Volume 40, Issue 7 e2020TC006465
Research Article
Open Access

The Segmented Campo Felice Normal Faults: Seismic Potential Appraisal by Application of Empirical Relationships Between Rupture Length and Earthquake Magnitude in the Central Apennines, Italy

Giulia Schirripa Spagnolo

Corresponding Author

Giulia Schirripa Spagnolo

Dipartimento di Scienze della Terra, Sapienza Università di Roma, Rome, Italy

Correspondence to:

G. Schirripa Spagnolo,

[email protected]

Search for more papers by this author
Marco Mercuri

Marco Mercuri

Dipartimento di Scienze della Terra, Sapienza Università di Roma, Rome, Italy

Search for more papers by this author
Andrea Billi

Andrea Billi

IGAG, Consiglio Nazionale delle Ricerche, Rome, Italy

Search for more papers by this author
Eugenio Carminati

Eugenio Carminati

Dipartimento di Scienze della Terra, Sapienza Università di Roma, Rome, Italy

Search for more papers by this author
Paolo Galli

Paolo Galli

IGAG, Consiglio Nazionale delle Ricerche, Rome, Italy

Dipartimento della Protezione Civile Nazionale, Rome, Italy

Search for more papers by this author
First published: 21 June 2021
Citations: 4

Abstract

During earthquakes, fault rupture can involve multiple segments in synchronous or cascade mechanisms, leading to an increasing magnitude of the mainshock or rate of aftershocks. Since the seismogenic portions of faults are inaccessible, studying the geometrical and mechanical interaction between exhumed fault segments can contribute to the understanding of multisegment and cascade earthquake scenarios at depth. We investigated a segmented active normal fault system in the Campo Felice area (central Italian Apennines), where fault scarps are well exposed. In this area, there are no instrumental-historical records of intermediate-strong earthquakes, although paleoseismology provided evidence for ancient Mw > 5 earthquakes. Geometry and kinematics of the studied faults as well as their physical linkage and mechanical interaction were assessed. Results of field surveys and geological-structural mapping, serial cross-sections, and throw versus distance diagrams highlight different stages of mechanical interaction between the Campo Felice faults. The suitability of three empirical equations relating earthquake rupture length and magnitude was tested in comparison to a new equation that we developed considering the last seven Mw > 5.5 earthquakes (1997–2016) from the central Apennines. Results show that the Campo Felice faults can produce earthquakes with maximum Mw of ∼5.8 and 6.2 with single or synchronous ruptures, respectively. In turn, Coulomb stress change modeling shows that the seismic hazard can increase considering a quasi-synchronous or cascade activation of the Campo Felice faults together with nearby faults.

Key Points

  • Single and multisegment seismic potential for the Campo Felice normal faults (5.7 ≤ Mw ≥ 6.2)

  • Geological mapping and structural analyses for the reconstruction of possible future seismic scenarios

  • Validation of rupture length-earthquake magnitude empirical equations for central Apennines and comparison with a new proposed equation

1 Introduction

Recent seismological studies have highlighted different coseismic rupture scenarios. In some seismic sequences, the mainshock is followed by aftershocks having comparable magnitude. Significant examples are from Chile (2014–2015 seismic sequences, Mw 8.1–8.4; Papadopoulos & Minadakis, 2016) and Italy (2016; Mw 6.2–6.5; Chiaraluce et al., 2017; Figure 1a). In these cases, the slip on the fault segment causing the mainshock led to a stress transfer (i.e., Coulomb stress change; King et al., 1994; Stein, 2003; Tung & Masterlark, 2018) on nearby fault segments or patches from the same fault and hence to strong aftershocks or to a new larger mainshock. In these scenarios, the triggering delay times are generally in the order of 0.1–1 year (Scholz, 2010).

Details are in the caption following the image

(a) Active faults and seismicity of the central Italian Apennines. The map reports the fault segments showing post-Last Glacial Maximum activity and the associated historical seismicity (dates refer to the main events of the last century). The dashed red box shows the location of the map of Figure 1b, which details the fault segments surrounding the study area. Inset: Location of study area within Europe. (b) The study area is characterized by compressive structures (blue in figure) such as the Mt. Cefalone-Mt. Rotondo thrust, and by normal faults (black in figure), some of them are inferred to be active (red in figure) by paleoseismological studies (Galli et al., 2008), that are indicated with bulldozers. The normal faults considered in this work are: Colle Cerasitto, Monte Cefalone (MCF), Monte Orsello (MOF), Tornimparte, and Ovindoli-Pezza faults. In this study, the Campo Felice faults include MOF and MCF. These two faults, which are the main subject of this work, border the Campo Felice basin. The yellow and green stars indicate the epicenters of the L'Aquila 2009 and Avezzano 1915 earthquakes, with their associated focal mechanisms (Galli et al., 2012). The different background colors show the main outcropping lithological successions (modified after Mercuri et al., 2020; Galli, 2020).

Coseismic deformation can also involve quasi-synchronous or cascade slip events over multiple adjacent fault segments during a single seismic event (Xu et al., 2018), all these contributing to the magnitude of the earthquake. The activation of multiple fault segments during an earthquake has been observed in several tectonic settings. For instance, under a strike-slip regime, the coseismic rupture during the 2016 Kaikoura earthquake (Mw 7.8, New Zealand) involved at least 21 fault segments for a total surface rupture length of about 180 km (Xu et al., 2018). The fault linkage caused a dramatic increase of the coseismic slip area and therefore of the released energy. Another significant example of multiple dynamic rupture is the 2012 Sumatra earthquake (Mw 8.6; Meng et al., 2012). The large magnitude of this earthquake resulted to be the consequence of the wide depth extent of the causative fault, high stress drop, and synchronous rupture of multiple fault segments that were oriented even orthogonally one to the other (Meng et al., 2012). Under extensional regime, three recent earthquake sequences in central Apennines (1997 Colfiorito, Mw 6.0; 2009 L'Aquila, Mw 6.3; 2016 Amatrice-Norcia, Mw 6.0 and 6.5; Figure 1a) had mainshocks characterized by coseismic ruptures over several normal fault segments (Chiaraluce et al., 200320112017; Delorme et al., 2019; Iezzi et al., 2019; Pucci et al., 2019; Smeraglia et al., 2017). The Amatrice-Norcia sequence started on August 24, 2016, with a Mw 6.0 earthquake that ruptured the two adjacent tips of the Mt. Gorzano-Mt. Vettore fault. On October 26, a Mw 5.9 event ruptured the northernmost tip of the Mt. Vettore fault, whereas on October 30, a Mw 6.5 earthquake re-ruptured the whole fault, propagating along strike and involving slip over further fault segments (Brozzetti et al., 2019; Galli et al., 2019; Iezzi et al., 2019; Scognamiglio et al., 2018; Villani et al., 2018; Figure 1a).

In assessing the seismic potential of fault systems, adjacent fault segments should be considered collectively to test the possibility of their coseismic simultaneous activation (Brozzetti et al., 2019; Iezzi et al., 2019; Morell et al., 2020; Scotti et al., 2021; Sgambato et al., 2020; Walker et al., 2021; Wesnousky, 2008; Xu et al., 2018). The capability of coseismically breaking inter-segment barriers depends on the degree of mechanical interaction. Obviously, the higher the linkage between fault segments, the more prone they are to break in cascading coseismic events (Manighetti et al., 2007). Thus, the knowledge of fault interaction is essential to decipher possible future rupture scenarios (Sgambato et al., 2020). Field structural studies focusing on the linkage between adjacent capable faults (e.g., Cowie & Roberts, 2001; Fossen & Rotevatn, 2016; Gupta & Scholz, 2000; Lunn et al., 2008; McClay & Khalil, 1998; Peacock & Sanderson, 19911994; Peacock et al., 2017; Rotevatn & Peacock, 2018; Schlische et al., 1996; Walsh & Watterson, 1991) can help detecting multisegment sources and assessing the seismogenic potential, based on their cumulative length (Gupta & Scholz, 2000; Iezzi et al., 2020; Peacock & Sanderson, 19911994; Walsh & Watterson, 1991), assuming that this length is comparable to the earthquake maximum coseismic surface rupture length (e.g., Alvarado et al., 2014; Iezzi et al., 2019; Mignan et al., 2015; Nicol et al., 2020; Tondi et al., 2020; Trippetta et al., 2019). The use of empirical relationships linking earthquake rupture size and magnitude (Leonard, 2010; Wells & Coppersmith, 1994) is suggested to assess maximum expected potential magnitude where no compelling data from instrumental, historical, and paleoseismological earthquake catalogs exist. Where possible, these equations should be verified and validated locally as the database on which they are founded usually includes a limited number of events from the studied area and may therefore be poorly applicable. The validation of these empirical equations, together with the assessment of fault segment linkage, is fundamental to better constrain the seismic hazard, particularly where active faults are well-exposed.

The seismically active central Apennines are an excellent natural laboratory for field based studies of active faults. In this study, we focus on the seismic potential of some active normal faults located in the Campo Felice basin (Long. 13°27″E, Lat. 42°13″N, 1,530 m a.s.l.; Figures 1, 2 and S1), by studying the exposed segmentation and linkage between adjacent fault segments. Italy has a dense seismological network and a rare if not unique historic record of earthquakes (Rovida et al., 2016). The richness of the seismotectonic datasets makes Italy a suitable country to validate the empirical equations between rupture size and earthquake magnitude, using fault length as input data. We apply this approach to the active normal faults of the Campo Felice area (Figures 1, 2 and S1). Although these faults, namely Monte Cefalone (MCF) and Monte Orsello (MOF) faults (Figures 1, 2, and S1), have not produced documented records of earthquakes in historical times, they are aligned with the seismogenic sources of the L'Aquila 2009 (Mw 6.3; e.g., Cheloni et al., 2010) and Avezzano 1915 (Mw 7; e.g., Galli et al., 2016) normal fault earthquakes (Figure 1). Moreover, paleoseismological evidence of the occurrence of paleo-earthquakes was documented along these faults (Benedetti et al., 2013; Galli et al., 2008; Giaccio et al., 2003); their geometry and potential mechanical interaction are, however, still unclear.

Based on our and previous geological mapping (Benedetti et al., 2013; Bigi et al., 1995; Ghisetti & Vezzani, 1996; Morewood & Roberts, 2000; Wilkinson et al., 2015), geometric and kinematic analyses (Bigi et al., 1995; Ghisetti & Vezzani, 1996; Morewood & Roberts, 2000; Wilkinson et al., 2015), serial cross-sectioning (Morewood & Roberts, 2000), and throw versus distance diagrams (Benedetti et al., 2013; Morewood & Roberts, 2000; Schlische et al., 1996; Wilkinson et al., 2015) in the Campo Felice area, we evaluated the linkage between fault segments and assessed the seismic potential of the faults in this area. Coupling field observations to Coulomb stress change modeling and employing both known and new empirical relationships between earthquake rupture length and magnitude, we present possible seismic scenarios for the area.

2 Geological Setting

2.1 Central Apennines

The central Apennines are the external domains of a NW-SE-oriented fold-thrust belt that developed since late Oligocene time by the westward subduction of the Adriatic plate below the eastwardly advancing European plate (Carminati & Doglioni, 2012; Dewey et al., 1989; Malinverno & Ryan, 1986). In the analyzed region, Late Miocene – Early Pliocene thrust faulting juxtaposed Lower Jurassic-Middle Miocene carbonates deposited in shallow-water or pelagic environments onto Upper Miocene synorogenic deposits (Cosentino et al., 2010; Ghisetti & Vezzani, 1996; Figure 1b). Since at least Pliocene time, the central Apennines was subjected to NE-SW oriented extension due to the diffuse back-arc stretching affecting the Tyrrhenian side of the orogen (Carminati & Doglioni, 2012; Malinverno & Ryan, 1986). At present, this extensional regime is accommodated by active normal faults striking mainly NW-SE, which have dissected the fold-thrust belt, partly inverting inherited thrust ramps (Cavinato & Celles, 1999; Di Luccio et al., 2010; Tortorici et al., 2019), generating several intermontane basins (e.g., Fucino, Sulmona, Aterno, Campo Imperatore, and Campo Felice basins; Cosentino et al., 2010; Galadini & Galli, 2000; Figure 1b). These active extensional faults generated most of the largest historically and instrumentally recorded earthquakes in central Italy.

2.2 Campo Felice Basin

The Campo Felice area is a tectonic intermontane extensional basin elongated in a NW-SE direction, filled by alluvial, lacustrine, and glacial deposits (Late Pleistocene to Holocene in age; Giraudi, 1995; Giraudi & Giaccio, 2017; Giraudi et al., 2011; Figures 1b and 2a). The Quaternary basin is surrounded by the Latium-Abruzzi, shallow-water carbonate platform succession (Servizio Geologico d'Italia-APAT, 2006; Figures 1b and 2a), which is capped by the synorogenic Tortonian to Messinian hemipelagic marls and, in places, by Messinian turbiditic siliciclastic deposits (Brandano, 2017; Cipollari & Cosentino, 1995; Figures 1b and 2a) here named Monte Ocre succession. In particular, in the type-area, Early Cretaceous levels and the unconformable Miocene deposits are widely exposed (Figure 1b). For a detailed description of the Monte Ocre stratigraphic succession, we refer the reader to the supplementary material (Figure S1).

Details are in the caption following the image

(a) Simplified geological map of the Campo Felice area showing Tornimparte, Monte Orsello (MOF), Monte Cefalone (MCF), and Ovindoli-Pezza faults. This figure includes a simplified stratigraphy of the study area and Schmidt nets (lower hemisphere) showing attitude of fault planes and striae. In the nets, black is for SW-dipping planes, green is for NE-SW-striking planes, and orange is for N-S-striking planes. Tracks of geological cross-sections are in blue (cross-sections are in Figures 5 and S2). Geographic coordinates are based on the UTM system (WGS84- zone 33T). Equidistance between contour lines of topography is 50 m. See Figure S1 for a complete version at the 1:20,000 scale of this map. (b) Outcrop of MOF escarpment (see Figure 2a for its location). (c) Outcrop of MCF fault escarpment (see Figure 2a for its location).

The main evidence of Late Miocene – Early Pliocene contractional tectonics in the Campo Felice area is the Monte Cefalone-Monte Rotondo thrust (e.g., Bigi et al., 1995; Figure 1b). This thrust borders the eastern side of the Monte Cefalone-Monte Serralunga ridge and extends southeastward reaching the Ovindoli and Celano villages (Figure 1b).

The Campo Felice basin is bounded toward the NE by two left-stepping en echelon NW-striking normal faults named Monte Cefalone and Monte Orsello (MCF and MOF, respectively, in Figure 1b), both showing evidence of Late Pleistocene-Holocene activity, documented through geomorphologic and paleoseismological analyses (Benedetti et al., 2013; Galadini & Galli, 2000; Galli et al., 2008; Giaccio et al., 2003; Giraudi, 1995; Giraudi & Giaccio, 2017).

The impressive MCF rock fault scarp (Bosi, 1975) allowed previous authors to infer: (a) through micro-morphologic investigations of the fault surface, the occurrence of two Mw > 6 paleoearthquakes, presumably between 860 and 1300 AD, and around 1900 BC, respectively (Giaccio et al., 2003); (b) through 36Cl dating, Holocene slip with at least four exhumation phases, possibly matching earthquakes with 6.2  <  Mw < 6.5, dated back to 9.4, 4.2, 3.4, and 1.1 ka, respectively (Benedetti et al., 2013); (c) through cosmogenic 36Cl concentration, an increase of slip rates in the last 4 ka with a peak between 0.5–2 ka (Goodall et al., 2021).

The MCF pertains to a longer fault system that comprises the Ovindoli-Pezza fault (OPF) to the SE and the Colle Cerasitto fault (CCF) to the NW (Figure 1). The northwestern part of OPF, dipping by 50° toward SW, borders the Piano di Pezza basin with an extensional kinematics (Villani et al., 2015). Here, paleoseismic trenching provided results roughly matching those of MCF, consisting of two Mw > 6 events occurred between 860 and 1300 AD, and around 1900 BC, respectively (Pantosti et al., 1996).

The CCF is ∼9.5 km long with an average strike of N140° (Salvi et al., 2003). Previous paleoseismological studies on the CCF showed: (a) gravitational displacements in addition to tectonic movements through analysis of COSMO-SkyMed InSAR data (Albano et al., 2015); (b) four shallow faulting events (6 < Mw < 7) in the past 20 kyr through GPR investigations (Salvi et al., 2003), three of which roughly matching the radiometric ages obtained for the events along OPF (Pantosti et al., 1996) and MCF (Giaccio et al., 2003). Based on these paleoseismological analyses, OPF, MCF, and CCF possibly ruptured together during some past earthquakes, releasing Mw ∼ 6.7 events with a recurrence time of ∼2.5 ka (Galli, 2020). Alternatively, these faults may have ruptured separately, thus originating earthquakes of lower magnitude (∼6.3; Galli, 2020). Paleoseismological studies provide an estimated throw-rate for these faults (MCF, OPF, and CCF) spanning between 0.8 and 1.3 mm/yr during the last 18 ka (Galadini & Galli, 2000; Papanikolaou et al., 2005).

So far, the MOF (Figure 1) has not been studied and mapped in detail. Therefore, its geometrical relationship with the MCF is not completely clear (Bigi et al., 1995; Ghisetti & Vezzani, 1996; Morewood & Roberts, 2000). There are no paleoseismological studies on the MOF, although its eastern prolongation across the Campo Felice basin cuts through local moraines dated ∼36 and ∼17 ka (Giraudi & Giaccio, 2017; Giraudi et al., 2011). This evidence suggests a late Quaternary activity for MOF as well. Moreover, Tondi and Cello (2003) suggested the MOF to be the seismogenic source of the 1786 L'Aquila earthquake.

To NW of MOF, the presence of the Tornimparte fault (TF), which continues northward for about 7 km, is inferred from the almost rectilinear contact between lower Cretaceous Gastropods limestones and Messinian synorogenic deposits (Servizio Geologico d'Italia-APAT, 2010). The principal surface of the TF is exposed in the northern portion and is characterized by predominant left-lateral strike-slip kinematic indicators (Bigi et al., 1995; Centamore & Dramis, 2010). The lack of faulted Quaternary deposits does not permit any inference regarding its activity under the present and recent NE-SW extensional stress field. Nevertheless, Tondi and Cello (2003) proposed the TF as the seismogenic source of the 1349 L'Aquila earthquake.

3 Methods

3.1 Field Mapping

The purpose of our field structural analysis is investigating the fault geometry and kinematics. During the 1:10,000 field surveys, we (a) focused in particular on the tip areas of the Campo Felice faults (MOF and MCF), and (b) carefully mapped and measured the lateral continuity of the faults to obtain reliable results on the potential surface rupture length. We labeled and mapped “normal faults” and “thrust faults” the structures that we observed directly and measured in the field. With “inferred faults” are labeled structures that we were not able to observe directly in the field but that we deduced through anomalous relationships within the sedimentary succession or whose occurrence we supposed from the prolongation of observed faults. We labeled as “faults” structures directly observed in the field without any kinematic indicator, for which we deduced extensional-oblique kinematics on the basis of the rocks cropping out in the footwall and hangingwall blocks (Figures 2 and S1). For geometric and kinematic analyses of faults and kinematic indicators, we used the lower hemisphere Schmidt stereographic projections. Based on our geological surveys and mapping at the 1:10,000 scale, on the topographic map by CTR Abruzzo (scale 1:10,000), on the geological map and stratigraphy by ISPRA (Foglio 359 “L'Aquila,” Servizio Geologico d'Italia-APAT, 2006), and on the digital terrain model with 10 m resolution by CTR Abruzzo, we realized the geological map shown in Figure S1. In detail, we mapped all the faults with trace length >∼0.5 km.

3.2 Geological Cross-Sections

The purpose of the construction of 12 serial geological cross-sections at the 1:20,000 scale is the estimation of the vertical component of fault displacements (i.e., fault throws) associated with the studied faults (i.e., TF, MOF, and MCF; Figures 2 and S1). All cross-sections are reported in Figure S2. The cross-section traces are oriented perpendicular to the fault strike, are 1–6 km long, and are spaced about 1 km (Figures 2 and S1). Bedding dip and thickness of formations (Figures 2 and S1) are constrained by our original field data and by previous works (Servizio Geologico d'Italia-APAT, 20042006, and 2010). With the exception of thickness variations in RDT deposits, geological data do not allow us to constrain eventual other thickness variations for Mesozoic and Cenozoic deposits along our geological cross-sections. For this reason, we assumed a constant thickness of these rock units across all cross-sections. Apparent dip of bedding was calculated using the nomogram. To evaluate fault throw, we leveraged on two easily recognizable stratigraphic markers characterized by small thickness and well constrained age. They are the Cenomanian Intra-bauxitic limestones (IBX), with a thickness of ∼50 m, and the early Aptian micritic limestones alternating with green clayey marls, with a thickness of ∼50 m (Figure S1). In each cross-section, we evaluated the throw of fault segments calculating the difference between elevations of hangingwall and footwall cutoffs with the fault segments.

With the exception of the north-easternmost portion of cross-sections L-L′, M-M′, and N-N′, where boreholes constrain the thickness of Late Pleistocene continental deposits to be 100 m (Giraudi et al., 2011), no geological and geophysical data are available to infer the depth of the bottom of these deposits in the remaining parts of the Campo Felice plain. To avoid over interpretations, we choose to leave blank (with question marks) the contact between the bedrock and continental deposits in the unconstrained areas. Inferred faults below recent deposits are drawn as SW-dipping by 55° and extensional kinematics, coherently with other nearby outcropping faults.

Due to uncertainties/variations of the dip angle of bedding, we attributed an asymmetric error to each throw measurement on fault segments. In particular, we calculated the maximum (Ti + δipos) and minimum (Ti + δineg) possible throw for each fault segment as follows. Ti is the calculated throw, whereas δipos and δineg are positive and negative errors, respectively. Even where fault throw was well constrained, we assumed a minimum error of 5 m to take into account possible instrumental uncertainties during measurements with the digital ruler. We also calculated the total throw and its associated error for the main faults (i.e., MCF and MOF). Unconstrained/unrecognized thickness variations of Mesozoic and Cenozoic formations could in principle affect throw calculations. However, since no post-RDT strata were used to calculate the throws, we are confident that the constant thickness assumption does not impact significantly on the evaluation of errors.

Since the throw measurements are independent, we used the method by D'Agostini (2004) to calculate final throw (Tfin) and their error (urn:x-wiley:02787407:media:tect21579:tect21579-math-0001) for each main fault:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0002

3.3 Fault Distance Versus Throw Diagrams

Fault distance versus throw diagrams were built to investigate the throw distribution along fault strike and to assess the stage of linkage (i.e., soft- vs. hard-linkages; Walsh & Watterson, 1991) between the TF, MOF, and MCF segments. We included in this analysis only the “inferred” fault segments deduced such as prolongation of studied faults. Hard-linkage occurs where fault segments are physically linked by transfer faults (Peacock, 2002). In this case, the fault trace abruptly changes its orientation in map view. Transfer faults linking two faults transmit the total amount of throw from one fault to the other (Boccaletti et al., 1998; Chorowicz & Sorlien, 1992; Moustafa, 2002). In turn, where a soft-linkage occurs, faults are not geometrically linked but they mechanically interact (Gupta & Scholz, 2000; Walsh & Watterson, 1991). In this case, the mechanical interaction between soft-linked faults produces a relay ramp, which is characterized by the reorientation of bedding in fault segment overstep areas to connect the footwall of a fault tip to the hangingwall of the adjacent one (Ferrill & Morris, 2001; Peacock & Sanderson, 19911994; Walsh & Watterson, 1991; Walsh et al., 1999). The evolution from soft-linkage to hard-linkage is characterized by breaching of the relay-ramp, where fractures and faults with horsetail geometry start to link the overstepping segments (Ferrill & Morris, 2001; McClay & Khalil, 1998; Peacock & Sanderson, 19911994). In addition to field observations (i.e., geometry and kinematics of minor faults, and attitude of bedding), diagrams displaying fault displacement versus along strike distance are regularly used to assess the degree of interaction between adjacent fault segments. In such diagrams, the displacement gradient becomes progressively steeper in proximity of soft-linked fault tips: The steeper the gradient, the larger the interaction between faults (e.g., Ferrill & Morris, 2001; Iezzi et al., 2020; Peacock & Sanderson, 19911994; Schlische et al., 1996; Spina et al., 2008; Williams & Chapman, 1983).

3.4 Seismic Potential

The purpose of empirical equations linking rupture size and maximum potential earthquake magnitude is the assessment of the fault seismic potential (i.e., maximum potential earthquake magnitude). We use three empirical equations valid for normal dip-slip faults from Galli et al., (2008), Leonard (2010), and Wells and Coppersmith (1994). Equations by Leonard (2010) and Wells and Coppersmith (1994) are general and used worldwide, whereas the equation by Galli et al. (2008) was specifically developed for the present extensional tectonics of Apennines.

The equation by Wells and Coppersmith (1994) is:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0003(1)
where Mw is the moment magnitude and L is the surface rupture length expressed in km. The standard deviation associated with this equation is 0.34. Equation 1 is empirically based on a worldwide database of 15 selected historical extensional earthquakes and is valid for earthquakes shallower than 40 km with 5.2 < Mw < 7.3, and for (rupture length) 2.5 < L (km) < 41 (Wells & Coppersmith, 1994). The database used by Wells and Coppersmith (1994) mainly includes earthquakes from USA, but it also included five earthquakes from Italy, namely the Avezzano 1915 Mw 7, Umbria 1979 Mw 5.9, Irpinia 1980 Mw 6.9, Lazio-Abruzzo 1984 Mw 5.8, and Umbria 1984 Mw 5.3.
The equation by Galli et al. (2008) is:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0004(2)
where L is the surface rupture fault length expressed in km obtained by mapping of active faults and by paleoseismological studies. This equation is empirically based on a database of 16 paleoearthquakes occurred in the past 2.4 ka in the Apennines with Mw > 5.5 (Galli et al., 2008).
The equation by Leonard (2010) is:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0005(3)
where L is the surface rupture length expressed in km. Equation 3 is an improvement and refinement of Equation 1. To formulate Equation 3, Leonard (2010) used worldwide databases from previous works (Hanks & Bakun, 2002; Henry & Das, 2001; Manighetti et al., 2007; Romanowicz & Ruff, 2002; Wells & Coppersmith, 1994). Limits of Equation 3 are the same ones of Equation 1.

In order to best estimate the magnitude expected by the faults studied in this work, we compared, through the chi-squared Χ2 statistical test (Mann & Wald, 1942), the aforementioned empirical equations (Galli et al., 2008; Leonard, 2010; Wells & Coppersmith, 1994) with data from the last seven Mw > 5.5 earthquakes occurred in the central Apennines from 1997 to July 2020 (Table S2). Through linear regression of these seven data, we propose an ad-hoc new empirical equation valid for central Italy that is reported in the Discussion section.

3.5 Coulomb Stress Change Modeling

Coulomb stress change modeling for different earthquakes scenarios was performed to infer post-seismic stress transfer between adjacent fault segments and eventually assess the possibility for a potential cascade or quasi-synchronous activation of adjacent faults during seismic events. We used Coulomb 3.3 software by Toda et al. (2011) and approximated the measured exposed fault length with the coseismic surface rupture length. The width of the rupture (W; i.e., the downdip rupture length) is calculated from the rupture length (L; i.e., the along strike length) using the following equation by Leonard (2010), valid for extensional faults with length >5.5 km:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0006(4)

Moreover, the fault bottomset depth (d) is calculated using the rupture width (W; see Equation 4) and the average dip angle of faults (∼52°, from field surveys): urn:x-wiley:02787407:media:tect21579:tect21579-math-0007.

Following Okada (1992), we model seismic events on the studied faults with an average coseismic slip (Dav) over the total surface of faults. We calculate the average coseismic slip value (Dav) using the following empirical equation by Leonard (2010), valid for extensional faults with length >5.5 km:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0008(5)

The assumptions of Coulomb stress modeling are as follows: (a) an elastic half-space with uniform isotropic elastic properties is used following Okada (1992); (b) coefficients of pore fluid flow and of static friction are kept constant (Stein et al., 1997); (c) coseismic slip is considered to remain constant along the whole fault (Okada, 1992; Stein et al., 1997); (d) the receiver faults are assumed as not slipping but optimally oriented for seismic failure (King et al., 1994; Lin & Stein, 2004; Stein et al., 1997; Toda et al., 2011).

During an earthquake, Coulomb stress diminishes along the source fault. As this stress cannot simply disappear, it is usually redistributed to other sites along the same fault or to nearby receiver faults. This increase in Coulomb stress could be sufficient to trigger earthquakes at new nearby locations (Stein, 2003). The equation of Coulomb stress change (e.g., Okada, 1992) is defined as:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0009(6)
where σc is the Coulomb stress change, τ and σn are the changes in shear and normal stress on a given fault plane, μ = 0.75 is the friction coefficient, and B = 0.47 is the Skempton's coefficient. Areas with positive Coulomb stress change are more prone to experience future earthquakes, possibly resulting in cascading or quasi-synchronous seismic events. It follows that the wider the lobes of Coulomb stress change, the larger the segment of faults that can seismically slip, and the higher the released magnitude (e.g., King et al., 1994; Okada, 1992).

4 Results

4.1 The Campo Felice Faults

In the Campo Felice area, Mesozoic carbonate beds dip toward NE by about 30° and are crosscut by two segmented main normal faults (MOF and MCF). The study area also includes the southeastern part of TF and northwestern part of OPF (Figure 2). All these main faults are associated with various synthetic, antithetic, or differently oriented subsidiary faults, with 1–10 m throw. MOF and MCF strike NW-SE and dip toward SW by about 50°, are arranged with an en echelon pattern in map view and show corrugations from the metric to kilometric scale (Figures 2 and S1). MOF consists of two ∼5 km long overstepping segments that juxtapose Cretaceous limestones on the footwall against Miocene limestones on the hangingwall (Figure 2). MCF is ∼9 km long and hosts Cretaceous limestones in the footwall and post Last Glacial Maximum (LGM) deposits in the hangingwall (Figures 2S1 and S2). Slickenlines and other kinematic indicators on the main fault scarps show dip-slip extensional kinematics (Figure 2). From a kinematic point of view, subsidiary faults can be divided into three main sets: Strike-slip faults striking N-S with high dip angles; oblique-slip faults striking NE-SW with high dip angles; and extensional/oblique slip faults striking NW-SE with intermediate-high dip angles (Figures 2a and 3).

Details are in the caption following the image

Focus on the area between Tornimparte (TF) and Monte Orsello (MOF) faults (see location in Figure 2a; see also Figures S1 and S2). AF (i.e., Accommodation Fault) indicates a secondary fault with relevant throw, accommodating the throw difference between TF and MOF. (a) Geological map of the area between TF and MOF with track of geological cross-section CC'. For the stratigraphy and legend, refer to Figure 2a. (b) Outcrop of a fault damage zone at the northern termination of MOF (location in Figure 3a). Schmidt net (lower hemisphere) shows, in orange and black, N-S-striking and E-W-striking minor faults, respectively. (c) Outcrop of minor strike-slip faults within the damage zone (location in Figure 3a). Red polygons and black arrows indicate fault surfaces and striae attitude, respectively. Schmidt net (lower hemisphere) shows, in orange and red, left- and right-lateral strike-slip faults, respectively. (d) Geological cross-section CC′, along which we calculated the cumulative throw of TF (∼330 m; Table 1) and MOF (∼500 m; Table 1).

To evaluate the potential linkage with adjacent faults in the field, we carefully analyzed the NW and SE tips of MOF. The area between MOF and TF (Figure 3) is characterized by intensely fractured rocks and abundant secondary faults. Fault planes show patches of cataclasite and fault breccias, suggesting fault activity at different depths. Throws accommodated by subsidiary faults could not be assessed due to the occurrence of heavily damaged rocks and lack of exposed reference horizons, except for the throw (∼200 m; Figure 3d) accommodated by the E-striking fault, which is labeled as Accommodation Fault (AF) in Figure 3a. Faults in this zone display a horsetail geometry (Kim et al., 2004; McGrath & Davison, 1995) radiating from the tip of MOF (Figures 3a and S1). In the area between MOF and MCF (Figures 4 and S1), which is characterized by less frequent secondary faults than those observed between MOF and TF (Figures 4a and 4b), bedding is re-oriented from a regional NW-striking NE-dipping attitude to a local NE-striking SE-dipping attitude, hence forming a local relay ramp (Figures 4a and 4c). Unfortunately, between MCF and OPF, only small outcrops of fractured and brecciated bedrocks as well as secondary faults with a few kinematic indicators occur. Such a poor exposure did not allow us to infer the linkage nature between these two faults.

Details are in the caption following the image

Focus on the area between Monte Orsello (MOF) and Monte Cefalone (MCF) faults (see location in Figure 2a; see also Figures S1 and S2). (a) Geological map with track of geological cross-section HH'. For stratigraphy and legend, the reader is referred to Figure 2a. (b) Outcrop of the stratigraphic boundary between the bauxitic level (IBX) in the bottom and Upper Cretaceous limestones on the top (location in Figure 4a). (c) Satellite lateral view (Google Earth) of the area shown in panel (a). Red lines indicate fault traces. (d) Geological cross-section HH′, along which we calculated the cumulative throw of MOF (∼670 m; Table 1) and MCF (∼80 m; Table 1).

Results from our field surveys and structural analyses of the Campo Felice faults are illustrated in the maps, cross-sections, diagrams, and photographs of Figures 2-5 and S1–S2.

4.2 Throw Versus Distance Diagrams

Cross-sections show domino SW-dipping normal faults across the studied area with bed dip domains toward NE (Figures 5 and S2). The maximum cumulated vertical component of displacements (i.e., throws) are ∼880, ∼685, and ∼1,480 m for TF, MOF, and MCF, respectively (Figure 5 and Table 1). In principle, considering the prevailing strike slip kinematic indicators measured along the TF, the throw measured along this fault could be a geometric artifact associated with out-of-section motion. However, based on its length (∼10 km; Servizio Geologico d'Italia-APAT, 2010) an exclusive strike-slip motion cannot produce a vertical throw of 880 m. Moreover, the fact that the average strike of strata at the hangingwall and footwall (i.e., ∼N160°; Figures 2, 3 and S1) parallels the average strike of the studied fault (i.e., ∼N165°; Figures 2, 3 and S1) allows us to exclude this scenario. Similarly, the average strike of bedding at the hangingwall and footwall of MOF and MCF is parallel to faults strike (Figures 2a and S1). This evidence excludes that a significant part of the vertical throw may have been accumulated during previous strike-slip phases.

Details are in the caption following the image

Geological cross-sections AA′, EE′, GG′, II′, and MM' (see also Figure S2) built across the Tornimparte, Monte Orsello, and Monte Cefalone faults in the Campo Felice area. See Figure 2 for cross-section tracks, stratigraphy, and legend. The color of arrows is: Red for faults, blue for fault displacement, and black for fault throw. The thickness of continental deposits is consistent with Giraudi et al. (2011).

Table 1. List of Cumulative Fault Throws and Associated Errors, Calculated From Cross-Sections of Figure S2
CROSS-SECTION TF throw (m) Error neg (m) Error pos (m) MOF throw (m) Error neg (m) Error pos (m) MCF throw (m) Error neg (m) Error pos (m)
A 880 −120 50 / / / / / /
B 570 −40 330 230 −150 150 / / /
C 330 −100 5 500 −150 10 / / /
D / / / 460 −75 5 / / /
E / / / 680 −40 5 / / /
F / / / 540 −50 50 50 −40 150
G / / / 685 −70 70 / / /
H / / / 670 −170 170 80 −40 20
I / / / 100 −20 10 1,480 −350 20
L / / / / / / 1,150 −100 40
M / / / / / / 1,000 −50 100
N / / / / / / 450 −80 5
  • Note. For each value of throw, we associated a confidence interval: Positive error (pos.) defining the possible maximum throw and negative error (neg.) defining the possible minimum throw. TF is Tornimparte Fault, MOF is Monte Orsello Fault, and MCF is Monte Cefalone Fault in Figures 2 and S1.

In Table 1, we reported all cumulative geological throws and related errors (see the methods section). The throws are plotted against distance along the fault strike in Figure 6. Moving from northwest to southeast, we observe that the TF throw reduces non-linearly toward the fault tip, where the throw gradient is steeper. The throw of MOF has a bell shape typical of most (normal) faults with maximum throw (see cross-sections EE' and GG' in Figure 5) in the central part of the fault, reducing along strike toward the tips, where the gradient is steeper (Figure 6). This is more evident at the southeastern tip than at the opposite tip (Figure 6). It should be considered that the throw reduction (∼200 m) along MOF observed between cross-sections EE′ and DD′ may have been accommodated by the AF fault (Figure 3), whose eastward continuation is hidden by Quaternary sediments. The throw distribution of MCF has a more pronounced antiformal shape than that observed for MOF, with higher values of throw, which is maximum in the central part of the fault and decreases toward the tips (Figure 6). Approaching the northwestern tip, the throw is very low (<50 m) and the throw gradient is steeper than that observed for the opposite tip (Figure 6).

Details are in the caption following the image

Fault throw profiles along the Tornimparte, Monte Orsello, and Monte Cefalone faults (above) and schematic map of these faults (below). In the map, intersections between cross-sections and faults are indicated with the labels of cross-sections (AA′, BB′, CC′, etc.).

4.3 Seismic Potential

In this section, most of the above-presented data and results are used and synthesized to infer the seismic potential of the Campo Felice normal faults. This is evaluated through empirical relationships between rupture length and earthquake magnitude, assuming that the exposed fault lengths correspond to the maximum future coseismic surface rupture lengths (e.g., Alvarado et al., 2014; Iezzi et al., 2019; Mignan et al., 2015; Nicol et al., 2020; Tondi et al., 2020; Trippetta et al., 2019). We apply Equations 1-3 to our measurements of the exposed Campo Felice fault lengths (MCF and MOF) to obtain the related maximum potential magnitude of earthquakes (Table 2). In particular, we assume both an independent motion of the two studied faults, using the exposed single lengths of MOF and MCF, and a synchronous motion of the two faults, using their cumulative exposed length calculated from the northwestern tip of MOF to the southeastern tip of MCF (Table 2). The resulting magnitude is comprised between Mw 5.7 and 6.4 (Table 2). In detail, Equation 1 provides Mw values of ∼6.0 for MOF, ∼6.1 for MCF, and ∼6.4 for synchronous motions of MOF and MCF, whereas Equations 2 and 3 provide Mw values of ∼5.7 for MOF, ∼5.9 for MCF, and ∼6.2 for synchronous motions of MOF and MCF (Table 2). These seismic potentials are further treated in the discussion section.

Table 2. List of Fault Lengths (L), Maximum Potential Earthquake Magnitude (Mw) Obtained From Known Empirical Equations 1-3, Compared With the Estimation From the New Equation 7, Discussed Below
Fault name L (km) Mw (1) Mw (2) Mw (3) Mw (7) W (km) d (km) Dav (m)
Monte Cefalone (MCF) 9 6.12 5.95 5.85 5.90 7.5 6 0.3
Monte Orsello (MOF) 7 5.97 5.79 5.68 5.75 6 5 0.3
Monte Cefalone-Monte Orsello (MCF-MOF) 15 6.41 6.27 6.19 6.22 10.5 8 0.5
  • Note. Fault width (W) was obtained by using Equation 4, depth of bottomset fault (d) was obtained through W (urn:x-wiley:02787407:media:tect21579:tect21579-math-0010), and average slip (Dav) derived from Equation 5.

4.4 Coulomb Stress Change Modeling

Coulomb stress change modeling was performed to investigate the post-seismic stress transfer between adjacent fault segments. Applying this modeling, we assumed the measured fault length to be equal to the maximum seismic surface rupture length. Geometry (fault length, L, and attitude) and kinematics of faults are retrieved from our field survey. Other input data, such as the downdip rupture length and coseismic average slip (i.e., W and Dav in Table 2), were calculated using the well-established empirical Equations 4 and 5 by Leonard (2010). The average coseismic slip (Dav) obtained for the studied faults is comprised between 0.3 and 0.5 m (Table 2). These results are consistent with the average slip that occurred during recent earthquakes with rupture lengths of about 10–15 km in the central Apennines (e.g., 0.3 m for the Mw 6.0 Colfiorito 1997 by Hunstad et al., 1999; 0.6 m for the Mw 6.3 L'Aquila 2009 earthquake by Cheloni et al., 2010; 0.4–0.6 m for the Mw 6.5 Norcia 2016 earthquake by Scognamiglio et al., 2018).

We assume a common depth of ∼5 km for the nucleation of earthquakes, consistent with the depth of the hypocenters of recent seismic events in the central Apennines (i.e., depths from 3 to 10 km; Chiaraluce et al., 2003; Cheloni et al., 2010; Galadini et al., 2018).

We performed Coulomb stress change models for both independent and synchronous coseismic slip along MOF and MCF, as summarized in Figure 7, which shows how these ruptures can cause a positive change of Coulomb stress on nearby optimally oriented receiver faults.

Details are in the caption following the image

Models of Coulomb stress change induced by supposed earthquakes at ∼5 km of depth and with an average coseismic slip on: (a) Monte Orsello fault (MOF) by 0.3 m of slip; (b) Monte Cefalone fault (MCF) by 0.3 m of slip; (c) MOF and MCF simultaneously by 0.5 m of slip (Table 2).

The model of coseismic rupture along MOF (L = 7 km, W = 6 km, and Dav = 0.3 m; Figure 7a) shows an area with positive Coulomb stress change (∼3 bar) at seismogenic depth on MCF and on the southern part of TF. The coseismic rupture modeled over MCF (L = 9 km, W = 7.5 km, and Dav = 0.3 m; Figure 7b) produces a positive Coulomb stress change (∼3 bar) at seismogenic depth over MOF and with minor intensity (∼1 bar), on the northern part of OPF. Finally, modeling synchronous coseismic ruptures along MOF and MCF (L = 15 km, W = 11 km, Dav = 0.5 m; Figure 7c), we obtain a positive Coulomb stress change (∼3 bar) at seismogenic depth over the southern part of TF and the northern part of OPF, and with minor intensity (∼1 bar) on CCF.

5 Discussion

5.1 Throw Rate

The detailed cumulated throw versus distance diagram of the MCF defines the distribution of the throw accommodated by this fault since its activation (Figure 6). This result can be compared to the post LGM (i.e., post 27–19 ka according to Clarks et al., 2009; post 30–15.5 ka in this area according to Galli et al., 2012) throw distribution measured by Wilkinson et al., (2015) for the southeastern part of the same fault segment (Figure 8). We observed a similar trend that reaches the maximum throw value (more than 1,000 m in our study and 14 m for the post LGM) in the same sector of the fault, with a distribution of throw along the fault strike that persisted almost unchanged through time (Iezzi et al., 2019; Walker et al., 2009). Upon the assumption of post-orogenic development for MCF (i.e., there is no evidence for previous tectonic phases on MCF; Figure 2a), this comparison allows us to infer the onset of the MCF activity (Figure 8). Indeed, by considering the maximum post-LGM throw (about 14 m), we obtain a maximum throw rate of 0.9 mm/yr during the last 16 ka (Wilkinson et al., 2015) for MCF, similarly to many Apennines recent or active faults (Galli & Peronace, 2014). In particular, our evaluation is consistent with Goodall et al. (2021), where the authors assessed a long-term average throw rate for MCF of 1.15 ± 0.36 mm/yr by cosmogenic 36Cl concentration. Assuming a long-term average throw-rate of 0.9 mm/yr for MCF and considering its maximum throw of ∼1,400 m (see Figure 6 and cross-section II' in Figure 5), we obtain that MCF was activated since ∼1.5 Ma. This result is consistent with the data published by Giaccio et al. (2012) for the nearby Aterno basin (L'Aquila), formed since about 1.8 Ma. Other studies proposed earlier onset for nearby faults in the study area (Cosentino et al., 2017; Gori et al., 2017; Nocentini et al., 2017). Such an earlier onset is still consistent with our inference (about 1.5 Ma for MCF). Note also that our dating calculation (about 1.5 Ma) may be slightly overestimated because the considered throw (∼1,400 m) has a high negative error bar (i.e., −350 m; Figures 6 and 8; Table 1) due to the uncertain thickness of recent deposits of the Campo Felice basin (see question mark beneath the Campo Felice basin in the II′ cross-section in Figure 5).

Details are in the caption following the image

Diagram showing the variation of fault throw with distance along the strike of the southeastern portion of the Monte Cefalone fault (MCF). Red is for cumulative geological (probably younger than 1.5 Ma) throw assessed in this work from cross-sections in Figures 5 and S2 (Table 1), whereas purple is for post LGM (post 16 ka) cumulated throw assessed by Wilkinson et al. (2015) (see Figure 7e of their work) for the same fault (MCF). Left y-axis is for post 1.5 Ma throws whereas right y-axis is for post 16 ka throws. The two throw datasets and patterns are geologically consistent.

5.2 Linkage Between Faults

In extensional settings, lateral linkage of faults is an important mechanism controlling fault growth (Peacock, 2002). Growth by linkage produces the increase of length of possible seismic ruptures. Subsequent to a hard linkage, indeed, the segment tips may not act anymore as persistent barriers to seismic ruptures and coseismic slip may therefore propagate from one fault to its neighbor (Manighetti et al., 2007; Moustafa, 2002; Zhang et al., 1991). Moreover, empirical studies have shown that normal faulting earthquakes are capable of rupturing multiple segment barriers that can be up to 5–7 km in length along the fault strike (Iezzi et al., 2019; Wesnousky, 2008). Examples of normal faulting earthquakes that ruptured simultaneously parallel faults spaced about 5 km are the 1954 Mw 7.2–6.8 events in Fairview Peak-Dixie Valley (Nevada, USA) and the 1959 Mw 7.5 event in Hebgen Lake (Montana, USA) (DePolo et al., 1991).

The evaluation of the stage of linkage between outcropping segments of seismogenic faults can be therefore important to assess future rupture scenarios also at hypocentral depths (Iezzi et al., 2019; Manighetti et al., 2007; Zhang et al., 1991). This can be achieved by integrating field observations (i.e., detailed fault mapping as wells as damage of rocks and kinematic analysis) with the throw versus distance diagrams shown, for instance, in Figure 6 (e.g., Moustafa, 2002; Spina et al., 20082009; Trudgill & Cartwright, 1994). We applied this approach to the studied normal Campo Felice faults (Figure 2).

MOF and MCF (Figure 2a) are characterized by pure dip-slip kinematics on SW-dipping main fault segments, coherently with both the present-day NE-SW oriented sub-horizontal minimum principal stress (e.g., Anzidei et al., 2009; D'Agostino et al., 2001; Devoti et al., 2010; Doglioni et al., 2015; Montone et al., 2004) and the focal mechanisms of recent earthquakes (e.g., Scognamiglio et al., 2010). In the relay zone between MOF and MCF, we observed only the presence of short faults (length < 100 m in map view, Figure 4a). In this area, a re-orientation of bedding occurs, defining a relay ramp (Figures 4a and 4c). The throw versus distance diagram is characterized by a maximum in the central part of both faults, with the throw decreasing moving along strike toward the fault tips (Figure 6). In Figure 6, we also observed different maximum throws accumulated by the Campo Felice faults. In particular, the cumulative throw of MCF is larger than that of MOF (although part of this difference could be due to the large negative error bar associated with the MCF datum). This is probably due to the fact that MCF, besides being longer than MOF, became active before MOF. The late linkage and common slip with MOF can explain and justify the different throw. Unfortunately, available geological data do not allow the recognition of Holocene or Pleistocene stratigraphic markers nearby the Campo Felice faults, so to reconstruct in detail their throw evolution. New geophysical or subsurface geological data are needed to unravel this issue. In conclusion, both the steep throw gradient of MOF and MCF in their overstepping zone and the increase of their cumulative throw in the same area (Figure 6) point to a linkage between these faults (Ferrill & Morris, 2001; Peacock & Sanderson, 19911994; Walsh & Watterson, 1991; Walsh et al., 1999). It is worth noting that the mechanical and geometrical interaction between fault segments observed in the field do not necessarily reflect the interaction between segments at seismogenic depths. Indeed, different evolutionary stages (from soft-linkage accommodated by relay ramp to hard-linkage accommodated by transfer faults) can develop through time and space, including depth (Fossen & Rotevatn, 2016; Peacock & Sanderson, 1994). A common observation is that the linkage increases with depth (e.g., Fossen & Rotevatn, 2016). Therefore, due to the assessed surface-shallow soft linkage between MOF and MCF, a hard-linkage between MOF and MCF is very likely to occur at seismogenic depth, forming a single seismogenic fault segment (e.g., Spina et al., 2008; Wesnousky, 2008). Based on this mechanical interaction between MOF and MCF and the extensional kinematic indicators measured along these faults (coherent with present-day tectonic setting; Figure 2), we will evaluate seismic potential and seismic scenarios for the Campo Felice area, assuming both an independent and a synchronous future activation of these two faults (Figure 7 and Table 2).

We have also investigated the northwestern tip of MOF (Figures 2 and 3) to understand its possible role during an earthquake. In this area, a ∼1.5-km left lateral step with the TF occurs. In the overlapping area, minor faults occur, showing a horsetail like geometry in map view, various orientations, and both dip- and strike-slip kinematics (Figure 3). The oblique slip on NE-SW-striking faults is coherent with the present NE-SW-oriented extension in the Apennines and with dip-slip kinematics on the main active SW-dipping faults. Conversely, oblique slip on SW-dipping faults and strike-slip kinematics on N-S-striking faults are poorly or not coherent with the present tectonic regime. Since there is no clear and systematic evidence of cross-cutting or abutting relationships between the different fault sets, an alternative explanation for this complex kinematics other than a pre-Late-Pliocene motion is the local re-orientation of stress field occurring at fault tips (Fossen & Rotevatn, 2016; Peacock & Sanderson, 1994). The very steep throw gradient of MOF and TF near their overstepping tips (Figure 6), the well-oriented strike of the TF in the present stress field, and the occurrence of minor faults with horsetail geometry linking MOF and TF suggest a shallow-surface soft-linkage with a breached relay-ramp between these faults. In this scenario, the AF fault may have transferred part of the throw (∼200 m; Figure 3d) from MOF to TF (Ferrill & Morris, 2001; McClay & Khalil, 1998; Peacock & Sanderson, 19911994). The reactivation of previous strike-slip faults as extensional faults is long known in the Apennines (Cello et al., 1997; Pace et al., 2002; Tondi & Cello, 2003). In such a context, the heavily fractured/damaged rocks at the northwestern tip of MOF can be explained not only with a tip damage zone (Kim et al., 2004; Peacock et al., 2017), but also with a strike-slip to extensional reactivation with fast dynamic shear propagation of coseismic rupture (Reches & Dewers, 2005) toward the TF. Unfortunately, due to its bad exposure, no paleoseismological studies were performed on the TF to exclude or support its reactivation as an extensional fault. Consequently, we will not assume the activation of TF in our models of future coseismic scenarios. Further studies are required to shed light on the role in the past and possibly in the future of the TF.

5.3 A New Magnitude-Length Empirical Equation

We previously used empirical equations from the literature (Table 2) to assess the seismic potential of Campo Felice faults. To understand which of these equations are the best for the central Apennines and Campo Felice area, we compared the magnitude deduced from seismic analyses with that predicted by Equations 1-3 for the last seven Mw > 5.5 earthquakes (1997–2016) from the central Apennines (Figure 9; Table S1). These earthquakes (Figure 1a) are: Colfiorito, September 26, 1997 (Mw 5.7 and 6; Chiaraluce et al., 2003), Sellano, October 14, 1997 (Mw 5.6; Chiaraluce et al., 2003), L'Aquila, April 6, 2009 (Mw 6.3; Guerrieri et al., 2010), Amatrice, August 24, 2016 (Mw 6.0; Galadini et al., 2018), Visso, October 26, 2016 (Mw 5.9; Galadini et al., 2018), and Norcia, October 30, 2016 (Mw 6.5; Galadini et al., 2018).

Details are in the caption following the image

Empirical equations used in this work to calculate the potential maximum moment magnitude (Mw) from exposed lengths (L) of Monte Orsello (MOF) and Monte Cefalone (MCF) faults as well as from the two faults together (see Table 2). Blue line is for Equation 1 by Wells and Coppersmith (1994), pink line is for Equation 2 by Galli et al. (2008), green line is for Equation 3 by Leonard (2010), and red line is for our new Equation 7 with standard deviation of 0.16 (dotted red lines). Stars are Mw assessed with Equations 1-3, and 7 from the exposed length of MOF, MCF, and their sum. Red dots are the last recent seven earthquakes that occurred in central Apennines with Mw > 5.5, namely the 2016 Norcia Mw 6.5, 2016 Visso Mw 5.9, 2016 Amatrice Mw 6.0, 2009 L'Aquila Mw 6.3, 1997 Colfiorito Mw 5.6, 1997 Colfiorito Mw 6, and 1997 Colfiorito Mw 5.7 earthquakes (Table S1).

Although L used in Equations 1-3 is the surface rupture length, the subsurface rupture lengths for the last seven Mw > 5.5 central Apennines earthquakes as obtained by inverting different types of datasets from previous literature (e.g., earthquake locations from Chiaraluce et al., 2003, InSAR from Galadini et al., 2018 and Guerrieri et al., 2010, and GPS data from Galadini et al., 2018; Table S2). The use of subsurface rupture lengths inferred from geodetic and geophysical interpretations for recent earthquakes in central Apennine is compelled by the fact that coseismic surface ruptures, strongly controlled by rock rheology (Carminati et al., 2020; Tondi et al., 2020), were not continuous in the field (Basili et al., 1998; EMERGEO Working Group, 2010; Falcucci et al., 2018; Pucci et al., 2017). However, in our view, rupture lengths constrained using geophysical data are best comparable with total lengths of exhumed faults, which are the results of tens to thousands of events. In addition, Trippetta et al. (2019) showed that the difference between evaluations of Mw obtained by using empirical equations for surface rupture lengths and for subsurface rupture lengths of Leonard (2010) and Wells and Coppersmith (1994) is less than 5% for major faults (with length > 10 km and earthquake Mw ≥ 5). This difference increases to about 15% for smaller faults (i.e., for fault length ≤ 2 km), corresponding to earthquake Mw less than about 4.8. Therefore, the difference in maximum potential Mw obtained by the empirical equations of Leonard (2010) and Wells and Coppersmith (1994) using surface and subsurface rupture lengths can be often negligible (Trippetta et al., 2019).

The chi-squared Χ2 statistical test (Mann & Wald, 1942; Table S2) shows that Equation 1, except for the case of the Mw 6.5 Norcia 2016 earthquake (see 1 in Figure 9), tends to overestimate recent instrumental data (Χ2 = 0.14%), whereas Equations 2 and 3 (Χ2 = 96% and Χ2 = 95%, respectively) mirror the data trend. Consequently, we recommend the use of Equations 2 and 3 to assess the seismic potential of active faults in the Apennines area for 5.5 < Mw ≤ 6.5 and for rupture lengths between about 6 and 21 km (Figure 9).

In addition, we used the aforementioned last seven Mw > 5.5 earthquakes from the central Apennines to build a new ad-hoc empirical equation (red line in Figure 9). The obtained empirical equation, characterized by a standard deviation of 0.16, is:
urn:x-wiley:02787407:media:tect21579:tect21579-math-0011(7)
where L is the subsurface rupture length.

Using Equation 7, we obtain Mw values of ∼5.9 for MOF, ∼5.8 for MCF, and ∼6.2 assuming synchronous motions of MOF and MCF (Table 2). These values are consistent with those calculated with Equations 2 and 3 (e.g., Galli et al., 2008 proposed an expected earthquake magnitude of ∼6.3 for MOF + MCF), but underestimate those (Mw 6.2 and Mw 6.5 for segments of MCF) inferred for paleoearthquakes by Benedetti et al. (2013) probably because the authors did not consider the post-seismic erosion of fault surface (Goodall et al., 2021; Kastelic et al., 2017). Moreover, the magnitudes proposed by Benedetti et al. (2013) were obtained measuring coseismic slip through 36Cl concentrations in limestones forming the MCF fault scarp (see also Schlagenhauf et al., 2011) and as pointed out by the authors themselves, the 36Cl method can slightly overestimate the seismic potential because single coseismic slips under about a 100 years' time interval cannot be resolved.

5.4 Possible Seismic Scenarios

We use Coulomb stress change models to verify whether coseismic slip on the investigated faults could induce a quasi-synchronous or cascade break on adjacent fault segments, thus increasing the released energy and raising regional shaking, damage, and casualties.

The maximum Coulomb stress change predicted by our modeling is ∼3 bar (Figure 7). This is coherent with previous real instances in the central Apennines. For example, several aftershocks of the L'Aquila 2009 seismic sequence (central Italy) occurred where the mainshock increased the Coulomb stress by 1 bar (Falcucci et al., 2011). During the seismic sequence of Amatrice 2016 (central Italy), the event of Visso (Mw = 5.9) and Norcia (Mw = 6.5) occurred where the previous Mw 6.0 Amatrice earthquake had caused a Coulomb stress increase of 0.35 and 2–3 bar, respectively (Galderisi & Galli, 2020; Tung & Masterlark, 2018). In addition to the case of the central Apennines, during the 1992 Landers (California, USA) seismic sequence, aftershocks mostly occurred where Coulomb stress had been increased by 0.5–1 bar due to the mainshock (King et al., 1994). Moreover, the largest aftershock occurred close to the area where the highest Coulomb stress increase had occurred (3 bar). During the 2016 Kaikoura (New Zealand) seismic sequence, the progressive seismic rupture evolved where the Coulomb stress had been increased by about 2 bar (Xu et al., 2018).

Based on our Coulomb stress modeling, we suggest that a singular coseismic rupture on MOF (Figure 7a) or MCF (Figure 7b) can trigger MCF or MOF, respectively. Moreover, owing to the potential hard-linkage at seismogenic depth between these two faults, we also modeled the synchronous rupture of MOF and MCF (Figure 7c), which resulted in a positive stress change on OPF and TF. Consequently, even if there is no evidence in the literature for a Quaternary activity of TF, Figure 7 shows that this fault could be loaded and hence potentially activated by synchronous seismic slip on MOF and MCF.

In previous studies (Galadini & Galli, 2000; Galli et al., 2008; Giraudi, 1995; Pantosti et al., 1996; Salvi et al., 2003), the MCF fault was considered as a part of the ∼35 km long NW-trending OPF – CCF fault system, which possibly generated earthquakes with Mw as high as 6.7 (Galli, 2020). Our modeling results (Figure 7b) highlight another potential scenario, since an earthquake along MCF could activate MOF or OPF more likely than CCF. However, being CCF very close to MCF (about 2 km), we believe that dynamic stress loading (i.e., the stress increases due to the passage of seismic waves; e.g., Freed, 2005; Kanamori & Brodsky, 2004) from an earthquake on MCF could activate CCF too. In summary, an earthquake (coseismic slip ≥ 0.3 m) over MCF could trigger OPF and CCF, as previously proposed (Galli et al., 2008; Pantosti et al., 1996; Salvi et al., 2003), but also MOF and (eventually) TF, as shown in this work (Figure 7). Similar processes (i.e., single or multisegment fault activations) were proposed for the nearby Norcia Fault (Galli et al., 2020). In particular, their research showed that the Norcia Fault could be seismically activated not only in conjunction with the Monte Vettore Fault, but also with the Cascia and Monte Alvagnano Faults.

6 Conclusions

In the last 25 years, central Apennines faults caused hundreds of victims during Mw ≥ 6.0 earthquakes, making their geological investigation and seismic potential evaluation primary goals for the scientific community. We focused our studies on the two Campo Felice faults, which are aligned between other seismogenic sources of the central Apennines and may represent a hitherto silent seismic connection between the catastrophic 1915, Mw 7 Avezzano and the destructive 2009, Mw 6.3 L'Aquila earthquakes (Figure 1). Results of our field surveys, geological-structural mapping, serial cross-sections, and throw versus distance diagrams highlight a shallow soft linkage between the two Campo Felice faults. Based on this mechanical interaction, we evaluate the maximum potential magnitude (Mw) of ∼5.8 and 6.2 assuming, respectively, an independent or synchronous activation of these faults. Furthermore, our Coulomb stress change modeling shows that these magnitude assessments could potentially increase considering a quasi-synchronous or cascade activation of the Campo Felice faults together with other conterminous faults. This study shows that, in addition to paleoseismological analyses as well as geophysical investigations, geological mapping and field structural analysis provide significant evidence to draw future seismic scenarios aimed at mitigating the earthquake threat. We emphasize two main conclusions: (a) The degree of linkage between adjacent segments of faults can be assessed through field studies and geological cross-sectioning, where this potential linkage becomes fundamental to correctly assess the seismic potential of fault systems; (b) empirical relationships between fault size and earthquake magnitude are a powerful tool to assess the seismic potential of faults and fault systems, but these relationships should be tested, where possible, against recent local earthquake data. Concerning this latter point, for the central Apennines, the empirical relationships by Galli et al. (2008) and Leonard (2010) and this work (Equation 7) appear as the most reliable and therefore, we recommend their use to estimate the seismic potential of active faults.

Acknowledgments

The authors thank Anna Maria Lombardi for her kind and constructive help with empirical equations relating rupture size and earthquake magnitude. Funding by Sapienza Progetti di Ateneo 2017 and 2019 (E. Carminati) is acknowledged. The authors thank I. Manighetti (Editor of JGR), F. Iezzi, S. Catalano, L. Jolivet (Editor of Tectonics), an anonymous associate Editor of JGR, an anonymous associate Editor of Tectonics, and two anonymous reviewers for useful suggestions that the authors largely used to improve the present version of this work. This work received the Licio Cernobori Award 2021 to Giulia Schirripa Spagnolo by the Gruppo Nazionale di Geofisica della Terra Solida GNGTS. Open Access Funding provided by Universita degli Studi di Roma La Sapienza within the CRUI-CARE Agreement.

    Data Availability Statement

    All data needed to evaluate the conclusions in the study are present in the supporting information. These data are also freely available in the Figshare external repository (https://doi.org/10.6084/m9.figshare.12102549.v5). Moreover, part of the data used to construct the diagram of Figure 7 (i.e., purple data in Figure 7) are available in Wilkinson et al. (2015).