Complex Shear Wave Velocity Structure Imaged Beneath Africa and Iceland

See allHide authors and affiliations

Science  03 Dec 1999:
Vol. 286, Issue 5446, pp. 1925-1928
DOI: 10.1126/science.286.5446.1925


A model of three-dimensional shear wave velocity variations in the mantle reveals a tilted low velocity anomaly extending from the core-mantle boundary (CMB) region beneath the southeastern Atlantic Ocean into the upper mantle beneath eastern Africa. This anomaly suggests that Cenozoic flood basalt volcanism in the Afar region and active rifting beneath the East African Rift is linked to an extensive thermal anomaly at the CMB more than 45 degrees away. In contrast, a low velocity anomaly beneath Iceland is confined to the upper mantle.

Tomographic models of three-dimensional seismic velocity variations continue to improve our understanding of the structure of flow in Earth's mantle (1–6). It has long been recognized that higher than average seismic velocities are present in regions of the mantle with long-term subduction (such as the circum-Pacific region and Asia) and that lower than average velocity structures are present in the deep mantle beneath Africa and the Pacific Ocean where subduction has not occurred since the Jurassic. The interpretation of high velocity anomalies as subducting slabs of relatively cold oceanic lithosphere has gained support from recent tomographic models, which show tabular structures of faster than average velocity beneath several subduction zones (7, 8) (Web figure 1). The complex shape and broadening of these slab-like structures may be a manifestation of the complex subduction zone process, large-scale plate motions, and the resistance to flow caused by the spinel-to-perovskite phase transition at 670 km depth (9). Although laboratory experiments (10, 11) and computer models (12, 13) demonstrate that hot upwellings may ascend into the mantle in a similarly complex manner, it is difficult to image narrow conduits of rapidly rising mantle plumes (14) in the deep mantle. In addition, tomographic models have not fully constrained the dimensions and shape of the large-scale upwellings beneath Africa and the Pacific Ocean (Fig. 1).

Figure 1

(left). Comparison of tomographic models along a cross section from the southwest to the northeast (left to right) of models: (A) s16u6l8 (6), (B) s12wm13 (3), (C) s16b30 (4), (D) SG (7), and (E) S20RTS (15).

We studied the structure of upwellings in the mantle using seismic tomographic model S20RTS (15). We focused on the large-scale low velocity anomaly in the deep mantle beneath Africa, which was imaged in earlier tomographic models (1, 2) (Fig. 1) and has been linked to surface manifestations of hot spots around Africa (16), the high elevation of southern Africa (17), and the ultra-low velocity anomalies at the base of the mantle beneath Africa (18). We compared this anomaly with the low velocity structure beneath Iceland. S20RTS is a degree-20 shear wave velocity model that incorporates surface wave phase velocities, body wave travel times, and free-oscillation splitting measurements (Table 1). These data cover the entire spectrum of seismic frequencies and, hence, constrain the long-wavelength and the short-wavelength seismic velocity variations.

Table 1

Seismic data used in S20RTS.

View this table:

We obtained phase velocity and travel time data high from digital global and regional network recordings (IRIS, GEOSCOPE, MedNet, CNSN, and PASSCAL) of moderate-size (magnitude,Mw > 5) earthquakes that occurred between 1980 and 1998. The surface wave data set contains over a million phase velocity measurements (on average 10 measurements per seismogram) of fundamental and higher mode Rayleigh waves with seismic periods ranging from 40 s to 250 s. Higher mode (up to the fourth overtone) phase velocity measurements, which constrain seismic structures in and below the upper mantle transition zone (400 to 1000 km depth), constitute a unique subset of these data (19). The body wave data set includes about 50,000 hand-picked absolute travel time measurements for a large variety of seismic phases that propagate through the lower mantle, including S,Sdiff multiple surface reflections (SS and SSS), core reflections (ScS,ScS2 , and ScS3 ), and core-phases (SKS and SKKS) (20). These phases provide a uniform sampling of the mid and lower mantle in the Northern and Southern hemispheres. The third type of data used were normal mode structure coefficients for multiplets below 3 mHz. These data are particularly useful to constrain the very long-wavelength (>2000 km) pattern of seismic velocity variation. It has been demonstrated that these data are a valuable addition to body wave and surface wave data to constrain velocity structures in the mid-mantle (21). We invert the entire data set for a model of shear velocity perturbation with respect to the Preliminary Reference Earth Model (PREM) (22) using an exact least-squares inversion technique (23).

