Multidisciplinary inferences on a newly recognized active east‐dipping extensional system in Central Italy

We use a multidisciplinary approach to gather preliminary evidence for a Quaternary east‐dipping extensional detachment in Central Italy. This structure crops out in the Sabini‐Eastern Simbruini (SES) and would be hidden at mid‐crustal depths beneath the L'Aquila 2009 (Mw6.3) epicentral area. The SES geometry is reconstructed through geological mapping, structural analysis and seismic line interpretation. The geometry of the mid‐crustal segment, referred to as the Ocre Segment (OS), is interpreted through seismological analyses of the largest aftershock (Mw5.4) of the L'Aquila 2009 sequence. The kinematic compatibility between the SES and the OS under a common SW–NE tensional field is tested through stress inversion of both geological and seismological data. The reliability of OS activation is tested through slip tendency analysis. Like other Italian cases, the SES and the OS are preliminarily interpreted as expressions at different depths of the same unknown east‐dipping extensional detachment, characterized by a ramp–flat–ramp geometry.


Introduction
The Abruzzo region of Central Italy, struck by the 6 April 2009 L'Aquila earthquake (EQ1, M w 6.3), has a high seismic hazard due to well-known west-dipping high-angle seismogenic upper-crustal normal faults (Pace et al., 2006). In contrast to the neighbouring regions of the Northern and Southern Apennines, no east-dipping normal fault system has been determined to be active and/or potentially seismogenic (Fig. 1a).
EQ1 ruptured the SW-dipping Late Quaternary Paganica normal fault from a depth of~9 km up to the surface and for a length of 25 km (Boncio et al., 2010;Chiaraluce et al., 2011;Doglioni et al., 2011). Its largest aftershock, which occurred on 7 April (EQ2, M w 5.4), nucleated within the Paganica footwall rock volume, at a depth of 14 km (Fig. 1b). In the literature, EQ2 has been associated with a NNW-SSE sub-vertical normal fault (Pino and Di Luccio, 2009;Di Luccio et al., 2010), with an east-dipping low-angle normal fault , with the extensional reactivation of an east-dipping Mio-Pliocene thrust (Valoroso et al., 2013) and with a NE-dipping normal fault antithetic to the Paganica fault (Guglielmino et al., 2013). Available EQ2 focal mechanisms (Pondrelli et al., 2010;Scognamiglio et al., 2010;Herrmann et al., 2011;D'Amico et al., 2013) show a highangle east-dipping plane with a relevant strike-slip component, making the interpretation of the EQ2 seismotectonic context more intriguing.
In this paper, we aim to constrain the fault generating EQ2 and its geometric-kinematic interlinks with the surrounding fault system. By integrating surface and subsurface geology and seismological data, we highlight a previously unrecognized east-dipping extensional system that crops out along the Latium-Abruzzo boundary and possibly released EQ2 at mid-crustal depths.

Seismotectonic framework
The intermountain Quaternary extensional belt of Central Italy consists of both high-angle (west-dipping) and moderate-to low-angle (east-dipping) normal and normal-oblique faults that cross-cut and offset Mio-Pliocene foldand-thrust structures and control continental basin growth and earthquake activity (Lavecchia et al., 1994;Doglioni et al., 1999;Ghisetti and Vezzani, 2002). The west-dipping faults are the best exposed and the most seismogenic (Galadini and Galli, 2000;Boncio et al., 2004;Roberts and Michetti, 2004). Since the last century, they have generated several destructive earthquakes including 1915 Avezzano (M w 7), 2009 L'Aquila (M w 6.3) and 2016 Central Italy (M w 6.5) (Fig. 1a). The east-dipping faults, which have less-evident field expressions, crop out along the western border of the Apennine belt (Boncio et al., 2000;Collettini et al., 2006;Mirabella et al., 2011;Di Naccio et al., 2013;Petricca et al., 2015) and define a regional NNW-SSE fault alignment known as the Etrurian Fault System (EFS; Brozzetti et al., 2009). Crustal transects and earthquake data across the EFS and the Apennines show that the eastand west-dipping normal faults both sole to a common detachment that dips eastward at a low angle to depths of 14-15 km (Barchi et al., 1998;Boncio et al., 2004;Chiaraluce et al., 2007;Eva et al., 2014).
A conceptual scheme of the spatial relationships between the east-dipping detachment and the high-angle antithetic faults is shown in the block diagram of Fig. 1c.
Potentially seismogenic east-dipping normal faults are also located in southern Italy (Fig. 1a), along the 'Campania-Lucania Extensional Fault System' (CLEFS in Brozzetti, 2011). These faults were first recognized after the 1980 Irpinia normal-fault earthquake (M w 6.8); their systematic occurrence is now well accepted (Maschio et al., 2005;De Matteis et al., 2012;Galli and Peronace, 2014). Eastward of the extensional belt, active E-W strike-slip faults are recognized. In instrumental times, they have released moderate earthquakes at mid-crustal depths within the Adriatic foreland (Adinolfi et al., 2015 and references therein).

EQ2 preferential seismic plane investigation
With the aim of constraining the preferential seismic plane responsible for EQ2, we first retrieved a new focal mechanism through Time Domain Moment Tensor (TDMT) full waveform inversion (Dreger and Helmberger, 1993;Dreger, 2003) (Fig. 2a). The obtained focal parameters are given in Table 1, together with data on the quality and stability of the solution. Considering the number of stations and the percentage of double couples, the retrieved solution is reliable and shows a good fit between the synthetic seismograms and the observed data (Fig. S1), as confirmed by uncertainty analysis (Fig. S2). Both nodal planes (N347°/ 58 and N077°/61) have significant strike-slip components (pitch~35°); the tensional axis is sub-horizontal and trends SW-NE; the focal mechanism can be classified as normal-oblique, following the kinematic classification in Zoback (1992).
We performed a kinematic rupture process analysis to model the Apparent Source Time Functions (ASTFs) that were retrieved by the waveform data ( Fig. 2b), as developed in Adinolfi et al. (2015). The ASTFs were calculated by deconvolution of the impulse response of the medium from the recorded data using the empirical Green's function (EGF) method (Hartzell, 1978) at 15 stations. We selected the 9 April 2009 M L 4.3 aftershock, which occurred at 03:14:53.07 (UTC), as the EGF.
We inverted the ASTFs to obtain a kinematic rupture model using the isochrone back-projection technique (Festa and Zollo, 2006). The method  (Brozzetti, 2011); stars = major intra-Apennine early-instrumental and instrumental earthquakes (Michele et al., 2016;Rovida et al., 2011). (b) Epicentral and hypocentral distributions (sections 1 and 2) of the L'Aquila sequence (data from Chiaraluce et al., 2011) with the focal mechanisms of the two major events (EQ1 = 6 April, M w 6.3; EQ2 = 7 April, M w 5.5) (Pondrelli et al., 2010); the histogram highlights the two-layer depth distribution of the aftershock sequence. (c) Sketch of the 3D geometric relationships between the EFS and the antithetic seismogenic high-angle normal faults in Central Italy (Lavecchia et al., 2011). back-projects the amplitude of the ASTFs along the isochrones on the fault plane to retrieve the slip distribution that is associated with a single receiver and stacks the retrieved maps to obtain the final slip model.
We investigated the fault plane responsible for the rupture, assuming a constant rupture velocity and comparing the misfit for the two planes indicated by the focal mechanism solution for different values of the rupture velocity (1.0-3.0 km s À1 ). Fig. 2c shows the difference between the actual misfit function and the misfit value for the minimum that is normalized by this latter value for the fault and auxiliary planes. The east-dipping nodal plane has a curve whose trend is smaller than the misfit for the other plane over the entire range of velocities between 5% and 10%. This result indicates that the ASTFs discriminate the fault plane with a significant reduction in the misfit function. For the east-dipping plane, the rupture velocity with the minimum misfit is 1.8 km s À1 . However, the small percentage variation (≤5%) of the cost function over the entire range of rupture velocities and the flattening of the function around the minimum indicate a large uncertainty in this parameter.
The ASTF shapes show an initial rapid increase that reveals a small fault dislocation that is relatively confined to a central nucleation zone (Fig. S3).

