic ratios of fault-normal to fault-parallel high-frequency ground motions. Co-seismic damage zones exhibit depth-dependent width variations, becoming broader near the Earth's surface consistent with field observations, even under uniform stress conditions. We discover a new delayed dynamic triggering mechanism in multi-fault systems, driven by reductions in elastic moduli and the ensuing stress heterogeneities in 3D tensile fault step-overs. This mechanism affects the static and dynamic stress fields and includes the formation of high shear-traction fronts around localized damage zones. The brittle damage facilitates rupture cascading across faults, linking delay times directly to damage rheology and fault zone evolution. Our results help explain near-fault high-frequency isotropic radiation and delayed rupture triggering, improving our understanding of earthquake processes, seismic wavefields and fault system interactions.

The fault damage zone is a region surrounding an earthquake fault interface where rocks are significantly fractured due to tectonic movements and historical large earthquakes on the fault. The rock fractures within the damage zone absorb and scatter seismic waves, causing amplitude decay in different frequency ranges. In this study, we use ambient noise attenuation tomography to image the fault damage zones in two tectonic settings: a transform fault in southern California and a thrust fault in western Sichuan. According to dynamic rupture models, a preferred rupture direction leads to asymmetric damage zones adjacent to the fault interface. In the Ramona array example for the San Jacinto Fault, the velocity contrast across the strike-slip fault interface leads to a preferred rupture direction towards northwest, resulting in more pronounced damage on the side with higher-velocity at depth. In the Hongkou array example for the Longmenshan Fault, significant rock damage is observed at ~ 1 km depth in the footwall side of the thrust fault interface due to upward rupture propagation from seismogenic depths. Combined with ambient noise differential adjoint tomography, a more detailed S-wave velocity model can be derived, facilitating the interpretation of tectonic structure across the fault interface and further constraining the asymmetric nature of the observed fault damage zones as predicted by dynamic rupture models.

We discuss general structural features of the Banning and Mission Creek strands (BF and MCF) of the southern San Andreas fault (SSAF) in the Coachella Valley, based on ambient noise and earthquake wavefields recorded by a seismic array with >300 nodes. Earthquake P arrivals show rapid changes in waveform characteristics over 20–40 m zones that coincide with the surface BF and MCF. These variations indicate that the BF and MCF are high-impedance contrast interfaces—an observation supported by the presence of seismic reflections. Another prominent but more diffuse change in SSAF structure is found ∼1 km northeast of the BF. This feature has average-to-low arrival times (P and S) and ambient noise levels (at <30 Hz), and likely represents a relatively fast velocity block sandwiched between broader MCF and BF zones. The maximal arrival delays (P ∼0.1 s and S ∼0.25 s) and the highest ambient noise levels (>2 times median) are consistently observed southwest of the BF—a combined effect of Coachella Valley sediments and rock damage on that side. Immediately northeast of the MCF, large S minus P delays suggest a broad high VP/VS zone associated with asymmetric rock damage across the SSAF. This general overview shows the BF and MCF as mature but distinctly different fault zones. Future analyses will further clarify these and other SSAF features in greater detail.