Potential earthquake-prone faults identified by dense seismic array monitoring in complex extensional settings

Potential earthquake-prone faults identified by dense seismic array monitoring in complex extensional settings | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Potential earthquake-prone faults identified by dense seismic array monitoring in complex extensional settings Francesco Scotto di Uccio, Titouan Muzellec, Antonio Scala, Grazia De Landro, and 8 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6727679/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 17 Jan, 2026 Read the published version in Scientific Reports → Version 1 posted 12 You are reading this latest preprint version Abstract Identifying active faults capable of generating moderate-to-large earthquakes is essential for seismic hazard assessment, yet remains challenging in extensional tectonic environments, where fault systems include multiple segments and bends accommodating strain. In this study, we demonstrate how a short-term deployment of densely distributed seismic arrays can provide critical insights into seismicity patterns and fault geometry in the Southern Apennines, Italy. Integrating arrays with advanced machine learning methodologies, we produce an enhanced seismic catalog that increases the content of the manual one by nearly one order of magnitude, achieving, in just one year, a resolution comparable to a decade of conventional monitoring. Our results reveal spatial consistency of seismicity down to the decameter scale, with hypocenter locations and b-value mirroring those from the previous decade. We distinguish a shallow, diffuse seismicity, likely influenced by hydrological loading from karst aquifers, from deeper seismic clusters characterized by greater spatial coherence. The distribution of deep seismicity, when integrated with a 3D tomographic model, delineates a complex, curving fault structure 50–60 km long, featuring a right-stepping jog several kilometers wide. Dynamic rupture simulations suggest that earthquakes nucleating on this fault could propagate through these structural complexities, potentially generating earthquakes up to magnitude 7.0. Earth and environmental sciences/Solid earth sciences/Geophysics Earth and environmental sciences/Solid earth sciences/Seismology Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Understanding the preparation and nucleation processes that drive the space-time evolution of moderate-to-large earthquake ruptures is essential for characterizing the conditions controlling earthquake initiation and growth. This knowledge is essential both for predicting the seismic wavefield generated during earthquakes and for mitigating seismic risk through targeted safety measures. These mitigation actions can be implemented on timescales corresponding to earthquake recurrence intervals (Kanamori, 2014 ; Bommer, 2021 ) and over shorter timescales during rupture propagation, as part of earthquake early-warning systems (Zollo et al., 2009 ; Allen & Melgar, 2019 ). In this framework, detailed information on fault geometry and mechanical conditions provides crucial information that can reduce epistemic uncertainty in ground-motion modeling and seismic hazard assessment (Maufroy et al., 2012). However, anticipating the geometry and even the presence of seismogenic faults presents a challenge, as demonstrated by many recent large earthquakes that occurred on previously unknown faults, such as those responsible for the 2019 M 7.1 Ridgecrest earthquake in California (Magen et al., 2020 ). This issue is even more challenging in extensional settings, where normal fault systems can be highly complex and comprise subparallel, segmented synthetic and antithetic structures that interact mechanically over diverse space and time scales. (Buttinelli et al., 2021 ; Bello et al., 2022 ). In the actively extending Apennines chain, even moderate-size events with magnitudes ranging from 6 to 6.5, that rupture 10 to 30 km-length faults, may lead to extensive casualties and building damage, as illustrated by the 2009 L’Aquila earthquake (Tertulliani et al., 2012 ) and the 2016 Amatrice-Norcia sequence (Galli et al., 2017 ). To better understand the geometry and stress state and to assess the risk related to moderate-size earthquake faults, dense multi-parametric monitoring infrastructures referred to as Near Fault Observatories, have been deployed in the Central and Southern Apennines of Italy and across Europe over the last 15 years (Chiaraluce et al., 2022 ). In this study, we focus on the multi-segmented Irpinia fault system, in the Southern Apennines, that generated the 1980 Ms 6.9 earthquake (Bernard & Zollo, 1989 ). This earthquake ruptured at least three fault segments, each of which were tens of km long over more than 40 s duration, resulting in a long-duration strong ground shaking that caused widespread building collapse and over 3000 fatalities (Rovida et al., 2019 ). The Irpinia region is classified as one of the highest seismic hazard areas in Italy (MPS Working group, 2004) with a relatively short return period for M 6 + earthquakes (Galli, 2020), and a probability greater than 30% of a M5.5 + earthquake occurring within the next decade (Cinti et al., 2004 ). In 2005, the Irpinia Near Fault Observatory (INFO) was created with the aim of developing a large research infrastructure in Earth Sciences to monitor the Irpinia fault system by a dense seismic network consisting of 39 stations, with inter-station distances ranging from 10 to 20 km (Iannaccone et al., 2010 ). This network allows for a local magnitude of completeness in seismic catalogs of \(\:{M}_{l}\:\) 1.1 and a magnitude detection threshold down to \(\:{M}_{l}\:\) 0.5 in its central sector (Vassallo et al., 2012 ). The epicenters of the earthquakes occurring from 2008 till September 2021 are reported as blue circles in Fig. 1 a, along with the stations of the INFO (yellow triangles in Fig. 1 a). Analysis of the background microseismicity reveals a rather sparse seismicity within the graben bounded by the main faults responsible for the 1980 Irpinia earthquake (De Landro et al., 2015 ). Occasionally, microseismicity clusters in seismic sequences lasting for a few days with modest magnitude mainshocks, that rupture smaller, sub-parallel structures to those activated during the Ms 6.9 event in 1980 (Scotto di Uccio et al., 2023; 2024). The tomographic modeling of first P and S arrival times (Amoroso et al., 2014 ) and ambient noise data (Vassallo et al., 2016 ) suggest that the area is permeated by deep fluids, predominantly CO₂ and brine (Amoroso et al. 2017 ). Additionally, geodetic data modeling reveals a non-linear elastic response of the shallow karst aquifers to hydrological loading, with the opening and closing of cracks correlated with fluctuations in seismicity rates in the area (D’Agostino et al., 2018; Tarantino et al., 2024 ). The long-term monitoring of microseismicity has provided evidence for structural segmentation and evolution of both crustal and source properties, constrained properties of the crust (De Landro et al., 2022 ), informed measurements of the apparent stress (Picozzi et al., 2019 ), and stress drop (Picozzi et al., 2022a ), and constrained the spatial variability of ground motion intensity (Picozzi et al., 2022b ). Uncertainties in earthquake location, however, hinder the clear identification of causative fault structures, making it ambiguous whether the sparse hypocentral distributions are due to limitations in resolution or to a genuinely chaotic orientation and spatial distribution of small structures hosting the microseismic events (De Landro et al., 2015 ; Camanni et al., 2025 ). To investigate new technological solutions aimed at improving seismic monitoring capability, we deployed a constellation of 20 small-aperture seismic arrays, each consisting of 10 stations (200 stations in total) integrating INFO during the period September 2021 - July 2022, in the framework of a temporary experiment, named DETECT (DEnse mulTi-paramEtriC observations and 4D high resoluTion imaging, red triangles in Fig. 1 a). DETECT is an international monitoring project coordinated by GFZ and the University of Napoli Federico II, that involved several Italian University and research institutes (INGV, CNR, Università di Salerno, and Università del Sannio). Within the DETECT arrays, the average inter-station distance ranged from several hundred meters to one kilometer, while the average distance between arrays was approximately 10 km. Each array was equipped with one broadband seismometer, one 1Hz sensor, and eight short-period (4.5 Hz natural frequency) geophones. The data collected were processed, standardized, and made publicly available (Strollo et al., 2025). The main goal of DETECT was the generation of enhanced catalogues of accurately located microseismic events. Machine learning (ML) and similarity-based methods have shown significant potential for increasing the size of seismic catalog by up to an order of magnitude, uncovering previously uncatalogued lower-magnitude events, even in areas for which the events were not included in the training dataset (Zhu & Beroza, 2018; Mousavi et al., 2020 ; Tan et al., 2021 ). However, these approaches can suffer from a high rate of false positives and missed detections. Such limitations can be mitigated using well-designed ML models trained on region-specific datasets (Michelini et al., 2021 ) or by assisting more conservative ML detectors with robust, network-based similarity techniques (e.g., Chamberlain et al., 2018 ), where events identified by ML algorithms can be used as templates for similarity-based detections (Scotto di Uccio et al., 2023). Indeed, template matching algorithms target earthquakes occurring close to a known set of events and typically achieve a better detection performance for earthquakes reporting low signal-to-noise ratio (Vuan et al., 2018 ; Essing & Poli, 2022 ; Scotto di Uccio et al., 2023). Accurate hypocenter determinations of earthquakes in ML-enhanced catalogs have shown to provide significant insights into fault geometries (Spallarossa et al., 2021 ; Michele et al., 2020 ; Cabrera et al., 2022 ; Park et al., 2022 ; Scotto di Uccio et al., 2024). For generating the seismic catalog, we followed the detection strategy proposed by Scotto di Uccio et al. (2023), integrating the deep learning detector EQTransformer (Mousavi et al., 2020 ; hereinafter EQT) and the template matching technique EQCorrscan (Chamberlain et al., 2018 ; hereinafter TM), using the enhanced catalog provided by EQT as the template set for the TM detection. During the detection stage, the dense array deployment was split into overlapping subnetworks allowing us to focus on the volume beneath the considered arrays, increasing the ability to detect local and low-magnitude events while limiting the false declarations caused by coherent non-seismic transients at distant stations (see Methods). We performed accurate hypocenter locations by refining the automatic phase arrival times among families of similar waveforms identified using hierarchical clustering (Muzellec et al., 2025 ). We used NLLoc (Lomax et al., 2009 ) and a 3D velocity model tailored for the area (De Landro et al., 2022 ) for obtaining an improved absolute hypocenter location. The locations were refined by performing double difference location using HYPODD (Waldhauser & Ellsworth, 2000 ), introducing catalog and cross-correlation differential travel times for highlighting the small-scale features of the fault structures. Location uncertainties based on bootstrap analysis of the seismic catalog resulted in uncertainties of a few hundreds of meters (Muzellec et al., 2025 , see Methods). Results The application of these techniques to continuous data from DETECT led to the identification of approximately 3,600 earthquakes occurred during the 11-months experiment duration. This enhanced catalog represents an ~ 8-fold increase in the number of earthquakes compared to the existing catalog provided by INFO within the same time window. The existing catalog is based on the use of a conventional network layout and seismic detection methods, as the visual inspection of records by operators analyzing data from the standard network stations (yellow triangles in Fig. 1 a). Focusing on the contribution of individual methodologies in detecting low-magnitude earthquakes, we found that the machine learning-generated catalog expanded the manual catalog by a factor of ~ 4. This represents a twofold improvement over the results reported by Scotto di Uccio et al. (2023), who applied the same deep learning detector to seismic sequences recorded by the INFO network. This highlights the efficacy of dense constellations of arrays in amplifying the detection capability of machine learning models for low-magnitude seismic events. Furthermore, the detection catalog derived from advanced strategies applied to dense arrays exhibits an earthquake count on the same order of magnitude as multiple years of conventional monitoring (Fig. 1 b). To characterize the earthquakes in the enhanced catalogs, we performed accurate relative event relocation. Refinement of phase arrival times using hierarchical clustering (Muzellec et al., 2025 ) yielded average corrections within ~ 0.1 s for both P and S picks (Figure S1 ). This result not only supports the high accuracy in the automatic identifications of phase arrival times but also provides critical adjustments for constraining the fine-scale properties of seismogenic sources. We successfully relocated 2,248 earthquakes (~ 65% of the detected events), nearly doubling the fraction of relocated events in enhanced catalogs for seismic sequences (Scotto di Uccio et al., 2024) and significantly surpassing typical relocation percentages in template-matching-derived catalogs (Ross et al., 2019 ; Cabrera et al., 2022 ). This improvement is attributed to the deployment of the dense seismic arrays, which facilitates the detection of a larger number of phase arrival times, even for low-magnitude events (~ 47,000 P-wave and ~ 72,000 S-wave picks). The increased number of picks resulted in a substantial dataset of differential travel times, which are fundamental to achieve high-precision earthquake locations. The left panel of Fig. 1 illustrates the epicenters of the relocated earthquakes, showcasing the spatial distribution of the relocated seismicity observed during DETECT (shaded green points) compared to the seismicity detected by INFO (blue points) from 2008 up to the start of the dense array monitoring (September 2021). The spatial distribution of epicenters from the enhanced catalogs reveals variable seismicity levels across different sectors of the Irpinia region. Most of the seismicity is concentrated in the Central and Southern sectors, while the Northern sector exhibits significantly lower activity, predominantly characterized by isolated events and a single swarm-like sequence of 30 earthquakes associated with a \(\:{M}_{l}\:\) 2.1 event located outside the coverage area of DETECT (northernmost green cluster in Fig. 1 a). When comparing the seismicity observed during the experiment with that recorded by INFO between 2008 and September 2021 (the start time of the experiment), we find consistency in the distribution of epicenters. This highlights the persistent occurrence of microseismicity in areas prone to