Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Corresponding author: E. Santoro, Dipartimento di Scienze della Terra, Università di Napoli “Federico II,” Largo S. Marcellino 10, IT-80138 Naples, Italy. (firstname.lastname@example.org)
 Morphotectonic analysis and fault numeric modeling of uplifted marine terraces along the Ionian Sea coast of the Southern Apennines allowed us to place quantitative constraints on middle Pleistocene-Holocene deformation. Ten terrace orders uplifted to as much as +660 m were mapped along ~80 km of the Taranto Gulf coastline. The shorelines document both a regional and a local, fault-induced contribution to uplift. The intermingling between the two deformation sources is attested by three 10 km scale undulations superimposed on a 100 km scale northeastward tilt. The undulations spatially coincide with the trace of NW-SE striking transpressional faults that affected the coastal range during the early Pleistocene. To test whether fault activity continued to the present, we modeled the differential uplift of marine terraces as progressive elastic displacement above blind oblique-thrust ramps seated beneath the coast. Through an iterative and mathematically based procedure, we defined the best geometric and kinematic fault parameters as well as the number and position of fault segments. Fault numerical models predict two fault-propagation folds cored by blind thrusts with slip rates ranging from 0.5 to 0.7 mm/yr and capable of generating an earthquake with a maximum moment magnitude of 5.9–6.3. Notably, we find that the locus of predominant activity has repeatedly shifted between the two fault systems during time and that slip rates on each fault have temporally changed. It is not clear if the active deformation is seismogenic or dominated by aseismic creep; however, the modeled faults are embedded in an offshore transpressional belt that may have sourced historical earthquakes.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 It is widely acknowledged that morphotectonic analysis constitutes a fundamental ingredient of active tectonic studies. Particularly, displaced marine terraces are used to unravel uplift rate histories and fault-induced landscape deformations at active coastal settings [e.g., Valensise and Ward, 1991; Ward and Valensise, 1996; Cucci and Cinti, 1998; Dubar et al., 2008; Ferranti et al., 2009; Gaki-Papanastassiou et al., 2009; Caputo et al., 2010; Saillard et al., 2011]. A paleo-shoreline shape differing from the original horizontal attitude which formed at the coeval sea level reflects deformation which may have been driven by large-scale and deep-seated geodynamic sources, by local processes occurring at shallower depths (e.g., fault slip and fold growth), or by a combination of both [e.g., Ferranti et al., 2009; Caputo et al., 2010]. If, within the shoreline deformation profile, the regional and local tectonic components can be separated, the possibility is offered to better constrain the geometric and kinematics parameters of faults, which cannot otherwise be investigated with conventional paleoseismic methods.
 A suitable setting where these notions can be tested is represented by the Apennines orogen in southern Italy, where recent coastal markers are ostensibly deformed [Caputo and Bianca, 2005; Ferranti et al., 2006; Antonioli et al., 2009]. At the end of the early Pleistocene, a major tectonic change occurred in southern Italy [Hippolyte et al., 1994]. At this time, east directed motion of the Apennines frontal thrust system became locked, probably because of involvement of the foreland continental lithosphere in the Adriatic subduction zone. This occurrence caused in turn the activation of NW-SE striking transpressional faults along the Ionian Sea margin of the Apennines (Figure 1), which served to accommodate further albeit slow convergence between the Adriatic foreland and the Apennines [Catalano et al., 1993]. The transpressional structures were coeval with extensional faults in the west (Figure 1), which are linked to back-arc thinning in the Tyrrhenian Sea at the rear of the Apennines (Figure 1, Section 1) [Hippolyte et al., 1994]. Around the same time, generalized uplift of the southernmost part of the Apennines and the Calabrian Arc commenced (Figure 1, Section 1) [Westaway, 1993].
 Whereas the extensional faults are relatively well exposed and are the sources of instrumental and destructive historical earthquakes (DISS Working Group, 2010, Database of Individual Seismogenic Sources, Version 3.1.1, available from INGV, http://diss.rm.ingv.it/diss/), post-early Pleistocene deformation in the eastern side of southern Italy is, however, difficult to assess. Because the frontal sector of the Apennines experienced a recent tectonic reorganization and has relatively low deformation rates and seismic activity [e.g., Chiarabba et al., 2005; 2008; D'Agostino et al., 2008; Devoti et al., 2008; Ferranti et al., 2008], several limitations are encountered when trying to unravel the recent deformation pattern. Geomorphic markers like marine terraces, being long-lived topographic feature capable of integrating slow deformation over long periods of time, can be used as paleogeodetic indicators to solve this problem.
 Previous studies on uplifted marine terraces along the Taranto Gulf, a major embayment of the Ionian Sea (Figure 1, Sections 1 and 2), interpreted local paleo-shoreline deformations superimposed on the regional uplift signal in the central part of the Gulf as due to fault-propagation folds linked to deep contractional structures [Ferranti et al., 2009; Caputo et al., 2010]. In the last decades, the increasing density of seismometer networks has allowed detailed study of earthquakes that occur along buried contractional faults (e.g., El Asnam, Algeria, 1980, Mw = 7.3 [Yielding et al., 1981]; Coalinga, California, 1983, M = 6.5 [Stein and King, 1984]; Kettleman Hills, California, 1985, M = 6.1 [Ekström et al., 1992]; Los Angeles Basin California, 1987, M = 6.0 [Hauksson and Jones, 1989]; Spitak, Armenia, 1988, M = 6.8 [Kikuchi et al., 1993]; Chi-Chi, Taiwan, 1999, Mw = 7.5 [Yu et al., 2001]). With the exception of the Chi-Chi earthquake, these events show a progressive decrease of the dislocation from the hypocenter toward the surface, which is only characterized by small fractures or secondary faulting (e.g., extensional faults associated with the bending moment).
 The investigation of subtle superficial deformation related to these structures through a morphometric analysis may return basic but robust information on the geometry and kinematics of otherwise invisible faults. The morphometric analysis is customarily blended with fault numeric modeling in order to place constraints on geometry, kinematics, and slip rates of blind faults [Rockwell et al., 1988; Valensise and Ward, 1991; Ward and Valensise, 1996; Keller et al., 1998; Benedetti et al., 2000; Ponza et al., 2010]. The underlying idea is to iteratively compare the geometry of a deformed zone, reconstructed by means of geomorphic markers, with the spatial distribution of surface displacements evaluated through fault numeric modeling.
 In this study, we focused on the uplifted marine terraces preserved along the southern half of the Taranto Gulf between the Sibari Plain and the southern border of the San Nicola Plain, where previous work proposed the existence of deformation induced by local thrust and transpressional faults (Figure 1). The paleo-shoreline profiles reconstructed from a detailed mapping of the terraces were used to calibrate the geometric and kinematic parameters of active structures deforming the coastal area. In order to shed light on the local deformation contribution, we derived the regional signal by interpolating the uplift pattern derived from MIS 5.5 terrace elevation along the entire Gulf, from Taranto city to the Sila massif (Figure 1, Section 1). The local tectonic signal in the central part of the Gulf, along the coast of the Pollino mountain range, was interpreted by performing elastic fault numeric modeling of the vertical coseismic displacement benchmarked on the paleo-shoreline profiles, with the regional signal removed.
 The results of our analysis highlight the usefulness of an integrated morphometric and fault modeling approach to investigate subdued landscape deformations and to define geometry, kinematic, and temporal behavior of potentially active blind structures. Moreover, the unprecedented resolution of the shoreline terraces has permitted to snapshot with extreme detail how fault activity is spatially and temporally partitioned between fault segments, a vital information to comprehend the mechanism of distributed deformation.
2 Tectonic Background
2.1 Regional Setting
 The Southern Apennines formed during Cenozoic slow relative convergence between Europe and Africa and fast subduction and roll-back of the Adriatic-Ionian slab associated with back-arc stretching in the Tyrrhenian Sea [Malinverno and Ryan, 1986; Faccenna et al., 2001]. These processes caused the eastward migration, toward the Apulian sector of the Adriatic foreland, of paired contractional and extensional belts [Patacca et al., 1990; Monaco et al., 1998; Menardi-Noguera and Rea, 2000; Van Dijk et al., 2000]. Today, regional GPS velocity fields point to a reduction in the rate of slab rollback and Tyrrhenian back-arc extension with respect to geologic rates, and whether subduction is ongoing is still an open question [Oldow et al., 2002; Hollenstein et al., 2003; D'Agostino and Selvaggi, 2004; D'Agostino et al., 2008; Devoti et al., 2008]. A Wadati-Benioff zone is well defined only beneath the Calabrian arc south of the Apennines (Figure 1, Section 1) and is represented by a steep, 200 km long plane that can be followed from the Ionian Sea down to a depth of 500 km below the Tyrrhenian Sea [Selvaggi and Chiarabba, 1995; Faccenna et al., 2005, 2011; Chiarabba et al., 2008]. The deep northeastern limit of the Ionian subduction is located in northern Calabria and represents the transition between Apulian continental crust and attenuated or oceanic Ionian crust (Figure 1, Section 1) [Catalano et al., 2001].
