The Irpinia region is one of the most seismically active areas of Italy owing to ongoing, late-orogenic extension in the axial zone of the Apennine mountain belt. However, the 3D architecture and the nature of the faults that drive this extension are still uncertain, posing challenges to seismic hazard assessment. Here, we address these uncertainties by integrating a new catalogue of high-resolution micro-seismicity (ML < 3.5) complemented by earthquake focal mechanisms, with existing 3D seismic velocity models and geological data. We found that micro-seismicity is primarily taking place along a segmented, approximately 60 km long, deep-seated, Mesozoic normal fault that was inverted during the shortening stages of the Apennine orogeny and then extensionally reactivated during the Quaternary. These findings suggest that multiple events of reactivation of long-lived faults can weaken their strength, making them prone to co-seismic remobilization under newly imposed strain fields in active mountain belts.

Supplementary material: A supplementary information file and table are available at https://doi.org/10.6084/m9.figshare.c.7401635

The Irpinia region, located in the axial zone of the Neogene southern Apennines mountain belt (e.g. Ippolito et al. 1975; Roure et al. 1991; Cello and Mazzoli 1999; Pescatore et al. 1999; Menardi Noguera and Rea 2000; Patacca and Scandone 2007; Ciarcia and Vitale 2024), is one of the most seismically active areas of Italy, owing to continuing, late-orogenic extension (c. 3 mm a−1, D'Agostino 2014; e.g. Ascione et al. 2013 and references therein). In 1980, the Irpinia region was struck by an Ms 6.9 earthquake (the largest Italian event in the last 100 years) that originated from a complex multi-segment rupture process (e.g. Bernard and Zollo 1989) and caused severe damage and thousands of fatalities in a wide epicentral area. During the last four decades, a moderate aftershock and background seismicity has continuously affected the crustal volume delimited by the faults that were activated during the 1980 Ms 6.9 Irpinia earthquake (e.g. De Matteis et al. 2012; Amoroso et al. 2014). Furthermore, since 2005, the monitoring of the Irpinia region by dense seismic networks (Irpinia Seismic Network (ISNet) and the Istituto Nazionale di Geofisica e Vulcanologia network (INGV)) revealed the occurrence of persistent seismicity with magnitude near zero (De Landro et al.2015; Amoroso et al. 2017; D'Agostino et al. 2018; Picozzi et al. 2019; Festa et al. 2021; Palo et al. 2023a, b; Scotto di Uccio et al. 2024; Tarantino et al. 2024).

However, the 3D architecture and the nature of the faults that are being remobilized nowadays, which may control this seismicity and drive the continuing extension, are not yet well understood. These uncertainties pose significant challenges for seismic hazard assessment in this highly active and densely populated region of southern Italy. In this work, we address these uncertainties by the integrated analysis of a newly reconstructed, high-resolution catalogue of approximately 15 years of micro-seismicity (ML < 3.5) complemented by newly calculated and compiled earthquake focal mechanisms, with existing 3D seismic velocity models (De Landro et al. 2022) and available geological models based on surface geology and seismic interpretation (e.g. Ascione et al. 2013, 2020).

The architecture of the axial zone of the southern Apennines is characterized by a two-layer structure (Mazzoli et al. 2014; Fig. 1). The upper layer consists of a far-travelled (>50 km) allochthonous assemblage forming an intensely deformed, thin-skinned fold and thrust belt. This overlies a less deformed, thick-skinned lower layer. The thin-skinned belt primarily involves Mesozoic–Cenozoic successions of the Apennine Platform (shallow-water carbonates) and of the Lagonegro Basin (shallow-water to pelagic limestones, radiolarian cherts and shales). Conversely, the deeper thick-skinned belt involves the 6–8 km thick shallow-water carbonate succession of the Apulian Platform, as well as the Lower Triassic siliciclastic deposits located at its base and the underlying basement. The recent tectonic emplacement of the thin-skinned belt is indicated by the occurrence, in its footwall, of Pliocene to Lower Pleistocene foreland basin deposits resting on top of the Apulian Platform carbonates and penetrated by numerous oil wells (Mazzoli et al. 2001). Since the middle Pleistocene the axial zone of the southern Apennines has been undergoing a phase of late-orogenic extension (Cello et al. 1982; Cinque et al. 1993; Hippolyte et al. 1994; Butler et al. 2004; Caiazzo et al. 2006). This extension led to the development of neo-formed Quaternary extensional faults that dissect the mountain belt (black faults in the section of Fig. 1) and is responsible for the continuing seismicity within the study area.

Importantly for this work, the thick-skinned belt (where the seismicity is mostly focused) is composed of deep-seated, steeply dipping reverse faults mainly formed by the reverse- or oblique-slip reactivation (‘inversion’) of pre-existing Mesozoic normal faults (Mazzoli et al. 2000, 2008; Butler et al. 2004; Shiner et al. 2004; Ascione et al. 2013, 2020; Amoroso et al. 2014, 2017). These structures control the so-called Apulian inversion belt (Mazzoli et al. 2008). The inversion tectonics model for the Apulian Platform shortening-related structures was derived by the interpretation of high-resolution seismic reflection profiles calibrated with numerous oil wells and subsequent depth conversion, cross-section balancing and restoration (e.g. Shiner et al. 2004). Shortening-related reactivation of inherited faults in the Apulian Platform carbonates was unravelled by the interpretation of high-quality seismic data also in the outer zone of the southern Apennines by Bitonte et al. (2021). The latter researchers also documented fault propagation into the foreland basin deposits as a result of such fault reactivation. Within the study area, the development of the Apulian inversion belt was controlled by the reverse- or oblique-slip reactivation of a SW-dipping Mesozoic normal fault (red fault in the cross-section of Fig. 1). This structure produced a prominent uplift of the Apulian Platform carbonates in its hanging wall (Ascione et al. 2013, 2020). The carbonate culmination is dissected by a deeply rooted Quaternary horst, the NW-dipping boundary fault of which was associated with the Ms 6.9 main shock of the 1980 Irpinia earthquake (Ascione et al. 2013, 2020; Amoroso et al. 2014).

The micro-seismicity catalogue on which this work is based was obtained by analysing a dataset consisting of about 2400 micro-earthquakes, with local magnitude (ML) ranging between 0.5 and 3.2. These events were recorded by 42 ISNet and INGV stations from August 2005 to December 2022. We used manually picked first P- and S-wave arrival times from the ISNet bulletin (http://isnet-bulletin.fisica.unina.it/cgi-bin/isnet-events/isnet.cgi) and integrated manually picks from INGV stations (see Supplementary material SM1). Initially, we located the events with a probabilistic method (NLLoc, Lomax et al. 2009) and a 3D velocity model optimized for the area (De Landro et al. 2022), which allowed us to obtain a first location catalogue with an average RMS residual of 0.15 s and horizontal location errors within 1.5 km (average 800 m) and vertical location errors within 2 km (average 1.2 km) for 85% of events. Successively, we refined the absolute 3D location with the double-difference approach (HypoDD, Michele et al. 2019; Waldhauser and Ellsworth 2000) by using catalogue (CT) and cross-correlation (CC) differential times (Schaff and Waldhauser 2005). The final residual RMS was 0.008 s for CC data and 0.03 s for CT data. The final horizontal and vertical location errors were within 100 m for most of the events. For further details on the location strategy and method used see Supplementary material MS2. The final dataset is available in the catalogue of De Landro (2024).

As a further constraint on the fault geometry and kinematics, the composite focal mechanisms of significant clusters of this new micro-seismicity catalogue were also calculated using the consolidated code FPFIT (Reasenberg and Oppenheimer 1985). Similarly to Festa et al. (2021) and Muzellec et al. (2024), we evaluated, for four selected hypocentre clusters, the composite focal mechanisms (red ‘beach balls’ in Fig. 2) by integrating the polarities of co-located events for the construction of more constrained mechanisms. Furthermore, where it was not possible to calculate composite focal mechanisms owing to the lack of hypocentre clusters, we integrated four additional focal mechanisms of single events (dark grey ‘beach balls’ in Fig. 2). These were selected among the focal mechanisms available from the ISNet bulletin and refined by using the 3D location. To validate this selection, we compared these mechanisms with those obtained by De Matteis et al. (2012), in which the author performed an extensive analysis of focal mechanisms and refined the stress field of the Irpinia region. (For further details on focal mechanism data, calculation, evaluation and selection, and uncertainties see Supplementary material SM3 and Table SM3 and De Landro (2024).)

As a constraint on the depth and geometry of the top of the Apulian Platform carbonates, we used the 3D P-wave velocity model of De Landro et al. (2022). This model was built using a linearized, tomographic approach in which about 13 000 P- and S-wave arrival times were inverted to retrieve, jointly, the location of about 1500 earthquakes of the ISNet catalogue and the P- and S-phase velocity parameters. Similarly to Amoroso et al. (2014, 2017), a multi-scale strategy was applied, starting from a coarser parameterization to a finer one of 3 × 3 × 1 km3, with a forward grid of 500 m step size. The model resolution was assessed by the integration of the resolution matrix, composed by intrinsic resolution and spreading function, and the ray density around each node.

Earthquake hypocentres within the study area mostly extend from about 3 to 14 km depth and are generally distributed along an about 20 km wide, 60 km long, NW–SE-elongated belt (Figs 1 and 2). Within this distribution, distinct features can be recognized from horizontal and vertical sections through the data.

On horizontal sections (Fig. 2a), from approximately 10 km depth and downward, hypocentres are roughly aligned along two NW–SE-elongated clusters, each c. 5–10 km wide and c. 30 km long (see red dashed lines in Fig. 2a). These two clusters are arranged in a geometry in which a SE cluster steps northwestward to the right into a NW cluster, the latter being associated with the largest concentration of hypocentres. Notably, at all depths, the volume in which this step occurs is characterized by a decreased density of hypocentres. From approximately 8 km depth and upwards, these two clusters become shorter and less clearly defined, whereas another NW–SE-elongated cluster more prominently emerges located to the SW of the NW cluster. This other cluster is c. 5–10 km wide and 30 km long, being best identifiable on the 8 and 6 km depth horizontal sections (see blue dashed lines in Fig. 2a).

On vertical sections perpendicular to the NW–SE map-view clusters (i.e. NE–SW; Fig. 2b), earthquake hypocentres forming the two major, right-stepping map-view clusters overall align into a primary, locally discontinuous, SW-dipping cluster c. 5–10 km wide, recognizable throughout the study area (see red dashed lines in Fig. 2b). This cluster widens upwards near the centre of the study area (section III in Fig. 2b). In contrast, the shallower NW–SE-elongated cluster in the NE sector of the study area corresponds to a NW-dipping, c. 5–10 km wide cluster in cross-section (see blue dashed lines in Fig. 2b).

From approximately 10 km depth upwards on horizontal sections, the two major NW–SE-elongated hypocentre clusters roughly coincide with a distinct NW–SE boundary in the P-wave velocity model, between higher velocities in the SW and lower velocities in the NE (Fig. 2a). In all vertical sections, this feature manifests as a prominent deepening of low P-wave velocities from SW to NW (Fig. 2b). This is best highlighted by the 4.5, 5 and 5.5 km s−1 isovelocity contours and approximately occurs across the primary SW-dipping cluster of hypocentres (Fig. 2b). It is worth noting that this deepening is largest in the central and northwestern parts of the study area (Fig. 2).

In the 10, 8 and 6 km depth sections (Fig. 2a), the boundary between higher and lower velocities shows a curved morphology in map view, taking on a more north–south orientation at the location of the step between the two deep NW–SE-elongated hypocentre clusters. This curved deepening of low velocities is clearly depicted by the 5.5 km s−1 isovelocity surface (Fig. 3).

The four composite and the four compiled single event earthquake focal mechanisms within the study area all display an approximately pure dip-slip, extensional kinematics (Fig. 2a and b; see also De Landro 2024 and Supplementary material Table SM3), in agreement with the regional strain field of the study area (De Matteis et al. 2012; Bello et al. 2021; Festa et al. 2021; Ricigliano Eqk report, RISSC-Lab 2024). In general, each focal mechanism has nodal planes striking roughly NW–SE sub-parallel to the overall orientation of the earthquake hypocentre clusters, although minor variations to these orientations exist. For example, some focal mechanisms include an approximately north–south-oriented nodal plane (e.g. focal mechanisms 1, 3, 5, 7 and 8, Fig. 2a), whereas others include an approximately east–west-oriented nodal plane (e.g. focal mechanism 2, Fig. 2a). Dip angles of the focal mechanism nodal planes are also overall consistent with the dip of the main earthquake hypocentre clusters (i.e. either SW- or NE-dipping; Fig. 2b). However, whereas some focal mechanisms are associated with two moderately dipping nodal planes (e.g. focal mechanisms 2, 3, 7 and 8, Fig. 2b), others display a pair of nodal planes in which one is steeply dipping and the other is gently dipping (focal mechanisms 1, 4, 5 and 6, Fig. 2b).

As we have seen, the deep geological structure within the study area is that of an Apulian Platform carbonates culmination bounded to the NE by, and uplifted along, an inverted Mesozoic normal fault (cross-section in Fig. 1). When integrating the seismological data presented above with this geological information, several coherent features become apparent (Fig. 4). In particular, (1) the shallow high P-wave velocities nearly coincide with the uplifted Apulian Platform carbonates and (2) the primary SW-dipping cluster of hypocentres approximately aligns with the inverted Mesozoic normal fault (red fault in the section of Fig. 1) responsible for this uplift. Because both seismological features are recognizable throughout the study area, we infer that a structural model in which Apulian Platform carbonates are uplifted along an inverted SW-dipping Mesozoic normal fault holds for the entire study area (Fig. 4). Furthermore, considering that both earthquake hypocentres and P-wave velocities are right-stepped (Figs 2 and 3), the inverted fault may be confidently interpreted as also stepped in map view (Fig. 4). This is further supported by results of gravity data modelling (Improta et al. 2003; see also De Landro et al. 2015), which show that the depocentre ahead of the inverted fault is also right-stepped in map view. This feature most probably reflects the original segmentation of the precursor normal fault, which comprised two fault segments separated by a relay zone (e.g. Camanni et al. 2023, and references therein).

The inverted Mesozoic normal fault coincides both in map view and in cross-section with an approximately 5–10 km wide zone of earthquake hypocentres (Figs 2 and 4), rather than with a well-delineated feature. We interpret this as due to the combined effect of this structure probably being a broad fault zone (a common feature for faults; e.g. Childs et al. 2009) rather than an individual fault surface (see also De Matteis et al. 2012 for a similar interpretation of the micro-seismicity in the Irpinia region) and of the uncertainty in the event's location. However, the overall dip and strike of the fault zone are consistent with those of the nodal planes of the focal mechanisms (i.e. NW–SE striking and SW-dipping, Fig. 4).

For the northern part of the study area (i.e. north of the step of the inverted fault), the attitude of the faults included in this new structural model (Fig. 4) is consistent with that derived from the 0 and 40 s sub-events of the 1980 Irpinia earthquake (Westaway and Jackson 1984; Bernard and Zollo 1989; Pantosti and Valensise 1990; Pingue et al. 1993; Amoruso et al. 2005, 2011). On the other hand, for the southern part of the study area this new structural model is consistent with the minority interpretation provided by Amoruso et al. (2005, 2011) for the 20 s sub-event (which most researchers interpreted to have occurred along a NE-dipping fault, in some case interpreted to be low angle; e.g. Bernard and Zollo 1989).

The micro-seismicity within the study area is associated with continuing late-orogenic extension, in agreement with focal mechanisms that display normal sense kinematics (e.g. this study; De Matteis et al. 2012; Festa et al. 2021; Fig. 4). This may appear counterintuitive, as we found that most seismicity is taking place along an inverted fault that preserves its reverse net displacement associated with the uplift of Apulian Platform carbonates (Figs 1 and 4). This fault is long-lived and has been active with a normal sense of movement in the Mesozoic and reactivated with a reverse- or oblique-slip kinematics during the Apennine orogeny. We suggest that these multiple slip events weakened the strength of the fault, making it prone to further remobilization (‘negative inversion’) under the newly imposed extensional strain field affecting the Irpinia region. Whereas in the SE part of the study area this is the only fault driving continuing extension, in the NW sector this fault is acting in association with two further neo-formed faults that define a Quaternary horst within the Apulian Platform carbonates (one of these two faults, i.e. the NE dipping one, generated the Ms 6.9 main shock of the 1980 Irpinia earthquake and contains most of the micro-seismicity; Fig. 4). This occurs in a sector where the net displacement of the inverted fault is maximum, as shown by the highest deepening of P-wave velocities that define the deepest top Apulian Platform carbonates in the fault footwall; Fig. 2). Based on this observation, we suggest that the need for extensional readjustment is most pronounced in this sector, thus requiring the involvement of all three faults.

These findings have broader implications for other active mountain belts characterized by a thick-skinned style of deformation and inversion tectonics (e.g. Taiwan: Lacombe and Mouthereau 2002; Camanni et al. 2014, 2016; Western Alps: Mosar 1999; Zagros: Tavani et al. 2020). When assessing seismic hazard in such mountain belts, particular attention should be directed towards long-lived faults. The results of this work also indicate that this should be done regardless of the consistency between the geological displacement of the fault and the kinematics of the newly imposed strain field.

We thank the Irpinia Near Fault Observatory revision team that works on the INFO bulletin construction (http://isnet-bulletin.fisica.unina.it/cgi-bin/isnet-events/isnet.cgi). The code FPFIT (Reasenberg and Oppenheimer 1985) was used for calculating the composite focal mechanisms. The Generic Mapping Tools (GMT) version 4 (Wessel et al. 2019), licensed under LGPL version 3 or later and available at https://www.generic-mapping-tools.org/, were used to produce some of the figures. Two anonymous reviewers and the Journal of the Geological Society editors are also acknowledged.

GC: conceptualization (lead), data curation (equal), formal analysis (equal), investigation (equal), methodology (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – original draft (lead); GDL: conceptualization (equal), data curation (equal), formal analysis (lead), funding acquisition (equal), investigation (equal), methodology (equal), project administration (equal), resources (equal), software (equal), supervision (equal), validation (equal), visualization (equal), writing – review & editing (lead); SM: conceptualization (supporting), investigation (equal), validation (equal), writing – review & editing (equal); MM: data curation (supporting), formal analysis (equal), investigation (supporting), software (supporting); TM: data curation (supporting), formal analysis (equal), investigation (supporting), software (supporting), writing – review & editing (supporting); AA: conceptualization (supporting), formal analysis (supporting), investigation (supporting), writing – review & editing (supporting); DPS: data curation (supporting), formal analysis (equal), investigation (supporting), software (supporting), writing – review & editing (supporting); ST: data curation (supporting), formal analysis (equal), investigation (supporting), software (supporting), writing – review & editing (supporting); AZ: conceptualization (supporting), funding acquisition (equal), investigation (equal), project administration (equal), resources (equal), supervision (equal), validation (equal), writing – review & editing (equal).

The authors acknowledge financial support from the Project TRHAM, ‘Relation Between 3D Thermo-Rheological Model And Seismic Hazard For The Risk Mitigation In The Urban Areas Of Southern Italy’ funded by the European Union, Next Generation EU, Mission 4, Component 2, CUP B53D23033710001 (Grant Number P2022P37SN); and from the Project FRACTURES, ‘Multiscale study of seismogenic processes in Campania–Lucania Apennines using machine learning algorithms and multiparametric observations’ funded by the European Union, Next Generation EU, Mission 4, Component 2, CUP B53D23006980006 (Grant Number 2022BEKFN2). The work of A.Z. was supported in part by Project ‘PE0000005–RETURN-SPOKE 3-CUP UNINA: E63C220002000002’.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

The micro-seismicity catalogue and the focal mechanisms used in this study are stored in the following repository under the CC BY 4.0 licence: https://zenodo.org/records/11208080 (De Landro 2024). The 3D P-wave velocity model used in this study is available in the following repository under the CC BY 4.0 licence: https://doi.org/10.5281/zenodo.14637925. This study used the Irpinia Near Fault Observatory (https://isnet.unina.it) data and products. Seismic data are available at the EIDA website (https://eida.ingv.it/it/) and at the EPOS Data Portal (https://www.epos-eu.org/dataportal), IRPINIA Seismic Velocity and Acceleration Waveforms (Continuous) provided by Università di Napoli Federico II and INGV, networks IX and IV. The seismic bulletin catalogue 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.