The causes for the formation of large igneous provinces and hotspot trails are still a matter of considerable dispute. Seismic tomography and other studies suggest that hot mantle material rising from the core-mantle boundary (CMB) might play a significant role in the formation of such hotspot trails. An important area to verify this concept is the South Atlantic region, with hotspot trails that spatially coincide with one of the largest low-velocity regions at the CMB, the African large low shear-wave velocity province. The Walvis Ridge started to form during the separation of the South American and African continents at ca. 130 Ma as a consequence of Gondwana breakup. Here, we present the first deep-seismic sounding images of the crustal structure from the landfall area of the Walvis Ridge at the Namibian coast to constrain processes of plume-lithosphere interaction and the formation of continental flood basalts (Paraná and Etendeka continental flood basalts) and associated intrusive rocks. Our study identified a narrow region (<100 km) of high-seismic-velocity anomalies in the middle and lower crust, which we interpret as a massive mafic intrusion into the northern Namibian continental crust. Seismic crustal reflection imaging shows a flat Moho as well as reflectors connecting the high-velocity body with shallow crustal structures that we speculate to mark potential feeder channels of the Etendeka continental flood basalt. We suggest that the observed massive but localized mafic intrusion into the lower crust results from similar-sized variations in the lithosphere (i.e., lithosphere thickness or preexisting structures).
Little is known about the source area of the huge amounts of magmatic material erupted during the formation of continental flood basalts (CFBs) and the role of the crust and lithosphere in this context. Some theories predict the existence of mantle plumes transporting heat and material from Earth’s interior to the lithosphere-asthenosphere boundary (e.g., Morgan, 1972; Torsvik et al., 2010; Burke, 2011; Steinberger and Torsvik, 2012; O’Connor et al., 2012). The interaction of plume-related temperature anomalies in the mantle with the lithosphere is complex, with results ranging from flood basalts in the oceans (oceanic plateaus), the eruption of CFBs (Coffin and Eldholm, 1994; Herzberg and Gazel, 2009), and doming caused by the dynamic uplift above the plume (Courtney and White, 1986). At deeper levels, thermal and mechanical erosion of the base of the lithosphere, the incorporation of plume material into the asthenosphere (e.g., Sobolev et al., 2011), the injection of plume-generated melts into the lithosphere and associated compositional changes (Hart and Zindler, 1986; Menzies, 1990), and thermal mobilization of low-temperature melt fractions within the lithosphere (White and McKenzie, 1989) have been identified as possible processes. Classical plume models predict widespread decompression melting after a plume head reaches the base of the lithosphere, causing intrusions in the crust and eruption of large volumes of basaltic magma. Although several hotspots (Tristan, Gough, Discovery, Shona, and Bouvet) are known in the South Atlantic, only the aseismic Walvis Ridge, interpreted as a hotspot trail forming the prolongation of the Etendeka-Paraná CFB (e.g., Renne et al., 1996), fits classical plume model predictions. Recently, large low shear-velocity provinces (LLSVPs) at the core-mantle boundary were proposed as deep thermochemical piles persisting over hundreds of millions of years. The edges of LLSVPs are seen as plume-generating zones (PGZs; Burke et al., 2008; Torsvik et al., 2010; Steinberger and Torsvik, 2012) causing vast amounts of melts, which migrate toward the surface through zones of structural weakness in the lithosphere (e.g., O’Connor et al., 2012; Buiter and Torsvik, 2014; Burov and Gerya, 2014). In the South Atlantic context, these are oceanic fracture zones and major fault zones created in the Pan-African orogeny.
Although the crustal structure of quite a number of oceanic large igneous provinces (LIPs) is relatively well known, only few deep-seismic sounding profiles covering CFB provinces on Earth are available (Columbia River, Deccan, Emeishan, Siberia; see, e.g., Ridley and Richards, 2010). Deep-seismic sounding profiles south of the Walvis Ridge show prominent high-velocity bodies in the lower continental crust parallel to the ocean margin. These were interpreted to reflect a magmatic belt produced by a fast-propagating narrow rift zone localizing lithospheric rupture (Bauer et al., 2000). An abrupt change from volcanic margins south of the Walvis Ridge to non-volcanic margins north of it has been observed independently (Gladczenko et al., 1997; Bauer et al., 2000; Contrucci et al., 2004). Despite these data, little is known about the deeper structure in the area, which is considered to be the center of CFB volcanism at ca. 136 Ma. Had a massive plume head been present, one would expect significant modification of the old continental crust (e.g., Gladczenko et al., 1997; Bauer et al., 2000; White and McKenzie, 1989, 1995).
EXPERIMENT, DATA, AND RESULTS
We present onshore seismic reflection and refraction data from the eastern prolongation of the Walvis Ridge into Africa to provide constraints on how the crust was modified during the initial magmatism in the coastal branch of the Damara orogen–Kaoko belt in northern Namibia (Porada et al., 1983). The seismic experiment consisted of a 320-km-long, coast-parallel refraction profile of which the central 200 km was also covered by near-vertical reflection seismics (Fig. 1; see the GSA Data Repository1).
Seismic Velocity Structure
A two-dimensional P-wave velocity (Vp) model (Fig. 2A) has been derived by a traveltime tomographic inversion of refracted phases (Zelt and Barton, 1998; Ryberg et al., 2012). The velocity model shows a narrow high-velocity feature in the lower crust (Fig. 2A). This high-velocity lower crustal zone (HVLC) is ∼100 km wide and is centered on the landward extension of the Walvis Ridge. It shows Vp of 7–7.6 km/s and has a maximum thickness of ∼20 km. The velocities and thickness are comparable with observations within the continent-ocean transition zone along the volcanic margin transects offshore to the south of the Walvis Ridge (Bauer et al., 2000). The HVLC velocities are elevated with respect to the global average for continental seismic velocities (Christensen and Mooney, 1995), while regions outside have values comparable to the global average (Fig. 3). Therefore, we interpret the HVLC body as intruded or underplated mafic to ultramafic rocks, generated by decompression melting above a hot mantle. Such conditions may have led to intrusions of MgO-rich gabbroic basalts (Farnetani et al., 1996; Geoffroy, 2005; Herzberg and Gazel, 2009; Shellnutt, 2013) in the lower crust with Vp >7 km/s and thickness >15 km (White and McKenzie, 1989). Similar high seismic velocities and lower crustal structures have been observed in other continental rifts, some of them with CFB (Catchings and Mooney, 1988). We evaluated the inland extension of this body by a three-dimensional inversion of P-wave travel times. A depth slice (Fig. DR2 in the Data Repository) at 31 km (lower crust) shows that this high-velocity body is well terminated inland (<100 km ENE of the coastline) and continues toward the ocean (Fromm et al., 2014), coinciding spatially with the landfall area of the Walvis Ridge. The estimated volume of the onshore part of the HVLC body is ∼200,000 km3. A similar velocity anomaly could be observed in the S-wave velocity (Vs) model (Fig. DR4 in the Data Repository). The Vp/Vs ratio is not anomalous (∼1.73).
Crustal Reflective Image
Our investigations were complemented by reflection seismic measurements imaging the crust in the central 200 km of the coast-parallel profile. Applying a new seismic processing technique (Bauer et al., 2013), we derived an improved image of the crustal structures (see the Data Repository). The reflectivity pattern along our profile shows laminar structures in the lower crust, as commonly observed in extensional tectonic settings (Meissner and Brown, 1991) (see Fig. 2B). Common interpretations of such laminations range from intrusive structures to heterogeneities and rheological features in the weak lower crust. We observed these structures within the HVLC, with laminar structures somewhat concentrated at the upper boundary of the zone (Fig. 2B, label L). These reflections are interpreted to represent either preexisting, older lower crustal structures, which were deformed during the intrusion event, or intrusive, mafic sills emplaced as new crustal material during the formation of the Walvis Ridge. Upper crustal structures in the northern part of the section (Fig. 2B, label D, down to 10 km depth) can be associated with fault zones related to the Kaoko belt–Damara orogeny. Directly above the HVLC, the middle and upper crust is devoid of such structures. Here, one would expect a system of dikes and magmatic sills above the magmatic intrusion (Karlstrom and Richards, 2011; Farnetani et al., 1996; Ewart et al., 2004). However, such dike systems, particularly when oriented coast-parallel, i.e., along the seismic line (Ewart et al., 2004), will not be imaged in the reflection seismic section. Nevertheless, elevated seismic velocities in this region, compared to those regions south and north, would be in accord with a high volume fraction of basaltic dikes in the crust. Another prominent feature in the seismic section is a linear band of high reflectivity starting from the HVLC body and continuing updip toward the south into the upper crust (Fig. 2B, label F). We speculate that this structure formed a feeder channel for melts from the lower crust or the upper mantle, possibly along a preexisting zone of mechanical weakness. This possible feeder channel links up to the southern Etendeka CFB region, located ∼100 km south of the area of investigation (Marsh et al., 2001; Ewart et al., 1998). Recent numerical modeling results show that such magma ascent, away from intrusions in the lower crust, can be explained by gravitational unloading pressure (Maccaferri et al., 2014). While the seismic image shows no coherent structures in the uppermost mantle, a flat Moho (typically expressed as the bottom of the lower crustal reflective band) at a depth of ∼40 km is observed. The flatness of the Moho indicates either that the plume head did not disturb the pre-rift Moho (unlikely), or that later equilibration processes led to the recovery of typical continental Moho (Nielsen and Thybo, 2009).