 Motion of the thin-skinned thrust front in the Southern Apennines reportedly ceased during early Pleistocene as the front was buried beneath the western margin of the Bradano foredeep basin (Figure 1) [Patacca and Scandone, 2001]. In the meantime, involvement of thick Apulian continental crust in the Southern Apennines subduction and the consequent change in local convergence direction caused the activation of transpressional faults breaking through the previously emplaced thin-skinned allochthon (Figure 1) [Knott and Turco, 1991; Catalano et al., 1993; Monaco et al., 1998; Van Dijk et al., 2000; Tansi et al., 2007]. Motion of these faults substantially occurred during the early Pleistocene. On the other hand, morphological analyses have identified that shortening in the frontal zone occurred during more recent time as well [Caputo and Bianca, 2005; Caputo et al., 2010], and may still be ongoing [Ferranti et al., 2009].
 Current deformation in southern Italy, as indicated by seismicity and geodesy, is however dominated by an extensional fault array along the Apennines axial zone (Figure 1). The extension direction determined by fault slip analyses [Hippolyte et al., 1994; Monaco and Tortorici, 2000; Maschio et al., 2005; Papanikolaou and Roberts, 2007], focal mechanisms of crustal earthquakes (Centroid Moment Tensor (CMT) and Regional Centroid Moment Tensor (RCMT) Catalogues) and GPS geodetic velocities [Hunstad et al., 2003; Ferranti et al., 2008] trends ~NE-SW in the Apennines and ~E-W in northern Calabria. The frontal belt of the southern Apennines, together with the Calabrian Arc, has experienced sustained uplift (~1 mm/yr) since the middle Pleistocene that formed flights of marine terraces hundred of meters above the present sea level [e.g., Westaway, 1993; Miyauchi et al., 1994; Cucci and Cinti, 1998; Westaway and Bridgland, 2007; Ferranti et al., 2009; Santoro et al., 2009; Caputo et al., 2010]. Uplift is viewed as an isostatic [Westaway, 1993; Wortel and Spakman, 2000; Gvirtzman and Nur, 2001] or dynamic response to removal of a high-density deep root, or as a dynamic response to mantle upwelling processes [Faccenna et al., 2011].
 A local tectonic component is also embedded in the uplift budget. Small amplitude and wavelength undulations in paleo-shorelines document the superposition of local, fault-induced deformations on a general northeastward tilt of northern Calabria [Cucci and Cinti, 1998; Ferranti et al., 2009; Caputo et al., 2010].
2.2 Tectonic Frame of the Ionian Margin of the Southern Apennines
 The mountain belt in northern Calabria and southern Lucania is floored by different tectonostratigraphic units (Figure 1), which were stacked since the Oligocene-Miocene along a roughly NW-SE trending thrust system [Knott, 1987; Cello et al., 1989]. The higher units are represented by remnants of an accretionary wedge and their satellite basin cover related to subduction of the Neotethys Ocean (Figure 1, Liguride and Sicilide complexes). The lower unit is a deformed stack of Mesozoic basinal rocks and platform carbonates, the latter forming the backbone of the Pollino Range (Figure 1). After late Miocene, all these units were thrust over the Apulian foreland platform, which was itself involved in thick-skinned thrusting [Menardi-Noguera and Rea, 2000]. The youngest units in the investigated area are represented by Pliocene-Middle Pleistocene fan-delta sediments [Colella, 1988] filling the Sibari Plain, marine clayey terrigenous deposits (Argille Subapennine Formation), extensively outcropping in San Nicola Plain and laterally correlative of the Bradano foredeep infill, and middle Pleistocene-Holocene coastal and continental deposits (Figure 1).
 The study area lies between the frontal thrust of the belt, buried beneath the Bradano foredeep, and the chain axial sector affected by the active normal faulting (Figure 1). Along the western border of the Sibari Plain, the extensional belt is represented by the Pollino-Castrovillari fault (PCF; Figure 1), a NNW striking, WSW dipping, ~15 km long normal fault, with several scarps cutting through middle Pleistocene-Holocene continental deposits [Michetti et al., 1997; Cinti et al., 2002]. Based on geomorphic and paleoseismological studies [Cinti et al., 1995, 1997], this fault is considered active; radiocarbon dating of organic material recovered in paleoseismological trenches evidenced that the two most recent earthquakes occurred in 2000–410 B.C. and 530–900 A.D. [Cinti et al., 2002].
 No active crustal normal fault is known along the Ionian Sea coast. Instead, Catalano et al.  mapped early-middle Pleistocene strike-slip faults in this sector of the Southern Apennines. Several WNW-ESE striking left-lateral strike-slip faults were recognized (Figure 1, CF, STF, PF, SRF, VF, and CNNF) that, together with ~ N-S striking, westward displacing, thrust faults, pushed Apulian carbonate wedges up through the overlaying allochthonous Liguride rocks. Based on seismic reflection profiles and borehole data (Figure 2a, Section 1), Del Ben et al.  and Ferranti et al.  suggested that the strike-slip fault zones also controlled the offshore morphostructural evolution (Figure 2a), where positive flower structures deform up the recent Pleistocene sequences. Ferranti et al. , through detailed structural analysis carried on the clayey sequences (Argille Subapennine) outcropping along the coastal area, documented that the strike-slip fault zones were activated with an ~NE to ENE-trending shortening axis (Figure 1) during the early-middle Pleistocene.
 Instrumental seismicity from the Italian Seismic Catalogue (CSI) is at low level and shows only scattered events in the region, with no recognizable hypocentral trend [Chiarabba et al., 2005; Frepoli et al., 2011]. Focal solutions of moderate (3.0 < M < 5.3) crustal strike-slip and thrust earthquakes suggest, however, that the area is still interested by transpressional deformation (Figure 1, Section 2) [Ferranti et al., 2009]. The P axis of the incremental strain tensor trends from ~NE-SW to ~E-W for both thrust and strike-slip earthquakes (Figure 1, Section 2), in agreement with the structural data from recent deposits (Figure 1).
 Similarly, the historical seismicity record of this region shows small and only occasional moderate events (CPTI11, Catalogo Parametrico dei Terremoti Italiani, version 2011) [Rovida et al., 2011] whose sources are still debated.
 The uplift process active since middle Pleistocene is well documented by marine terraces extensively found in the area, ranging in elevation from ~5 to ~650 m above sea level (asl) [Bianca and Caputo, 2003; Cucci, 2004; Cucci and Cinti, 1998; Ferranti et al., 2009; Santoro et al., 2009; Caputo et al., 2010]. Based on absolute (Electro Spin Resonance) [Santoro et al., 2009] and relative (amino acid racemization) [Cucci, 2004] dating, although constrained from only two sites, a terrace mapped between ~100 and ~130 m asl (T4 in the present paper) was correlated to the Marine Isotopic Stage (MIS) 5.5, aged at ~124 ka. By applying the average uplift rate deduced from the MIS 5.5 marine surface elevation (~1 mm/yr), the remaining terraces were indirectly attributed to the Quaternary eustatic highstands, extending the chronology back to MIS 15 (~580 ka) [Santoro et al., 2009].
 Several authors have pointed out the existence of local paleo-shoreline deformations superimposed on the regional uplift process. Cucci and Cinti  and Cucci  explained the increase in elevation of terraces along the western border of the Sibari Plain as the consequence of coseismic footwall uplift of the Pollino-Castrovillari normal fault. Based on geomorphic, structural, and geophysical data, Ferranti et al.  argued that the paleo-shoreline undulations along the Pollino Range were related to middle Pleistocene-Holocene activity of blind thrusts linked to transpressional shear zones. Finally, Caputo et al.  interpreted the increase of uplift and tilting rate moving southward from the Apulia foreland toward the Pollino Range as due to the combination of re-activated out-of-sequence blind thrusts of the Valsinni fault (Figure 1) and deep-crustal shortening.
 The terraces are locally affected by deep-seated slope gravitational deformation (Figure 1, DSSGD) triggered by differential uplift and eastward coastal tilting [Bentivenga et al., 2004; Ferranti et al., 2009; Caputo et al., 2010]. These gravitational phenomena formed several not-eustatic terraces that have contributed to foster different interpretations, within published papers, regarding the number of terrace orders, chronological attribution, and uplift rates.
3 Uplifted Marine Terraces
3.1 Terrace Mapping and Paleo-Shorelines Reconstruction
 Marine terraces were mapped along the coast between the Sibari and San Nicola Plains, which bound the Pollino Range to the south and north, respectively. The observations discussed here build on, and refine, the works published by Ferranti et al.  and Santoro et al. , by slightly adjusting their marine terrace positions based on additional observations, and by extending their terrace map to the southern sector of the San Nicola Plain. In the overlapping area, some terraces position and lateral correlation errors were identified and corrected during the new field analysis presented here.
