Introduction

Anatolia is a micro-plate between Eurasia and Arabia. Its tectonics is dominated by the collision of the Arabia and Eurasia plates on its eastern boundary (Fig. 1), which leads to the westward motion of the Anatolian Block, as constrained by Global Positioning System (GPS) measurements1,2,3. The extrusion of Anatolia is also greatly influenced by a complex subduction system in the eastern Mediterranean on its western and southern boundaries4. Anatolia is considered to be relatively simple in terms of deformation and can be explained by a simple block model, with a uniform pattern of westward ground velocity of the block3,5. Its northern frontier is the North Anatolia Fault, a right-lateral strike-slip boundary fault between the Anatolia and Eurasia plates, whereas its boundary with Arabia Plate is the left-lateral East Anatolian Fault5. The Anatolian Block has a more or less uniform crustal thickness of 40–50 km6,7,8, and the thickness of lithosphere in the region is 80–100 km9,10.

Fig. 1: Geodynamics of Anatolian Plate.
figure 1

a Regional tectonic map of Anatolia. Green lines are plate boundaries in the PB2002 model60, and white arrows show velocity vectors of the Anatolia and Arabia plates relative to the Eurasia Plate61. Volcanoes62 are displayed with pink triangles with vents. b Locations of the suspected slab tears (inverted blue triangles) beneath Anatolia36. c Surface velocity vectors (blue lines with uncertainty ellipses) extrapolated in and around the Anatolia Plate from GPS measurements2. d Seismic anisotropy estimated by SKS/SKKS shear-wave splitting measurements18. The black lines represent the fast directions of propagation, and the sizes of the red circles are proportional to the splitting times.

Full size image

Seismic anisotropy in the crust is mostly attributed to the preferred orientation of anisotropic minerals, oriented cracks, lateral or vertical variations of compositions and structures11,12. In the upper mantle, seismic anisotropy is mainly attributed to stress or strain-induced alignment of anisotropic minerals13,14,15,16,17, and has been used to infer the mantle flow and its relation with plate motions18,19. The SKS-based studies concluded that the anisotropic pattern is mostly controlled by a complex mantle flow, and the local perturbations originate from slab tears19. Seismic anisotropy can be deduced from laboratory experiments (elastic anisotropic behavior of rocks and minerals), numerical modeling (solution of the full elastic (anisotropy) wave equation), and seismological investigations15,20,21. In the Anatolia region, the fast axis derived by SKS/SKKS shear-wave splitting measurements18 is rather homogeneous with a direction of N30∘E (Fig. 1d). Other studies investigating the anisotropic structures of the region mostly focused on Pn polarization22,23, providing the average anisotropy over the crust and upper lithosphere, or on surface-wave anisotropy of the crust24. Those studies found a complex anisotropic pattern within the upper crust, with mostly EW fast directions of anisotropy in eastern Anatolia, and NS to N170∘E fast directions of anisotropy in western Anatolia.

Results and discussion

In this study, we investigate the vertical layering of anisotropy in the crust as well as the lithospheric and asthenospheric mantle beneath Anatolia by building maps of the fundamental-mode Rayleigh-wave phase velocity in the region using both ambient noise and teleseismic records from the available regional networks (Fig. S1). We adopt a two-station technique in measuring the interstation dispersion curves by cross-correlation approach (Figs. S2 and S3), resulting in a set of 1589 dispersion curves from ambient noise24 and 2604 curves from teleseismic events (Fig. S4). Over 5750 teleseismic events were used to constrain our dispersion curves. Those events occurred between 1999 and 2015, with magnitude Mw > 5.5, located near the great circle linking two stations, and with an epicentral distance greater than ten times the interstation distance. Ambient noise cross-correlations of seismic records from regional networks provide measurements of short-period (5–25 s) dispersion curves, mostly sensitive to crustal and upper lithospheric structures25,26. We augment noise data with earthquake records to measure dispersion curves of longer periods (20–300 s), which are sensitive to the crust, lithosphere, and asthenosphere27,28. The overall period range is 5–300 s (Figs. 2 and S5), and the range of depth sensitivity of our dataset is 5–400 km.

Fig. 2: Layering of seismic anisotropy at selected periods.
figure 2

Results of Rayleigh-wave phase velocity and anisotropy for the crust (25 s, a) and the lithospheric (60 s, d) and asthenospheric (150 s, g) mantle, with corresponding depth-dependent sensitivity kernels in (b), (e), and (h), respectively, in which the color-shaded areas indicate the estimated depth ranges of main sensitivity. The Moho depths beneath eastern and western Anatolia are also marked. Anisotropy averaged over the whole Anatolia region in crustal (5–35 s, c), lithospheric (50–100 s, f), and asthenospheric (120–300 s, i) mantle depth ranges.

Full size image

We then invert the collection of dispersion curves for maps of both isotropic and azimuthally anisotropic anomalies of Rayleigh-wave phase velocity at selected periods (5–200 s every 5 s, and 210–300 s every 10 s)29,30,31. Parametrization and resolution are discussed in Supplementary Information (Figs. S6–S12). The resulting anisotropic Rayleigh-wave phase-velocity maps (Fig. S13) allow us to analyze the patterns of seismic anisotropy in different layers (Fig. S14) beneath Anatolia and understand possible contributions to the anisotropy.

Crustal anisotropy

The crustal thickness in Anatolia is mostly uniform, gradually changing from 27 km in the west to 50 km in the east6,7,8,32,33,34. This depth range is well sampled by Rayleigh waves of periods of 5–35 s (Fig. 2).

A recent study24 has been focusing on the variations of crustal velocities and anisotropies derived from surface waves beneath Anatolia. Their results revealed a connection between the variations in isotropic and anisotropic crustal velocities and the locations of faults and melts. They also observed a general EW fast direction of anisotropy at large scale, supplemented by local variations in the anisotropic pattern.

In our updated model, large-scale patterns exhibit mostly EW-oriented fast axis of anisotropy in western, central, and souteastern Anatolia (Fig. 2a), which may be related to the westward extrusion of Anatolia and the stress induced in the crust. The average anisotropy (average azimuthal direction and average amplitude over all points in the model for a specific period) in the whole region in the period range of 5–35 s is mostly EW (Fig. 2c), inconsistent with the shear-wave splitting derived anisotropy orientation of N30∘E (Fig. 1d), but in good agreement with the westward surface motion of Anatolia from GPS observations (Fig. 1c). The lateral variation observed in the crustal anisotropy pattern is also absent in the previous shear-wave splitting measurements (Fig. 1d). These discrepancies imply that the shear-wave splitting measurements in Anatolia are not strongly influenced by crustal structures, although considerable (non-negligible) contribution from the crustal structures on splitting measurements had been reported in other tectonic regions35.

Lithospheric anisotropy

In the period band of 50–100 s, Rayleigh waves are mostly sensitive to depth range 50–120 km28, they sample the lithospheric mantle. In this period range in our model, the anisotropic pattern is relatively simple (Fig. 2). A NS-oriented anisotropy, coupled with negative velocity anomaly, is present in eastern Anatolia and in the northern Arabian Plate, whereas the anisotropy in the region south of central and in western Anatolia shows a slightly different pattern, having mostly EW and NE-SW orientations with local variations that may be attributed to mantle heterogeneities. For instance, slab tears36,37 are suspected to be present beneath Anatolia (Fig. 1b), and the probable locations of the slab tears are in good agreement with the velocity anomaly and the change in anisotropy in central Anatolia (37∘E, 40∘N in Fig. 2d), highlighted by a relatively strong heterogeneity in anisotropy in western Anatolia (29∘E, 37∘N in Fig. 2d and g).

Although the average anisotropy over the whole region we obtained in periods of 60–70 s is consistent with the shear-wave splitting measurements (Fig. 1d), other periods sampling the lithosphere (60 s and 80–100 s, Fig. S13) yield directions that are perpendicular to the anisotropy displayed by the shear-wave splitting measurements.

The most prominent feature at periods of 50–100 s is the strong dichotomy observed in our models. Beneath central and western Anatolia (Fig. 2d), EW anisotropy is present, whereas beneath eastern Anatolia and northern Arabia, mostly NS anisotropy is found (Figs. 2d and S13).

Such clear contrast is absent in the shear-wave splitting measurements (Fig. 1d). The strong dichotomy between Anatolia and Arabia in terms of anisotropy for periods of 50–100 s as well as the small-scale lateral variations of anisotropy would also rule out the lithosphere as the main contributor of the shear-wave splitting measurements in the region.

Our result also reveals a strong difference between the pattern of seismic anomaly and fast direction observed beneath Anatolia and Arabia for periods of 50–100 s. The N020∘E direction of anisotropy beneath eastern Anatolia agrees well with the hypothesis of lithospheric delamination caused by the thermal erosion of the lithosphere in the region38,39,40. Beneath the northern Arabia Plate and eastern Turkey where North Anatolia Fault and East Anatolia Fault meet the Bitlis suture, a strong negative velocity anomaly is found (Figs. 2d and S13), indicating the presence of a warmer mantle or asthenospheric material. The slow anomaly is well correlated with the region of wide-spread post collisional volcanism34,40, and the anisotropy appears stronger near eastern Turkey where the age of the volcanism is relatively young (≤6 Ma). The amplitudes of the anisotropy are rather limited in most parts of the region south of Bitlis and East Anatolia Fault, suggesting the presence of melts and probable vertical mantle flow. In our 2D tomographic inversion, vertical flow would have a predominant vertical component which would result in minor amplitudes in the horizontal components of the anisotropy. In Figs. 2d and g, around 37∘E, where slab tear is proposed, limited amplitude of anisotropy might be explained by vertical flow.

Asthenospheric anisotropy

At periods sampling the asthenosphere (120–300 s, Fig. 2g), the anisotropic pattern shows a NW-SE orientation in the central and western parts of Anatolia, while in a small region in eastern Anatolia displays NE-SW orientation. In northwestern Anatolia, the anisotropic pattern is mostly of NW-SE orientation. The isotropic part of our model is still well consistent with previous P-wave tomographic studies in the depth range of 200–320 km36. In central and western Anatolia, in particular in the coastal areas, we observe a slight change in the anisotropic pattern, which is mostly in EW direction. Local variations of velocity and anisotropy may still be attributed to mantle heterogeneities, such as the Hellenic and Cyprean slabs and suspected slab tears (Fig. 1b). A major slab tear is thought to be present in central Anatolia (37∘E, 40∘N)36. Some small-scale perturbations of the anisotropic pattern and slow velocity anomaly in this area are also present in our model. In regions with fast velocities associated with subducting Aegean and Cyprus slabs (Figs. 2g and 3e), the anisotropy orientation is mostly NE-SW; whereas in regions with low velocities, potentially related to slab tears, the anisotropy is NW-SE. This may be linked to a mixture of horizontal flow in the region with a mantle toroidal flow in eastern Mediterranean, in addition to the vertical flow related to the slab system. The low-velocity anomaly beneath the presumed slab tears (37∘E, 40∘N and 29∘E, 36∘N) may be related to the upwelling of hot mantle in the complex slab system, which is also suggested by the anisotropic pattern. The presence of a slab tear may be supported by the toroidal flow in the asthenospheric mantle, indicated by a fan-shaped anisotropic pattern diverging near the tear (Fig. 3c). In the asthenosphere sampled by Rayleigh waves in the period band of 120–300 s, the average anisotropy in our model has a mostly NW-SE orientation (Fig. 2g–i), perpendicular to the NE-SW direction of anisotropy measured by shear-wave splitting, but in agreement with the average anisotropy in the region in the lithosphere (Fig. 2d and f). Therefore, it seems that the asthenosphere and lithosphere together in the upper mantle still make a limited contribution to the average anisotropy as measured by shear-wave splitting. The uniformly NE-SW direction of shear-wave splitting anisotropy in the whole region suggests that the source of anisotropy is probably deep from a large-scale process.

Fig. 3: Layering of seismic anisotropy in the crust and upper mantle.
figure 3

Sketches of the anisotropy in a the crust (blue arrows), b the lithosphere mantle (green arrows), c the asthenosphere mantle (orange arrows), and d the mantle transition zone (red arrows). e Perspective view of the shear-wave splitting measurements (black lines on the map) and the morphology of the slabs36 under Anatolia.

Full size image

Origin of anisotropy

As shown in our Rayleigh-wave phase-velocity maps, small-scale lateral variations of the anisotropy in the Anatolian crust as well as the underlying lithospheric and asthenospheric upper mantle are inconsistent with the largely uniform anisotropy pattern derived from shear-wave splitting measurements. On the other hand, large-scale patterns in the fast directions of anisotropy in the crust and upper mantle in our model are coherent and suggest that these two layers of anisotropy (crust and mantle) are coupled together. At periods of 3–35 s, sampling the crust, the fast direction of anisotropy is found to be mostly in EW direction for the whole region. In western Anatolia, similar EW directions of fast axis are found, at periods of 50–100 s, sampling the lithospheric mantle. In the asthenosphere, the fast directions of anisotropy are quite different from N120∘E azimuth. In eastern Anatolia, beneath the Iranian Block and Arabia Plate, depth dependency in anisotropic fabric is also found between the crustal level (5–35 s) and the asthenospheric flow (sampled by periods >50 s) where lithosphere is likely thinner32,41. In our model shown in Figs. 2 and S13, the amplitude of anisotropy has a variation of 0.2–1.2% over the period band we investigated (5–300 s), with a mean value of about 1%. However, the shear-wave splitting measurements display a delay time of 1–1.5 s over the whole region. A 1% anisotropy over 40-km-thick crust would result in a delay time of 0.12 s, which is insignificantly small. Similarly, a 1% anisotropy over a 100-km-thick lithosphere and 270-km asthenosphere would result in delay times of 0.20 and 0.50 s, respectively. In order to explain the delay time of 1.5 s we need to have a consistent fast direction of anisotropy with an average anisotropy of ~3.2% over a 250-km-thick layer. This implies that the cause for the main anisotropic pattern observed by shear-wave splitting measurements should have a deep origin, with a long-wavelength signal and a very large lateral extension. One possible candidate shall be the remnant of previously subducted Tethyan lithosphere lying in the mantle transition zone36,42,43,44. The required anisotropy strength of over 3% is obtainable with the presence of olivine polymorphs and dense hydrous magnesium silicate as well as a limited amount of fluids16,45,46,47, which is compatible with the presence of cold slabs48 lying in the mantle transition zone49,50. Although the existence of α-olivine in the slabs in the Mediterranean region has never been confirmed, our result rules out the crust, lithosphere, and asthenosphere as the main contributor of anisotropy. Strong lateral variations of anisotropy are observed at those depths, which is inconsistent with the uniform pattern seen in shear-wave splitting measurements.

To reconcile our surface-wave results and the observed splitting data47, we propose that there is a strong anisotropic layer within the mantle transition zone, probably the olivine polymorphs or dense hydrous magnesium silicate51. Strain-induced deformation and crystallographic orientation of bridgmanite in the uppermost lower mantle surrounding the slab is also another possible source52,53, if the anisotropy can not be fully accommodated in the transition zone. Therefore, the current study provides plausible evidence for the existence of prominent anisotropy in the stagnant slabs at the bottom of the upper mantle beneath Anatolia.

The scenarios are likely that the specific orientation (N30∘E) of the anisotropy within the slab could be explained by the frozen anisotropy preserved within the oceanic Tethyan lithosphere, or acquired anisotropy resulting from mantle flow within the transition zone54,55,56.

Conclusion

In this study, we obtained the fundamental-mode Rayleigh-wave anisotropic phase-velocity maps for Anatolia from a combined dataset of ambient noise and earthquake records. Our model displays important features related to structural variations and dynamic processes in the region. Isotropic anomalies map the main tectonic blocks that are consistent with results from previous studies, whereas lateral and vertical variations of anisotropy provide additional constraints on the regional geodynamical evolution. In our anisotropic phase-velocity maps we resolve two distinct anisotropic layers beneath Anatolia, suggesting a mechanical decoupling between the crust and lithospheric mantle in this region. In the crust, the anisotropic pattern is in good agreement with the westwards extrusion of the Anatolia Plate as illustrated by GPS observations. On the other hand, the vertical continuity of the anisotropic patterns in the lithospheric and asthenospheric mantle beneath the Anatolia and Arabia plates, as well as small-scale local variations in the anisotropy, are inconsistent with the shear-wave splitting measurements, implying a deeper origin of the anisotropy. A uniform layer of anisotropy over the whole region is suggested to explain the delay times observed by shear-wave splitting measurements: a thick layer (≥250 km) with strong anisotropy (around 3%) is necessary, which argues for a remnant stagnant slab horizontally lying in the mantle transition zone beneath Anatolia, with the presence of highly anisotropic material such as α-olivine or hydrous magnesium silicate. Finally, our model also maps an anisotropy pattern consistent with a toroidal flow in the asthenospheric mantle combined with a strong velocity contrast in the presence of slab tear beneath eastern Anatolia.

Methods

In this study, we construct the anisotropic phase-velocity maps of fundamental-mode Rayleigh waves following a two-step approach. In the first step, we use the two-station method to measure the interstation Rayleigh-wave dispersion curve. For this purpose, three years (2015–2017) worth of vertical-component teleseismic waveforms at 142 seismic stations deployed in Anatolia are retrieved from IRIS-DMC (Incorporated Research Institutions for Seismology Data Management Center) and KOERI (Kandilli Observatory and Earthquake Research Institute), which offer relatively dense distribution of stations. This resulted in 2604 interstation dispersion curves with good coverage of the study region under the condition that the angle between the arc connecting a pair of stations and that connecting the earthquake and the station pair is less than 5 degrees. This dataset is further augmented by 1589 interstation dispersion curves (Fig. S1) derived from ambient noise24, for a total of 4193 dispersion curves. To measure the interstation dispersion curve, vertical-component Rayleigh-wave records at a pair of stations are cross-correlated (Fig. S2). The cross-correlation function is filtered by a frequency-dependent Gaussian filter to minimize the effects of noise and interferences, and then transformed into the frequency domain where its complex phase is determined and used to estimate the phase velocity. For each station pair, phase velocities obtained from all earthquakes are assembled and the average dispersion curve as well as the standard deviation are calculated (Fig. S3). This approach allows us to obtain dispersion curves over the broad period range of 5–300 s. Bumps or portions of dispersion curves that are not smooth, measurements that do not agree with the average dispersion curve in terms of velocity, first and second derivatives, are considered as outliers and are discarded. The entire collection of dispersion curves are displayed in Fig. S4. After removal of outliers, the numbers of interstation pairs retained for phase-velocity map construction are different for different periods. Figure S5 shows the interstation path distribution for selected periods.

In the second step, we invert for the anisotropic phase-velocity maps at selected periods using the 3460 interstation dispersion curves. The anisotropic phase velocity at each point of the model can be expressed as follows57:

$$\delta C=\delta {C}_{{\mathrm{iso}}}+{A}_{2\psi }{\mathrm{cos}}(2\psi )+{B}_{2\psi }{\mathrm{sin}}(2\psi )+{A}_{4\psi }{\mathrm{cos}}(4\psi )+{B}_{4\psi }{\mathrm{sin}}(4\psi ).$$
(1)

Where δCiso represent the isotropic phase-velocity perturbation, and the last four terms are the so-called 2ψ and 4ψ anisotropic anomalies. It has been shown that the 2ψ terms are 2–5 times larger than the 4ψ terms24, and thus in our inversions, the 4ψ terms are not considered. This is also justified by the observation that the 4ψ terms are not necessary to explain the data since their inclusions in inversion have insignificant effect on χ2 58.

After deriving the dispersion curves for the 3460 interstation paths, we invert them for both isotropic and anisotropic (2ψ and 4ψ) Rayleigh-wave phase-velocity maps at selected periods. At each point of the model, the total velocity anomaly can be parameterized with five coefficients: one for the isotropic phase-velocity variation, δCiso, 2 for the 2ψ-anomaly, A2ψ and B2ψ, and 2 for the 4ψ-anomaly, A4ψ and B4ψ:

In the inversion for Anatolia, we parameterize the model by a triangular grid of knots43. The grid spacing as well as the number of knots are variable with period depending on the interstation path coverage. At shorter periods (<30 s), the coverage is poorer with fewer paths, resulting in relatively large grid spacing and fewer knots over a reduced area (e.g., 78.226 km and 306 knots at 20 s). At longer periods, on the other hand, the coverage is better with more abundant path, leading to smaller grid spacing and more knots (e.g., 66.543 km and 425 knots at 80 s). The average interstation dispersion measured in the first step can be related to the local phase-velocity perturbation in Eq. (1) via an integral over the interstation arc:

$$\delta \bar{{C}_{i}}={\int_{\varphi }}{\int_{\theta }}{K}_{i}(\varphi ,\theta )\ \delta C(\varphi ,\theta )\ d\theta \ d\varphi \ ,$$
(2)

where Ki(φ, θ) is the sensitivity kernel.

In this study, we did not attempt to invert our dispersion curves into a 3D model. Therefore, sensitivity kernels used to estimate the depths of penetration of the surface waves at the periods investigated in this study are provided in Fig. S14.