OS fault segment building
To reconstruct the geometry of the fault segments activated by EQ1 and EQ2, referred to as the Paganica fault and the Ocre Segment (OS), respectively ( Fig. 3), we adopted a semi-automatic procedure using the Midland Valley MOVE software. Primary data were the Late Quaternary fault traces and fault/slip data , the L'Aquila 2009 relocated events (Chiaraluce et al., 2011) and the EQ2 focal mechanism calculated herein. Triangulated fault surfaces were created by interpolating plan-view and section-view fault traces (Fig. S4). The latter were drawn across evident hypocentral alignments and clusters and, whenever possible, were matched to the corresponding outcropping fault.
The built OS fault model allows for a moderate east-dipping surface (average dip~45°), which develops at depths between 11 and 16 km beneath the intersection with the SW-dipping Paganica fault. Along strike, the OS extends for~15 km, rotating from WNW (in the north) to NNW (in the south). The OS also shows along-dip variability, with local changes in dip of the triangulated meshes from 35°t o 60° (Fig. S4c,d). According to its moment magnitude (M W 5.4), EQ2 would have ruptured an area of 6 9 7 km 2 (Wells and Coppersmith, 1994) lying on the southern NNW-SSE-striking portion of the OS (Fig. S4c).

Structural regional context of the OS
We identified the area where the mid-crustal OS might have its surface expression in a Late Quaternary east-dipping extensional system cropping out in the Sabini-Eastern Simbruini (SES) sector ( Fig. 4), along the Latium-Abruzzo regional boundary. In the previous literature, the SES almost exclusively is cut by SW-dipping normal faults (Cosentino et al., 2010;. In this paper, based on new field observations and fault/slip data synthetized in an updated geological sketch map (Fig. S5), we highlight the systematic presence of high-to moderate-angle (60°-40°) NE-dipping normal faults. These faults, which locally bound small asymmetric Late Quaternary continental basins, accommodate a horizontal displacement of~1500 m. Moreover, the interpretation of a short commercial seismic line perpendicular to the SES (1-84-CC-3, Videpi Project, 2016) helps, in spite of the poor quality of the seismic image ( Fig. S6), to highlight an east-dipping basal detachment. This detachment is traceable from the surface to~2.5 s TWT (~6.7 km) and delimits the high-angle east-and west-dipping extensional faults at depth (Fig. S6). A 3D view of the reconstructed SES east-dipping geometry is given in the block diagram of Fig. 4b.
With the aim of investigating the possibility of a geometric-kinematic compatibility between the SES and the OS, we constructed a regional geological section (A-A 0 in Figs. 4a . They accommodate a cumulative net extension of~6000 m with an average strain rate of~2.5 mm a À1 . New fault slip data collected in the SES and in the IAFS were projected along the section trace and, for completeness, integrated with some of our previous data (Ferrarini et al., 2015).
The SES and IAFS geological stress tensors were separately calculated using the inversion procedure proposed in Delvaux and Sperner (2003) (Fig. 5b). A normal-fault regime with a sub-horizontal and~SW-NE-trending r 3 axis was obtained for both (Table 2). A co-axial tensional seismological stress tensor was also computed for the L'Aquila 2009 sequence, integrating the EQ2 focal solution computed in this study with a focal mechanism dataset (M W ≥ 3.5) available in the literature (Herrmann et al., 2011).
The inversion results show the coaxiality among the stress tensors computed for the outcropping eastdipping SES and west-dipping IAFS and for the fault system activated at depth during the L'Aquila seismic sequence ( Table 2). The inversion procedure highlights a very small misfit angle (8°) between the input EQ2 slip vector and the resolved shear stress on the EQ2 east-dipping preferential seismic plane. In addition, the analysis of slip vector vs. the stress ratio Φ (Angelier, 1994) shows that the EQ2 resolved shear stress falls in the range of predicted values (Fig. S7). This implies that the relevant strike-slip component characterizing EQ2 (rake~35°) is kinematically compatible with the L'Aquila 2009 tensional tensor, possibly due to the activation of a preexisting, nearly N-S-striking plane.

OS potential fault activity
To investigate the likelihood of the OS being reactivated within the reconstructed stress field, we calculated its slip tendency (sensu Morris et al., 1996), expressed as T = Ts/ Ts max , where Ts is the shear to normal stress ratio and Ts max is the maximum calculated T S . Starting from the reconstructed OS fault model (Figs. 3 and S4) and the seismologic stress tensor attitude (Table 2), we calculated the slip tendency on each OS mesh (Fig. 6). We assumed r 1 equal to the lithostatic pressure (q r gz), where q r was related to the different stratigraphic horizons (cross-section in Fig. 7a). We derived r 3 from the differential stress (r 1 -r 3 ) (Sibson, 1974) at the hypocentral depth of the OS (14 km) and r 2 from the calculated stress ratio (r 2 = 0.8r 1 ). We assumed a frictional strength consistent with normal tectonics (Collettini and Sibson, 2001), a hydrostatic fluid regime and cohesionless fault surfaces. Input data and results are given in Table 3 and Fig. 4b. Under such conditions, the OS meshes with dips >~30°show a good mechanical tendency to be reactivated under the present stress field (T ≥ 0.6). Conversely, reactivation is not expected (T ≤ 0.5) on the lowangle OS meshes (<~25°). This stress state is represented well in the Mohr circle in Fig. 6b, where the straight line represents the limiting friction line that separates the favourably (red to pinkish) and unfavourably (green to bluish) oriented domains. If suprahydrostatic conditions were assumed, as proposed by Di Luccio et al.   Delvaux and Sperner (2003). Key: dark and pink arrows = measured slip directions and resolved shears, respectively (the corresponding misfit angles vs. the number of data points are represented in the histograms); nt = total number of data (e.g. plane/slikenline pairs); n = number of successfully inverted data; r 1 , r 2 , r 3 = principal stress axes; Φ = stress ratio = (r 2 Àr 3 )/(r 1 Àr 3 ); the quality ranking factors (QR) and the stress inversion parameters with associated uncertainties are listed in Table 2. [Colour figure can be viewed at wileyonlinelibrary.com] Table 2 Geological and seismological stress tensor parameters calculated in Fig. 5b for the SES (Sabini-Eastern Simbruini) and IAP (Intra-Apennines) fault systems and for the L'Aquila 2009 sequence. nt = total number of data (e.g., plane/slikenline pairs); n = inverted data; r 1 , r 2 , r 3 = principal stress axes; Φ = stress ratio = (r 2 Àr 3 )/(r 1 Àr 3 ); QR = quality ranking: A-QRw as in Sperner et al. (2003) and A-QRfm as in Heidbach et al. (2010).

Proposed model and conclusions
The results provided in this paper support the following observations: 1 seismological evidence of a moderate-to-steep (35°-60°) east-dipping normal-oblique fault segment, the OS, located at mid-crustal depth beneath the Paganica fault and partially activated by EQ2 (Figs 2, 3 and S4); 2 geological evidence of previously under-evaluated moderately eastdipping (40°-60°) extensional structures cropping out within the SES and propagating down-dip to depths of 9-10 km (Figs. 4, S5 and S6); 3 geological and seismological kinematic compatibility of the SES with the intra-Apennine west-dipping faults (IAFS) and with the L'Aquila focal mechanisms under a common tensional stress field with a SW-NE-trending leaststress axis (Fig. 5); 4 no evidence of any transcurrent stress regime at the EQ2 hypocentral depth, but rather evidence of a tensional regime with a SW-NE-trending least-stress axis that can activate favourably oriented pre-existing planes, although with a considerable oblique component (Fig. S7); 5 the likelihood of the entire OS being potentially seismogenic within the reconstructed stress field (Fig. 6).
Considering the above points, we wonder whether, similar to the EFS in the northern Apennines (Fig. 1c), the SES and the OS might be connected at depth, being expressions at different structural levels of the same discontinuity, which dips eastward with a ramp-flat-ramp geometry. The IAFS, which includes the Paganica fault responsible for EQ1, would be antithetic to such a structure, here referred to as the Latium-Abruzzo Extensional Detachment (e.g., LAED). This working hypothesis is presented in Fig. 7a, along the trace of the geological section of Fig. 5a.
The proposed LAED geometry is subdivided into four sectors with different degrees of interpretation and data constraints. Sectors I and IV extend across the western (shallow) and eastern (deep) LAED ramps, respectively, and are constrained by geological and seismological data discussed in this paper (Figs 3-5, S5 and S6).
Sectors II and III extend across the central LAED flat that connects the two lateral ramps, possibly developing at the interface between the Middle Triassic quartzites and phyllites and the underlying Late Palaeozoic-Early Triassic low-grade metamorphic basement (Fig. 7a). The western half of the LAED flat (sector II) is model-driven due to the lack of instrumental activity. The eastern half of the LAED flat (sector III) is inferred from the depth distribution of the background seismicity from 2004 to 2012, which shows a sharp cut-off at~10 km (Bagh et al., 2007;Chiarabba et al., 2015). A subhorizontal flat behind the Paganica fault is also hypothesized by Atzori et al. (2013), based on the interpretation of an anomalous far-field interferometric signal that preceded the 2009 L'Aquila sequence. On the same flat, Borghi et al. (2016) located a precursory M w 5.9 slow-slip seismic event that occurred on 12 February 2009. Furthermore, the presence of an east-dipping low-angle discontinuity at the base of the SW-dipping intra-Apennine seismogenic master fault is highlighted by preliminary aftershock locations of the 2016 Central Italy seismic sequence (Michele et al., 2016) (Fig. 7d). According to our interpretation, the LAED would represent a regional right-lateral en echelon arm of the Etrurian Fault System (EFS) (Fig. 7b,c). Like the EFS, it would delimit at the base of the intra-Apennine crustal volume undergoing active extension (Fig. 7a). Unlike the EFS breakaway zone, where Late  Fig. 5a with the hypocentral distribution of the 2009 L'Aquila events relocated by Chiaraluce et al. (2011), assuming a half-width of 5 km. Key for the stratigraphic sequence (after Patacca et al. (2008) and Cosentino et al. (2010)): 1 = Late Miocene siliciclastic foredeep deposits (average thickness 2000 m); 2 = Jurassic-Cretaceous to Early Miocene carbonates (up to 3500-4000 m); 3 = Late Triassic dolostone and evaporites (up to 3000 m); 4 = Middle Triassic quartzites and phyllites (~1500 m); 5 = Late Palaeozoic-Early Triassic low-grade metamorphic basement (average thickness 6000 m); 6 = Middle Palaeozoic crystalline basement; the q density values are from Boncio et al. (2004). Key for fault structures: a = Late Miocene-Early Pliocene thrust faults; b and c = east-and west-dipping normal fault systems with associated antithetic faults; LAED = Latium-Abruzzo Extensional Detachment, subdivided into four sectors with different quality rankings in the depth interpretation. (b) Sketch map of the east-dipping Latium-Abruzzo Extensional System and of the antithetic intra-Apennine west-dipping faults as reconstructed in this paper. The red stars represent the epicentres of the two major events (M w 6.3 and 5.4) of the L'Aquila sequence 2009. The yellow star locates the major event of the Central Italy 2016 seismic sequence (30 September, M w 6.5). (c) Map highlighting the regional right-lateral en echelon segmentation pattern of the major east-dipping extensional fault systems (EFS) that crop out along the inner border of the seismogenic intra-Apennine extensional domain (modified from Lavecchia et al., 2011). They are here identified as the Tuscan EFS, Umbria EFS and Latium-Abruzzo EFS; the SES, first highlighted in this paper, belongs to the newly defined Latium-Abruzzo Extensional Detachment (LAED). (d) Sketch of the fault pattern along a section that extends across the 2016 Central Italy seismic sequence (trace B-B 0 in Fig. 7b) (after Lavecchia et al., 2016). The aftershock sequence recorded from 25 August to 15 September 2016 (yellow dots) is from section 1 in Bonini et al. (2016). Note that the aftershock sequence reveals both a high-angle west-dipping master fault, where the main event nucleated, and a low-angle east-deepening basal plane, closely recalling the geometry and structural style proposed in this paper for the LAED. Miocene flysch terranes are juxtaposed against Triassic evaporites , the tectonic elision cropping out along the LAED breakaway zone, e.g., the SES, is limited to the contact between Early Cretaceous and Early Jurassic formations (Figs. S5 and S6). Nevertheless, we advance the hypothesis that, at depth, the LAED might accrue significant normal-sense displacement, coupled with slip on the synthetic SES and with the progressive eastward shift and extension of the hangingwall volume cross-cut by the IAFS (Fig. S8).
