Main

Deep to intermediate seismicity in the Mediterranean is restricted to the Hellenic and Calabria Wadati–Benioff zones and locally beneath the Alboran region8, although fast seismic velocity anomalies9,10,11 run discontinuously all along the alpine Tethyan suture, from North Africa to the Calabria–Apennines, twisting beneath the Alps, the Carpathians–Dinarides–Hellenides, and further to the east joining the Bitlis–Zagros zone. Most of these fast anomalies are not connected to the surface. This indicates that the once almost-continuous subduction zone is now fragmented, owing to the entrance at the trench of small continental fragments under the slow (∼1 cm yr−1) Nubia–Eurasia convergence9,12. Eventually, this process generates new plate boundaries, producing the complex pattern observed today (Fig. 1). For example, the collision along the Zagros–Bitlis belt induced the westward motion of Anatolia and the formation of the North Anatolian fault3. Soon after, Aegea separated from Anatolia, moving southward towards the Hellenic trench. To the west, the arrival of the African continent at the trench locked the Calabria–North Africa subduction zone, generating a new convergent boundary along the pre-existing passive African margin (Fig. 1). Although the overall kinematics of these processes are now fairly well established, the dynamics are still debated. Proposed models include gravitational potential energy contrasts within the crust13, plate (edge force) interactions, subduction12 and large-scale mantle flow14.

Figure 1: Topography and deformation indicators for the Mediterranean.
figure 1

a, GPS measurements (yellow arrows, Eurasia fixed reference frame24) and seismicity (dots, colour-coded by depth) with magnitude M > 5 (ref. 8), colour-coded by depth. z, Elevation. b, Residual topography relative to a regional mean, estimated by correcting for isostatic adjustment using the CRUST2.0 model30; with plus and minus symbols showing uplift and subsidence, respectively (see text for references).

PowerPoint slide

Full size image

To quantify the contribution of mantle circulation to surface tectonics, we compute instantaneous, three-dimensional spherical mantle flow driven by temperature (density) anomalies as inferred from seismic tomography, assuming that velocity anomalies are simply related to temperature5. Flow produced by density anomalies within the mantle stresses the lithospheric plates and produces both horizontal and vertical motions. For the Mediterranean case, we leave the Adria and Anatolia–Aegea microplates free of shear stress (white arrows in Fig. 2a) so that they can be pulled by the mantle, while the lithospheres of Nubia, Eurasia and Arabia are prescribed to move with NUVEL-1A model velocities15 (grey vectors, applied only at the surface), leaving the underlying mantle flowing freely. The normal stresses generated by viscous flow at the surface are used to infer the equivalent ‘dynamic’ topography. Assuming no erosion, the dynamic topography can be compared to the ‘residual’ topography (Fig. 1b), calculated by removing the isostatic adjustment from the topography6,7.

Figure 2: Reference flow model for the Mediterranean region.
figure 2

The model is based on the tomography model of ref. 11. a, Layer viscosity η and predicted surface velocities of Adria and Anatolia–Aegea (white vectors). Geodetic velocities (orange vectors) and prescribed velocities (grey vectors, from NUVEL-1A) are in the Eurasia fixed reference frame. Plate boundaries are treated as weak narrow belts, with a simplified geometry merging the eastern Alps with the Dinarides. b, Predicted dynamic topography (zdyn is dynamic elevation, boundary in green). c, d, Horizontal (white vectors) and vertical flow (background colour shading) field at 250 km (c) and 600 km depth (d). e, Cross-sections from Massif Central to Calabria and to Anatolia (marked in a), with temperature (normalized by a reference temperature) in the background.