generating moderate-to-low magnitude earthquakes along the Irpinia fault system. Most of the earthquakes recorded by the dense arrays occur in regions reporting a high density of events already identified in the manual catalog. We use the Chamfer distance (Yu et al., 2025 ) to measure the connection between relocated hypocenters from this analysis and the seismicity depicted by the long-term INFO catalog. We obtain an average nearest-neighboring distance of 0.9km and 1.1km for earthquakes occurring in the Central and Southern sectors, respectively, lower than what observed from the earthquakes in the INFO catalog during the DETECT experiment (1.3km and 1.6km, respectively). The low seismicity rate observed in the Northern sector, which is typically interested by the occurrence of seismic sequences populating the INFO catalog, prevents from extracting robust estimates. Moreover, an absence of seismicity is evident in the Southeastern edge of the region (Fig. 1 a), a feature previously reported by De Landro et al. ( 2015 ) inspecting a catalog of six years of microseismicity with moment magnitude ranging between 0.9 and 3.1. This phenomenon has been attributed to the presence of a contact-zone between geological units with differing rheological properties in response to the NE–SW stress regime acting on the chain. To evaluate the statistical parameters of the enhanced catalog, we calculated the local magnitudes of the events using a scale calibrated for the area (Bobbio et al., 2009 ), resulting in values ranging in \(\:{M}_{l}\) [-1.9 to 2.9]. We estimated the statistical parameters of the Gutenberg-Richter distribution for the earthquakes in the enhanced seismic catalog, including its magnitude of completeness (Mc) following the method of Wiemer et al. (2000). Figure 1 b illustrates the Gutenberg-Richter distribution for the DETECT catalog (11 months) compared to the INFO catalog spanning 13 years of conventional monitoring. Both distributions are plotted with overlaid discrete magnitude bins for direct comparison. The analysis of the completeness magnitude reveals that monitoring seismogenic sources in Irpinia using temporary dense array deployments significantly reduces the detection and completeness thresholds. Specifically, Mc​ decreases by approximately 1.5 magnitude units, down to -0.3 for the enhanced catalog, as compared to corresponding value characterizing the catalog from conventional monitoring (Mc 1.1, Vassallo et al., 2012 ). Moreover, the comparison of the b-values for the two catalogs shows consistent slopes of the Gutenberg-Richter distribution, suggesting scale-invariant seismic generation processes from \(\:{M}_{l}\) 0 to 4. Specifically, we found b DETECT = − 1.06 ± 0.10 ​and b INFO = − 1.03 ± 0.08, indicating that the temporary deployment of dense arrays not only enhances detection capabilities but also provides robust statistical consistency with long-term monitoring data. To identify spatial and temporal clustering of seismicity, we conducted a DBSCAN clustering analysis (Ester et al., 1996 ; Schoenball & Ellsworth, 2017 ) imposing a minimum of 10 events to declare a cluster of earthquakes. This analysis identified 22 seismicity clusters, with the three largest clusters comprising approximately 100 earthquakes each. These events occurred in the Central and Southern sectors as part of three swarm-like sequences culminating in events smaller than \(\:{M}_{l}\:\) 2.0. Analysis of the characteristic depths of clustered seismicity revealed a distinct contrast between deep (> 5 km) and shallow seismicity. Shallow seismicity appears sparser and lacks systematic clustering in space and time, while deeper seismicity predominantly forms spatially compact concentrations. Approximately 45% of earthquakes deeper than 5 km are part of a cluster, compared to only 15% for shallower events. Figure 2 illustrates the clustered seismicity alongside characteristic cross-sections showing the relocated hypocenters. The events are projected on vertical planes oriented according to the orthogonal direction to the strike of the fault segments generating the 1980 Irpinia earthquake (Bernard & Zollo, 1989 ). Figure 2 shows seismicity clusters in the Northern, Central and Southern sectors, (section A-A’, B-B’ and C-C’, respectively), including events within \(\:\pm\:\) 10 km from the vertical planes, with isolated earthquakes represented as shaded black dots. Most clusters are concentrated in the Central Irpinia sector, where the most abundant families occur at depths between 8 and 15 km, consistent with typical earthquake depths in the Southern Apennines (Festa et al., 2021 ; De Landro et al., 2022 ; Scotto di Uccio et al., 2024). In contrast, the Southern sector hosts deeper clusters, with some earthquakes occurring below 15 km. Notably, the larger cluster, consisting of 110 events, is located in this sector (pink dots in Fig. 2 ) and illuminates a new structure, west of the main Irpinia fault zone. Discussion and Conclusions A main result of the DETECT experiment is that deploying a constellation of dense seismic arrays combined with advanced detection techniques significantly enhances the seismic catalogs as compared to the sole use conventional networks while improving the quality and accuracy of earthquake locations and source parameter determinations. By integrating advanced detection techniques with a high-density station deployment over one year, we achieved large catalogs comparable in size to those produced by more than a decade of continuous monitoring using conventional local seismic networks. To isolate the enhancement carried by seismic arrays, we considered the findings by Scotto di Uccio et al. (2023) on seismic sequences in the area recorded by the permanent Near Fault Observatory. Our results indicate that the use of arrays can increase the number of events detected by machine learning techniques by a factor of 2 to 4 as compared to the ordinary network. The DETECT experiment reveals a stable pattern of seismicity throughout the considered period, consistent with long-term monitoring, both in terms of statistical properties, characterized by the b -value of the Gutenberg-Richter relationship, and of spatial distribution of earthquakes, highlighting the spatial invariance of seismic activity on a shorter time scale. When focusing on one year of seismicity monitored by the dense experiment, the overall pattern closely resembles that observed over fifteen years of monitoring, while the magnitude of detected events systematically decreases by 1 to 1.5 units. We identified a clear separation between shallow and deep earthquake behavior, with a discriminating depth around 5 km, suggesting distinct physical mechanisms governing their occurrence. Shallow events appear sparsely located and are primarily concentrated in the central part of the region within the volume enclosed by two boundary faults of 1980 Irpinia earthquake (De Landro et al., 2015 ). Here, hydrological loading influences the stress field in karst aquifers, inducing seasonal oscillations of strain along the NE-SW direction, perpendicular to the trend of the Apennines (D’Agostino et al., 2018; Tarantino et al., 2024 ). In this context, seismicity may be linked to the opening and closing of fluid-permeated microcracks, generating localized earthquakes with a sparse spatial distribution. Additionally, the shallow portion of this area exhibits a lower-than-average stress drop (Picozzi et al., 2021), highlighting the strong influence of fluids—likely brine and CO₂ on stress release mechanisms (Amoroso et al., 2017 ; De Landro et al. 2022 ). Deeper events show a more clustered pattern, with sequences mostly characterized by small magnitude mainshocks followed by aftershocks. Aftershock magnitude generally falls within the 0–1 magnitude range, corresponding to rupture extent of approximately 5–15 meters, according to the self-similar, constant stress drop scaling from former studies on the area (Festa et al., 2021 ; Picozzi et al., 2022a ; Scotto di Uccio et al., 2024). This observation supports the idea that seismicity in the area is primarily driven by stress release following strain accumulation during the inter-seismic period, occurring within a high-fractured medium. Many of these events appear isolated when recorded by the standard network, which suffers from lack of resolution, required to identify such deep and small aftershocks whose signals remain obscured by background noise. This prevents the reconstruction of the sequences in their full complexity, which is crucial for understanding their generation mechanisms and, on a larger scale, the characteristics of the seismic cycle of the fault zone across multiple scales. When interpreting seismic events jointly with the 3D tomographic model of De Landro et al. ( 2022 ), seismicity reveals a right stepping bend that was not clearly identified in previous studies (Fig. 3 ). This feature is consistent with the step-over described in Camanni et al. ( 2025 ), that interprets this fault as a segmented, deep-seated, normal fault Mesozoic in age that was inverted during the Apennine orogeny and is currently reactivated in extension. To the southeast, seismicity extends well into the Apulian carbonate platform, aligning with and surrounding a previously identified, southeast-dipping, long-lived and multiply reactivated major fault (Fig. 3 ; Amoruso et al., 2005, 2011; Amoroso et al., 2014 ). Northward shallowing of high vₚ velocity zones in the tomographic model delineates a bend interpreted as the basement culmination in the hanging wall of this segmented, reactivated fault. This interpretation is further supported by Bouguer gravity anomaly data for the study area (e.g., Improta et al., 2003). Whether this fault directly accommodates the observed seismicity or whether the events occur on nearby sub-parallel small-scale faults within a highly fractured medium—while the main fault remains locked—requires further investigation. Given its total length of 50–60 km, this fault has the potential to generate an event of up to M 7.0 (Bernard & Zollo, 1989 ; Wells & Coppersmith, 1994 ) in cases when rupture propagates across the fault offset and involves the entire fault system. To assess whether a rupture can dynamically propagate across the bend, we conducted 2D numerical simulations using a spectral element solution of the antiplane elastodynamic equation (Festa & Vilotte, 2006 ; Scala et al., 2019 ). We assumed a regional stress field with the maximum principal stress oriented vertically and the minimum compressive stress directed horizontally, perpendicular to the Apennine chain (De Matteis et al., 2012 ). When projecting this stress field onto the segmented fault system, the two longest segments appear well-oriented for rupture propagation (De Matteis et al., 2012 ), whereas the oblique segment is less favorably aligned for rupture continuation. We explored rupture scenarios by varying key parameters, such as the main stress components, frictional coefficients, and rupture nucleation locations. Illustrative examples of rupture evolutions under these conditions are presented in Fig. 4 . Our findings reveal that, in most tested cases, rupture propagates through the entire segmented fault structure, highlighting a significant potential for generating large M6.5 + earthquakes. However, under conditions where the stress drop along bended fault segment remains limited (a few percent of the available strength excess), rupture tends to arrest at the kink. This occurs because the available elastodynamic energy becomes insufficient to overcome the fracture energy needed for frictional weakening and continued rupture propagation (Ma & Archuleta, 2006 ). To refine these rupture scenarios and better quantify seismic hazard, further investigations are required. These should include detailed characterization of stress fields and frictional conditions through laboratory and field measurements, coupled with advanced numerical modeling to systematically evaluate rupture sensitivity to fault geometry complexities. Additionally, analyzing the source properties (e.g., stress drop, rupture dimension, and frictional parameters) of ongoing microseismic activity would offer critical insights. Specifically, high-resolution seismic catalogs could be leveraged to constrain frictional heterogeneity and stress-state variations along fault segments. Combining these observational and modeling approaches will be fundamental to reducing uncertainties in hazard assessments and developing reliable earthquake rupture forecasts. Methods Earthquake detection We followed the same earthquake detection followed by Scotto di Uccio et al. (2023) for seismic sequences in the same region. The workflow is grounded on the use of the machine learning detector EQTransformer (Mousavi et al. 2020 ), which provided a diverse set of templates to be used as the basis for further similarity-based detection using the template matching technique EQCorrscan (Chamberlain et al. 2018 ). After splitting the network into 6 subnetworks of 6 arrays each, with an overlapping of 3 arrays between consecutive subnetworks, we performed earthquake detection independently for each subnetwork and integrated the declaration among the subnetworks according to the detection times. For each subnetwork, we applied EQTransformer on daily continuous data streams, resampled to 100 Hz and filtered in the frequency band [1–45] Hz using the parameterization of Scotto di Uccio et al. (2023), requiring a probability value of 0.1 for P and S arrival times and using 50% overlap between consecutive time windows. Detections were declared when at least 5 picks were associated within time windows of 10 s and were visually confirmed. The machine learning catalog was used as template set for a similarity search using EQCorrscan. Each template event contained only the picked stations, selecting 1.6 s long windows extracted around the automatic picks (including 0.15 s of pre-pick waveforms). We decimated the traces to 25 Hz and filtered in the frequency band [2–9] Hz. For the similarity detection threshold, we selected the sum of cross-correlation coefficient (SCC) between the portion of continuous streams and the templates. We declared an event when the SCC overcomes a similarity threshold with at least one template, fixed at 8 times the MAD of the cross-correlation coefficients between the template and the one-hour chunk of continuous streams. Moreover, we performed cross-correlation picking for the detections, requiring a minimum similarity coefficient of 0.6 at channel level. We applied the automatic selection criteria of Scotto di Uccio et al. (2023) for limiting the false declarations inside the catalog and visually inspected the remaining declarations. Event location and magnitude Using automatic picks, we performed a preliminary location of the earthquakes. We used NLLoc (Lomax et al., 2009 ) and a 1D layered velocity model tailored for the area (Matrullo et al. 2013 ) for retrieving the absolute location of the events. For these initial locations we determined location uncertainties of few kilometers for horizontal and vertical locations and decimals of seconds for the time residuals. To improve the quality of the automatic picks and obtain high quality double difference locations, we applied a refined picking procedure based on cross-correlation and hierarchical clustering (Muzellec et al., 2025 ). The refinement picking procedure is performed at each individual station by correlating waveforms around the picks of events located in 20km x 20km squares. Seismic records were bandpass filtered [1–15] Hz and polarized using a polarization filter (Ross & Ben-Zion, 2014 ; Muzellec et al., 2025 ). We correlated the seismic waveforms on the three components independently in a time window of 0.25 s around the initial picks. The normalized cross-correlation coefficient threshold to define the family was set to 0.8. 10,657 P-wave picks and 34,924 S-wave picks survived the refinement picking procedure. We located the hypocenters with NLLoc (Lomax et al., 2009 ) in 3D velocity models optimized for the area (De Landro et al., 2022 ). The refined picks allowed an increase in the accuracy of the absolute location by reducing the median location error from 1 km to 0.4 km (Figure S2) and the median root-mean-square (RMS) from 0.2 s to 0.1 s (Figure S2). We computed the cross-correlation (CC) differential times based on the refined locations. We finally relocated the seismicity with HypoDD (Waldhauser & Ellsworth, 2000 ) including both the CC and catalog differential times. The CC differential travel times were calculated between events within families built during the refinement picking procedure. The application of the double difference location resulted in a reduction of the median RMS from 0.1 s to 0.007 s. To estimate the error from the DD location procedure, we iteratively used the SVD inversion method on a selection of 100 events by considering a maximum distance of 1 km. Figure S3 shows the histograms of the location errors estimated by DD location with the SVD inversion. The DD location procedure provides location uncertainties featuring a median error of 45 m, 60 m and 75 m, for the x, y and z directions respectively, strongly reducing the absolute location uncertainties, characterized by a median error of 400 m. We computed the local magnitude for equivalent Wood-Anderson displacement, obtained from integration of velocity waveforms, assigning absolute locations to those that were not relocated, using the empirical local magnitude relationship of Bobbio et al. ( 2009 ). Numerical simulations We conducted numerical simulations using the spectral element code developed by Festa & Vilotte ( 2006 ), modified to model 2D antiplane ruptures. We discretized the elastic medium using quadrangular elements, with an eighth-order polynomial approximation to solve the elastodynamic equations within each element. We included perfectly matched layers in the modelling to prevent wave reflections at domain boundaries (Festa & Vilotte, 2005 ). The fault consists of three branches, measuring 32 km, 9.6 km, and 20 km in length, respectively. We imposed a remote regional stress field with the maximum principal stress s 1 oriented vertically and the minimum compressive stress s 3 aligned horizontally, perpendicular to the trend of the Apennines (De Matteis et al., 2012 ). The two parallel fault branches have a strike aligned with the intermediate stress direction s 2 , while the oblique segment forms a 35°S angle relative to s 2 ​ direction (Fig. 4 ). We projected the stress field onto the three segments assuming a fault dip of 60°. We explored s 1 values between 60 and 80 MPa, s 3 ​ values between 20 and 30 MPa, a static friction coefficient ranging from 0.6 to 0.7, and a dynamic friction coefficient between 0.25 and 0.35. These parameters were constrained by the maximum lithostatic load in the region and the expected stress drops, which range between 1 and 15 MPa, as inferred from previous events (Zollo et al., 2014 ; Picozzi et al., 2021). The intermediate stress s 2 ​ was set based on the R value from De Matteis et al. ( 2012 ). Rupture nucleated by locally increasing the initial tangential stress above the failure threshold within a small patch, whose size was comparable to the nucleation length defined by Uenishi & Rice ( 2003 ). We modeled the fracture process using a linear slip-weakening law, where stress decreases from the yield level to the dynamic level over a critical slip distance D c =1m. Declarations Competing interests The authors declare no competing interests. Funding declarations Francesco Scotto di Uccio was supported by the project Transform2 funded by the European Commission under project number 101188365 within the HORIZON-INFRA-2024-DEV-01 call. Titouan Muzellec was supported by the Italian national project PON GRINT CIR01_00013 - Rafforzamento Capitale Umano. Gaetano Festa was supported by the project Geo-Inquire funded by the European Commission under project number 101058518 within the HORIZON-INFRA-2021-SERV-01 call. Author Contribution F.S.d.U. performed earthquake detection, catalog analysis and contributed to writing the original version of the manuscript. T.M. curated refinement of phase arrival times and performed earthquake relocation. A.S. and C.S. generated models and meshes for numerical simulations and revised the original version of the manuscript. G.D.L, G.C., M.P., A.Z. and G.B. contributed to formal analysis, discussion and revision of the original version of the manuscript. F.C. and L.E. contributed to data storage and accessibility. G.F. performed numerical simulations, contributed to the discussion and writing the original version of the manuscript. All authors contributed to finalize the manuscript. Data Availability Seismic waveforms recorded during the DETECT experiment are available at https://geofon.gfz.de/doi/network/ZK/2021. At the time of the submission, data are available under request at https://geofon.gfz.de/doi/network/access. Data will be publicly available from 2025-09-01. Earthquake detection is performed using EQTransformer (https://github.com/smousavi05/EQTransformer, accessed on 2025-05-09) and EQCorrscan (https://github.com/eqcorrscan/EQcorrscan, accessed on 2025-05-09). Earthquake relocations were performed using NonLinLoc (https://alomax.free.fr/nlloc, accessed on 2025-09-25) and HYPODD (https://github.com/fwaldhauser/HypoDD, accessed on 2025-05-09). The enhanced seismic catalog is available at the following link: https://zenodo.org/records/15372023. INFO catalog is available at the Irpinia Near Fault Observatory website (http://isnet‐bulletin.fisica.unina.it/cgi‐bin/isnet‐events/isnet.cgi) and at the EPOS Data Portal (https://www.epos‐eu.org/dataportal)‐IRPINIA Seismic Events provided by Università di Napoli Federico II (accessed on 2025-09-05). Maps and images in Figures 1, 2 and 4 were produced using PyGMT (Uieda et al., 2021) and Matplotlib (Hunter, 2007). Figure 3 was constructed with Surfer® from Golden Software, LLC (version 27, www.goldensoftware.com), used under an educational license. References Kanamori, H. The diversity of large earthquakes and its implications for hazard mitigation. Annu. Rev. Earth Planet. Sci. 42 (1), 7–26 (2014). Bommer, J. J. Review of Seismic Hazard and Risk Analysis. Seismological Soc. Am. 92 (5), 3248–3250 (2021). Zollo, A. et al. Earthquake early warning system in southern Italy: Methodologies and performance evaluation. Geophysical Res. letters , 36 (5). (2009). Allen, R. M. & Melgar, D. Earthquake early warning: Advances, scientific challenges, and societal needs. Annu. Rev. Earth Planet. Sci. 47 (1), 361–388 (2019). Maufroy, E. et al. 3D numerical simulation and ground motion prediction: Verification, validation and beyond–Lessons from the E2VP project. Soil Dyn. Earthq. Eng. 91 , 53–71 (2016). Magen, Y., Ziv, A., Inbal, A., Baer, G. & Hollingsworth, J. Fault rerupture during the July 2019 Ridgecrest earthquake pair from joint slip inversion of InSAR, optical imagery, and GPS. Bull. Seismol. Soc. Am. 110 (4), 1627–1643 (2020). Buttinelli, M. et al. The impact of structural complexity, fault segmentation, and reactivation on seismotectonics: Constraints from the upper crust of the 2016–2017 Central Italy seismic sequence area. Tectonophysics 810 , 228861 (2021). Bello, S. et al. High-detail fault segmentation: Deep insight into the anatomy of the 1983 Borah Peak earthquake rupture zone (2022). Mw 6.9, Idaho, USA). Lithosphere, 2022(1). Tertulliani, A., Leschiutta, I., Bordoni, P. & Milana, G. Damage distribution in L’Aquila city (Central Italy) during the 6 April 2009 earthquake. Bull. Seismol. Soc. Am. 102 (4), 1543–1553 (2012). Galli, P., Castenetto, S. & Peronace, E. The macroseismic intensity distribution of the 30 October 2016 earthquake in central Italy (Mw 6.6): Seismotectonic implications. Tectonics 36 (10), 2179–2191 (2017). Chiaraluce, L. et al. The Near Fault Observatory community in Europe: a new resource for faulting and hazard studies. Ann. Geophys. 65 (3), DM316 (2022). Bernard, P. & Zollo, A. The Irpinia (Italy) 1980 earthquake: detailed analysis of a complex normal faulting. J. Geophys. Research: Solid Earth . 94 (B2), 1631–1647 (1989). Rovida, A., Locati, M., Camassi, R., Lolli, B. & Gasperini, P. (2019). Italian Parametric Earthquake Catalogue CPTI15, version 2.0. Cinti, F. R., Faenza, L., Marzocchi, W. & Montone, P. Probability map of the next M ≥ 5.5 earthquakes in Italy. Geochemistry Geophys. Geosystems , 5 (11). (2004). Iannaccone, G. et al. A prototype system for earthquake early-warning and alert management in southern Italy. Bull. Earthq. Eng. 8 , 1105–1129 (2010). Vassallo, M., Festa, G. & Bobbio, A. Seismic ambient noise analysis in southern Italy. Bull. Seismol. Soc. Am. 102 (2), 574–586 (2012). De Landro, G. et al. High-precision differential earthquake location in 3-D models: Evidence for a rheological barrier controlling the microseismicity at the Irpinia fault zone in southern Apennines. Geophys. Supplements Monthly Notices Royal Astron. Soc. 203 (3), 1821–1831 (2015). di Scotto, F., Scala, A., Festa, G., Picozzi, M. & Beroza, G. C. Comparing and integrating artificial intelligence and similarity search detection techniques: application to seismic sequences in Southern Italy. Geophys. J. Int. 233 (2), 861–874 (2023). di Scotto, F. et al. Characterization and evolution of seismic sequences in the normal fault environment of the Southern Apennines. Journal Geophys. Research: Solid Earth , 129 (8), (2024). e2023JB028644. Amoroso, O., Ascione, A., Mazzoli, S., Virieux, J. & Zollo, A. Seismic imaging of a fluid storage in the actively extending Apennine mountain belt, southern Italy. Geophys. Res. Lett. 41 (11), 3802–3809 (2014). Vassallo, M., Festa, G., Bobbio, A. & Serra, M. Low shear velocity in a normal fault system imaged by ambient noise cross correlation: The case of the Irpinia fault zone, Southern Italy. J. Geophys. Research: Solid Earth . 121 (6), 4290–4305 (2016). Amoroso, O. et al. From velocity and attenuation tomography to rock physical modeling: Inferences on fluid-driven earthquake processes at the Irpinia fault system in southern Italy. Geophys. Res. Lett. 44 (13), 6752–6760 (2017). D'Agostino, N. et al. Crustal deformation and seismicity modulated by groundwater recharge of karst aquifers. Geophys. Res. Lett. 45 (22), 12–253 (2018). Tarantino, S. et al. Non-linear elasticity, earthquake triggering and seasonal hydrological forcing along the Irpinia fault, Southern Italy. Nat. Commun. 15 (1), 9821 (2024). De Landro, G. et al. Decade-long monitoring of seismic velocity changes at the Irpinia fault system (southern Italy) reveals pore pressure pulsations. Sci. Rep. 12 (1), 1247 (2022). Picozzi, M., Bindi, D., Zollo, A., Festa, G. & Spallarossa, D. Detecting long-lasting transients of earthquake activity on a fault system by monitoring apparent stress, ground motion and clustering. Sci. Rep. , 9 (1). (2019). Picozzi, M. et al. Spatiotemporal Evolution of Microseismicity Seismic Source Properties at the Irpinia Near-Fault Observatory, Southern Italy. Bull. Seismol. Soc. Am. 112 (1), 226–242 (2022a). Picozzi, M. et al. Spatiotemporal Evolution of Ground-Motion Intensity at the Irpinia Near-Fault Observatory, Southern Italy. Bull. Seismol. Soc. Am. 112 (1), 243–261 (2022b). Camanni, G. et al. Remobilization of inverted normal faults drives active extension in the axial zone of the southern Apennine mountain belt (Italy). J. Geol. Soc. 182 (2), jgs2024–jgs2184 (2025). Zhu, W. & Beroza, G. C. PhaseNet: a deep-neural-network-based seismic arrival-time picking method. Geophys. J. Int. 216 (1), 261–273 (2019). Mousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y. & Beroza, G. C. Earthquake transformer—an attentive deep-learning model for simultaneous earthquake detection and phase picking. Nat. Commun. 11 (1), 3952 (2020). Tan, Y. J. et al. Machine-learning‐based high‐resolution earthquake catalog reveals how complex fault structures were activated during the 2016–2017 central Italy sequence. Seismic Record . 1 (1), 11–19 (2021). Michelini, A. et al. INSTANCE–the Italian seismic dataset for machine learning. Earth Syst. Sci. Data . 13 (12), 5509–5544 (2021). Chamberlain, C. J. et al. EQcorrscan: Repeating and near‐repeating earthquake detection and analysis in Python. Seismol. Res. Lett. 89 (1), 173–181 (2018). Vuan, A., Sugan, M., Amati, G. & Kato, A. Improving the detection of low-magnitude seismicity preceding the Mw 6.3 L’Aquila earthquake: Development of a scalable code based on the cross correlation of template earthquakes. Bull. Seismol. Soc. Am. 108 (1), 471–480 (2018). Essing, D. & Poli, P. Spatiotemporal evolution of the seismicity in the alto tiberina fault system revealed by a high-resolution template matching catalog. J. Geophys. Research: Solid Earth , 127 (10), (2022). e2022JB024845. Spallarossa, D. et al. An automatically generated highresolution earthquake catalog for the 2016–2017 Central Italy seismic sequence, including P and S phase arrival times. Geophys. J. Int. 225 (1), 555–571 (2021). Michele, M., Chiaraluce, L., Di Stefano, R. & Waldhauser, F. Fine-scale structure of the 2016–2017 Central Italy seismic sequence from data recorded at the Italian National Network. Journal of Geophysical Research: Solid Earth , 125 (4), e2019JB018440. (2020). Cabrera, L., Poli, P. & Frank, W. B. Tracking the spatio-temporal evolution of foreshocks preceding the Mw 6.1 2009 L’Aquila earthquake. J. Geophys. Research: Solid Earth , 127 (3), (2022). e2021JB023888. Park, Y., Beroza, G. C. & Ellsworth, W. L. Basement fault activation before larger earthquakes in Oklahoma and Kansas. Seismic Record . 2 (3), 197–206. oclc.org/10.1785/0320220020 (2022). https://doi-org.stanford.idm. Muzellec, T., De Landro, G., Camanni, G., Adinolfi, G. M. & Zollo, A. The complex 4D multi-segmented rupture of the 2014 Mw 6.2 Northern Nagano Earthquake revealed by high-precision aftershock locations. Tectonophysics , 230641. (2025). Lomax, A., Michelini, A., Curtis, A. & Meyers, R. A. Earthquake location, direct, global-search methods. Encyclopedia Complex. Syst. Sci. 5 , 2449–2473 (2009). Ross, Z. E., Trugman, D. T., Hauksson, E. & Shearer, P. M. Searching for hidden earthquakes in Southern California. Science 364 (6442), 767–771 (2019). Yu, Y., Ellsworth, W. L. & Beroza, G. C. Accuracy and precision of earthquake location programs: Insights from a synthetic controlled experiment. Seismol. Res. Lett. 96 (3), 1860–1874 (2025). Bobbio, A., Vassallo, M. & Festa, G. A local magnitude scale for southern Italy. Bull. Seismol. Soc. Am. 99 (4), 2461–2470 (2009). Ester, M., Kriegel, H. P., Sander, J. & Xu, X. August). A density-based algorithm for discovering clusters in large spatial databases with noise. kdd (Vol . 96 (34), 226–231 (1996). Schoenball, M. & Ellsworth, W. L. A systematic assessment of the spatiotemporal evolution of fault activation through induced seismicity in Oklahoma and southern Kansas. J. Geophys. Research: Solid Earth . 122 (12), 10–189 (2017). Festa, G. et al. Insights into mechanical properties of the 1980 Irpinia fault system from the analysis of a seismic sequence. Geosciences 11 (1), 28 (2021). Wells, D. L. & Coppersmith, K. J. New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. Bull. Seismol. Soc. Am. 84 (4), 974–1002 (1994). Festa, G. & Vilotte, J. P. Influence of the rupture initiation on the intersonic transition: Crack-like versus pulse‐like modes. Geophysical Res. Letters , 33 (15). (2006). Scala, A., Festa, G., Vilotte, J. P., Lorito, S. & Romano, F. Wave interaction of reverse-fault rupture with free surface: Numerical analysis of the dynamic effects and fault opening induced by symmetry breaking. J. Geophys. Research: Solid Earth . 124 (2), 1743–1758 (2019). De Matteis, R. et al. Fault delineation and regional stress direction from the analysis of background microseismicity in the southern Apennines. Italy Bull. Seismological Soc. Am. 102 (4), 1899–1907 (2012). Ma, S. & Archuleta, R. J. Radiated seismic energy based on dynamic rupture models of faulting. J. Geophys. Research: Solid Earth . 111 , B5 (2006). Matrullo, E., De Matteis, R., Satriano, C., Amoroso, O. & Zollo, A. An improved 1-D seismic velocity model for seismological studies in the Campania–Lucania region (Southern Italy). Geophys. J. Int. 195 (1), 460–473 (2013). Ross, Z. E. & Ben-Zion, Y. Automatic picking of direct P, S seismic phases and fault zone head waves. Geophys. J. Int. 199 (1), 368–381 (2014). Waldhauser, F. & Ellsworth, W. L. A double-difference earthquake location algorithm: Method and application to the northern Hayward fault, California. Bull. Seismol. Soc. Am. 90 (6), 1353–1368 (2000). Festa, G. & Vilotte, J. P. The Newmark scheme as velocity–stress time-staggering: an efficient PML implementation for spectral element simulations of elastodynamics. Geophys. J. Int. 161 (3), 789–812 (2005). Zollo, A., Orefice, A. & Convertito, V. Source parameter scaling and radiation efficiency of microearthquakes along the Irpinia fault zone in southern Apennines, Italy. J. Geophys. Research: Solid Earth . 119 (4), 3256–3275 (2014). Uenishi, K. & Rice, J. R. Universal nucleation length for slip-weakening rupture instability under nonuniform fault loading. J. Geophys. Research: Solid Earth . 108 , B1 (2003). Uieda, L. et al. PyGMT: A Python interface for the generic mapping tools. (2021). Hunter, J. D. Matplotlib: A 2D graphics environment. Comput. Sci. Eng. 9 (03), 90–95 (2007). Additional Declarations No competing interests reported. Supplementary Files ScottodiUccioetalSupportingInformation.docx Cite Share Download PDF Status: Published Journal Publication published 17 Jan, 2026 Read the published version in Scientific Reports → Version 1 posted Editorial decision: Revision requested 11 Aug, 2025 Reviews received at journal 11 Aug, 2025 Reviewers agreed at journal 21 Jul, 2025 Reviewers agreed at journal 14 Jul, 2025 Reviews received at journal 26 Jun, 2025 Reviewers agreed at journal 17 Jun, 2025 Reviewers agreed at journal 03 Jun, 2025 Reviewers invited by journal 30 May, 2025 Editor invited by journal 27 May, 2025 Editor assigned by journal 26 May, 2025 Submission checks completed at journal 25 May, 2025 First submitted to journal 22 May, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6727679","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":488517658,"identity":"acc9d9a3-4db1-4867-8b8f-654df2bae19a","order_by":0,"name":"Francesco Scotto di Uccio","email":"data:image/png;base64,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","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":true,"prefix":"","firstName":"Francesco","middleName":"Scotto di","lastName":"Uccio","suffix":""},{"id":488517659,"identity":"742074af-42f9-4119-910e-c3c57fbda2ce","order_by":1,"name":"Titouan Muzellec","email":"","orcid":"","institution":"University of Vienna","correspondingAuthor":false,"prefix":"","firstName":"Titouan","middleName":"","lastName":"Muzellec","suffix":""},{"id":488517660,"identity":"d2fa5988-0b1c-48c9-ac28-55f942ab182a","order_by":2,"name":"Antonio Scala","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Antonio","middleName":"","lastName":"Scala","suffix":""},{"id":488517661,"identity":"7c7e58c6-e0c7-491c-9a39-6b1dd4cb8e99","order_by":3,"name":"Grazia De Landro","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Grazia","middleName":"","lastName":"De Landro","suffix":""},{"id":488517662,"identity":"c8e73121-7547-4930-8b03-63e2b8d7d445","order_by":4,"name":"Giovanni Camanni","email":"","orcid":"","institution":"University of Modena and Reggio Emilia","correspondingAuthor":false,"prefix":"","firstName":"Giovanni","middleName":"","lastName":"Camanni","suffix":""},{"id":488517663,"identity":"dda17fd5-5325-48a2-ab52-254c2c4c5ac8","order_by":5,"name":"Francesco Carotenuto","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Francesco","middleName":"","lastName":"Carotenuto","suffix":""},{"id":488517664,"identity":"052fefa1-f933-44b4-a284-3ff578fd2f68","order_by":6,"name":"Luca Elia","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Luca","middleName":"","lastName":"Elia","suffix":""},{"id":488517665,"identity":"97e1a1b1-c10c-4b9a-b899-4bf4a16832c4","order_by":7,"name":"Matteo Picozzi","email":"","orcid":"","institution":"National Institute of Oceanography and Experimental Geophysics","correspondingAuthor":false,"prefix":"","firstName":"Matteo","middleName":"","lastName":"Picozzi","suffix":""},{"id":488517666,"identity":"613133de-e78e-40d5-a3c9-9dd717a6706d","order_by":8,"name":"Aldo Zollo","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Aldo","middleName":"","lastName":"Zollo","suffix":""},{"id":488517667,"identity":"a853c1fc-5e58-43fe-be10-5c7cfe432370","order_by":9,"name":"Claudio Strumia","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Claudio","middleName":"","lastName":"Strumia","suffix":""},{"id":488517668,"identity":"e96a20b4-9232-41a0-9e43-768607facdac","order_by":10,"name":"Greg Beroza","email":"","orcid":"","institution":"Stanford University","correspondingAuthor":false,"prefix":"","firstName":"Greg","middleName":"","lastName":"Beroza","suffix":""},{"id":488517669,"identity":"00aee12e-a4c3-40f6-b45c-5b45ff84b9fa","order_by":11,"name":"Gaetano Festa","email":"","orcid":"","institution":"University of Naples \"Federico II\", Department of Physics","correspondingAuthor":false,"prefix":"","firstName":"Gaetano","middleName":"","lastName":"Festa","suffix":""}],"badges":[],"createdAt":"2025-05-22 20:08:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6727679/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6727679/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41598-026-35586-3","type":"published","date":"2026-01-17T16:29:30+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":87527506,"identity":"d2e0aca7-dddb-4cfc-be64-568f58997825","added_by":"auto","created_at":"2025-07-24 20:00:46","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":788134,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003eLeft panel\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e: Epicentral distribution of the earthquakes recorded by INFO (blue dots) within 2008 and September 2021 (start of the DETECT survey), along with the seismic stations (yellow triangles). Epicentral distribution of the relocated earthquakes occurred within the DETECT experiment (green dots), along with the stations of the dense arrays (red triangles). \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eRight panel: \u003c/strong\u003e\u003c/em\u003e\u003cem\u003eFrequency-magnitude distribution for the earthquakes in the INFO catalog from 2008 to September 2021 (blue dots and bars) and analogous distribution for the earthquakes recorded within the DETECT survey (green dots and bars). The magnitude of completeness of the seismic catalog is improved by more than one unit (from Mc 1.1 to -0.3), while we observe compatible b-values between the catalogs between decadal conventional and short-term dense monitoring.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/d6913a90b596a8f418a93423.png"},{"id":87527914,"identity":"dedd501d-bf10-4a18-8e95-dcf276cffeea","added_by":"auto","created_at":"2025-07-24 20:08:46","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":686533,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eClustering analysis of DETECT seismicity using DBSCAN. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eLeft panel\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e: Epicentral distribution of clustered (colored dots, 22 clusters with a minimum number of 10 earthquakes) and isolated (black dots) seismicity. \u003c/em\u003e\u003cem\u003e\u003cstrong\u003eRight panel\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e: cross-section representations of relocated hypocenters for the seismicity in the Northern (Section A-A’), Central (Section B-B’) and Southern (Section C-C’) sectors, color coded according to clustered and isolated seismicity. For each cross-section, earthquakes within ± 10 km are represented. Most of the earthquakes occurring at depth \u0026gt; 5 km belong to a seismic cluster (45%), while the percentage drops to 15% for shallower earthquakes\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/878af3038d3b5fed22fb7571.png"},{"id":87528086,"identity":"6000e373-e7cf-4b2b-bdeb-18027344114e","added_by":"auto","created_at":"2025-07-24 20:16:46","extension":"jpeg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":881445,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eA comprehensive view of the fault trace reconstructed by integrating the enhanced micro-seismicity catalog (events with depth greater than 5 km are represented by black dots if above the iso-surface and brown dots if below) and the P-wave velocity tomography (De Landro et al. 2022). Here we show the iso-velocity surface at 5.5 km/s and two vertical sections crossing the fault surface at the boundary of the low velocity anomaly (red shaded area).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage3.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/d95c1fd9b508741650a74a16.jpeg"},{"id":87527915,"identity":"7e6e25fb-5db5-4bd3-9324-1dacdbe999e4","added_by":"auto","created_at":"2025-07-24 20:08:46","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":229384,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eExamples of dynamic antiplane ruptures that propagate across the bend (left panel) or arrest after turning the kink (right panel).\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/4ae1973518edb7ef620fc9ee.png"},{"id":100614658,"identity":"661b6e09-998d-4a01-8d64-352395219607","added_by":"auto","created_at":"2026-01-19 17:22:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3266124,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/0627b31d-3221-435c-9cf7-0b73c74d08a0.pdf"},{"id":87527511,"identity":"f6d524e4-96a7-4747-8b85-4c602cb41b64","added_by":"auto","created_at":"2025-07-24 20:00:46","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":288803,"visible":true,"origin":"","legend":"","description":"","filename":"ScottodiUccioetalSupportingInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-6727679/v1/13bc3755d61332eeef7f271d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Potential earthquake-prone faults identified by dense seismic array monitoring in complex extensional settings","fulltext":[{"header":"Introduction","content":"\u003cp\u003eUnderstanding the preparation and nucleation processes that drive the space-time evolution of moderate-to-large earthquake ruptures is essential for characterizing the conditions controlling earthquake initiation and growth. This knowledge is essential both for predicting the seismic wavefield generated during earthquakes and for mitigating seismic risk through targeted safety measures. These mitigation actions can be implemented on timescales corresponding to earthquake recurrence intervals (Kanamori, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Bommer, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) and over shorter timescales during rupture propagation, as part of earthquake early-warning systems (Zollo et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Allen \u0026amp; Melgar, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In this framework, detailed information on fault geometry and mechanical conditions provides crucial information that can reduce epistemic uncertainty in ground-motion modeling and seismic hazard assessment (Maufroy et al., 2012).\u003c/p\u003e\u003cp\u003eHowever, anticipating the geometry and even the presence of seismogenic faults presents a challenge, as demonstrated by many recent large earthquakes that occurred on previously unknown faults, such as those responsible for the 2019 M 7.1 Ridgecrest earthquake in California (Magen et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). This issue is even more challenging in extensional settings, where normal fault systems can be highly complex and comprise subparallel, segmented synthetic and antithetic structures that interact mechanically over diverse space and time scales. (Buttinelli et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Bello et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn the actively extending Apennines chain, even moderate-size events with magnitudes ranging from 6 to 6.5, that rupture 10 to 30 km-length faults, may lead to extensive casualties and building damage, as illustrated by the 2009 L\u0026rsquo;Aquila earthquake (Tertulliani et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e) and the 2016 Amatrice-Norcia sequence (Galli et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). To better understand the geometry and stress state and to assess the risk related to moderate-size earthquake faults, dense multi-parametric monitoring infrastructures referred to as Near Fault Observatories, have been deployed in the Central and Southern Apennines of Italy and across Europe over the last 15 years (Chiaraluce et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn this study, we focus on the multi-segmented Irpinia fault system, in the Southern Apennines, that generated the 1980 Ms 6.9 earthquake (Bernard \u0026amp; Zollo, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1989\u003c/span\u003e). This earthquake ruptured at least three fault segments, each of which were tens of km long over more than 40 s duration, resulting in a long-duration strong ground shaking that caused widespread building collapse and over 3000 fatalities (Rovida et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). The Irpinia region is classified as one of the highest seismic hazard areas in Italy (MPS Working group, 2004) with a relatively short return period for M 6\u0026thinsp;+\u0026thinsp;earthquakes (Galli, 2020), and a probability greater than 30% of a M5.5\u0026thinsp;+\u0026thinsp;earthquake occurring within the next decade (Cinti et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2004\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn 2005, the Irpinia Near Fault Observatory (INFO) was created with the aim of developing a large research infrastructure in Earth Sciences to monitor the Irpinia fault system by a dense seismic network consisting of 39 stations, with inter-station distances ranging from 10 to 20 km (Iannaccone et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). This network allows for a local magnitude of completeness in seismic catalogs of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\:\\)\u003c/span\u003e\u003c/span\u003e1.1 and a magnitude detection threshold down to \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\:\\)\u003c/span\u003e\u003c/span\u003e0.5 in its central sector (Vassallo et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The epicenters of the earthquakes occurring from 2008 till September 2021 are reported as blue circles in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea, along with the stations of the INFO (yellow triangles in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). Analysis of the background microseismicity reveals a rather sparse seismicity within the graben bounded by the main faults responsible for the 1980 Irpinia earthquake (De Landro et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Occasionally, microseismicity clusters in seismic sequences lasting for a few days with modest magnitude mainshocks, that rupture smaller, sub-parallel structures to those activated during the Ms 6.9 event in 1980 (Scotto di Uccio et al., 2023; 2024).\u003c/p\u003e\u003cp\u003eThe tomographic modeling of first P and S arrival times (Amoroso et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) and ambient noise data (Vassallo et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) suggest that the area is permeated by deep fluids, predominantly CO₂ and brine (Amoroso et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Additionally, geodetic data modeling reveals a non-linear elastic response of the shallow karst aquifers to hydrological loading, with the opening and closing of cracks correlated with fluctuations in seismicity rates in the area (D\u0026rsquo;Agostino et al., 2018; Tarantino et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). The long-term monitoring of microseismicity has provided evidence for structural segmentation and evolution of both crustal and source properties, constrained properties of the crust (De Landro et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), informed measurements of the apparent stress (Picozzi et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and stress drop (Picozzi et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e), and constrained the spatial variability of ground motion intensity (Picozzi et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2022b\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eUncertainties in earthquake location, however, hinder the clear identification of causative fault structures, making it ambiguous whether the sparse hypocentral distributions are due to limitations in resolution or to a genuinely chaotic orientation and spatial distribution of small structures hosting the microseismic events (De Landro et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Camanni et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo investigate new technological solutions aimed at improving seismic monitoring capability, we deployed a constellation of 20 small-aperture seismic arrays, each consisting of 10 stations (200 stations in total) integrating INFO during the period September 2021 - July 2022, in the framework of a temporary experiment, named DETECT (DEnse mulTi-paramEtriC observations and 4D high resoluTion imaging, red triangles in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). DETECT is an international monitoring project coordinated by GFZ and the University of Napoli Federico II, that involved several Italian University and research institutes (INGV, CNR, Universit\u0026agrave; di Salerno, and Universit\u0026agrave; del Sannio).\u003c/p\u003e\u003cp\u003eWithin the DETECT arrays, the average inter-station distance ranged from several hundred meters to one kilometer, while the average distance between arrays was approximately 10 km. Each array was equipped with one broadband seismometer, one 1Hz sensor, and eight short-period (4.5 Hz natural frequency) geophones. The data collected were processed, standardized, and made publicly available (Strollo et al., 2025).\u003c/p\u003e\u003cp\u003eThe main goal of DETECT was the generation of enhanced catalogues of accurately located microseismic events. Machine learning (ML) and similarity-based methods have shown significant potential for increasing the size of seismic catalog by up to an order of magnitude, uncovering previously uncatalogued lower-magnitude events, even in areas for which the events were not included in the training dataset (Zhu \u0026amp; Beroza, 2018; Mousavi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Tan et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). However, these approaches can suffer from a high rate of false positives and missed detections. Such limitations can be mitigated using well-designed ML models trained on region-specific datasets (Michelini et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) or by assisting more conservative ML detectors with robust, network-based similarity techniques (e.g., Chamberlain et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), where events identified by ML algorithms can be used as templates for similarity-based detections (Scotto di Uccio et al., 2023). Indeed, template matching algorithms target earthquakes occurring close to a known set of events and typically achieve a better detection performance for earthquakes reporting low signal-to-noise ratio (Vuan et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Essing \u0026amp; Poli, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Scotto di Uccio et al., 2023).\u003c/p\u003e\u003cp\u003eAccurate hypocenter determinations of earthquakes in ML-enhanced catalogs have shown to provide significant insights into fault geometries (Spallarossa et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Michele et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Cabrera et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Park et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Scotto di Uccio et al., 2024). For generating the seismic catalog, we followed the detection strategy proposed by Scotto di Uccio et al. (2023), integrating the deep learning detector EQTransformer (Mousavi et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; hereinafter EQT) and the template matching technique EQCorrscan (Chamberlain et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; hereinafter TM), using the enhanced catalog provided by EQT as the template set for the TM detection. During the detection stage, the dense array deployment was split into overlapping subnetworks allowing us to focus on the volume beneath the considered arrays, increasing the ability to detect local and low-magnitude events while limiting the false declarations caused by coherent non-seismic transients at distant stations (see Methods). We performed accurate hypocenter locations by refining the automatic phase arrival times among families of similar waveforms identified using hierarchical clustering (Muzellec et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). We used NLLoc (Lomax et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and a 3D velocity model tailored for the area (De Landro et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) for obtaining an improved absolute hypocenter location. The locations were refined by performing double difference location using HYPODD (Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), introducing catalog and cross-correlation differential travel times for highlighting the small-scale features of the fault structures. Location uncertainties based on bootstrap analysis of the seismic catalog resulted in uncertainties of a few hundreds of meters (Muzellec et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, see Methods).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eThe application of these techniques to continuous data from DETECT led to the identification of approximately 3,600 earthquakes occurred during the 11-months experiment duration. This enhanced catalog represents an ~\u0026thinsp;8-fold increase in the number of earthquakes compared to the existing catalog provided by INFO within the same time window. The existing catalog is based on the use of a conventional network layout and seismic detection methods, as the visual inspection of records by operators analyzing data from the standard network stations (yellow triangles in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea). Focusing on the contribution of individual methodologies in detecting low-magnitude earthquakes, we found that the machine learning-generated catalog expanded the manual catalog by a factor of ~\u0026thinsp;4. This represents a twofold improvement over the results reported by Scotto di Uccio et al. (2023), who applied the same deep learning detector to seismic sequences recorded by the INFO network. This highlights the efficacy of dense constellations of arrays in amplifying the detection capability of machine learning models for low-magnitude seismic events. Furthermore, the detection catalog derived from advanced strategies applied to dense arrays exhibits an earthquake count on the same order of magnitude as multiple years of conventional monitoring (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb).\u003c/p\u003e\u003cp\u003eTo characterize the earthquakes in the enhanced catalogs, we performed accurate relative event relocation. Refinement of phase arrival times using hierarchical clustering (Muzellec et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) yielded average corrections within ~\u0026thinsp;0.1 s for both P and S picks (Figure \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). This result not only supports the high accuracy in the automatic identifications of phase arrival times but also provides critical adjustments for constraining the fine-scale properties of seismogenic sources. We successfully relocated 2,248 earthquakes (~\u0026thinsp;65% of the detected events), nearly doubling the fraction of relocated events in enhanced catalogs for seismic sequences (Scotto di Uccio et al., 2024) and significantly surpassing typical relocation percentages in template-matching-derived catalogs (Ross et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Cabrera et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This improvement is attributed to the deployment of the dense seismic arrays, which facilitates the detection of a larger number of phase arrival times, even for low-magnitude events (~\u0026thinsp;47,000 P-wave and ~\u0026thinsp;72,000 S-wave picks). The increased number of picks resulted in a substantial dataset of differential travel times, which are fundamental to achieve high-precision earthquake locations.\u003c/p\u003e\u003cp\u003eThe left panel of Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e illustrates the epicenters of the relocated earthquakes, showcasing the spatial distribution of the relocated seismicity observed during DETECT (shaded green points) compared to the seismicity detected by INFO (blue points) from 2008 up to the start of the dense array monitoring (September 2021).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe spatial distribution of epicenters from the enhanced catalogs reveals variable seismicity levels across different sectors of the Irpinia region. Most of the seismicity is concentrated in the Central and Southern sectors, while the Northern sector exhibits significantly lower activity, predominantly characterized by isolated events and a single swarm-like sequence of 30 earthquakes associated with a \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\:\\)\u003c/span\u003e\u003c/span\u003e2.1 event located outside the coverage area of DETECT (northernmost green cluster in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e\u003cp\u003eWhen comparing the seismicity observed during the experiment with that recorded by INFO between 2008 and September 2021 (the start time of the experiment), we find consistency in the distribution of epicenters. This highlights the persistent occurrence of microseismicity in areas prone to generating moderate-to-low magnitude earthquakes along the Irpinia fault system. Most of the earthquakes recorded by the dense arrays occur in regions reporting a high density of events already identified in the manual catalog. We use the Chamfer distance (Yu et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) to measure the connection between relocated hypocenters from this analysis and the seismicity depicted by the long-term INFO catalog. We obtain an average nearest-neighboring distance of 0.9km and 1.1km for earthquakes occurring in the Central and Southern sectors, respectively, lower than what observed from the earthquakes in the INFO catalog during the DETECT experiment (1.3km and 1.6km, respectively). The low seismicity rate observed in the Northern sector, which is typically interested by the occurrence of seismic sequences populating the INFO catalog, prevents from extracting robust estimates. Moreover, an absence of seismicity is evident in the Southeastern edge of the region (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003ea), a feature previously reported by De Landro et al. (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e) inspecting a catalog of six years of microseismicity with moment magnitude ranging between 0.9 and 3.1. This phenomenon has been attributed to the presence of a contact-zone between geological units with differing rheological properties in response to the NE\u0026ndash;SW stress regime acting on the chain.\u003c/p\u003e\u003cp\u003eTo evaluate the statistical parameters of the enhanced catalog, we calculated the local magnitudes of the events using a scale calibrated for the area (Bobbio et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), resulting in values ranging in \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\)\u003c/span\u003e\u003c/span\u003e [-1.9 to 2.9]. We estimated the statistical parameters of the Gutenberg-Richter distribution for the earthquakes in the enhanced seismic catalog, including its magnitude of completeness (Mc) following the method of Wiemer et al. (2000). Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb illustrates the Gutenberg-Richter distribution for the DETECT catalog (11 months) compared to the INFO catalog spanning 13 years of conventional monitoring. Both distributions are plotted with overlaid discrete magnitude bins for direct comparison. The analysis of the completeness magnitude reveals that monitoring seismogenic sources in Irpinia using temporary dense array deployments significantly reduces the detection and completeness thresholds. Specifically, Mc​ decreases by approximately 1.5 magnitude units, down to -0.3 for the enhanced catalog, as compared to corresponding value characterizing the catalog from conventional monitoring (Mc 1.1, Vassallo et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Moreover, the comparison of the b-values for the two catalogs shows consistent slopes of the Gutenberg-Richter distribution, suggesting scale-invariant seismic generation processes from \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\)\u003c/span\u003e\u003c/span\u003e 0 to 4. Specifically, we found b\u003csub\u003eDETECT\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;1.06\u0026thinsp;\u0026plusmn;\u0026thinsp;0.10 ​and b\u003csub\u003eINFO\u003c/sub\u003e\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;1.03\u0026thinsp;\u0026plusmn;\u0026thinsp;0.08, indicating that the temporary deployment of dense arrays not only enhances detection capabilities but also provides robust statistical consistency with long-term monitoring data.\u003c/p\u003e\u003cp\u003eTo identify spatial and temporal clustering of seismicity, we conducted a DBSCAN clustering analysis (Ester et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e1996\u003c/span\u003e; Schoenball \u0026amp; Ellsworth, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) imposing a minimum of 10 events to declare a cluster of earthquakes. This analysis identified 22 seismicity clusters, with the three largest clusters comprising approximately 100 earthquakes each. These events occurred in the Central and Southern sectors as part of three swarm-like sequences culminating in events smaller than \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{M}_{l}\\:\\)\u003c/span\u003e\u003c/span\u003e2.0.\u003c/p\u003e\u003cp\u003eAnalysis of the characteristic depths of clustered seismicity revealed a distinct contrast between deep (\u0026gt;\u0026thinsp;5 km) and shallow seismicity. Shallow seismicity appears sparser and lacks systematic clustering in space and time, while deeper seismicity predominantly forms spatially compact concentrations. Approximately 45% of earthquakes deeper than 5 km are part of a cluster, compared to only 15% for shallower events. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the clustered seismicity alongside characteristic cross-sections showing the relocated hypocenters. The events are projected on vertical planes oriented according to the orthogonal direction to the strike of the fault segments generating the 1980 Irpinia earthquake (Bernard \u0026amp; Zollo, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1989\u003c/span\u003e). Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows seismicity clusters in the Northern, Central and Southern sectors, (section A-A\u0026rsquo;, B-B\u0026rsquo; and C-C\u0026rsquo;, respectively), including events within \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\pm\\:\\)\u003c/span\u003e\u003c/span\u003e10 km from the vertical planes, with isolated earthquakes represented as shaded black dots. Most clusters are concentrated in the Central Irpinia sector, where the most abundant families occur at depths between 8 and 15 km, consistent with typical earthquake depths in the Southern Apennines (Festa et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; De Landro et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Scotto di Uccio et al., 2024). In contrast, the Southern sector hosts deeper clusters, with some earthquakes occurring below 15 km. Notably, the larger cluster, consisting of 110 events, is located in this sector (pink dots in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) and illuminates a new structure, west of the main Irpinia fault zone.\u003c/p\u003e"},{"header":"Discussion and Conclusions","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003cp\u003eA main result of the DETECT experiment is that deploying a constellation of dense seismic arrays combined with advanced detection techniques significantly enhances the seismic catalogs as compared to the sole use conventional networks while improving the quality and accuracy of earthquake locations and source parameter determinations. By integrating advanced detection techniques with a high-density station deployment over one year, we achieved large catalogs comparable in size to those produced by more than a decade of continuous monitoring using conventional local seismic networks. To isolate the enhancement carried by seismic arrays, we considered the findings by Scotto di Uccio et al. (2023) on seismic sequences in the area recorded by the permanent Near Fault Observatory. Our results indicate that the use of arrays can increase the number of events detected by machine learning techniques by a factor of 2 to 4 as compared to the ordinary network.\u003c/p\u003e\u003cp\u003eThe DETECT experiment reveals a stable pattern of seismicity throughout the considered period, consistent with long-term monitoring, both in terms of statistical properties, characterized by the \u003cem\u003eb\u003c/em\u003e-value of the Gutenberg-Richter relationship, and of spatial distribution of earthquakes, highlighting the spatial invariance of seismic activity on a shorter time scale. When focusing on one year of seismicity monitored by the dense experiment, the overall pattern closely resembles that observed over fifteen years of monitoring, while the magnitude of detected events systematically decreases by 1 to 1.5 units.\u003c/p\u003e\u003cp\u003eWe identified a clear separation between shallow and deep earthquake behavior, with a discriminating depth around 5 km, suggesting distinct physical mechanisms governing their occurrence. Shallow events appear sparsely located and are primarily concentrated in the central part of the region within the volume enclosed by two boundary faults of 1980 Irpinia earthquake (De Landro et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Here, hydrological loading influences the stress field in karst aquifers, inducing seasonal oscillations of strain along the NE-SW direction, perpendicular to the trend of the Apennines (D\u0026rsquo;Agostino et al., 2018; Tarantino et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). In this context, seismicity may be linked to the opening and closing of fluid-permeated microcracks, generating localized earthquakes with a sparse spatial distribution. Additionally, the shallow portion of this area exhibits a lower-than-average stress drop (Picozzi et al., 2021), highlighting the strong influence of fluids\u0026mdash;likely brine and CO₂ on stress release mechanisms (Amoroso et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; De Landro et al. \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDeeper events show a more clustered pattern, with sequences mostly characterized by small magnitude mainshocks followed by aftershocks. Aftershock magnitude generally falls within the 0\u0026ndash;1 magnitude range, corresponding to rupture extent of approximately 5\u0026ndash;15 meters, according to the self-similar, constant stress drop scaling from former studies on the area (Festa et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Picozzi et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2022a\u003c/span\u003e; Scotto di Uccio et al., 2024). This observation supports the idea that seismicity in the area is primarily driven by stress release following strain accumulation during the inter-seismic period, occurring within a high-fractured medium. Many of these events appear isolated when recorded by the standard network, which suffers from lack of resolution, required to identify such deep and small aftershocks whose signals remain obscured by background noise. This prevents the reconstruction of the sequences in their full complexity, which is crucial for understanding their generation mechanisms and, on a larger scale, the characteristics of the seismic cycle of the fault zone across multiple scales.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWhen interpreting seismic events jointly with the 3D tomographic model of De Landro et al. (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), seismicity reveals a right stepping bend that was not clearly identified in previous studies (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). This feature is consistent with the step-over described in Camanni et al. (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), that interprets this fault as a segmented, deep-seated, normal fault Mesozoic in age that was inverted during the Apennine orogeny and is currently reactivated in extension. To the southeast, seismicity extends well into the Apulian carbonate platform, aligning with and surrounding a previously identified, southeast-dipping, long-lived and multiply reactivated major fault (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e; Amoruso et al., 2005, 2011; Amoroso et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Northward shallowing of high \u003cem\u003evₚ\u003c/em\u003e velocity zones in the tomographic model delineates a bend interpreted as the basement culmination in the hanging wall of this segmented, reactivated fault. This interpretation is further supported by Bouguer gravity anomaly data for the study area (e.g., Improta et al., 2003). Whether this fault directly accommodates the observed seismicity or whether the events occur on nearby sub-parallel small-scale faults within a highly fractured medium\u0026mdash;while the main fault remains locked\u0026mdash;requires further investigation. Given its total length of 50\u0026ndash;60 km, this fault has the potential to generate an event of up to \u003cem\u003eM\u003c/em\u003e 7.0 (Bernard \u0026amp; Zollo, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1989\u003c/span\u003e; Wells \u0026amp; Coppersmith, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e1994\u003c/span\u003e) in cases when rupture propagates across the fault offset and involves the entire fault system.\u003c/p\u003e\u003cp\u003eTo assess whether a rupture can dynamically propagate across the bend, we conducted 2D numerical simulations using a spectral element solution of the antiplane elastodynamic equation (Festa \u0026amp; Vilotte, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2006\u003c/span\u003e; Scala et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). We assumed a regional stress field with the maximum principal stress oriented vertically and the minimum compressive stress directed horizontally, perpendicular to the Apennine chain (De Matteis et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). When projecting this stress field onto the segmented fault system, the two longest segments appear well-oriented for rupture propagation (De Matteis et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), whereas the oblique segment is less favorably aligned for rupture continuation.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eWe explored rupture scenarios by varying key parameters, such as the main stress components, frictional coefficients, and rupture nucleation locations. Illustrative examples of rupture evolutions under these conditions are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e. Our findings reveal that, in most tested cases, rupture propagates through the entire segmented fault structure, highlighting a significant potential for generating large M6.5\u0026thinsp;+\u0026thinsp;earthquakes. However, under conditions where the stress drop along bended fault segment remains limited (a few percent of the available strength excess), rupture tends to arrest at the kink. This occurs because the available elastodynamic energy becomes insufficient to overcome the fracture energy needed for frictional weakening and continued rupture propagation (Ma \u0026amp; Archuleta, \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTo refine these rupture scenarios and better quantify seismic hazard, further investigations are required. These should include detailed characterization of stress fields and frictional conditions through laboratory and field measurements, coupled with advanced numerical modeling to systematically evaluate rupture sensitivity to fault geometry complexities. Additionally, analyzing the source properties (e.g., stress drop, rupture dimension, and frictional parameters) of ongoing microseismic activity would offer critical insights. Specifically, high-resolution seismic catalogs could be leveraged to constrain frictional heterogeneity and stress-state variations along fault segments. Combining these observational and modeling approaches will be fundamental to reducing uncertainties in hazard assessments and developing reliable earthquake rupture forecasts.\u003c/p\u003e\u003c/div\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003eEarthquake detection\u003c/h2\u003e\u003cp\u003eWe followed the same earthquake detection followed by Scotto di Uccio et al. (2023) for seismic sequences in the same region. The workflow is grounded on the use of the machine learning detector EQTransformer (Mousavi et al. \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), which provided a diverse set of templates to be used as the basis for further similarity-based detection using the template matching technique EQCorrscan (Chamberlain et al. \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). After splitting the network into 6 subnetworks of 6 arrays each, with an overlapping of 3 arrays between consecutive subnetworks, we performed earthquake detection independently for each subnetwork and integrated the declaration among the subnetworks according to the detection times. For each subnetwork, we applied EQTransformer on daily continuous data streams, resampled to 100 Hz and filtered in the frequency band [1\u0026ndash;45] Hz using the parameterization of Scotto di Uccio et al. (2023), requiring a probability value of 0.1 for P and S arrival times and using 50% overlap between consecutive time windows. Detections were declared when at least 5 picks were associated within time windows of 10 s and were visually confirmed. The machine learning catalog was used as template set for a similarity search using EQCorrscan. Each template event contained only the picked stations, selecting 1.6 s long windows extracted around the automatic picks (including 0.15 s of pre-pick waveforms). We decimated the traces to 25 Hz and filtered in the frequency band [2\u0026ndash;9] Hz. For the similarity detection threshold, we selected the sum of cross-correlation coefficient (SCC) between the portion of continuous streams and the templates. We declared an event when the SCC overcomes a similarity threshold with at least one template, fixed at 8 times the MAD of the cross-correlation coefficients between the template and the one-hour chunk of continuous streams. Moreover, we performed cross-correlation picking for the detections, requiring a minimum similarity coefficient of 0.6 at channel level. We applied the automatic selection criteria of Scotto di Uccio et al. (2023) for limiting the false declarations inside the catalog and visually inspected the remaining declarations.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eEvent location and magnitude\u003c/h3\u003e\n\u003cp\u003eUsing automatic picks, we performed a preliminary location of the earthquakes. We used NLLoc (Lomax et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) and a 1D layered velocity model tailored for the area (Matrullo et al. \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) for retrieving the absolute location of the events. For these initial locations we determined location uncertainties of few kilometers for horizontal and vertical locations and decimals of seconds for the time residuals.\u003c/p\u003e\u003cp\u003eTo improve the quality of the automatic picks and obtain high quality double difference locations, we applied a refined picking procedure based on cross-correlation and hierarchical clustering (Muzellec et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The refinement picking procedure is performed at each individual station by correlating waveforms around the picks of events located in 20km x 20km squares. Seismic records were bandpass filtered [1\u0026ndash;15] Hz and polarized using a polarization filter (Ross \u0026amp; Ben-Zion, \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Muzellec et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). We correlated the seismic waveforms on the three components independently in a time window of 0.25 s around the initial picks. The normalized cross-correlation coefficient threshold to define the family was set to 0.8. 10,657 P-wave picks and 34,924 S-wave picks survived the refinement picking procedure.\u003c/p\u003e\u003cp\u003eWe located the hypocenters with NLLoc (Lomax et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) in 3D velocity models optimized for the area (De Landro et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The refined picks allowed an increase in the accuracy of the absolute location by reducing the median location error from 1 km to 0.4 km (Figure S2) and the median root-mean-square (RMS) from 0.2 s to 0.1 s (Figure S2). We computed the cross-correlation (CC) differential times based on the refined locations. We finally relocated the seismicity with HypoDD (Waldhauser \u0026amp; Ellsworth, \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2000\u003c/span\u003e) including both the CC and catalog differential times. The CC differential travel times were calculated between events within families built during the refinement picking procedure. The application of the double difference location resulted in a reduction of the median RMS from 0.1 s to 0.007 s.\u003c/p\u003e\u003cp\u003eTo estimate the error from the DD location procedure, we iteratively used the SVD inversion method on a selection of 100 events by considering a maximum distance of 1 km. Figure S3 shows the histograms of the location errors estimated by DD location with the SVD inversion. The DD location procedure provides location uncertainties featuring a median error of 45 m, 60 m and 75 m, for the x, y and z directions respectively, strongly reducing the absolute location uncertainties, characterized by a median error of 400 m.\u003c/p\u003e\u003cp\u003eWe computed the local magnitude for equivalent Wood-Anderson displacement, obtained from integration of velocity waveforms, assigning absolute locations to those that were not relocated, using the empirical local magnitude relationship of Bobbio et al. (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eNumerical simulations\u003c/h3\u003e\n\u003cp\u003eWe conducted numerical simulations using the spectral element code developed by Festa \u0026amp; Vilotte (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2006\u003c/span\u003e), modified to model 2D antiplane ruptures. We discretized the elastic medium using quadrangular elements, with an eighth-order polynomial approximation to solve the elastodynamic equations within each element. We included perfectly matched layers in the modelling to prevent wave reflections at domain boundaries (Festa \u0026amp; Vilotte, \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2005\u003c/span\u003e). The fault consists of three branches, measuring 32 km, 9.6 km, and 20 km in length, respectively. We imposed a remote regional stress field with the maximum principal stress s\u003csub\u003e1\u003c/sub\u003e oriented vertically and the minimum compressive stress s\u003csub\u003e3\u003c/sub\u003e aligned horizontally, perpendicular to the trend of the Apennines (De Matteis et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). The two parallel fault branches have a strike aligned with the intermediate stress direction s\u003csub\u003e2\u003c/sub\u003e, while the oblique segment forms a 35\u0026deg;S angle relative to s\u003csub\u003e2\u003c/sub\u003e​ direction (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). We projected the stress field onto the three segments assuming a fault dip of 60\u0026deg;.\u003c/p\u003e\u003cp\u003eWe explored s\u003csub\u003e1\u003c/sub\u003e values between 60 and 80 MPa, s\u003csub\u003e3\u003c/sub\u003e​ values between 20 and 30 MPa, a static friction coefficient ranging from 0.6 to 0.7, and a dynamic friction coefficient between 0.25 and 0.35. These parameters were constrained by the maximum lithostatic load in the region and the expected stress drops, which range between 1 and 15 MPa, as inferred from previous events (Zollo et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Picozzi et al., 2021). The intermediate stress s\u003csub\u003e2\u003c/sub\u003e​ was set based on the R value from De Matteis et al. (\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). Rupture nucleated by locally increasing the initial tangential stress above the failure threshold within a small patch, whose size was comparable to the nucleation length defined by Uenishi \u0026amp; Rice (\u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). We modeled the fracture process using a linear slip-weakening law, where stress decreases from the yield level to the dynamic level over a critical slip distance D\u003csub\u003ec\u003c/sub\u003e=1m.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003ch2\u003eCompeting interests\u003c/h2\u003e\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e\u003cp\u003edeclarations\u003c/p\u003e\u003cp\u003eFrancesco Scotto di Uccio was supported by the project Transform2 funded by the European Commission under project number 101188365 within the HORIZON-INFRA-2024-DEV-01 call. Titouan Muzellec was supported by the Italian national project PON GRINT CIR01_00013 - Rafforzamento Capitale Umano. Gaetano Festa was supported by the project Geo-Inquire funded by the European Commission under project number 101058518 within the HORIZON-INFRA-2021-SERV-01 call.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eF.S.d.U. performed earthquake detection, catalog analysis and contributed to writing the original version of the manuscript. T.M. curated refinement of phase arrival times and performed earthquake relocation. A.S. and C.S. generated models and meshes for numerical simulations and revised the original version of the manuscript. G.D.L, G.C., M.P., A.Z. and G.B. contributed to formal analysis, discussion and revision of the original version of the manuscript. F.C. and L.E. contributed to data storage and accessibility. G.F. performed numerical simulations, contributed to the discussion and writing the original version of the manuscript. All authors contributed to finalize the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eSeismic waveforms recorded during the DETECT experiment are available at https://geofon.gfz.de/doi/network/ZK/2021. At the time of the submission, data are available under request at https://geofon.gfz.de/doi/network/access. Data will be publicly available from 2025-09-01. Earthquake detection is performed using EQTransformer (https://github.com/smousavi05/EQTransformer, accessed on 2025-05-09) and EQCorrscan (https://github.com/eqcorrscan/EQcorrscan, accessed on 2025-05-09). Earthquake relocations were performed using NonLinLoc (https://alomax.free.fr/nlloc, accessed on 2025-09-25) and HYPODD (https://github.com/fwaldhauser/HypoDD, accessed on 2025-05-09). The enhanced seismic catalog is available at the following link: https://zenodo.org/records/15372023. INFO catalog is available at the Irpinia Near Fault Observatory website (http://isnet‐bulletin.fisica.unina.it/cgi‐bin/isnet‐events/isnet.cgi) and at the EPOS Data Portal (https://www.epos‐eu.org/dataportal)‐IRPINIA Seismic Events provided by Universit\u0026agrave; di Napoli Federico II (accessed on 2025-09-05). Maps and images in Figures 1, 2 and 4 were produced using PyGMT (Uieda et al., 2021) and Matplotlib (Hunter, 2007). Figure 3 was constructed with Surfer\u0026reg; from Golden Software, LLC (version 27, www.goldensoftware.com), used under an educational license.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eKanamori, H. The diversity of large earthquakes and its implications for hazard mitigation. \u003cem\u003eAnnu. Rev. Earth Planet. Sci.\u003c/em\u003e \u003cb\u003e42\u003c/b\u003e (1), 7\u0026ndash;26 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBommer, J. J. Review of Seismic Hazard and Risk Analysis. \u003cem\u003eSeismological Soc. Am.\u003c/em\u003e \u003cb\u003e92\u003c/b\u003e (5), 3248\u0026ndash;3250 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZollo, A. et al. Earthquake early warning system in southern Italy: Methodologies and performance evaluation. \u003cem\u003eGeophysical Res. letters\u003c/em\u003e, \u003cb\u003e36\u003c/b\u003e(5). (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllen, R. M. \u0026amp; Melgar, D. Earthquake early warning: Advances, scientific challenges, and societal needs. \u003cem\u003eAnnu. Rev. Earth Planet. Sci.\u003c/em\u003e \u003cb\u003e47\u003c/b\u003e (1), 361\u0026ndash;388 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaufroy, E. et al. 3D numerical simulation and ground motion prediction: Verification, validation and beyond\u0026ndash;Lessons from the E2VP project. \u003cem\u003eSoil Dyn. Earthq. Eng.\u003c/em\u003e \u003cb\u003e91\u003c/b\u003e, 53\u0026ndash;71 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMagen, Y., Ziv, A., Inbal, A., Baer, G. \u0026amp; Hollingsworth, J. Fault rerupture during the July 2019 Ridgecrest earthquake pair from joint slip inversion of InSAR, optical imagery, and GPS. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e110\u003c/b\u003e (4), 1627\u0026ndash;1643 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eButtinelli, M. et al. The impact of structural complexity, fault segmentation, and reactivation on seismotectonics: Constraints from the upper crust of the 2016\u0026ndash;2017 Central Italy seismic sequence area. \u003cem\u003eTectonophysics\u003c/em\u003e \u003cb\u003e810\u003c/b\u003e, 228861 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBello, S. et al. High-detail fault segmentation: Deep insight into the anatomy of the 1983 Borah Peak earthquake rupture zone (2022). Mw 6.9, Idaho, USA). Lithosphere, 2022(1).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTertulliani, A., Leschiutta, I., Bordoni, P. \u0026amp; Milana, G. Damage distribution in L\u0026rsquo;Aquila city (Central Italy) during the 6 April 2009 earthquake. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e102\u003c/b\u003e (4), 1543\u0026ndash;1553 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGalli, P., Castenetto, S. \u0026amp; Peronace, E. The macroseismic intensity distribution of the 30 October 2016 earthquake in central Italy (Mw 6.6): Seismotectonic implications. \u003cem\u003eTectonics\u003c/em\u003e \u003cb\u003e36\u003c/b\u003e (10), 2179\u0026ndash;2191 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChiaraluce, L. et al. The Near Fault Observatory community in Europe: a new resource for faulting and hazard studies. \u003cem\u003eAnn. Geophys.\u003c/em\u003e \u003cb\u003e65\u003c/b\u003e (3), DM316 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBernard, P. \u0026amp; Zollo, A. The Irpinia (Italy) 1980 earthquake: detailed analysis of a complex normal faulting. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e94\u003c/b\u003e (B2), 1631\u0026ndash;1647 (1989).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRovida, A., Locati, M., Camassi, R., Lolli, B. \u0026amp; Gasperini, P. (2019). Italian Parametric Earthquake Catalogue CPTI15, version 2.0.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCinti, F. R., Faenza, L., Marzocchi, W. \u0026amp; Montone, P. Probability map of the next M\u0026thinsp;\u0026ge;\u0026thinsp;5.5 earthquakes in Italy. \u003cem\u003eGeochemistry Geophys. Geosystems\u003c/em\u003e, \u003cb\u003e5\u003c/b\u003e(11). (2004).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIannaccone, G. et al. A prototype system for earthquake early-warning and alert management in southern Italy. \u003cem\u003eBull. Earthq. Eng.\u003c/em\u003e \u003cb\u003e8\u003c/b\u003e, 1105\u0026ndash;1129 (2010).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVassallo, M., Festa, G. \u0026amp; Bobbio, A. Seismic ambient noise analysis in southern Italy. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e102\u003c/b\u003e (2), 574\u0026ndash;586 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDe Landro, G. et al. High-precision differential earthquake location in 3-D models: Evidence for a rheological barrier controlling the microseismicity at the Irpinia fault zone in southern Apennines. \u003cem\u003eGeophys. Supplements Monthly Notices Royal Astron. Soc.\u003c/em\u003e \u003cb\u003e203\u003c/b\u003e (3), 1821\u0026ndash;1831 (2015).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003edi Scotto, F., Scala, A., Festa, G., Picozzi, M. \u0026amp; Beroza, G. C. Comparing and integrating artificial intelligence and similarity search detection techniques: application to seismic sequences in Southern Italy. \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e233\u003c/b\u003e (2), 861\u0026ndash;874 (2023).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003edi Scotto, F. et al. Characterization and evolution of seismic sequences in the normal fault environment of the Southern Apennines. \u003cem\u003eJournal Geophys. Research: Solid Earth\u003c/em\u003e, \u003cb\u003e129\u003c/b\u003e(8), (2024). e2023JB028644.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAmoroso, O., Ascione, A., Mazzoli, S., Virieux, J. \u0026amp; Zollo, A. Seismic imaging of a fluid storage in the actively extending Apennine mountain belt, southern Italy. \u003cem\u003eGeophys. Res. Lett.\u003c/em\u003e \u003cb\u003e41\u003c/b\u003e (11), 3802\u0026ndash;3809 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVassallo, M., Festa, G., Bobbio, A. \u0026amp; Serra, M. Low shear velocity in a normal fault system imaged by ambient noise cross correlation: The case of the Irpinia fault zone, Southern Italy. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e121\u003c/b\u003e (6), 4290\u0026ndash;4305 (2016).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAmoroso, O. et al. From velocity and attenuation tomography to rock physical modeling: Inferences on fluid-driven earthquake processes at the Irpinia fault system in southern Italy. \u003cem\u003eGeophys. Res. Lett.\u003c/em\u003e \u003cb\u003e44\u003c/b\u003e (13), 6752\u0026ndash;6760 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eD'Agostino, N. et al. Crustal deformation and seismicity modulated by groundwater recharge of karst aquifers. \u003cem\u003eGeophys. Res. Lett.\u003c/em\u003e \u003cb\u003e45\u003c/b\u003e (22), 12\u0026ndash;253 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTarantino, S. et al. Non-linear elasticity, earthquake triggering and seasonal hydrological forcing along the Irpinia fault, Southern Italy. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e15\u003c/b\u003e (1), 9821 (2024).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDe Landro, G. et al. Decade-long monitoring of seismic velocity changes at the Irpinia fault system (southern Italy) reveals pore pressure pulsations. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cb\u003e12\u003c/b\u003e (1), 1247 (2022).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePicozzi, M., Bindi, D., Zollo, A., Festa, G. \u0026amp; Spallarossa, D. Detecting long-lasting transients of earthquake activity on a fault system by monitoring apparent stress, ground motion and clustering. \u003cem\u003eSci. Rep.\u003c/em\u003e, \u003cb\u003e9\u003c/b\u003e(1). (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePicozzi, M. et al. Spatiotemporal Evolution of Microseismicity Seismic Source Properties at the Irpinia Near-Fault Observatory, Southern Italy. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e112\u003c/b\u003e (1), 226\u0026ndash;242 (2022a).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePicozzi, M. et al. Spatiotemporal Evolution of Ground-Motion Intensity at the Irpinia Near-Fault Observatory, Southern Italy. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e112\u003c/b\u003e (1), 243\u0026ndash;261 (2022b).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCamanni, G. et al. Remobilization of inverted normal faults drives active extension in the axial zone of the southern Apennine mountain belt (Italy). \u003cem\u003eJ. Geol. Soc.\u003c/em\u003e \u003cb\u003e182\u003c/b\u003e (2), jgs2024\u0026ndash;jgs2184 (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhu, W. \u0026amp; Beroza, G. C. PhaseNet: a deep-neural-network-based seismic arrival-time picking method. \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e216\u003c/b\u003e (1), 261\u0026ndash;273 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y. \u0026amp; Beroza, G. C. Earthquake transformer\u0026mdash;an attentive deep-learning model for simultaneous earthquake detection and phase picking. \u003cem\u003eNat. Commun.\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e (1), 3952 (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTan, Y. J. et al. Machine-learning‐based high‐resolution earthquake catalog reveals how complex fault structures were activated during the 2016\u0026ndash;2017 central Italy sequence. \u003cem\u003eSeismic Record\u003c/em\u003e. \u003cb\u003e1\u003c/b\u003e (1), 11\u0026ndash;19 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichelini, A. et al. INSTANCE\u0026ndash;the Italian seismic dataset for machine learning. \u003cem\u003eEarth Syst. Sci. Data\u003c/em\u003e. \u003cb\u003e13\u003c/b\u003e (12), 5509\u0026ndash;5544 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChamberlain, C. J. et al. EQcorrscan: Repeating and near‐repeating earthquake detection and analysis in Python. \u003cem\u003eSeismol. Res. Lett.\u003c/em\u003e \u003cb\u003e89\u003c/b\u003e (1), 173\u0026ndash;181 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVuan, A., Sugan, M., Amati, G. \u0026amp; Kato, A. Improving the detection of low-magnitude seismicity preceding the Mw 6.3 L\u0026rsquo;Aquila earthquake: Development of a scalable code based on the cross correlation of template earthquakes. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e108\u003c/b\u003e (1), 471\u0026ndash;480 (2018).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEssing, D. \u0026amp; Poli, P. Spatiotemporal evolution of the seismicity in the alto tiberina fault system revealed by a high-resolution template matching catalog. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e, \u003cb\u003e127\u003c/b\u003e(10), (2022). e2022JB024845.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSpallarossa, D. et al. An automatically generated highresolution earthquake catalog for the 2016\u0026ndash;2017 Central Italy seismic sequence, including P and S phase arrival times. \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e225\u003c/b\u003e (1), 555\u0026ndash;571 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMichele, M., Chiaraluce, L., Di Stefano, R. \u0026amp; Waldhauser, F. Fine-scale structure of the 2016\u0026ndash;2017 Central Italy seismic sequence from data recorded at the Italian National Network. \u003cem\u003eJournal of Geophysical Research: Solid Earth\u003c/em\u003e, \u003cem\u003e125\u003c/em\u003e(4), e2019JB018440. (2020).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCabrera, L., Poli, P. \u0026amp; Frank, W. B. Tracking the spatio-temporal evolution of foreshocks preceding the Mw 6.1 2009 L\u0026rsquo;Aquila earthquake. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e, \u003cb\u003e127\u003c/b\u003e(3), (2022). e2021JB023888.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark, Y., Beroza, G. C. \u0026amp; Ellsworth, W. L. Basement fault activation before larger earthquakes in Oklahoma and Kansas. \u003cem\u003eSeismic Record\u003c/em\u003e. \u003cb\u003e2\u003c/b\u003e (3), 197\u0026ndash;206. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003eoclc.org/10.1785/0320220020\u003c/span\u003e\u003cspan address=\"oclc.10.1785/0320220020\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2022). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi-org.stanford.idm.\u003c/span\u003e\u003cspan address=\"https://doi-org.stanford.idm.\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMuzellec, T., De Landro, G., Camanni, G., Adinolfi, G. M. \u0026amp; Zollo, A. The complex 4D multi-segmented rupture of the 2014 Mw 6.2 Northern Nagano Earthquake revealed by high-precision aftershock locations. \u003cem\u003eTectonophysics\u003c/em\u003e, 230641. (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLomax, A., Michelini, A., Curtis, A. \u0026amp; Meyers, R. A. Earthquake location, direct, global-search methods. \u003cem\u003eEncyclopedia Complex. Syst. Sci.\u003c/em\u003e \u003cb\u003e5\u003c/b\u003e, 2449\u0026ndash;2473 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoss, Z. E., Trugman, D. T., Hauksson, E. \u0026amp; Shearer, P. M. Searching for hidden earthquakes in Southern California. \u003cem\u003eScience\u003c/em\u003e \u003cb\u003e364\u003c/b\u003e (6442), 767\u0026ndash;771 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYu, Y., Ellsworth, W. L. \u0026amp; Beroza, G. C. Accuracy and precision of earthquake location programs: Insights from a synthetic controlled experiment. \u003cem\u003eSeismol. Res. Lett.\u003c/em\u003e \u003cb\u003e96\u003c/b\u003e (3), 1860\u0026ndash;1874 (2025).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBobbio, A., Vassallo, M. \u0026amp; Festa, G. A local magnitude scale for southern Italy. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e99\u003c/b\u003e (4), 2461\u0026ndash;2470 (2009).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEster, M., Kriegel, H. P., Sander, J. \u0026amp; Xu, X. August). A density-based algorithm for discovering clusters in large spatial databases with noise. \u003cem\u003ekdd (Vol\u003c/em\u003e. \u003cb\u003e96\u003c/b\u003e (34), 226\u0026ndash;231 (1996).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchoenball, M. \u0026amp; Ellsworth, W. L. A systematic assessment of the spatiotemporal evolution of fault activation through induced seismicity in Oklahoma and southern Kansas. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e122\u003c/b\u003e (12), 10\u0026ndash;189 (2017).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFesta, G. et al. Insights into mechanical properties of the 1980 Irpinia fault system from the analysis of a seismic sequence. \u003cem\u003eGeosciences\u003c/em\u003e \u003cb\u003e11\u003c/b\u003e (1), 28 (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWells, D. L. \u0026amp; Coppersmith, K. J. New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e84\u003c/b\u003e (4), 974\u0026ndash;1002 (1994).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFesta, G. \u0026amp; Vilotte, J. P. Influence of the rupture initiation on the intersonic transition: Crack-like versus pulse‐like modes. \u003cem\u003eGeophysical Res. Letters\u003c/em\u003e, \u003cb\u003e33\u003c/b\u003e(15). (2006).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eScala, A., Festa, G., Vilotte, J. P., Lorito, S. \u0026amp; Romano, F. Wave interaction of reverse-fault rupture with free surface: Numerical analysis of the dynamic effects and fault opening induced by symmetry breaking. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e124\u003c/b\u003e (2), 1743\u0026ndash;1758 (2019).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDe Matteis, R. et al. Fault delineation and regional stress direction from the analysis of background microseismicity in the southern Apennines. \u003cem\u003eItaly Bull. Seismological Soc. Am.\u003c/em\u003e \u003cb\u003e102\u003c/b\u003e (4), 1899\u0026ndash;1907 (2012).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa, S. \u0026amp; Archuleta, R. J. Radiated seismic energy based on dynamic rupture models of faulting. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e111\u003c/b\u003e, B5 (2006).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMatrullo, E., De Matteis, R., Satriano, C., Amoroso, O. \u0026amp; Zollo, A. An improved 1-D seismic velocity model for seismological studies in the Campania\u0026ndash;Lucania region (Southern Italy). \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e195\u003c/b\u003e (1), 460\u0026ndash;473 (2013).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRoss, Z. E. \u0026amp; Ben-Zion, Y. Automatic picking of direct P, S seismic phases and fault zone head waves. \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e199\u003c/b\u003e (1), 368\u0026ndash;381 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWaldhauser, F. \u0026amp; Ellsworth, W. L. A double-difference earthquake location algorithm: Method and application to the northern Hayward fault, California. \u003cem\u003eBull. Seismol. Soc. Am.\u003c/em\u003e \u003cb\u003e90\u003c/b\u003e (6), 1353\u0026ndash;1368 (2000).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFesta, G. \u0026amp; Vilotte, J. P. The Newmark scheme as velocity\u0026ndash;stress time-staggering: an efficient PML implementation for spectral element simulations of elastodynamics. \u003cem\u003eGeophys. J. Int.\u003c/em\u003e \u003cb\u003e161\u003c/b\u003e (3), 789\u0026ndash;812 (2005).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZollo, A., Orefice, A. \u0026amp; Convertito, V. Source parameter scaling and radiation efficiency of microearthquakes along the Irpinia fault zone in southern Apennines, Italy. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e119\u003c/b\u003e (4), 3256\u0026ndash;3275 (2014).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUenishi, K. \u0026amp; Rice, J. R. Universal nucleation length for slip-weakening rupture instability under nonuniform fault loading. \u003cem\u003eJ. Geophys. Research: Solid Earth\u003c/em\u003e. \u003cb\u003e108\u003c/b\u003e, B1 (2003).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eUieda, L. et al. PyGMT: A Python interface for the generic mapping tools. (2021).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHunter, J. D. Matplotlib: A 2D graphics environment. \u003cem\u003eComput. Sci. Eng.\u003c/em\u003e \u003cb\u003e9\u003c/b\u003e (03), 90\u0026ndash;95 (2007).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6727679/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6727679/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIdentifying active faults capable of generating moderate-to-large earthquakes is essential for seismic hazard assessment, yet remains challenging in extensional tectonic environments, where fault systems include multiple segments and bends accommodating strain. In this study, we demonstrate how a short-term deployment of densely distributed seismic arrays can provide critical insights into seismicity patterns and fault geometry in the Southern Apennines, Italy.\u003c/p\u003e\u003cp\u003eIntegrating arrays with advanced machine learning methodologies, we produce an enhanced seismic catalog that increases the content of the manual one by nearly one order of magnitude, achieving, in just one year, a resolution comparable to a decade of conventional monitoring. Our results reveal spatial consistency of seismicity down to the decameter scale, with hypocenter locations and b-value mirroring those from the previous decade. We distinguish a shallow, diffuse seismicity, likely influenced by hydrological loading from karst aquifers, from deeper seismic clusters characterized by greater spatial coherence. The distribution of deep seismicity, when integrated with a 3D tomographic model, delineates a complex, curving fault structure 50\u0026ndash;60 km long, featuring a right-stepping jog several kilometers wide. Dynamic rupture simulations suggest that earthquakes nucleating on this fault could propagate through these structural complexities, potentially generating earthquakes up to magnitude 7.0.\u003c/p\u003e","manuscriptTitle":"Potential earthquake-prone faults identified by dense seismic array monitoring in complex extensional settings","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-24 20:00:41","doi":"10.21203/rs.3.rs-6727679/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-12T03:36:09+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-08-11T07:39:04+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"263599174694608047322419835025853883591","date":"2025-07-21T11:21:53+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"41107754505104926215197938671385306215","date":"2025-07-14T17:56:13+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-06-26T23:37:08+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"310640987315273957242348123915437985704","date":"2025-06-17T04:58:39+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"260847066615051046643097230263329815195","date":"2025-06-03T16:24:15+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-05-30T21:43:23+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-05-27T14:58:12+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-05-26T09:21:24+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-26T01:25:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"Scientific Reports","date":"2025-05-22T19:57:04+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"scientific-reports","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"scirep","sideBox":"Learn more about [Scientific Reports](http://www.nature.com/srep/)","snPcode":"","submissionUrl":"","title":"Scientific Reports","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Scientific Reports","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"c17dd4b0-de0a-46fa-aedb-bc423918e835","owner":[],"postedDate":"July 24th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":51857064,"name":"Earth and environmental sciences/Solid earth sciences/Geophysics"},{"id":51857065,"name":"Earth and environmental sciences/Solid earth sciences/Seismology"}],"tags":[],"updatedAt":"2026-01-19T16:46:30+00:00","versionOfRecord":{"articleIdentity":"rs-6727679","link":"https://doi.org/10.1038/s41598-026-35586-3","journal":{"identity":"scientific-reports","isVorOnly":false,"title":"Scientific Reports"},"publishedOn":"2026-01-17 16:29:30","publishedOnDateReadable":"January 17th, 2026"},"versionCreatedAt":"2025-07-24 20:00:41","video":"","vorDoi":"10.1038/s41598-026-35586-3","vorDoiUrl":"https://doi.org/10.1038/s41598-026-35586-3","workflowStages":[]},"version":"v1","identity":"rs-6727679","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6727679","identity":"rs-6727679","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}