Although the LAED seismogenic role is questionable, preliminary results obtained in this paper (e.g. EQ2/fault association and slip tendency analysis) suggest the possible occurrence of small/moderate (M 4.5-5.5) extensional earthquakes on mid-crustal LAED segments with dip >~45° (Fig. 6). As in the L'Aquila 2009 case, these mid-crustal events might be triggered by major uppercrustal earthquakes on the antithetic high-angle faults, due to stress transfer (De Natale et al., 2011;Serpelloni et al., 2012) and/or fluid migration (Doglioni et al., 2014).
The hypothesis of a regional lowangle east-dipping normal fault beneath the central Apennines might be relevant to the definition of the active extensional style in the region, with consequent implications for seismic hazard evaluations. In addition, the data and interpretation provided in this paper might contribute to the worldwide discussion on the geometry, kinematics and seismogenic behaviour of continental extensional detachments (Axen, 1999).

Acknowledgements
This work was financed by DiSPUTer, University of Chieti "G. d'Annunzio" (research funding to Giusy Lavecchia) and by the Department of Physics, University of Naples "Federico II" (research funding to Aldo Zollo). The 3D fault modelling and slip tendency analysis were performed using the Move Software Suite 2016, which was granted by Midland Valley's Academic Software Initiative. We thank Lauro Chiaraluce for providing the relocated L'Aquila 2009 dataset and Martin Vall ee for the source code that was used to compute the ASTFs. Many thanks also to the Terra Nova editor, Carlo Doglioni, and to Massimiliano Barchi and an anonymous referee for their criticism and helpful suggestions. The data are listed in the references, tables and supplements. Table 3 Rock properties and stress conditions computed for the Ocre Segment (reference depth 14 km) and adopted for the slip tendency calculation in Fig. 6.

Rock properties Values References
Frictional strength, l 0.6 Collettini and Sibson (2001) Cohesion, c 0 Pore fluid factor, k = P f /qgz 0.4 (hydrostatic fluid pressure state) Sibson (2000) Density, q (kg m À3 10 3 ) as given in Fig. 7a This paper Stress ratio Φ = (r 2 Àr 3 )/(r 1 Àr 3 ) 0 variance reduction (VR) are also reported. We used broadband velocity waveforms that were recorded at RSNC between 100 km and 400 km from the earthquake epicentre to consider records with good signal-tonoise ratios and to avoid the use of saturated waveforms. The Green's functions were computed with the frequency-wave number integration method (Saikia, 1994) in the 1-D regional velocity model proposed by Herrmann et al. (2011). After fixing the depth and the location of the earthquake, we inverted data from 9 stations and obtained a solution with a variance reduction of~89% and a percentage of double couple (DC) of 91%. Figure S2. Uncertainty assessment of the source angles and Kagan angle of the calculated TDMT solution (Table 1 of the main document). The uncertainty analysis was performed by calculating the theoretical error ellipsoid for a fixed source position and time as proposed by Sokos and Zahradn ık (2013) with ISOLA software (Sokos and Zahradnik, 2008). The source-station configuration, frequency range and crustal model define the ellipsoid shape and orientation, whereas the absolute size of the error ellipsoid is determined by the variance of the data. In each panel, the nodal planes are indicated with red dashed lines and are derived from the moment tensor solution calculated in this study. The grey histograms represent the potential and acceptable solutions, as function of the data variance. The angle values associated with the constrained seismic preferential plane are reported in underlined bold italic. The uncertainty values are small, indicating a well constrained focal mechanism solution. Referring to the preferential seismic plane, (a) the strike value varies between 347°and 344°; (b) the dip value centred on 58°varies over a range of 4°, whereas (c) the rake (-34°) varies mainly between -32°a nd -34°. Moreover, (d) the Kagan angle, that is, the angle between acceptable solutions compared with the optimal solution, shows values whose mean is about 1°with a standard deviation of 0.5. This additional parameter underlines the low uncertainty of the focal