Along the Italian peninsula adjoin two crustal domains, peri-Tyrrhenian and Adriatic, whose boundary is not univocal in central Italy. In this area, we attempt to map the extent of the Moho in the two terrains from variations of the travel time difference between the direct P wave and the P-to-S wave converted at the crust-mantle boundary, called PsMoho. We use teleseismic receiver functions computed at 38 broad-band stations in this and previous studies, and assigned each of the recording sites to the Adriatic or peri-Tyrrhenian terrains based on station location, geologic and geophysical data and interpretation, and consistency of delays with the regional Moho trend. The results of the present study show that the PsMoho arrival time varies from 2.3 to 4.1 s in the peri-Tyrrhenian domain and from 3.7 to 5.5 s in the Adriatic domain. As expected, the lowest time difference is observed along the Tyrrhenian coastline and the largest values are observed in the axial zone of the Apennine chain. A key new result of this study is a sharp E-W boundary in the Adriatic domain that separates a deeper Moho north of about 42°N latitude from a shallower Moho to the south. This feature is constrained for a length of about 40 km by the observations available in this study. The E-W boundary requires a revision of prior mapping of the Moho in central Italy and supports previous hypotheses of lithosphere segmentation.
 Peninsular Italy extends in the Mediterranean Sea from 38° to 46° latitude North and from 8° to 18° longitude East. Its geologic setting is dominated by the Apennine chain that extends along the whole peninsula. This chain built up mostly during the Neogene and early Pleistocene following the deformation of the African continental margin of the Tethyan ocean [e.g., Malinverno and Ryan, 1986; Albarello et al., 1995; Vezzani et al., 2010].
 In peninsular Italy, the topography of the Moho discontinuity, that is the object of this study, has been investigated through active seismic profiles collected during the Deep Seismic Sounding (DSS) experiments in the 1960s–1990s [Cassinis et al., 2003, and references therein] and the CROP Project in the 1980s–1990s [Scrocca et al., 2003, and references therein], and with passive seismology methods such as tomography and teleseismic receiver functions [e.g., Piana Agostinetti et al., 2002; Mele and Sandvol, 2003; Mele et al., 2006; Di Luzio et al., 2009; Di Stefano et al., 2009; Piana Agostinetti and Amato, 2009]. Active and passive seismic data have been combined in Di Stefano et al. .
 The Moho map proposed by Cassinis et al.  had the merit, unlike the majority of the maps derived from other studies, of distinguishing the crustal domains that characterize Italy and surrounding areas: continental crust in the European and African/Adriatic domains; oceanic/suboceanic crust in the Ligurian and Tyrrhenian Seas; transitional crust in the peri-Tyrrhenian side of peninsular Italy and northern Sicily. The boundary between the Adriatic and peri-Tyrrhenian crust runs along peninsular Italy and northern Sicily (Figure 1). Recently, Di Stefano et al. [2009, 2011] have proposed two boundaries that differ from each other and from that of Cassinis et al.  in central Italy, as shown in Figure 1. In this area, where the three boundaries deviate one from the other and one of them is partially unconstrained, we attempt to reconstruct the extent of the Adriatic and peri-Tyrrhenian crust.
 To map the Adriatic and peri-Tyrrhenian Moho, we use the teleseismic receiver functions method that is based on the identification of the P wave converted to S at the Moho discontinuity (called PsMoho in the following). The delay time of the PsMoho with respect to the direct P arrival is affected primarily by Moho depth: the larger/smaller the delay, the deeper/shallower the Moho beneath the recording site; therefore, we interpret variations in the PsMoho time in terms of variations of Moho depth. We integrate new data with previous receiver functions computed by Mele et al.  and Di Luzio et al. .
 The stations used in central Italy are assigned to one or the other crustal domain based on location, geologic and geophysical data, and consistency with the regional trend of the Moho. The PsMoho-P times are interpolated with the Ordinary Kriging statistical method to map the extent and the lateral variations of the Adriatic and peri-Tyrrhenian Moho.
2. Geologic Setting
 In peninsular Italy, the peri-Tyrrhenian area is characterized by a stretched transitional crust with positive Bouguer anomalies [e.g., Morelli, 1981], high heat flow [e.g., Della Vedova et al., 2001] and relatively low uppermost mantle velocity [e.g., Mele et al., 1998]. On the contrary, the Adriatic domain is a more stable area with low heat flow, low-to-moderate positive Bouguer anomalies and normal-to-high uppermost mantle velocity. Since Mele and Sandvol , the Adriatic Moho was inferred to deepen to about 50 km beneath the Apennine chain.
 The geologic setting of central Italy is characterized by Meso-Cenozoic platform and basin units of the Apennine chain verging NE-ward above the Bradanic foredeep and the Adriatic/Apulian foreland (Figure 2). To the west, Plio-Quaternary marine-to-continental deposits and Pleistocene volcanics cover large sectors of the internal Apennines that were downthrown by extensional faults since the late Miocene [e.g., Patacca et al., 1990]. The foreland sequence outcrops in the Gargano promontory and Tremiti Islands (Figure 2), mainly characterized by the carbonate units of the Apulian Platform (AP). Part of the Apulian Platform was involved in the Apennine deformation during the Pliocene-Early Pleistocene; it is exposed in the Maiella Massif and surroundings (Apennine external units of Figure 2) [Bally et al., 1986; Mostardini and Merlini, 1986; Patacca et al., 2008; Cosentino et al., 2010].
 The Apulian Platform Top (APT) is a regional key-horizon distinctive of the Adriatic crust; it was used to follow the westward dipping of the foreland monocline beneath the foredeep and the Apennines [Mariotti and Doglioni, 2000]. This horizon, made of Miocene limestones and/or evaporites, is reached at depths ranging from about 1 km in the peri-Adriatic region to about 3 km in the axial zone of the Apennines by the exploration wells plotted in Figure 2. In the CROP11 profile, a high-amplitude pair of reflectors interpreted as the APT horizon is followed from the Adriatic coast to the Fucino basin [Patacca et al., 2008]; west of the Fucino basin the CROP11 profile is not interpreted. In this work, the APT will be used to constrain the extent of the Adriatic crust.
3. Method of Analysis, Seismologic Data, and Observations
 Since the first observations in the 1950s [Cook et al., 1962, and references therein], teleseismic P waves converted to S at major velocity discontinuities of the Earth were used to infer the gross seismic structure under a recording station. The PsMoho is often the highest-energy signal in the coda of the direct P arrival due to the large velocity contrast between the crust and the mantle, and is used to build regional Moho maps [e.g., Priestley et al., 1988; Kind et al., 1995; Jones and Phinney, 1998; Al-Damegh et al., 2005; Lloyd et al., 2010]. Data usable for these studies are three-component, possibly broad-band recordings of teleseismic events with epicentral distance of 30° to 90°.
 The PsMoho phase arrives few seconds after the direct P and most of the times it is hard to observe in the seismogram. The method used to identify the PsMoho consists in deconvolving the vertical component of the ground motion from the horizontal component rotated into the radial direction (source-to-receiver path) where Ps conversions have the largest amplitude [Langston, 1979]. Deconvolution filters out most of the common features such as source, travel path effects, and instrumental response, producing a simpler time series called receiver function. This last is composed by the first positive P pulse followed by Ps conversions and reverberations. Deconvolution also enables to compare receiver functions from various seismic sources that are stacked together to enhance the coherent signals.
 The time delay between PsMoho and P (tPs hereinafter) can be used to estimate the depth of the Moho (H) for given bulk crustal velocities Vp and Vs and P-wave incidence angle (expressed through the ray parameter p):
 In this work, we have collected teleseisms with minimum magnitude Mw 5.5 recorded in the 2004–2009 period by 29 permanent stations of the Italian Seismic Network. The epicentral distance is computed from the center of the study area. Given the abundance of seismic sources in the distance range 80° ± 10°, we selected these events because steeper incidence angles yield larger energy of the incoming P wave.
 After a selection of the recordings in terms of the signal-to-noise ratio, we cut a window of 30 s from the seismograms of 148 events (Figure 3a), starting 5 s before the P onset. To compute receiver functions, we applied the time-domain deconvolution technique of Ligorria and Ammon ; a Gaussian low-pass filter with width parameter α = 2.0 was used to remove the high-frequency noise.
 In the receiver functions of 24 stations, a positive peak arriving 2.3–5.2 s after P is interpreted as the Ps wave converted at the Moho discontinuity; five stations were discarded due to noisy or inconsistent observations. We also used the tPs computed by Mele et al.  at the permanent station AQU and 12 temporary stations installed for few months in 1995 (0–4C, 6–9C, 11C, 12C, and 14C), and by Di Luzio et al.  at the permanent station FRES (Figure 3b). For most of these stations, only events from the north-east and 80° ± 10° distance were available [Mele et al., 2006].
 In the present study, most of the observations are naturally clustered between 330° and 100° backazimuth (Figure 3a) and this prevented to analyze the crustal response as a continuous function of azimuth. For this reason, and for consistency with previous works, we stacked the receiver functions of events occurred in the NE quadrant. This ensures also to sample the same Moho structure beneath each station.
 Depending on the working state and quality of the recording site, the number of receiver functions varies from 4 (4C, GUAR, CIGN) to 56 (INTR). In Figure 3b are shown the stacks of five stations arranged along a SW-NE profile that crosses the boundaries between the peri-Tyrrhenian and Adriatic crust.
4. Mapping the Peri-Tyrrhenian and Adriatic Moho
 In order to estimate Moho depth from the PsMoho delay, a bulk crustal velocity must be provided for all stations. However, previous works propose conflicting models in the study area, especially at midcrustal depth. As an example, we show in Figure 4 two seismic tomography sections where high-velocity anomalies are imaged on both sides of the Fucino basin [Chiarabba et al., 2010] and two sections interpreted from active seismic data where low velocity is inferred in the same area [Cassinis et al., 2003; Patacca et al., 2008].
 Because of the uncertainty in the regional velocity structure, in the present study we use the PsMoho delays as indicative of Moho depth variations. The delay of the Moho conversion is read from the stack trace of each station and mapped in Figure 5a. PsMoho delays span from 2.3 to 5.5 s, and the conversion points at the Moho occur NE of the stations, at an average distance of 10 km. In this map, we attributed each station (i.e., observation points of tPs) to the Adriatic or peri-Tyrrhenian terrain based on location with respect to the proposed boundaries; where the boundaries deviate from each other, the attribution is based on surface and shallow geology (well logs) or on the consistency of tPs with the Moho trend defined by the Tyrrhenian stations 0–4C and the Adriatic stations 6–14C and FRES [Mele et al., 2006; Di Luzio et al., 2009].
 Stations located west of the three boundaries are assigned to the peri-Tyrrhenian crust (from north to south: MAON, LATE, CESX, MNS, TOLF, MTCE, ROM9, RDP, CERT, GUAR, GIUL), while stations located east of the boundaries are assigned to the Adriatic crust (TERO, CAMP, CAFR, LPEL, CIGN, SGRT, MSAG). Other stations can be attributed to the Adriatic crust based on the following aspects: (i) MIDA and CERA, located close to two explorations wells that reached the APT horizon at about 3 km of depth, and to outcrops of the deformed Apulian domain (see area framed in Figure 2); (ii) INTR, located along the segment of the CROP11 profile where the reflection package interpreted as the Apulian Platform Top is recognized beneath the Apennine units [Patacca et al., 2008]; and (iii) CAMP and FAGN, where the relatively large tPs (5.0 and 5.2 s) is consistent with the westward deepening Adriatic Moho.
 The attribution of stations matches the boundaries of Cassinis et al.  and Di Stefano et al. , while it is inconsistent with the boundary proposed by Di Stefano et al.  (Figure 5a). Stations FIAM, VVLD, and POFI are uncertain because their location is not constrained by geologic evidence and the tPs matches the trend of the Moho in both crustal domains.
 Figure 5b displays a contouring of tPs. We used the Ordinary Kriging prediction method [Matheron, 1970] to model the spatial trend of a single variable; to avoid a priori bias, data were interpolated without using barrier polylines between the Mohos. The basic assumption, when using statistics to handle heterogeneity in Earth systems, is that properties are not random, but have some spatial continuity or are correlated over some distance. The distribution of data is analyzed to create a semivariogram model that allows to compute the parameter value in unsampled locations. The Kriging model generates the predicted surface after selecting the best suitable model based on regression statistics. Observed versus simulated tPs resulting from the cross-validation procedure are plotted in the inset of Figure 5b. In the map of Figure 5b, smaller differential times occur in the western sector of the peninsula, characterized by brown colors (tPs between 3.3 and 3.8 s), matching the attribution of the peri-Tyrrhenian stations. As to the Adriatic stations, the contouring highlights two regions with different tPs that define a sharp transition of the Moho surface along the 42°N latitude: tPs changes from 4.6–4.7 s to the north to 3.7–3.8 s to the south. The receiver function stacks of the five stations straddling the Moho transition are shown in Figure 5b.
 In central Italy, we have distinguished stations located in the peri-Tyrrhenian and in the Adriatic terrain to reconstruct the variations of the Moho in these crustal domains.
 A key finding of this study is a sharp variation of tPs in the Adriatic domain, at about 42°N latitude: from north to south, tPs changes from 4.6–4.7 s at stations 9C, 11C, and 12C to 3.7–3.8 s at stations INTR and LPEL, within a distance of 15 km (Figures 5a and 5b). At stations 9C, 11C, and 12C, Di Luzio et al.  have estimated a Moho depth of 38 ± 1 km using a local bulk crustal Vp of 6.3 km/s derived from the interpretation of the CROP 11 profile. This is a good crustal average commonly used in literature. Adopting such Vp value in equation (1), we estimate a Moho depth of 30 and 31 km beneath stations LPEL and INTR, respectively, i.e., the Adriatic Moho is ∼8 km shallower south of the 42°N parallel. The E-W Moho transition can be constrained for about 40 km with the observations available for this study (Figure 5b). It is worth to underline that the PsMoho delays of the Adriatic stations are consistent on either side of the Moho transition: 4.6–5.5 s are observed at all stations located north of 9C–12C while 3.8–4.2 s are observed at all stations located south of INTR and LPEL (Figures 5a and 5b).
 The E-W step of the Adriatic Moho supports previous ideas of lithosphere segmentation in central Italy [Royden et al., 1987; Doglioni et al., 1994]. Royden et al.  based their model on the morphology of the Apennine foredeep basin (correlated with Bouguer gravity anomalies) and of the outermost thrust of the chain; both show differential offsets from north to south reflecting a different amount of lithosphere retreat (Figure 5c). Doglioni et al.  hypothesized that a differential lithosphere rollback occurs between the central Adriatic and the Puglia region, caused by the difference in the lithospheric thickness inherited from the Mesozoic rifting: the downgoing of the 40 km thicker Puglia lithosphere slowed down since the middle Pleistocene favoring the uplift of the foreland and the Moho in the Gargano promontory (Figure 5d). The present study results confirm that the Adriatic Moho is shallower over the whole sector below 42°N latitude, not only beneath the Gargano promontory, and is rather flat: tPs values recorded at stations MSAG and SGRT are similar to those recorded at the surrounding stations, including 14C, located in the Tremiti islands, where Mele et al.  have estimated a Moho depth of 33 km.
 The step of the Moho in central Italy is not displayed in the Moho map of Piana Agostinetti and Amato , obtained with the receiver functions stacking technique of Zhu and Kanamori . This map is a smoothed image of Moho depth variations where about 1/3 of the stations are of high quality (class 1–2 defined by the authors). In the study area the temporary stations 9C, 11C, 12C, and the permanent station LPEL, i.e., four of the five stations that constrained the E-W Moho step, are not used by these authors; this produces a low-resolution image of the Adriatic Moho around the 42°N parallel. It is worth noting that INTR, that is the only station shared by the two studies around the Moho step, has the same average Moho depth (Table 1).
Table 1. Seismic Stations Used in This Study Listed in Alphabetical Order With PsMoho Time Delays (tPs), Assigned Crustal Domain (AD: Adriatic; TR: peri-Tyrrhenian), and Moho Depthsa
In the last column are listed for comparison the Moho depths of Piana Agostinetti and Amato  (PA-A 2009); in parenthesis is the quality class of each station defined by these authors, decreasing from 1 to 5. The Adriatic stations located within 50 km from INTR are highlighted in boldface.
Di Stefano et al.  have used the Moho depths estimated by Piana Agostinetti and Amato  to integrate active seismic data and reconstruct the Moho topography in Italy. In the study area, the Tyrrhenian/Adriatic boundary of Di Stefano et al.  is in contrast with the PsMoho delays: several stations located west of this boundary have tPs of 5.0 s and more (Figure 5a), corresponding to Moho depths larger than 40 km, that are not consistent with the peri-Tyrrhenian Moho.
 We have presented a revised mapping of the peri-Tyrrhenian and Adriatic Moho in central Italy supplementing our previous receiver function studies (14 stations) with results obtained from 24 additional stations. We have compared the cumulative receiver function results with constraints from wells data and active source imaging to assign each station to either crustal domain. The new result of the present study is evidence for a sharp E-W transition in the Adriatic Moho that rises of ∼8 km south of ∼42°N parallel. This feature can be constrained for a length of ∼40 km with the data available in this study. The E-W transition requires a major revision to prior mapping of crustal domains and supports previously hypothesized lithosphere segmentation.