Horizontal cross sections through model S20RTS illustrate that the low shear wave velocity anomaly beneath Africa has a complex three-dimensional shape (Fig. 2A). The African anomaly covers an extensive region in the lowermost mantle (4000 × 2000 km2 at the CMB) beneath the southeastern Atlantic Ocean and it is connected to patches of relatively low velocity beneath central Africa, northwestern Africa, and the southern Indian Ocean. Structures with relatively high velocity anomalies beneath North and South America, the polar regions, eastern Asia and the Indian Ocean surround the African anomaly. The African anomaly is narrower at mid-mantle depths (2350 to 1100 km) and it is centered progressively further to the east and northeast with increasing height above the CMB. Vertical cross sections further illustrate the continuity of the African anomaly at least 2000 km into the mantle (Fig. 2, B through D). Figure 2B shows that the African anomaly extends from the CMB beneath southern Africa into the upper mantle beneath the East African rift, while the tilt of the African anomaly toward the east and northeast is also obvious in the cross section of Fig. 2, C and D. This tilt may be enhanced by a lateral offset at a depth of about 2000 km in our model.

Figure 2

(right). (A) Horizontal cross sections through S20RTS at the CMB (2890 km depth), and at depths of 2600, 2350, 2100, 1850, 1600, 1350, and 1100 km. Relatively high velocity and low velocity regions are indicated by blue and red colors, respectively, with an intensity that is proportional to the percentage amplitude of the velocity perturbations compared with shear wave velocities from the PREM at these depths. Green lines represent plate boundaries and black lines outline land masses. (B toE) Cross sections along 140° wide great circle arcs. The dashed line represents the 670-km seismic discontinuity. Superposed are maps showing corresponding great circle arcs along which the cross sections are made. White circles on the great circle arc are plotted at 20° intervals and correspond to the bold tick marks shown in the cross sections. The color scale bar indicates the percent that the shear wave velocities are higher (positive percentages) or lower (negative percentages) than the PREM.

Model S20RTS indicates that the low velocity anomaly beneath Iceland is smaller in volume than the African anomaly (Fig. 3). The 660-km phase transition marks a lower boundary of the low velocity anomaly beneath Iceland (with up to 2.5% lower shear wave velocities compared to PREM velocities at this depth). The low velocity anomaly in the transition zone beneath Iceland may explain the change in the depth of the 410-km and 660-km phase transitions versus the predicted transition depths from PREM and mineral physics studies (24). However, a connection to the CMB by a narrow plume conduit, as recently suggested using P wave travel time tomography (25), is not modeled in S20RTS.

Figure 3

(left). (Ato D) Vertical cross sections through Iceland plotted with the same color scale and labeling as used in Fig. 2.

The finite extent of Backus-Gilbert resolution kernels (26) (Fig. 4 and Web figure 2) for three locations in the mantle beneath Africa indicates that small-scale volumetric heterogeneity (<500 km) is not resolved and that velocity contrasts are underestimated in S20RTS. However, shear wave velocity anomalies at points X, Y, and Z were independently derived because Backus-Gilbert kernels, determined for these points, do not overlap. Hence, the continuity of the low velocity anomaly and its tilt beneath Africa are not caused by preferential southwest to northeast body wave sampling.

Figure 4

(right). Backus-Gilbert resolution kernels for locations X, Y, and Z at 2500, 1500, and 700 km depth, respectively, in the mantle beneath Africa. Horizontal cross sections through the Backus-Gilbert kernels are shown on the left. The relative amplitude of the kernels (from 0.1 to 1.0) is represented by nine colors, which range from yellow to red to pink. Regions where the Backus-Gilbert kernels have a relative amplitude lower than 0.1 are not shaded. The radial dependence of the kernels is shown on the right.

A large-scale thermal upwelling (or an assemblage of several boundary layer instabilities) from the CMB is a consistent interpretation of the African anomaly. Although the images of the African anomaly cannot completely constrain the significance of compositional heterogeneity in the mantle (27) or the evolution of such an upwelling (28), they do indicate that the shape of the upwelling at its base is complex and that the upwelling does not ascend vertically into the mantle. The shear of the upwelling may be a consequence of the migration of the African plate to the northeast since the breakup of Gondwanaland, interaction with mid-mantle heterogeneity, the 670-km phase transition or a putative physical boundary near 2000 km depth (3,29).

The weakening of the low velocity anomaly between 670 and 1000 km depth indicates that the upwelling is obstructed. Eventually, however, an upwelling may form in the transition zone beneath eastern Africa, which propagates along the base of the lithosphere into the East African rift region. The central location of the upwelling in the deep mantle beneath southern Africa can explain the anomalously high elevation of southern Africa and of its contiguous ocean basins and the high long-wavelength geoid over Africa and the Atlantic Ocean (17). The continuity of this upwelling into the upper mantle region beneath East Africa compels a link between a relatively hot CMB region and flood basalt volcanism that formed the Ethiopian traps and contributed to the rifting in the Red Sea, the Gulf of Aden, and along the East African Rift (30).

  • * To whom correspondence should be addressed. E-mail: jeroen{at}


View Abstract

Stay Connected to Science

Navigate This Article