mechanism solution. Figure S3. (a) EQ2 slip map obtained by isochrone back-projection computed with a rupture velocity of 1.8 km/s. The ASTFs shapes discriminate the ENE-dipping plane of the computed TDMT focal mechanism (Fig. 2a in the main document) as the one for which the misfit function is minimized (Fig. 2c in the main document). After testing different rupture velocities, we selected the final kinematic model that minimizes the L1-norm between the synthetic and observed ASTFs (Fig. 2b in the main document). The fault geometry was fixed at the solution that is derived from the computed focal mechanism (ENE-dipping plane). We adopted the deconvolution technique that was proposed by Vall ee (2004). The deconvolution was performed at 15 stations (Fig. 2a in the main document) for the direct S wave in the frequency range 0.01-2 Hz based on the limits that were imposed by the expected corner frequency of the master event and the stability of the S-wave polarization. Figure S4. (a,b) 3D-model of the Paganica fault segment activated by EQ1 and its aftershock sequences in the time interval from 6 April to 30 September 2009 (dataset from Chiaraluce et al., 2011). To reconstruct the 3D surfaces, we adopted a semi-automatic procedure using the Midland Valley MOVE software (vers.2016). First, we traced closely spaced (1.25 km semi-width) and differently oriented sets of sections across the epicentral area, in the MOVE georeferenced frame. Second, we drew the fault trace connecting the aftershock volume and the surface fault in the MOVE section-view window. Third, we used the Delaunay Triangulation function to build the fault surfaces. The extent of the seismogenic patch activated by the 6 April main shock and by its aftershock sequence with respect to the overall Paganica fault surface is highlighted in Fig. S4b. (c,d) 3D-model of the Ocre Segment (OS) activated by EQ2 and by the other deep events (depth >11 km) of the L'Aquila 2009 sequence. (c) Triangulated meshes of the OS surface reconstructed according to the same work-flow adopted for the Paganica fault in Fig. S4b. The white dashed line is a schematic representation of the EQ2 (M w 5.4) rupture area calculated according to scale law (about 6 9 7 km from Wells and Coppersmith, 1994). (d) Histograms of the strike and dip of the OS triangulated surfaces, showing the OS along-strike and along-dip variability. Figure S5. Geological map of the Sabini-Eastern Simbruini (SES) and Carseolani Mts. area, simplified from the Carta Geologica d'Italia, 1:100,000 scale, sheets http://193.206.192.231/carta_geologica_ italia/default.htm) and redrawn on a GIS-platform with original field data from the authors. The map is overlain on a 20 m digital elevation model. The structural sites surveyed along the SES normal faults are indicated by yellow dots and numbered from 1 to 19; the corresponding fault/slip data are represented in a number of stereographic plots (Schmidt net, lower hemisphere) as both (a) cyclographic trace/lineation and (b) poles to major normal faults and contouring of the associated slip vectors. Figure S6. Two portions of the commercial seismic line 1-84-CC3 (VIDEPI Project, 2016; http://unmig. mise.gov.it/deposito/videpi/allegati/ 1214.pdf) extending in an average SW-NE direction across the Sabini-Eastern Simbruini (SES) extensional system (traces in Fig. S5), with corresponding line drawing and geological interpretation of the main reflectors and seismic facies. In spite of the generally poor quality of the line, some closely spaced and relatively continuous packages of reflections, separated by areas with lowcontinuity signals, can be detected and traced along the entire section. Taking into account the outcropping stratigraphic units and fault planes (Fig. S5), the average stratigraphic thickness given in the literature (Patacca et al., 2008;Cosentino et al., 2010) and the corresponding seismic and mechanical stratigraphy (Di Luzio et al., 2009 and references therein), these reflectors can be referred to the tops and the stratigraphic units reported in the legend. Figure S7. Slip vector variability range for the preferential fault plane of the EQ2 focal mechanism (Table 1), in the L'Aquila 2009 computed extensional stress regime (Table 2), as function of the stress ratio Φ (after Angelier, 1994). Key: