Raffaele Di Stefano
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Raffaele Di Stefano has a total of 86 co-authors. Here are the top co-authors by number of shared publications:
- Lauro Chiaraluce
Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy
10 shared publications - Claudio Chiarabba
Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy
6 shared publications - Maria Grazia Ciaccio
Istituto Nazionale di Geofisica e Vulcanologia, sezione di Roma1, Roma, Italy
6 shared publications - Luisa Valoroso
Istituto Nazionale di Geofisica e Vulcanologia (INGV) Rome Italy
5 shared publications E. Kissling
4 shared publications- Maddalena Michele
Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy
4 shared publications
Latest Publications
- Open Access
- 3 August 2026
- Published by Springer Nature in Earth Science Informatics
- Vol. 19 (9) , 152
Researcher-Curated knowledge governance for trustworthy scientific AI in the geosciences: the rockGPT approach
Large Language Models (LLMs) are increasingly adopted in scientific research, yet their use raises important challenges related to transparency, provenance, reproducibility, and scientific knowledge governance. In the geosciences, where conclusions often depend on integrating multidisciplinary evidence, ensuring that AI-generated responses can be traced to verifiable scientific literature is particularly important. This paper presents and evaluates a methodology for trustworthy scientific knowledge management that places researcher-curated knowledge governance at its core. The approach is based on four complementary principles: researcher-curated knowledge, retrieval-based grounding, traceability and provenance, and institutional governance. The methodology was implemented through RockGPT, a self-hosted geoscience-oriented AI system combining open-source LLMs with Retrieval-Augmented Generation (RAG) and a curated corpus of scientific publications on Italian seismotectonics and the Central Apennines. The system integrates Ollama, OpenWebUI, and six open-source LLM families (Qwen, LLaMA, DeepSeek, Mixtral, Gemma, and Phi). Evaluation combined a comparative quantitative assessment with a qualitative governance-oriented analysis. Across all evaluated models, access to the researcher-curated corpus consistently improved response quality, with the largest gains observed in tasks requiring synthesis across multiple documents. Beyond these quantitative improvements, the qualitative analysis showed that the proposed approach supports key requirements for scientific AI, including transparency, provenance, reproducibility, data sovereignty, and human oversight. The results indicate that reliable scientific AI depends not only on model capabilities but also on how scientific knowledge is selected, curated, governed, and connected to generated responses. The proposed methodology demonstrates how researcher-curated knowledge governance can provide a foundation for transparent, reproducible, and institutionally controlled AI systems for scientific knowledge management in the geosciences and beyond
- 3 August 2026
- Published by Springer Nature in Earth Science Informatics
- Vol. 19 (9) , 152
Open AccessResearcher-Curated knowledge governance for trustworthy scientific AI in the geosciences: the rockGPT approach
Large Language Models (LLMs) are increasingly adopted in scientific research, yet their use raises important challenges related to transparency, provenance, reproducibility, and scientific knowledge governance. In the geosciences, where conclusions often depend on integrating multidisciplinary evidence, ensuring that AI-generated responses can be traced to verifiable scientific literature is particularly important. This paper presents and evaluates a methodology for trustworthy scientific knowledge management that places researcher-curated knowledge governance at its core. The approach is based on four complementary principles: researcher-curated knowledge, retrieval-based grounding, traceability and provenance, and institutional governance. The methodology was implemented through RockGPT, a self-hosted geoscience-oriented AI system combining open-source LLMs with Retrieval-Augmented Generation (RAG) and a curated corpus of scientific publications on Italian seismotectonics and the Central Apennines. The system integrates Ollama, OpenWebUI, and six open-source LLM families (Qwen, LLaMA, DeepSeek, Mixtral, Gemma, and Phi). Evaluation combined a comparative quantitative assessment with a qualitative governance-oriented analysis. Across all evaluated models, access to the researcher-curated corpus consistently improved response quality, with the largest gains observed in tasks requiring synthesis across multiple documents. Beyond these quantitative improvements, the qualitative analysis showed that the proposed approach supports key requirements for scientific AI, including transparency, provenance, reproducibility, data sovereignty, and human oversight. The results indicate that reliable scientific AI depends not only on model capabilities but also on how scientific knowledge is selected, curated, governed, and connected to generated responses. The proposed methodology demonstrates how researcher-curated knowledge governance can provide a foundation for transparent, reproducible, and institutionally controlled AI systems for scientific knowledge management in the geosciences and beyond
- Open Access
- 14 March 2026
- Published by Copernicus GmbH
