Seismica

Investigating the D" Reflector Beneath the Indian Ocean with Source Arrays

Christine Thomas, Björn Holger Heyn, Lena Tölle, Rûna van Tent — Tue, 08 Jul 2025 00:00:00 -0400

We used seismic P-wave reflections to search for the discontinuity at the top of the D" region beneath the Indian Ocean. Due to a lack of seismic receiver arrays to target this region, we build source arrays using earthquakes in Indonesia and taking advantage of the long-running history of GEOSCOPE stations located in the western Indian Ocean and Antarctica, as well as three additional stations (Seychelles and Antarctica). Despite restricting the earthquake depth range, source-array stacks were difficult to interpret due to complications arising from differing earthquake depths, violating the plane wave assumption. Therefore, we use a source-array scatter imaging method, that does not rely on travel-times calculated for a plane wave. Using this technique in conjunction with source normalization, we found evidence for a D" P-wave reflector for several stations with reflector depths varying between 230-160 km above the CMB South of Australia and 190 to 270 km above the CMB beneath the Indian Ocean, where the depth of the reflector in the north of our study area is consistent with previously imaged D" depths using S-waves and agrees with receiver array data. We suggest that earlier imaged subducted lithosphere in this region is responsible for our D" reflections.

The 1804 Alborán seismic series: Search for the source using seismic scenarios and static stress interactions

Yolanda de Pro Díaz, José Jesús Martínez Díaz, Carolina Canora Catalán — Mon, 14 Jul 2025 00:00:00 -0400

Linking historical earthquakes with the faults that caused them is crucial for seismic hazard assessment. Historical documentation describing the effects of an earthquake is a useful information source, from which we can compile the observed intensity field of the earthquake. In this work, we use intensity data from the catastrophic 1804 Alborán earthquake (south of Iberia) along with intensity simulations and coseismic stress transfer analysis to search for this earthquake's seismic source. We build intensity simulations for each fault proposed as a potential source, and compare these simulations with the intensity field. We also propose the possibility of the Alborán 1804 earthquake triggering the Dalías earthquake (European macroseismic intensity (IEMS-98) IX), which occurred seven months after, and analyze stress transfer between the possible sources of both earthquakes. Our results point to a conjunct rupture of the northern Al-Idrissi Fault segment and the North–South Faults as the most likely source for the Alborán earthquake.

Picking Induced Seismicity with Deep Learning (piSDL)

Janis Heuel, Vincent Maurer, Michael Frietsch, Andreas Rietbrock — Tue, 19 Aug 2025 00:00:00 -0400

Training deep-learning picking models with several published data sets can be easily done through the Python toolbox SeisBench. Most of the data sets contain earthquakes recorded at local, regional and teleseismic distances, with only limited data in the low magnitude, close distance region. Applying current published PhaseNet models to induced seismicity data leads to only a few events being detected and trained PhaseNet models are not able to outperform well-established workflows in seismology.
Here we present a new seismological data set and trained PhaseNet models for picking induced seismicity with deep-learning (piSDL). PhaseNet was trained with 171,182 three component waveforms from 40,576 events. Noise samples were added in the training data set to reduce the number of false picks. In this study, we noticed that a good earthquake training data set and noise samples from the analysed area are both important to detect more seismic events with a newly trained PhaseNet model. We validated our new PhaseNet models at a geothermal site in Rittershoffen (France). The models trained with the new data set and noise samples clearly outperform PhaseNet’s original published model and traditional methods in seismology by detecting up to 62% more events compared to a seismicity catalogue published by an agency.

Masked graph neural network for rapid ground motion prediction in Italy

Tue, 16 Sep 2025 00:00:00 -0400

This study presents an updated version of TISER-GCN, a graph neural network (GCN) designed to predict maximum intensity measurements (IMs) from 10-second seismic waveforms starting at the earthquake origin time, without prior knowledge of location, distance, and magnitude. The improved model was applied to nearly 600 seismic stations from the INSTANCE benchmark dataset, significantly expanding the original TISER-GCN setup, which was limited to 39 stations in a smaller area of central Italy. Input data consist of three-component waveforms selected to ensure high quality and minimize saturation. Results show that masking stations where the P-wave arrives within the first 10 seconds , combined with the integration of additional information, reduces the mean squared error (MSE) by up to 6% for peak ground acceleration (PGA) and 5.5% for peak ground velocity (PGV), compared to the unmasked baseline. Moreover, the proposed approach yields near-zero median residuals across all IMs, mitigating the systematic underestimation observed when using a ground motion model specifically developed for Italy. These findings indicate that the model provides accurate predictions of ground motions, comparable to those obtained with the original TISER-GCN, which, however, requires a fixed seismic network geometry.

ScarpLearn: an automatic scarp height measurement of normal fault scarps using convolutional neural networks

Wed, 16 Jul 2025 00:00:00 -0400

Geomorphic markers such as displaced surfaces, offset rivers or scarps are witnesses to the neotectonic activity of the faults.
The characterization (such as fault detailed surface trace, the scarp height, etc.) of these geomorphological markers is currently a time-consuming step with expert-dependent results, often qualitative and with uncertainties that are difficult to estimate. To overcome those issues, we present a proof of concept study for the use of deep learning in morphotectonics, specifically on fault markers. We developed a Bayesian supervised machine learning method using one-dimentional (1D) convolutional neural networks (CNN) trained on a database of simulated topographic profiles across normal fault scarps, called ScarpLearn. From a topographic profile, ScarpLearn is able to automatically give the cumulative scarp height with an uncertainty. We have developed two versions: one designed for more generalized cases involving profiles with multiple fault scarp (ScarpLearn), and another specifically trained to handle profiles featuring a single fault scarp (ScarpLearn_1F). We apply ScarpLearn for the characterization of active normal faults in extensional settings such as the Trans-Mexican Volcanic Belt and Malawi Rift system. From those specific case studies, we explore the progress (computation time, accuracy, uncertainties) that machine learning methods bring to the field of morphotectonics, as well as the current limits (such as
bias). Our results show that we are able to develop a CNN model that is estimating scarp heights on topographic profiles from 5m resolution digital elevation model. We compared the results obtained with ScarpLearn and other non deep-learning methods. ScarpLearn achieves similar accuracy while being much faster and having smaller uncertainties. We invite readers to use and to extend our study: codes to build the synthetic scarp database and for the CNN model ScarpLearn are available at: https://gricad-gitlab.univ-grenoble-alpes.fr/poussel/scarplearn.

Societal Impact of COVID-19 Crisis on the Ambient Seismic Noise in Metropolitan France

Wed, 17 Sep 2025 00:00:00 -0400

The COVID-19 pandemic led to restrictions on human mobility worldwide. In France, numerous phases of lockdowns and curfews were instituted in an attempt to limit the consequences of this pandemic. Through these various phases of restrictions, we analysed changes in human activity based on the study of ambient seismic noise level in metropolitan France. We propose a different approach to previous studies, studying variations in the seismic noise level between the pandemic years 2020 and 2021 with respect to 2019, before the pandemic, taken as a reference. We focused our work between 4 and 8 Hz, where human induced noise sources are significant. We took advantage of the wide instrumental coverage of metropolitan France to distinguish the effects of restrictions in urbanized and rural areas. Whether in urban or rural areas, the effects of lockdowns and curfews coincide with reductions in seismic noise levels. The magnitude of the noise level reduction is greater for the first than for the last lockdowns. We also observe a signature of curfew periods and analyse variations according to time of day and day of the week. Changes in road traffic during lockdowns and curfews are a major factor contributing to the observed variations in ambient seismic noise.

