Search

Seismicity at active volcanoes provides crucial constraints on the dynamics of magma systems and complex fault activation processes preceding and during an eruption. We characterize time‐dependent spectral features of volcanic earthquakes at Axial Seamount with unsupervised machine learning (ML) methods, revealing mixed frequency signals that rapidly increase in number about 15 hr before eruption onset. The events migrate along pre‐existing fissures, suggesting that they represent brittle crack opening driven by influx of magma or volatiles. These results demonstrate the power of unsupervised ML algorithms to characterize subtle changes in magmatic processes associated with eruption preparation, offering new possibilities for forecasting Axial's anticipated next eruption. This analysis is generalizable and can be employed to identify similar precursory signals at other active volcanoes. Plain Language Summary: Our research used observations of small earthquakes to understand the dynamic behaviors of magma and fault systems before and during a volcano eruption. Specifically, we used ML techniques to search for patterns in the waveforms that may inform us of their associated physical processes. At Axial Seamount, an active underwater volcano, we discovered distinct patterns in earthquake signals preceding and during the 2015 eruption. Based on event spectral patterns, we identified signals of mixed‐frequency earthquakes that rapidly increase in number about 15 hr before the eruption starts and migrate along pre‐existing eruptive fissures. The spectral pattern involves a mixture of low frequency energy following the first arrivals, which we interpret to represent opening of cracks and being filled with magma or gases. Our study demonstrates that we can use ML algorithms to detect subtle changes in volcanic signals and help us better understand the processes leading up to an eruption. This may help us in forecasting Axial's upcoming eruption and can be applied to other active volcanoes too. Key Points: Unsupervised learning separated regular earthquakes and precursory mixed frequency earthquakes (MFEs) based on different spectral patternsThe regular earthquakes have strong tidal modulation, corresponding to failures on the caldera ring faults triggered by tidal stress changesThe MFEs intensify 15 hr before eruption and migrate along pre‐existing fissures, likely associated with eruption preparation processes [ABSTRACT FROM AUTHOR]

Artificial intelligence (AI) seismology has witnessed enormous success in a variety of fields, especially in earthquake detection and P and S‐wave arrival picking. It has become widely accepted that DL techniques greatly help routine seismic monitoring by enabling more accurate picking than traditional pickers like STA/LTA. However, a completely automatic AI‐driven earthquake monitoring framework has not been reported due to the concerns of potential false positives using DL pickers. Here, we propose a novel AI‐facilitated near real‐time monitoring framework using a recently developed deep learning (DL) picker (EQCCT) that has been deployed in the Texas seismological network (TexNet). For the West Texas area, TexNet's seismic monitoring relies on the EQCCT picker to report earthquake events. For earthquakes with a magnitude above two, the picks are further validated by analysts to output the final TexNet catalog. Due to the fast‐increasing seismicity caused by continuing oil&gas production in West Texas, this AI‐facilitated framework significantly relieves the workload of TexNet analysts. We show the mean absolute error (MAE) of automatic magnitude estimation for the magnitude‐above‐two earthquakes is smaller than 0.15 in West Texas and MAEs of hypocenter locations within 2.6 km in both distance and depth estimates. This research provides more evidence that DL pickers can play a fundamental role in daily earthquake monitoring. Plain Language Summary: AI has been a significant part of almost every aspect of our life. However, it is not widely accepted that AI can be a game changer in geoscience. In this work, we provide a compelling example that AI can significantly lower the overload of earthquake analysts in everyday work and boost our earthquake detectability, thereby enhancing our understanding of earthquake statistics, physics, and potential nucleation mechanisms. In a nutshell, AI has already been demonstrated to be successful in earthquake analysis but never in real‐time monitoring, which requires a high success rate and zero tolerance to large‐earthquake detection errors. Here, we show that AI‐assisted earthquake monitoring workflow can be almost 100% accurate for moderate‐to‐large earthquakes. Key Points: We propose an AI‐facilitated near real‐time monitoring framework that has been deployed at TexNet using a recently developed deep learning (DL) pickerDue to the fast‐increasing seismicity in West Texas, this AI‐facilitated framework significantly relieves the workload of TexNet analystsThe mean absolute error (MAE) of automatic magnitude estimation for the magnitude‐above‐two earthquakes is smaller than 0.15 in West Texas [ABSTRACT FROM AUTHOR]

The north‐south seismic belt of China poses a high risk of earthquakes, necessitating the need for accurate and rapid prediction of intensity measures (IMs) to prevent and mitigate potential damage. We have developed a new multi‐task model, CRAQuake, to predict IMs for the north‐south seismic belt of China. Using initial arrival seismic waves recorded at a single station as input, CRAQuake simultaneously predicts six IMs without relying on pre‐configured parameters such as earthquake source, path, and location. The model was trained on 4,281 sets of strong motion records data sets at 822 stations and tested to show highly correlated results with the target IMs. The prediction performance continues to improve as the input initial arrival seismic wave time window increases. CRAQuake promises to enhance the accuracy and timeliness of IMs prediction in the north‐south seismic belt of China. Plain Language Summary: The north‐south seismic belt of China, a region at high risk for earthquakes, necessitates the accurate and rapid prediction of earthquake intensity measures (IMs) to minimize potential damage. We have developed a powerful tool, CRAQuake, to address this critical need. This advanced model leverages initial seismic waves recorded at a single station to simultaneously predict six different IMs without relying on preset information like earthquake source, path, or location. Trained on a vast data set of strong motion records from 822 stations, our testing has shown that CRAQuake's predictions are highly aligned with the actual IMs. Furthermore, increasing the time window of the initial seismic waves used as input significantly improves the model's prediction accuracy. With CRAQuake, we can look forward to more accurate and timely predictions of IMs in the north‐south seismic belt of China, empowering us to better prepare for and respond to earthquakes. Key Points: CRAQuake: a model for the simultaneous prediction of six different intensity measures based on data‐driven techniquesModel performance tested by earthquake events in the north‐south seismic belt of ChinaThe model can improve accuracy and timeliness as the input seismic wave is continuously updated [ABSTRACT FROM AUTHOR]

Earthquake early warning (EEW) is of great significance in mitigating seismic disasters. Traditional EEW algorithms, which are knowledge‐driven approaches, rely on seismologists' analysis. The limited intensity measures were extracted by seismologists from P‐wave signals. And there is considerable uncertainty for predicting epicentral distance, magnitude, peak ground acceleration (PGA), and peak ground velocity (PGV). Currently, data‐driven deep learning methods with the strong learning abilities do not consider knowledge information from seismologists in EEW; thus, there is unexplored potential in enhancing the performance of deep learning models for EEW. Here, we construct the Data‐knowledge driven Hybrid deep Learning network (DHLnet) for EEW using the waveform input, knowledge embedding, convolutional neural network and graph convolutional network, aiming to integrate knowledge information from knowledge‐driven methods and the strong learning ability of data‐driven deep learning methods, that is, improving the performance of EEW. For the same test data set, compared with knowledge‐driven methods and data‐driven deep learning models, we demonstrate that DHLnet enhances the timeliness and robustness in predicting the epicentral distance, magnitude, PGA, and PGV during 10 s time window following the arrival of P‐wave. Furthermore, to validate the generalization and robustness of the DHLnet in EEW, we applied the trained DHLnet to an independent data set, within first few seconds after an earthquake occurs, DHLnet can provide robust magnitude estimation, epicentral distance estimation and high alarm accuracy. The potential of the proposed network is to enhance the performance of EEW systems and provides new insights into the exploration of deep learning methods for EEW domain. Plain Language Summary: Earthquake early warning (EEW) relies heavily on crucial parameters like epicentral distance, magnitude, and peak ground motion (peak ground acceleration [PGA] and velocity [PGV]). To quickly and accurately determine these parameters, traditional EEW algorithms, which are knowledge‐driven approaches, rely on seismologists' analysis based on earthquake rupture physics, and establish empirical EEW parameter prediction equations. Currently, data‐driven deep learning models with strong learning ability in EEW are mainly used to extract features from raw seismic waveforms, and do not take existing knowledge information from traditional EEW algorithms into account. Therefore, there is underutilized potential in improving the generalization, reliability and interpretability of deep learning model for EEW. Here, a Data‐knowledge driven Hybrid deep Learning network (DHLnet) for EEW is proposed and demonstrate that compared with knowledge‐driven methods and data‐driven deep learning models, DHLnet has better performance on predicting epicentral distance, magnitude, PGA and PGV. Key Points: A Data‐knowledge driven Hybrid deep Learning network (DHLnet) is proposed for earthquake early warning (EEW)The DHLnet mainly consists of the knowledge embedding, convolutional neural network and graph convolutional networkWe demonstrate that DHLnet outperforms knowledge‐driven EEW methods and data‐driven deep learning models [ABSTRACT FROM AUTHOR]

Recent geophysical studies have highlighted the potential utility of integrating both seismic and infrasound data to improve source characterization and event discrimination efforts. However, the influence of each of these data types within an integrated framework is not yet well‐understood by the geophysical community. To help elucidate the role of each data type within a merged structure, we develop a neural network which fuses seismic and infrasound array data via a gated multimodal unit for earthquake‐explosion discrimination within the Korean Peninsula. Model performance is compared before and after adding the infrasound branch. We find that the seismoacoustic model outperforms the seismic model, with the majority of the improvements stemming from the explosions class. The influence of infrasound is quantified by analyzing gated multimodal activations. Results indicate that the model relies comparatively more on the infrasound branch to correct seismic predictions. Plain Language Summary: Earthquakes and explosions can produce energy that travel as waves through the ground, seismic, and the air, infrasound. As these waves travel to the station where they are detected, they can be changed so drastically by the medium that it makes it difficult to determine what caused them. In these instances, it has been shown that using both seismic and infrasound data works better to characterize an event than using them independent of one another. However, due to the differences in how the air and ground influence the movement of energy, it is not well‐known how these types of data work in unison to give us more information about an event. In this study, we use a machine learning model trained on both seismic and infrasound data to help us better understand how they can be used together to determine their source. Key Points: Discrimination performance within the Korean Peninsula is improved after fusing seismoacoustic data within a deep learning architectureNeural network framework provides insight into how information in multimodal data combine to distinguish between different event types [ABSTRACT FROM AUTHOR]

The application of deep‐learning‐based seismic phase pickers has surged in recent years. However, the efficacy of these models when applied to monitoring volcano seismicity has yet to be fully evaluated. Here, we first compile a data set of seismic waveforms from various volcanoes globally. We then show that the performances of two widely used deep‐learning pickers deteriorate systematically as the earthquakes' frequency content decreases. Therefore, the performances are especially poor for long‐period earthquakes often associated with fluid/magma movement. Subsequently, we train new models which perform significantly better, including when tested on two data sets where no training data were used: volcanic earthquakes along the Cascadia subduction zone and tectonic low‐frequency earthquakes along the Nankai Trough. Our model/workflow can be applied to improve monitoring of volcano seismicity globally while our compiled data set can be used to benchmark future methods for characterizing volcano seismicity, especially long‐period earthquakes which are difficult to monitor. Plain Language Summary: Earthquake activity at volcanic regions is often monitored to indicate volcanic activity. Identifying the time when the energy radiated from an earthquake source arrives at a seismometer is essential for locating the earthquake, which can be difficult for volcanic earthquakes because of high noise levels, high event rates, and obscured onsets. Previous studies have demonstrated that deep learning can excel in picking the arrival times of regular earthquakes. However, it is unclear how sensitive these detectors are to earthquakes in volcanic regions. Here, we first compile a data set of earthquakes from various volcanoes globally. We then show that existing deep‐learning‐based detectors can miss a large fraction of these earthquakes, especially those without an abrupt change in signal amplitude. We then provide two new models which can better detect volcanic earthquakes than existing models. Our model/workflow can be applied to improve monitoring of volcanic earthquakes globally. Key Points: We compile a data set of seismic waveforms from various volcanic regions globallyWe show that existing deep‐learning phase pickers' performances deteriorate with decreasing earthquake frequency contentOur retrained models perform better and are more generalizable for monitoring volcano seismicity, especially long‐period earthquakes [ABSTRACT FROM AUTHOR]

Applying machine learning to continuous acoustic emissions, signals previously deemed noise, from laboratory faults and slowly slipping subduction‐zone faults, demonstrates hidden signatures are emitted that describe physical details, including fault displacement and friction. However, no evidence currently exists to demonstrate that similar hidden signals occur during seismogenic stick‐slip on earthquake faults—the damaging earthquakes of most societal interest. We show that continuous seismic emissions emitted during the 2018 multi‐month caldera collapse sequence at the Kı̄lauea volcano in Hawai'i contain hidden signatures characterizing the earthquake cycle. Multi‐spectral data features extracted from 30 s intervals of the continuous seismic emission are used to train a gradient boosted tree regression model to predict the GNSS‐derived contemporaneous surface displacement and time‐to‐failure of the upcoming collapse event. This striking result suggests that at least some faults emit such signals and provide a potential path to characterizing the instantaneous and future behavior of earthquake faults. Plain Language Summary: Applications of machine learning have revealed that continuous acoustic emissions from laboratory earthquake experiments contain continuous, hidden signatures that describe the fault slip. In these laboratory studies, the acoustic emissions signals that were previously deemed to be dominantly noise, are found to be rich with details that can describe physical properties such as the fault displacement, friction, and fault thickness. By applying similar machine learning approaches, it was discovered that signatures of surface displacement exist in the seismic emissions from slowly slipping subduction zone faults. However, there has yet to be similar evidence observed during seismogenic stick‐slip on earthquake faults–the damaging earthquakes of most societal interest. Here we study a repeating caldera collapse sequence with short enough repeat times to mimic the experiments performed in the laboratory. We find that seismic emissions from seismogenic fault slip associated with the 2018 caldera collapse at the Kı̄lauea volcano in Hawai'i, also contain hidden signatures informing of instantaneous surface displacement associated with fault slip at depth, as well as time‐to‐failure of the upcoming slip event. These observations suggest that at least some seismogenic faults emit such signals, and provide a potential path to characterizing the instantaneous and future behavior of earthquake faults. Key Points: Features extracted from 30 s of continuous seismic waveforms contain information about contemporaneous GNSS ground displacementsA gradient boosted tree model predicts GNSS ground displacements and timing of the next caldera collapse eventThe observations suggest some seismogenic faults emit precursory signals, providing a potential path to characterize behavior of faults [ABSTRACT FROM AUTHOR]

Estimating an earthquake's magnitude and location may not be necessary to predict shaking in real time; instead, wavefield‐based approaches predict shaking with few assumptions about the seismic source. Here, we introduce GRAph Prediction of Earthquake Shaking (GRAPES), a deep learning model trained to characterize and propagate earthquake shaking across a seismic network. We show that GRAPES' internal activations, which we call "seismic vectors", correspond to the arrival of distinct seismic phases. GRAPES builds upon recent deep learning models applied to earthquake early warning by allowing for continuous ground motion prediction with seismic networks of all sizes. While trained on earthquakes recorded in Japan, we show that GRAPES, without modification, outperforms the ShakeAlert earthquake early warning system on the 2019 M7.1 Ridgecrest, CA earthquake. Plain Language Summary: Have you ever heard something through the grapevine? It often takes you by surprise to hear a message from someone other than the original source. You might have felt an earthquake in a similar way: experiencing shaking (the message) at your location rather than movement along a fault (the source). We apply grapevine‐style communication to earthquake early warning (EEW). The goal of EEW is to warn people to prepare for earthquake shaking before damaging seismic waves arrive at their location. We build on recent work that used deep learning and large earthquake data sets to predict earthquake shaking. We developed a deep learning algorithm called GRAPES which predicts shaking in a manner similar to a game of seismic telephone: when a seismic sensor detects shaking, it sends a message to its neighboring sensors, warning them to expect shaking soon. These sensors then pass on the message to their more distant neighbors along the grapevine. We show that the messages GRAPES learned to send between sensors are like seismic status updates: "I'm seeing this type of seismic wave right now". We applied GRAPES to the 2019 M7.1 Ridgecrest, CA earthquake and it predicted shaking accurately and quickly. Key Points: A deep learning network trained to predict ground motion learned an internal representation of the seismic wavefieldIndividual neurons within the network activate with the arrival of P waves, S waves, surface waves, coda waves, and ambient noiseWhile trained on earthquakes in Japan, the model generalizes well to predicting ground motions for the 2019 Ridgecrest, CA earthquake [ABSTRACT FROM AUTHOR]

The incompleteness of earthquake catalogs is a well‐known issue caused by our technical limitation in detecting the small‐to very small‐magnitude seismicity falling near or below the background seismic noise. The detection of small‐magnitude events is fundamental for improving our knowledge of geometry and kinematics of seismogenic sources and the spatio‐temporal characteristics of seismicity, thus leading to better models for seismic hazard. Template‐matching (TM) is a powerful technique that, based on similarity measure (cross‐correlation) of seismic waveforms, allows to detect hidden earthquakes that are similar to known events (called templates). The high computational effort often limits such technique to small areas and for short time frames (less than 1 year). In this work, we present the first application of template‐matching at regional scale for the Italian Peninsula, focusing on the Southern Apennines. We use about 3,600 high‐quality events as templates, scanning 6‐year long continuous recordings (2009–2014), at more than 180 stations of the INGV network. About 20,000 new events are found, showing a comparable quality to the template catalog in terms of hypocentral solution, reaching a decrease of the magnitude of completeness of about one unit. To highlight the improved quality of the TM catalog, we report two main examples regarding the Sannio‐Matese area, where TM allowed us to unravel relevant details on the spatio‐temporal distribution of the local seismicity. Plain Language Summary: The natural and man‐made seismic noise is commonly higher than the signal of small‐size earthquakes, which represents the vast majority of the seismic activity. Identifying these hidden events is crucial for gaining a better understanding of fault structures, their location, and overall characteristics. A powerful technique for the retrieval of hidden seismicity is template‐matching, based on the comparison of known earthquakes with the continuous recording of a network of seismic stations. However, due to the high computational demand, the use of template‐matching is usually limited to small areas (e.g., volcanoes) and/or small time periods. In this study, we provide the first application of template‐matching at regional scale for the Italian Peninsula, specifically in the area of Southern Apennines. We analyze 6 years of continuous recording from more than 180 stations. Starting from about 3,600 templates, we detected over 20,000 previously hidden earthquakes with uncertainties on the earthquake location that are low and comparable to those of the initial templates. To demonstrate the potential of such an enriched earthquake catalog for future research, we focused on the Sannio‐Matese region. For the earthquake sequence in the Matese area that occurred between 2013 and 2014, template‐matching revealed a consistent level of background seismicity before the onset of the sequence, whereas no activity was reported in the official catalog. In the nearby Sannio area, we identified several earthquake clusters at unusual depths, indicating a sudden increase in the depth of seismic activity. These observations could significantly contribute to improving the seismic hazard model for the area. Key Points: First application of template‐matching at regional scale in the Southern Apennines (Italy) for the retrieval of hidden seismicityThe study results in 20,000 new detections, a five‐fold increase with respect to the starting catalogFocusing on the Sannio‐Matese area, we show how such a high‐resolution catalog enhances the spatio‐temporal characterization of seismicity [ABSTRACT FROM AUTHOR]

Template matching has proven to be an effective method for seismic event detection, but is biased toward identifying events similar to previously known events, and thus is ineffective at discovering events with non‐matching waveforms (e.g., those dissimilar to existing catalog events). In principle, this limitation can be overcome by cross‐correlating every segment (possible template) of a seismogram with every other segment to identify all similar event pairs, but doing so has been previously considered computationally infeasible for long time series. Here we describe a method, called the 'Matrix Profile' (MP), a "correlate everything with everything" calculation that can be efficiently and scalably computed. The MP returns the maximum value of the correlation coefficient of every sub‐window of continuous data with every other sub‐window, as well as the best‐correlated sub‐window location. Here we show how MP methods can obtain valuable results when applied to months and years of continuous seismic data in both local and global case studies. We find that the MP can identify many new events in Parkfield, California seismicity that are not contained in existing event catalogs and that it can efficiently find clusters of similar earthquakes in global seismic data. Either used by itself, or as a starting point for subsequent template matching calculations, the MP is likely to provide a useful new tool for seismology research. Plain Language Summary: Detecting and cataloging earthquakes through analysis of seismic data—the shapes of seismic waves, recorded by seismometers—is foundational to our understanding of Earth's interior structure and processes, as well as geological hazards such as earthquakes. Methods to improve the efficiency and sensitivity of earthquake detection while maintaining accuracy are critical, as seismic data volumes have grown exponentially in recent decades. Recently, methods that use recorded earthquakes as template patterns to identify in seismic data, have proven capable of detecting several times more earthquakes than traditional methods. However, such methods require knowledge of the template earthquakes ahead of time, and are best at identifying earthquakes whose waveforms are similar to the templates. We present here a new method, called the 'Matrix Profile' (MP), that takes short windows of data from a seismic data stream, and compares them to all other parts of that data stream, identifying parts of the data that are highly similar. The MP effectively identifies earthquakes, even those which are hidden within noise, does not require any templates to be provided upfront, and can be efficiently calculated. We demonstrate its success at detecting earthquakes in different types of seismic data, and provide practical guidelines for applying the method and interpreting the results. Key Points: The matrix profile finds the maximum autocorrelation of every subsequence in a time seriesPeaks of high similarity in the matrix profile can be used to identify seismic events and similar event pairsThe matrix profile method has similar sensitivity to template matching but does not require a priori templates [ABSTRACT FROM AUTHOR]

Myanmar bears a high risk of destructive earthquakes, yet detailed seismicity catalogs are rare. We designed a deep‐learning‐based data processing pipeline and applied it to the data recorded by a large‐aperture (∼400 km) seismic array in central Myanmar to produce a high‐resolution earthquake catalog. We precisely located 1891 earthquakes at shallow (<50 km) depth, a 2‐fold increase compared to the traditional procedures. The new catalog reveals the Kabaw Fault seismicity disappears south of ∼22.8°N, where the deeper (20–40 km) seismicity appears west of the southern Kabaw Fault. Such seismicity contrast along the strike of the Kabaw Fault possibly implies an along‐strike change of deformation responses to the shortening process by the India plate oblique subduction. The middle segment of the Sagaing Fault is likely locked and prone to hosting large earthquakes according to the derived low b‐value. Plain Language Summary: Myanmar is a highly seismically active region, yet fault geometry and activities remain poorly understood because of limited modern seismological investigations. Here, we designed a set of machine‐learning algorithms to detect small earthquakes and determine their locations precisely. The seismic data are recorded by a temporary seismic network deployed in central Myanmar. We obtained twice as many earthquakes as the previous research used the regular procedure. Our improved earthquake data set unveils seismic activity changes along the Kabaw Fault through the changes in earthquake locations, depths, and magnitude‐frequency relations. Kabaw Fault is an import boundary fault in the subduction system of the Indo‐Burma Range. This subtle change was not previously observed but means a significant alternation in deformation style along the subduction strike. Moreover, our improved data set indicates that the Sagaing Fault, the most active fault in Myanmar, is prone to generating large earthquakes in the future. This implication warns the nearby populated cities, like Mandalay, of a significant megaquake threat. Key Points: We detect 1891 shallow earthquakes in Myanmar with a deep‐learning‐empowered pipeline, a 2‐fold increase against the routine procedureN‐S seismicity discrepancy is observed near the Kabaw Fault and may imply different responses to E‐W shortening by the Indian subductionLow b‐value derived from the new catalog on the middle Sagaing Fault indicates a high risk of destructive earthquakes [ABSTRACT FROM AUTHOR]

Distributed Acoustic Sensing (DAS) is becoming a powerful tool for earthquake monitoring, providing continuous strain‐rate records of seismic events along fiber optic cables. However, the use of standard seismological techniques for earthquake source characterization requires the conversion of data in ground motion quantities. In this study we provide a new formulation for far‐field strain radiation emitted by a seismic rupture, which allows to directly analyze DAS data in their native physical quantity. This formulation naturally accounts for the complex directional sensitivity of the fiber to body waves and to the shallow layering beneath the cable. In this domain, we show that the spectral amplitude of the strain integral is related to the Fourier transform of the source time function, and its modeling allows to determine the source parameters. We demonstrate the validity of the technique on two case‐studies, where source parameters are consistent with estimates from standard seismic instruments in magnitude range 2.0–4.3. When analyzing events from a 1‐month DAS survey in Chile, moment‐corner frequency distribution shows scale invariant stress drop estimates, with an average of Δσ = (0.8 ± 0.6) MPa. Analysis of DAS data acquired in the Southern Apennines shows a dominance of the local attenuation that masks the effective corner frequency of the events. After estimating the local attenuation coefficient, we were able to retrieve the corner frequencies for the largest magnitude events in the catalog. Overall, this approach shows the capability of DAS technology to depict the characteristic scales of seismic sources and the released moment. Plain Language Summary: A new formulation for far‐field strain radiation from seismic ruptures is derived, leading to a direct interpretation of Distributed Acoustic Sensing (DAS) data to retrieve source properties (seismic moment and source size), via a spectral modeling. This approach is validated on real data recorded in two different tectonic environments, the Chilean margin and the southern Apennines, in Italy. Despite the unique directional sensitivity and peculiar signal characteristics, we demonstrated the high potential of DAS systems in characterizing the seismic ruptures over different space scales, with accuracy increased by redundancy of information from the very‐high spatial resolution in the recording of seismic waves. Key Points: A theoretical description of strain far field radiation from a seismic rupture is introducedSource parameters were evaluated from strain data for earthquakes in magnitude range 2.0–4.3Distributed Acoustic Sensing allows for investigation of source parameters and site effects with fine spatial resolution [ABSTRACT FROM AUTHOR]

The magnitude‐frequency distribution (MFD) describes the relative proportion of earthquake magnitudes and provides vital information for seismic hazard assessment. The b‐value, derived from the MFD, is commonly used to estimate the probability that a future earthquake will exceed a specified magnitude threshold. Improved MFD and b‐value estimates are of great importance in the central and eastern United States where high volumes of fluid injection have contributed to a significant rise in seismicity over the last decade. In this study, we recalculate the magnitudes of 8,775 events for the 2011 Prague, Oklahoma sequence using a relative magnitude approach that depends only on waveform data to calculate magnitudes. We also compare the distribution of successive magnitude differences to the MFD and show that a combination of the magnitude difference distribution (MDFD) and relative magnitudes yields a reliable estimate of b‐value. Using the MDFD and relative magnitudes, we examine the temporal and spatial variations in the b‐value and show that b‐value ranges between ∼0.6 and 0.85 during the aftershock sequence for at least 5 months after the M 5.7 mainshock, though areas surrounding the northeast part of the sequence experience higher b‐values (0.7–0.85) than the southwestern part of the Meeker‐Prague fault where b‐value is the lowest (0.6–0.7). We also identify a cluster of off‐fault events with the highest b‐values in the catalog (0.85). These new estimates of MFD and b‐value will contribute to understanding of the relations between induced and tectonic earthquake sequences and promote discussion regarding the use of b‐value in induced seismic hazard estimation. Plain Language Summary: The magnitude‐frequency distribution and the b‐value are statistical parameters that describe how earthquake magnitudes are distributed in a catalog. These parameters are vital for hazard estimation of induced seismicity in the central and eastern United States, but they are not well understood in this region due to reasons including, but not limited to, catalog incompleteness, magnitude uncertainty between different magnitude types, and variable injection history. In this study, we reevaluate the previously cataloged magnitude estimates using a relative magnitude method to estimate the temporal and spatial variation in b‐value for the 2011 Prague, Oklahoma sequence. With the relative magnitude method, we reestimate magnitude for 81% of the events in the Prague sequence which are then utilized for b‐value estimation. We demonstrate the presence of consistently low b‐values during the aftershock sequence for at least ∼5 months following the mainshock and identify a trend of decreasing b‐value along the Meeker‐Prague fault as distance from the mainshock increases. These observations allow us to compare spatial and temporal b‐value variations of induced sequences to those observed in tectonic sequences. Key Points: The relative magnitude method allows for recalculation of earthquake magnitude without requiring location or geological dataRelative magnitudes combined with the b+ estimator gives a reliable estimate of b‐valueWe identify patterns of spatial and temporal variations in b‐value that are consistent with tectonic sequences [ABSTRACT FROM AUTHOR]

Seismic wave arrival time measurements form the basis for numerous downstream applications. State‐of‐the‐art approaches for phase picking use deep neural networks to annotate seismograms at each station independently, yet human experts annotate seismic data by examining the whole network jointly. Here, we introduce a general‐purpose network‐wide phase picking algorithm based on a recently developed machine learning paradigm called Neural Operator. Our model, called Phase Neural Operator, leverages the spatio‐temporal contextual information to pick phases simultaneously for any seismic network geometry. This results in superior performance over leading baseline algorithms by detecting many more earthquakes, picking more phase arrivals, while also greatly improving measurement accuracy. Following similar trends being seen across the domains of artificial intelligence, our approach provides but a glimpse of the potential gains from fully‐utilizing the massive seismic data sets being collected worldwide. Plain Language Summary: Earthquake monitoring often involves measuring arrival times of P‐ and S‐waves of earthquakes from continuous seismic data. With the advancement of artificial intelligence, state‐of‐the‐art phase picking methods use deep neural networks to examine seismic data from each station independently; this is in stark contrast to the way that human experts annotate seismic data, in which waveforms from the whole network containing multiple stations are examined simultaneously. With the performance gains of single‐station algorithms approaching saturation, it is clear that meaningful future advances will require algorithms that can naturally examine data for entire networks at once. Here we introduce a multi‐station phase picking algorithm based on a recently developed machine learning paradigm called Neural Operator. Our algorithm, called Phase Neural Operator, leverages the spatial‐temporal information of earthquake signals from an input seismic network with arbitrary geometry. This results in superior performance over leading baseline algorithms by detecting many more earthquakes, picking many more seismic wave arrivals, yet also greatly improving measurement accuracy. Key Points: We introduce a multi‐station phase picking algorithm, Phase Neural Operator (PhaseNO), that is based on a new machine learning paradigm called Neural OperatorPhaseNO can use data from any number of stations arranged in any arbitrary geometry to pick phases across the input stations simultaneouslyBy leveraging the spatial and temporal contextual information, PhaseNO achieves superior performance over leading baseline algorithms [ABSTRACT FROM AUTHOR]

Earthquake early warning (EEW) systems aim to forecast the shaking intensity rapidly after an earthquake occurs and send warnings to affected areas before the onset of strong shaking. The system relies on rapid and accurate estimation of earthquake source parameters. However, it is known that source estimation for large ruptures in real‐time is challenging, and it often leads to magnitude underestimation. In a previous study, we showed that machine learning, HR‐GNSS, and realistic rupture synthetics can be used to reliably predict earthquake magnitude. This model, called Machine‐Learning Assessed Rapid Geodetic Earthquake model (M‐LARGE), can rapidly forecast large earthquake magnitudes with an accuracy of 99%. Here, we expand M‐LARGE to predict centroid location and fault size, enabling the construction of the fault rupture extent for forecasting shaking intensity using existing ground motion models. We test our model in the Chilean Subduction Zone with thousands of simulated and five real large earthquakes. The result achieves an average warning time of 40.5 s for shaking intensity MMI4+, surpassing the 34 s obtained by a similar GNSS EEW model. Our approach addresses a critical gap in existing EEW systems for large earthquakes by demonstrating real‐time fault tracking feasibility without saturation issues. This capability leads to timely and accurate ground motion forecasts and can support other methods, enhancing the overall effectiveness of EEW systems. Additionally, the ability to predict source parameters for real Chilean earthquakes implies that synthetic data, governed by our understanding of earthquake scaling, is consistent with the actual rupture processes. Plain Language Summary: Earthquake Early Warning (EEW) systems can evaluate the shaking impact soon after an earthquake occurs and send alerts to areas where strong shaking has not yet arrived. However, predicting shaking intensity for large earthquakes face challenges such as magnitude underestimation and rupture complexity, leading to inaccurate prediction. In our previous study, we proposed a machine learning model, called M‐LARGE, which can predict the magnitude of large earthquakes accurately and rapidly by utilizing surface deformation patterns recorded by HR‐GNSS. Here, we expand the M‐LARGE model to also predict the source location and fault size. This capability allows us to construct a fault, enabling forecasts of strong shaking 35–40 s before the shaking onset, longer than other similar methods. Our approach has the potential to improve the accuracy of EEW and reduce the impact of seismic hazards. Because the model is fully trained with simulated data, the successful prediction of real Chilean earthquakes indicates that our assumptions in simulating large earthquakes are sufficiently consistent with the rupture dynamics of actual events. Key Points: We train a ML model that utilizes HR‐GNSS data to predict source parameters and shaking intensity for large (M7.0+) earthquakesThe model yields a long median warning time of 35–40 s for MMI 4+ when tested on synthetic and real HR‐GNSS recordsOur model fills a critical gap in current EEW systems for large events and can operate in parallel with other methods for better warnings [ABSTRACT FROM AUTHOR]

Elevated seismic noise for moderate‐size earthquakes recorded at teleseismic distances has limited our ability to see their complexity. We develop a machine‐learning‐based algorithm to separate noise and earthquake signals that overlap in frequency. The multi‐task encoder‐decoder model is built around a kernel pre‐trained on local (e.g., short distances) earthquake data (Yin et al., 2022, https://doi.org/10.1093/gji/ggac290) and is modified by continued learning with high‐quality teleseismic data. We denoise teleseismic P waves of deep Mw5.0+ earthquakes and use the clean P waves to estimate source characteristics with reduced uncertainties of these understudied earthquakes. We find a scaling of moment and duration to be M0 ≃ τ4, and a resulting strong scaling of stress drop and radiated energy with magnitude (Δσ≃M00.21 ${\Delta }\sigma \simeq {M}_{0}^{0.21}$ and ER≃M01.24 ${E}_{R}\simeq {M}_{0}^{1.24}$). The median radiation efficiency is 5%, a low value compared to crustal earthquakes. Overall, we show that deep earthquakes have weak rupture directivity and few subevents, suggesting a simple model of a circular crack with radial rupture propagation is appropriate. When accounting for their respective scaling with earthquake size, we find no systematic depth variations of duration, stress drop, or radiated energy within the 100–700 km depth range. Our study supports the findings of Poli and Prieto (2016, https://doi.org/10.1002/2016jb013521) with a doubled amount of earthquakes investigated and with earthquakes of lower magnitudes. Plain Language Summary: The vibration of the Earth's ground recorded at seismometers carries the seismic signatures of distant earthquakes superimposed to the Earth's natural or anthropogenic noise surrounding the seismic station. We use artificial intelligence technology to separate the weak signals of distant earthquakes from other sources of ground vibrations unrelated to the earthquakes. The separated signal provides new insights into earthquakes, especially those within the Earth's deep interior, most of which have not been investigated due to noise levels. In contrast with shallow earthquakes, deep earthquakes are less efficient at radiating energy, though they exhibit a higher rate of increase in both stress drop and radiated energy as they grow. This may suggest that deep earthquakes tend to be more confined fault surfaces. A dual mechanism between nucleation in the subduction‐zone core and propagation of larger events in the dry mantle explains our observations. Key Points: A neural network is used to double the number of earthquakes for source analyses compared to previous studies by improving the data qualityDenoising teleseismic waves improves the source signature in Mw5.0–6.0 events and reduces uncertainties over all magnitudesLarge deep earthquake ruptures are dissipative and compact relative to crustal earthquakes [ABSTRACT FROM AUTHOR]

Seismic activity in the Northern Canadian Cordillera is characterized by diffuse earthquakes that extend hundreds of km northwest from the Yakutat collision zone. We use 25 months of broadband seismic data from Mackenzie Mountain Earthscope Project (MMEP), USArray Transportable Array (TA), and permanent Canadian National Seismic Network stations to present a local earthquake catalog with high sensitivity to small regional events. Deep learning techniques are adopted for both seismic phase detection and association. Event relocations are performed to provide well constrained estimates of earthquake depth distributions. Clusters of seismicity spanning the upper crust are located in the central Richardson Mountains, along the Tintina fault, and in the northeast Selwyn Basin. These clusters suggest that the core of the Richardson Anticlinorium is tectonically active and that the Tintina fault is a locus for low levels of active deformation. We interpret seismicity in the northeast Selwyn Basin as primarily occurring in the hanging wall of the Plateau thrust fault and suggest that some combination of localized duplex structures and lithological strength contrasts both within the Selwyn Basin and between abutting Paleozoic shelf sequences may be responsible for seismicity in the Mackenzie Mountain foreland. Plain Language Summary: The Northern Canadian Cordillera is situated inboard of a transition from a right‐lateral tectonic regime in coastal southeast Alaska to a subduction setting beneath southern Alaska. It presents a unique case study to examine the length scales over which tectonic deformation is able to permeate into a continental interior. In this study we use data from a relatively dense regional network of seismometers in order to present the most spatially complete catalog of earthquakes in the Northern Canadian Cordillera to date. We use machine learning techniques for the detection and classification of small magnitude earthquakes and use our results to analyze regional tectonics. Active faults that host earthquakes are identified and we use the earthquake depth distributions to further distinguish between mechanisms for regional stress transfer. We find evidence that the Tintina fault is presently active along a confined area at a low rate. We also find that low magnitude earthquakes do not generally occur within the Mackenzie Mountain Fold‐Thrust belt but are prevalent within the foreland Selwyn Basin. We also detect a pattern of relatively deep (∼30 km) earthquakes in the northern Richardson Mountains. Key Points: Small magnitude earthquakes extending to 25 km depth are detected throughout the Selwyn Basin in the Mackenzie Mountain forelandThe Tintina fault hosts a range of low magnitude earthquakes in the uppermost 15 km of crust within a 400 km zone along strikeDeformation in the Richardson Mountains is localized onto a narrow band of faults centered on the Richardson Anticlinorium [ABSTRACT FROM AUTHOR]

We investigate spatiotemporal changes of intermediate‐depth earthquakes in the double seismic zone beneath Central and Northeastern Japan before and after the 2011 magnitude 9 Tohoku earthquake. We build a template‐matching catalog 1 year before and 1 year after the Tohoku earthquake using Hi‐net recordings. The new catalog has a six‐fold increase in earthquakes compared to the Japan Meteorological Agency catalog. Our results show no significant change in the intermediate‐depth earthquake rate prior to the Tohoku earthquake, but a clear increase in both planes following the Tohoku earthquake. The regions with increased intermediate‐depth earthquake activity and the post‐seismic slips following the Tohoku earthquake are spatially separate and complementary with each other. Aftershock productivity of intermediate‐depth earthquakes increased in both planes following the Tohoku earthquake. Overall, aftershock productivity of the upper plane is higher than the lower plane, likely indicating that stress environments and physical mechanisms of intermediate‐depth earthquakes in the two planes are distinct. Plain Language Summary: Intermediate‐depth earthquakes occur at 70–350 km depth below the Earth's surface. Because of the high pressure and temperature at such depths, the physical mechanism of intermediate‐depth earthquakes is still poorly understood. In this study, we try to obtain insights into this problem by investigating the behaviors of intermediate‐depth earthquakes before and after the 2011 magnitude 9 Tohoku earthquake. We build an intermediate‐depth earthquake catalog in Central and Northeastern Japan around the Tohoku earthquake. This new catalog contains six times more earthquakes and is more complete than the standard catalog used in previous studies. After analyzing the earthquakes from the new catalog, we find no significant precursory acceleration of intermediate‐depth earthquake activities prior to the Tohoku earthquake. However, we observe a clear increase in intermediate‐depth earthquake rate following the Tohoku earthquake. We also observe that the aftershock productivity of intermediate‐depth earthquakes increased following the Tohoku earthquake. The subduction slab beneath our studied area is characterized by a double seismic zone, and we find that aftershock productivity in the upper plane of seismicity is higher than that in the lower plane. This phenomenon may be due to differences in the environments or physical mechanisms in the two planes of intermediate‐depth earthquakes. Key Points: We build a new catalog of intermediate‐depth earthquakes in Japan 1 year before and 1 year after the 2011 M9 Tohoku earthquakeThere is no significant increase in intermediate‐depth earthquake activities prior to the Tohoku earthquake but a clear increase after itAftershock productivities of double seismic zones increase after the Tohoku earthquake, and it in upper plane is higher than in lower plane [ABSTRACT FROM AUTHOR]

We introduce the Ensemble Earthquake Early Warning System (E3WS), a set of Machine Learning (ML) algorithms designed to detect, locate, and estimate the magnitude of an earthquake starting from 3 s of P‐waves recorded by a single station. The system is made of six Ensemble ML algorithms trained on attributes computed from ground acceleration time series in the temporal, spectral, and cepstral domains. The training set comprises data sets from Peru, Chile, Japan, and the STEAD global data set. E3WS consists of three sequential stages: detection, P‐phase picking, and source characterization. The latter involves magnitude, epicentral distance, depth, and back azimuth estimation. E3WS achieves an overall success rate in the discrimination between earthquakes and noise of 99.9%, with no false positive (noise mis‐classified as earthquakes) and very few false negatives (earthquakes mis‐classified as noise). All false negatives correspond to M ≤ 4.3 earthquakes, which are unlikely to cause any damage. For P‐phase picking, the Mean Absolute Error is 0.14 s, small enough for earthquake early warning purposes. For source characterization, the E3WS estimates are virtually unbiased, have better accuracy for magnitude estimation than existing single‐station algorithms, and slightly better accuracy for earthquake location. By updating estimates every second, the approach gives time‐dependent magnitude estimates that follow the earthquake source time function. E3WS gives faster estimates than present alert systems relying on multiple stations, providing additional valuable seconds for potential protective actions. Plain Language Summary: We designed E3WS, an Earthquake Early Warning System (EEWS) using the first 3 s of seismic waves recorded by a single station. E3WS is the first EEWS entirely made of Artificial Intelligence algorithms. Our results show an improvement in magnitude estimates compared to existing single‐station based EEWS, such as Urgent Earthquake Detection and Alarm System and ElarmS, and slightly better localization. In order to track the magnitude of M > 7.0 earthquakes, whose source duration is typically longer than 6 s, we test E3WS on large earthquakes in real‐time scenarios. E3WS provides time‐dependent magnitude estimates that follow the earthquake source time functions. E3WS is highly exportable. Thanks to its simplicity, it can be installed in 32‐bit Single Board Computers like a Raspberry Pi 4, and of course in more complex 64‐bit processors. E3WS is not only theoretical, it is already running in alpha test mode as the core of the EEWS of Peru in construction. We believe that E3WS will provide valuable seconds to populations living at constant risk, giving the user extra seconds that may give them enough time to "drop, cover, and hold on" or to perform mitigation actions like stopping traffic, stopping elevators or evacuating the ground floor of buildings. Key Points: Earthquake Early Warning System (EEWS) with faster and robust preliminary estimates of magnitude and location to generate a first alertFirst EEWS designed purely with Artificial IntelligenceEarthquake Early Warning with continuous updates that follow the earthquake source time function [ABSTRACT FROM AUTHOR]

An approximate 236‐year interval in major seismic activity near the southern Xianshuihe fault terminated on 5 September 2022 until a destructive M6.8 Luding earthquake. The strike‐slip mainshock, accompanied by potent normal‐faulting aftershocks, fell short of the previously anticipated M7+ event. Utilizing deep‐learning and source inversion techniques, we analyzed dense near‐field seismograms to examine detailed fault structures and rupture characteristics. Our high‐quality relocated catalog with 7,388 events revealed that the earthquake ruptured an unmapped multiscale fault network. The analysis of the mainshock and 43 ML ≥ 2.8 aftershocks suggests normal‐faulting events relate to the vertical movement of the Gongga Mountain. The sequence length, seismic gaps, and asperities derived from seismicity and waveform modeling imply that the rupture of the M6.8 quake was incomplete, indicating a high risk of a future M6+ event at the northern Moxi segment. These findings hold crucial implications for assessing future seismic hazards in this region. Plain Language Summary: On 5 September 2022, a destructive M6.8 earthquake struck the southern Xianshuihe fault, Luding, China. This event ended the 236‐year‐long silence in the area, resulting in 93 deaths and severe economic damage. This strike‐slip mainshock, accompanied by unusual normal‐faulting aftershocks, fell into earlier predictions of an approximate 248‐year cycle of M7+ earthquakes in this area. However, its magnitude was less than anticipated, leading us to question whether another significant earthquake could potentially occur. To investigate this, we used a novel deep‐learning method named DiTingPicker, which was trained with ∼5 million seismograms, and waveform modeling techniques to examine the detailed fault structures and mainshock rupture characteristics using dense local seismic stations. We compiled a 415‐day‐long earthquake catalog, showing this sequence ruptured an unmapped multiscale fault network of interlaced northwest‐trending and northeast‐trending faults. Our study suggests the normal‐faulting aftershocks are related to the vertical movement of the Gongga Mountain. By comparing the rupture lengths of the historical 1786 M7.8 earthquake and the 2022 event, our research suggests a potential risk for another M6+ earthquake at the unruptured northern Moxi segment. Key Points: A 415‐day seismic catalog that characterizes fine‐scale fault structures and seismic gaps was compiled using deep learningA hidden developing fault network was revealed and distinct normal‐fault aftershocks were spotted west of the strike‐slip mainshockSeismicity and source inversion suggest the M6.8 quake's rupture was incomplete and a future M6+ event at the unruptured Moxi segment [ABSTRACT FROM AUTHOR]

Using P‐wave seismograms, we trained a seismic source classifier using a Convolutional Neural Network. We trained for three classes: earthquake P‐wave, underground nuclear explosion (UNE) P‐wave, and noise. With the current absence of nuclear testing by countries that have signed the Comprehensive Test Ban Treaty, high quality seismic data from UNEs is limited. Even with limited training data, our model can accurately characterize most events recorded at regional and teleseismic distances, finding over 95% signals in the validation set. We applied the model on holdout datasets of the North Korean test explosions to evaluate the performance on unique region and station‐source pairs, with promising results. Additionally, we tested on the Source Physics Experiment events to investigate the potential for chemical explosions to act as a surrogate for nuclear explosions. We anticipate that machine‐learning models like our classifier system can have broad application for other seismic signals including volcanic and non‐volcanic tremor, anomalous earthquakes, ice‐quakes or landslide‐quakes. Plain Language Summary: We train a global seismic event classifier using machine learning on underground nuclear test explosion seismic data. Our classifier model can successfully discriminate (with over 95% accuracy) between underground nuclear explosion, earthquake, and noise signals from stations both regionally and far‐field. Since this model was trained on a relatively small data set (for machine learning applications) we expect that similar methods can be applied to event or discrimination of other unique seismic sources like those from volcanoes, landslides, or glaciers. Key Points: We successfully discriminate underground nuclear explosions with a Convolutional Neural network (CNN) trained on P‐wave seismogramsRobust global seismic event discrimination is possible with machine learning trained on regional and teleseismic dataA CNN trained with historical nuclear explosion data can be applied with high accuracy to other regions, like the six Democratic People's Republic of Korea's test explosions [ABSTRACT FROM AUTHOR]

Seismology is witnessing explosive growth in the diversity and scale of earthquake catalogs. A key motivation for this community effort is that more data should translate into better earthquake forecasts. Such improvements are yet to be seen. Here, we introduce the Recurrent Earthquake foreCAST (RECAST), a deep‐learning model based on recent developments in neural temporal point processes. The model enables access to a greater volume and diversity of earthquake observations, overcoming the theoretical and computational limitations of traditional approaches. We benchmark against a temporal Epidemic Type Aftershock Sequence model. Tests on synthetic data suggest that with a modest‐sized data set, RECAST accurately models earthquake‐like point processes directly from cataloged data. Tests on earthquake catalogs in Southern California indicate improved fit and forecast accuracy compared to our benchmark when the training set is sufficiently long (>104 events). The basic components in RECAST add flexibility and scalability for earthquake forecasting without sacrificing performance. Plain Language Summary: We explore the potential for deep learning in earthquake forecasting. Prior work has relied heavily on statistical models that do not scale to fully utilize the currently available large earthquake data sets. Here we build on recent developments in deep learning for forecasting event sequences in general to create an implementation for earthquake data. The new approach allows us to incorporate larger data sets, potentially with more information about each earthquake. We also avoid a specific functional form, so the method naturally adapts to additional information about events, like magnitude or variations in behavior over time. As we add more data, results show continued improvements. This ability to incorporate and improve continually as training data sets increase indicates that there is more information in the earthquake catalogs than has yet been used for earthquake forecasting. Key Points: We introduce a deep learning model for earthquake forecasting and explore its performance on synthetic and regional earthquake data setsIt is flexible in the sense that a predefined functional form is not requiredIt is scalable in two senses: it is efficient on large data sets, and its performance relative to benchmarks improves with more training data [ABSTRACT FROM AUTHOR]

Following laboratory experiments and friction theory, slow slip events and seismicity rate accelerations observed before mainshocks are sometimes interpreted as evidence of a nucleation phase. However, such precursory observations still remain scarce and are associated with different time and length scales, raising doubts about their actual preparatory nature. We study the 2017 Valparaiso Mw = 6.9 earthquake, which was preceded by aseismic slip accompanied by an intense seismicity, suspected to reflect its nucleation phase. We complement previous observations, which have focused only on precursory activity, with a continuous investigation of seismic and aseismic processes from the foreshock sequence to the post‐mainshock phase. By building a high‐resolution earthquake catalog and searching for anomalous seismicity rate increases compared to aftershock triggering models, we highlight an over‐productive seismicity starting within the foreshock sequence and persisting several days after the mainshock. Using repeating earthquakes and high‐rate GPS observations, we highlight a transient aseismic perturbation starting 1‐day before the first foreshock and continuing after the mainshock. The estimated slip rate over time is lightly impacted by large magnitude earthquakes and does not accelerate toward the mainshock. Therefore, the unusual seismic and aseismic activity observed during the 2017 Valparaiso sequence might be interpreted as the result of a slow slip event starting before the mainshock and continuing beyond it. Rather than pointing to a possible nucleation phase of the 2017 Valparaiso mainshock, the identified slow slip event acts as an aseismic loading of nearby faults, increasing the seismic activity, and thus the likelihood of a large rupture. Plain Language Summary: Both laboratory experiments and friction theory show that earthquakes do not begin abruptly but are preceded by an accelerating slip associated with a seismicity increase. On the field, however, such precursory observations still remain scarce and are associated with different characteristic time and length scales, raising doubts that they actually reflect the same nucleation phenomena. We study the 2017 Valparaiso M = 6.9 earthquake, which was preceded by both a slow slip and an intense seismicity suspected to reflect such nucleation phase. We complement previous studies, that have focused only on precursory activity, with a continuous investigation of seismic and slow slip before and after the mainshock. Using refined earthquake detection tools, we highlight a seismicity excess starting before and persisting several days after the mainshock. Using repeating earthquakes and high‐resolution GPS, we show that the slow slip does not accelerate toward the mainshock, but continues after it. Therefore, rather than pointing to a possible accelerating nucleation phase of the Valparaiso mainshock, we suggest that the slow slip drives an enhanced seismic activity that is not mainshock‐directed. Within such slow‐slip driven seismicity, the probability of triggering a large earthquake (subsequently considered as the mainshock) is increased. Key Points: We use a high resolution seismic catalog and GPS to investigate seismic and aseismic process before and after the Valparaiso mainshockAn unusually high seismicity and a slow slip event is continuously observed from the foreshock sequence up to days after the mainshockRather than a nucleation phase of the mainshock, the slow slip event acts as an aseismic loading of nearby faults during the entire sequence [ABSTRACT FROM AUTHOR]

A high‐precision earthquake catalog is made from continuous waveform data recorded at 360 portable seismographs deployed for one month in the Changbaishan‐Tianchi volcanic area based on an automatic earthquake detection and location workflow. Then we investigate detailed 3‐D structures of P and S wave velocities (Vp and Vs) and Vp/Vs ratio of the crust beneath the study region. Our results reveal a high‐Vp/Vs zone accompanied with active volcanic earthquakes beneath the Tianchi caldera at ∼5 km depth, which may reflect a magma chamber. The intense seismic activities within 2 km depth beneath the Tianchi caldera from 2002 to 2005 occurred directly above the magma chamber revealed by this study, which could be related to upward migration of high‐pressure volatiles originating from the magma chamber. A potential middle‐crustal magma chamber might provide continuous magmatic materials feeding the upper‐crust magma chamber and lead to ruptures of surrounding rocks, resulting in two swarms of volcanic seismicity during the observation period. We deem that the diverse eruption styles of the Changbaishan volcano are closely associated with the multilevel structure of magma chambers in the upper and middle crust. Plain Language Summary: It is difficult for the earthquake monitoring system to identify and locate microearthquakes by using traditional seismic detection methods due to their weak signals, leading to a relatively low quantity and quality of seismic data in a study region, which cannot meet the demands of high‐resolution seismic tomography. In this study, we detect and locate microseismic events beneath the Changbaishan‐Tianchi volcanic area using 1‐month waveform data recorded at our 360 portable seismographs deployed in the volcanic area. Then we investigate high‐resolution 3‐D crustal structures of P and S wave velocities (Vp, Vs) and Vp/Vs ratio beneath the study area. Our results reveal a possible crustal magma reservoir at ∼5 km depth beneath the Tianchi caldera, which is accompanied with volcanic earthquakes during our observation period. In addition, small earthquakes occurred actively at a depth of ∼2 km during 2002–2005, which are located right above the magma reservoir. We deem that the joint effect of magma reservoirs in the upper and middle crust could explain the active volcanic seismicity and distinctive eruption behaviors of the Changbaishan volcano. Key Points: A magma chamber at ∼5 km depth beneath the Tianchi caldera of the Changbaishan volcano (CBV) is revealedThe CBV unrest during 2002–2005 was related to reactivation of the crustal magma chamberDistinctive eruption behaviors of the volcano are related to multilevel magma chambers in the crust [ABSTRACT FROM AUTHOR]

Oceanic transform faults connect spreading centers and are imprinted with previous tectonic events. However, their tectonic interactions are not well understood due to limited observations. The Discovery transform fault system at 4°S, East Pacific Rise (EPR), represents a young transform system, offering a unique opportunity to study the interplay between faulting and other tectonic events at an early phases of an oceanic transform system. Discovery regularly hosts M5–6 characteristic earthquakes, and the seafloor north of Discovery includes a 35 km‐long rift zone that records a complex history of rifting, faulting and volcanism, suggesting that the transform faults likely interact with regional tectonic activity. We apply a machine‐learning enabled workflow to locate 21,391 earthquakes recorded during a 1‐year ocean bottom seismometer experiment in 2008. Our results indicate that seismicity on the western Discovery fault is separated into seven patches with distinct aseismic and seismic slip modes. Additionally, we observe a patch of off‐fault seismicity near where seafloor abyssal hills intersect the rift zone. This seismicity may have been caused by varying opening rates as spreading rate decreases from north to south in the rift zone. Our findings suggest that the Discovery system is still evolving, and that system equilibrium has not been reached between rifting and faulting. These results reflect the complex yet rarely observed interactions between fault slip, plate rotation, and rifting which are likely ubiquitous at oceanic transform systems. Plain Language Summary: Oceanic transform faults are major plate boundaries connecting mid‐ocean ridges. Despite their important role in plate tectonics, their interactions with adjacent mid‐ocean ridges and surrounding oceanic plates are not well understood. The Discovery transform fault system at 4°S, East Pacific Rise, is a young oceanic transform system formed approximately 1 My ago, offering a unique opportunity to study the interplay between faulting and other tectonic events at an early phase of an OTF. Discovery faults have quasi‐periodical magnitude (M) 5–6 earthquakes. Using ocean bottom seismometer data recorded over 1 year, we find that seismicity of the western Discovery fault can be grouped into seven patches, indicating division of alternating slip modes that either releases tectonic strain by M > 5 earthquakes or creep steadily. North of the western Discovery fault, a ∼10 km wide rift zone, abundant seamounts, and abyssal hills form an interactive tectonic complex. We observe a patch of off‐fault seismicity coinciding with seafloor abyssal hills near their intersection with the rift zone. This off‐fault seismicity indicates ongoing deformation within the oceanic plate and possible spatial variations in rifting rates. Our results suggest that the Discovery system is still evolving with rifting and faulting accommodating plate spreading simultaneously. Key Points: The western Discovery transform fault has seven patches that are likely dominated by alternating seismic and aseismic slip modesMachine‐learning method helps to identify off‐fault seismicity along abyssal hills, indicating ongoing deformation within the oceanic plateThe Discovery transform system is young and still evolving, forming an interactive system with faulting, rifting, and plate rotation [ABSTRACT FROM AUTHOR]

We deployed a dense geodetic and seismological network in the Atacama seismic gap in Chile. We derive a microseismicity catalog of >30,000 events, time series from 70 GNSS stations, and utilize a transdimensional Bayesian inversion to estimate interplate locking. We identify two highly locked regions of different sizes whose geometries appear to control seismicity patterns. Interface seismicity concentrates beneath the coastline, just downdip of the highest locking. A region with lower locking (27.5°S–27.7°S) coincides with higher seismicity levels, a high number of repeating earthquakes and events extending toward the trench. This area is situated where the Copiapó Ridge is subducted and has shown previous indications of both seismic and aseismic slip, including an earthquake sequence in 2020. While these findings suggest that the structure of the downgoing oceanic plate prescribes patterns of interplate locking and seismicity, we note that the Taltal Ridge further north lacks a similar signature. Plain Language Summary: Deformation along plate boundaries can occur seismically (i.e. through earthquakes) as well as aseismically (i.e. slipping slowly), and it is important to understand where each of these modes is dominant. Along the Chilean subduction contact, North‐Central Chile is the only place where aseismic deformation episodes have been observed so far. In order to study these processes in detail, we deployed and operated dense geodetic and seismological networks in this region. Analyzing the data collected by these networks, we find notable relationships between seismic and aseismic processes. Thousands of small earthquakes are found at the boundaries of locked regions, whereas no small earthquakes are found at their interior. Thus, implying such regions are mechanically coupled, that is, currently accumulating elastic deformation energy that will 1 day be released during a large earthquake. Along the North‐Central Chilean plate boundary, there is one region (around 27.5°S) that shows many signs of aseismic deformation. It is located where a chain of seamounts is being subducted, which is likely responsible for the different behavior of this segment. Key Points: Microseismicity catalog and map of interplate locking derived for the Atacama 1922 seismic gap in North‐Central ChileSeismicity in vicinity of plate interface coincides with downdip edge of high couplingSeismo‐geodetic signals due to the subduction of the Copiapó ridge are prominent but negligible for the subducting Taltal Ridge [ABSTRACT FROM AUTHOR]

Earth's crust is continuously subjected to oscillatory stress perturbations due to the solid Earth and ocean tides. The seismic response to such stress modulations carries information on earthquake physics and crustal properties. Experimental and observational studies suggested but could not demonstrate that the strength of tidal modulation of seismicity increases before large earthquakes. We tested this hypothesis by (a) developing a new, comprehensive 10‐year long earthquake catalog preceding the 6 July 2019 magnitude 7.1 Ridgecrest, CA earthquake, and (b) applied our novel method for extracting a statistical signal of tidal modulation. Our results show enhanced tidal sensitivity of seismicity along the fault starting about 1.5 years before the mainshock, corroborating the hypothesis. This observation suggests that small magnitude earthquakes may be used to gain insight into subtle changes in fault conditions, bringing new promise for studying the earthquake preparation process. Plain Language Summary: The solid Earth experiences tides, like the ocean, it deforms under the gravitational attraction of the Sun and Moon. This deformation induces an oscillatory stress change in the crust with small peak‐to‐peak amplitudes of the order of 1 kPa (about 1% of the atmospheric pressure). Tidal stresses weakly influence the rate of earthquake occurrence and the characteristics of this modulation carry information on earthquake physics and crustal properties. On the basis of experimental and field observation studies, it has been proposed that the modulation of seismicity by the tides increases before a large earthquake. We tested this hypothesis by analyzing 10 years of seismicity before the M7.1 2019 Ridgecrest, CA earthquake. We first built a new, comprehensive earthquake catalog with our automated method and used our novel method to extract the signal of tidal modulation throughout the study period. We found that seismicity became strongly modulated by the tides about 2 years before the mainshock, thus corroborating the hypothesis. Key Points: Seismicity in the Ridgecrest area is modulated by the solid Earth tidesPeak seismicity occurs when tidal Coulomb stress or stress rate favors ruptureStrength of tidal triggering is gradually increasing in the rupture area about 2 years before the Mw 7.1 2019 mainshock [ABSTRACT FROM AUTHOR]

Despite the fact that earthquake occurrence can be strongly influenced by the architecture of pre‐existing faults, it remains challenging to obtain information about the detailed subsurface geometries of active fault systems. Current geophysical methods for studying such systems often fail to resolve geometrical complexities at sufficiently high spatial resolutions. In this work, we present a novel method for imaging the detailed 3D architectures of seismically active faults based on high‐precision hypocenter catalogs, using nearest neighbor learning and principal component analysis. The proposed approach enables to assess variations in fault instabilities and kinematics. We apply the method to the relatively relocated St. Léonard (max. ML = 3.2) and Anzère (max. ML = 3.3) microearthquake sequences in the Southwestern Swiss Alps, revealing strike‐slip fault systems with interconnecting stepovers at depths of 3–7 km and lengths ranging from 0.5 to 2 km. In combination with additional information about fault instabilities and kinematics, we observe significantly reduced earthquake migration velocities and fault locking processes within the stepovers. Understanding such processes and their role in the propagation of strain across stepovers is of great relevance, as these structures can potentially limit earthquake ruptures but also represent possible locations for the nucleation of larger ruptures. Our proposed method is expected to be broadly useful for further applications such as monitoring hydraulic fracture stimulations or geothermal exploration of natural, fluid‐bearing faults. Conducting similar high‐resolution spatiotemporal analyses of microseismic sequences has the potential to greatly enhance our comprehension of how the 3D fault architecture impacts seismogenic fault reactivation. Plain Language Summary: Earthquakes commonly occur on planar geological structures, called faults. Profound knowledge of the presence and geometries of earthquake‐generating faults is crucial to understand the regional earthquake hazards. Especially for regions with distributed and rather small earthquakes, it is however challenging to detect such faults, because of the fact that these faults often do not reach the surface. Here, we present a new method that uses the locations of small earthquakes to resolve the geometries of earthquake‐generating faults at depth. This is only possible due to the recent advances in earthquake location methods, which can reduce the uncertainties of the locations from the kilometers down to some tens of meters for precise measurements. With our method, we image the complex geometries of two fault systems in the Southwestern Swiss Alps at high resolution. The knowledge obtained by applying our method cannot only help to detect previously unknown, potentially hazardous earthquake‐generating faults but might also improve our understanding of earthquake processes in general. Key Points: We present a novel approach that uses relocated hypocenters to image 3D geometries and instabilities of active faultsApplication to two earthquake sequences in the Southwestern Swiss Alps provides detailed insights into these strike‐slip faults systemsThis study documents the dynamics of earthquake‐migration processes across stepover faults at high resolution [ABSTRACT FROM AUTHOR]

By providing unrivaled resolution in both time and space, volcano seismicity helps to chronicle and interpret eruptions. Standard earthquake detection methods are often insufficient as the eruption itself produces continuous seismic waves that obscure earthquake signals. We address this problem by developing an earthquake processing workflow specific to a high‐noise volcanic environment and applying it to the explosive 2008 Okmok Volcano eruption. This process includes applying single‐channel template matching combined with machine‐learning and fingerprint‐based techniques to expand the existing earthquake catalog of the eruption. We detected an order of magnitude more earthquakes, then located, relocated, determined locally calibrated magnitudes, and classified the events in the enhanced catalog. This new high‐resolution earthquake catalog increases the number of observations by about a factor of 10 and enables the detailed spatiotemporal seismic analysis during a large eruption. Plain Language Summary: Volcanic eruptions are some of the most challenging events to study seismically. The eruption itself continually excites seismic waves that hide the signatures of the accompanying earthquakes. The jetting of the plume, explosions, and impact of material on the ground all combine to create a blind spot in the seismological record. In this work, we develop a workflow that can overcome these limitations and reveal events that would otherwise remain hidden in the high‐noise, continuous seismic record. We utilize a combination of machine‐learning and related methods to increase the number of observed events by almost one order of magnitude. We determine the locations of the events using modern, high‐precision methods and use these locations with a customized, calibrated local magnitude scale for the region, which is necessary for volcanoes where seismic waves decay much faster than in the regions where standard magnitude scales were developed. We also automatically separate ordinary earthquakes from the more unusual long‐period ones and see an increase in detectability of both populations. The combined workflow now allows us to illuminate potentially the most interesting, and certainly most obscure, parts of the volcanic seismic cycle. Key Points: We develop a complete workflow suitable for co‐eruptive seismicity, which is inaccessible by normal seismic methodsWe combine different detection algorithms to increase the number of events of an explosive volcanic eruption by almost an order of magnitudeAutomated classification detects both long‐period and ordinary volcanic earthquakes [ABSTRACT FROM AUTHOR]

Earthquakes induced by human activities can impede the development of underground resources. Significant induced events (M5) have caused both economic and human losses. The recent ML 5.6 (MW 5.1) event near Peace River, Alberta occurs in a region of in situ bitumen recovery. We find that 3.4 cm of ground deformation was caused by reverse fault slip (∼29 cm), possibly related to Peace River Arch faulting. Events are located within the shallow basement, nearby to significant wastewater injection into Paleozoic strata. We find a statistical relationship between earthquakes and injection operations. These events were likely related to the in situ bitumen development: dominantly from wastewater disposal induced pore pressure increases, with smaller poroelastic contributions from bitumen recovery. The assessment of this earthquake as induced will likely have implications for future energy development, management, and regulation—including carbon capture and blue hydrogen. Plain Language Summary: Earthquakes can be caused by underground fluid injection; cases of M5 induced events have caused damage and harm. One of the largest recorded earthquakes in Alberta (ML 5.6) occurred in a region of underground oil sand development. Here, ground shaking and deformation information are combined into an interpreted result: that ancient faults were reactivated with reverse slip. The fault slip is largely within the crystalline basement, with a small portion extending into basal sediments. Nearby injection operations dispose of petroleum‐related wastewater in these basal sediments. This earthquake was likely triggered by the injection process: injection increases pore pressure, which diffuses laterally along permeable sediments, until encountering fractured rock, which channelizes flow into the crystalline basement—the increase of pore pressure within the fault continues until reaching a critical point for slip initiation. This event likely being induced will have important implications for future operations. Key Points: On 30 November 2022 one of Alberta's largest recorded events (ML 5.6, MW 5.1) occurred in a region of in situ bitumen recoveryA well oriented fault was reactivated with reverse slip (29 cm), causing up to 3.4 cm of ground deformationThe event was likely induced by pore pressure from disposal, with smaller poroelastic stress changes from bitumen production [ABSTRACT FROM AUTHOR]

We use a dense seismic array of 1,108 vertical‐component geophones within a 600‐m footprint to detect thousands of small earthquakes near an active strand of the San Jacinto fault zone in southern California during a 26‐day period. We first correct site effects using multichannel cross‐correlations of the P‐waves of 256 cataloged earthquakes, and then perform beamforming analysis on the continuous waveforms in a slowness range from −0.4 to 0.4 s/km in both the east and north directions. At each time step, we identify the beam slowness with maximum amplitude and apply a picking algorithm to identify 13,408 events. These detections include over 55.6% of the events in the Quake Template Matching (QTM) catalog for all of southern California during the same time period and 70% of those within 100 km of the array. In addition, we detect over 10,000 new events, not in the QTM catalog. Many of these events can also be seen in records from nearby borehole seismic stations. Measured slownesses for the catalog and newly‐discovered events group into clusters that can be associated with QTM earthquake locations, but with slowness values considerably distorted from predictions based on a 1‐D velocity model, presumably owing to strong velocity heterogeneity near the San Jacinto Fault. Amplitudes of the detected events obey a Gutenberg‐Richter distribution with a b‐value close to one. Foreshocks are common among these detected events, increasing in rate before mainshocks following an inverse Omori's law. Plain Language Summary: We show that an array of over 1,000 seismometers located on a small patch of the San Jacinto Fault in southern California can detect many times more earthquakes than those in standard earthquake catalogs. Although accurate locations cannot be obtained for most of these newly identified events, their amplitudes indicate they obey similar scaling with size as larger magnitude earthquakes. Many of the larger events are preceded by foreshocks and the foreshock rate increases in the time leading up to the mainshocks. Our results suggest that future deployments of small‐aperture seismometer arrays could be used to complement existing seismic networks to probe earthquake activity in great detail. Key Points: We detect over 13,000 seismic events during 26 days in 2014 using a dense nodal array on the San Jacinto fault zone in southern CaliforniaMost of the detections are newly discovered earthquakes not in existing catalogsForeshocks are common among these earthquakes, increasing in the rate before their mainshocks following an inverse Omori's law [ABSTRACT FROM AUTHOR]

We conduct a detailed study of the foreshock sequence preceding the 2010 Mw 6.7 Yushu, Qinghai earthquake in the Tibetan plateau by examining continuous waveforms recorded at a seismic station near the mainshock rupture zone. By using a deep learning phase picker—EQTransformer and a matched‐filter technique, we identify 120 foreshocks with magnitude ranging from −0.7 to 1.6, starting with a Mw 4.6 foreshock approximately 2 hr before the Mw 6.7 Yushu mainshock. Our analyses show that the foreshock sequence follows a typical Omori's law decay with a p‐value of 0.73 and the Gutenberg‐Richer frequency‐magnitude b‐value of 0.66. We do not find any evidence of accelerating events leading up to the Yushu mainshock. Hence, they could be considered as aftershocks of the Mw 4.6 earthquake. We further invert for the focal mechanisms and rupture directions for both the largest foreshock and the mainshock. The Mw 4.6 foreshock likely occurred on a NE‐SW trending fault conjugating to the NW‐SE trending fault of the mainshock. Coulomb stress analysis suggests the Mw 4.6 foreshock induces negative stress on the mainshock source area. These observations do not support either the pre‐slip or the cascade triggering model for foreshock generation. The occurrence of the foreshock, mainshock and large aftershocks appear to be modulated by the Earth's tidal forces, likely reflecting the role of high pore‐fluid pressures. Our observations, together with other recent studies, suggest that extensional step‐overs and conjugate faults along major strike‐slip faults play an important role in generating short‐term foreshock sequences. Plain Language Summary: We use both deep learning and matched‐filter methods to detect foreshocks that occur 2 hr before the magnitude 6.7 Yushu, Qinghai earthquake in Western China on 13 April 2010. The foreshock sequence begins with a magnitude 4.6 foreshock and the activity gradually decays with time. We do not observe an increase of magnitude or number of the events right before the mainshock. Therefore we interpret this as a typical aftershock sequence of the magnitude 4.6 earthquake. We also resolve the fault geometry and rupture direction. The results suggest the magnitude 4.6 earthquake ruptures along a fault that is nearly perpendicular to the fault broken during the magnitude 6.7 mainshock. Due to the fault geometry and relative location, the magnitude 4.6 foreshock might have inhibited the occurrence of the mainshock by reducing the stress on the fault. These observations do not support either the pre‐slip or cascade triggering models for foreshock generation. We calculate the theoretical tidal stress and strain in the study area and observe a correlation of earthquake activity and the tidal stress change. This correlation might provide evidence for fluid driven process. In conclusion our research shows complex geometry along major strike‐slip faults can influence foreshock behaviors. Key Points: We identified missing foreshocks of the 2010 Mw 6.7 Yushu, Qinghai earthquake by using deep learning and matched‐filter techniquesThe foreshock sequence can be considered as aftershocks of Mw 4.6 foreshock given no evidence of an acceleration leading to the mainshockThe Mw 4.6 foreshock likely occurred on an extensional step‐over which induced negative stress change at the Yushu mainshock hypocenter [ABSTRACT FROM AUTHOR]

We use 12 temporary and 9 permanent broadband seismometers that were operating for ∼400 days from October 2018 to November 2019 to generate the first published earthquake catalog and local magnitude function for southern New Caledonia (SNC). Local hypocenters mostly have depths <20 km, but east of the New Hebrides‐Vanuatu subduction zone they deepen to >100 km. Our local magnitude estimates ML for 107 earthquakes in the subduction zone are consistently 1.1 units smaller than Mw and mb over a range of Mw from 4.5 to 7.5, as determined by the United States Geological Survey. Our catalog has 460 earthquakes with Mw≥3.7 in the subduction zone and the largest event in SNC has ML 3.8. Seismicity rates in SNC are low, but ML> 5 earthquakes are 2.7 times more frequent than elsewhere in the northern Australian plate. The probability of an ML> 5 event in 50 years is 0.6 in SNC. The hazard of damaging seismic shaking in SNC is dominated by local moderate‐magnitude earthquakes, rather than large‐magnitude subduction events. The predicted peak ground acceleration for Nouméa at 10% probability of exceedance in 50 years is 0.08 g. Residual analysis of ground accelerations demonstrates that the hazard for Nouméa from subduction zone earthquakes is currently overestimated and that new regionally specific ground motion prediction equations (GMPEs) are needed. Our results highlight the inadequacy of current GMPEs in subduction zone footwall settings and the need for additional studies of this type of setting. Plain Language Summary: We used seismometers from a local network near Nouméa, which is the capital city of New Caledonia, to record earthquakes for 14 months. About 13% of the earthquakes were shallow crustal events (<20 km depth) that clustered near our network. The others clustered ∼250 km away in the New Hebrides‐Vanuatu subduction zone (NHV), where large earthquakes (recorded on global networks) regularly occur. Earthquakes beneath southern New Caledonia (SNC) are smaller and less frequent than in nearby Vanuatu, but they are more hazardous for local communities because they are at shallow depths and nearby. Our work demonstrates the necessity of long‐term monitoring of local earthquakes in SNC to assess hazard there. Our analysis shows that current models used to predict shaking from large earthquakes in the NHV over‐estimate hazard in New Caledonia, and this may have implications for other regions in similar tectonic settings. Key Points: We use 14 months of data from a local network to determine location and magnitude of 1,374 earthquakes near southern New CaledoniaGround motions in New Caledonia, which is in a subduction footwall, are smaller than expected from ground motion prediction equationsEarthquakes in New Caledonia are smaller and less frequent than in the nearby subduction zone, but induce greater local shaking hazard [ABSTRACT FROM AUTHOR]

In this study, the variation in frequency magnitude distribution factor (b‐value) before the main shock has been assessed for the 26 November 2021 Mizoram earthquake (Mw 6.1) in northeast India. The study area covering the Indo‐Burma Ranges is divided into square grids of equal dimension and spatial mapping of the b‐value for each square grid has been carried out. The maximum likelihood method is implemented for the estimation of the b‐value of each grid. The epicentral location of the 26 November 2021 Mizoram earthquake (Mw 6.1) is found to be in the intermediate b‐value square grid. The temporal variation in the b‐value indicates a considerable decline in the b‐value before the occurrence of the main event. The depth‐wise variation in the b‐value suggests an antithetical relationship between the b‐value and crustal stress. The spatio‐temporal and depth‐wise fluctuation in the b‐value results have been strengthened by employing a non‐parametric Kolmogorov–Smirnov (K–S) test. Based on our study, we infer that interplate earthquakes occur mostly in the upper crust of the study region. [ABSTRACT FROM AUTHOR]

The spatial distribution of the b‐value has been determined using a homogeneous earthquake catalogue from 1964 to 2018 to examine the characteristics of structural heterogeneities and their bearing on the style of faulting in the Indo‐Burma region of Northeastern India. The study region is associated with an uneven distribution of structural heterogeneities and demonstrates the mixed type of faulting, nature where the majority of earthquakes resulted mainly by thrust and strike‐slip fault. Our critical analysis demonstrates that the existence of two prominent zones is associated with high b‐values and low b‐values where the mixed type of faulting resulted in the genesis of seismicity in the study area. The high b‐values in the study area contain a large number of smaller earthquakes whilst the low b‐values comprise a smaller number of larger earthquakes distributed along with the Naga‐Disang Thrust, Kabaw Fault, and Sagaing fault systems. A close comparison between b‐values and different faulting styles reveals that the area containing a low b‐value suggests stronger seismogenic zones, where an accumulation of strain energy leads to the genesis of seismicity by thrust faulting with slight slip component whilst Indo‐Burma Ranges regions connected with b‐values (0.5–0.9) demonstrate strike‐slip to the normal mode of faulting, indicating relatively weak seismogenic zones where tensional forces are prevalent. The depth sections illustrate that lower b‐values for the upper part of the subducting slab correspond to more stress generated by tectonics compared to that of the lower part, and interestingly the deeper part of the subducting Indian slab is coupled with conspicuously very high‐b values, suggesting the weakening of the Indian subducting slab through dehydration. [ABSTRACT FROM AUTHOR]

Hydraulic fracturing (HF) operations are widely associated with induced seismicity in the Western Canadian Sedimentary Basin. This study correlates injection parameters of 12,903 HF stages in the Kiskatinaw area in northeast British Columbia with an enhanced catalog containing 40,046 earthquakes using a supervised machine learning approach. It identifies relevant combinations of geological and operational parameters related to individual HF stages in efforts to decipher fault activation mechanisms. Our results suggest that stages targeting specific geological units (here, the Lower Montney formation) are more likely to induce an earthquake. Additional parameters positively correlated with earthquake likelihood include target formation thickness, injection volume, and completion date. Furthermore, the COVID‐19 lockdown may have reduced the potential cumulative effect of HF operations. Our results demonstrate the value of machine learning approaches for implementation as guidance tools that help facilitate safe development of unconventional energy technologies. Plain Language Summary: Hydraulic fracturing (HF), a technique used in unconventional energy production, increases rock permeability to enhance fluid movement. Its use has led to an unprecedented increase of associated earthquakes in the Western Canadian Sedimentary Basin in the last decade, among other regions. Numerous studies have investigated the relationship between induced earthquakes and HF operations, but the connection between specific geological and operational parameters and earthquake occurrence is only partly understood. Here, we use a supervised machine learning approach with publicly available injection data from the British Columbia Oil and Gas Commission to identify influential HF parameters for increasing the likelihood of a specific operation inducing an earthquake. We find that geological parameters, such as the target formation and its thickness, are most influential. A small number of operational parameters are also important, such as the injected fluid volume and the operation date. Our findings demonstrate an approach with the potential to develop tools to help enable the continued development of alternative energy technology. They also emphasize the need for public access to operational data to estimate and reduce the hazard and associated risk of induced seismicity. Key Points: We use supervised machine learning to investigate the relationship between hydraulic fracturing operation parameters and induced seismicityGeological properties and a limited number of operational parameters predominantly influence the probability of an induced earthquakeThe approach has the potential to guide detailed investigations of injection parameters critical for inducing earthquakes [ABSTRACT FROM AUTHOR]

Oceanic transform faults accommodate plate motions through both seismic and aseismic slips. However, deformation partition and slip mode interaction at these faults remain elusive mainly limited by rare observations. We use 1‐year ocean bottom seismometer data collected in 2008 to detect and locate earthquakes at the westernmost Gofar transform fault. The ultra‐fast slipping rate of Gofar results in ∼30,000 earthquakes during the observational period, providing an excellent opportunity to investigate interrelations between the slip mode, seismicity, and fault architecture at an unprecedented resolution. Earthquake distribution indicates that the ∼100‐km‐long Gofar transform fault is distinctly segmented into five zones, including one zone contouring a M6 earthquake that was captured by the experiment. Further, a barrier zone east of the M6 earthquake hosted abundant foreshocks preceding the M6 event and halted its active seismicity afterward. The barrier zone has two layers of earthquakes at depth, and they responded to the M6 earthquake differently. Additionally, a zone connecting to the East Pacific Rise had quasi‐periodic earthquake swarms. The seismicity segmentation suggests that the Gofar fault has multiple slip modes occurring in adjacent fault patches. Spatiotemporal characteristics of the earthquakes suggest that complex fault architecture and fluid–rock interaction play primary roles in modulating the slip modes at Gofar, possibly involving multiple concurrent physical processes. Plain Language Summary: Oceanic transform faults are apparently simple tectonic plate boundaries. However, their structures are surprisingly complex as manifested through various seismic and aseismic slip modes. The deformation partition mechanism is not well understood due to a lack of near‐field observations. Here, we use 1‐year ocean bottom seismometer data to study earthquakes at the westernmost Gofar transform fault and use these earthquakes to infer the fault slip modes. Spatiotemporal evolution of the earthquakes suggests that the fault has five distinctive zones along strike, including one zone hosted a magnitude (M) 6 earthquake captured by the experiment. The remaining zones are dominated by either seismic or aseismic slip. Such distinct variations of slip mode along strike likely originated from the complex, heterogeneous fault structure, and extensive fluid–rock interactions. Key Points: The westernmost Gofar transform fault is composed of distinct seismic and aseismic zonesThese fault zones are controlled by different slip modes and they possibly interact with each other via multiple mechanismsThe slip mode variations may result from the complex fault architecture and fluid–rock interactions at multiple scales [ABSTRACT FROM AUTHOR]

Accurate earthquake location and magnitude estimation play critical roles in seismology. Recent deep learning frameworks have produced encouraging results on various seismological tasks (e.g., earthquake detection, phase picking, seismic classification, and earthquake early warning). Many existing machine learning earthquake location methods utilize waveform information from a single station. However, multiple stations contain more complete information for earthquake source characterization. Inspired by recent successes in applying graph neural networks (GNNs) in graph‐structured data, we develop a Spatiotemporal Graph Neural Network (STGNN) for estimating earthquake locations and magnitudes. Our graph neural network leverages geographical and waveform information from multiple stations to construct graphs automatically and dynamically by adaptive message passing based on graphs' edges. Using a recent graph neural network and a fully convolutional neural network as baselines, we apply STGNN to earthquakes recorded by the Southern California Seismic Network from 2000 to 2019 and earthquakes collected in Oklahoma from 2014 to 2015. STGNN yields more accurate earthquake locations than those obtained by the baseline models and performs comparably in terms of depth and magnitude prediction, though the ability to predict depth and magnitude remains weak for all tested models. Our work demonstrates the potential of using GNNs and multiple stations for better automatic estimation of earthquake epicenters. Plain Language Summary: The location and magnitude of an earthquake can be best determined by relating the motion recorded at multiple stations in a network, and it would be beneficial to combine the waveforms from multiple seismic stations for source characterization. Machine learning‐based approaches have recently become prevalent in seismological tasks such as earthquake source characterization. Because of the irregular spatial distribution of seismic stations, graph convolutional neural networks (a deep learning architecture which handles graph‐structured data) have great potential in combining both spatial and temporal information from different seismic stations. In this work, waveforms recorded at multiple stations are passed through neural networks with connective links based on the similarity of waveform features and geographic locations. The model is tested on two data sets and compared to two published baselines (a graph convolutional neural network and a fully convolutional network). Compared with the baselines, STGNN achieves improved accuracy for epicenter estimation and comparable accuracy for depth and magnitude estimation. Key Points: We design a multistation‐based graph neural network for earthquake source characterizationBoth geographic distance and waveform feature similarity are considered in a graph convolutional neural network for feature combinationOur network is capable of automatically selecting and combining relevant seismic stations to characterize earthquake source parameters [ABSTRACT FROM AUTHOR]

High rate Global Navigation Satellite System (GNSS) processed time series capture a broad spectrum of earthquake strong motion signals, but experience regular sporadic noise that can be difficult to distinguish from true seismic signals. The range of possible seismic signal frequencies amidst a high, location‐varying noise floor makes filtering difficult to generalize. Existing methods for automatic detection rely on external inputs to mitigate false alerts, which limit their usefulness. For these reasons, geodetic seismic signal detection makes for a compelling candidate for data‐driven machine learning classification. In this study we generated high rate GNSS time differenced carrier phase (TDCP) velocity time series concurrent in space and time with expected signals from 77 earthquakes occurring over nearly 20 years. TDCP velocity processing has increased sensitivity relative to traditional geodetic displacement processing without requiring sophisticated corrections. We trained, validated and tested a random forest classifier to differentiate seismic events from noise. We find our supervised random forest classifier outperforms the existing detection methods in stand‐alone mode by combining frequency and time domain features into decision criteria. The classifier achieves a 90% true positive rate of seismic event detection within the data set of events ranging from MW4.8–8.2, with typical detection latencies seconds behind S‐wave arrivals. We conclude the performance of this model provides sufficient confidence to enable these valuable ground motion measurements to run in stand‐alone mode for development of edge processing, geodetic infrastructure monitoring and inclusion in operational ground motion observations and models. Plain Language Summary: Continuously operating, high sample rate Global Navigation Satellite System (GNSS) sensors that experience ground shaking from an earthquake can provide valuable data regarding the nature of the ground motion. If this data is streamed in real‐time, these observations can complement existing traditional seismic infrastructure measurements that are used for earthquake early warning or rapid ground motion assessments. However, the data from these sensors can be noisy and have non‐earthquake artifacts that are difficult to tell apart from true seismic signals. In this work we used a nearly 20‐years archive of high sample rate GNSS velocities occurring during known seismic events to train, validate and test a machine learning model for earthquake detection. This machine learning approach is taken from existing algorithms used for a wide variety of challenging classification problems where a label can be applied to a sample. We demonstrate that this data‐driven method, without any external information, is more likely to detect these signals with less false alarms when compared to existing methods. The added confidence this algorithm provides will allow these valuable measurements to be included in operational seismic assessment and warning decision criteria. Key Points: We assembled a labeled data set of 5 Hz Global Navigation Satellite System (GNSS) velocity time series from 77 earthquakes over nearly 20 yearsWe trained a supervised random forest classifier for detecting seismic motion that outperforms existing detection methodsImproved detection enables lightweight, high rate GNSS velocity processing to be included in operational ground motion observations [ABSTRACT FROM AUTHOR]

Dense arrays deployed near the hydraulic fracturing (HF) wells greatly enhance the understanding of injection‐induced seismicity. In this study, we revisit the continuous recordings that are acquired by 69 three‐component nodes at an HF site in Alberta, Canada, taking advantage of a machine learning‐based seismic detection and location workflow. The obtained new earthquake catalog contains 21,619 events with relative location errors of <1 m, which exceeds ∼20% of the number of earthquakes (18,040) reported previously (Igonin et al., 2021). This high‐resolution catalog reveals the distribution of earthquake sequences at much improved spatiotemporal resolution and illustrates several previously unmapped faults/fractures. Earthquake frequency‐magnitude distribution reveals that the average b value increases with depth from ∼1.1 above 3.5 km to ∼2.5 at greater depths. Further spatial analysis of seismic clusters indicates that the b value varies laterally (∼1–1.7) at shallow depths and is inversely related to the proximity to injection wells in conjunction with the change in structural types (i.e., reactivated fault and pre‐existing fracture). The seismic sequence on the north‐south oriented faults also shows a distinctive occurrence pattern and temporal affinity to the fracturing network reactivation between the two stages of HF operations. The change in faulting behavior could reflect a shift of dominating triggering mechanisms and physical processes from (a) the rapid diffusion of pore fluid pressure along pre‐existing fracture corridors to (b) the cascade migration of earthquake sequences in response to the cumulative Coulomb stress perturbation on the fluid‐lubricated, critically stressed faults. Plain Language Summary: The potential for fluid‐related industrial activities (e.g., hydraulic fracturing, HF) to induce felt earthquakes has long been recognized. In the past few years, a number of felt events (i.e., M ∼ 4) have occurred in western Canada, which are often linked to HF operations (e.g., Kao et al., 2018, https://doi.org/10.1190/tle37020117.1; Mahani et al., 2019, https://doi.org/10.1785/0220190040). To better understand processes that induce earthquakes, the Tony Creek dual Microseismic Experiment (ToC2ME) deployed a dense seismic array around an HF well in late 2016 (26 October to 30 November). In this study, we utilize a recently developed earthquake detection and location workflow to revisit the ToC2ME data set and build a high‐resolution earthquake catalog. We observe over 20,000 earthquakes with magnitudes down to −2 with high location accuracy at meter scale. The improved earthquake locations allow us to better delineate the fractures and faults activated during the HF. Our results further reveal the complexity of the triggering mechanism of HF‐induced earthquakes according to the statistical analysis of the frequency‐magnitude distribution of earthquakes. Our study suggests that multiple factors, including pore fluid pressure due to fluid injection and perturbation in the stress field caused by earthquakes, can jointly control the spatiotemporal variations of regional seismicity during multi‐stage HF operations. Key Points: We revisited the Tony Creek dual Microseismic Experiment data set and built a high‐resolution catalog with a new earthquake detection and location workflowThe 21,619 events revealed spatially varying b values associated with changes in structural type and pore‐fluid pressureMultiple triggering mechanisms may control earthquake occurrence during multi‐stage hydraulic fracturing operations [ABSTRACT FROM AUTHOR]

An Ms 6.4 earthquake occurred in Yangbi County, Yunnan Province, on May 21, 2021. The magnitude is only 0.1 difference when compared with the 2014 Ms 6.5 earthquake in Ludian County, Yunnan Province, but the disasters show a big difference. Seismic signal includes the comprehensive information of source, propagation path, and site effect, which directly reflect the effect of the earthquake on the ground. The study mainly investigated the difference between seismic signals of two events in time domain and time-frequency-energy space based on variational mode decomposition and Hilbert transform to clarify the reasons for disaster differences. Most damaged buildings are column-and-tie timber structures with Earth walls with local collapse or vertical crack at the corner of the Earth wall in the Yangbi earthquake. Multistorey buildings are intact, and a few small landslides are triggered, whereas, in the Ludian earthquake, multistorey buildings with the frame structure and masonry structure occurred with serious X-type cracking, large interlayer displacement, and even complete collapse, and large landslides with length about 165 m are triggered. 53LLT and 53YBX are two strong motion stations in the Ludian earthquake and Yangbi earthquake, respectively, and are near the disaster investigation points. Peak ground acceleration in 53LLT is up to 949.2 cm/s2, which is larger than 720.3 cm/s2 in 53YBX, and the duration in 53LLT is longer than 53YBX. The instantaneous energy of 53YBX is concentrated within 0–10 Hz and 10–20 Hz, respectively; however, the instantaneous energy of 53LLT is concentrated at 5 Hz. Cumulative energy of 53YBX is mainly distributed at 0–5 Hz and 10–20 Hz, which is around the natural frequency of column-and-tie timber structure, while the energy of 53LLT is concentrated at 1.6 Hz and is much greater than other stations. Thus, large energy in high frequency caused the instantaneous and cumulative damage to short-period structures in Yangbi County and may cause the resonance of column-and-tie timber structure, and large energy in low frequency caused the instantaneous and cumulative damage to long-period structures of Longtou Village in Ludian earthquake. The result is meaningful to understanding the seismic disasters from the characteristics of seismic signals and can provide a reference for the seismic design of structures. [ABSTRACT FROM AUTHOR]

Outer‐rise faults are predominantly concentrated near ocean trenches due to subducted plate bending. These faults play crucial roles in the hydration of subducted plates and the consequent subducting processes. However, it has not yet been possible to develop high‐resolution structures of outer‐rise faults due to the lack of near‐field observations. In this study we deployed an ocean bottom seismometer (OBS) network near the Challenger Deep in the Southernmost Mariana Trench, between December 2016 and June 2017, covering both the overriding and subducting plates. We applied a machine‐learning phase detector (EQTransformer) to the OBS data and found more than 1,975 earthquakes. An identified outer‐rise event cluster revealed an outer‐rise fault penetrating to depths of 50 km, which was inferred as a normal fault based on the extensional depth from tomographic images in the region, shedding new lights on water input at the southmost Mariana subduction zone. Plain Language Summary: Estimating water input at subduction zones plays a crucial role in understanding the material cycles of the Earth and subduction zone dynamics. As outer‐rise faults provide the primary channel for water to penetrate into the incoming plate, investigating the generation and extent of outer‐rise faults is thus an effective way to understand the hydration degree. However, high‐resolution structure of outer‐rise faults is not common due to the lack of near‐field observations. In this study, we apply a machine‐learning phase detector (EQTransformer) to a new ocean bottom seismometer data set at the southernmost Mariana Subduction Zone. After careful analysis of earthquake location and clustering, we find a deep outer‐rise fault extending to 50 km in depth. We interpret the fault as a normal fault based on the depth range of a cluster of events with high waveform similarity. Such a deep outer‐rise fault implies much higher water input at southern most Mariana than had been previously estimated. Key Points: A machine‐learning phase detector (EQTransformer) was applied to nearfield ocean bottom seismometer data at the southernmost Mariana trenchThe outer‐rise seismicity at the southernmost Marina trench varied along the trench, with one identified clusterWe identified an outer‐rise fault that penetrates to depths of 50 km [ABSTRACT FROM AUTHOR]

Earthquake phase association algorithms aggregate picked seismic phases from a network of seismometers into individual sesimic events and play an important role in earthquake monitoring and research. Dense seismic networks and improved phase picking methods produce massive seismic phase datasets, particularly for earthquake swarms and aftershocks occurring closely in time and space, making phase association a challenging problem. We present a new association method, the Gaussian Mixture Model Association (GaMMA), that combines the Gaussian mixture model with earthquake location, origin time, and magnitude estimation. We treat earthquake phase association as an unsupervised clustering problem in a probabilistic framework, where each earthquake corresponds to a cluster of P and S phases with a hyperbolic moveout of arrival times and a decay of amplitude with distance. We use the multivariate Gaussian distribution to model the collection of phase picks of an event; and the mean of the multivariate Gaussian distribution is given by the predicted arrival time and amplitude from the causative event. We carry out the pick assignment to each earthquake and determine earthquake source parameters (i.e., earthquake location, origin time, and magnitude) under the maximum likelihood criterion using the Expectation‐Maximization algorithm. The GaMMA method does not require typical association steps of other algorithms, such as grid‐search or supervised training. The results for both synthetic tests and for the 2019 Ridgecrest earthquake sequence show that GaMMA effectively associates phases from a temporally and spatially dense earthquake sequence while producing useful estimates of earthquake location and magnitude. Plain Language Summary: Earthquakes are monitored by seismic networks consisting of several to hundreds of seismometers. An earthquake detection workflow usually has two important steps: detecting/picking seismic phases at each seismometer and associating picked phases across multiple seismometers in a network. Deep‐learning‐based phase pickers have greatly improved phase detection performance and can automatically generate many more seismic phases than conventional algorithms. These massive numbers of automatic phase picks pose a challenge for the phase association task. We have developed a new phase association method using a Bayesian Gaussian Mixture Model. We treat the phase association problem as a unsupervised clustering problem meaning that we aim to cluster detected phases into different groups based on individual earthquakes that produce these phases. The Gaussian mixture model makes it easy to consider multiple phase parameters, such as phase arrival time, phase amplitude, phase picking quality score, and phase type, to improve phase association. We test our method on both synthetic data and the 2019 Ridgecrest earthquake. The results show that our method can effectively associate phases from a temporally and spatially dense earthquake sequence and generate a more complete earthquake catalog than catalogs created using conventional methods. Key Points: We proposed an new approach to solve phase association as an unsupervised clustering problem using the Bayesian Gaussian Mixture ModelWe used the multivariate Gaussian distribution to represent both phase arrival time and amplitude to improve associationOur unsupervised method is fast without the need for conventional grid‐search or supervised training [ABSTRACT FROM AUTHOR]

Earthquake monitoring by seismic networks typically involves a workflow consisting of phase detection/picking, association, and location tasks. In recent years, the accuracy of these individual stages has been improved through the use of machine learning techniques. In this study, we introduce a new end‐to‐end approach that improves overall earthquake detection accuracy by jointly optimizing each stage of the detection pipeline. We propose a neural network architecture for the task of multi‐station processing of seismic waveforms recorded over a seismic network. This end‐to‐end architecture consists of three sub‐networks: a backbone network that extracts features from raw waveforms, a phase picking sub‐network that picks P‐ and S‐wave arrivals based on these features, and an event detection sub‐network that aggregates the features from multiple stations to associate and detect earthquakes across a seismic network. We use these sub‐networks together with a shift‐and‐stack module based on back‐projection that introduces kinematic constraints on arrival times, allowing the neural network model to generalize to different velocity models and to variable station geometry in seismic networks. We evaluate our proposed method on the STanford EArthquake Dataset (STEAD) and on the 2019 Ridgecrest, CA earthquake sequence. The results demonstrate that our end‐to‐end approach can effectively pick P‐ and S‐wave arrivals and achieve earthquake detection accuracy rivaling that of other state‐of‐the‐art approaches. Because our approach preserves information across tasks in the detection pipeline, it has the potential to outperform approaches that do not. Plain Language Summary: Earthquakes are monitored using multiple seismic stations in a seismic network. A typical earthquake detection workflow consists of two stages: first, earthquake signals are detected at each station; then these detections are associated across multiple stations to determine whether an earthquake occurred. These two stages are independently optimized in conventional algorithms, thus the overall earthquake detection performance can be limited due to information loss between each stages. In this work, we proposed an end‐to‐end approach to jointly optimize both stages (i.e., signal‐station detection and multi‐station association) inside one combined neural network architecture. The results on the 2019 Ridgecrest, CA earthquake sequence show that our end‐to‐end approach achieves earthquake detection accuracy rivaling that of other state‐of‐the‐art approaches. Because our approach preserves information across tasks in the detection pipeline, it has the potential to outperform approaches that do not. Key Points: We optimize the phase picking and phase association tasks using an end‐to‐end neural network architectureWe design a shift‐and‐stack module based on back‐projection to introduce kinematic constraints on arrival times and station geometryOur approach preserves information across tasks in the detection pipeline and has the potential to outperform approaches that do not [ABSTRACT FROM AUTHOR]

Seismology focuses on the study of earthquakes and associated phenomena to characterize seismic sources and Earth structure, which both are of immediate relevance to society. This article is composed of two independent views on the state of the integrated, coordinated, open, networked (ICON) principles (Goldman et al., 2021, https://doi.org/10.1029/2021eo153180) in seismology and reflects on the opportunities and challenges of adopting them from a different angle. Each perspective focuses on a different topic. Section 1 deals with the integration of multiscale and multidisciplinary observations, focusing on integrated and open approaches, whereas Section 2 discusses computing and open‐source algorithms, reflecting coordinated, networked, and open principles. In the past century, seismology has benefited from two co‐existing technological advancements—The emergence of new, more capable sensory systems and affordable and distributed computing infrastructure. Integrating multiple observations is a crucial strategy to improve the understanding of earthquake hazards. However, current efforts in making big datasets available and manageable lack coherence, which makes it challenging to implement initiatives that span different communities. Building on ongoing advancements in computing, machine learning algorithms have been revolutionizing the way of seismic data processing and interpretation. A community‐driven approach to code management offers open and networked opportunities for young scholars to learn and contribute to a more sustainable approach to seismology. Investing in new sensors, more capable computing infrastructure, and open‐source algorithms following the ICON principles will enable new discoveries across the Earth sciences. Plain Language Summary: Seismological observations can provide critical insights into the physical processes of the Earth's interior and associated near‐surface consequences, resulting from both natural and anthropogenic activities across spatial and temporal scales. This commentary discusses the current status and opportunities of integrated, coordinated, open, and networked (ICON) principles in seismology. As an applied discipline, seismology is highly data‐dependent and inherently relies on sensing (data acquisition) and computing (data processing and modeling) technologies. Integrating data from multiple scales and domains has improved our understanding of earthquakes and is also beneficial to adjacent disciplines, such as reservoir engineering and rock mechanics. When more open data and models are produced and integrated by coordinated acquisition systems and networked programming efforts, seismology will enable new discoveries in the Earth sciences. Key Points: We comment on the current status and potential for synergy and challenges of implementing integrated, coordinated, open, networked principles in seismologyThe integration of multi‐parametric and multi‐scale observations across disciplines benefits Earth imaging and earthquake understandingHigh‐performance computing and open‐source algorithms offer networked opportunities for a broader community to contribute to seismology [ABSTRACT FROM AUTHOR]

Recent studies suggested that transient and long‐term stress changes caused by Earth's surface processes (e.g., extreme weather events, annual variations on groundwater storages) can affect earthquake activities in the subsurface. However, these studies may be limited by the completeness of standard earthquake catalogs, especially during or right after extreme weather events. Here we apply the template matching method to build a more complete earthquake catalog in Taiwan spanning seven months before and 12 months after 2009 typhoon Morakot, which brought the highest rainfall in southern Taiwan in the past 60 years and triggered numerous landslides. We then use the enhanced catalog to investigate possible influences of typhoon‐driven Earth's surface processes (atmospheric pressure, precipitation, and erosion) on local seismicity. We find that the seismicity rate of a 40‐day earthquake sequence in northeastern Taiwan was reduced significantly right after the passage of typhoon Morakot's eye center. In the typhoon‐triggered landslide zone in southern Taiwan, we find a slight increase in background seismicity rate in the next year after Morakot, matching the results of a recent study. However, we do not observe a clear change in the Gutenberg‐Richter b‐value in this zone, which is different from the recent study. Station outages during and right after Morakot prevents us from better understanding short‐term precipitation effect on local seismicity. Overall, except for a reduction in seismicity rate near the typhoon's low‐pressure eye center in northeastern Taiwan, we do not observe other clear seismicity changes that can be attributed to surface changes induced by typhoon Morakot. Plain Language Summary: It is important to study cascading nature hazards (typhoons/hurricanes, landslides, earthquakes, etc.) because of their threats to human life and property. 2009 typhoon Morakot is the deadliest typhoon in Taiwan, which brought the highest rainfall in the past 60 years and led to numerous landslides. To investigate the possible influences of Morakot on subsurface earthquake activities, we build a more complete earthquake catalog for Taiwan spanning seven months before and 12 months after 2009 typhoon Morakot. This enhanced earthquake catalog includes many small earthquakes that cannot be felt by local residents in their ordinary life and were not in the standard earthquake catalog. In northeastern Taiwan, our results show that a 40‐day earthquake sequence was apparently shut down right after the passage of typhoon Morakot. If this observation is not a pure coincidence, it may suggest that the low‐pressure system of typhoon Morakot was capable of modulating the subsurface seismicity behavior. In southern Taiwan, the heavy‐rainfall and landslide region, we do not observe clear seismicity changes that can be attributed to surface changes induced by typhoon Morakot. Key Points: We use a template matching method to rebuild an earthquake catalog for Taiwan 7 months before and 12 months after 2009 typhoon MorakotThe seismicity rate of a 40‐day earthquake sequence in northeastern Taiwan was reduced significantly after the passage of MorakotThe seismicity rate increased slightly but without clear b‐value change after Morakot in typhoon‐driven landslide zone in southern Taiwan [ABSTRACT FROM AUTHOR]

Seismic attenuation is generally greater in the western United States (WUS) than the central and eastern United States (CEUS), but the nature of this transition or location of this boundary is poorly constrained. We conduct crustal seismic (Lg) attenuation tomography across a region that stretches from the CEUS across the Rocky Mountains to the Basin and Range using a total of 115,870 amplitude measurements from 106 earthquakes recorded on 544 stations across five frequency bands spanning 0.5–16 Hz. Similar to previous studies, we find higher attenuation in the WUS (Q0 ∼ 190) than the nominally CEUS (Q0 ∼ 250) and comparatively high attenuation on the Gulf Coast (Q0 ∼ 175). Our models defy simple east versus west regionalization, however. Heterogeneity within the Rocky Mountain region—low attenuation in the Colorado Plateau interior and Wyoming Craton (Q0 ∼ 230) compared to high attenuation in the southern Rockies (Q0 ∼ 110)—exceeds the gross differences between the CEUS and western United States. These province‐scale patterns are readily interpreted in terms of intrinsic attenuation. The boundary between the Colorado Plateau and Basin and Range hosts the highest attenuation imaged in the study area (Q0 ∼ 90), consistent with localized scattering across contrasting crustal structure. Focused high attenuation in the southern Rockies may represent the effects of represent in situ partial crustal melt. Within the CEUS, second‐order bands of comparatively high attenuation align with the Proterozoic Yavapai‐Mazatzal suture zone and Midcontinent Rift. This complex attenuation structure defies broad regionalization and suggests a need for path‐specific models near these boundaries and for critical infrastructure. Plain Language Summary: Shaking dies away, or attenuates, with distance from an earthquake. Attenuation is slower in the eastern United States than the West, which greatly increases the area over which shaking is felt and damage can occur. Low seismicity rates in central North America have made it challenging to determine and understand the boundary between East and West, however. Recent manmade earthquakes in parts of Oklahoma and Colorado provide new data to map attenuation across the nation, and we find that there is not simple dichotomy between East and West after all. Instead, ancient tectonic events and modern processes combine to create a complex tapestry of attenuation. Understanding this complexity, however, will make it easier to forecast and prepare for seismic hazard. Key Points: Complex attenuation structure in the Rocky Mountain region defies broad regionalizationSharp, high‐scattering boundaries between adjacent provinces; these may require path‐specific ground‐motion predictionsEvidence for in situ crustal melt in the southern Rockies and focused alteration and/or scattering in central and eastern United States sutures and tectonic zones [ABSTRACT FROM AUTHOR]

The Raton Basin has been an area of injection induced seismicity for the past two decades. Previously, the reactivated fault zone structures and spatiotemporal response of seismicity to evolving injection have been poorly constrained due to sparse publicly available seismic monitoring. The application of a machine‐learning phase picker to 4 years of continuous seismic data from a local array enables the detection and location of ∼38,000 earthquakes. The events from 2016 to 2020 are ∼2.5–6 km below sea level and range from ML < −1 to 4.2. Most earthquakes occur within previously identified ∼N‐S zones of seismicity, however our new catalog illuminates that these zones are composed of many short faults with variable orientations. The two most active zones, the Vermejo Park and Tercio zones, are potentially linked by small intermediate faults. In total, we find ∼60 short (<3 km long) basement faults with strikes from WNW to NNE. Faulting mechanisms are predominantly normal but some variability, including reverse dip‐slip and oblique‐slip, is observed. The Trinidad fault zone, which previously hosted a Mw 5.3 earthquake in 2011, is quiescent during 2016–2020, likely in response to both slow accumulation of tectonic strain after the 2011 sequence, and the significant decrease (80% reduction) in nearby wastewater injection from 2012 to 2016. Unlike some other regions, where induced seismicity was triggered in response to higher injection rates, the Raton Basin's frequency‐magnitude and spatiotemporal statistics are not distinguishable from tectonic seismicity. The similarity suggests that seismicity in the Raton Basin is predominantly releasing tectonic stress. Plain Language Summary: The Raton Basin, located on the Colorado‐New Mexico border, has a low level of historical seismicity but experienced a significant increase of earthquakes in the past two decades, coinciding with increased wastewater injection and coal‐bed methane production. We use four years of continuous local seismic data to detect ∼38,000 earthquakes from 2016 to 2020. We find that the previously identified 10–20 km long zones of seismicity are composed of many shorter faults, <3 km long, with variable orientations and slip mechanisms. We observed very little seismicity in the zone that previously hosted the largest earthquake (Mw 5.3 in 2011) in the basin and attribute this finding to both the slow accumulation of tectonic strain following the 2011 sequence and the significant decrease (80% reduction) of wastewater injection to this zone from 2012 to 2016. Overall, the frequency‐magnitude and spatial‐temporal earthquake behavior of the Raton Basin is not distinguishable from tectonic seismicity. This statistical similarity suggests Raton Basin seismicity predominantly releases tectonic stress accumulated over geologic time. Key Points: Seismicity is hosted by short, <3 km long, reactivated basement faults with variable strikes and faulting stylesThe fault zone that hosted a Mw 5.3 event in 2011 exhibits low seismicity following an 80% reduction of the wastewater‐injection rateSimilarity of spatiotemporal‐magnitude basin statistics to tectonic sequences suggests earthquakes mainly release stored tectonic stress [ABSTRACT FROM AUTHOR]

Although infrequent, large (Mw7.5+) earthquakes can be extremely damaging and occur on subduction and intraplate faults worldwide. Earthquake early warning (EEW) systems aim to provide advanced warning before strong shaking and tsunami onsets. These systems estimate earthquake magnitude using the early metrics of waveforms, relying on empirical scaling relationships of abundant past events. However, both the rarity and complexity of great events make it challenging to characterize them, and EEW algorithms often underpredict magnitude and the resulting hazards. Here, we propose a model, M‐LARGE, that leverages deep learning to characterize crustal deformation patterns of large earthquakes for a specific region in real‐time. We demonstrate the algorithm in the Chilean Subduction Zone by training it with more than six million different simulated rupture scenarios recorded on the Chilean GNSS network. M‐LARGE performs reliable magnitude estimation on the testing data set with an accuracy of 99%. Furthermore, the model successfully predicts the magnitude of five real Chilean earthquakes that occurred in the last 11 years. These events were damaging, large enough to be recorded by the modern high rate global navigation satellite system instrument, and provide valuable ground truth. M‐LARGE tracks the evolution of the source process and can make faster and more accurate magnitude estimation, significantly outperforming other similar EEW algorithms. This is the first demonstration of our approach. Future work toward generalization is outstanding and will include the addition of more training and testing data, interfacing with existing EEW methods, and applying the method to different tectonic settings to explore performance in these regions. Plain Language Summary: Great earthquakes are infrequent but devastating natural disasters. To mitigate their effects, earthquake early warning (EEW) systems aim to provide advance warning of strong shaking and tsunami. However, many of the most sophisticated EEW algorithms operating globally have a difficult time characterizing large earthquakes quickly and accurately enough to issue a meaningful warning—this is most evident from the failure of EEW during the 2011 M9 Tohoku Oki, Japan earthquake. Here, we propose a model, M‐LARGE, that learns earthquake's surface deformation patterns from millions of simulations for a specific region, and then applies it to unseen events in the same region. Our model shows a high accuracy of 99% performing on the testing data set and accurately estimates the magnitude of five real large historical events in Chile. The M‐LARGE outperforms currently operating similar EEW algorithms. This is the first demonstration of our approach, we note that for actual EEW operation, future work including the addition of more training and testing data, interacting with existing EEW methods, and testing the method in alternative tectonic settings are necessary prior to real‐time operations. Key Points: We use synthetic data to train a deep learning model to predict magnitude from crustal deformation patterns in simulated real‐timeThe model, M‐LARGE, has an accuracy of 99% and accurately estimates the magnitude of five real large events, outperforming other methodsM‐LARGE's rapid and accurate magnitude prediction suggesting that significant warning times are possible during real large earthquakes [ABSTRACT FROM AUTHOR]

Recently deep learning (DL), as a new data-driven technique compared to conventional approaches, has attracted increasing attention in geophysical community, resulting in many opportunities and challenges. DL was proven to have the potential to predict complex system states accurately and relieve the "curse of dimensionality" in large temporal and spatial geophysical applications. We address the basic concepts, state-of-the-art literature, and future trends by reviewing DL approaches in various geosciences scenarios. Exploration geophysics, earthquakes, and remote sensing are the main focuses. More applications, including Earth structure, water resources, atmospheric science, and space science, are also reviewed. Additionally, the difficulties of applying DL in the geophysical community are discussed. The trends of DL in geophysics in recent years are analyzed. Several promising directions are provided for future research involving DL in geophysics, such as unsupervised learning, transfer learning, multimodal DL, federated learning, uncertainty estimation, and active learning. A coding tutorial and a summary of tips for rapidly exploring DL are presented for beginners and interested readers of geophysics. [ABSTRACT FROM AUTHOR]