In this paper, we propose a new model of the crustal structure and seismotectonics for central Sicily (southern Italy) through the analysis of the depth distribution and kinematics of the instrumental seismicity, occurring during the period from 1983 to 2010, and its comparison with individual geological structures that may be active in the area. The analysed data set consists of 392 earthquakes with local magnitudes ranging from 1.0 to 4.7. We defined a new, detailed 1-D velocity model to relocate the earthquakes that occurred in central Sicily, and we calculated a Moho depth of 37 km and a mean VP/VS ratio of 1.73. The relocated seismic events are clustered mainly in the area north of Caltanissetta (e.g. Mainland Sicily) and in the northeastern sector (Madonie Mountains) of the study area; only minor and greatly dispersed seismicity is located in the western sector, near Belice, and along the southern coast, between Gela and Sciacca. The relocated hypocentral distribution depicts a bimodal pattern: 50 per cent of the events occur within the upper crust at depths less than ∼16 km, 40 per cent of the events occur within the middle and depth crust, at depths between 16 and 32 km, and the remaining 10 per cent occur at subcrustal depths. The energy release pattern shows a similar depth distribution.
On the basis of the kinematic analysis of 38 newly computed focal plane solutions, two major geographically distinct seismotectonic domains are distinguished: the Madonie Mountain domain, with prevalent extensional and extensional-oblique kinematics associated with upper crust Late Pliocene–Quaternary faulting, and the Mainland Sicily domain, with prevalent compressional and compressional-oblique kinematics associated with thrust faulting, at mid to deep crust depth, along the north-dipping Sicilian Basal Thrust (SBT). The stress inversion of the Mainland Sicily focal solutions integrated with neighbouring mechanisms available in the literature highlights a regional homogeneous compressional tensor, with a subhorizontal NNW–SSE-striking σ1 axis. In addition, on the basis of geodetic data, the Mainland Sicily domain may be attributed to the SSE-ward thrusting of the Mainland Sicily block along the SBT plane. Seismogenic shearing along the SBT at mid-crustal depths was responsible for the unexpected Belice 1968 earthquake (Mw 6.1), with evident implications in terms of hazard assessment.
The Sicilian region represents a portion of the Apennine–Maghrebide fold-and-thrust belt (Fig. 1a) developed in an area dominated by both the convergence between the European and Nubia (northern Africa) plates and the extensional-compressional processes linked to the opening of the Tyrrhenian basin. The compressional belt comprises southward-verging crustal-scale tectonic units and related back-thrusts that have been progressively overthrust since Oligocene times (Ghisetti & Vezzani 1984; Butler et al. 1992; Lentini et al. 1994; Catalano et al. 1996; Guarnieri et al. 2002; Tavarnelli et al. 2003). The outermost and youngest group of tectonic units consists of late Pliocene–Quaternary south-verging fold-and-thrust structures developed at the hanging wall of a major north-dipping thrust discontinuity (Sicilian Basal Thrust, SBT; in Lavecchia et al. 2007a), whose surface tip line extends with a southward arcuate shape from Sciacca to Gela and Catania (Fig. 1a). Active and possibly seismogenic compression of the SBT and its hanging wall splays in Quaternary times is debated. Some investigators assume cessation of active convergence since early middle Pleistocene times (Butler et al. 1992; Torelli et al. 1998; D’Agostino & Selvaggi 2004; Ghisetti et al. 2009); others consider the SBT still active (Catalano et al. 2004; Jenny et al. 2006; Lavecchia et al. 2007a).
Historical and instrumental seismic data support the seismogenic compression hypothesis as it relates to the western (Monaco et al. 1996; Anderson & Jackson 1987; Pondrelli et al. 2006) and eastern regions of Sicily (Cocina et al. 1997; Neri et al. 2005; Lavecchia et al. 2007a and references therein), whereas no seismotectonic information is available on the active state of strain in central Sicily, specifically in the area of Caltanissetta (Fig. 1a). Such portion of Sicily, hereinafter named Mainland Sicily (Fig. 1a), appears as an aseismic domain in the most recent official seismic zoning of the Italian territory (model ZS9 in Meletti et al. 2008) and in the corresponding individual seismic source model (Basili et al. 2008). This appearance as an aseismic domain is due mainly to three different observations: (i) since early instrumental times, Mainland Sicily has not been struck by a significant earthquake; (ii) the background seismicity recorded over the past 20 yr was low and infrequent, and (iii) the seismic station coverage was less favourable with respect to other Italian areas.
The aim of this work is to clarify the active state of strain and evaluate the seismogenic sources in central Sicily through the investigation of the geometry and kinematics of the potentially active sources. The first part of the paper aims to improve and extend the knowledge on hypocentre locations occurring in the study area. The collection of seismic recordings related to 392 local earthquakes (1.0 ≤ML≤ 4.7) occurred in the period from 1983 to 2010, and the creation of a joint data set composed of both arrival times from national bulletins (time interval 1983–2005) and repicked seismic data (time interval 2005–2010) was performed. To create a uniform starting seismic catalogue, the joint data set was preliminarily relocated using an 1-D velocity model optimized for the Italian territory (Chiarabba et al. 2005). Subsequently, a new 1-D velocity model was computed for central Sicily and was used to perform the final relocation of the whole data set. The second part of the study is dedicated to the calculation of first motion polarity focal mechanisms and to the tectonic interpretation. The new focal mechanisms are integrated with fault solutions from the literature to better constrain the seismotectonic regime of the region. The obtained results provide further insight into the regional state of stress of the SBT and surrounding regions.
On the basis of evidence from fault plane solution mechanisms of moderate-to-large crustal earthquakes (Mw≥ 4.0, depth ≤ 40 km; Neri et al. 2005; Lavecchia et al. 2007a; Billi et al. 2010; Visini et al. 2010), five major geographically and kinematically seismogenic domains may be distinguished in Sicily. They are shown schematically in Fig. 1(b). From north to south, they are: (A) the Southern Tyrrhenian E–W striking domain, located off-shore of northern Sicily, undergoing N–S compression; (B) the northern Sicily domain including the Madonie-Nebrodi (B1) and the Peloritani (B2) areas, undergoing N–S and WNW–ESE tension, respectively; (C) the Mainland Sicily domain undergoing primarily nearly N–S compression in the Mazara-Belice (C1) and Etna areas (C3) and strike-slip deformation in the Caltanissetta area (C2); (D) the southern Sicily domain, extending from Sciacca to Gela and Catania, not associated with any significant instrumental earthquake and (E) the Hyblean Foreland domain in southeastern Sicily with primarily strike-slip deformation.
Although some seismological studies have been performed in the Madonie–Nebrodi–Peloritani and Hybleans domains (e.g. Musumeci et al. 2003; Langer et al. 2007), no direct information concerning the velocity structures and kinematics in the Mainland Sicily domain is available. Neri et al. (2005) and Jenny et al. (2006) identify a unique large province, inclusive of the Mainland Sicily and Southern Tyrrhenian domains, undergoing NNW–SSE compression. Lavecchia et al. (2007a) interpreted Mainland Sicily as an independent seismogenic province undergoing N–S compression at mid-crustal depths, at the hangingwall of the N-dipping SBT. The shallower upper crustal portion of the SBT province has an arcuate shape at surface and coincides with the Sciacca–Gela–Catania thrust front (Fig. 1a). Wells and seismic reflection profiles reveal the uppermost few kilometres of the SBT downdip geometry (Lickorish et al. 1999), while at depth, the geometry may be only inferred. A thick-skinned style is proposed by some authors (Guarnieri et al. 2002; Lavecchia et al. 2007a), and they assume a northward deepening of the SBT to depths of nearly 35–40 km beneath central and northern Sicily.
Although Mainland Sicily is not considered an active seismogenic zone, three large historical earthquakes (Mw≥ 6.0) may be associated with thrust activity. From west to east, they are Belice 1968 (Mw 6.1), central Sicily 361 (Mw 6.6) and Catania 1818 (Mw 6.0; Fig. 1a). The 1968 earthquake occurred in the Belice area (west of Mainland Sicily), where no other significant earthquake is reported in the historical catalogue and where very scarce background seismicity is recorded, apart from the 1981 Mazara del Vallo earthquake (Mw 4.9). This event, nucleated nearly 40 km west from the Belice earthquake, was associated with N–S compression along a north-dipping thrust plane at depths of 15–20 km. On the other hand, the Belice sequence was elongated in a roughly E–W direction (Gasperini et al. 1999; Valensise & Pantosti 2001), and the hypocentres were distributed along a N-dipping thrust plane from near the surface to a depth of ∼30 km (Monaco et al. 1996). The focal solution of the main event and of two major aftershocks were evaluated as both thrusting on a north-dipping plane (Anderson & Jackson 1987) and right lateral oblique compression on a NNE-striking plane (McKenzie 1972; Gasparini et al. 1985). In both cases, the average P-axis was evaluated being nearly N–S and subhorizontal. The central area of the Mainland Sicily domain was highly damaged by the 361 earthquake (Mw 6.6), whose macroseismic epicentre was located at Caltanissetta, according to archaeological evidence of damages reported at the great historic Roman Villa of Casale in the town of Piazza Armerina (III–IV A.D.; Working Group CPTI 2004). Although some investigators attributed the event to N–S compression (Jenny et al. 2006; Visini et al. 2010), others proposed a northward relocation in the Madonie Mountains extensional area (Barreca et al. 2010). The eastern area of the Mainland Sicily domain was struck by the Catania 1818 (Mw 6.0) earthquake. Although located at the southern flank of Mt Etna, such events are commonly interpreted as tectonic and independent from the volcanic activity (Azzaro et al. 2000) and possibly associated with N–S/NNW–SSE compression (Lavecchia et al. 2007a).
The area located at the hangingwall of the shallowest portion of the SBT, close to the surface trace of the Sciacca–Gela–Catania thrust front, is characterized by a moderate level of seismic activity with a few events with Mw < 5.0 on the western side of the arc (1578, 1652, 1727, 1740, 1817, 1933) and by several events with Mw∼ 5.5 in the eastern side, from Gela (Fig. 2) to Mineo (1624, 1878, 1896, 1903, 1909) and Catania (1959; Working Group CPTI 2004). Not far from Sciacca, two destructive events may have occurred between the 4th and 3rd century B.C. and between the 6th and 13th century A.D. These events are indicated by archeo-seismological evidence at the Selinunte temples (Guidoboni et al. 2002).
The historical seismic record in Sicily may be considered complete only from the 17th century (1680 ± 100) for events with M≥ 5.5 (Lavecchia et al. 2007b and references therein) and the 19th century (1825 ± 25) for events with M≥ 4.5. Therefore, earthquakes with longer recurrence times cannot be excluded in Mainland Sicily, and the identification of their potential seismogenic sources is important to seismic hazard assessment.
Seismological data and processing
We analyse the crustal instrumental seismicity that occurred in central Sicily between 1983 and 2010, recorded by the National Seismic Network (RSN) and managed by the Istituto Nazionale di Geofisica e Vulcanologia (INGV). The boundary of the study area is enclosed in the rectangular area outlined in Fig. 2. The RSN improved significantly over the last 5 yr throughout Italian territory and, in particular, in Sicily. An increasing number of three-component extended band (Lennartz 5 s) and/or broad-band (Trillium 40 s) sensors, replacing the Kinemetrics S-13 short period sensors, were deployed resulting in better station coverage. In 1983, the RSN had 17 stations deployed in Sicily, in the Aeolian Islands and southern Calabria; however, today, the RSN has approximately 70 stations (triangles in Fig. 2).
Following previous investigators (Amato & Mele 2008; Schorlemmer et al. 2010) that noted the improved earthquake detection capabilities of the national network and the lowering of the earthquake catalogue completeness threshold, we observe that the denser coverage of broad-band seismic stations in Sicily since 2005 has significantly lowered the magnitude of detectable earthquakes. This observation is well evident in the diagram within the bottom-left inset of Fig. 2, which shows the cumulative number of earthquakes recorded within the study area per year and per two ranges of magnitude (ML≤ 3.0 and ML > 3.0, respectively). In this figure, the two small steps indicate the occurrence of two seismic sequences in central Sicily (between 1995 April and June and between 2001 November and December), and the third one indicates the improvement of earthquake detection capabilities due to the development of seismic network and which can be seen from 2005 onward.
We collected the available arrival times of earthquakes during the period from 1983 December to 2005 April, as reported by the CSI catalogue (Castello et al. 2005) and by the Italian Seismic Bulletin (ISB 2003–2010; http://bollettinosismico.rm.ingv.it). However, we analysed the waveforms and repicked the arrival times of events recorded by the RSN from 2005 April to 2010 April. Similarly to the weighting procedure used for bulletin′s traveltimes and on the basis of the picking accuracy, we assigned to each P or S arrival a weight, varying from 0 (accuracy of 0.05 s) to 4 (accuracy of 2.0 s or more). All P- and S-readings in single earthquakes having weights of four were excluded from the relocation procedure and from further analysis. We collected a total data set comprising 3566 P- and 2159 S-phases, associated with 392 earthquakes, with local magnitudes ranging from 1.0 to 4.7.
To standardize the compiled data set, in terms of location procedure and quality of results, we performed a preliminary relocation using the velocity model proposed by Chiarabba et al. (2005) for the Italian territory and using the Hypoellipse code (Lahr 1989). The obtained epicentral distribution of events is presented in the map of Fig. 2, while their depth distribution is shown in the bottom-right side inset. In the histogram, the hypocentral depths range 0–75 km, with the majority of events located within 0–15 km and 20–35 km depth intervals.
With the purpose of determining the VP/VS ratio and calculating a 1-D minimum velocity model, we selected from the total data set the events that satisfied the following requirements: (i) a minimum number of eight readings (at least five easily readable P and three clear S arrivals), (ii) maximum travel time residuals (rms) of 0.8 s and (iii) maximum GAP of 180°. In so doing, we obtained a selected data set, which consists of 215 best-constrained events (red epicentres in Fig. 2).
Considering the time readings of both P and S phases from the selected database, a mean velocity ratio VP/VS was computed using the cumulative Wadati diagram. A VP/VS ratio of 1.728 with a 95 per cent confidence interval bounded by the values 1.718 and 1.739, having a squared correlation coefficient (R2) greater than 0.98, was estimated (Fig. 3). This value of the VP/VS ratio is much lower than the ones found in the northern, central and southern Apennines (VP/VS= 1.80, VP/VS= 1.85, VP/VS= 1.83 and VP/VS= 1.82) from Piccinini et al. (2009), Monna et al. (2003), Bagh et al. (2007) and Maggi et al. (2010), while it is slightly lower than the values calculated by Langer et al. (2007) for the Calabro–Peloritan zone (VP/VS= 1.75) and the value obtained for the Hyblean Foreland by Musumeci et al. (2003) (average value VP/VS= 1.76).
Minimum 1-D velocity model
Events location in central Sicily suffers from the lack of a specific velocity model for the study area. As is well known, the use of inadequate velocity parameters during the location process can introduce systematic errors in the hypocentre location (Thurber 1992; Eberhart-Phillips & Michael 1993), which may results in incorrect seismotectonic interpretations. Therefore, the identification of a more suitable velocity model for central Sicily is necessary. As both the density of the network and the number of earthquakes do not permit 3-D tomography with an adequate resolution, we pointed to a 1-D inversion of the velocity structure.
Starting velocity models
A common problem in the inversion procedure is the dependence of the results on the initial guess. Following Kissling et al. (1994), we carefully chose the starting models used for the inversion of the velocity structure. Therefore, we first collected all the a priori available information regarding the impedance structure of Sicily (velocities and layer thickness). Seven 1-D P-wave velocity models were taken from the literature and sketched in Fig. 4 (from a to g). Model (a) is a regional model, computed from the analysis of seismological data recorded in central and southern Italy (Chiarabba & Frepoli 1997); it consists of four layers and assumes a Moho depth of 35 km depth for the entire Sicily region. Models (b) and (c) were computed across eastern and western Sicily by a joint interpretation of wide-angle seismic refraction profiles and Bouguer anomalies (Chironi et al. 2000). The eastern model (b) consists of three layers, with a Moho depth of 30 km and a sub-Moho P-wave velocity of 7.0 km s–1; the western model (c) consists of four layers, with a Moho depth of 35 km and a sub-Moho P-wave velocity of 7.2 km s–1. Model (d) was obtained from the integrated structural–kinematic and seismological analysis of the crustal structure in central-southern Sicily (Lavecchia et al. 2007a). Models (e), (f) and (g) were obtained from crustal tomography investigations in the Calabrian Arc-Sicilian region (Barberi et al. 2004; Billi et al. 2010). Among these three models, two consist of nine layers, and the third consists of eight layers, with a Moho depth at 40, 43 and 39 km and a sub-Moho P-wave velocity of 7.25, 7.50 and 7.25 km s–1, respectively.
Minimum 1-D velocity model for central Sicily
For the identification of optimum minimum 1-D P-velocity model for central Sicily, we have used the widely known software VELEST (Kissling 1995). In this approach, the hypocentre locations, the velocity structure and the station corrections are derived using a simultaneous inversion of P and S waves.
The inversion was performed using the selected database of 215 events (Fig. 2) and the seven starting velocity models previously discussed (Fig. 4). S-wave readings were included only to better constrain the earthquake location.
We obtained the final model performing attempts by trial and error with different combinations of damping factors and by adjusting the layer thickness of the initial models (a–g in Fig. 4) to better estimate the depth of the main discontinuities. Afterwards, we further tested average models that were derived from ones we selected with different layering. In particular, we created some models with thin layers near the Moho discontinuity to constrain its depth and balanced opportunely the damping factors to avoid both overdamping and an unrealistic velocity model solution.
The steps that lead to the final model consisted of several inversions. For each inversion, we analysed the rms trends versus the number of iterations and chose as the best model the one corresponding to the minimum misfit of traveltime residuals. Even when we used a wide range of variability and combinations of values within each different layer (Fig. 4), the output models were quite similar, implying a stable solution. The final velocity model obtained from the inversion is sketched in Fig. 5(a) and is enclosed by the grey area, which represents the range of variability of the computed minimum 1-D velocity models with a misfit less than or equal to 0.14 s, value compatible with the noise level of traveltime data. Among these computed velocity models, the best one was chosen on the basis of goodness of the fit (0.12 s), the reduction of rms with respect to the starting and final traveltime residuals (85 per cent from 0.79 s to 0.12 s) and the a priori geological–geophysical information.
The new velocity model for central Sicily has five layers above the Moho, which is located at a depth of 37 km (Fig. 5a). The velocity pattern is fairly well resolved in the upper, middle and lower crust (approximately 96 per cent of the ray paths across these layers; Table 1), although the uppermost and deepest layers are also sufficiently constrained. The new 1-D velocity model shows relatively low VP values in the upper crust (∼5.0 km s–1 down to 12 km) with respect to the values obtained for other parts of Italy, for example, the Apennines chain (Piccinini et al. 2009; Monna et al. 2003; Chiarabba et al. 2009; Bagh et al. 2007; Maggi et al. 2010), and these values are in agreement with the average velocities retrieved for the entire Sicilian region (Chiarabba & Frepoli 1997), for the Hyblean Foreland (Musumeci et al. 2003) and the Nebrodi–Peloritani area (Langer et al. 2007), where average values of 5.1, 5.2 and 4.9 were found, respectively. The Moho depth at 37 km is in agreement with the ranges of values (35–40 km) proposed by different authors (Finetti 2005; Tesauro et al. 2008; Grad et al. 2009); conversely, the velocity beneath the base of the crust (7.79 km s–1 at depths >37 km) is slightly greater than the average value (7.56 km s–1) computed by Chiarabba & Frepoli (1997) and close to the value (8.0 km s–1) derived from a DSS profile across Sicily (Cassinis et al. 2005).
The station corrections are positive in central Sicily, consistently associated with the thick deposits of the soft sediments of the Caltanissetta basin, while negative values are evident in northern Sicily and in the Hyblean Foreland (Fig. 5b). The positive values of the ECNV and MNO station corrections, placed near the Mt Etna volcano, correlate well with the presence of the thermal effects associated with the volcanic activity. A geological interpretation of the inferred P-wave velocity model of central Sicily is shown in Fig. 5(c). The low P-wave velocities in the upper crust might be associated with the presence of a thick Tertiary sedimentary cover in the uppermost 5 km and with Mesozoic carbonates, evaporites and sedimentary Permo-Trias terrains at depths between 5 and 12 km; a metamorphic Permo-Trias basement (from 12 to 18 km) above a crystalline Palaeozoic basement (from 18 to 22 km) might characterize the middle crust, whereas a classic granodiorite composition may be assumed for the lower crust at depths between 22 and 37 km. The P-wave velocities of the upper mantle (over 37 km depth) would be associated with a peridotitic composition.
Earthquake final relocations
We performed the final relocations of the entire data set of 392 earthquakes that occurred in central Sicily from 1983 to 2010, using the Hypoellipse code (Lahr 1989). Considering the presence in Sicily of five well distinguished crustal and seismotectonic domains (Fig. 1b), we associated with each of them the appropriate velocity model. These models were derived from the literature (Fig. 6, and references therein), apart from the new one that is computed here for Mainland Sicily, and were applied to all the seismic stations lying within the boundary of the study area. The final epicentral relocations distribution is shown in the map of Fig. 6. We observe two major concentrations of earthquakes in the northeastern sector of the study area (Madonie Mountains) and in the area north and northwest of Caltanissetta; a minor and greatly dispersed seismicity is evident in the western sector, close to the Belice valley, and along the southern coast, between Gela and Sciacca. A depth view of the seismicity is presented in Fig. 7, where the earthquakes are projected along the traces of two sections striking NNE–SSW (S1) and NNW–SSE (S2), across the study area (Fig. 6), assuming a semi-width of 20 km. The relocated hypocentres (Fig. 7C) are compared with the bulletin locations (Fig. 7A) and with the ones preliminarily relocated (Fig. 7B). An improvement of the final locations with respect to the original and preliminary ones is evident. In both sections, the original and preliminary locations show a clustering of the hypocentres at depths of approximately 13 and 34 km, respectively, probably as a result of strong discontinuities in the adopted velocity and/or fixed depths in the location procedures. Conversely, the final depth distribution does not show any systematic shift of the hypocentres, and the seismicity is distributed in a N-dipping wedge shape volume.
We estimated the quality of the final locations accounting for the azimuthal gap, the root mean square of the traveltime residuals (rms) and the horizontal (ErrH) and vertical (ErrZ) errors determination (Table 2). Hypocentral solutions with azimuthal gaps <180°, rms < 0.5 s and ErrH and ErrZ errors lower than 1 km were considered the best location quality (A). On the other hand, hypocentral solutions with gaps > 180°, rms > 1.0 s and ErrH and ErrZ greater than 10.0 km gave the worst locations quality (D). In total, 134 events have quality A, 221 quality B, 31 events have quality C and only 6 events are of quality D. The comparison between the quality of the preliminary relocations (Fig. 2) and the final relocations (Fig. 6) is presented in the histograms of Fig. 8. In particular, the P- and S-phase residuals (Figs 8a–a′ and b–b′), the rms residuals (Figs 8c–c′), and the horizontal and vertical errors (Figs 8d–d′ and e–e′) are compared to the number of events. Most of the final relocations have rms of less than 0.5 s and horizontal and vertical errors of less than 1 km, demonstrating the improvement in location determination. Moreover, the comparison of mean (M) and standard deviation (SD), computed on the former parameters and marked within the histograms of Fig. 8, highlights the improvement of our final location quality compared to the preliminary ones.
The hypocentral distribution of the best relocated events (quality A and B) in terms of number of events and energy release, evaluated using the magnitude–energy relationship (log E = 9.9 + 1.9Ml + 0.024Ml2 valid for ML≤ 4.5, Richter 1958) for classes of depths, is shown in the two histograms within the left panel of Fig. 6, which depict a bimodal pattern. The 50 per cent of events occurring at depths less than ∼16 km realize 50 per cent of the energy; the 40 per cent occurring at depths between 16 and 32 km realize 40 per cent of total energy at depths between 24 and 28 km.
Assuming the base of the seismogenic layer as the depth corresponding to 90 per cent of the seismicity on the cumulative curves of the total number of events (D90NE) and energy release (D90ER; black arrows on histograms of Fig. 6), a cut-off depth of approximately 32 km is estimated. In fact, the integration of histograms with the earthquake distribution in map and section view shows that such a value is representative only of Mainland Sicily, whereas in the northern sector of the study area (the Madonie Mountains), 90 per cent of the seismicity is shallower than 16 km (Fig. 7).
The repicking process also permitted us to collect P-wave first motion polarities and compute focal mechanism solutions using the FPFIT standard procedure (Reasenberg & Oppenheimer 1985). We considered the polarities of all events occurring within the central Sicily study area from 1983 to 2010, and we evaluated the focal mechanisms for 60 events. From this data set, we selected 38 solutions with a minimum number of eight clear observations homogenously distributed on the focal sphere (approximately 70 per cent of events have a number of polarities ≥ 11) and with less than two discrepant polarities (approximately 60 per cent of events do not have discrepant polarities; Fig. 9). The selection was further checked on the basis of the two quality factors (Q) of the FPFIT code, decreasing from A to C, which are the degree of polarity misfit (Qp) and the range of uncertainties of strike, dip and rake (Qf) in a solution. The focal mechanisms parameters are listed in Table 3; 13 solutions have quality A–A; 21, A–B or B–A and 4, B–B.
The relocated epicentres are projected in the map of Fig. 9(a). Eleven focal mechanisms lie within the Plio-Quaternary Madonie Mountains extensional domain (B1 in Fig. 1), northward of the Outer Extensional Front, OEF, in Fig. 9); 23 lie in the central portion of the Mainland Sicily domain (C2 in Fig. 1), which is located between the OEF and the SBT front.
Following the Frohlich (1992) classification scheme, based on the plunge of T-, B- and P-axes, the solutions are subdivided into five kinematic categories (thrust, thrust-strike, strike, normal, normal-strike) plus an unclassified category and are represented schematically in a ternary diagram (see the uppermost left side inset of Fig. 9a). Although the overall kinematics of the considered earthquakes are of mixed type, when considering the earthquake spatial distribution in map and in section view (Figs 9a and b), it is possible to identify three spatially distinct groups with rather homogeneous kinematics. One group is located within the Madonie Mountains domain; the other two (one prevalent, the other subordinate), in Mainland Sicily. The three groups also differ in their depth distribution, as is well highlighted by the depth histograms of Fig. 10. The Madonie Mountains group (MM in Figs 9 and 10) is characterized by seismic events (2.0 ≤ML≤ 4.3) with prevalent normal and normal/oblique kinematics nucleated in the 0–15 km depth range. The Mainland Sicily prevalent group (MS1 in Figs 9 and 10) is characterized by seismic events (2.1 ≤ML≤ 4.7) with reverse and oblique/reverse kinematics, located mainly at depths between 25 and 30 km, along the downdip projection of the N-dipping SBT plane. The Mainland Sicily minor group (MS2 in Figs 9 and 10) consists of a small cluster of low magnitude events (2.0 ≤ML≤ 3.1) with a predominance of normal and oblique/normal solutions, located at depths between 15 and 25 km, within the SBT footwall.
To verify the kinematic compatibility among the population of focal mechanisms of each group, the simple right dihedral graphic construction (Angelier & Mechler 1977) was applied (Figs 10a–c). The graphic construction consists of a grid stereographic projection that indicates, at each gridpoint, the number of dihedrals delimited by the nodal planes that contain the P- or the T-axis; the necessary condition for belonging to a single stress tensor is that 100 per cent of the data fall within the area defined by the dihedral superposition. Such condition is fully satisfied for the P-dihedra grid of the MS1 group (100 per cent at one grid node) and for the T-dihedra grid of the MS2 and MM groups.
To highlight the prevailing kinematics of each group, an average focal solution was also computed by using the FaultKinWin 1.1. software (Allmendinger 2001) and by applying both the Bingham statistical procedure that does not weight the data for the seismic moment, and the moment tensor summation procedure that, conversely, does weight the data (Figs 10a′, b′ and c′).
The average focal mechanism of the Mainland Sicily prevalent group (MS1) is reverse with a subordinate strike-slip component (Fig. 10a′); a minor difference in the maximum shortening direction results from applying the Bingham or the moment tensor summation statistics, trend and plunge of the P-axis ranging from 144/14 to 133/22.
The average focal mechanism of the Mainland Sicily minor group (MS2) is normal with a subhorizontal E–W trending T-axis. (Fig. 10 b′); evidently, to this small group of rather homogenous low-magnitude (2.0 ≤ML≤ 3.1) events, only the Bingham statistics were applied.
The average focal mechanism of the Madonie Mountain group (MM) is normal with a subordinate strike-slip component (Fig. 10c′). The attitude of the maximum stretching direction computed with the Bingham statistics (nearly ENE–WSW subhorizontal T-axis) differs ∼ 30° from that derived from the moment tensor summation (nearly NNE–SSW subhorizontal T-axis); this difference implies that the NNW–SSE normal fault trend characterizes the most frequent low-magnitude events, whereas a nearly E–W fault trend characterizes rarer high-magnitude events.
To compute the seismogenic stress tensor in Mainland Sicily, we performed a stress inversion analysis of a compilation of focal mechanisms composed of two major groups of data. One group consists of all the events analysed in this paper lying within the area in between the OEF and the SBT (Fig. 9a), with the exclusion of only the small MS2 group data set. The MS2 group was excluded because of its kinematic incompatibility with the MS1 group, well highlighted by the dihedral grid construction (Fig. 10), and because of its physically independent geometric position at the SBT footwall (Fig. 9b). The other group consists of the events available in the literature (Table 3) located eastward and westward of the study area, within the boundary of SBT seismogenic province as defined by Lavecchia et al. 2007a (grey shaded area in Fig. 11a). This group consists of four events within the westward area (Mazara del Vallo and Belice valley Mw≥ 4.0) and of six events within the eastward area (two events with Mw≥ 4.0 and four deep Etnean events within Mw≥ 3.0; Anderson & Jackson 1987; Frepoli & Amato 2000; Neri et al. 2005; Pondrelli et al. 2006). The inversion was performed by applying the Gephart & Forsyth (1984) algorithm to a total of 26 events. The computed stress tensor is almost purely compressional, with a nearly horizontal, NNW–SSE trending σ1-axis (07/341) and a subvertical σ3-axis attitude (83/152); the shape factor (R=σ2-’σ1/σ3-’σ1) equal to 0.6 indicates a near triaxial tensor. The solution gives acceptably constrained stress orientations (the principal axes do not overlap) at the 95 per cent confidence level, with a minimum average misfit m= 7.4°.
We reevaluated the seismicity and the crustal dynamics of central Sicily, a portion of the Apennine–Maghrebide fold-and-thrust belt not yet well studied from a seismological and seismotectonic point of view, through the analysis of the earthquakes that occurred from 1983 to 2010. The studied data set consists of 392 events, with local magnitudes ranging between 1.0 and 4.7, almost all located at crustal depths. The final relocation, performed in several steps that are discussed in the text, was achieved with standard procedures, but it benefited from the calculation of a new 1-D velocity model for the central portion of Mainland Sicily. Such a model, until now absent from the literature, locates the upper–middle crust, the middle–lower crust and the crust–mantle boundary at depths of 12, 22, and 37 km, respectively, and it defines the corresponding impedance structure well (Fig. 5). The depth value of the Moho discontinuity is in good agreement with those available in the literature that range from a minimum of ∼32 to a maximum of ∼42 km (Cassinis et al. 2005 and references therein). In particular, we found the Moho depth to be ∼37 km, which is in good agreement with that defined by Finetti (2005) and confirmed by Tesauro et al. (2008) and Grad et al. (2009), as well as with that calculated by Chironi et al. (2000) from the modelling of the regional pattern of the Bouguer anomaly. This finding is also consistent with the crustal thickness inferred by Lavecchia et al. (2007a), as determined by an integrated analysis of tectonic and geophysical data.
The velocity model calculated here for central Sicily, integrated with those from the literature for the surrounding regions (Hybleans, Belice, Etna, Madonie-Peloritani, Aeolian Islands), which are substantially different from a structural-tectonic point of view, was used for the final relocation of the considered seismological data set (Figs 6 and 7C). Such relocation, although improved in quality, still shows a rather diffuse distribution of the seismicity and indicates some previously unrecognized clusters of events. In map view (Fig. 6), one cluster is evident in the NE corner of the study area (Madonie Mountain domain), and the other cluster is evident in its central portion, north and northwest of Caltanissetta (Mainland Sicily domain). Conversely, two seismic gaps are evident: one in the westernmost sector of the study area, close to the Belice valley area, where the destructive 1968 earthquake occurred, and the other one along the southern coast of Sicily, between Sciacca and Gela, where previous authors (Lavecchia et al. 2007a; also the DISS Working Group, http://diss.rm.ingv.it/dissNet/) identified a narrow seismogenic source associated with the shallow portion of the SBT.
Ninety per cent of the relocated earthquakes fall at depths lower than 32 km, the remaining ones being sparsely distributed at depths as great as ∼70 km (see the insets of Fig. 6). A bimodal depth distribution, in terms of not only the number of events and energy release but also prevailing kinematics (histograms in Fig. 10), is well evident. The shallower events are located at depths less than 16 km and therefore are contained mainly within the upper crustal layer; they are mainly due to the extensional seismicity of the Madonie Mountains domain. The deeper ones are located at depths of less than 32 km and therefore are contained mainly within the lower crustal layer; they are substantially related to the Mainland Sicily compressional seismicity (MS1 domain in Fig. 9).
The northward deepening of the relocated earthquakes, evident in the section view of Fig. 7(C) and in the earthquake depth contour areas of Fig. 11(c), recalls the northward deepening of the SBT, whose depth contour lines were reconstructed by Lavecchia et al. (2007a). When overlapping the hypocentral distribution of the here calculated focal mechanism with the extrapolation at depth of the SBT, the control exerted by such a still-active tectonic feature on the earthquake distribution is very well evident (Figs 9b, 11c and 11d). In fact, the compressional seismicity of Mainland Sicily (MS1 domain) lies just in the surroundings of the SBT plane, at mid- to lower-crustal depths and within its hangingwall. Conversely, the isolated small extension cluster of the MS2 domain clearly lies within the SBT footwall (Fig. 9b) and recalls other small, strike-slip and normal deformation volume, recorded in the Hyblean forelands at the focal depth of 17–25 km below the active thrust zone of the Sicily mountain chain (Scarfì 2003).
The coexistence of seismogenic extension in the Madonie domain at upper crustal depths (< ∼15 km) and of compression in mainland Sicily at mid-to-lower crustal depths, with areas of partial geometric overlap between the two (Fig. 11d), recalls the distribution of the seismogenic province in central Italy, where the pede-Apennine NNW–SSE striking mid-to-deep crustal thrust structures coexist in the same area with the shallower and coaxial Apennine extensional faults and deepens beneath them (Lavecchia et al. 2003; Lavecchia et al. 2007b).
Whether the maximum NNW–SSE average shortening direction computed in this paper in Mainland Sicily (Fig. 10a′) is coaxial with the average Madonie maximum stretching direction (Fig. 10b′) is not clear from the analysis of our focal mechanisms. In fact, a nearly WSW–ENE trend of T-axis is obtained when applying the Bingham statistics whereas an average NNE–SSW trend is obtained when weighting more the largest events. Both directions are partially supported when considering outside data. The WSW–ENE trend of T-axis is consistent with the tensional strain pattern from GPS data reported by Cuffaro et al. (2011) and with field fault slip data reported by Billi et al. (2010) and Barreca et al. (2010). Conversely, the NNE–SSW direction is subparallel with T-axes derived from RCMT focal mechanisms (Pondrelli et al. 2006) and from geodetic data reported by Devoti et al. (2011). More information is needed to underscore the true meaning of this observed contradiction, but what is certainly confirmed is the existence of an extensional domain in northeastern Sicily, as proposed by Lavecchia et al. 2007a, b and by Billi et al. 2010.
The compressional tectonic pattern defined in Mainland Sicily from the new focal mechanisms (black beach balls in Fig. 11a), which is almost purely compressive with a NNW–SSE oriented P-axis, is consistent with all the available information (red beach balls in Fig. 11a). In fact, it is similar to the fault plane solutions of the Belice 1968 sequence (Mw 6.1) and to the 1981 Mazara earthquake (Mw 4.9), which occurred in western Sicily, as well as to the average focal mechanism computed with small magnitude events located at middle and lower crustal depths beneath the Etna edifice (Visini et al. 2010). The stress inversion of all the focal mechanisms falling within the boundary of the SBT province (transparent grey in Fig. 11a) has further demonstrated the presence of a largely homogenous NNW–SSE compressive stress field (Fig. 11b). The obtained tensor is consistent with the ones previously inverted from focal mechanisms by Neri et al. (2005), as well as with the principal strain axes evaluated by Visini et al. (2010). Moreover, the oblique convergence of central Sicily relative to the GPS site Noto, within the Hyblean Foreland, was noted by Ferranti et al. (2008). More recently, on the basis of geodetic data, Devoti et al. (2011) computed an active NNW–SSE convergence of ∼1.1 ± 0.2 mm yr–1 between the Mainland Sicilian domain and the Hyblean–Malta Plateau, and they also noted a slight subsidence of the rigid plateau block (0.2 ± 0.2 mm yr–1), opposed to an uplift of 1.2 ± 0.5 mm yr–1 in northwestern Sicily. Such a deformation pattern well supports the hypothesis of the active thrusting of Mainland Sicily above the SBT.
The active compressional activity in Mainland Sicily (Fig. 11b) is nearly coaxial to that of the E–W striking compressional domain offshore of northern Sicily (Fig. 1). Some authors have interpreted the latter as proof of a backward shift of the Sicilian Apennine–Maghrebian active thrust front and for a definitive abandonment of the SBT emerging along the Sciacca–Gela–Catania thrust front (Chiarabba et al. 2005). Other authors consider the compressional strip as an independent process, because of an incipient southward dipping subduction of the Tyrrhenian thinned lithosphere beneath the Sicilian crust (Goes et al. 2004; Billi et al. 2007). Available data do not allow the problem to be solved unequivocally. Seismological and geodetic data show clearly that the offshore strip and the Mainland Sicily domains are separated by the interposed northern Sicily Madonie–Nebrodi–Peloritani extensional belt (Devoti et al. 2011 and references therein); therefore, the hypothesis of a large active compressional domain from to southern Tyrrhenian to southern Sicily (Jenny et al. 2006) can be reasonably excluded. We suggest that two compressional domains, although coaxial, would be an expression of two different geodynamic processes: the compressive strip offshore northern Sicily would absorb most of the convergence rate between the convergent Nubia and Eurasia plates, whereas the Mainland Sicily compression domain, at the outer front of the Apennine–Maghrebides fold-and-thrust belt, and the northern Sicily extensional domain might represent a still-ongoing activity of the outward moving Tyrrhenian–Apennine extension–compression pair, which since late Miocene times has been responsible for coeval inward extension and outward compression at a crustal scale (Lavecchia 1988).
This paper provides new data and insight to better understand the seismicity and tectonic framework of central Sicily. The computation of a new 1-D velocity model through the inversion of seismic data, the relocation of events by using the new model and the computation of focal mechanisms have especially contributed to suggest the importance of the geometry and the seismogenic role played by the SBT, the outermost thrust of the outward-verging Apennine–Maghrebian thrust system. The SBT deepens at a low-angle (average 30°) northward, cross-cutting the middle–lower crustal transition (22 km) approximately beneath Caltanissetta and reaching the Moho discontinuity at depth of ∼ 35 km, beneath the southern portion of the Madonie Mountains domain (Fig. 11d). The data analysed in this paper show that the seismogenic thrust is located mainly at mid-to-lower crustal depths between 20 and 32 km, thus helping to confirm the thick-skinned configuration of the active thrust structure at the outermost Sicilian front. Such configuration, already proposed in the literature (Guarnieri et al. 2002 and Lavecchia et al. 2007a), also fits well with a crustal reflection seismic profile recently recorded across central Sicily by Accaino et al. (2010), from the Tyrrhenian shore to the Sicily Channel.
In the recent literature (Basili et al. 2008; Meletti et al. 2008), as well as in the seismic zonation of the Italian territory that is available online (Working Group DISS 2010), the earthquake potential on the SBT is still largely under evaluated. The Mainland Sicily territory, which is located along the surface projection of the lower crustal portion of the SBT, is classified as an ‘aseismic region’, and a narrow compressional zone characterizes only the shallow portion of the SBT along the Sciacca–Gela–Catania front in Southern Sicily. Our results show that a large seismogenic province, here called SBT province (grey transparent area in Fig. 11a), is undergoing homogenous NNW–SSE compressive stress, at the hangingwall of the SBT, with most of the seismic shearing being released just along the basal thrust plane in the 20–35 km depth interval. Such results, although still not fully accepted in the literature, which sometimes considers the SBT and the associated splays as an abandoned structure (Chiarabba et al. 2005), are fully supported by geodetic data, which indicate the SSE-verging thrusting of Mainland Sicily at relevant geodetic rates (Devoti et al. 2011). The implications in terms of seismic hazard assessment are evident as, in historical times, the SBT has been capable of releasing at least three large earthquakes with Mw≥ 6.0 (361, 1818, 1968; Fig. 1a) and several more moderate events (Lavecchia et al. 2007a). The theory of a possible association with the highly destructive 1693 earthquake (Fig. 1a) of eastern Sicily (Io = XI MCS) (Visini et al. 2009) should also be considered.
We thank the Editor Ingo Grevemeyer and the two anonymous reviewers for the constructive remarks that greatly improved the work. We also thank Paolo Boncio, Paolo Favali and Stephen Monna for their valuable comments and discussions.