Enhancing Earthquake Location in the Central Apennines (Italy): A Hybrid Approach Combining Arrivals from Line-Sensor Telecom Fiber Interferometry and Traditional Point-sensors
The integration of existing telecommunication fiber-optic infrastructure into seismic monitoring networks offers a transformative opportunity to densify observations in seismically active regions. We present the results of a multi-year monitoring experiment (2021–2026) utilizing a 39-km telecom fiber link from the Italian telecommunication company Open Fiber between Ascoli Piceno and Teramo in the Central Apennines, Italy. The system employs an ultra stable laser to measure seismic-induced deformation of the fiber, operating on a dedicated wavelength in coexistence with commercial data traffic.A significant challenge in utilizing fiber-optic data for earthquake location is the transition from traditional point-sensor geometry to distributed sensing. To address this, we implemented a hybrid localization approach using a modified version of the NonLinLoc (NLL) algorithm. We move beyond traditional discrete measurements (point sensors) by treating the cable as a continuous "line sensor." Following the NLL algorithm, the most effective strategy is translating both point and line geometries into a unified framework of 3D travel-time maps. Once the sensors are translated into these maps, their combined use for location becomes independent of the sensor type, allowing for a seamless merging of traditional seismic station data and fiber-optic pickings. We applied this methodology to the real seismic catalog recorded from the fiber's installation in mid 2021 until January 2026 in the Ascoli-Teramo area, a region where the Italian seismic network is relatively sparse. Specifically, we analyzed signals from: 1) several small seismic sequences occurring at short distances (up to approximately 20 km) from the fiber cable, including the Civitella del Tronto (TE) sequence that followed a Mw 3.9 event (September 22, 2022); and 2) more distant earthquakes (ranging from approximately 20 to 50 km from the fiber) with local magnitudes exceeding ML 2.5, distributed along the Central Apennines axis. For events where the fiber signal allowed for the correct identification of P- and S-wave arrival times, we applied the NLL algorithm using the integrated network. In this work, we present several of these examples and associated tests to discuss how the inclusion of fiber-derived arrival times can provide further hypocentral constraints. This study aims to highlight the scalability of fiber interferometry combined with non-linear inversion as a robust tool for seismic surveillance in populated and high-risk tectonic environments.
- 14 March 2026
- Published by Copernicus GmbH
Open AccessEnhancing Earthquake Location in the Central Apennines (Italy): A Hybrid Approach Combining Arrivals from Line-Sensor Telecom Fiber Interferometry and Traditional Point-sensors
The integration of existing telecommunication fiber-optic infrastructure into seismic monitoring networks offers a transformative opportunity to densify observations in seismically active regions. We present the results of a multi-year monitoring experiment (2021–2026) utilizing a 39-km telecom fiber link from the Italian telecommunication company Open Fiber between Ascoli Piceno and Teramo in the Central Apennines, Italy. The system employs an ultra stable laser to measure seismic-induced deformation of the fiber, operating on a dedicated wavelength in coexistence with commercial data traffic.A significant challenge in utilizing fiber-optic data for earthquake location is the transition from traditional point-sensor geometry to distributed sensing. To address this, we implemented a hybrid localization approach using a modified version of the NonLinLoc (NLL) algorithm. We move beyond traditional discrete measurements (point sensors) by treating the cable as a continuous "line sensor." Following the NLL algorithm, the most effective strategy is translating both point and line geometries into a unified framework of 3D travel-time maps. Once the sensors are translated into these maps, their combined use for location becomes independent of the sensor type, allowing for a seamless merging of traditional seismic station data and fiber-optic pickings. We applied this methodology to the real seismic catalog recorded from the fiber's installation in mid 2021 until January 2026 in the Ascoli-Teramo area, a region where the Italian seismic network is relatively sparse. Specifically, we analyzed signals from: 1) several small seismic sequences occurring at short distances (up to approximately 20 km) from the fiber cable, including the Civitella del Tronto (TE) sequence that followed a Mw 3.9 event (September 22, 2022); and 2) more distant earthquakes (ranging from approximately 20 to 50 km from the fiber) with local magnitudes exceeding ML 2.5, distributed along the Central Apennines axis. For events where the fiber signal allowed for the correct identification of P- and S-wave arrival times, we applied the NLL algorithm using the integrated network. In this work, we present several of these examples and associated tests to discuss how the inclusion of fiber-derived arrival times can provide further hypocentral constraints. This study aims to highlight the scalability of fiber interferometry combined with non-linear inversion as a robust tool for seismic surveillance in populated and high-risk tectonic environments.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
CARS - Catalog of Relative Seismic Locations of 1981-2018 Italian Seismicity
The Istituto Nazionale di Geofisica e Vulcanologia (INGV) monitors the Italian peninsula seismicity by using the data recorded by the Italian National Seismic Network (RSN) together with the ones gathered by other permanent regional networks (PRN). Earthquakes are real-time located by the INGV surveillance system and manually revised by the Italian Seismic Bulletin (BSI) group analysts.Starting from a catalog composed by homogeneous absolute locations (CLASS; Latorre et al., 2023), obtained by using a 3D regional-scale velocity model, we generated a catalog of relative seismic locations (CARS) of about 310,000 events occurred in Italy during 1981-2018.We inverted absolute P- and S- waves arrival times derived from data collected by RSN plus PRN for the period 1981-2008 and only by RSN for 2009-2018 to apply the double-difference relocation algorithm (Waldhauser and Ellsworth, 2000).For the second period, we combined the absolute travel times with relative ones obtained by waveforms cross-correlations analysis performed on pairs of similar events. The time domain cross‐correlation method proposed by Schaff et al., 2004 and Schaff & Waldhauser, 2005 was applied to seismograms of all pairs of events separated by 10 km or less and recorded at common stations. Seismograms were filtered in the 1–15 Hz frequency range using a four pole, zero phase band‐pass Butterworth filter. The correlation measurements were performed on 1.0 s long window for P- and S-waves. We collected a total of ~17 million P- and ~23 million S-wave delay times, retaining all measurements with correlation coefficients greater than 0.7.1D velocity models characterising 18 different (geologically, seismically and tectonically homogeneous) Italian macroareas (after Pastori et al. B2-2019-2021, Wp1-task4) were used in the location procedure.To cope with the memory limits (15,000 events with less than 200 readings) of the HypoDD code so to use it in a steady operational mode, we first subdivided the study region in the 18 macroareas related to the velocity models, then we additional discretize each in 100x100km2 cells, overlapped by the 80% in longitude and latitude. We repeatedly produced hypocentral locations of the same events that we merged by computing a final weighted mean location.We present the double-difference catalog of Italian seismicity, allowing to depict alignments clearer with respect to the starting catalog (CLASS), to be related to seismogenic faults and/or to regional structures along the whole Italian peninsula.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessCARS - Catalog of Relative Seismic Locations of 1981-2018 Italian Seismicity
The Istituto Nazionale di Geofisica e Vulcanologia (INGV) monitors the Italian peninsula seismicity by using the data recorded by the Italian National Seismic Network (RSN) together with the ones gathered by other permanent regional networks (PRN). Earthquakes are real-time located by the INGV surveillance system and manually revised by the Italian Seismic Bulletin (BSI) group analysts.Starting from a catalog composed by homogeneous absolute locations (CLASS; Latorre et al., 2023), obtained by using a 3D regional-scale velocity model, we generated a catalog of relative seismic locations (CARS) of about 310,000 events occurred in Italy during 1981-2018.We inverted absolute P- and S- waves arrival times derived from data collected by RSN plus PRN for the period 1981-2008 and only by RSN for 2009-2018 to apply the double-difference relocation algorithm (Waldhauser and Ellsworth, 2000).For the second period, we combined the absolute travel times with relative ones obtained by waveforms cross-correlations analysis performed on pairs of similar events. The time domain cross‐correlation method proposed by Schaff et al., 2004 and Schaff & Waldhauser, 2005 was applied to seismograms of all pairs of events separated by 10 km or less and recorded at common stations. Seismograms were filtered in the 1–15 Hz frequency range using a four pole, zero phase band‐pass Butterworth filter. The correlation measurements were performed on 1.0 s long window for P- and S-waves. We collected a total of ~17 million P- and ~23 million S-wave delay times, retaining all measurements with correlation coefficients greater than 0.7.1D velocity models characterising 18 different (geologically, seismically and tectonically homogeneous) Italian macroareas (after Pastori et al. B2-2019-2021, Wp1-task4) were used in the location procedure.To cope with the memory limits (15,000 events with less than 200 readings) of the HypoDD code so to use it in a steady operational mode, we first subdivided the study region in the 18 macroareas related to the velocity models, then we additional discretize each in 100x100km2 cells, overlapped by the 80% in longitude and latitude. We repeatedly produced hypocentral locations of the same events that we merged by computing a final weighted mean location.We present the double-difference catalog of Italian seismicity, allowing to depict alignments clearer with respect to the starting catalog (CLASS), to be related to seismogenic faults and/or to regional structures along the whole Italian peninsula.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
Accurate Earthquake Locations of the Adriatic Thrust Fault of the 2021 Seismic Sequence with sP Depth Phases.
We re-located 70 earthquakes belonging to the seismic sequence started on 2021 March 27th, with a mainshock of Mw 5.2 at 13:47 UTC, in the Central Adriatic region (Italy off-shore) by using the on-purpose designed code by Zhao et al. (2007, 2011), modeling the sP converted phases.The mainshock of the 2021 seismic sequence occurred about 20 km north of the Palagruza island, 80 km from the Gargano promontory and about 40 km from the Croatian island of Lastovo. It was felt in many central-southern Italian regions, from Ancona to Foggia, and in Central Dalmatia. All the epicenters of this seismic sequence lie in the open sea, about 100km to the SE and about 50 km to the NW of the 2003 Jabuka seismic sequence, and the 1988 Palagruza seismic sequence, respectively.Though the seismicity in the central Adriatic Sea has been recorded by improving seismic networks, especially in recent decades, the precise location of the Adriatic offshore earthquakes was hampered mainly by the large distance of the closest stations, and by the large gap in the distribution of seismic stations.The possibility to model the sP depth phases enables us to estimate the epicentral parameters and focal depths of these offshore earthquakes more accurately, thanks to the peculiar ray-path that mimics the presence of a receiver approximately on top of the hypocenter. The refined earthquake locations allow us to make inferences on the structure responsible for the seismicity of the 2021 seismic sequence, a thrust fault NW-SE striking and ~35° NE-dipping, and on its seismotectonic context.The use of depth-phase arrival times to constrain the off-network events' locations is of particular interest to Italy due to both the peculiar shape of the peninsula and the extreme scarcity of seafloor stations, whose cost and management are very expensive and complex.We present the first attempt to apply this off-network location technique to the Italian offshore seismicity with the aim of improving the hazard estimation of these hard-to-monitor regions.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessAccurate Earthquake Locations of the Adriatic Thrust Fault of the 2021 Seismic Sequence with sP Depth Phases.
We re-located 70 earthquakes belonging to the seismic sequence started on 2021 March 27th, with a mainshock of Mw 5.2 at 13:47 UTC, in the Central Adriatic region (Italy off-shore) by using the on-purpose designed code by Zhao et al. (2007, 2011), modeling the sP converted phases.The mainshock of the 2021 seismic sequence occurred about 20 km north of the Palagruza island, 80 km from the Gargano promontory and about 40 km from the Croatian island of Lastovo. It was felt in many central-southern Italian regions, from Ancona to Foggia, and in Central Dalmatia. All the epicenters of this seismic sequence lie in the open sea, about 100km to the SE and about 50 km to the NW of the 2003 Jabuka seismic sequence, and the 1988 Palagruza seismic sequence, respectively.Though the seismicity in the central Adriatic Sea has been recorded by improving seismic networks, especially in recent decades, the precise location of the Adriatic offshore earthquakes was hampered mainly by the large distance of the closest stations, and by the large gap in the distribution of seismic stations.The possibility to model the sP depth phases enables us to estimate the epicentral parameters and focal depths of these offshore earthquakes more accurately, thanks to the peculiar ray-path that mimics the presence of a receiver approximately on top of the hypocenter. The refined earthquake locations allow us to make inferences on the structure responsible for the seismicity of the 2021 seismic sequence, a thrust fault NW-SE striking and ~35° NE-dipping, and on its seismotectonic context.The use of depth-phase arrival times to constrain the off-network events' locations is of particular interest to Italy due to both the peculiar shape of the peninsula and the extreme scarcity of seafloor stations, whose cost and management are very expensive and complex.We present the first attempt to apply this off-network location technique to the Italian offshore seismicity with the aim of improving the hazard estimation of these hard-to-monitor regions.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
Moment vs local magnitude scaling of small-to-moderate earthquakes from seismic moment estimation of 10 years (2009-2018) of Italian seismicity
We computed moment (Mw) and local magnitude (ML) of about 250,000 earthquakes occurred in Italy from 2009 to 2018 and recorded at seismic stations of the Italian National Network managed by INGV.For moment magnitude computation, we start from raw velocity waveforms and invert the displacement spectra of more than 2,000,000 S-waves manually picked. We use the probabilistic method of Supino et al. [2019] to estimate the a-posteriori joint probability density function of the source parameters: seismic moment M0, corner frequency fc and high-frequency decay γ. Mw is obtained from M0 using the Kanamori [1977] equation.We start from the same waveforms to compute local magnitude using two designed on purpose codes, PyAmp and PyML [Di Stefano et al., 2023], and an attenuation law specific for the Italian region, Di Bona et al. [2016], obtaining ML values characterized by quality and homogeneity.Both magnitude catalogs can be reproduced due to the availability in open databases of all the input and output parameters used for processing.We observe a self-similar scaling between fc and M0 for Mw larger than ~2.0. For smaller magnitudes, S-wave spectra show an almost constant corner frequency (~10 Hz), which does not scale with the earthquake source (seismic moment). We interpret this as the constant cut-off frequency of the anelastic attenuation, which acts as a low-pass filter and produces an apparent corner frequency. The latter is lower than expected, and corresponds to an apparent larger source duration.Because of the conservation of total displacement integral after a low-pass filtering, signals must exhibit a maximum amplitude lower than expected to “compensate” the apparent larger source duration. ML values are therefore expected to be underestimated while moment magnitudes, by definition, are not affected by this as they are proportional to the displacement integral.Coherently, the comparison of our Mw and ML estimates shows the systematic underestimation of ML with respect to Mw for small magnitude events. The deviation from a 1:1 scaling relationship between ML and Mw overlaps the magnitude range where the constant apparent corner frequency arises in the M0-fc scaling (ML <~ 2).Regarding the upcoming of a new generation of earthquake catalogs characterized by very low completeness magnitudes (MC << 2), our results suggest that a robust analysis of the statistical features of these catalogs (e.g., event size distribution) should consider the use of a precise magnitude estimate such as Mw instead of ML.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessMoment vs local magnitude scaling of small-to-moderate earthquakes from seismic moment estimation of 10 years (2009-2018) of Italian seismicity
