An inverse problem involves deducing the causes from observed effects within a system, often requiring the solution of partial differential equations that describe the underlying physics. In geophysical seismic imaging, the objective is to reconstruct subsurface structures, such as velocity and density fields, by analyzing seismic waveforms recorded at the Earth’s surface. This process involves solving the non-linear wave equation to model seismic wave propagation through the Earth. Full Waveform Inversion (FWI) is a deterministic technique that employs gradient-based methods. Despite its potential, FWI faces challenges such as non-uniqueness, local minima, and computational complexity, highlighting the critical need for advanced methods to address these issues and quantify uncertainties in subsurface property estimation.
Bayesian inference provides a robust framework for solving inverse problems and estimating uncertainties by applying Bayes' theorem. This approach derives a posterior probability density function for model parameters based on observed data. In this study, we present an innovative method that parametrizes unknowns using Generative Adversarial Networks (GANs), enabling the creation of realistic subsurface representations by learning the prior distribution in a latent space. Once trained, the GAN remains fixed, serving as a generative prior for Bayesian posterior sampling.
We compare and evaluate four posterior sampling methods, i.e. the Metropolis-adjusted Langevin Algorithm (MALA), variational Bayesian inference using normalizing flows (NF), inference neural networks (INN), and Stein Variational Gradient Descent (SVGD). The performance of these methods is assessed in terms of computational efficiency and accuracy in capturing the posterior distribution. By integrating deep generative priors with advanced Bayesian sampling techniques, we demonstrate significant improvements in handling the high dimensionality and non-linearity inherent in geophysical inverse problems. This work contributes to the development of advanced methods for seismic imaging and uncertainty quantification, aligning with the need for robust, data-driven approaches in the field of geophysics.
How to cite: Xie, Y., Chauris, H., and Desassis, N.: Generative AI and Bayesian methods for seismic imaging and uncertainty estimation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7358, https://doi.org/10.5194/egusphere-egu25-7358, 2025.
Recent advances in seismology underscore the potential of innovative instrumentation and data-driven methods to overcome long-standing challenges in source localization. Traditional techniques such as P-wave polarization or arrival time analysis often suffer from reduced precision in complex wavefields, where scattering and heterogeneities distort seismic signals. These limitations highlight the need for methods that leverage emerging technologies and provide robust uncertainty quantification.
In this study, we used a novel approach for determining the back azimuth of seismic sources using six degrees of freedom (6-DoF) ground motion measurements, enabling precise source localization from single-point data. We tested the method in a controlled experiment, tracking the migration of a vibroseis truck across 160 distinct locations. Each of the 480 recorded sweep signals, lasting 15 seconds and spanning a frequency range of 7–120 Hz, was analyzed to derive back azimuth values.
One key innovation in this approach is the use of a wavefield fingerprinting algorithm to isolate SV-type waves, significantly improving the precision of back azimuth estimates. This step addresses the inherent challenges posed by the sensitivity of rotational sensors primarily to S waves and the scattering effects that degrade localization accuracy as the source moves farther from the receiver. By isolating the SV wavefield, our method reduced deviations in back azimuth estimates to a maximum of 2.2 degrees, compared to deviations of up to 48.6 degrees when the entire wavefield was analyzed.
Our findings not only demonstrate the value of combining advanced monitoring instruments with wavefield-specific processing techniques but also highlight the importance of integrating uncertainty quantification into seismic analyses. This approach offers a pathway to more robust localization methods, especially for applications requiring high-resolution imaging or real-time seismic monitoring in complex tectonic environments.
How to cite: Izgi, G., Eibl, E. P. S., Krüger, F., and Bernauer, F.: Accurate Back Azimuth Determination Using 6-DoF Measurements and Wavetype Fingerprinting, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4181, https://doi.org/10.5194/egusphere-egu25-4181, 2025.
Enriching seismic network catalogues is a key object to understand the seismotectonic processes and seismic hazard. In areas with poor knowledge of seismotectonic patterns, the installation of dense local seismic networks is essential. The enhanced density of seismic network coverage improves the detection of seismic events of low energy. Nowadays, Machine Learning (ML) techniques are becoming widely used in seismology, in addition to standard automatic procedures (i.e., STA/LTA-based algorithms).
In this study, we assess the performance of automatic P- and S-wave picking and earthquake detection algorithms for the period from 2013 to 2022, applied to the data recorded by the OTRIONS seismic network (FDSN code OT), a local network installed in 2013 in the Apulia region (Southern Italy) by UniBa and INGV.
The aim is to provide an automated data analysis system to collect a catalogue of the seismic activity of the Gargano area. For the period 2013-2022, a catalogue has been collected by employing CASP (Complete Automatic Seismic Processor), a software based on STA/LTA algorithms for automated event detection, picking and location. The obtained CASP automated catalogue has been manually revised to identify false events and quarry blasts.
Now, for the same period, the goal is to compile a new seismic catalogue for the Gargano area by using PhaseNet, an ML algorithm for phase detection, based on a deep neural network. We used GaMMA algorithm for phase association. Finally, NonLinLoc software was used to locate the events.
The results revealed a significant increase in the number of detected events with respect to the CASP processing. To evaluate the reliability of the results obtained by PhaseNet and GaMMA, a manual revision has been carried out on a sub-dataset of the collected event catalogue and compared with the CASP manual catalogue for the same period: we observed a significant increase in the earthquake detection. This increase also relates to events whose reliability has been verified.
From a seismotectonic point of view, the newly detected seismicity confirms the seismicity pattern of the Gargano Promontory, characterized by a deepening of the earthquakes trend moving northwards in the area with a clear and well defined cut off of the seismicity in the lower crust. This peculiar result is one of the most intriguing findings of the study and could provide important indications on the thermo-rheological characteristics of the lower crust.
Finally, to improve the knowledge of the seismogenic structures of the Gargano area, the new package of NonLinLoc, NLL-SSST-coherence, was used to looking for seismogenic structures. Preliminary results show that the Gargano area is characterised by widespread seismicity.
How to cite: Ferreri, A. P., Cecere, G., Filippucci, M., Ninivaggi, T., Panebianco, S., Romeo, A., Satriano, C., Serlenga, V., Selvaggi, G., Stabile, T. A., and Tallarico, A.: Enhancing the seismic catalog of the Gargano Area (Southern Italy) with machine learning-based event detection and earthquake relocation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13258, https://doi.org/10.5194/egusphere-egu25-13258, 2025.
Understanding rupture mechanisms, seismicity propagation, distribution, and migration after a major earthquake relies on the quality of earthquake catalogs, particularly their detection capabilities, location accuracy, and magnitude completeness. On February 27, 2010, a Mw 8.8 earthquake struck the Maule region in south-central Chile, causing widespread damage and substantial loss of life. As the largest well-instrumentally recorded earthquake in Chile, this event offers a unique opportunity to revisit an old dataset, refine the aftershock sequence analysis, and gain deeper insights into subduction zone dynamics.
Here we analyze ~10 months of continuous seismic data from the International Maule Aftershock Deployment (IMAD), a temporary network with about 156 stations deployed throughout the rupture zone. Using the recent Back-Projection and Matched Filtering (BPMF) workflow, which integrates PhaseNet, a deep-learning-based phase picker, we detected more than 100,000 earthquakes with at least 10 P and S-wave arrival phases. We relocated these events using NonLinLoc-SSST-Coherence, a probabilistic algorithm. A subset of them served as templates for template matching, producing a final catalog of about 375,000 events. This represents nearly a ninefold increase in detected events compared to prior catalogs and achieves a magnitude of completeness of Mw ~1.7, lowering it by over one order of magnitude.
Our catalog significantly enhances the spatio-temporal resolution, revealing intricate seismic structures (e.g., fault geometries) and dynamic post-seismic activity. Our improved relocations draw these key features, including the shallower seismic zone in the Pichilemu-Vichuquén region (33.5°S–35°S) and deeper seismic clusters near Concepción (37°S–38°S) in unprecedented detail. Temporal b-value variations (0.8–1.1) reveal zones of high-stress accumulation and the activation of multiple fault systems, highlighting the heterogeneous nature of post-seismic deformation. This high-resolution dataset underscores the potential of modern methodologies and algorithms, unveiling features from older data with improved clarity and detail.
How to cite: Flores-Allende, R., Seydoux, L., Beaucé, É., Bonilla, L. F., Gueguen, P., and Satriano, C.: An enhanced earthquake catalog of the 2010 Mw 8.8 Maule aftershock sequence using modern tools, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6874, https://doi.org/10.5194/egusphere-egu25-6874, 2025.
Seismic activity provides critical insights into subsurface processes such as tectonic movements, volcanic activity, and fluid migrations, with accurate earthquake locations being essential for enhancing our understanding of these seismic behaviors. However, seismic array geometry significantly influences earthquake location accuracy. Over the past two decades, seismologists have improved earthquake catalogs by expanding seismic networks and densifying station coverage in seismically active regions, leading to more precise event detection and location accuracy. Following major seismic events, temporary seismometer deployments refine monitoring and analysis, particularly for aftershocks, enhancing the understanding of the region's seismic behavior and potential risks. In this study, we introduce a novel method that benefits from these temporary deployments to relocate hypocenters determined by a permanent seismic array, using hypocenters derived from a combination of permanent and temporary arrays with better geometry. Our method employs a random forest algorithm to learn how to relocate seismic
events detected with the permanent, low-density seismic array. We developed the method in the case of Mayotte Island, where, following the eruption in 2018, scientists deployed ocean-bottom seismometers (OBS) and land seismic sensors to build high-quality catalogs that provide a better understanding of the region's dynamics. Our findings show a significant reduction in root-mean-square error between the hypocenters located with permanent seismic stations and those located with a combination of permanent and temporary seismic stations, demonstrating the method's effectiveness in reducing systematic biases and enhancing location accuracy. This method is applicable across a range of contexts, particularly in scenarios characterized by suboptimal seismic station geometry, offering a robust framework for enhancing the location accuracy of seismic events.
How to cite: Mohammadi, F., Seydoux, L., Retailleau, L., and Satriano, C.: Machine learning enhanced earthquake relocation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10327, https://doi.org/10.5194/egusphere-egu25-10327, 2025.
Recent studies have demonstrated the potential of deep learning (DL) techniques for denoising seismic signals and improving signal analysis, but they are not yet widely adopted in seismic monitoring. Denoising models are typically applied to short segments of triggered data. To use the full potential of state-of-the-art denoising models for seismicity catalogue generation, the methods need to be applicable to continuous data. Several challenges arise in this case, in particular during dense (aftershock) sequences, as models may fail to consistently detect signals near or across the window edges, or require overlapping windows that lead to several parallel denoised waveform solutions.
We investigate the optimal integration of denoising approaches into operational network settings to enhance seismic catalogues, focusing on improving detections, phase picks, and peak amplitude measurements. As we are most interested in characterising weak events that are commonly missing or poorly described in existing catalogues, special attention is given to them. We train and compare a range of promising algorithms, including a method that operates on time-frequency representation of the data and outputs segmentation masks to separate event and noise signals.
We evaluate the approach on seismic data recorded by the Swiss network, and train a model on recorded noise and about 25k earthquake signals, corresponding to most of the available high-quality, local recordings.
To assess the benefit of the denoiser, we test it on a dense seismic sequence recorded by different types of seismic sensors under diverse site and noise conditions. We employ the denoiser to detect event signals, and produce continuous denoised data, which then serve as the input for standard phase pickers and event associators. We compare the derived catalogue to those obtained with i) standard and ii) DL tools, both applied on raw data. We demonstrate i) significantly deeper catalogues in the first case, and ii) catalogues comparable to those obtained with DL pickers, but with enhanced characterisation, including event location and magnitudes.
How to cite: Dahmen, N., Clinton, J., and Meier, M.-A.: Operational Seismic Denoising Workflow to Enhance Seismic Catalogues, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20208, https://doi.org/10.5194/egusphere-egu25-20208, 2025.
Machine learning (ML) has seen widespread use in seismology recently, with a significant focus on earthquake monitoring. ML models are now available for phase picking, first motion polarity determination, etc. Implementing them in standard monitoring software (e.g. SeisComP) could significantly improve the automatic earthquake monitoring and save time for human analysts, whilst leveraging all the existing benefits of existing mature monitoring frameworks. An important first step for moving the ML models from research into production has been the Python package SeisBench (Woollam et al., 2022; DOI: 10.1785/0220210324), which allows users to benchmark and access ML models and datasets. The scdlpicker SeisComP module (Tillman et al., 2023; DOI: 10.5194/egusphere-egu23-10046) created an interface between SeisComP and the trained ML pickers in SeisBench to allow event-based re-picking (i.e., not real-time phase onset detection) as demonstrated using teleseismic earthquakes and the GEOFON network. Here, we build on top of the existing scdlpicker module to provide both P and S picks at local distances, and add pick uncertainty and P-pick first motion polarity. We demonstrate the performance of this extended module in routine earthquake monitoring at the Swiss Seismological Service (SED) and show the improvements over classical pickers currently in use. We show that the ML pickers improve the automatic monitoring in both the number and the quality of the picks, leading to better automatic locations and magnitudes. We show that the ML picker’s characteristic function provides a good proxy of the human analyst assigned pick uncertainty. Additionally, this extended SeisComP module provides the ML-determined first-motion polarity for each pick, fully characterizing the pick itself (pick time, pick uncertainty, first motion polarity) in the same way a manual analyst would do. This allows the adoption of streamlined workflows in which the automatic (i.e. ML) picks would only be reviewed (and in most cases accepted) rather than re-picked from scratch by the human analyst (as currently done at SED).
How to cite: Jozinović, D., Clinton, J., Massin, F., Diehl, T., and Saul, J.: Advancing Operational Earthquake Monitoring at Local and Regional Scales with Machine Learning-Enhanced SeisComP Tools - as Demonstrated in Switzerland, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17226, https://doi.org/10.5194/egusphere-egu25-17226, 2025.
This work introduces a novel framework for full-waveform seismic source inversion using simulation-based inference (SBI). Traditional probabilistic approaches often rely on simplifying assumptions about data errors, which we show can lead to inaccurate uncertainty quantification. SBI addresses this limitation by building an empirical probabilistic model of the data errors using machine learning models, known as neural density estimators, which can then be integrated into the Bayesian inference framework. We apply the SBI framework to point-source moment tensor inversions as well as joint moment tensor and time-location inversions. We construct a range of synthetic examples to explore the quality of the SBI solutions, as well as to compare the SBI results with standard Gaussian likelihood-based Bayesian inversions. We then demonstrate that under real seismic noise, common Gaussian likelihood assumptions for treating full-waveform data yield overconfident posterior distributions that underestimate the moment tensor component uncertainties by up to a factor of 3. We contrast this with SBI, which produces well-calibrated posteriors that generally agree with the true seismic source parameters, and offers an order-of-magnitude reduction in the number of simulations required to perform inference compared to standard Markov chain Monte Carlo techniques. Finally, we apply our methodology to a pair of moderate magnitude earthquakes in the North Atlantic. We utilise seismic waveforms recorded by the recent UPFLOW ocean bottom seismometer array as well as by regional land stations in the Azores, comparing full moment tensor and source-time location posteriors between SBI and a Gaussian likelihood approach. We find that our adaptation of SBI can be directly applied to real earthquake sources to efficiently produce high quality posterior distributions that significantly improve upon Gaussian likelihood approaches.
How to cite: Saoulis, A., Piras, D., Spurio Mancini, A., Joachimi, B., and Ferreira, A.: Can you trust your uncertainties? Improving Bayesian earthquake source inversions using simulation-based inference, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6145, https://doi.org/10.5194/egusphere-egu25-6145, 2025.
Finding optimal wave sampling methods has far-reaching implications in wave physics, such as seismology, acoustics, and telecommunications. A key challenge is surpassing the Whittaker-Nyquist–Shannon (WNS) aliasing limit, establishing a frequency below which the signal cannot be faithfully reconstructed. However, the WNS limit applies only to periodic sampling, opening the door to bypass aliasing by aperiodic sampling. In this work, we investigate the efficiency of a recently discovered family of aperiodic monotile tilings, the Hat family, in overcoming the aliasing limit when spatially sampling a wavefield. By analyzing their spectral properties, we show that monotile aperiodic seismic (MAS) arrays, based on a subset of the Hat tiling family, are efficient in surpassing the WNS sampling limit. Our investigation leads us to propose MAS arrays as a novel design principle for seismic arrays. We show that MAS arrays can outperform regular and other aperiodic arrays in realistic beamforming scenarios using single and distributed sources, including station-position noise. While current seismic arrays optimize beamforming or imaging applications using spiral or regular arrays, MAS arrays can accommodate both, as they share properties with both periodic and aperiodic arrays. More generally, our work suggests that aperiodic monotiles can be an efficient design principle in various fields requiring wave sampling.
How to cite: Grushin, A. and Mordret, A.: Beating the aliasing limit with aperiodic monotile arrays, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7397, https://doi.org/10.5194/egusphere-egu25-7397, 2025.
Posters on site: Fri, 2 May, 14:00–15:45 | Hall X1
The Artius broadband node represents a transformative innovation in seismic instrumentation, designed to bridge the gap between traditional broadband seismometers and popular nodal systems. While broadband seismometers offer unparalleled sensitivity and frequency range, their cost and complexity often limit large-scale and dense deployments. Conversely, geophones provide cost-effective solutions for high-frequency applications but lack sensitivity to low-frequency seismic signals, which are critical for many research and monitoring purposes. Artius provides a cost-effective compromise, delivering the increased sensitivity and a true broadband frequency range at an economic price point.
Designed by Güralp Systems, Artius integrates a compact, highly sensitive broadband seismometer with an environmentally sealed anodized aluminium enclosure, ensuring optimal performance and robustness across diverse geophysical applications. Boasting a response of 30 seconds to 200 Hz, Artius greatly outperforms geophone-based systems while remaining perfectly suited to rapid, temporary deployments where it can be either pushed or staked into the ground and connected to an external power supply. Artius pushes the limits of versatility, facilitating real time data monitoring, as well as passive data collection. Artius also features an onboard SEEDlink server, ensuring compatibility with all standard seismological monitoring techniques and distinguishing it from existing solutions.
To manage the ever-growing seismic datasets generated by instruments like Artius, Güralp's Discovery software provides an advanced platform for seismic data acquisition, processing, and analysis. Discovery was designed to be proficient in handling large datasets and integrates seamlessly with Artius supporting both offline playback and real-time data analysis.
Artius is designed to be docked into an eight-node capacity docking station for data validation and mass data download. The docking station also serves as a “huddle” system for configuration and testing prior to deployment, ensuring each node is performing optimally prior to deployment. The Artius nodes are intended to be deployed in large arrays, perfect for passive seismology, ambient noise studies, and earthquake studies.
By combining advanced instrumentation like the Artius broadband node with innovative data processing capabilities provided by Discovery, Güralp Systems is advancing the frontiers of observational seismology. These technologies empower researchers to tackle new challenges in seismic data analysis.
How to cite: Mohr, S., Hill, P., Lindsey, J., Restelli, F., Watkiss, N., Calver, J., and Kerkenyakova, A.: Güralp Artius - A novel triaxial broadband node designed for Large-N arrays, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17158, https://doi.org/10.5194/egusphere-egu25-17158, 2025.
The recent advances in seismic data acquisition technology allow to transform fiber-optic cables into dense arrays of geophones that sample nearly continuously the seismic wavefield. This technology, known as Distributed Fiber-Optics Sensing (DFOS) (or Distributed Acoustic Sensing, DAS), has the potential to make cost-effective microseismic monitoring operations in borehole installations. Unlike conventional geophones, fiber-optic cables can be easily installed behind well casings without interfering with injection or production activities, eliminating the need for drilling dedicated monitoring wells.
Despite these benefits, DFOS data is generally characterized by higher noise levels when compared to conventional seismometers. The development of efficient denoising techniques is therefore a critical step to improve the Signal-to-Noise Ratio (SNR) of DFOS recordings, enhancing the capability to detect and analyse microseismic events. Traditional filtering techniques often struggle to recover low-amplitude signals, leading to limited noise reduction performances. In this study, we propose an effective denoising workflow based on an adaptation of spectral-subtractive algorithms, typically used in the context of speech enhancement for audio signals.
We validate this approach first simulating synthetic DFOS data resembling realistic data acquisition geometries and noise conditions. Then, our denoising workflow is further applied to real DFOS data recorded during the April 2022 stimulation campaign at the FORGE (US) EGS project.
Our results from both synthetic and real DFOS data show significant SNR improvements, showcasing the robustness of our method even when the original data show poor SNR conditions. This algorithm outperforms standard filtering techniques, offering a promising solution for enhancing DFOS data and improving the detection of previously hidden signals.
How to cite: Pascucci, G., Gaviano, S., and Grigoli, F.: A Speech Enhancement-based Method for Denoising Microseismic Distributed Fiber-Optic Sensing (DFOS) data, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1192, https://doi.org/10.5194/egusphere-egu25-1192, 2025.
Machine-learning algorithms are widely applied to facilitate human tasks. For instance, in seismology, they can help deliver high-resolution seismic catalogs including very small magnitude events that usually remain undetected by human analysts and by standard monitoring procedures.
The new frontier of modern seismology is to exploit deep neural networks (DNN) to automatically detect P- and S-wave arrival times to obtain good-quality seismic event locations in terms of hypocentral errors. Increasing the number of events and seismic phases is essential to build complete earthquake catalogs to be used in seismological analyses (such as seismic hazard estimation, seismic tomography, fault zone structure determination, rupture mechanism study, etc.).
Machine Learning methods are being integrated into large Research Infrastructures (RIs), like the European Plate Observing System (EPOS ERIC), which brings together 10 different scientific domains in Solid Earth Sciences. In this contribution, we present results from a specific Sponsored Research Activity promoted by the RI EPOS and dedicated to ML-driven methods for phase picking in seismic time series.
To ensure the correct recognition of seismic waves, neural networks trained on large and representative training datasets are essential. Here, we investigate the influence of the training dataset on the DNN PhaseNet performance, applying the method to three case studies. In the first case, the DNN trained with the Italian seismicity dataset (INSTANCE) is used to build a catalogue on the Fucino basin (Central Italy) study area; in the second case, we use the AQ2009 training dataset (based on the L’Aquila 2009 aftershocks) to analyse the 2016-2017 Amatrice-Visso-Norcia seismic sequence; and finally, the CREW training dataset (that contains P- and S-waves reflected on the mantle Earth and recorded at large epicentral distance) to detect P- and S- waves of teleseismic (regional) data acquired by the Adria Array project network.
The use of different training datasets greatly improves the performance of the neural network in recognizing P and S phases, reducing the number of false positives and providing more accurate and precise P and S-phase arrival times.
How to cite: Fonzetti, R., Bailo, D., De Gori, P., Valoroso, L., Anselmi, M., Bagh, S., Trani, L., and Chiarabba, C.: How do automatic phase pickings based on deep neural networks perform on different-scale case studies?, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5992, https://doi.org/10.5194/egusphere-egu25-5992, 2025.
The Central Apennines region (Italy) is regularly affected by seismic activity. In 2009, a Mw 6.1 earthquake occurred close to L’Aquila city. Although the crisis is over, there is still seismic activity in the region (ML<2.8). To improve our understanding of the structure and dynamics of seismogenic fault zones in Central Apennines, within the GRANDE (hiGh Resolution imAging of Normal faults Damage zonEs) experiment, we deployed nine dense linear arrays of seismic nodes (157 nodes) crossing two well studied fault damage zones at the surface (Campo Imperatore and Monte Marine fault systems). Previous geophysical and geological surveys precisely characterized these fault damage zones down to a few tens of meters (Fondriest et al., 2020; Cortinovis et al., 2024), but how this fault-related damage extend in depth is yet poorly understood. This temporary network was installed in May 2024, for a one-month period of recording in continue. We used a software using machine-learning to pick, detect and create a preliminary seismic event catalogue. Several tests were realized to check the nature of the event (anthropic or seismic) and the quality of detection. Then, we applied a relocation program to obtain a better location of the events and create a high-quality event catalogue for the one-month recording period. Moreover, we cross-correlated the signal at pairs of stations to retrieve a local tomographic picture of the studied areas. This new and original seismological dataset allow deepening our understanding of the structure of fault damage zones from surface to depth and improve our knowledge about dynamics of large earthquakes rupture, interseismic strain accumulation and release for one of the most hazardous faults in Europe.
How to cite: Vassallo, M., Basques, M., Pavoni, M., Tarantino, S., Barone, I., Fondriest, M., Di Giulio, G., Boaga, J., Poli, P., Courbis, R., and Hillers, G.: The GRANDE Array project: temporary seismic network in the Central Apennines, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18368, https://doi.org/10.5194/egusphere-egu25-18368, 2025.
Distributed Acoustic Sensing (DAS) technology offers a valuable opportunity to enhance seismological monitoring of ocean floors. Typically, in a standard monitoring network, most seismic stations are installed on land, with only a few ocean-bottom seismometers available. Consequently, monitoring oceanic seismicity is intrinsically challenging due to the lack of observations near seismic sources. In this context, DAS systems, for instance, those installed on fibers connecting islands and land-islands, can help bridge this observational gap. However, integrating the spatially dense DAS data (meter scale) with the often sparser seismometer data (kilometer scale) is not straightforward. Specifically, inverting DAS alongside seismometer P- and S-wave arrival times can lead to biased location results due to the numerical disparity between datasets and/or outliers not identified. Moreover, employing the full set of DAS arrival times can be computationally intensive, limiting its feasibility for real-time monitoring and integration into routine seismological software.
Automated weighting methods can help mitigate bias introduced by arrival time outliers in data inversion. This is particularly useful for DAS data, where user control over individual channels is limited. However, suppose the goal is to use DAS as a complementary tool to a seismometer network, and meter-scale spatial density is not essential. In that case, DAS data selection/sub-sampling can improve computational efficiency. To this end, we propose an approach that, for a given seismological network and DAS system, a) automatically identifies “reliable” DAS channels using a machine learning classifier trained on specific data attributes and b) further selects a subset of DAS channels to achieve similar interchannel spacing to the network. The proposed strategy generates a final set of DAS P- and S-wave arrival times with a number of observations comparable to the network. To test the benefits of this procedure on oceanic seismic monitoring, we use data from a fiber optic cable northeast of Gran Canaria and seismometers operated by the Instituto Geográfico Nacional (IGN) in the Canary Islands. We then compare event locations obtained using: a) IGN-only P- and S-wave arrival times, b) DAS-only P- and S-wave arrival times (unselected), and c) IGN and selected DAS P- and S-wave arrival times using the proposed method. Event locations are estimated using a hierarchical Markov Chain Monte Carlo approach.
Preliminary results show promising improvements in the location uncertainty of oceanic seismicity when the proposed data integration approach is applied. Additionally, a refined location catalog, incorporating a more detailed velocity model, is compiled using standard monitoring software alongside the proposed data selection approach.
How to cite: Bozzi, E., Piana Agostinetti, N., Ugalde, A., Cubas Armas, M., Rodriguez, T., Latorre, H., J. Vidal-Moreno, P., and Saccorotti, G.: A Pragmatic Approach for Integrating Submarine DAS Data with Onshore Seismometer Networks: A Case Study in the Canary Islands, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1193, https://doi.org/10.5194/egusphere-egu25-1193, 2025.
In the process of coal mining, pressure will be induced in the mining face, resulting in the stress concentration of surrounding rock, which will affect the safety and orderly normal operation of coal energy mining. The all-fiber optic micro-seismic monitoring technology has the advantages of high sensitivity, wide range and high safety, and can monitor the pressure activity of coal mine in real time. Taking the II1012 mining face of Taoyuan Coal Mine as the engineering background, the all-fiber optic micro-seismic monitoring work is carried out, and the data are analyzed by the methods of micro-seismic event detection, identification, classification and location. The characteristics of micro-seismic activity during the first pressure, periodic pressure and square pressure are studied. The results show that: Large energy events in the first pressure stage play a major role in roof failure, and the first pressure interval is 25.10 m. In the periodic pressure stage, the influence of micro-seismic activity on the roof is greater than that on the floor, but the influence on the floor is increasing. The large energy events increase significantly in the square pressure stage, which is easy to promote the frequent occurrence of high intensity and stress micro-seismic activities. The occurrence of micro-seismic events in mine pressure phases have advanced characteristics. There is a positive correlation between the intensity of micro-seismic activity and the rate of recovery, and the all-fiber optic micro-seismic has a good response to the mine pressure. The research work provides theoretical basis and technical support guidance for the safe production of the II1012 mining face in Taoyuan Coal Mine and other similar mining faces in other coal mines, reduces the risk of geological disasters caused by micro-seismic events during the pressure period, and further guarantees the safe and normal orderly development of the subsequent production work of the mining face. It is of great significance to the safe mining of coal energy and the supply of production and life.
How to cite: Wang, K.: Research on Coal Mine Pressure Characteristics Based on All-fiber Optic Micro-seismic Monitoring, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17795, https://doi.org/10.5194/egusphere-egu25-17795, 2025.
In recent decades, geothermal systems have gained increasing importance and attention. They have the potential to greatly contribute to the transition toward green energy and the establishment of a climate-neutral economy. Enhanced Geothermal Systems (EGS) represent a significant advancement in energy production methodologies. EGS utilize hydraulic stimulation techniques to inject and extract fluids, thereby enabling the harnessing of geothermal energy, which is crucial for electricity generation.
In addition to the existing natural seismicity, this production loop of hot and cold fluids may generate induced seismic events, specifically those caused by pressure changes that affect active faults or lead to stress variations within the rock volume. For these reasons, EGS could potentially induce medium to severe earthquakes that might impact nearby communities if not properly monitored and managed, or if strict monitoring methods are not followed to mitigate risks at EGS sites, particularly during operational stages.
Various physical and mechanical properties are recorded in real-time during operational stages. With the continual advancement of deep learning methods, these time series data can be analyzed individually and collectively for short-term forecasting of expected seismic magnitudes from future earthquakes.
Specifically, this work presents an experimental technique that leverages the spatio-temporal capabilities of graph neural networks by connecting these time series within a dynamic graph structure for short-term predictions of the maximum expected magnitude. This method is effective in identifying relationships that traditional approaches can sometimes overlook. Our preliminary results indicate that our algorithm can indeed assist in risk mitigation at EGS sites, potentially serving as a valuable complement to the current state-of-the-art frameworks (i.e., Traffic Light Systems, TLS) used globally.
How to cite: Bagagli, M., Grigoli, F., and Bacciu, D.: Leveraging Deep-Learning Methods for Operational Analysis at Enhanced Geothermal Systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17747, https://doi.org/10.5194/egusphere-egu25-17747, 2025.
Posters virtual: Mon, 28 Apr, 14:00–15:45 | vPoster spot 1
EGU25-2973 | ECS | Posters virtual | VPS21
Post stack inversion of seismic data based on Semi-supervised learningMon, 28 Apr, 14:00–15:45 (CEST) | vP1.5
Seismic inversion in geophysics is a method that uses certain prior information, such as known geological laws and well logging and drilling data, to infer the physical parameters of underground media, such as wave impedance, velocity, and density, from seismic observation data, and thereby obtain the spatial structure and physical properties of underground strata. Seismic inversion is a highly complex problem with multiple solutions, and with the advancement of collection equipment, the volume of geophysical observation data is increasing at an astonishing rate. This presents new challenges for the accuracy and speed of seismic data inversion methods. There is an urgent need to develop intelligent and efficient inversion technologies for seismic inversion.
Deep learning networks have powerful nonlinear fitting capabilities and can be used to solve complex nonlinear problems, such as seismic inversion. However, the predictive ability of deep learning networks largely depends on the quantity of training data. In the early stages of oil and gas exploration and development, the amount of well logging label data available for training is very limited, which poses a challenge for the application of deep learning in seismic inversion. Semi-supervised learning seismic inversion methods consider both data mismatch issues and well logging data mismatch issues, and can better adapt to inversion problems in real-world scenarios. Unlike supervised learning approaches, semi-supervised learning does not require a large amount of labeled data, thus it can better handle situations of data scarcity or mismatch.
This paper utilizes a semi-supervised learning workflow to perform inversion on post-stack seismic data and has conducted experimental validation on the Marmousi 2 model. The experimental results show that, compared to supervised learning networks, the semi-supervised learning network still exhibits good predictive performance with a limited amount of data, demonstrating better stability in the presence of noise and geological variations, and effectively learns the mapping relationship between seismic data and artificial intelligence. Furthermore, as the amount of training data increases, the performance of the network also improves, confirming the importance of data quantity for training deep learning networks. The application results of the network on actual data indicate that the network has broad application prospects and feasibility. However, since the network is based on a channel-by-channel inversion method, there is still a lack of representation in terms of lateral continuity, which requires further exploration and improvement in subsequent research.
How to cite: zou, C., zhang, J., tang, B., and huang, Z.: Post stack inversion of seismic data based on Semi-supervised learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2973, https://doi.org/10.5194/egusphere-egu25-2973, 2025.
