To investigate the kinematics of the Adriatic region, we integrate continuous and episodic GPS measurements with Mw > 4.5 earthquake slip vectors selected from the Regional Centroid Moment Tensor catalogue. Coherent motion of GPS sites in the Po Valley, in Apulia, and in the Hyblean Plateau allows us to estimate geodetically constrained angular velocities for these regions. The predictions of the GPS-inferred angular velocities are compared with the earthquake slip vectors, showing that the seismically expressed deformation at the microplate boundaries is consistent with the observed geodetic motion. The remarkable consistency between geodetic, seismological, and geological evidence of active tectonics suggests that active deformation in the central Adriatic is controlled by the relative motion between the Adria and Apulia microplates. The microplates' angular rotation rates are then compared with the rotation rates calculated with a simple block model supporting the hypotheses (1) that Apulia forms a single microplate with the Ionian Sea and possibly with the Hyblean region and (2) that Adria and Apulia rotate in such a way as to accommodate the Eurasia-Nubia relative motion. We suggest that the present-day microplate configuration follows a recent fragmentation of the Adriatic promontory that during the Neogene rigidly transferred the Africa motion to the orogenic belts that now surround the Adriatic region.
 Large regions of relatively low seismicity have been recognized within several plate boundary zones [McKenzie, 1972; Thatcher, 1995; Stein and Sella, 2002]. Seismotectonic and space geodetic studies have investigated the kinematics of these regions which are frequently characterized by rapid rotations, relative to adjacent major plates, about nearby Eulerian poles. Resolving the motion of the microplates within these plate boundaries is important not only for regional studies but also to discriminate among end-member models that describe distributed deformation either as the result of tractions applied at the edges of the crustal blocks or as the result of shear tractions exerted by diffuse deformation at the base of the lithosphere [e.g., England and Jackson, 1989]. Seismotectonic investigations in the Mediterranean region had a central role in the development of the microplate concept [McKenzie, 1970], in the description of distributed deformation [McKenzie and Jackson, 1983], and in the confirmation of early microplate hypotheses by space geodesy [e.g., Smith et al., 1994; McClusky et al., 2000].
 The Adriatic region (Figure 1), a relatively aseismic area encircled by active orogenic belts, has always attracted and puzzled the interest of the researchers involved in the studies of the Alpine-Mediterranean plate boundary zone [Argand, 1924; McKenzie, 1972; Channell et al., 1979; Anderson and Jackson, 1987]. Whereas stratigraphic and paleomagnetic studies have described the Meso-Cenozoic evolution of the Adriatic region as a rigid promontory of the African plate [Channell et al., 1979; Channell, 1996], seismological and space geodetic studies now converge in the definition of the Adriatic region as moving independently relative to both Eurasia and Africa. Slip vectors of earthquakes around the Adriatic Sea and space geodesy have been used to establish a counterclockwise rotation relative to Eurasia around an Eulerian pole located between the Po Valley and the French Alps [Anderson and Jackson, 1987; Ward, 1994; Calais et al., 2002]. This apparent discrepancy raises the possibility of recent plate boundary reorganization and offers the opportunity to investigate the factors contributing to the fragmentation of rigid continental promontories and indentors in collision zones.
 A general acceptance of the Adria microplate hypothesis however has been hindered by two main factors. The first is the lack of a clear plate separation between Adria and Africa. According to the position and rate of published Eulerian poles of Adria relative to Eurasia, several mm/a are expected at the southern margin of Adria and somehow expressed in the southern Adriatic by the distribution of seismicity and active faulting. Attempts based on limited observational support have been made to place the boundary in the Strait of Otranto [Anderson and Jackson, 1987], along the Apulia Escarpment [Battaglia et al., 2004; Serpelloni et al., 2005], or by considering diffuse deformation in the Ionian Sea [Goes et al., 2004; Stein and Sella, 2005]. The second factor, which has remained relatively obscure since the proposition of an independent Adria microplate, concerns the relation between the rotation of Adria and the accommodation of the relative motion between Eurasia and Nubia (the African plate west of the East African Rift).
 In this study we present a new GPS velocity field which, integrated with the analysis of earthquake slip vectors, allows us (1) to assess the crustal motion of the Adriatic region and (2) to propose a new model to explain this motion in the contest of the convergence between Eurasia and Nubia. On the basis of a rigorous statistical analysis of the data, we propose a set of Eulerian poles which describe the kinematics and the active deformation of the Adriatic region in terms of relative motion between two microplates, Adria and Apulia. We propose that Apulia forms a single microplate with the Ionian Sea and possibly with the Hyblean Plateau thus resolving the need of a direct boundary between Adria and Nubia. We then use a simple block model to illustrate how the observed microplate's motion may be related to and accommodate the convergence between Eurasia and Nubia. The proposed present-day kinematic model of the plate boundary is then evaluated in relation to the evolution and fragmentation of the Adriatic promontory.
2. Tectonic Settings and Previous Works on the Adriatic Region
 The Meso-Cenozoic geological evolution of the central Mediterranean, as the boundary between the Eurasian and African plates, has been largely controlled by the paleogeographic inheritance of the African and Eurasian margins [Argand, 1924]. The most relevant feature is the Adriatic promontory, a large piece of continental crust of African affinity [Channell et al., 1979], which colliding with the Eurasian plate formed the orogenic belts that now surround the Adriatic Sea. The reconstruction of past relative plate positions [Dewey et al., 1989; Mazzoli and Helman, 1994; Wortmann et al., 2001; Rosenbaum et al., 2002] and the absence of differential paleomagnetic rotations [Channell, 1996; Van der Voo, 1993; Rosenbaum et al., 2004] have been used as arguments to support the hypothesis of the Adriatic region behaving as a rigid promontory of the African plate.
 Outcrops of the relatively stable and undeformed regions of the Adriatic Promontory are exposed in Apulia, Istria, Hyblean Plateau and Southern Alps (Figure 2) [Channell et al., 1979]. In contrast the margins of the Adriatic promontory underwent intense deformation during the Alpine Orogeny resulting in the formation of the orogenic belts that encircle the Adriatic region [Rosenbaum et al., 2004]. The relatively slow convergence between Eurasia and Nubia coexisted with the simultaneous subduction of the Ionian lithosphere (Figure 2) northwestward beneath the Tyrrhenian Sea and northeastward beneath the Hellenic Arc [Jolivet and Faccenna, 2000]. Estimates of past trench positions in the Tyrrhenian Sea during the Neogene-Quaternary [Malinverno and Ryan, 1986; Patacca et al., 1990; Nicolosi et al., 2006] suggest that back-arc opening rates as high as 50–70 mm/a has been much faster than the plate convergence during the same time interval (Figure 2). The main effect of these opposite subduction zones around the Ionian Sea has been the reduction of the lithosphere which could rigidly transfer the motion of Africa in the collision zone along the northern rim of the Adriatic promontory.
 In an Eurasian reference frame crustal velocities in the Adriatic regions are almost perpendicular with respect to the Eu-Nu plate motion direction (Figure 1). Calais et al.  and Battaglia et al.  have shown that the velocities of GPS sites within the Adriatic region are consistent with a counterclockwise rigid rotation relative to Eurasia about a pole located in the Po Valley in agreement with the hypothesis of Anderson and Jackson . Substantial uncertainty however remains regarding the southward extent of Adria, its southern boundary with Nubia, and the role of shortening in the northern Apennines. A separation between the northern and the southern Adriatic region along a tectonic structure also known as the Gargano-Dubrovnik line, has been proposed [Westaway, 1990; Calais et al., 2002; Oldow et al., 2002; Battaglia et al., 2004] on the basis of the distributed seismicity which characterizes the central Adriatic [Console et al., 1993; Favali et al., 1993].
 Code and phase data for the time interval 1996.0–2008.3 have been reduced with the Gipsy-Oasis software. Point positioning [Zumberge et al., 1997] and precise orbits and clocks from JPL (ftp://sideshow.jpl.nasa.gov) were used to analyze GPS data followed by ambiguity resolution. For continuous sites this step was performed using the Ambizap strategy [Blewitt, 2008]. The Ambizap algorithm uses a fixed point theorem to identify linear combinations of network parameters that are theoretically invariant under ambiguity resolution [Blewitt, 2008]. Such strategy allows for very rapid, multiple reanalysis of large networks to assess various models and makes trivial the addition of extra stations or subnetworks to an existing solution. Daily ambiguity-fixed solutions of episodic measurements have been processed by integrating survey measurements with 30–40 continuous sites every day and using the GIPSY ambigon-p1 script. Each ambiguity-fixed daily solution is aligned to a realization of the ITRF2005 reference frame [Altamimi et al., 2007] using a seven-parameter transformation. The realization of ITRF2005 is obtained using daily precise point-positioning solutions of the 67 stations in Eurasia, Nubia and central Mediterranean with the longest observation intervals and minimum hardware changes, and parameters from JPL (x-files) that permits daily coordinate transformation into ITRF2005. Average weighted root-mean-square of the residuals (WRMS) after transformation are 1.6, 2.2, and 6.2 mm for the north, east and up components, respectively.
 Time series in the ITRF2005 reference frame for all the stations were locally rotated to NEU components and cleaned from outliers using the strategy described by Nikolaidis , and then analyzed for their noise properties, linear velocities, periodic signals and antenna jumps using maximum likelihood estimation (MLE). This technique allows simultaneous estimation of the noise properties structure together with the parameters of a time-dependent model of the data. Quantities estimated in the MLE analysis are linear trend, offset at designated times, annual and semiannual periodic signals and power law noise index and amplitude [Williams, 2003]. The MLE analysis has been performed using the CATS software [Williams, 2007] and all parameters estimated with full flicker noise covariances. Their uncertainties can thus be considered realistic estimates on the basis of analysis of the noise at individual stations. Average RMS of the residuals are 1.6, 2.0 and 5.9 mm for the north, east and up components, respectively. Only sites with an observation period longer than 2.5 years have been considered, as shorter intervals may result in biased estimates of linear velocities [Blewitt and Lavalle, 2002]. Velocity uncertainties for survey-style measurements were estimated adopting an error model which combine white + random walk noise using a value of 1 mm/a0.5 for the random walk component [Hammond and Thatcher, 2004; Langbein and Johnson, 1997]. Station positions, horizontal linear velocities in various reference frames and associated uncertainties, time span of the observations and number of daily solutions for each site are listed in Table S1 of the Auxiliary Material.
4. Analysis of the Velocity Field
 The Euler vectors of the major plates and microplates are estimated by weighted least-squares minimization of their velocities relative to the ITRF2005. We formed the data covariance matrix using the velocity uncertainties and their site-by-site NS-EW correlation. To verify the quality of the fits, the results of the various inversions are analyzed using the F ratio tests described by Nocquet et al. . The F ratio test compares the χ2 from two least squares estimations in order to decide which model best fits the data. The null hypothesis follows the rigid plate assumption that all the sites velocities can be modeled with a single pole of rotation. The F ratio test is here used in two ways: in the first approach we test if the velocity of an individual site (or an earthquake slip vector) is consistent with a previously defined data set. If the null hypothesis is verified the tested site is consistent with the velocity of the other sites at the given confidence level. The second approach [Stein and Gordon, 1984] tests if the reduction in χ2 deriving from the use of a two-plate model instead of a single plate, is statistically significant. If the null hypothesis is verified the decrease in χ2 is consistent with the increasing number of model parameters at the given confidence level. The result of this test is not modified if velocity uncertainties have been systematically over or underestimated.
4.1. Realization of Nubia and Eurasia Reference Frames
 We considered 60 sites on the Eurasian plate (Figure 3) far from actively deforming boundaries and from the Fennoscandian platform, where horizontal velocities may be significantly affected by glacial isostatic adjustment (GIA). However possible effects of GIA in central Europe have been shown to have a minor influence on horizontal velocities [Nocquet et al., 2005]. The inversion for the Eulerian pole yields a χν2 (χ2 divided by the degrees of freedom) of 1.47 with a WRMS of 0.3 mm/a for both the east and north components (residuals to the Eulerian vectors are given in Table S1 of Auxiliary Material; angular velocities in geographic coordinates relative to ITRF2005 are given in Table 1). For the realization of the Nubia reference frame we considered 22 sites (Figure 3). This data set yields a pole of rotation relative to ITRF2005 (Nu, Table 1) with a χν2 of 0.9 with a WRMS of 0.5 and 0.3 mm/a on the east and north components, respectively.
Table 1. Plate Angular Velocities Relative to ITRF2005 in Geographic Coordinatesa
Azimuth (deg clockwise north)
Plate abbreviations: Eu, Eurasia; Nu, Nubia; Ad, Adria; Ap, Apulia; Hy, Hyblean Plateau; Ap+Hy, Apulia and Hyblean considered as a single plate. Ω is positive for counterclockwise rotation; σmaj and σmin are the 1σ lengths in degrees of the major and minor axes of the pole error ellipse, with the azimuth of the major axis given clockwise from north; σΩ is the 1σΩ uncertainty; Nsites is the number of stations used to estimate plate rotation parameters; and χν2 values are the sum of squared weighted residuals, normalized by the degrees of freedom.
Ap + Hy
 The residual values for the Eurasia and Nubia plates suggest that the rigid plate assumption is respected and that the selected sites define reference frames that meet a condition of no internal plate deformation at the level of GPS velocity uncertainties. The relative angular velocity between the Eurasia and Nubia plates is derived by differencing their ITRF2005 angular velocities, with appropriate error propagation (Table 2). Figure 3 shows the position of the Eurasia-Nubia Eulerian pole together with other recently published solutions.
Table 2. Relative Plate Angular Velocities in Geographic Coordinates
The first plate rotates counterclockwise relative to the second.
(Ap + Hy)-Nu
 To verify that the angular rotations are not critically affected by few sites, in a second inversion we exclude the GPS sites which cumulatively provide more than 50% of the formal data importance [Minster et al., 1974] for each plate. We thus exclude 14 sites for Eurasia and 2 sites for Nubia and repeat the estimation of the Euler vectors. The resulting Eulerian poles do not significantly differ with respect to the full data set estimates and the Eurasia-Nubia relative motion calculated in the central Mediterranean differs by less than 0.3 mm/a in velocity and 2° in azimuth. Our analysis also confirms (Figure 3) the significant discrepancy between the GPS Eurasia-Nubia pole of rotation and the Nuvel1A Eurasia-Africa pole [DeMets et al., 1994], which average plates motion over the last 3 Ma, supporting the hypothesis of a post-3-Ma change in relative plate motion between Nubia and Eurasia [McClusky et al., 2003; Calais et al., 2003].
4.2. Microplates Motion Within the Eurasia-Nubia Plate Boundary: Adria and Apulia
Figure 4 shows maps of the GPS velocity field in various reference frames drawn in an Oblique Mercator projection about the Nubia-Eurasia pole of rotation so that relative plate motion is aligned with the horizontal axes. GPS sites lying on the region previously defined as the Adriatic Promontory show significant residuals in both the Eurasian and Nubian reference frames. In an Eurasian reference frame (Figure 4a) crustal velocities around the Adriatic region are almost perpendicular with respect to the Eu-Nu plate motion direction and show significant gradients in the regions characterized by active seismicity and deformation. As remarked by Anderson and Jackson  the distribution of seismicity defines a relatively aseismic and rigid region along the Po Valley, the Adriatic Sea and the Apulian Promontory. Inverting the velocities of all the sites which lie on the crustal domains previously defined as the Adriatic Promontory (including Apulia and excluding the Hyblean Plateau), for a single Eulerian vector we obtain a best fit pole which describes the rotation of Adria relative to Eurasia at 44.63°N, 6.47°E (Ω = −0.25°/Ma) with a relatively large χν2 of 5.2 and a WRMS of 0.9 and 0.6 mm/a for the east and north components, respectively. Subtracting the obtained rigid rotation the distribution of residuals shows patches of systematic residual motions in the Po Valley and in Apulia. These results suggest that the two regions (Po Valley + central Adriatic and Apulia) are not rotating around a common Eulerian pole. The inversion of GPS velocities for two distinct blocks (Adria, Ad, and Apulia, Ap), significantly increases the quality of the fit and eliminates systematic patterns of residuals (Figures 4b and 4d). The inversion of 22 sites lying south of the Alps and far from the active Apennines and Southern Alps front, yields a pole of rotation EuΩAd (Figure 4b) for the northern part of the Adriatic (Ad) located at the western end of the Po Valley with a significantly improved fit (χν2 = 1.2, WRMS 0.3 and 0.4 mm/a for the east and north components). The fit for an Eulerian pole of Apulia (selecting 7 GPS velocities on the Apulian Promontory) yields a very good fit and low residuals (χν2 = 1.2, WRMS 0.4 and 0.3 mm/a for the east and north components, respectively). We tested the significance of the χ2 decrease from a solution with a single Adriatic block to a solution where the Adriatic region is divided in two blocks using the previously described F ratio test [Stein and Gordon, 1984]. The F statistics, defined as (((χ1plate2 − χ2plates2)/3)/χ2plates2/55), is 61.87 implying that the χ2 decrease is significant at a confidence level higher than 99.99%. The improvement in fit to the GPS data resulting from the assumption of an Apulian microplate distinct from Adria thus largely exceeds that expected purely by chance because of the introduction of the three additional parameters associated with an additional Euler vector. Thus the additional microplate and its motion are kinematically resolved, and our results support the hypothesis of the separation between the northern and southern Adriatic region. The residual velocities in Apulia relative to Adria (Figure 4b), define a clockwise rotation around an Eulerian pole located on the Dalmatian coast near Dubrovnik.
 In a Nubia reference frame (Figure 4c) GPS sites in the Hyblean region show a systematic eastward motion, consistently with 1.5–2 mm/a of oblique opening in the Sicily Channel rift zone. These same sites show minor residuals (0.5–1 mm/a) relative to Apulia (Figure 4d) which may suggest a kinematic continuity between Apulia and the Hyblean Plateau through the undeformed portion of the Ionian Sea. The application of the F test to the Hyblean sites shows that (1) the null hypothesis (Hyblean velocities consistent with the Apulian angular velocity) cannot be rejected at a significance level higher than 25% and (2) the reduction in χ2 deriving from the inversion of two distinct poles for the Hyblean and the Apulian sites, respectively, is not statistically significant at a confidence level higher than 72%. Figure 5 shows that the Eulerian poles describing the rotation of Apulia (NuΩAp) and the Hyblean Plateau (NuΩHy) relative to Nu, are very close and their error ellipses overlap. This observations suggest that, to first-order, Apulia and the Hyblean Plateau may kinematically belong to the same crustal block, which include the Apulian region, the Ionian Sea area and the southernmost part of Sicily (Hyblaen Plateau).
 Another important observation is the presence of systematic residuals in the northern Apennines (relative to Adria (Figure 4b)) and in Calabria (relative to Apulia, Figure 4d), directed toward the Adriatic and the Ionian Sea, respectively. These residuals are observed only where active crustal compression, subcrustal seismicity or a Wadati-Benjoff zone have been documented [Selvaggi and Amato, 1992; Selvaggi and Chiarabba, 1995], suggesting a possible outward motion of the northern Apennines and the Calabrian Arc as a result of a still active subduction of the Adriatic and Ionian lithosphere beneath the northern Apennines and the Calabrian Arc, respectively.
Figure 6 shows a possible and simplified interpretation of the present-day kinematics and microplates distribution of the Adriatic promontory. In Figure 6 the inactive portions of the Tyrrhenian subduction (see Figure 2) zone are in gray while in black we emphasize those sectors where the Adriatic-Ionian lithosphere may still passively sink in the mantle [Malinverno and Ryan, 1986] and subduction may be active.
Table 3 reports for comparison the poles of rotation of Adria relative to Eurasia previously proposed by various researchers. Most of these poles are however obtained using sites in Apulia (MATE), which we suggest to be a distinct microplate with respect to Adria.
Table 3. Comparison of Eurasia-Fixed Angular Velocities of the Adria Microplate
WA94 (Matera and Medicina Very Long Baseline Interferometry sites)
WE92 (earthquake slip vectors)
CA02 (2 GPS sites and AJ87 slip vectors)
BA03-AP (including sites on Apulia)
GR05-Adria2 (2 GPS sites)
GR05-Adria4 (4 GPS sites)
GR05-Adria5 (5 GPS sites)
GR05-Adria6 (including MATE)
DA05 (14 GPS sites)
SE05 (9 GPS sites including 4 on Apulia)
This study (22 GPS sites in the Po Valley)
This study (earthquake slip vectors)
5. Earthquake Slip Vectors
5.1. Data Selection
 We selected 58 events (Table S2 of the Auxiliary Material) along the active margins of the Adriatic region from the Italian CMT data set [Pondrelli et al., 2006] (data are available at http://www.bo.ingv.it/RCMT/Italydataset) which merges global Harvard CMT moment tensors with regional CMTs from the Mediterranean area for Mw ≥ 4.5. Since moment tensor catalogues contain data with differing quality, we selected the most reliable moment tensors using the criteria described by Frohlich and Davis . We have considered three parameters: the CLVD component fCLVD that measures how much the moment tensor deviates from a pure double couple, the relative error Erel that is a measure of the relative size of the moment tensor and its standard error tensor, and the number of unfixed moment tensor components Nfree. We selected moment tensor solutions with fCLVD ≤ 0.20, Erel ≤ 0.15 and Nfree = 6. Following Seno et al.  we assigned an error of 15° to slip vectors derived from solutions with fCLVD ≤ 0.14 and 20° for fCLVD > 0.14. Surface ruptures, aftershock distributions and local tectonic setting have been used as a guide to resolve the ambiguity among the two focal planes. However the large majority of the selected focal mechanisms are characterized by dip-slip motions with an azimuthal discrepancy between the two slip vectors less than the associated uncertainties, so that errors in the choice of focal plane would have only minor effects. About 40% of the slip vector data set considered in this study coincides with the 23 slip vectors data set used by Anderson and Jackson  and Calais et al.  to estimate the Eulerian pole of Adria relative to Eurasia. Selected earthquakes, their id, location, azimuth and associated uncertainty are listed in Table S2 of the Auxiliary Material.
 The estimate of the motion of Adria relative to Eurasia using earthquake slip vectors, relies on the underlying assumption that the considered slip vectors directly represent the motion of Adria relative to Eurasia (i.e., the neighboring regions belong to Eurasia). This assumption may not be fully respected (see also Babbucci et al.  for a discussion) because (1) shortening in the northern Apennines may bias the direction of extensional slip vectors, (2) GPS sites along the Tyrrhenian side of the Apennines show a systematic residual motion toward NW of ≈1 mm/a relative to Eurasia (Figure 4a), and (3) active deformation on the northeastern boundary of Adria seems to be widely distributed within the Balkan Peninsula. The extent to which the above assumption is violated (i.e., the neighboring regions do not move with Eurasia), can invalidate the use of slip vectors to derive the motion of Adria relative to Eurasia. Our approach is thus meant to represent a test of consistency between the GPS-derived motion of Adria relative to Eurasia and the seismically expressed deformation along its margins. The pole of rotation of Adria relative to Eurasia predicts extension along the Apennines and shortening along the Alps and the Dinarides. Active shortening documented in the outer part (Adriatic) of the northern Apennines [Selvaggi et al., 2001] cannot thus be explained by this pole of rotation, but is probably related to the passive sinking of the Adriatic lithosphere beneath the Apennines. We thus exclude a priori the compressional focal solutions in the northern Apennines.
5.2. Slip Vectors Analysis
Figure 7 shows the observed slip vectors and the GPS-predicted directions of motion of Adria and Apulia relative to Eurasia (small circles around their poles of rotation). The general consistency qualitatively suggests that although neighboring regions do not fully correspond with “stable” Eurasia, observed seismic deformation reflects to first-order the motion of the Adriatic and the Apulian blocks relative to Eurasia. This consistency is observed also in the regions where deformation along the Adria boundary is more distributed (i.e., inland of the Dalmatian coast).
 A more quantitative comparison can be obtained from the estimate of the Eulerian pole position (but not its angular velocity) directly from the slip vectors and from a test of consistency between the observed slip vectors and those predicted by the GPS pole. We have estimated the position of the Eulerian pole which best fits the observed slip vectors using the minimization function for azimuthal information described by Chase  and DeMets et al.  and a grid search algorithm. Following our two-microplate Adriatic model we divided the slip vectors data set in two groups: the first data set includes the slip vectors around the Po Valley and the northern Adriatic Sea (i.e., the Adria microplate defined in Figure 6), whereas the second data set contains the slip vectors from earthquakes around the Apulia microplate. The best fit pole obtained from the slip vectors of earthquakes around the northern Adriatic block, is very close and not statistically different from the pole obtained using GPS velocities (Figure 7 and Table 3). The seismic deformation around the Po Valley and the northern Adriatic thus appears fully consistent with the motion of Adria relative to Eurasia. In this sense extension along the Apennines seems more reasonably related to the northeastward motion of Adria rather than to back-arc extension above a retreating subduction zone.
 Thrusting and shortening in the northern Apennines have been attributed to the passive sinking of the Adriatic lithosphere with little influence from regional plate motion [Malinverno and Ryan, 1986]. Compressional earthquakes in the northern Apennines [Frepoli and Amato, 1997], are generally deeper than the normal faulting events along the crest of the Apennines and are generally located at hypocentral depths larger than 20 km [Pondrelli et al., 2006]. Active compression, driven by the subduction of the Adriatic lithosphere beneath the Apennines, can thus coexist with active extension associated with the divergent motion of the Adriatic microplate relative to the Tyrrhenian Sea (Figure 6). The compressional focal solutions in the northern Apennines, in our opinion, thus represent a second-order process probably representative of the slowing and decaying subduction of the Adriatic lithosphere [Selvaggi and Amato, 1992; Levin et al., 2007].
 The geometry of the slip vectors around Apulia does not provide enough information for a reliable estimate of the Eulerian pole with compact confidence limits. On the other hand a test of consistency between the slip vectors predicted by the GPS pole of rotation and the observed slip vectors around Apulia shows that only 4 slip vector residuals out of 18 fall outside their 2σ (95% confidence level) uncertainty, and only 3 do not pass the F ratio test at the 95% confidence level. Active seismic deformation around the Apulia microplate thus appears to reflect its motion relative to Eurasia although a higher misfit and some systematic residuals in the southern Apennines (predicted directions of motion are rotated counterclockwise relative to the observed slip vectors) and Dinarides may reflect the residual motion of the Tyrrhenian coast relative to Eurasia and more distributed deformation in the Balkan Peninsula.
 The above analysis suggests (1) an independent Adriatic microplate can be resolved from slip vectors, and the resulting Eulerian pole lies within the uncertainty of the corresponding GPS pole; and (2) although the slip vectors around the Apulian region have not enough resolution to provide independent constraints, they are consistent with the GPS-based inference of a kinematically independent Apulian microplate.
6. Active Deformation of the Central Adriatic Sea and the Southern Apennines
 Our GPS velocity solution indicates that two distinct rotating blocks (Adria, Ad; Apulia, Ap) reproduce significantly better the kinematic observations of the Adriatic compared to a single rigid Adriatic block. The motion predicted by their relative pole of rotation should produce deformation which is investigated in this section. Figure 8 shows the central Adriatic with the GPS velocity field (relative to Ad), the AdΩAp pole between Adria and Apulia, the relative convergence directions (small circles around the pole of rotation) and the Regional Centroid Moment Tensor focal mechanisms. The pole of rotation predicts ≈1.5 mm/a of NW-SE shortening in the Gargano Promontory rotating northward to a NNE direction. This prediction is consistent with seismological observation showing NW-SE shortening and right-lateral strike slip on E-W faults in the Gargano Promontory and in the Tremiti Islands [Milano et al., 2005] and N-S to NNE-SSW shortening in the central Adriatic. The observed active deformation does not confirm the NW-SE extension outlined in the Gargano region by remeasurements of the triangulation network [Hunstad et al., 2003] which probably originated from local monument instability, and does not support the hypotheses of active NW-SE extension and left-lateral strike slip on the Mattinata Fault of Billi et al. . Assuming that all the relative convergence is absorbed in the Gargano Promontory, we can put an upper bound of 1.1–1.4 mm/a on the right-lateral strike slip taken up by the east striking Mattinata Fault.
 The velocities which define the rigid clockwise rotation of Apulia rotate to increasing westerly directions causing crustal extension across the Apennines. The highest velocity gradients are here observed along the axial part of the Apennines in the area of maximum historical and instrumental seismic activity, consistent with remeasurements of the triangulation network which documented extensional strain rates in excess of 100 nanostrain per year [Hunstad et al., 2003].
 The role of E-W trending right-lateral strike-slip faulting, responsible for the 2002 Molise earthquake sequence (Mw 6.0, 5.9) and the 1990 Potenza earthquake (Mw 6.0) and associated to the 1456 A.D. earthquake [Fracassi and Valensise, 2007], is not clear. Their contraction P axes are however consistent with the NW-SE shortening along the Apennines caused by the relative motion between Apulia and Adria. Whereas the NW-SE convergence between Ad and Ap is accommodated by thrusting and crustal thickening in the Adriatic and the Gargano region, it appears that more to the south a first-order organization of the faulting is realized by E-W right-lateral strike slip between Apulia and the Apennines and NW-SE normal faults along the crest of the Apennines [Fracassi and Valensise, 2007]. Figure 8 shows a striking parallelism between the predicted convergence directions between Adria and Apulia and the shortening directions defined by the focal mechanisms and by the anticline axes which deform late Pleistocene-Holocene sediments [Ridente and Trincardi, 2006]. In the central Adriatic the relative block motion between Ad and Ap seems thus to be the principal factor controlling the deformation in this part of the Adriatic. The lack of GPS sites north of Gargano promontory, hampers a detailed understanding of deformation distribution. The distribution of seismicity and late Pleistocene-Holocene anticlines, however suggest that the deformation may be diffused over at least 200 km in the north direction and not concentrated on a single tectonic element, such as the proposed Gargano-Dubrovnik line. Because of the proximity of the Eulerian pole the direction and magnitude of relative convergence changes rapidly approaching the Dinarides, where the observed deformation reflects the relative motion between the Adriatic blocks and the Balkan Peninsula [Bennett et al., 2008].
 Different hypotheses have been suggested to explain the active deformation in the central Adriatic and the occurrence of E-W right-lateral faults. Scrocca  suggest that active thrusting in the central Adriatic represents the easternmost contractional front of the Apennines and that E-W right-lateral strike slip accommodates a more pronounced eastward rollback in the northern-central Apennines subduction relative to the southern Apennines. This hypothesis implies a significant eastward shift of the conventionally recognized [Bigi et al., 1990] easternmost thrust front of the Apennines. Right-lateral slip on E-W faults have also been related to intraplate deformation of Adria and NW-SE compression, ultimately related with Africa-Europe plate convergence [Di Bucci and Mazzoli, 2003]. Alternatively we think that the consistency between geodetic, seismological and geological observations, supports a central Adriatic deformation model based on the relative motion between the Adriatic and the Apulian blocks (Figure 8). The way in which these two blocks rotate relative to Eurasia, determines a maximum of 4–5 mm/a of NE-SW extension in the Apennines southwestward of their relative pole of rotation (AdΩAp) (Figure 6). This maximum in crustal extension rate is observed in the GPS velocity field, in the shear strains calculated from remeasurements of the triangulation network [Hunstad et al., 2003] and in the moment tensors of historical seismicity summed over the last 650 years [Selvaggi, 1998].
7. Is There an Independent Apulia-Ionia-Hyblean Microplate?
 The hypothesis that the Adriatic region behaves as two separate submicroplates successfully explains the space geodetic and seismological observations but does not specifically address the problems of the southern boundary of the Adriatic region with Nubia. In our opinion the southern Adriatic, the Ionian Sea and the Hyblean region form a single independent microplate rotating, relative to Nubia, around the GPS-derived Eulerian pole located in the Gulf of Sirte (Figure 5). This hypothesis offers a reasonable explanation for the absence of a well defined southern Adria boundary as the relative motion between Adria and Nubia is now taken up by the rotation of the additional Apulian-Ionian-Hyblean microplate. The Eulerian pole (AdΩAp) describing the motion between the two Adriatic microplates, predicts a maximum relative velocity of about 2 mm/a, which is accommodated by a zone of diffuse deformation in the central Adriatic (Figure 8), and which therefore represents the “missing” southern boundary of Adria. The counterclockwise rotation of Adria relative to Eurasia therefore finds its explanation in the context of a simple model of plate convergence accommodated by the opposite rotations of the two microplates (Figure 6).
 The idea of an independent microplate, including the Ionian Sea and the southern part of the Adriatic Sea, has been suggested by Westaway , who proposed an Eulerian pole describing the rotation relative to Nubia located few hundreds of kilometers south of GPS-derived pole of rotation (Figure 5). The absence of land above sea level south of the Apulian Promontory precludes a direct test of the kinematic continuity between the southern Adriatic, the Ionian Sea and the Hyblean region. Several lines of evidence however suggest a rigid connection between these regions. The seismicity distribution (Figure 1) does not show any evident activity along the Apulia Escarpment, the most prominent morphological structure which has been considered as a possible southern boundary of the Adriatic microplate [Battaglia et al., 2004; Serpelloni et al., 2005]. This evidence is supplemented by seismic reflection profiles [Argnani, 2006] showing flat-lying Plio-Quaternary sediments resting on the escarpment, without any deformation. The argument of a continuous lithosphere between the southern Adriatic and the Ionian Sea is also supported by the efficient propagation of Sn waves from Africa to seismic stations in Apulia, outlining a continuous lithospheric mantle [Mele, 2001].
 The western boundary of the Ionian microplate was traced by Westaway  along the Malta escarpment, a major morphological element with evidence of recent activity [Argnani and Bonazzi, 2005] and supposed to be responsible for the destructive 1693 eastern Sicily earthquake [Piatanesi and Tinti, 1998]. Our results show a limited motion of the Hyblean region relative to Apulia (Figure 4d), and a maximum accumulation rate of 0.5–1 mm/a along the Malta escarpment, which contrast with the larger 1.5–2.0 mm/a extension rate found for the Sicily Channel Rift Zone (Figure 4c). We therefore prefer an interpretation in which the main boundary between the Apulian-Ionian-Hyblean block and Nubia runs along the Sicily Channel Rift Zone (Figure 6).
 The Ionian Sea is the lower subducting plate of the Calabrian subduction zone. Deep earthquakes between 100 and 600 km characterize the northward steeply dipping slab beneath the Tyrrhenian Sea [Selvaggi and Chiarabba, 1995, and references therein]. Although the deep earthquakes and seismic tomography [Lucente et al., 1999] provide clear evidence of the slab, the distribution of seismicity does not show any evidence of shortening in the wedge or plate interface deformation beneath the Calabrian wedge [Chiarabba et al., 2005]. This evidence prompted the idea of a detached slab kinematically unconnected with the Ionian Sea lithosphere [Westaway, 1993; Wortel and Spakman, 2000]. The kinematics of the Ionian Sea is thus critical for evaluating the rate of convergence that is possibly accommodated in the subduction zone and the hazard associated with the rupture of the subduction thrust [Gutscher et al., 2006]. If one regards the Ionian Sea as moving with Nubia this reference frame should be considered (Figure 4c), and at least 4–5 mm/a of motion of the Calabrian Arc are consequently accommodated by oblique convergence in the Ionian wedge (see also Devoti et al., 2008]. If alternatively the Ionian Sea is rigidly attached to Apulia and the Hyblean Plateau (see Figure 6), the reference frame attached to this region (Figure 4d) should be considered and a significantly lower motion of 1.5–2.5 mm/a of the Calabrian Arc toward the trench is observed. This range of values, which represent the motion of the Calabrian Arc relative to the Ionian subducting plate, suggests that subduction may still be active and constrains the shortening taken up in the accretionary wedge which is eventually accommodated by long-term slip on the subduction interface. On the other hand the observed subduction rate is almost an order of magnitude lower than the back-arc opening rate averaged over the Plio-Quaternary, suggesting a marked decrease in the efficiency of the tectonic processes related to the long-lived subduction of the Ionian slab [Mattei et al., 2007].
 On the eastern margin of the Ionian Sea any effects of an independent Ionian block moving at few mm/a relative to Eurasia and Nubia, would be masked by the Aegean Sea moving to SSE at ≈30 mm/a relative to Nubia [Kreemer and Chamot-Rooke, 2004].
8. A Simple Block Model
 In the previous sections we proposed a regional kinematic model for the central Mediterranean based on two independent microplates rotating within the Eurasia-Nubia plate boundary zone (Figure 6). A direct verification of the southward continuity of the Apulian-Ionian block with the Ionian Sea and the Hyblean Plateau is hindered by the lack of islands in the Ionian Sea. Therefore we use a simple block model to illustrate how the plate convergence can be accommodated with a counterclockwise rotation of Adria relative to Eurasia, and a clockwise rotation of a single Apulian-Ionian-Hyblean microplate relative to Nubia. The purpose of this modeling is to illustrate the general features of crustal motion, not to produce a detailed simulation of the velocity field as pursued by more sophisticated block modeling [Reilinger et al., 2006].
 The model is illustrated in Table 4 and Figure 9 and is similar to that proposed by Taymaz et al.  for the Aegean Sea region. The plate boundary is composed of two rigid blocks (L and R) pinned with pivots (poles of rotation AΩL and BΩR) which constrain their ends to move with the adjacent plate on either side. The two blocks are also pinned to each other at the other end (LΩR). The distant EuΩNu pole of rotation allows us to use a flat Earth approximation on an Oblique Mercator projection (the same as used in Figure 4) in which the plate convergence (calculated at the position of the NuΩAp+Hy pole) is parallel to the X axis, and is assumed uniform in the study area. The rotation rates of the left and right blocks, relative to their adjacent major plates, are a simple function of the convergence rate and block geometry, and can be compared to the angular velocities of the EuΩAd and NuΩAp+Hy poles, respectively. The rotation rates of the left (AΩL) and right (BΩR) blocks are:
where aR and aLare the slat lengths projected on the X axis and ϕL and ϕL are the angles made with Y axis. We fix the position of the three poles of rotation which constrain the motion of blocks, impose the convergence vector AVB derived from the EuΩNu Euler vector, and calculate the resulting microplate rotation rates. If the Ionian region forms a single microplate with Apulia and the Hyblean region and if, together with Adria, they rotate to accommodate plate convergence according to the model in Figure 9a, we expect that their observed rotation rates will be similar to the rotation rates calculated with equations (1) and (2). If, alternatively, the Ionian region is not moving with Apulia and the rotations of microplates are controlled by other processes, we do not expect that observed rates will be necessarily similar to the calculated rates. Figure 9 shows indeed that the observed and calculated rotation rates are consistent.
Table 4. Calculated and Observed Microplate Rotation Rates
Observed ± 1σ (deg/Ma)
0.31 ± 0.02 (EuΩAd)
−0.27 ± 0.04 (NuΩAp+Hy)
−0.55 ± 0.05 (AdΩAp+Hy)
 Our results therefore support (1) the presence of a single Apulian-Ionian-Hyblean microplate which rotates around an Eulerian pole located in Gulf of Sirte and (2) a simple model of the plate boundary in which the relative motion between Eurasia and Nubia is accommodated by microplates rotating in such a way as to accommodate plate convergence.
9. Fragmentation of the Adriatic Promontory
 Several lines of evidence suggest that during the Meso-Cenozoic evolution of the Alpine-Mediterranean plate boundary zone the Adriatic region behaved as a rigid promontory of Africa colliding with Eurasia, leading to continental collision, crustal thickening and formation of orogenic belts encircling the Adriatic region [Argand, 1924; Channell et al., 1979]. The present-day occurrence of independent microplates suggests that an important phase of plate fragmentation has separated the original Africa-fixed continental promontory into the present-day configuration.
 The main processes which have controlled the evolution of this part of the Mediterranean are (1) the relatively slow Eurasia-Africa convergence [Dewey et al., 1989] and (2) the simultaneous opposite subductions of the Ionian lithosphere beneath the Aegean Sea and beneath the Calabrian Arc. Both subduction zones have been characterized by back-arc extension and outward motion of the trench [Jolivet and Faccenna, 2000] which have progressively reduced the width of the Adriatic Promontory available to transfer the motion of Africa to the orogenic belts encircling the Adriatic Promontory. The present-day configuration shows in fact only a thin lithospheric corridor in correspondence of the Strait of Otranto where the orogenic belts on the opposite side of the Adriatic are separated by less than 250 km (Figure 2). We suggest that the origin of the fragmentation of the Adriatic Promontory can be sought in the progressive narrowing of the Adriatic “corridor”, up to the point in which the available geometry was no longer compatible with rigid transfer of the Africa motion, resulting in the formation of two independent microplates.
 Is there a relationship between the microplate rigid behavior of Adria, Apulia and the Ionian region and their lithospheric structure? Velocity structures obtained from surface waves are often a good indicator of contrasts in lithosphere properties on the continents [Ritzwoller and Levshin, 1998; Ritsema and van Heijst, 2000]. Ancient, cold, inactive cratonic lithosphere generally has a faster velocity structure than the actively deforming mountain belts [Emmerson et al., 2006]. Figure 9b shows the central Mediterranean part of the Sv wave speed tomographic model DKP2005 of Debayle et al. , which uses regional waveforms of the fundamental and first four higher modes to image the upper mantle at depths up to 400 km. Although the data are smoothed on a horizontal length scale of about 400 km, there is a clear correlation between geological structures and the velocity contrast that marks the boundary of the Adriatic promontory. A clear correlation exist between low S wave velocities and regions of Neogene-Quaternary extension such as Tyrrhenian Sea, Pannonian basin and the Aegean Sea. Regions of intact, relatively undeformed lithosphere correspond conversely to high relative velocity. The most evident high velocity runs along the Adriatic Sea and joins the Ionian Sea through Apulia closely following the paleogeographic distribution of the Adriatic promontory. This correlation suggests that the rigid rotating Adria and Apulia microplates owe their behavior to the relatively rigid lithosphere of the Adriatic promontory Before the fragmentation of the African promontory, Adria was a rigid indenting crustal blocks with associated collisional processes active along its northern and eastern rim. The progressive reduction in width of the colliding indenter caused a fragmentation of the promontory, with the crustal blocks still behaving rigidly but rotating independently around nearby Eulerian poles. Given the Adriatic and Apulia lithospheric structures and the success of the simple block model in Figure 9a in reproducing their observed rotation rates, it seems reasonable that the main forces driving microplates motion are applied at the edges and that other sources, such as basal tractions or horizontal gradients of gravitational potential energy, are not required.
 Given the slow rotation rate of Adria relative to Nubia (<0.3°/Ma), any finite rotation of the microplate is probably under the limit of detection of paleomagnetic investigations (≈10°), and is of limited usefulness to infer the time when Adria started moving independently from Africa. More useful constraints can therefore be sought in the geological history of the orogenic belts surrounding the Adriatic region. Mountain building in the Western Alps has been related to the convergence between the Adriatic promontory and the Eurasian continent [Schmid et al., 1996, 2004, and references therein]. Stratigraphic and structural evidence suggest that shortening in the Jura thrust-and-fold belt was active as late as the early Pliocene, somewhere between 9 and 3.3 Ma [Becker, 2000], whereas thrusting is still active along the southern front of the eastern Alps [Galadini et al., 2005; Benedetti et al., 2000; D'Agostino et al., 2005]. In the southern part of the Adriatic promontory a strong phase of extension and magmatism affected the Sicily Channel Rift Zone from the Messinian [Argnani, 1990; Torelli et al., 1995] with an extension direction consistent with the present-day pole of rotation of the Apulian-Ionian-Hyblean block. We thus speculate that the end of convergence in the Western Alps, and the opening of the Sicily Channel Rift Zone may be associated with the fragmentation of the Adriatic promontory in two main blocks: the northern one (Adria) started a counterclockwise rotation relative to Eurasia, while the southern one (Apulia, Ionian, Hyblean) started a clockwise rotation relative to Nubia according to the model in Figure 9. At present the motion of Nubia relative to Eurasia is thus no longer rigidly transferred along the Adriatic promontory and absorbed uniformly along its northern rim, but is accommodated by the opposite rotations of the two microplates.
 We have investigated the kinematics of the Adriatic region, its role within the central Mediterranean plate boundary and the relationship with the Eurasia-Nubia plate convergence. Using episodic and continuous GPS observations and earthquake slip vectors we confirm that the crustal motion of the northern Adriatic region can be described as a rigid rotation around an Eulerian pole located at the western margin of the Po Valley. We propose that this microplate is limited to the south by a region of diffuse deformation in the central Adriatic which separates the Adria microplate from a newly defined microplate which includes the Apulian promontory, the Ionian Sea and, possibly, the Hyblean region in southern Sicily.
 The pattern of active deformation in the central Adriatic and the occurrence of E-W right-lateral strike slip in the southern Apennines are explained in terms of the relative motion between these microplates. A rigid block model illustrates that the observed microplate motions are consistent with a simple model of the plate boundary in which two oppositely rotating microplates accommodates the relative plate motion, providing also a simple explanation for the lack of a well-defined boundary between Adria and Nubia. At the scale of the central Mediterranean plate boundary the present-day microplate configuration appears to be the result of the progressive narrowing and final fragmentation of the Adriatic promontory.
 We suggest that the subduction of the Ionian and Adriatic lithosphere beneath the back-arc Tyrrhenian basin, which has dominated the Neogene evolution of this part of the Mediterranean, now plays a second-order role in controlling the active deformation of the Eurasia-Nubia plate boundary, but may still be locally active in the northern Apennines and in the Calabrian Arc.
 Overall the fragmentation of the Adriatic promontory, the resulting formation of microplates and the accommodation of the Eurasia-Nubia convergence seem to be the main factors controlling the velocity field and the active deformation within the plate boundary.
 We would like to thank the staff involved in the development and maintenance of the RING GPS network and all the agencies that made available the GPS observations used in this work. We thank James Jackson and Dan McKenzie for numerous discussions and Don Argus for a careful review of the paper. James Jackson is thanked also for a review of an early draft of the paper. We thank Geoff Blewitt for the Ambizap algorithm and for useful discussions. We thank Enrico Serpelloni and Marco Anzidei for INGV campaigns data.