We computed moment (Mw) and local magnitude (ML) of about 250,000 earthquakes occurred in Italy from 2009 to 2018 and recorded at seismic stations of the Italian National Network managed by INGV.For moment magnitude computation, we start from raw velocity waveforms and invert the displacement spectra of more than 2,000,000 S-waves manually picked. We use the probabilistic method of Supino et al. [2019] to estimate the a-posteriori joint probability density function of the source parameters: seismic moment M0, corner frequency fc and high-frequency decay γ. Mw is obtained from M0 using the Kanamori [1977] equation.We start from the same waveforms to compute local magnitude using two designed on purpose codes, PyAmp and PyML [Di Stefano et al., 2023], and an attenuation law specific for the Italian region, Di Bona et al. [2016], obtaining ML values characterized by quality and homogeneity.Both magnitude catalogs can be reproduced due to the availability in open databases of all the input and output parameters used for processing.We observe a self-similar scaling between fc and M0 for Mw larger than ~2.0. For smaller magnitudes, S-wave spectra show an almost constant corner frequency (~10 Hz), which does not scale with the earthquake source (seismic moment). We interpret this as the constant cut-off frequency of the anelastic attenuation, which acts as a low-pass filter and produces an apparent corner frequency. The latter is lower than expected, and corresponds to an apparent larger source duration.Because of the conservation of total displacement integral after a low-pass filtering, signals must exhibit a maximum amplitude lower than expected to “compensate” the apparent larger source duration. ML values are therefore expected to be underestimated while moment magnitudes, by definition, are not affected by this as they are proportional to the displacement integral.Coherently, the comparison of our Mw and ML estimates shows the systematic underestimation of ML with respect to Mw for small magnitude events. The deviation from a 1:1 scaling relationship between ML and Mw overlaps the magnitude range where the constant apparent corner frequency arises in the M0-fc scaling (ML <~ 2).Regarding the upcoming of a new generation of earthquake catalogs characterized by very low completeness magnitudes (MC << 2), our results suggest that a robust analysis of the statistical features of these catalogs (e.g., event size distribution) should consider the use of a precise magnitude estimate such as Mw instead of ML.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
Adjoint Tomography of the Italian Lithosphere
The evolution and state of geological structure at Earth’s surface is best understood with an accurate characterization of the subsurface, where fluid distribution plays a key role. We present high-resolution seismic tomographic images of tectonic and geological features of the Italian lithosphere based on ground motion recordings and obtained through an iterative procedure. Enhanced accuracy is enabled by state-of-the-art three-dimensional wavefield simulations in combination with an adjoint-state method. The resulting tomographic model characterizes the subsurface structure in terms of compressional and shear wavespeed values at remarkable resolution, corresponding to a minimum period of ~10 s. As primary findings of our work, images of the lithospheric structure in Central Italy are consistent with recent studies on the distribution of fluids and gas (CO2) within the Italian subsurface, allowing us to infer the presence of deep melted material that induces shallow gas fluxes, or traps and deep storage of gas that can be correlated with seismicity. We illuminate Mt. Etna volcano and support the hypothesis of a deep reservoir (~30 km) feeding an intermediate-depth magma-filled intrusive body, which in turn is connected to a shallow chamber. We also investigate the intriguing features of the Adriatic plate offshore of the eastern Italian coast. Tomographic evidence reveals a structure of the plate made of two distinct microplates with different fabric and behavior, and separated by the Gargano deformation zone, indicating a complex lithosphere and tectonic evolution.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessAdjoint Tomography of the Italian Lithosphere
The evolution and state of geological structure at Earth’s surface is best understood with an accurate characterization of the subsurface, where fluid distribution plays a key role. We present high-resolution seismic tomographic images of tectonic and geological features of the Italian lithosphere based on ground motion recordings and obtained through an iterative procedure. Enhanced accuracy is enabled by state-of-the-art three-dimensional wavefield simulations in combination with an adjoint-state method. The resulting tomographic model characterizes the subsurface structure in terms of compressional and shear wavespeed values at remarkable resolution, corresponding to a minimum period of ~10 s. As primary findings of our work, images of the lithospheric structure in Central Italy are consistent with recent studies on the distribution of fluids and gas (CO2) within the Italian subsurface, allowing us to infer the presence of deep melted material that induces shallow gas fluxes, or traps and deep storage of gas that can be correlated with seismicity. We illuminate Mt. Etna volcano and support the hypothesis of a deep reservoir (~30 km) feeding an intermediate-depth magma-filled intrusive body, which in turn is connected to a shallow chamber. We also investigate the intriguing features of the Adriatic plate offshore of the eastern Italian coast. Tomographic evidence reveals a structure of the plate made of two distinct microplates with different fabric and behavior, and separated by the Gargano deformation zone, indicating a complex lithosphere and tectonic evolution.