PowerPoint slide

Full size image

Figure 2 shows mantle velocities for the reference model in a Eurasia fixed reference frame. At shallow depths (∼100–300 km, Fig. 2c), the regional flow field shows a large-scale, toroidal current from Arabia to the western Mediterranean perturbed by density anomalies associated with the Hellenic slab, Calabrian slab and the western Alpine slab. Subduction zones generate downward pull towards the inner portion of their respective arcs, where deep mantle slabs accumulate (Fig. 2c), ponding in the transition zone (Fig. 2e). Downward flow is accompanied by an upwelling return flow between slabs and at their edges. In cross-section, this pattern corresponds to an undulated flow confined to the uppermost portion of the mantle (Fig. 2e). Details of the flow and topography depend upon assumptions such as the seismic velocity anomaly to density scaling, and the coupling of the slab to the plate, but the overall patterns are insensitive to assumptions about rheology (Supplementary Fig. 3), or the adopted tomographic model (Supplementary Fig. 4). Figure 2e shows that flow is at an angle to the ∼25–30 mm yr−1 retreating Hellenic slab, but parallel down-dip to the stationary Calabria slab. This agrees with theoretical expectations of slab-induced flow16, indicating differences in slab force transmission. The subduction-related downward flow induces pronounced (∼1 km) negative dynamic topography within the non-compensated17, dynamically subsiding, southern Tyrrhenian and Cretean sea backarc basins18.

The pull exerted by the retreating Hellenic slab, coupled with an upwelling in the Middle East, helps propel Anatolia westward (Fig. 2a)—why it moved westward has been the subject of controversy over the past few decades4. Our model shows a reasonable fit with geodetic data, indicating that Anatolia’s motion is about equally driven by slab anomalies (Fig. 3a), which leads to large-scale toroidal flow at depth, and by the indenting Arabia plate (Supplementary Fig. 3a), which leads to lithospheric push. The Bitlis suture in eastern Anatolia acts as a more efficient strain guide when the plates are coupled (that is, no weak zone on the plate boundary; Fig. 3b). The mantle upwelling beneath the Near East is a particularly pronounced feature, extending from north Arabia to eastern Anatolia and the Caucasus. A comparison of dynamic and residual topography indicates that this upwelling is responsible for the uplift and elevation of the whole region, notably of the Anatolia high plateau, and also for intraplate volcanism19 (Fig. 1b). Our results indicate that this upward flow is probably rooted in the transition zone (Fig. 2e), and could be triggered by the subducted Hellenic (Tethyan) slab. However, its far-field linkage with the large-scale upwelling over the east Africa–southeast Arabia plate cannot be ruled out7.

Figure 3: Additional flow models for the Mediterranean.
figure 3

Surface velocities (as in Fig. 2a) are superimposed on dynamic topography (as in Fig. 2b). a, Flow (white vectors) generated by density anomalies only (that is, surface held fixed outside the microplates). b, Reference model with Bitlis and Alps collisional zones coupled (weak zones removed within dashed regions). Geodetic velocities (orange vectors) and prescribed velocities (grey vectors, from NUVEL-1A) are in the Eurasia fixed reference frame.

PowerPoint slide

Full size image

Our model fails to reproduce the fast southward motion of Aegea with respect to Anatolia, because the mantle anomaly is positioned mostly beneath the western side of the Hellenic trench (Supplementary Fig. 1a), such that little southward pull is induced. This may be due partially to a simplified treatment of the trench without faults, and rheologically more realistic cases with stiff slabs and resulting asymmetric force transmission, or density anomalies inferred from slab seismicity slightly improves our solution (Supplementary Fig. 3d). This indicates that the motion of Aegea is probably dominated by crustal gravitational potential energy variations not included here. Indeed, the crustal column of the Aegean is probably thickened and weakened enough to be prone to extension towards the Hellenic trench13.

Between the Calabria–Apennines and the Hellenides–Dinarides subduction zones, the mantle is upwelling, reflecting both passive return flow squeezed by two opposite-facing retreating slabs, and active flow due to shallow-mantle, low-velocity anomalies (Fig. 2). The details of the dynamic topography depend on the adopted tomographic model (compare Fig. 2b and Supplementary Fig. 4a, e), but the general features are robust and explain the uplift observed over southern Italy20 and intraplate rifting and volcanism in the Sicily channel21. Free air gravity analysis22 and mantle flow computations23 extend this dynamic signature also into the northern Apennines. The opposite-facing slabs anchor the small Adria plate. As inferred from geodesy, Adria indeed moves slowly and its Euler vector is not coherent with the motion of Nubia24,25. We found that Adria kinematics can be better fitted by adding the eastward pull of the Hellenic slab to the northward Nubia motion (compare Figs 2a and 3a). This provides an alternative explanation to the in-plane block-interaction model25. Geodetic data show that the northern portion of Adria is moving north-northeast (Fig. 1), revealing a recent change from the westward motion deduced by fault kinematics in the southern Alps. Our results show that this earlier motion could be related to the pull exerted by the mantle beneath the Alps (Fig. 3a), whereas the present-day motion can be reproduced by coupling the plate boundary beneath the Alps (removal of the northern weak zone; Fig. 3b).

Just as in the Middle East, a vigorous upwelling is present in the western Mediterranean, from southern Iberia to the Massif Central. Here, it is apparent that this flow is connected to the western Mediterranean high-velocity anomaly, forming an upwelling–downwelling cell restricted to the uppermost mantle (Fig. 2c and e). Residual topography (Fig. 1b) matches these non-compensated features beneath eastern Iberia, suggesting that the origin of Quaternary alkaline volcanism, and of the uplift of the ∼850-m-high plateau-like mesetas26 is related to mantle dynamics. The Massif Central represents a similar, but more focused, feature in which the long-lived alkaline volcanic province21 has been previously related to a mantle plume. In contrast, we suggest that volcanism can be caused by decompression melting related to a large-scale return flow at the trailing edge of the western Mediterranean slab (Fig. 2e). SKS wave splitting anisotropy supports this hypothesis, because splitting directions beneath the Massif Central are parallel to the flow generated by the retreating Calabria slab27.

To isolate the contribution of mantle flow, we simplified our model by excluding crustal effects such as gravitational potential energy contributions. Nevertheless, based on the fairly high-resolution (down to ∼100 km) tomography beneath the Mediterranean upper mantle9,10,11, we are able to analyse the signal generated by regional mantle anomalies within a complex tectonic region. Mantle circulation induced by slab fragments within a mobile zone seems to be more important than previously thought, and generates vigorous upwellings even far from the subduction zone (Fig. 4). We can contrast those short-wavelength currents originating between the base of the lithosphere and the transition zone with those associated with the longer-wavelength, subdued mantle anomalies below ∼400 km. In the Mediterranean region, irrespective of the adopted rheological stratification, such small-scale convection appears to be dominant in the origin and mobility of crustal fragments and for the large uncompensated topography (Fig. 4).

Figure 4: Cartoon illustrating the architecture of the subduction zones and the related pattern of mantle convection in the Mediterranean region.
figure 4

Vigorous convection (red arrows) is mainly confined to the uppermost portion of the upper mantle and related to subduction (white arrows) within the Africa–Eurasia convergence (black arrows) . We note also the large-scale toroidal flow beneath the Middle East.

PowerPoint slide

Full size image

This multi-scale pattern of convection may well apply to other mobile belts such as the North American Cordillera or the Himalayan–Tibetan belt. Also in these regions, discontinuously subducting slab segments could excite vigorous flow confined to the uppermost mantle—which may explain the lateral escape of crustal fragments and produce pronounced topographic signals. We speculate that this vigorous small-scale convection could be the ultimate cause of the distribution of convergent deformation inside continental domains.

Methods Summary

To model mantle flow, we solve the equations for instantaneous fluid circulation using the global, finite-element code CitcomS28. Mechanical boundary conditions are of mixed type: although the velocities of large plates are prescribed at the surface, the regions shown by white vectors in Fig. 2a are left free of shear stress (that is, free slip, allowing for dynamically consistent regional plate motions). Internally, flow is additionally driven by density anomalies, which are inferred by scaling velocity anomalies from seismic tomography to temperature29. Results shown here are based on a mesh resolution of ∼20 km, and tests indicate that overall model uncertainties are mainly due to imperfect knowledge of input models, such as tomography, and are not due to computational limitations.

Our reference density structure is based on recent, global P-wave tomography11, which provides, in general, high-resolution (down to ∼100 km) images of subduction zones. The rheology of the mantle is Newtonian viscous and most models have only radial-viscosity variations besides weak zones confined to the lithosphere (down to 100 km depth), have 100 km width, and a viscosity reduction to 0.01 of the ambient viscosity. The reference, radial-viscosity structure is 5 × 1022 Pa s in the lithosphere, 1021 Pa s from 100–660 km, and 5 × 1022 Pa s in the lower mantle. Tests using different rheology and input model choices are described in the Supplementary Information.