Benchmarking seismic phase associators: Insights from synthetic scenarios

Jorge Puente Huerta, Christian Sippl, Jannes Münchmeyer, Ian W. McBrearty — Tue, 09 Sep 2025 00:00:00 -0400

Reliable seismicity catalogs are fundamental for seismological analysis. Following phase picking, phase association groups arrivals into sets with consistent origins (i.e., events), determines event counts, and identifies outlier picks. To handle the substantial increase in the quantity of seismic phase picks from improved picking methods and larger deployments, several novel phase associators have recently been proposed. This study presents a detailed benchmark analysis of five seismic phase associators, including classical and machine learning-based approaches: PhaseLink, REAL, GaMMA, GENIE, and PyOcto. We use synthetic datasets mimicking real seismicity characteristics in crustal and subduction zone scenarios. We evaluate performance for different conditions, including low- and high- noise environments, out-of-network events, very high event rates, and variable station density. The results reveal notable differences in precision, recall, and computational efficiency. GENIE and PyOcto demonstrate robust performance, with almost perfect performance for most scenarios, but under the most challenging conditions with high noise levels and event rates, performance drops while F1 scores still remain above 0.8. PhaseLink's performance declines with noise and event density, particularly in subduction zones, dropping to near zero in the most complex cases. GaMMA outperforms PhaseLink but struggles with accuracy and scalability in high-noise, high-density scenarios. REAL performs reasonably but loses recall under extreme conditions. PyOcto and PhaseLink show the quickest runtimes for smaller-scale datasets, while REAL and GENIE are more than an order of magnitude slower for these cases. At the highest pick rates, GENIE’s runtime disadvantage diminishes, matching PyOcto and scaling effectively. Our results can guide practitioners compiling seismicity catalogs and developers designing novel associators.

Impact of Seismic Attenuation Corrections on Source Parameter Estimation

Dino Bindi, Matteo Picozzi, Adrien Oth, Daniele Spallarossa — Fri, 22 Aug 2025 00:00:00 -0400

We estimate the stress drop ∆σ for 551 earthquakes from the 2019 Ridgecrest sequence in Southern California using a spectral decomposition. To assess the impact of propagation model assumptions, we apply a 2D cell-based approach that accounts for lateral attenuation variations and compare results with previous models using distance and depth-dependent attenuation. The 95% confidence interval for azimuthal-dependent attenuation over an 80 km radius is 0.290 at 2 Hz and 0.473 at 14 Hz (log10 units). While the 2D model reveals significant azimuthal variations, the overall ∆σ distribution remains similar to that from a simple distance-dependent model, at least for the analyzed data set. High ∆σ is observed near the M7.1 and M6.4 events, while lower values appear at shallower depths, especially toward the Coso region and near the left-lateral fault junction of the M6.4 sequence. All models consistently identify a high-∆σ region at 4-8 km depth between stations CLC and WRC2, north of the M7.1 hypocenter, where the main fault bends. While spatial comparisons reveal more localized differences, the most pronounced impact arises when the attenuation model incorporates depth dependence.

Introducing the Rapid Earthquake Damage Estimation (RED-E) System for Improved Life Safety Outcomes During Earthquake Early Response in Canada

Megumi Patchett, Tiegan Hobbs, Lucinda Leonard — Sat, 06 Sep 2025 00:00:00 -0400

In the wake of a major earthquake in Canada, responders can expect to encounter critical gaps in situational awareness in the first 48-72 hours that will hamper effective decision-making. To address this challenge, Natural Resources Canada is developing the Rapid Earthquake Damage Estimation (RED-E) system. This modelling system aims to produce maps of structural, human, and economic impacts within tens of minutes of a significant seismic event, similar to the United States Geological Survey's PAGER product but with enhanced details specific to Canada. This paper presents our research on optimizing the RED-E system through the User-Centered Design approach. End-user consultation throughout the development of RED-E will ensure that its outputs are practical and actionable for first responders, emergency managers, and infrastructure operators. Key findings from initial consultations underscore the need for immediate post-earthquake situational awareness, although complete understanding may take days to weeks. End-users expressed a preference for RED-E outputs in diverse formats, with road disruption modelling and secondary hazard assessments being particularly valuable. This study outlines the essential requirements for the outputs of RED-E and documents initial prototypes, showcasing the potential of the system to transform early post-seismic emergency response efforts across Canada, aiding in prioritization and resource allocation until ground-truth data become available.

Forecasting 3D Rupture Dynamics of the Alto Tiberina Low-Angle Normal Fault, Italy

Thu, 21 Aug 2025 00:00:00 -0400

The seismic potential of active low-angle normal faults (LANFs, <30° dip) remains enigmatic under Andersonian faulting theory, which predicts that normal faults dipping less than 30° should be inactive. The Alto Tiberina fault (ATF) in the northern Apennines, a partly creeping 17°-dipping LANF, has not been associated with any historical earthquakes but could potentially generate earthquakes up to Mw~7. We investigate the mechanical preconditions and dynamic plausibility of large ATF earthquakes using 3D dynamic rupture and seismic wave propagation simulations constrained by multidisciplinary data from the Alto Tiberina Near Fault Observatory (TABOO-NFO). Our models incorporate the complex non-planar ATF fault geometry, including hanging wall secondary faults and a recent geodetic coupling model. We show that potential large earthquakes (up to Mw~7.4) are mechanically viable under Andersonian extensional stress conditions if the ATF is statically relatively weak (μ_s=0.37). Large earthquakes only nucleate on favorably oriented, steeper fault sections (dip ≥30°), and remain confined to the coupled portion, limiting earthquake magnitude. These ruptures may dynamically trigger an intersecting synthetic branch but are unlikely to affect more distant antithetic faults. Jointly integrating fault geometry and geodetic coupling is crucial for forecasting dynamic rupture nucleation and propagation.

Depth-varying azimuthal anisotropy and mantle flow in the Patagonian slab window

Hannah Mark, Douglas Wiens, Walid Ben Mansour, Zhengyang Zhou — Fri, 11 Jul 2025 00:00:00 -0400

Subduction of spreading ridges forms slab windows which perturb the local structure and dynamics of the upper mantle. Slab windows may alter the pattern of mantle flow and serve as portals for the exchange of mantle material between upper mantle reservoirs that are otherwise separated by the boundary of the subducting slab. Here, we use Rayleigh waves to derive an azimuthally anisotropic regional seismic velocity model for the Patagonian slab window and use the anisotropy model to infer patterns of upper mantle flow and deformation. Anisotropic fast directions are primarily trench-parallel in the upper ~40 km of the mantle throughout the region, likely reflecting the history of subduction and compression along the South American margin. At greater depths sensed by long-period Rayleigh waves, fast directions within the youngest part of the slab window are consistent with cross-basin mantle flow between the Atlantic and Pacific, as previously suggested by shear wave splits. Overall, the anisotropic velocity model reveals complex, depth-dependent patterns of mantle deformation and flow within the Patagonian slab window.

Towards surface wave tomography with 3D resolution and uncertainty

Wed, 13 Aug 2025 00:00:00 -0400

Surface-wave tomography is crucial for mapping upper-mantle structure in poorly instrumented regions such as the oceans. However, data sparsity and errors lead to tomographic models with complex resolution and uncertainty, which can impede meaningful physical interpretations. Accounting for the full 3D resolution and robustly estimating model uncertainty remains challenging in surface-wave tomography.
Here, we propose an approach to provide direct control over the model resolution and uncertainty and to produce these in a fully three-dimensional framework by combining the Backus-Gilbert-based SOLA method with finite-frequency theory.
Using a synthetic setup, we demonstrate the reliability of our approach and illustrate the artefacts arising in surface-wave tomography due to limited resolution. We also indicate how our synthetic setup enables us to discuss the theoretical model uncertainty (arising due to assumptions in the forward theory), which is often overlooked due to the difficulty in assessing it.
We show that the theoretical uncertainty components may be much larger than the measurement uncertainty, thus dominating the overall uncertainty. Our study paves the way for more robust and quantitative interpretations in surface-wave tomography.

A repeating earthquake catalog for Northern Chile

Jonas Folesky, Jörn Kummerow, Laurens Jan Hofman — Fri, 08 Aug 2025 00:00:00 -0400

Repeating earthquakes are nearly identical seismic events that are assumed to occur reiteratively on the same fault-patch with highly consistent focal mechanisms. Their magnitude-dependent recurrence intervals can be used to estimate local fault-slip and to infer spatial and temporal patterns of aseismically creeping zones at depth.
We construct here the first long-term repeating earthquake catalog for Northern Chile. Using waveforms for 180,000 earthquakes from a recent regional seismicity catalog as templates, a GPU-based template matching is performed to search for repeating earthquakes in the continuous, multiannual seismological data of the permanent IPOC station network between 2006 and 2024. The resulting repeater catalog contains 10,684 events grouped into 3153 families. We observe a notable variability of size and behavior of the families, ranging from long-lasting, regular sequences to short-term, burst-type repeaters. Two megathrust earthquakes in 2007 and 2014 have a strong effect on the spatio-temporal distribution of repeaters. We compute the first time-dependent slip map for the interface between the subducting Nazca slab and the overriding South American plate.
The catalog facilitates future detailed analysis of rupture processes, source structures and the spatio-temporal evolution of slow slip at depth. It also allows for comparative studies with other subduction zones, such as Japan.

Global EarthquakE ScEnarios (GEESE): An OpenQuake Engine-Based Rupture Matching Algorithm and Scenarios Database for Seismic Source Model Testing and Rapid Post-Event Response Analysis

Tue, 29 Jul 2025 00:00:00 -0400

The Global EarthquakE ScEnarios (GEESE) algorithm retrieves from a seismic hazard input model the ruptures matching a set of criteria (e.g., magnitude, location, focal mechanism). We applied the GEESE algorithm to create a publicly available database (version 1.0) of finite rupture models for past earthquakes which can be used for scenario seismic hazard and risk analysis applications. To this end, we selected earthquakes with a moment magnitude larger than 7.0 and hypocentral depth less than 200 km in the ISC-GEM catalogue (version 10.0) and retrieved the best matching ruptures from the seismic hazard models in the GEM Mosaic. The GEESE algorithm also automatically computes a set of ground-motion fields using each matched rupture, which are also provided in the database. The ability of the GEESE algorithm to test whether a Mosaic model can generate a rupture sufficiently representative of a queried event is a useful means of evaluating the Mosaic model's seismic source characterisation (SSC). Sufficiently matching ruptures are retrieved from the Global Mosaic for 90 percent of the tested ISC-GEM events. The GEESE algorithm can also be used in post-event response analysis to rapidly obtain an initial finite rupture when only minimal event information is initially available. A demonstration of these capabilities of the GEESE algorithm is provided using the 2023 Morocco earthquake, the 1994 Northridge earthquake, and the 2023 Kahramanmaras earthquake.

Seismicity and Surface Deformation in Kamanjab Inlier, Northern Namibia

Mon, 14 Jul 2025 00:00:00 -0400

The last two decades have seen the onset of felt earthquakes, including occasionally damaging events, in the Kamanjab Inlier, a block of Paleoproterozoic crystalline basement in northern Namibia. The Geological Survey of Namibia (GSN) and the Council for Geoscience, South Africa (CGS) deployed a temporary network of 10 seismic stations within the Kamanjab Inlier from June to September 2018 and cataloged ~1500 events. We used a neural network-based earthquake phase detector, EQTransformer, to enhance the published GSN catalog to >9000 detections. The double-difference earthquake relocation of ~4500 events reveals two distinct major and three minor spatial clusters that we interpret as local discrete faults that intersect the NE-dipping seismogenic fault of the 4 April 2021 Mw 5.4 earthquake, which is the largest instrumentally recorded earthquake in Namibia to date. We name the Mw 5.4 host fault "Anker Fault" and constrain its orientation using Sentinel 1 Interferometric Satellite Aperture Radar (InSAR) to image surface uplift and subsidence patterns. Given the sudden onset of the 2018 seismic activity and the absence of dams, mineral or energy exploration projects nearby, we eliminated the possibility of anthropogenic triggering. We suggest that the proximal cause for 2018 seismicity is shallow groundwater migration, possibly associated with nearby hot springs and modulated by tidal forces. The Kamanjab Inlier area has shown an increase in the number and magnitude of earthquakes from 2018 to 2021, which could pose a seismic hazard in the future. Our study introduces an earthquake detection and relocation workflow that can be adopted for regions with limited instrumentation.

Alaska Upper Crustal Velocities Revealed by Air-to-Ground Coupled Waves From the 2022 Hunga Tonga-Hunga Ha’apai Eruption

Tue, 08 Jul 2025 00:00:00 -0400

Pressure changes in the atmosphere couple to the solid earth, producing ground motions that contain information about local crustal elastic parameters. This type of air-to-ground coupled wave was observed globally following the largest explosion of the instrumental age, the climactic eruption of the Hunga Tonga-Hunga Ha’apai volcano on 15^th January, 2022. We utilize this unprecedented source, along with the presence of colocated seismometers, infrasound sensors, and barometers in Alaska, to examine coupling and reveal elastic parameters beneath the stations. We derive coupling spectra by forming seismic--to--pressure amplitude ratios as a function of frequency, and identify passbands of high coherence between the pressure and seismic records. By relating coupling spectra in high-coherence bands to elastic parameters, we estimate mean shear wave velocities under stations to a depth encompassing much of the upper crust. Our velocity estimates from low-frequency coupling exhibit good agreement with a previously existing tomographic velocity model from Berg et al. (2020), while estimates from high-frequency coupling show considerable scatter when compared to proxy V_s30, even though the overall values are reasonable. In addition to providing velocity estimates, our results also indicate that, for the broadband pressure signals from the Hunga Tonga-Hunga Ha’apai eruption, microseismic noise exerts a strong effect on the frequency bands where coupling is observed, and that the air-to-ground coupled waves exhibit significant complexity not necessarily described by theory. Our results show that coupling observations provide a simple forward observation of mean seismic velocities beneath seismoacoustic stations, without the need to resort to complex inversion schemes. It is remarkable that pressure waves generated thousands of kilometers away are able to reveal the seismic velocity structure of Alaska to several kilometers depth.

The 28 January 2020, Mw7.7, Cayman Trough / Oriente Fault, Supershear Earthquake Rupture

Sun, 31 Aug 2025 00:00:00 -0400

The largest magnitude strike-slip event of the instrumental seismology era along the northern Caribbean plate boundary, with a moment magnitude of 7.7, occurred on 28 January 2020 on the Oriente transform fault, along the northern edge of the Cayman Trough, west of Cuba. We use local, regional, and global seismic waveforms and coseismic geodetic offsets, to produce high-resolution rupture models for both the low-frequency (~ 0.02 Hz) and high-frequency (~ 1 Hz) components of the rupture using a finite fault kinematic inversion and back-projection imaging, respectively. We document a rupture that propagated predominantly unilaterally westward, with an initial phase at subshear speed for 20–-30~s and over 40 to 50~km, followed by an acceleration to supershear speed that persisted all the way to the western end of the rupture, for 40~s and over about 200~km. Supershear rupture speed is consistent with strong motion observations of low ground acceleration levels in the near-field of the fault and low aftershock production in numbers and moment release. The rupture followed a very linear, unsegmented portion of the Oriente fault that had not experienced significant seismic activity for at least a century. Observational evidence and models indicate that the 28 January 2020, Mw7.7 earthquake, supershear over most of its length, had a smooth rupture process along a simple linear fault segment where earthquake nucleation is infrequent and interseismic locking depth shallow, two characteristics that may explain this unusually large magnitude supershear event.

Adding strain rate information into a short-term seismicity model improves forecasting performances: the case of Campi Flegrei, Italy

Giuseppe Petrillo, Matteo Taroni — Tue, 16 Sep 2025 00:00:00 -0400

Campi Flegrei is a large active volcanic caldera in southern Italy, currently undergoing a prolonged phase of unrest that began in 2005, characterized by ground uplift and an increase in seismicity. Classical short-term seismicity models, such as the temporal Epidemic Type Aftershock Sequence (ETAS) model, rely exclusively on earthquake catalog data and do not incorporate external forcing mechanisms like crustal deformation. In this study, we extend the ETAS model by integrating strain rate information derived from GNSS measurements, allowing the background rate to vary in time through a data-driven convolution with an empirically estimated response kernel. Using eleven years of observations (2013-2024), we compare the forecasting performance of the classical and deformation-driven ETAS models. Our results show that including strain rate significantly improves forecasting ability, as evidenced by a lower Akaike Information Criterion (AIC). This finding suggests that incorporating geodetic signals into seismicity models enhances their physical realism and predictive skill, providing a promising path toward Operational Earthquake Forecasting in active volcanic regions.

Supershear source model of the 2025 M7.8 Myanmar earthquake and paleoseismology of the Sagaing Fault: regions of significant overlap with past earthquakes

Wed, 30 Jul 2025 00:00:00 -0400

The 2025 Mw 7.8 earthquake on the central Sagaing Fault is one of the most destructive seismic events in Myanmar's recorded history, producing near-fault shaking exceeding Modified Mercalli Intensity X and impacting tens of millions of people across Southeast Asia. We present a detailed kinematic rupture model of the event based on joint inversion of regional strong motion waveforms and Sentinel-1 SAR pixel offsets. The rupture extended over ~450 km with an average slip of 3--5 m, predominantly within the upper 10 km of the crust. Inversions favor a maximum rupture speed of ~4.8 km/s, consistent with supershear propagation inferred from near-field waveform observations. We also report on paleoseismic evidence from a key site at the epicenter of the 2025 earthquake near Mandalay, which reveals five surface-rupturing earthquakes over the past millennium, with similar average displacement. Our results indicate a pattern of overlapping large ruptures along the central fault, with implications for rupture segmentation, recurrence, and seismic hazard. Given the exceptional exposure to earthquakes and high strain rates, our findings underscore the need for urgent attention to earthquake preparedness and infrastructure resilience in central Myanmar.

Supershear-subshear-supershear rupture sequence during the 2025 Mandalay Earthquake in Myanmar

Shiro Hirano, Ryosuke Doke, Takuto Maeda — Fri, 01 Aug 2025 00:00:00 -0400

We investigated the rupture dynamics of the 2025 M_w7.7 Mandalay, Myanmar earthquake, using a video recording of surface rupture, strong motion recordings, waveform simulation, and satellite imagery. Our assessment, based on the S-wave observation in the video and rupture arrival time at a seismic station 246 km south of the hypocenter, suggests that rupture decelerated to subshear speeds (~3 km/s) from initial supershear propagation (~6 km/s) before reaching the camera location. This deceleration is also supported by comparison between the fault-normal acceleration patterns seen in the video and that simulated by kinematic rupture modeling. Additionally, satellite imagery indicated a local minimum in slip (2-3 m) approximately 40-60 km south of the epicenter, suggesting a region of reduced stress drop that likely caused the temporary deceleration. Beyond this point, the rupture appears to have re-established supershear propagation.

A 3D finite-element mesh for modeling large-scale surface deformation induced by subduction megathrust earthquakes: Application to Chile