- Open Access
- 15 May 2023
- Published by Copernicus GmbH
Diffusion processes in minor normal faulting seismic sequences monitored by the Alto Tiberina Near Fault Observatory (Northern Apennines, Italy)
The analysis of microseismicity has a fundamental role in understanding earthquakes, giving insights on the long- and short-term driving forces and processes preparing and generating the seismicity occurrence and its evolution in space and time.Recent advances in detection and location algorithms, paired with dense seismic networks, and supported by higher computing capacity, allow dramatic increase in the quality and quantity of low magnitude earthquakes recorded resulting in high resolution earthquakes catalogs in terms of both location attributes and completeness.Such catalogs enable us to analyze small magnitude (M<4) sequences having the advantage of a high frequency of occurrence, with unprecedented resolution in illuminating minor (few kilometers of extent) fault systems and seismicity patterns.We present a detailed analysis of two seismic sequences occurred within an extensional sector of the Northern Apennines between 2010 and 2014: the Città di Castello and Pietralunga sequences (maximum magnitude ≤ 3.6). The area is within the Alto Tiberina Near Fault Observatory (TABOO-NFO), a multidisciplinary monitoring infrastructure dedicated to the investigation of the fault slip behavior of this very low angle normal fault.The very high microseismic activity, the availability of a dense network and a complex tectonic setting involving shallow (H2O) and deep (CO2) fluids circulation, result in an ideal location to apply modern detection and analysis techniques to study microseismicity in detail.We build the high-resolution catalog starting from the raw waveforms recorded by a seismic network composed of ~60 stations covering an area of 80x80km and applying a deep learning phase picker. The events are located with a probabilistic nonlinear algorithm and finally relocated with the double differences algorithm after undergoing a quality selection based on location parameters. The resulting catalog for these sequences counts 6 times the number of events documented in previously available standard catalogs.The spatiotemporal distribution of events shows different characteristics, ranging from foreshock-mainshock-aftershock to more swarm-like patterns but almost all these patterns are compatible with pore-pressure diffusion (1 − 2m2s-1) processes and exhibiting along-strike migration. These are very similar behavior with respect to the ones observed during the larger extensional sequences occurred in the Apennines in recent years.The case of fluid driven seismicity is coherent with the seismotectonic setting of the area showing large CO2 degassing phenomena and the presence of geologic formations prone to develop fluids overpressure. The comparison of the spatial distribution of events with a three-dimensional deterministic seismostratigraphic model based on different (non-seismic) geophysical data, highlights in fact a ubiquitous involvement of the Triassic Evaporites as hosting lithology, indicating a strong mechanical control and corroborating their seismogenic role.
- 15 May 2023
- Published by Copernicus GmbH
Open AccessDiffusion processes in minor normal faulting seismic sequences monitored by the Alto Tiberina Near Fault Observatory (Northern Apennines, Italy)
The analysis of microseismicity has a fundamental role in understanding earthquakes, giving insights on the long- and short-term driving forces and processes preparing and generating the seismicity occurrence and its evolution in space and time.Recent advances in detection and location algorithms, paired with dense seismic networks, and supported by higher computing capacity, allow dramatic increase in the quality and quantity of low magnitude earthquakes recorded resulting in high resolution earthquakes catalogs in terms of both location attributes and completeness.Such catalogs enable us to analyze small magnitude (M<4) sequences having the advantage of a high frequency of occurrence, with unprecedented resolution in illuminating minor (few kilometers of extent) fault systems and seismicity patterns.We present a detailed analysis of two seismic sequences occurred within an extensional sector of the Northern Apennines between 2010 and 2014: the Città di Castello and Pietralunga sequences (maximum magnitude ≤ 3.6). The area is within the Alto Tiberina Near Fault Observatory (TABOO-NFO), a multidisciplinary monitoring infrastructure dedicated to the investigation of the fault slip behavior of this very low angle normal fault.The very high microseismic activity, the availability of a dense network and a complex tectonic setting involving shallow (H2O) and deep (CO2) fluids circulation, result in an ideal location to apply modern detection and analysis techniques to study microseismicity in detail.We build the high-resolution catalog starting from the raw waveforms recorded by a seismic network composed of ~60 stations covering an area of 80x80km and applying a deep learning phase picker. The events are located with a probabilistic nonlinear algorithm and finally relocated with the double differences algorithm after undergoing a quality selection based on location parameters. The resulting catalog for these sequences counts 6 times the number of events documented in previously available standard catalogs.The spatiotemporal distribution of events shows different characteristics, ranging from foreshock-mainshock-aftershock to more swarm-like patterns but almost all these patterns are compatible with pore-pressure diffusion (1 − 2m2s-1) processes and exhibiting along-strike migration. These are very similar behavior with respect to the ones observed during the larger extensional sequences occurred in the Apennines in recent years.The case of fluid driven seismicity is coherent with the seismotectonic setting of the area showing large CO2 degassing phenomena and the presence of geologic formations prone to develop fluids overpressure. The comparison of the spatial distribution of events with a three-dimensional deterministic seismostratigraphic model based on different (non-seismic) geophysical data, highlights in fact a ubiquitous involvement of the Triassic Evaporites as hosting lithology, indicating a strong mechanical control and corroborating their seismogenic role.
