Discovering the connection between processes observed to occur at the surface of the Earth and its internal dynamics remains an essential goal in the Earth sciences. Deep mantle structure, as inferred from seismic tomography or subduction history, has been shown to account well for the observed surface gravity fieldand motions of tectonic plates1,2,3. But the origin of certain large-scale features, such as the anomalous elevation of the southern and eastern African plateaux, has remained controversial. Whereas the average elevation of most cratons is between 400 and 500 m, the southern African plateau stands more than 1 km above sea level, with the surrounding oceans possessing a residual bathymetry in excess of 500 m (ref. 4). Global seismic tomography studies have persistently indicated the existence of a large-scale low-velocity anomaly beneath the African plate5,6,7,8,9,10 and here we show that mantle flow induced by the density variations inferred from these velocity anomalies can dynamically support the excess elevation of the African ‘superswell’. We also find that this upwelling mantle flow—which is most intense near the core–mantle boundary—constitutes a significant driving force for tectonic plates in the region.
Large-scale anomalous elevation of a continent can originate from: (1) changes in the average density and/or thickness of the lithosphere; and (2) vertical motion of the continent in the absence of faulting or folding. A significant source of vertical motion is dynamic topography11.
Most previous explanations for the high residual topography of the southern and eastern African plateaux and surrounding oceans (Fig. 1a) have invoked isostatic mechanisms. These have included heating of the lithosphere, as indicated by high heat-flow measurements in the southern African craton12, and lithospheric thinning, as evidenced by studies of the gravity field and Cenozoic volcanism under the eastern African plateau13. Although certainly some of these mechanisms will be at play, they have failed to successfully account for the existence of the African superswell.
Here we focus on dynamic topography as an explanation for the African superswell. Dynamic topography is a deformation of the surface of the Earth, supported by the vertical stresses at the base of the lithosphere, that are generated by flow in the mantle below1,14,15,16,17,18,19. This is in contrast to the more familiar mechanism of isostatically supported topography which is in equilibrium at the Earth's surface and would exist even in a static mantle. Dynamic topography is expected to contribute to the elevation of the African continent because of upwelling (associated with anomalously low seismic velocities) that coincides with a high in the long-wavelength geoid1. But we note that detection of dynamic topography, particularly over the oceans, remains controversial.
We predict the dynamic topography from a calculation of instantaneous flow in the Earth's mantle20. The flow is a solution to the equations for the conservation of mass and momentum in an incompressible newtonian viscous fluid. The buoyancy forces that drive the flow are prescribed by a model of lateral density heterogeneity within the mantle. These calculations yield the radial stress acting at the base of the lithosphere, and allow us to compute the resulting dynamic topography simply by dividing the stress by thedensity contrast between the lithosphere and the overlying medium and by the gravitational acceleration. Because we are focusing on thedynamic topography of Africa, we use a density contrast (3,200 kg m−3) appropriate for the continent–air interface. This leads us to underestimate by ∼25% the dynamic topography over ocean basins, for which the appropriate density jump is smaller because of the larger density of water.
We explore two different models of mantle density heterogeneity. The first is based on the history of subduction for the past 200 Myr. This type of model is based on the idea that in an Earth primarily heated from within, cold subducted slabs will be the main source of thermal buoyancy in the mantle and will dominate the structure of the flow. This model has been extremely successful at explaining the geoid2, present and past plate motions3, and the history of uplift and subsidence of continents11,21. In this model, upwellings are passive, and are characterized by the broad-scale return flow away from downwelling regions.
Active upwellings, generated in the basal thermal boundary layer, may also be an integral part of the flow. To address this possibility, we consider a second model of mantle density heterogeneity, one that is inferred from seismic tomography. This model will include a signal from active upwellings in addition to the slab signature. Global seismic tomography models have consistently imaged a coherent low-velocity anomaly under the African plate5,6,7,8,9,10. One of the most recent high-resolution seismic models9 shows a continuous feature from the surface to the core–mantle boundary. In the lower mantle, between 1,000 and 2,890 km depth, the anomaly is located under southern Africa. In the shallower portions of the mantle, at depths less than 1,000 km, the anomaly becomes thinner and jogs northeastwards, under the Great Rift Valley (eastern African plateau). This type of thinning is expected for an upwelling encountering a lower-viscosity layer22—the upper mantle has an average viscosity more than an order of magnitude less than that of the lower mantle2,23,24,25.
The density heterogeneity model based on subduction history does not adequately reproduce the residual topography (Fig. 1b). Instead, it produces a very broad-scale topographic high in the entire Atlantic basin, which is the result of the passive, upwelling return flow that balances the downwelling flow induced by subduction-related buoyancy. This broad upwelling cannot explain the anomalous elevation of the African superswell. If the anomalous elevation is to be related to dynamic topography, other density anomalies must be present in the mantle, in particular a concentrated low-density anomaly that we can associate with a large-scale upwelling.
To calculate the dynamic topography induced by the tomographically inferred density field, we use only the velocity anomalies below 325 km depth. At depths less than 325 km many of the anomalies found in the upper mantle are associated with cratons, and are probably chemical rather than thermal in nature. To avoid any contamination of our results with deep lithospheric structure, we restrict ourselves to depths greater than 325 km. We convert the remaining velocity anomalies to density by using a constant conversion factor, 0.4 kg m−3 km−1 s, from 325 km to the core–mantle boundary. This is a value that is well within the bounds of laboratory experiments and geodynamical models26. We find that the dynamic topography predicted from this density field reproduces remarkably well the anomalous residual topography over the African plate (Fig. 1c). The anomalous topography under the eastern African plateau is the result of the portion of the density anomaly in the upper mantle (325–525 km). The topographic high corresponding to the southern African plateau is the direct result of the density anomaly at depths greater than 1,000 km. Without including the lower-mantle anomaly (below 1,000 km), the pattern and amplitude of dynamic topography is not reproduced.
Our results agree in amplitude, as well as pattern, with the observed residual topography (peak predicted amplitudes ∼700 m, average residual elevation of the African craton ∼650 m). The agreement in amplitude may be partly fortuitous, because there are two known sources of amplitude bias that will tend to counteract each other. First, intentionally conservative seismic anomalies in Grand's model9 will underpredict the amplitude of velocity heterogeneity in the mantle and consequently the amplitude of surface dynamic topography. Dynamic topography calculated with this model in other regions (such as the circum-Pacific subduction zones) is smaller than that predicted by other models2,27 by as much as 650 m. Second, mantle flow calculations tend to overpredict dynamic topography. Resolving the discrepancy between models of dynamic topography and its observed amplitude remains a fundamental problem in geodynamics. One recent approach has been to allow the development of dynamic topography at internal phase or chemical boundaries17,18,19.
In this context it is important to note that the geoid predicted by Grand's model agrees as well with observations as most other tomographic models for the same parametrization and boundary conditions. The agreement is good both at the global and regional level, and the model correctly predicts the geoid high over Africa (see Supplementary information).
With the caveat that mantle flow models tend to overpredict the magnitude of dynamic topography, we suggest that the deep low-velocity anomaly under Africa, as seen consistently in tomographic models, is the primary cause of the African superswell. The shallower portions of this lower-mantle feature are at least partly responsible for the high elevations of the eastern African plateau. This deep-mantle feature and its upper-mantle extension might also be the source for the high heat flow and volcanism in this region.
Having established the effect that mantle flow has on surface topography, we turn our attention to the possible contribution of a large-scale upwelling to plate driving forces. Such forces are, in general, gravitational in nature and result from the large lateral density contrasts that arise from the cooling of the oceanic lithosphere and its subsequent subduction. But, because plates are coupled to mantle flow, other density anomalies in the mantle (such as the large-scale upwelling described here) will also have an influence on the shear tractions at the base of the lithosphere. To consider its effect, we separate plate driving forces into four categories; lithospheric thickening, upper-mantle slabs, lower-mantle slabs, and a general large-scale upwelling corresponding to the low velocity anomalies in Grand's tomographic model. We calculate the instantaneous flow from each driving load (that is, density distribution) independently, compute the shear tractions acting at the base of the lithosphere and find, in the no-net-torque approximation, the driving torques exerted on each plate3.
For plates in the Atlantic basin we find that the torque magnitudes related to the large-scale upwelling are large when compared to other driving forces—in fact, much larger (almost an order of magnitude) than lithospheric thickening and comparable to the torques from upper-mantle slabs (Fig. 2). Most of the large-scale upwelling contribution (∼90%) is derived from the lower-mantle portion of the anomaly. Clearly then, the driving torques exerted by a large-scale upwelling, even one constrained to the lower mantle, cannot be ignored, provided that there is no barrier to flow between upper and lower mantle. Given that there is a significant and visible contribution to the radial stresses from this upwelling, this result is not unexpected.
We have shown here that the low-velocity anomaly under the African plate, seen in tomographic models, can be viewed as an image of a large-scale active upwelling under this region. This density anomaly induces a surface boundary deformation (dynamic topography) that explains the existence of the African superswell including the southern African plateau and surrounding oceans, and the eastern African plateau. The upwelling also constitutes a significant driving force for plates in its vicinity. The upwelling is most intense in the vicinity of the core–mantle boundary, presumably its point of origin. We therefore conclude that large-scale upwellings originated from the core–mantle boundary, such as the one we have examined here, are important features of global mantle flow, and have a strong influence on surface processes. Such upwellings were probably also important in the past. As previously shown11,21, the inferred evolution of dynamic topography predicted from known sources of buoyancy, such as those related to the history of subduction, can explain much of the uplift history implied by the Cenozoic continental flooding record11,21. It may now be possible to similarly examine the evolution of deep-mantle upwellings by comparison with the uplift history of the African plate, in a manner similar to that suggested by White and Lovell for plume activity28.
We thank S. Grand for providing his model, and A. Nyblade for providing the data for Fig. 1a. We also thank H. Pollack and A. Nyblade for comments. The manuscript was significantly improved by comments from U. Christensen and Y. Ricard. C.L.-B. was supported by a National Science Foundation postdoctoral fellowship; P.G.S. and C.L.-B. were supported by the Carnegie Institution of Washington.