Fri, 05 Sep 2025 00:00:00 -0400

Megaearthquakes (Mw > 8) cause continental-scale, long-lasting surface deformation, mainly due to viscoelastic relaxation of the asthenosphere. To investigate the links between this deformation and the slip history along subduction interfaces—including earthquakes, postseismic slip, and interseismic coupling—large 3D spherical finite-element meshes are required.
This technical report introduces the various steps to build Chile_Mesh_v1.0, a customizable mesh for the Chilean subduction zone, designed as a robust platform for testing various viscoelastic rheologies. It spans ~8500 km in longitude, ~7300 km in latitude, encompassing the entire South American plate, and from the surface to 2900 km depth. Special care was taken to reproduce the complex slab geometry, especially in flat-slab regions such as the Pampean and Peruvian segments, following the Slab2 model.
We show that accurately modeling both coseismic and postseismic deformation over large scales requires realistic meshed domains, extending down to the Core-Mantle boundary and thousands of kilometers from the trench. In some cases, depth-reduced meshes can be used to model viscoelastic postseismic deformation, but they fail to simultaneously capture coseismic deformation accurately. We hope this open-access mesh proves valuable for researchers studying subduction dynamics in Chile and supports the development of similar models for other regions.

A seismological large-N multisensor experiment to study the magma transfer of intracontinental volcanic fields: The example of the Eifel, Germany

Thu, 02 Oct 2025 00:00:00 -0400

The understanding of the magma system beneath intracontinental volcanic fields depends critically on our ability to resolve small-sized anomalies distributed over large areas of hundreds of kilometres. Magmatic reservoirs co-exist at different depths in the upper mantle and crust and may consist of extensive zones of crystal mush, swarms of sills and dikes of different ages and states, pore space saturated by volatiles or melt, or larger-volume, differentiated magma.
Passive seismological experiments with a large number of sensors deployed with small interstation spacings, combining different types of sensors and fibre-optic sensor technology, have great promise for addressing the resolution to capture the distributed magmatic system.
We report on a one-year, large-N experiment in the Quaternary volcanic fields of the Eifel, Germany, where more than 494 seismic stations were deployed and combined with a 64-km-long DAS cable and permanent stations. A cloud-based, open-source GIS system was implemented to address logistical challenges and ensure data quality combined with seismological analysis and visualisation tools. We present initial results to test the potential of such an extensive waveform database and automated processing for locating small earthquakes and imaging crustal and upper mantle anomalies using techniques such as ambient noise cross-correlation, receiver functions, and SKS splitting.

A brief exploration of open-source gradient-based numerical optimization Python libraries for full-waveform inversion

Oscar Mojica — Tue, 19 Aug 2025 00:00:00 -0400

Geoscientists favor Python for its user-friendly interface and scientific packages that support application implementation. Python's capabilities make it particularly useful for seismic full waveform inversion (FWI), which can see its implementation time reduced by making use of its extensive library collection. We compare four open-source gradient-based optimization Python packages - scipy.optimize, sotb-wrapper, NLopt, and PyROL - for solving the FWI optimization problem. The comparison is based on the packages' core features, documentation, and learning curves evaluated through the implementation of a 2D time-domain FWI application, built using the Devito modeling engine along with the aforementioned optimization packages. We detail how one can use a particular solver from each package for the solution of a bound-constrained optimization problem such as FWI. The open-source FWI template models used to obtain the numerical results are provided.

A Global-scale Database of Seismic Phases from Cloud-based Picking at Petabyte Scale

Sat, 06 Sep 2025 00:00:00 -0400

We present the first global-scale database of 4.3 billion P- and S-wave picks extracted from 1.3 PB continuous seismic data via a cloud-native workflow. Using cloud computing services on Amazon Web Services, we launched ~145,000 containerized jobs on continuous records from 47,354 stations spanning 2002-2025, completing in under three days. Phase arrivals were identified with a deep learning model, PhaseNet, through an open-source Python ecosystem for deep learning, SeisBench. To visualize and gain a global understanding of these picks, we present preliminary results about pick time series revealing Omori-law aftershock decay, seasonal variations linked to noise levels, and dense regional coverage that will enhance earthquake catalogs and machine-learning datasets. We provide all picks in a publicly queryable database, providing a powerful resource for researchers studying seismicity around the world. This report provides insights into the database and the underlying workflow, demonstrating the feasibility of petabyte-scale seismic data mining on the cloud and of providing intelligent data products to the community in an automated manner.