- 15 February 2023
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 94 (3) , 1566-1578
Temporal Variations of Seismicity Rates and Gutenberg–Richterb-Values for a Stochastic Declustered Catalog: An Example in Central Italy
One important aspect of the seismicity is the spatiotemporal clustering; hence, the distinction between independent and triggered events is a critical part of the analysis of seismic catalogs. Stochastic declustering of seismicity allows a probabilistic distinction between these two kinds of events. Such an approach, usually performed with the epidemic‐type aftershock sequence (ETAS) model, avoids the bias in the estimation of the frequency–magnitude distribution parameters if we consider a subset of the catalog, that is, only the independent or the triggered events. In this article, we present a framework to properly include the probabilities of any event to be independent (or triggered) both in the temporal variation of the seismic rates and in the estimation of the b‐value of the Gutenberg–Richter law. This framework is then applied to a high‐definition seismic catalog in the central part of Italy covering the period from April 2010 to December 2015. The results of our analysis show that the seismic activity from the beginning of the catalog to March 2013 is characterized by a low degree of clustering and a relatively high b‐value, whereas the following period exhibits a higher degree of clustering and a smaller b‐value.
- 15 February 2023
- Published by Seismological Society of America (SSA) in Seismological Research Letters
- Vol. 94 (3) , 1566-1578
Temporal Variations of Seismicity Rates and Gutenberg–Richterb-Values for a Stochastic Declustered Catalog: An Example in Central Italy
One important aspect of the seismicity is the spatiotemporal clustering; hence, the distinction between independent and triggered events is a critical part of the analysis of seismic catalogs. Stochastic declustering of seismicity allows a probabilistic distinction between these two kinds of events. Such an approach, usually performed with the epidemic‐type aftershock sequence (ETAS) model, avoids the bias in the estimation of the frequency–magnitude distribution parameters if we consider a subset of the catalog, that is, only the independent or the triggered events. In this article, we present a framework to properly include the probabilities of any event to be independent (or triggered) both in the temporal variation of the seismic rates and in the estimation of the b‐value of the Gutenberg–Richter law. This framework is then applied to a high‐definition seismic catalog in the central part of Italy covering the period from April 2010 to December 2015. The results of our analysis show that the seismic activity from the beginning of the catalog to March 2013 is characterized by a low degree of clustering and a relatively high b‐value, whereas the following period exhibits a higher degree of clustering and a smaller b‐value.
- 2 December 2022
- Published by Seismological Society of America (SSA) in Bulletin of the Seismological Society of America
- Vol. 113 (1) , 480-493
The Adriatic Thrust Fault of the 2021 Seismic Sequence Estimated from Accurate Earthquake Locations Using sP Depth Phases
An earthquake sequence occurred in the Central Adriatic region during March–June 2021. This sequence started on 27 March with a mainshock of moment magnitude ( ) 5.2 occurring at 13:47 coordinated universal time (UTC). No foreshock was observed before this mainshock. The sequence lasted approximately three months, until the end of June 2021. Approximately 200 seismic events were recorded by the regional seismic network during this time, including four M ≥ 4.0 earthquakes. The 27 March 2021 earthquake was one of the strongest instrumentally recorded events in the area bounded approximately by the Ancona–Zadar line to the north and the Gargano–Dubrovnik line to the south. The mainshock originated at a focal depth of 9.9 km. The seismicity spread from the mainshock up‐dip and down‐dip along a northeast‐dipping plane. Here, we investigate the geometry of the fault activated by this seismic sequence by using sP depth phases. We aim to significantly reduce the large uncertainties associated with the hypocentral locations of offshore earthquakes beneath the Adriatic Sea—an area that plays a fundamental role in the geodynamics of the Mediterranean. These refined earthquake locations also allow us to make inferences with regards to the seismotectonic context responsible for the analyzed seismicity, thus identifying a structure (here referred to as the Mid‐Adriatic fault) consisting of a northwest–southeast‐striking thrust fault with a ∼35° northeast‐dipping plane. The use of depth‐phase arrival times to constrain off‐network event locations is of particular interest in Italy due to both the peculiar shape of the peninsula and the extreme scarcity of seafloor stations, the cost and management of which are very expensive and complex. Here, we present the first attempt to apply this off‐network locating technique to the Italian offshore seismicity research with the aim of improving hazard estimations in these hard‐to‐monitor regions.- 2 December 2022
- Published by Seismological Society of America (SSA) in Bulletin of the Seismological Society of America
- Vol. 113 (1) , 480-493
The Adriatic Thrust Fault of the 2021 Seismic Sequence Estimated from Accurate Earthquake Locations Using sP Depth Phases
An earthquake sequence occurred in the Central Adriatic region during March–June 2021. This sequence started on 27 March with a mainshock of moment magnitude ( ) 5.2 occurring at 13:47 coordinated universal time (UTC). No foreshock was observed before this mainshock. The sequence lasted approximately three months, until the end of June 2021. Approximately 200 seismic events were recorded by the regional seismic network during this time, including four M ≥ 4.0 earthquakes. The 27 March 2021 earthquake was one of the strongest instrumentally recorded events in the area bounded approximately by the Ancona–Zadar line to the north and the Gargano–Dubrovnik line to the south. The mainshock originated at a focal depth of 9.9 km. The seismicity spread from the mainshock up‐dip and down‐dip along a northeast‐dipping plane. Here, we investigate the geometry of the fault activated by this seismic sequence by using sP depth phases. We aim to significantly reduce the large uncertainties associated with the hypocentral locations of offshore earthquakes beneath the Adriatic Sea—an area that plays a fundamental role in the geodynamics of the Mediterranean. These refined earthquake locations also allow us to make inferences with regards to the seismotectonic context responsible for the analyzed seismicity, thus identifying a structure (here referred to as the Mid‐Adriatic fault) consisting of a northwest–southeast‐striking thrust fault with a ∼35° northeast‐dipping plane. The use of depth‐phase arrival times to constrain off‐network event locations is of particular interest in Italy due to both the peculiar shape of the peninsula and the extreme scarcity of seafloor stations, the cost and management of which are very expensive and complex. Here, we present the first attempt to apply this off‐network locating technique to the Italian offshore seismicity research with the aim of improving hazard estimations in these hard‐to‐monitor regions.- Open Access
- 23 March 2022
- Published by Springer Nature in Communications Earth & Environment
- Vol. 3 (1) , 1-12
Adjoint tomography of the Italian lithosphere
The evolution and state of geological structure at Earth’s surface is best understood with an accurate characterization of the subsurface. Here we present seismic tomographic images of the Italian lithosphere based on ground motion recordings and characterized by compressional and shear wavespeed structure at remarkable resolution, corresponding to a minimum period of ~10 s. Enhanced accuracy is enabled by state-of-the-art three-dimensional wavefield simulations in combination with an adjoint-state method. We focus on three primary findings of our model Im25. It highlights the distribution of fluids and gas (CO2) within the Italian subsurface and their correlation with seismicity. It illuminates Mt. Etna volcano and supports the hypothesis of a deep reservoir (~30 km) feeding a shallower magma-filled intrusive body. Offshore of the eastern Italian coast, it reveals that the Adriatic plate is made of two distinct microplates, separated by the Gargano deformation zone, indicating a complex lithosphere and tectonic evolution.
- 23 March 2022
- Published by Springer Nature in Communications Earth & Environment
- Vol. 3 (1) , 1-12
Open AccessAdjoint tomography of the Italian lithosphere
The evolution and state of geological structure at Earth’s surface is best understood with an accurate characterization of the subsurface. Here we present seismic tomographic images of the Italian lithosphere based on ground motion recordings and characterized by compressional and shear wavespeed structure at remarkable resolution, corresponding to a minimum period of ~10 s. Enhanced accuracy is enabled by state-of-the-art three-dimensional wavefield simulations in combination with an adjoint-state method. We focus on three primary findings of our model Im25. It highlights the distribution of fluids and gas (CO2) within the Italian subsurface and their correlation with seismicity. It illuminates Mt. Etna volcano and supports the hypothesis of a deep reservoir (~30 km) feeding a shallower magma-filled intrusive body. Offshore of the eastern Italian coast, it reveals that the Adriatic plate is made of two distinct microplates, separated by the Gargano deformation zone, indicating a complex lithosphere and tectonic evolution.