 Ten terrace orders, labeled from T0 to T9 (lowest to highest), were distinguished between ~5 and ~660 m asl (Figure 2a). Higher remnants were noticed, but due to the very scattered distribution and difficulty in correlation, they were discarded from further analysis.
 A lateral correlation of individual terrace remnants was attempted through aereophotography analysis (1:17.000 scale) and field surveys (1:10.000 and 1:5.000 scale topographic maps and orthophotos, respectively). Geomorphic correlations (Table 1) were performed considering the altimetric distribution of inner margins (intersection line between abrasion platform and landward sea cliff), the abrasion platform width (planimetric distance between inner and outer margins) and slope, and the platform-cliff ratio (ratio between width of the terrace and height of the on-looking cliff) [Verma, 1973]. Data listed in Table 1 are corrected for the younger colluvial cover on the terraces.
Table 1. Main Morphologic and Sedimentary Parameters of the T0-T9 Marine Terracesa
Inner Margin Altimetric Range (m)
Average Width (m)
Average Slope (degree)
Maximum Deposit Thickness (m)
Altimetric Range Between Successive Marine Terrace Orders (m) c
Coastal sectors where higher uplift rates are suggested by an increase in inner margin elevations and in altimetric difference between successive marine terraces are in bold. PSN = San Nicola Plain; PLL = Pollino Range; SB = Sibari Plain; Tot = Total study area.
 The morphologic parameters of terraces vary along the surveyed coastal sector (Table 1 and Figure 2a). Typically, the terraces are characterized by an increase in width and platform-cliff ratio and by a decrease in platform slope moving southward and northward from the Pollino Range toward the Sibari and San Nicola Plains, respectively. This morphologic heterogeneity chiefly arises from a laterally uneven coastal morphology and from the different mechanical properties of the outcropping lithologies [Amato et al., 1997]. Indeed, along the Sibari and San Nicola coastal stretches, the abrasion processes active during highstands of protracted duration developed more extensive platforms due to the easily erodible Pliocene-Pleistocene clay and sands underlying the marine terraces. On the other hand, along the Pollino slope, the steeper coast and the outcrop of more resistant Miocene and Mesozoic rocks of the accretionary wedge (Figure 1) inhibited the modeling of wide platforms.
 In addition, as a consequence of the lower uplift rates in the San Nicola Plain (see the northward decrease in inner margin elevations in Table 1), the increase in platform width and the lesser number of uplifted marine terraces (Figure 2b and Table 1) point to the re-occupation of a platform during successive interglacial transgressions. Particularly, proceeding northward, T2 and T4 lose a clear geomorphologic expression and are reduced to subtle morphologic steps on the surface of higher terraces (T5 and T3, respectively; Figure 2b).
 Correlation accuracy between terrace remnants was enhanced through a detailed sedimentological analysis of the terraced deposit, each generally characterized by a transgressive-regressive cycle. Correlation between separate exposures of the coastal deposit hosted by the terraces, which in fortunate cases are found along river valley sides, was used to construct semi-quantitative cross section of the depositional sequences and to define the sedimentary parameters characterizing each terrace order (for a full facies analysis of the deposit, see Santoro et al. ).
 Sedimentological analysis of the terraced deposits was instrumental to determine the position of the terrace inner margin and thus the paleo-shoreline elevation. Commonly, inner margins are considered as the best indicator of the highest position reached by the sea during the eustatic highstand which modeled the terrace. In this work, we considered the highest elevation reached by foreshore deposits in the coastal depositional sequence as the maximum highstand sea level. However, the paleo-sea level position can be modified or obscured by post-emergence erosive and depositional (colluvial and/or alluvial) processes. For each terrace, the estimation of continental cover thickness, as well as the occurrence and the amount of erosion of the primary coastal deposit, was estimated on a site-by-site basis, using natural cuts or boreholes (details in Santoro et al. ).
 Along some parts of the coast, the development of deep-seated slope gravitational deformations (Figure 1, DSSGDs) acts against uplift and obscures a truthful reconstruction of the paleo-shoreline pattern [Ferranti et al., 2009]. For this reason, terraces related to the DSSGDs were discarded in the paleo-shoreline reconstruction.
 In Figure 3, data points indicate sites where inner margin elevation was measured, with error bars representing the estimated uncertainty derived from a combination of accuracy in marker identification and precision in measurement. The maps used to trace markers have 10 m and locally 5 m contours. Thus, uncertainty in elevation estimates is probably within ±5 m. GPS and altimeters used to retrieve spot locations of inner margins and of other markers have a similar ±5 m uncertainty. The largest error in nominal paleo-shoreline elevation stems from the choice of the appropriate bathymetric correction for different markers (constructional, erosional), which must be evaluated at individual locations. Assigned paleo-bathymetric uncertainty is ±2 m, where a complete reconstruction of the foreshore depositional sequence associated to individual terraces was possible; ±5 m, where coastal sequences are partially exposed; ±10 m, where the terraced deposits are completely eroded but the shoreline angle can be derived from the platform projection; and +20 m where no relation between platforms and their shoreline angle can be established. Error bars attached to data points incorporate both measurement errors and paleo-bathymetric uncertainty.
 The paleo-shoreline elevations have been projected along section A-B parallel to the present and ancient coastline (Figure 2a). As noted previously [Ferranti et al., 2009; Santoro et al., 2009], the paleo-shorelines show the superposition of local deformations on the regional uplift process. Three major undulations of maximum ~20 km wavelength and up to ~200 m amplitude are localized along the northern and southern flanks of the Pollino Range and beneath the Sibari Plain (Figures 2a and 3, labeled as 1, 2, and 3, respectively). As shown in Figure 3, high densities of observation points in the central and northern sector are available and provide a tight constrain on the deformed shape of the coastline. Higher position uncertainties exist in the southern sector due to the erosional removal of older terraces and to the cover of lower orders by late Quaternary continental deposits filling the Sibari Plain.
3.2 Terrace Chronology
 The terrace chronology presented in Table 2 is based on a critical review of all the available dating in the Taranto Gulf. Along the southern part of the Gulf, in the coastal area extending from the San Nicola Plain to the Sila massif, the absence of the characteristic Last Interglacial faunal assemblages (Senegalese Fauna and, particularly, the Persististrombus latus has enhanced the application of different absolute and relative dating methods in order to shed light on the terraces chronology [Brükner, 1980; Dai Pra and Hearty, 1988; Amato et al., 1997; Cucci, 2004; Zander et al., 2006; Santoro et al., 2009; Caputo et al., 2010]. Unfortunately, agreement among authors dealing with this problem is still lacking. The existing interpretation conflicts chiefly arise from the choice of a proper calibration of the racemization extent of amino acids from mollusk shells. The chronological scheme we present in this section is based on the dates we consider more reliable, as briefly summarized below. More technical aspects and our reasons for discarding other available dates are discussed in Appendix A.
Mean values of the nominal inner margin elevations along the Pollino Range and the Sibari and San Nicola plains derive from observed morphological inner margins corrected for erosive phenomena, continental cover, and paleo-bathymetry of sea level markers.
Sequential markers couplets for calculation of the time partitioned uplift rates.
 Two different aminozone schemes were proposed for the Taranto Gulf area by Hearty et al.  and Belluomini et al. . Ostensible differences exist between the two reported aminozone schemes and are one of the elements that enhanced substantially different marine terrace chronological interpretations in the region [Amato et al., 1997; Santoro et al., 2009; Caputo et al., 2010]. For the interpretation of the racemization data available in the study area, we rely on the Hearty et al.  Mediterranean aminozone scheme because of the following reasons: (1) it is based on 389 amino acid racemization analysis in the Mediterranean area, firmly calibrated against seven Cladocora caespitosa U/Th ages; and (2) it is in agreement with global isoleucine epimerization data relative to the Last Interglacial [Murray-Wallace, 1995]. Conversely, Belluomini et al.  calibrated their aminozones with U/Th ages on mollusk shells that can suffer substantial age underestimation problems due to open radiometric system condition during diagenetic processes [Kaufman et al., 1971; McLaren and Rowe, 1996]. Besides, Belluomini et al.  results disagree both with racemization data from Mediterranean sites where chronology is tightly constrained, and with global isoleucine data [Murray-Wallace, 1995].
 In the northern part of the San Nicola Plain, between Basento and Cavone Rivers, Brükner  applied U-series dating technique on two mollusk samples collected from our terrace T2 (Figure 2b, site BR80). He obtained an age of 75 ± 7 ka from the base of the terrace and one of 63 ± 3 ka at the top of the coastal regressive system. Based on these ages and on paleontological evidences, he attributed this terrace (named as T1 in Brükner ) to the MIS 5.1.