Our results show clear evidence of massive magmatic overprinting of the continental crust at the landward edge of the Walvis Ridge, a feature that extends into the South Atlantic Ocean. Viewed along the coastline, this modified crust is confined to the width of the Walvis Ridge, and does not continue very far ENE into the continent. All of this suggests that the Tristan hotspot (Fig. 1), which is linked with this location by the Walvis Ridge, was responsible for this magmatic modification of the crust. It can also be considered as prima causa for the Etendeka flood basalts. In the context of plumes generated at the edges of the LLSVPs (PGZs), it has been proposed that the Tristan hotspot developed volcanic trails on both sides of the South Atlantic mid-ocean ridge because of its proximity to the spreading axis (O’Connor et al., 2012). However, breakup of the South Atlantic probably followed a Pan-African structural heritage (Buiter and Torsvik, 2014), and deep, hotspot-related mantle processes were less important in constraining the geometry and evolution of the rift (e.g., Franke, 2013). Our data suggest that the localized imprint of the hotspot (or plume) onto the African crust is more in line with the model of an impinging, smaller plume (Courtillot et al., 2003) near the rift axis at the time of breakup. This plume (Tristan hotspot) could have used the weakened, preexisting lithospheric structures of Pan-African age (Porada, 1979) and/or rift-related thinned lithosphere to produce the CFB. At the same time, the deep continental crust was intruded only very locally by mafic melts (north-south extent of HVLC body ∼100 km; see Fig. 2; see the Data Repository). The South Atlantic was opening from the south (White and McKenzie, 1989; Koopmann et al., 2014b; Franke, 2013), creating the magmatic-dominated passive margin off southwestern Africa (Bauer et al., 2000). When the opening reached the location of the present Walvis Ridge, the plume-related asthenospheric upwelling turned into a spot-like feature, causing the intrusion of the HVLC body. This would explain why magmatic activity as expressed by the seismic velocity signature there is similar to that of the volcanic margin off central Namibia further south.
The relatively small size of the HVLC body seems to require similar small-scale structures in the upper mantle (varying lithosphere thickness or preexisting lithospheric structure), otherwise one would expect a much broader-scale magmatic modification of the crust. Short-wavelength (or localized) variations in lithosphere topography in the case of plume impingement in the absence of tectonic forcing, as suggested by numerical modeling, could explain our observations (see Burov and Gerya, 2014). Modeling of rift development has shown that excess magmatism (i.e., onshore HVLC body described here) can be associated with rift segment boundaries (“transfer zones” of Koopmann et al., 2014a). In our case, this is the Florianópolis fracture zone linked to the Walvis Ridge and Paraná-Etendeka LIP. Consequently, the opening of the South Atlantic was driven by the Jurassic-Cretaceous Southern Ocean plate movements driven by spreading systems between South America and South Africa and Antarctica, which had different spreading directions and spreading rates (Jokat et al., 2003).
We conclude that the high-velocity lower crustal body embedded in normal continental crust near the coast of northwestern Namibia is compatible with a localized impact of a plume head. There is a structural control by complex, small-wavelength features deeper in the lithosphere and/or preexisting tectonic features, e.g., of the Damara orogen. Our findings question the active role of a plume during Gondwana breakup. The superposition of the northward-propagating South Atlantic rift and a locally impinging plume head were equally responsible for the increased magmatic activity in the crust of northwestern Namibia. This led to the formation of a locally confined magmatic body in the lower crust and associated feeding systems for the Etendeka flood basalts.
The project is part of the Priority Program SPP1375, South Atlantic Margin Processes and Links with Onshore Evolution (SAMPLE), financed by the Deutsche Forschungsgemeinschaft (grants RY 12/9-1, RY 12/9-2) and the GeoForschungsZentrum Potsdam (GFZ). Without the kind cooperation of the Geological Survey of Namibia, this project could not have been accomplished. We thank reviewers E. Gazel, W.D. Mooney, K. Burke, and S.A. Stein for helpful comments. Figures were prepared using the Generic Mapping Tools (GMT) (Wessel and Smith, 1998). Seismic instruments were provided by the Geophysical Instrument Pool Potsdam (GIPP) (GFZ Potsdam).