We used a deep learning workflow to enhance earthquake detection during the 2025 seismic unrest between Santorini and Amorgos islands to track the evolution of the crisis in near real-time. We analysed the continuous seismic waveforms daily (1/2 - 3/3/25) as the crisis unfolded. Our analysis enhanced the earthquake catalogue from around 4,000 to 80,000 earthquakes. The enhanced catalogue allowed this international expert group to identify the volcanic-tectonic character, clearly revealing burst-like, spasmodic seismicity swarms, which is a pattern associated with fluid-driven processes from early stages of the crisis. Detailed moment tensor inversions in early events characterised by a significant non-double couple component indicated the involvement of magmatic or high-pressure hydrothermal fluids driving the unrest. Concurrent DL-enhanced tomography efforts identified a third, deep magmatic reservoir beneath Anydros Islet, consistent with pressure-driven processes. To date, volcanic-tectonic swarms in which >200 earthquakes of ML > 4 occurred within only a few weeks, largely within episodic bursts of seismicity, have not been observed elsewhere.

Urban environments currently host more than 55% of the global population. However, the subsurface of such environments is not well characterized. Conventional geophysical surveys pose logistical challenges, such as restricted access and limitations on the use of active seismic sources, which impede such surveys to be performed. To overcome these challenges, Distributed Acoustic Sensing (DAS) can be deployed on existing, unused telecommunication fibre optic cables (dark fibres) and repurpose them as dense seismic arrays. In this way, the ambient seismic wavefield can be continuously recorded for high-resolution, passive subsurface imaging. However, the seismic noise field present in urban environments is complex, and mostly dominated by anthropogenic activity (i.e., trains, cars…), resulting in transient and moving seismic sources. Besides, dark fibres tend to have complicated layouts. A thorough understanding of the urban seismic noise field recorded by dark fibre DAS arrays is needed to understand the retrieved energy and its potential for seismic imaging.In this work, we investigate the interaction between diverse noise sources, complex fibre geometries and DAS directional sensitivity, and its impact on the application of ambient noise interferometry for imaging in complex urban environments. Our study area is located in the megacity of Istanbul (Türkiye), a highly densely populated urban area sitting in a region of high earthquake risk. The subsurface structure beneath Istanbul is poorly known, with very limited information available regarding subsurface material properties and faults directly underneath the city. Since May 2024, we have been continuously and simultaneously recording passive DAS data along two dark fibres located on the Eastern side of Istanbul; one crossing the densely populated district of Kartal and another one connecting the coastal section of Kartal with the Princess Islands archipelago, directly offshore.We start by analysing the ambient noise field recorded along fibre segments with different orientations and at diverse time periods; trying to isolate low-frequency seismic energy generated by natural sources. Combining measurements along both fibres, we apply beamforming approaches to understand the distribution of noise sources with respect to our array, and explore optimal channel-pair configurations to retrieve Rayleigh and Love waves, by taking into account the directional sensitivity of the DAS measurement. Ultimately, our goal is to develop a methodological framework for obtaining a reliable subsurface velocity model using seismic ambient noise in urban areas.

When a porous rock is subjected to overall compressive loading, either increasing pore pressure or decreasing confining pressure could result in rock failure. The stress path and the applied pressure change rate may affect the initiation and propagation of fractures within brittle materials. Understanding the physical mechanisms leading to failure is crucial for underground engineering applications and geo-energy exploration and storage. We conducted triaxial compression experiments on porous Bentheim sandstone samples at different stress paths and pressure change rates. First, at a constant confining pressure of 35 MPa and pore pressure of 5 MPa, intact cylindrical samples were axially loaded up to about 85% of the peak strength. Subsequently, the axial piston position was fixed, and then either the pore pressure was increased or the confining pressure was decreased at two different rates (0.5 MPa/min or 2 MPa/min), leading to final catastrophic failure. The mechanical results revealed that samples subjected to higher rates of decreasing effective confining pressure exhibited larger stress drop rates, higher slip rates, higher total breakdown work, higher rates of acoustic emissions (AEs) before failure, and higher post-failure AE decay rates. In contrast, the applied stress path did not significantly affect rock failure characteristics. Comparison of located AE events with post-mortem microstructures of deformed samples shows a good agreement. The AE source type determined from the P-wave first-motion polarity shows that shear failure dominated the fracture process when approaching failure. Gutenberg-Richter b-values revealed a significant decrease before failure in all tests. Our results indicate that, in contrast to the stress path, the rate of effective stress change strongly affects fracturing behavior and AE rate changes.

Urban areas are highly vulnerable to geohazards due to their dense populations and infrastructure, often resulting in severe consequences for human life and economic stability. Improving our understanding of near-surface and shallow subsurface structures in urban environments is therefore essential for effective seismic hazard assessment and risk mitigation. However, conventional geophysical surveys in cities are frequently limited by logistical constraints. In this context, repurposing existing telecommunication optical fibers (so-called dark fibers) as dense seismic sensing arrays using Distributed Acoustic Sensing (DAS) offers a powerful alternative for urban subsurface investigations.The megacity of Istanbul (Turkey) is located in one of the most tectonically active regions worldwide and is exposed to significant seismic hazard. Since May 2024, we have been continuously recording passive seismic data using DAS along an amphibious fiber-optic cable deployed in the urban district of Kartal (eastern Istanbul) and extending offshore. In this study, we focus on one month of data acquired along a 3 km-long urban segment of the fiber.Here, we exploit high-frequency urban noise for passive seismic interferometry. We analyze ambient seismic noise primarily generated by anthropogenic sources, such as urban traffic, in a frequency range up to 12Hz. We adapt ambient noise interferometry processing strategies to address the challenges posed by dense urban environments and DAS array geometries, including the selection of suitable fiber sections, channels, and source–receiver configurations. First, we retrieve high-frequency surface waves along different segments of the fiber. Then, we use these arrivals within an Eikonal tomography framework to map local phase velocities. Finally, we invert the surface-wave dispersion to constrain the shallow subsurface velocity structure, contributing to a better understanding of shallow structures and material properties relevant to seismic hazard assessment. Ultimately, this work aims to establish efficient methodologies for imaging the urban subsurface using existing infrastructure.

Urban areas are highly vulnerable to the impacts of geohazards due to their dense populations and complex infrastructure, with potentially severe consequences for human life and economic stability. Improving our knowledge of near-surface and shallow subsurface structures in urban environments is therefore essential for effective seismic hazard assessment and risk mitigation. However, conventional geophysical surveys in cities are often limited by logistical constraints, including strong anthropogenic activity, restricted access, legal limitations, and risks associated with instrument deployment. In this context, repurposing existing telecommunication optical fibers (so-called dark fibers) as dense seismic sensing arrays using Distributed Acoustic Sensing (DAS) offers a powerful alternative for urban subsurface investigations. This approach enables continuous, high-resolution seismic monitoring without the need for extensive field instrumentation.The megacity of Istanbul (Turkey) is located in one of the most tectonically active regions worldwide and is exposed to significant seismic hazard. Since May 2024, we have been continuously recording passive seismic data using Distributed Acoustic Sensing (DAS) along an amphibious fiber-optic cable, is deployed in the urban district of Kartal (eastern region of Istanbul) and immediately offshore. In this study, we focus on the 3 km-long urban segments of the fiber. We analyze ambient seismic noise generated by various anthropogenic sources, such as train and vehicle traffic and other urban activities, and evaluate their suitability for high-frequency, DAS-based passive seismic interferometry in a complex and heterogeneous urban setting.We develop and adapt processing strategies for ambient-noise interferometry that address the challenges of dense urban environments and DAS array geometries, including the identification of suitable fiber sections, channels, and source-receiver configurations, as well as preprocessing schemes designed for strongly anthropogenic noise.The objective is to retrieve high-resolution, urban-scale subsurface velocity models that improve our understanding of shallow structures and material properties relevant to seismic hazard. Ultimately, this work aims to establish efficient methodologies for imaging the urban subsurface using existing infrastructure, contributing to improved geohazard assessment and supporting sustainable urban development in seismically active regions.

A total of 600 m3 of fresh water was injected into the 4 km-deep KTB pilot borehole near Windischeschenbach, Germany, in November 2023. The main goal of the GEOREAL hydraulic stimulation experiment was to stimulate fractured metamorphic rocks in the vicinity of the well-characterised SE2 fault zone at 4 km depth with high flow rates and to compare rock and fluid responses to a previous one-year-long injection experiment in 2004/5 into the same formation. The intention was to quantify the potential for heat extraction at a site where temperature and pressure conditions are representative for crystalline rocks for large parts of Germany.Due to an unforeseen incident during packer deployment, however, a reduced fluid volume of 600 m3 was injected, pressurising the entire pilot hole (KTB-VB), including the cased section to 3.85 km depth and the 150 m open borehole section below. Flow rates were variable ranging from 10 to 220 l/min. Pressure records were obtained at the well heads of KTB-VB and the main borehole (KTB-HB) at 200 m lateral distance at the surface which was also used for downhole microseismic monitoring. Hydraulic data were analysed and compared to results from 2004/5. The hydraulic parameters show similar values as in 2004/5 when no packer was blocking the borehole. Due to an unknown leak in the casing cement of the 35-year-old KTB-VB, the GEOREAL experiment had to be stopped earlier than planned. No microseismic events were detected during the injection using conventional event-detection methods. Hydraulic results for the KTB-VB injections two decades apart suggest that no significant permeability changes have occurred in the SE2 fault zone indicating that fracture healing does not play a major role in the reservoir. The pressure recordings from the KTB-HB showed no change caused by the injection into the KTB-VB, suggesting that either the injected volume was too small to cause a hydraulic response, too much fluid leaked into the formation, the hydraulic connection observed previously ceased to exist, its geometry is unfavourable for this configuration, or hydraulic diffusivity is smaller than suggested for previous experiments.

Predicting large earthquakes remains a significant challenge due to the complexity of fault systems and the variability of preparatory processes. We introduce an unsupervised machine learning framework to categorize seismicity patterns and identify, when present, seismicity transients preceding large earthquakes. We focus on five large earthquakes and extract seismo-mechanical features per families of events, defined as clustered events in space, time and magnitude. Here we show that for those cases displaying a preparatory phase, specific long-lasting families belonging to a critical category signalling an upcoming earthquake occur during the preparatory phase. Compared to other periods, critical categories reflect a higher spatial-temporal localization, earthquake interaction and strain release. The method will not detect such a transient for earthquakes with no detectable seismic preparatory phase. Finally, we demonstrate that the method is capable of identifying preparatory phases (when present), showing potential for operational earthquake forecasting.

Coastal areas are among the most densely populated areas on Earth, with 50% to 70% of the population projected to live in these regions in the next 50-100 years. Many large cities are located along tectonically active coastal areas, and the combination of increasing population, sea-level rise, and extreme weather events expose coastal regions to significant geohazard risk. Therefore, detailed characterization of the structure, physical properties and dynamics of the shallow subsurface in coastal urban areas is critical for geohazard assessment and mitigation. However, this task remains challenging, mostly due to limited access to the subsurface for the deployment of conventional sensors. In this context, Distributed Acoustic Sensing (DAS) deployed on existing, unused (“dark”) telecommunication networks offers an unprecedented opportunity to efficiently investigate subsurface seismic structure at high spatial and temporal resolution over tens of kilometers.In this study, we establish an amphibious fiber-optic sensing testbed to investigate the subsurface structure and dynamics of the megacity of Istanbul (Türkiye) and the eastern Marmara Sea, one of Europe's highest earthquake risk areas. Istanbul is located approximately 20 km north of the North Anatolia Fault Zone (NAFZ), one of the World's most active faults. Since 2015, the GFZ Helmholtz Centre for Geosciences is operating the Geophysical Observatory at the Northern Anatolian Fault (GONAF) in collaboration with the Turkish Disaster and Emergency Management Presidency (AFAD). The observatory consists of 10 boreholes equipped with seismometer strings and partly with strainmeters, providing key information on seismicity and deformation processes in the Marmara Sea. Despite this efforts, high-resolution imaging of the NAFZ, and continuous recording of near-fault seismicity, aseismic deformation and slow-slip events remains challenging. Detailed data on near-city fault complexity, potential hidden faults directly underneath the urban area, and the spatial variability of subsurface material properties at high resolution is also still lacking. By integrating fiber-optics sensing, we expand and enhance the observatory by simultaneously providing critical data on offshore fault structure and seismicity and enabling efficient investigation of structure and seismic hazard along the coast.Since May 2024, continuous passive seismic data have been recorded along two dark fibers in eastern Istanbul: a 17 km-long cable crossing the coastal district of Kartal, and a 34 km-long cable immediately offshore, connecting the coast with the Princess Islands. Both natural (i.e. ocean waves) and anthropogenic (traffic) seismic noise, as well as local and regional earthquakes have been captured by both fibers, enabling the characterization of the testbed and its potential and limitations. We apply ambient seismic noise interferometry approaches across multiple spatial scales and frequency bands for multi-resolution imaging, and explore the potential for temporal monitoring of subsurface variations associated with earthquake processes and environmental changes. We also assess the capabilities of the testbed to detect near-fault seismic events and improve seismicity catalogs. Ultimately, our study will provide a framework to leverage dark fibers in densely populated coastal areas for efficient subsurface imaging and near-fault monitoring, with significant potential to improve geohazard assessment.

Distributed Acoustic Sensing (DAS) deployed on submarine telecommunication cables provides continuous and spatially dense measurements in offshore environments where conventional instrumentation is sparse. In this study, we analyze DAS data recorded along a submarine fiber-optic cable in the eastern Marmara Sea (Türkiye), located in close proximity to the North Anatolian Fault.The dataset consists of continuous strain-rate recordings along a ~34 km-long cable connecting the Istanbul mainland to the Princes’ Islands, which is a component of the integration of fiber-optic sensing into GONAF (Geophysical Observatory of the Northern Anatolian Fault) operated by GFZ in collaboration with the Turkish Disaster and Emergency Management Authority (AFAD). Since May 2024, passive DAS data have been continuously recorded along the marine cable in the Marmara Sea. As an initial step, we focus on understanding the cable geometry and data characteristics, including channel selection, spatial variability, and waveform behavior along the fiber.Clear and coherent wavefields are observed for multiple events, allowing the tracking of seismic wave propagation along the cable over tens of kilometers. Variations between cable segments indicate differences in coupling conditions and local recording characteristics.Possible saturation effects are currently being investigated in the analyzed recordings. So far, no obvious signal clipping has been observed within the current data range, although further quantitative analysis and recordings of stronger ground motion are required to better evaluate the dynamic response of the system.These first observations highlight the potential of submarine DAS for offshore seismic monitoring and provide a basis for future studies focusing on earthquake detection, characterization, and integration with existing seismic networks.

Urban sustainable development and improved resilience to geohazards requires an exhaustive understanding of the geological structure, physical properties and dynamics of the shallow subsurface underneath urbanized areas at the sub-kilometer scale. Yet, our current understanding of the urban subsurface is limited by our ability to image its structure and temporal variations at high resolution using classical geophysical approaches. Recently, the application of conventional ambient noise interferometry analysis to dynamic strain data recorded using Distributed Acoustic Sensing (DAS) deployed on unused telecommunication fiber-optic cables (dark fibers) has emerged as an attractive alternative for cost-efficient, regional scale (10’s of km) seismic imaging and monitoring at high spatial and temporal resolution. Still, its application to urban environments remains vastly underutilized. One of the most significant hurdles is the lack of adequate and efficient data exploration and processing tools to address and harness the unique challenges associated with DAS-based urban seismic noise recordings, which include the complexity of the noise field, unconventional array geometries and non-uniform coupling conditions, and increasingly massive data volumes. The InDySE project (Interrogating the Dynamic Shallow Earth) aims at addressing these challenges to develop and validate the next-generation of subsurface imaging and monitoring platforms in urban areas based on the combination of existing fiber-optic networks and high-frequency infrastructure noise. The project comprises (1) developing high-performance computational tools, advanced processing workflows and machine learning approaches for efficient data exploration, selection and processing using existing datasets, (2) field experiments in target areas to retrieve high-resolution velocity models and monitor changes in seismic velocities, (3) integrating the resultant high-resolution seismic models with complementary datasets such as deformation maps derived from InSAR measurements. One of the selected study areas is the highly populated metropolitan area of Istanbul (Turkey), where understanding the structure, properties and dynamics of the shallow subsurface at high-resolution is critical for evaluating geohazard exposure. Among our objectives will be illuminating potential hidden faults underneath the city, obtaining high-resolution maps of geological materials and subsurface properties that can be translated into maps of local site response to large earthquakes, and tracking seismic velocity changes linked to hydrological dynamics that are known to be responsible for ground movements such as landsliding and subsidence. Ultimately, InDySE aims at developing efficient approaches for using dark fiber and ambient noise in urban subsurface investigations with implications in geohazard assessment.

Abstract We investigate the influence of fault roughness on physical damage prior to large laboratory rock failure and the evolution of the local stress field surrounding the fault zone as macroscopic shear slip approaches. To achieve this, we analyze acoustic emission (AE) data from displacement‐driven rock friction experiments conducted on porous sandstone samples containing either a saw‐cut (smooth) or a rough fault. Using high‐quality AE‐derived focal mechanisms and two stress tensor inversion approaches–one considering double‐couple (DC) components and the other one incorporating non‐DC components, we examine the temporal evolution of the local stress tensor for both smooth and rough faults. Our results show no significant differences between the two stress inversion methods, indicating that non‐DC components have no significant influence on the resulting stress tensors in our experiments. As macroscopic shear slip approaches, the principal stress axes surrounding the fault zone gradually rotate, regardless of the initial fault roughness. The observed evolution of stress tensors correlates with the evolving partitioning between volumetric and shear deformation, as derived from moment tensor inversion of AEs. Compared to the smooth fault, the rough fault exhibits higher local stress heterogeneity and more erratic fluctuations in AE source‐related parameters as loading progresses.

In rocks and other consolidated geomaterials, static or dynamic excitation leads to a fast softening of the material, followed by a slower healing process in which the material recovers all or part of its initial stiffness as a logarithmic function of time. This requires us to exit the framework of time-independent elastic properties, linear or not, and investigate non-classical, non-linear elastic behavior and its time dependency. Softening and healing phenomena can be observed during seismic events in affected infrastructure as well as in the subsurface. Since the transient material changes are not restricted to elastic parameters but also affect hydraulic and electric parameters as well as material strength – documented for instance by long lasting changes in landslide rates – it is of major interest to characterize the softening and recovery phases. To characterize this behavior in a controlled environment, we perform experiments on Bentheim sandstone in a Materials Testing System triaxial cell with pore pressure and confining pressure control. Our sample is subjected to various static loading cycles in both dry and water-saturated conditions, while an active acoustic measurement setup allows us to monitor minute P-wave velocity changes, which can then be directly tied to dynamic elastic modulus changes. Our transducer array allows us to observe the dynamic softening as well as the recovery processes in the sample during repeated loading phases of various time lengths. Observations indicate high spatial, frequency and lapse-time sensitivity of the observed velocity changes, indicating a rich landscape of concurrent effects and physical phenomena affecting our sample during these simple experiments. To investigate the spatial and directional dependency of the velocity changes, we restrict the analysis to direct and reflected ballistic waves. Our observations indicate that, while stress-induced classical effects are clearly anisotropic as expected, the non-classical effects do not exhibit significant anisotropy. This allows us to rule out a number of physical phenomena as the cause for the non-classical effects. Most importantly, we conclude that the microscopic structures responsible for the reversible softening and healing processes are different from the cracks that induce the anisotropic acousto-elastic effect.

Abstract Recent advances in artificial intelligence have enhanced the detection and identification of transient low-amplitude signals across the entire frequency spectrum, shedding light on deformation processes preceding natural hazards. This study investigates low-frequency, low-amplitude signals preceding the 2023 MW 7.8 Kahramanmaraş earthquake in Türkiye. Using a deep neural network, we extract key features from the spectrograms of continuous seismic signals and employ unsupervised clustering to reveal distinct transient patterns. We identify an increased occurrence of low-frequency tremor-like signals during the six months preceding the mainshock. However, the location of these signals suggests that their origin is not tectonic, but rather related to anthropogenic activities at cement plants along the Narlı Fault, where the MW 7.8 mainshock nucleated. Such findings highlight the importance of understanding the origin of patterns detected by machine-learning methods and the large variety of seismic signals due to anthropogenic activities. Furthermore, the search for the origin of the tremor-like signals motivated an investigation into the local seismicity around the Narlı Fault. The resulting extended seismicity catalog suggests that seismicity in this area arises from a combination of tectonic and anthropogenic processes.

Rock avalanches are catastrophic events that can be triggered by various geological and climatic factors. Large-scale rock avalanches have been observed near fault zones, indicating a potential relationship between fault creep motions and the initiation of rock avalanches. This study proposes a physical mechanism that explains how fault creep leads to the initiation of near-fault rock avalanches through stress redistribution. The Muztag rock avalanche which occurred near the Muztag fault in Muztag Ata, Kashgar, China is revisited using field measurements and numerical modelling. We consider the Muztag rock avalanche model with its initial slope toe supported by a portion of the fault's hanging wall. Site-specific numerical simulations using discontinuous deformation analysis (DDA) reveal that fault creep initially generates localized stress concentrations at the slope toe. Once the slope toe completely detaches from the hanging wall, the slope toe fails when the accumulated stress exceeds the local strength, resulting in a sharp stress drop. This stress redistribution triggers consecutive failure at the bottom of the rear edge and middle slope, forming a through-going shear sliding surface, which leads to the initiation of the overall rock avalanche. The kinematic processes of the avalanche, including sliding distance and deposit thickness, as modelled by DDA, are consistent with the post-failure characteristics of the Muztag rock avalanche. These findings suggest that tectonic fault creep motions can play a significant role in inducing near-fault rock avalanches.

Fault Zone Head Waves (FZHW) are a key diagnostic tool to identify bimaterial interfaces along fault zones. We detect and analyse FZHW recorded in the waveforms from the local MONGAN (MONitoring of the GANos Fault) seismic network along the Ganos section of the North Anatolian Fault Zone, northwestern T & uuml;rkiye, between October 2017 and July 2019. MONGAN covers the Ganos fault with different interstation distances ranging from 25 m to similar to 4 km. To detect FZHW, an automatic detector is used as a preliminary analysis method followed by manual revision and particle-motion analyses to distinguish between FZHW and direct P waves. FZHWs are predominantly detected at the southern side of the fault. The observed FZHWs have a moveout (triangle t) with respect to the direct P arrivals, increasing with distance travelled along the fault and indicating a deep bimaterial interface down to the bottom of the seismogenic crust. The average velocity contrast is estimated to be 5.9 per cent across the fault. Near fault-recordings indicate that the Ganos Fault is offset by similar to 250 m with respect to the surface trace obtained from literature. To a lesser extent, FZHW are also observed in the northern stations from the fault, indicating a shallow wedge-shaped low-velocity portion constituted by highly fractured material to either side along the southwestern section of the Ganos Fault between the fast Eocene block to the north and the slow Miocene block to the south. The seismic velocity contrast and geological complexity have important implications for the rupture evolution during future earthquakes on the Ganos fault in that they would progress predominantly westward, away from Istanbul and Tekirda & gbreve;. Furthermore, an asymmetric aftershock distribution skewed to the northern block can be expected, with subsequent implications for site-dependent risk there. Our results allow to revise focal mechanism solutions by separating FZHW from direct-P wave for previous Sea of Marmara earthquakes.

A longstanding question in geoscience concerns whether earthquakes show a preparatory process and precursory seismic activity. Some models hold that in the intermediate-term (from months to years), seismicity and/or aseismic transients in fault slip and in other fault properties occur. During the last decades, improvements in earthquake monitoring, the integration of geodesy capturing slow deformation, and the incorporation of novel data analysis techniques including machine learning and artificial intelligence have improved our ability to better discern how earthquake sequences evolve before a mainshock. The few available observations of transient deformation preceding well-recorded earthquake sequences show a high variability, thus our potential for improving earthquake forecasting is still limited. The body of knowledge available from mechanical models, numerical simulations, experimental work and field observations highlighted a wealth of structural, tectonic and boundary conditions which may control the dynamics of earthquake sequences. These suggest that several processes can affect earthquake preparation on different temporal and spatial scales, ultimately yielding highly varying transient observations prior to mainshocks. These observations also highlight that existing theoretical and conceptual models of the preparation/nucleation process may not fully capture the governing physics.We analyzed seismicity transients prior to the occurrence of the 2023, MW 7.8 Kahramanmaraş/Türkiye earthquake. We identified seismic precursory activity composed of a handful of isolated spatio-temporal clusters occurring in a complex fault network within 65 km of the future earthquake epicenter. Some of these clusters contributed to acceleration of seismicity rates in an area surrounding the future mainshock and starting ca. 8 months before the event. Within that area, we also observed a decrease in Gutenberg-Richter b-values. Comparable seismic transients were not observed in the region at least since 2014. The complex preparatory process differs significantly from the cascade of close (

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

Earthquake forecasting is a highly complex and challenging task in seismology ultimately aiming to save human lives and infrastructures. In recent years, Machine Learning (ML) methods have demonstrated progressive achievements in earthquake processing and even labquake forecasting. Developing a more general and accurate ML model for more complex and/or limited datasets is obtained by refining the ‘ML models’ and/or enriching the ‘input data’. In this study, we present an event-based approach to enrich the input data by extracting spatio-temporal seismo-mechanical features that are dependent on the origin time and location of each event. Accordingly, we define and analyze a variety of features such as: (a) immediate features, defined as the features which benefit from very short characteristics of the considered event in time and space, (b) time-space features, based on the subsets of acoustic emission (AE) catalog constrained by time and space distance from the considered event, and (c) family features, extracted from topological characteristics of the clustered (family) events extracted from clustering analysis in different time windows. We use AE catalogs recorded by tri-axial stick-slip experiments on rough fault samples to compute event-based features. Then, a random forest classifier is applied to forecast the occurrence of a large magnitude event (MAE>3.5) in the next time window. Results show that to obtain a more accurate forecasting model, one needs to separate background and clustered activities. Based on our results, the classification accuracy when the entire catalog data is used reaches 73.2%, however, it shows a remarkable improvement for separated background and clustered populations with an accuracy of 82.1% and 89.0%, respectively. Feature importance analysis reveals that not only AE-rate, seismic energy and b-value are important, but also family features developed from a topological tree decomposition play a crucial role for labquake forecasting.

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