 Along the southern border of the San Nicola Plain, between the Canna and San Nicola Rivers, Dai Pra and Hearty  evaluated the leucine epimerization in seven Glycymeris samples collected at the outer margin of our terrace T3 (43 m asl; Figure 2b, site DP88). They obtained a D-alloisoleucine L-isoleucine ratio (alle/ille ratio) of 0.29 ± 0.02 which they interpreted as suggestive of a MIS 5.3/5.1 age.
 Because the age of T2 is constrained in the northern part of the plain by Brükner  U-series dating, we attribute the data from site DP88 from the overlying terrace T3 in the southern part of the plain to MIS 5.3 (Table 2).
 To the south, approaching the border of the Pollino Range, Amato et al.  estimated the alle/ille ratios on three Glycymeris specimens sampled at 102 m asl on their terrace IV (Figure 2b, site AM97-2). On the basis of an alle/ille ratio of 0.34 ± 0.01 and of geomorphological and geochronological considerations, they correlate, according to Hearty et al. , their terrace IV to the MIS 5.5. The detailed location of this sample is questioned by our field mapping. We did not find evidence of the small remnant of terrace IV that Amato et al.  mapped just a few meters above our terrace T4 (Figure 2b). Instead, we retain that the Glycymeris sp. sampled at site AM97-2 (102 m asl) comes from the uppermost onlaps of T4 beach deposits above the coeval sea cliff. Notwithstanding the small discrepancy, the dating result of Amato et al.  allows attributing terrace T4 to the MIS 5.5, in agreement with the ages discussed above for the lower terraces in the north.
 The attribution of T4 to the Last Interglacial is supported by chronological constraints in the southern Taranto Gulf. Santoro et al.  reported an ESR age of 135 ± 20 ka for a Cardium sp. collected from T4 close to Vaccarizzo Albanese village on the northern slope of the Sila massif (Figure 1, site SA09), which permitted to correlate T4 with the MIS 5.5. This age is in agreement with the work of Cucci , who evaluated the racemization (D/L ratio) in aspartic acid from four in situ Glycymeris sp. collected from terrace T4 close to Trebisacce village, on the eastern slope of the Pollino Range (Figure 2a, site CU04). He obtained a good reproducibility among the samples and a D/L ratio of 0.73 ± 0.041 with a predicted age of 130 ka. Because the aspartic acid has a faster racemization rate than leucine [Wehmiller et al., 2010], this D/L ratio cannot be interpreted based on the Hearty et al.  aminozone scheme and the predicted age was, presumably, obtained indirectly through statistical and kinetic racemization models [e.g., Wehmiller et al., 2010; Hearty, personal communication, 2012; Cucci, personal communication, 2012].
 Terraces T1, T5, T6, T7, T8, and T9 were indirectly dated by applying the mean uplift rate evaluated from T2, T3, and T4 inner margin elevations to theoretical terraces formed in correspondence with eustatic interglacials of the Waelbroeck et al.  sea level curve and Lisiecki and Raymo  δ18O curve. The mean uplift rate was evaluated at the SW border of the Sibari Plain and at the San Nicola Plain, away from the major shoreline undulations (Figure 3). Comparing these theoretical values with our paleo-shoreline elevations, we obtained the following correlations: T1 (MIS 3.5), T5 (MIS 7.1), T6 (MIS 7.5), T7 (MIS 9.5), T8 (MIS 11), and T9 (MIS 15.1). All the terraces but T1 are correlated with highstands during interglacials when the glacio-hydro-isostatic condition was very similar to the present-day Holocene transgression. As a result, we are confident about the correlation accuracy. For the terrace T1, which corresponds to a highstand during an interstadial, we acknowledge that a greater uncertainty, related to unknown glacio-hydro-isostatic effects, is embedded in our indirect dating.
 For the lowest terrace (T0), by using the mean uplift rate value along the southern Pollino coast, we found that the best age correlation was with MIS 1. In order to estimate accurate age and related paleo-sea level for T0, we used the sea level curve of Lambeck et al.  that takes into account the isostatic response of the crust to the removal of glacial masses both in the Northern Hemisphere and in the Alps and to the load of the released water column on the continental shelf in the Taranto Gulf. We considered T0 formation was possible when, during the Holocene transgression, sea level rise and uplift rates were comparable. The predicted age for T0 is ~6 ka (Table 2).
4 Quantitative Estimate of Regional and Local Source Contribution to Uplift
 A first step toward determining the fault-induced signal and derive fault parameters from the paleo-shoreline record requires splitting the total uplift of individual terraces in its regional and local tectonic component. The two components, local and regional, are characterized by different wavelengths, which vary from a few tens to 100 km, respectively, and their understanding requires assessment of different spatial distributions of markers.
 Because the regional uplift signal is driven by deep crustal or mantle sources [Westaway, 1993; Ferranti et al., 2006; 2010; Faccenna et al., 2011], its large wavelength geomorphological signature was reconstructed by using the elevation of the MIS 5.5 terrace along an ~300 km long coastal sector of the Taranto Gulf (Figure 1, Section 2; and Figure 4, Section 1).
 Specifically, we used MIS 5.5 elevation data from sectors where the paleo-shorelines are not appreciably deformed by local sources, and namely from the northernmost Taranto Gulf [Ferranti et al., 2006], the Sila Massif [Santoro et al., 2009], and the San Nicola Plain (this work) (Figure 1, Section 2; and Table 3). The elevation of the MIS 5.5 terrace in the northernmost Taranto Gulf is well constrained based on the huge amount of available geomorphology, stratigraphy, and paleontology observations [e.g., Gignoux, 1913; Gigout, 1960a, 1960b, 1960c; Cotecchia et al., 1969] and geochronology data (see Appendix A for further details) [Dai Pra and Stearns, 1977; Hearty and Dai Pra, 1985; Belluomini et al., 2002]. From the northernmost Taranto Gulf, the elevation of the MIS 5.5 terrace smoothly increases toward the Apennines to the south [Ferranti et al., 2006], suggestive of the lack of local deformation sources.
Table 3. MIS 5.5 Coastal Marker Data Along the Northern Taranto Gulf [Ferranti et al., 2006], the Southern San Nicola Plain (This Work), and the Sila Massif [Santoro et al., 2009]a
Vertical Uncertainty (m)
Shoreline Elevation (m asl)
Uplift Rate (mm/yr)
NTG = northern Taranto Gulf; SNP = San Nicola Plain; SM = Sila Massif.
Castellaneta-Ponte del Re
Boenzi et al. ; Caldara ; Dai Pra and Hearty ; Hearty and Dai Pra 
 Regarding the Sila massif, Galli et al.  recognized a local deformation of the MIS 5.5 shoreline that appears to be down dropped in the hanging wall of the Rossano-Corigliano Fault (RCF; Figure 1).. According to Molin et al. , this evidence indicates a recent reactivation of the RCF as a normal fault. In order to avoid this local deformation signal, we included in our regional analysis only the MIS 5.5 terraces preserved between Corigliano Calabro and Spezzano Albanese villages (Figure 1), outside the RCF deformation zone and where the MIS 5.5 shoreline does not show any evidence of deformation [Santoro et al., 2009]. Furthermore, as mentioned above, in this coastal sector, the MIS 5.5 coastal marker was recognized through the application of the ESR technique (Figure 1) [Santoro et al., 2009]. Finally, we discarded data from the coast between the Pollino Range and the San Nicola Plain, which is characterized by two local deformation sources, hereafter named Pollino and Valsinni anticlines (Figures 2a and 3).
 The linear regression (R2 = 0.979) through the MIS 5.5 elevation data highlights a northeastward tilt of the coast with a mean regional gradient of 0.4 m km−1 and a tilt rate (gradient of the coastline divided by the MIS 5.5 age) of 3 × 10−3 m km−1 ka−1 (Figure 4, Section 1). Cucci and Cinti , using only the terraces preserved between the Pollino Range and the San Nicola Plains, evaluated a slightly higher tilt rate (4.5–6.0 × 10−3 m km−1 ka−1). As noted above, this discrepancy likely arises from incorporation of local-source uplift in their estimation (Figures 2a and 3).
 The accuracy of our regional uplift trend was verified by plotting it against the observed T4 (MIS 5.5) paleo-shoreline elevation between the Pollino Range and the San Nicola Plain (Figure 4, Section 2). Indeed, if we force the regional curve to pass close to the tilted (but not folded) terraces of San Nicola Plain, it apparently runs very close by the inflection point between uplift and subsidence for the Pollino and Valsinni anticlines. The inflection point position for the two anticlines was derived using the elastic deformation models of Okada  that predict, for a blind thrust with a geometry similar to that of the northern Calabria faults, a coseismic surface deformation partitioned on the average between an ~84% uplift and ~16% subsidence. It is evident from Section 2 in Figure 4 that the model condition is closely matched by the combined terrace observation data and computed regional trend.
 Finally, relying on the assumption that the regional uplift is steady and uniform, we derived regional and local uplift signals for lower and higher terraces. Specifically, we used the MIS 5.5 tilt rate of 3 × 10−3 m km−1 ka−1 to evaluate the regional uplift gradient (MIS 5.5 tilt rate multiplied by Terrace Age; Figure 4, Section 3) and the local uplift component (elevation difference between the terrace and the regional uplift curve) for individual terrace orders of known or inferred age.
5 Fault Modeling
 The model built to parameterize the active structures of northern Calabria is based on standard dislocation modeling of slip on buried, rectangular-shaped faults embedded in an elastic half-space [Okada, 1985]. Fault location, geometry, dimension, and kinematics were constrained from seismic reflection data, fault kinematic analysis, focal mechanisms, and through scalar relationships [Wells and Coppersmith, 1994]. The numeric models were performed by using the Fault Studio software (a technical report of the version 1.1 is provided by R. Basili, INGV, http://www.earth-prints.org/handle/2122/1039) a MapBasic application which uses a geographical interface and was developed to calculate geometrical and kinematics parameters of seismogenic sources. The best geometric and kinematics parameters were chosen based on iterative comparison between the observed geomorphic markers deformation and displacement models output.
5.1 Fault Geometry and Kinematics Parameters
 Geometric parameters for the modeling of transpressive structures at the coast were derived from the interpretation of offshore seismic reflection profiles (Figure 2a) [Ferranti et al., 2009; Mazzella, 2011]. Because of its ~100 km length extending from the Sibari Plain to southern Apulia, and the strike almost orthogonal to the tectonic structures, the F75-89 multichannel seismic reflection profile was chosen as the most representative to investigate the target faults (Figure 5). In order to extract data about dip and minimum depth of the investigated faults, the F75-89 seismic profile was depth converted (see Appendix B and Mazzella , where full details of methods and results are discussed).
 Inspection of the profile indicates that the recent deformation is controlled by steep reverse faults (Figure 5, Section 2) extending down to ~8–10 km depth and locally deforming the Apulian carbonate rocks (Figure 1, Section 1). At a shallower level, these thrusts displace the lower Pleistocene clay and apparently fold the overlying sequence up to the sea bottom (Figure 5, Section 2). The recent activity of these faults is well documented by the growth geometry of the Pliocene-Quaternary succession (Figure 5, Section 2), characterized by a variable thickness moving from the Amendolara High (~0.5 km) to the Sibari Basin (~3.5 km; Figure 5, Section 2).
 The major morphostructural feature in the profile is the Amendolara high (Figure 5, Section 2) that forms a positive structure bounded by SW (AMF) and NE (SPBF) verging thrusts. Careful mapping of displaced early Pleistocene reflectors [Mazzella, 2011] points to vertical displacement rates ranging from 0.26 to 0.30 mm/yr accrued along the AMF and the SPBF, respectively.
 Following Ferranti et al. , the faults highlighted in the F75-89 profile were laterally correlated to the onshore structures in order to extract fault geometric parameters for modeling (Table 4). Specifically, the AMF is correlated with the PF, the STF, and the SRF, and the SPBF is traced into the VF (Figure 2a and Table 4). Based on the above correlation, the depth-converted profile was used to derive dip as well as minimum depths of the modeled fault surfaces, and the relative uncertainty range.
Table 4. Geometric and Kinematics Parameters of the Modeled Fault Segments, Obtained Through Iterative Comparison Between Paleo-Shorelines Shape and Vertical Displacement Modelsa
Co-seismic Slip (m)
Minimum Depth (km)
Maximum Depth (km)
Slip Rate (mm/yr)
Bracket values indicate the range of possible values extracted from seismic reflection profile F75-89. Slip rates (mm/yr) are referred to the last 400 ka and were evaluated only for the fault segments that deform the terraces (STF1and VF1). The Mw is derived from scalar relationships of Wells and Coppersmith  (see the text for further details). AMF = Amendolara Fault; PF = Pagliara Fault; SPBF = Spulico Basin Fault; SRF = Saraceno Fault; STF = Satanasso Fault; VF = Valsinni Fault.
 Kinematics parameters (rake) of the modeled transpressive faults rely on structural analysis carried out in recent deposits (Figure 1) [Ferranti et al., 2009] and on strain tensor data derived from available focal mechanisms (Figure 1, Section 2). At the southeastern tip of the STF and SRF, the early Pleistocene clays are affected by left transpressive faults active with an ~NE to ENE trending shortening axis (Figure 1, sites FE09-1 and FE09-2). South of this site, middle Pleistocene sands and conglomerates are tilted and cut by meters-scale reverse faults indicating NE-SW shortening (Figure 1, site FE09-3). Similarly, focal solutions of small to moderate crustal earthquakes have a P axis of the incremental strain tensor trending from ~N-S to ~E-W, with the two events closer to the study area having a NE-SW axis (Figure 1, Section 2).
 In light of the small scatter in both the finite and incremental shortening trend estimates, we determined the preferred orientation carrying out a set of fault models with different shortening axis trend changing from NNE-SSW to E-W, tested against the MIS 5.5 terrace. We found that the model with a N30°E shortening axis better fits the MIS 5.5 deformation profile. Based on this, we proceeded in the model steps holding fix the fault rake derived by using a N30°E axis orientation and the geometric parameters established from the seismic profile. Without a priori information, we used for the slip per event an arbitrary value of 1 m.
 Finally, for the Pollino-Castrovillari normal fault (PCF), we used the parameters listed in the DISS database (DISS Working Group, 2010) that were derived from the geomorphic and paleoseismological studies of Cinti et al. [1995, 1997] (Table 5). Cinti et al.  did not find evidence of horizontal movements along the fault, and thus we used for the model a pure dip slip motion (Table 5) and a mean slip per event of 1.2 m.
LD = literature data; ER = empirical relationship; AR = analytical relationship.
Location (Lat/Lon; ED50)
Based on geological data from Cinti et al. .
Based on geological data from Cinti et al. .
Calculated using the scalar relationships from Wells and Coppersmith .
Minimum depth (km)
Based on geological data from Cinti et al. .
Maximum depth (km)
Derived from dip, width and minimum depth.
Based on geological data from Cinti et al. .
Based on geological data from Cinti et al. .
Based on geological and seismotectonic information.
Slip per event (m)
0.8 - 1.6
Based on paleoseismological data from Cinti et al. .
Slip rate (mm/yr)
0.2 - 0.6
Based on geological data and markers dislocation.
Vertical slip rate (mm/yr)
0.2 - 0.5
Based on geological observations of displaced markers.
Recurrence time (yr)
Based on paleoseismological data from Cinti et al. .
Calculated using the scalar relationships from Wells and Coppersmith .
5.2 Output Fault Models for the Pollino Coastal Fault Segments
 In order to determine the geometry, position, and number of fault segments that potentially deform the terraces along the coast of the Pollino Range, we generated several fault models using the range of geometric values reported in Table 4, and iteratively comparing vertical displacement models with paleo-shorelines profiles. In a first step, using the mean geometric parameters derived from the seismic profile F75-89 (Table 4), we modeled all the fault systems that cross the coastal area. We found that only the Satanasso and Valsinni faults (STF and VF, respectively; Figure 6, Section 1) are involved in warping of the marine terraces. Indeed, the surface deformational areas predicted by the numeric models for CF, PF, and CNNF are not geographically coincident with paleo-shoreline dislocation. Therefore, we regard these latter faults as inactive and discard them from further modeling. In a second step, we focused our attention on the STF and VF in order to iteratively determine the best geometric parameters as well as the number and the position of fault segments that best account for the paleo-shoreline deformation. Because the fold wavelength is controlled by dip and width of the fault plane, we used our deformed paleo-shorelines (geographic position and distance between anticline and syncline axis) to calibrate the width, maximum depth, and dip of the fault surfaces by moving in the value ranges reported in Table 4, with steps of 1 km and 1°, respectively. The width we obtained following this procedure can be assumed as the subsurface rupture width of Wells and Coppersmith . Once this parameter was fixed, we were able to evaluate the moment magnitude and, hence, the length (subsurface rupture length) trough the empirical relationships of Wells and Coppersmith . The aspect ratio of the fault dimensions obtained in this way is within the natural variability as shown by world-wide earthquake data sets [e.g., Nicol et al., 1996].
 Strike and dip direction of the two modeled faults were derived from geological maps, geological cross sections, and geophysical data (reference faults). Both the best position and the number of potentially active fault segments for each fault were chosen starting from the reference fault and proceeding with a step of 1 km along their strike (Figure 6, Section 2). Besides, we envisaged the possibility that the terraces were deformed by structures which do not coincide with the reference faults, such as blind or off-trace segments. For this reason, we looked for the position of the best fit fault segments in a NE-SW direction orthogonal to the reference faults, moving in a 4 km large window with a 0.5 km step (Figure 6, Section 3). At each new position of the fault segments, strike, dip, and rake were slightly adjusted to account for the internal geometric variations of the fault.
 Altogether, 558,613 fault models were scrutinized with a dedicated Visual Basic application to find the model that minimizes the difference between marine terraces elevation and predicted vertical displacement. This difference was estimated through a mean square deviation approach. The mathematical basis of the iterative fault model is described in Appendix C. All the tested fault models were evaluated against the best morphological marker available in the area, that is, the MIS 5.5 paleo-shoreline (T4), whose elevation was corrected site by site for the regional uplift component as described in section 4. In this way, only the local vertical displacement (elevation difference between paleo-shoreline and regional uplift curve) was modeled (Figure 4, Section 2).
5.3 Comparison with Observational Data
 Along the Pollino coastal range, the best model solution for the local coastal deformation is represented by the two fault segments STF1 and VF1 (Figure 7 and Table 4). The imposed ~NNE-SSW trending shortening axis outcomes in almost pure reverse motion along the fault segments (see theoretical focal mechanisms in Figure 7).
 Slip on the fault segments results in the active uplift of two fault-propagation folds beneath the coast (Pollino and Valsinni anticlines; Figures 7 and 8a, Section 1). The Pollino anticline is underlain by the southwest displacing STF1. This fault is modeled as a blind thrust extending from 1 to 7 km depth with a dip of 40°, a length of 16.9 km, a slip rate of 0.5 mm/yr, and a maximum expected moment magnitude of 6.3 (Table 4). Vertical uplift caused by the STF1 is restricted to a narrow sector centered at the boundary between the Pollino Range and the Sibari Plain, with only a very limited subsidence north and south of the anticline (Figure 8a, Section 2).
 On the other hand, the Valsinni anticline has grown above the northeastern displacing VF1 (Figure 7), which has a length of 8.7 km, a plane extending from 3 to 8 km with a southward dip of 60°, a slip rate of 0.7 mm/yr, and a maximum expected moment magnitude of 5.9 (Table 4). Like the STF1, slip on the VF1 is responsible for a narrow belt of uplift, albeit of a slightly lesser amplitude, centered in the northern Pollino Range at the border with the San Nicola Plain (Figure 8a, Section 2). No significant subsidence is expected north of the anticline (Figure 8a, Section 2).
 The modeled slip on the thrust faults reproduces with fair accuracy not only the MIS 5.5 paleo-shoreline shape, through which the model itself was tested, but also the elevation trend of terraces from T2 to T8 (Figures 8b and 8c, Sections 1 and 2). Conversely, because of their more limited spatial extent and the larger elevation uncertainties due to alluvial cover and human settlements, it was not possible to apply the fault model to the two younger terraces preserved along the southern Pollino coast (Figure 3): T1 (MIS 3.5; ~60 ka) and T0 (MIS 1; ~6 ka). Nevertheless, the sudden increase in elevation of T1, and to a lesser extent of T0 moving southward across the STF1 (Figures 8c, Sections 3 and 4, respectively), suggests fault activity also during the latest Pleistocene-Holocene.
 In order to evaluate the fault-induced uplift of T1 and T0 across the STF1, we estimated a nominal elevation due to the regional displacement for these terraces at the San Nicola Plain using the regional uplift rate derived at this location from the MIS 5.5 terrace (~0.6 mm/yr). Considering age and eustatic depth of T1 (60.5 ka; −36 m below sea level (bsl)) and T0 (5.9 ka; −5.3 m bsl), we obtained nominal elevations of ~1.4 m asl and approximately −1.7 m bsl, respectively.
 Because the uncertainties on elevation estimates can be significant on the younger terraces, we preferred to consider for T1 the minimum local vertical displacement (Figure 8c, Section 3; T1: 15 m). As far as T0, because of the large uncertainty on the Holocene displacement, which can incorporate un-modeled glacial-hydro-isostatic compensation components, we did not derive quantitative rate estimate across the STF1. Notwithstanding, the position of T1 and T0 above the elevation predicted by the regional uplift curve (Figure 8c, Sections 3 and 4) is a robust evidence of vertical displacement accrued by the STF1 during the Holocene. This contention agrees with previous studies in the Sibari Plain [Ferranti et al., 2011] and at Roseto Capo Spulico [Ferranti and Antonioli, 2009], where uplifted coastal markers suggest uplift rates higher than the regional signal during the Holocene (Figure 2a).
 With regard to the southernmost anticline (Figures 2a and 3, labeled as 3), highlighted by the paleo-shoreline trends along the western border of the Sibari Plain, no transpressive system is known which can explain this local deformation. The only structure having the potential to form the anticline is the Pollino-Castrovillari normal fault through footwall uplift. We tested this hypothesis by means of a vertical elastic displacement model benchmarked against the MIS 5.5 terrace (Figure 8a, Sections 1 and 2). Our model for the Pollino-Castrovillari Fault, whose parameters are summarized in Table 5, predicts a zone of uplift in the sector where we found one of the major positive undulations in the paleo-shorelines profile (Figures 7 and 8a, Section 1). In Section 1 of Figure 8a, it is possible to appreciate the good correlation between the model displacement curve and the increase in elevation of the MIS 5.5 paleo-shoreline moving southward from the Pollino range toward the Sibari Plain. Our model supports the results of Cucci and Cinti  and Cucci , which proposed the repeated reactivation of the Pollino-Castrovillari fault during the late Pleistocene.
6.1 Quantitative Contribution of Regional and Local Sources to Deformation
 Reconstruction and modeling of the deformation experienced by middle Pleistocene to Holocene paleo-shorelines along the Ionian margin of the Southern Apennines provided quantitative estimates of the regional and fault-induced components contributing to the total uplift. The two components specifically interlace along the Pollino Range coast, where the current paleo-shoreline shape can be interpreted as due to the superposition, on the regional northeastward tilt, of fault-propagation folds. Our results confirm the contention of Caputo and Bianca , Ferranti et al. , and Caputo et al.  that shortening in the Southern Apennines went on during the middle-late Pleistocene and, possibly, is still ongoing.
 The regional and local uplift signals reported in Figure 9 for coastal stretches straddling individual fault segments are mean values for the last ~400 ka and are based on our reconstruction of the regional uplift trend for terraces from T2 (MIS 5.1; ~80 ka) to T8 (MIS 11; ~401 ka). As pointed out previously [Cucci and Cinti, 1998; Ferranti et al., 2009; Santoro et al., 2009], the uplift process is mainly controlled by a regional source: the local fault uplift rate reaches a maximum of only 27% of the total budget along the VF1 (Figure 9). Fault-induced uplift occurred at broadly comparable rates for the STF1 (0.24 mm/yr) and the VF1 (0.28 mm/yr).
 We are aware that the position of the modeled fault segments and the estimation of their contribution to the uplift rate are in large part determined by the spatial distribution of marine terraces used to evaluate the model accuracy, and subordinately by the imposed kinematic parameters. On the other hand, landward and seaward of the coast, the lack of appropriate geomorphologic and geologic markers does not allow constraining the activity of the modeled faults. Nonetheless, the vertical displacement rate derived for each fault from the morphotectonic analysis is consistent with results from the seismic profile interpretation across the offshore extension of the STF1 and VF1 (Figure 5). Restoration of the vertical displacement experienced by middle Pleistocene reflectors documents 0.26 and 0.30 mm/yr along the AMF and SPBF, respectively [Mazzella, 2011], in close agreement with the rates stipulated along their coastal counterparts (Figure 9).
 Finally, for the Pollino-Castrovillari fault, the total displacement accrued by marine terraces in the Sibari Plain is partitioned between an 88% of regional uplift and a remaining 12% of footwall elastic displacement.
6.2 Temporal Changes and Spatial Migration of Fault-Induced Deformation
 The availability of multiple paleogeodetic markers represented by stepped terraces dated back to the last ~400 ka offers a means to evaluate the change in displacement distribution on the modeled fault segments during progressive time intervals. We found that fault displacement rate not only changed through time for individual faults (and for the cumulative fault array) but also varied in a coordinate manner among them. Figures 10a and 10b show the time-partitioned vertical displacement rate related to the STF1 and VF1 in the time lag spanning the formation of two successive marine terraces.
 For each time period, the vertical displacement rate (VDR) for individual faults was obtained from the following equation:
 where and are the number of seismic events predicted by fault modeling to explain paleo-shoreline deformations of terraces Tx and Ty, respectively; Cvd is the coseismic vertical displacement (mm); and T is the time elapsed (year) between the formation of terraces Tx and Ty. VDR values are reported in Figure 10.
 The fault-related vertical uplift rate was not constant during time but occurred as an alternation of periods with more rapid (up to ~1.9 mm/yr) and slower or null displacement (Figure 10). From ~330 to 200 ka, the STF1 accrued the largest deformation, but it was characterized by a sharp decrease in displacement rate since then. Conversely, the VF1, which was markedly active before ~330 ka, was almost quiescent between ~330 and 200 ka and resumed activity afterwards. Thus, periods of sustained deformation on one fault strand were coincident with quiescence on the other (Figures 10a and 10b). This observation is consistent with well-known models of stress shadowing between subparallel faults at a temporal scale (~100 ka) larger than typical seismic cycles of crustal structures [Dolan et al., 2007].
 During the last 80 ka, the STF1 and VF1 are characterized by comparable, limited displacement rates (0.26 and 0.20 mm/yr, respectively). This result is replicated for the STF1 by an estimated 0.20 mm/yr vertical displacement rate during the last 60 ka (Figure 8c, Section 3). Although the accurate physical significance of this observation is not completely understood, and caution must be exercised because of the lack of precise age and position constraints for the younger terraces, we speculate that the coastal fault systems along the Pollino coast may be witnessing as a whole a period of healing. Building on this, we further suggest that crustal deformation is being transferred to laterally adjacent fault strands, possibly in the offshore sector, where earthquakes appear concentrated [Ferranti et al., 2009].
6.3 Seismotectonic Implications
 The documentation of late Pleistocene shortening on the eastern side of northern Calabria brings important implications for the seismotectonic and seismic hazard evaluation of the region, which is characterized by a low level of historical and instrumental seismicity. However, a few Mw 5–6 historical events localized in the coastal area north of the Sila massif and in its offshore (CPTI, 2011) testify of the presence of seismogenic structures able to generate moderate but damaging earthquakes. The damage distribution and the effects associated with some of these events show that they may have been sourced by the fault system that includes the modeled coastal segments and stretches to the east in the offshore area [Ferranti et al., 2009]. Although we stress the difficulty of locating offshore historical and instrumental earthquakes, due to the uncompleted macroseismic field and the poor seismological constraints [e.g., Fracassi et al., 2012], this contention is consistent with the hypothesized recent shift of significant activity from the coastal to the offshore portion of the fault system.
 A major uncertainty related to this issue regards the seismic behavior of the studied structures. When we consider the expected 5.9 to 6.3 moment magnitude (Mw) of the modeled fault segments (Table 4) derived from scalar relationships [e.g., Wells and Coppersmith, 1994] and the significant extent of the combined coastal and offshore fault system (Figure 2a), the contrast with the lack of historical and instrumental seismic moment released in the region is apparent (Figure 1). Obviously, this discrepancy might be mitigated by the observation that fault modeling of geologic or geomorphic markers based on elastic deformation [Okada, 1985] often results in an overestimation of Mw, since it does not consider the possibility that energy is released through microseismic or aseismic processes.
 Moreover, historical reconstruction of settlement distribution in the Apennines indicates that the beginning of the fourteenth century marks a general cultural and economical development [Cinti et al., 2002]; therefore, the occurrence of an earthquake in the magnitude range estimated for the fault system since 1300 A.D. would probably have been reported in the historical record, even in sparsely settled regions. In northern Calabria, several small settlements already existed prior to 1300 A.D. [Cinti et al., 2002], and because of the region's relatively favorable climatic and topographic conditions, it is likely that cultural centers able to report large earthquakes in the area were settled at least since 1200 A.D.
 Thus, by combining historical, geomorphic and modeling data, we suggest that major (Mw 5.9–6.3) seismic events along the PCF, STF1, and VF1 since 1200 A.D. may be unknown because (1) the recurrence time exceeds the completeness of historic catalogue for the related magnitude; and (2) an unknown percentage of the accrued deformation is gradually released through aseismic and/or microseismic processes, thus decreasing the expected maximum coseismic magnitude.
 An integrated study based on morphometric analysis and fault numeric modeling was carried to shed light on the middle Pleistocene-Holocene deformation affecting the southernmost area of the Apennines frontal sector. In a region characterized by a scarce geomorphic expression of active faulting and a low level of instrumental seismicity, the morphometric approach proved to be useful in order to unravel the more recent tectonic history. A site-by-site sedimentological analysis of the middle Pleistocene terraced deposit permitted to extract, from the paleomorphology linked to the previous tectonic phases, the subtle deformation signal related to the more recent tectonic evolution. This signal has been split in a regional (deep-crustal or slab-related) and a local (fault-induced) component.
 The regional uplift process, determining an overall northeastward tilt of the coastal area (tilt rate: 3 × 10−3 m km−1 ka−1), accounts for ~80% of the total uplift. The local signal, conversely, is represented by anticline folds with a maximum ~20 km wavelength and an ~200 m amplitude. Taking advantage of existing geological and geophysical data, this local component can be modeled through an elastic fault numeric approach. Fault model results indicate two anticlines controlled by buried thrusts (spatially coinciding with the Satanasso and Valsinni transpressional faults mapped at the surface) that accrued deformation under a NNE-SSW trending shortening axis. Our multidisciplinary investigation points to activity of the thrust segments up to the late Pleistocene-Holocene times. In the time period investigated, local, fault-induced, vertical deformation was not constant but occurred as an alternation of more rapid (up to ~1.9 mm/yr) and slower or null periods of vertical displacement rates. Remarkably, we found not only that the fault segments activity was not constant at the 400 ka scale but also that deformation shifted during time between the two modeled fault segments.
 Based on our results and on available geophysical data, it is reasonable to hypothesize that the modeled coastal fault segments together with other segments in the laterally adjacent offshore belt could be seismogenic, thus posing a serious threat to the adjacent coastal sectors. The lack of significant historical and instrumental seismicity in the region can be explained by the release of an unknown percentage of accrued strain by means of aseismic or microseismic slip. Thus, acknowledging that the elastic fault model can overestimate the maximum expected magnitudes, we propose that the modeled fault zones can be responsible of moderate high intensity earthquakes, probably not registered in the historical record due to recurrence time interval exceeding the record length.
Geochronological Issues on the Age of Uplifted Terraces
 In Appendix A, we fully explain the reasons for preferring the Hearty et al.  aminozone scheme and for discarding some published dates along the coast of the Taranto Gulf.
A1. Aminozone Schemes
 In the northern Taranto Gulf, Hearty et al.  established local aminozones based on three U-series dating of Cladocora caespitosa sampled at Il Fronte, Mare Piccolo. Corals incorporate uranium into their aragonite exoskeletons in equilibrium with the uranium oceanic content during the life cycle, and then behave as radiometric closed systems until aragonite is transformed into calcite. For this reason, unrecrystallized fossil corals (calcite content <3%) are, generally, considered the best material for U/Th dating. The above mentioned U/Th ages were first presented by Hearty and Dai Pra ; in situ coral samples were collected in a level rich in Persististrombus latus. The three Cladocora caespitosa U/Th ages (117 ± 7 ka, 128 ± 7 ka, and 121 ± 7 ka) result in a mean value of 122 ± 4 ka and are all concordant with the attribution of Il Fronte section to the MIS 5.5 (~124 ka). These age determinations substantially refined previous corals U/Th ages presented by Dai Pra and Stearns  for the same section (87 ± 4 ka, 106 ± 8 ka, 130 ± 10 ka, and 154 ± 13 ka). It is not easy to determine whether the major scatter of these older dates represents several marine episodes centered on the Last Interglacial or analytical variability dependent on the major uncertainty about U-series method at the time, when potential problems of this technique applied to coral samples were not yet fully realized. Notwithstanding, we believe that the attribution of the Il Fronte section to the Last Interglacial, based on the available U-series ages, is strongly constrained. Regarding the aminozones, Hearty et al.  correlated the Last Interglacial deposits in the Taranto Gulf to their aminozone E and, specifically, to a Glycymeris alle/ille ratio of 0.38 ± 0.02. Overall, the extent of leucine epimerization for the aminozone E was based on 41 Glycymeris and 28 Arca specimens. Four additional aminozones were established in this area: C, 0.30 ± 0.02 (MIS 5.1/5.3); F, 0.50 ± 0.02 (MIS 7); G, 0.58 ± 0.02 (MIS 9); and K, 1.09 ± 0.09 (early Pleistocene). Besides, Hearty et al. , based on seven Cladocora caespitosa U/Th ages and 389 amino acid racemization analyses, extended the aminozone definition to the entire Mediterranean area.
 The Hearty et al.  aminozone scheme for the Taranto Gulf fits with amino acid data coming from Mediterranean sites where an independent chronology is well established. Because the extent of racemization is a temperature-dependent process, we focus our discussion on MIS 5.5 stratigraphic sections characterized by a similar mean annual temperature (MAT). For example, when we compare the Taranto Gulf with the San Grauet section of the Mallorca island (MATs: 16.9° and 16.8°, respectively), we found the same extent of leucine epimerization (alle/ille ratios for aminozone E of 0.38 and 0.37, respectively) in Glycymeris specimens. We remark that the correlation of the San Grauet section with the MIS 5.5 has been well established with U-series dating applied to one sample of Cladocora caespitosa (129 ± 7 ka) [Hearty et al., 1986] and three samples of Glycymeris violacescens (112 ± 7 ka, 102 ± 7 ka, and 135 + 14/−10 ka) [Hoang and Hearty, 1989]. Besides, a reasonable agreement also exists with data from Cagliari in southern Sardinia. Here a stratigraphic section hosting Persististrombus latus is independently attributed to the Last Interglacial based on Cladocora caespitosa U/Th ages (138 ± 8 ka) [Ulzega and Hearty, 1986] and electro spin resonance (ESR) applied on Arca Noae, Arca Tetragona, and Glycymeris (mean age: 112 ka) [Orrù et al., 2011]. In the same section, a larger leucine epimerization ratio (alle/ille: 0.42) with respect to the Taranto Gulf is in agreement with the slightly higher MAT (17.5°).
 Finally, the aminozone scheme of Hearty et al.  for the Taranto Gulf is further supported by its accordance with global isoleucine epimerization data relative to the Last Interglacial [Murray-Wallace, 1995].
Belluomini et al.  proposed a new aminozone scheme for the Taranto Gulf based on 65 alle/ille ratios determinations on Glycymeris, Arca, and Cerastoderma specimens, calibrated with U/Th ages of three samples of Cladocora caespitosa and four mollusk shells (Glycymeris and Pinna). They established the following correlation between alle/ille ratios and eustatic highstands: MIS 3 (0.20–0.31), 5.1 (0.34–0.38), 5.3 (0.41–0.44), and 5.5 (0.48–0.55).
 We believe, however, that severe pitfalls are embedded within the Belluomini et al.  aminozones definition. The scheme could be influenced by an age underestimation caused by a continuous or recent uranium uptake by the analyzed mollusks acting as radioactive open systems during diagenetic processes. Unlike corals, mollusk shells contain little authigenic U and their bulk U content essentially represents early diagenetic uptake [Kaufman et al., 1996]. It has been suggested that this uptake occurs within a few thousand years [Broecker, 1963] and perhaps up to 10 ka after death [Ivanovich et al., 1983]. If after this early U uptake mollusks act as closed systems for the duration of their diagenetic history, they can be used for chronological purposes [Szabo and Rosholt, 1969]. In their seminal work, Kaufman et al. , by reviewing U-series ages of 400 mollusks, concluded on the basis of internal inconsistencies with absolute and stratigraphic ages that 230Th/U ages of mollusks are generally unreliable. Similarly, McLaren and Rowe  demonstrated the unreliability of mollusks U-series ages in the Mediterranean area. On the other hand, consistent and reproducible results were obtained by Kaufman et al.  and Hillaire-Mercel et al. . It was shown that U-uptake mechanisms may differ between taxa [Hillarie-Mercel et al., 1996; McLaren and Rowe, 1996] and species of the same taxa [Hoang and Hearty, 1989], possibly in relation to the porosity of the shell and/or to the specific rate of organic lining decay. Therefore, it is necessary to check the validity of mollusks U-series dating through the analysis of different samples of various taxa collected from the same stratigraphic level as well as by comparing results with ages derived from different dating methods.
 These kind of tests are absent in Belluomini et al.  who analyzed only four samples (Glycymeris and Arca), collected from three different stratigraphic levels and with ages substantially lower than the Cladocora caespitosa U/Th ages obtained by Hearty and Dai Pra  at Il Fronte section. Furthermore, Hoang and Hearty  had already demonstrated the unreliability of U/Th ages applied on Glycymeris in a stratigraphic section located inside the area studied later by Belluomini et al. .
 Finally, if we accept the aminozone scheme and the U/Th ages of Belluomini et al. , a straightforward problem arises by their disagreement both with Mediterranean sites where chronology is firmly established (see for instance the sites of Cagliari and Mallorca discussed above) and global isoleucine data [Murray-Wallace, 1995].
A2. Discarded Ages
 In the study area, five sites where ages are available were not considered in our chronological scheme (sites AM97-1, CA10, and ZA06-1-2-3). Between the Basento and Cavone Rivers, just north of the study area, our terraces T2, T3, and T4 were sampled by Zander et al.  for optically stimulated luminescence (OSL) dating techniques (Figure 2b, sites ZA06-1, ZA06-2 and ZA06-3, respectively). The application of single aliquot regeneration (SAR) protocols and multiple aliquot additive (MAA) protocols on coarse-grain potassium feldspars and SAR protocols on quartz did not bring significant information about terrace chronology. Regarding the MAA quartz results, the authors stated that they are not reliable because the strong exponential shape of the saturating growth curves indicates a close to saturation level. With regard to the OSL dating results on potassium feldspars, in spite of the general agreement between SAR and MAA protocols, they fail in distinguishing by a chronological point of view terraces T2 and T3 (La Petrulla and San Teodoro I terraces in Zander et al. ). Although a small age increase is detectable in the OSL feldspar results for our terrace T4 (San Teodoro II terrace in Zander et al. ) with respect to T2 and T3, SAR and MAA ages do not show a stratigraphic order in the coastal depositional sequence. Because of the evident age underestimation, probably caused by unstable luminescence centers in the feldspars and fading effects, the authors admitted the impossibility to establish a reliable chronostratigraphic frame for the terrace flight. According to their conclusions, we did not consider their ages in our chronological scheme.
 Between the Agri and Cavone Rivers, Caputo et al.  obtained an AMS 14C age of >42 ka from a Pecten sp. collected at the outer margin of our terrace T2 (Figure 2b, site CA10). Radiocarbon dating of material aged >25–30 ka B.P. can be problematic because of several principal influences. Perhaps the most significant is contamination of carbonaceous material by exogenous carbon, the extent to which it has become chemically associated or cross-linked and the success of radiocarbon pretreatment chemistry in removing it. Furthermore, the calibration of 14C age >25–30 ka B.P. is problematic because of the uncertainty about the 14C concentration in the Earth's atmosphere in this period. For example, Giaccio et al.  and Pyle et al.  brought evidences of larger changes in atmospheric 14C historic levels than previously attested, probably related to geomagnetic excursions. For these reasons and as suggested by Caputo et al.  themselves, their AMS 14C age poses only a minimum age limit to the relative coastal depositional sequence and does not bring substantial information about terrace chronology.
 On the northern slope of the Pollino Range, Amato et al.  estimated the alle/ille ratios on four samples of Glycymeris sp. collected from the outer margin of our terrace T5 (110 m asl), about 4 km NW of Roseto Capo Spulico village (Figure 2a, site AM97-1). The alle/ille ratio of 0.29 ± 0.02 for the site AM97-1 corresponds, according to the Hearty et al.  aminozone scheme, to MIS 5.1/5.3. Based on geomorphological considerations, the authors recognized a substantial underestimation in this age estimate. Hence, this site was not considered in our chronological scheme.
Depth Conversion of Seismic Profile F75-89
 The TWT F75-89 profile was depth converted based on a standard multiphase velocity analysis. In order to define the actual thickness of single layers and assign them a velocity value, starting from the velocity readings at individual shot points where available, we computed every 200 ms the interval velocity at each point, by the Dix formula [Dix, 1955] simplified for horizontal strata:
where vint is the interval velocity for each layer; v1 and v2 are stacking velocity values for the upper and lower layer boundaries; and t1 and t2 are two-way travel time values down to the upper and lower boundaries. With these values, we generated an isovelocity contour map using a Surfer-based triangulation with linear interpolation. We geometrically overlaid this map to the seismic profile in order to manually improve the automatic contour, eliminating computational errors through picking and smoothing of the curvatures. Finally, the manual contour map was re-interpolated to produce a final contour of isovelocities. On this base, the contour was manually converted into fields of color, each of them identifying a velocity body; this image was imported in a dedicated software (depthcon2000) that recognizes the chromatic scale associated to a specific velocity and converts the time image (both the interpolated and original seismic profile) into a depth calibrated image.
 Identification in the depth profile of prominent horizons was calibrated with nearby oil exploration well logs (Francesca, Franca and Lucia wells; Figure 2a Section 1). Occasional biostratigraphic and lithologic observations offer excellent markers that can be laterally traced along the profile and allow to assess the deformation rates across faults and folds.
Iterative Fault Model
 The iterative fault model is based on the following mathematic theory. Considering a paleo-shoreline of np points and a certain number (nL) of vertical deformation models (hereafter Layers; L), we want to determine the coefficients (number of seismic events) of the Layers linear combination that best approximate the marine terraces elevation. The solution of the problem is represented by the absolute minimum of the following equation:
where (xi, yi) are the points of a shoreline, Ri is the elevation of the shoreline in the ith point, Lij is the elevation of the layer (vertical deformation model) jth in the ith point, and (α1, α2, … αnl) are the linear combination coefficients that we want to determine. Given that equation (C1) is the function of a hyperbolic paraboloid, it is characterized by only a minimum, that is, the absolute minimum that guarantees the solution of the problem. The approximations between vertical displacement models and terraces elevation were estimated using the following equation:
 In order to determine the absolute minimum of equation (2), we used the following procedure:
 For each l = 1,…,nl, we calculated the partial derivative of function F:
 We obtained a system of nl equations in nl unknowns. The absolute minimum was found solving the following system: