Resolving the age-distance relation of volcanism along the Walvis Ridge (southern Atlantic Ocean) is essential to understanding relative motion between the African plate and the Tristan-Gough mantle plume since the opening of the South Atlantic. However, tracking the location of the Tristan-Gough plume might not be practicable if most of the complex morphology of the massive Walvis Ridge is related to the proximity of the South Atlantic mid-ocean ridge. Here we use new 40Ar/39Ar basement ages for the Tristan-Gough hotspot track, together with information about morphology and crustal structure from new swath maps and seismic profiles, to infer that separated age-progressive intraplate segments track the location of the Tristan-Gough mantle plume. The apparent continuity of the inferred age-distance relation between widely separated age-progressive segments implies a connection to a stable or constantly moving source in the mantle.


Progressive changes in the age, paleolatitude, morphology, structure, and geochemistry of volcanic edifices along major hotspot tracks can provide information about the depth of origin of mantle plumes and their relative motion with tectonic plates, which has important implications for understanding and modeling plate motion and the role and character of mantle flow (Gordon et al., 1978; Molnar and Stock, 1987; Tarduno et al., 2003; Steinberger et al., 2004; Doubrovine et al., 2012; Courtillot et al., 2003; Sharp and Clague, 2006; Cande and Stegman, 2011; Koppers, 2011; Koppers et al., 2012; Rohde et al., 2013b; O’Connor et al., 2013). However, the relative importance of deep plume and shallow plate tectonic and asthenosphere processes in controlling the formation of hotspot trails remains controversial (Natland and Winterer, 2005). Here we combine our discovery of discrete chains of intraplate volcanic centers buried in the Walvis Ridge (southern Atlantic Ocean) with new information about the age-distance relation of volcanism along the Tristan-Gough hotspot trail to disentangle for the first time a buried discontinuous plume track through the Walvis Ridge.


The only age-progressive volcanic trail on the African plate connecting an active hotspot with continental flood basalts is the Walvis Ridge and a younger province of seamounts and small ridges (O’Connor and Duncan, 1990; O’Connor and le Roex, 1992). The Walvis Ridge and associated continental flood basalts are part of the global catalogue of large igneous provinces consisting of aseismic ridges, continental flood basalt provinces, oceanic plateaus, and volcanic rifted margins.

The Walvis Ridge and Rio Grande Rise were forming together until ca. 70 Ma, while the spreading Mid-Atlantic Ridge was located close to the hotspot (O’Connor and Duncan, 1990; O’Connor and le Roex, 1992). The spreading ridge and hotspot continued diverging and the Tristan-Gough hotspot trail mostly formed thereafter on the African plate as a broad swath of scattered seamounts and small volcanic ridges (O’Connor and Duncan, 1990).

A widespread volcanic-tectonic event in the Eocene seems to have uplifted parts of the Rio Grande Rise plateau above sea level, triggering late-stage volcanism (e.g. Mohriak et al., 2010). Thus, much of the age and morphology of the Walvis Ridge are likely to express the relative motion between the Tristan-Gough mantle plume and the mid-ocean ridge rather than the African plate (Sleep 2002, 2006) as predicted by the classic plume hypothesis.


Morphology and Crustal Structure

Bathymetric data suggest that seamounts along the Tristan-Gough hotspot track are confined to the younger seamount province extending from the southwest end of the Walvis Ridge. Nevertheless, thick sediments cover and surround much of the Walvis Ridge, especially closer to the Namibian coast; this masks any significant variations in basement structure that might be present. Because the density of sediment is less that of volcanic basement (Wessel et al., 2010), the sediment-hidden large-scale basement structures of the Walvis Ridge are better captured by the gravity data (Figs. 1 and 2).

The gravity map reveals a series of buried chain-like structural highs along the Walvis Ridge between the African continent and ∼6°E (Fig. 1; see also an enlarged version in Fig. DR1 in the GSA Data Repository1). High-resolution bathymetric maps acquired during the RV Maria S. Merian MSM 17 cruises (Jokat, 2012) show that these structures are discrete seamounts as far as ∼8°E, and become more ridge like to ∼6°E (Fig. 2). This part of the Walvis Ridge seems to have formed subaerially, as evidenced by denudational features, cliffs as high as 150 m, and deeply incised valleys reflecting paleodrainage; sand wave abrasion platforms (Fig. 2B) show the later subsidence of the Walvis Ridge below sea level (J. Behrmann, 2014, personal commun.).

Seismic refraction profiles provide further evidence that these buried structural highs near the continental margin are closely spaced seamounts or small volcanic ridges (Fromm et al., 2014; T. Fromm, 2015, personal commun.). These edifices are on thickened oceanic crust (to 30 km) beneath a cover of extrusive rocks such as hyaloclastites and basalt lava flows (Fromm et al., 2014). The excess melt crystallized within the crust–upper lithosphere, implying that buoyant plume material was guided to the surface by local plate structure. An important inference based on multibeam mapping and seismic information is, therefore, that the Walvis Ridge erupted as a series of discrete intraplate point sources on thick oceanic crust.

Although oceanic crust north of the Walvis Ridge was likely also thickened by hotspot flow, it was transferred to the South American plate by a major spreading ridge jump ca. 93 Ma and formed the São Paulo Plateau (Pérez-Díaz and Eagles, 2014; T. Fromm, 2015, personal commun.). The sharp transition across the Rio Grande Fracture Zone (Fig. 2) from thick (30 km) Walvis crust to thin oceanic crust (6 km) in the Angola Basin is therefore the result of a tectonic event.

The gravity maps show that a wide low-relief plateau segment extends from the western end of the buried intraplate volcanic centers as far as the transect of Deep Sea Drilling Project (DSDP) ocean drilling sites (Fig. 1). South of these DSDP sites, the Walvis Ridge splits into three roughly parallel volcanic ridges (Fig. 1; Fig. DR1). The two ridges to the east formed in an intraplate setting (Fig. DR1b) and seem to consist also of chains of discrete buried seamount and/or small ridges, probably over thickened oceanic crust. This implies that both of these intraplate ridges, rather than just the eastern spur, reflect the track of the plume orifice on the African plate after it had crossed the spreading axis in the vicinity of the DSDP transect (Sleep 2002).

South of these three ridges, the Walvis Ridge developed as a province of intraplate seamounts and minor ridges. A seismic profile across this part of the Walvis Ridge at ∼35°E (Fig. 1) shows that the oceanic crust in the Cape Basin (4 km) was ∼2 km thinner that in the Angola Basin, indicating that plume flow influenced crustal formation only in the Cape Basin (Keßling, 2008; O’Connor et al., 2012,). Evidence that buoyant plume material did not affect (i.e., thicken) oceanic crust in both basins implies that the area influenced by the plume material was as small as ∼100 km (Keßling, 2008; O’Connor et al., 2012).

Gravity and seismic modeling support the prediction that hotspot material used the existing fracture zone system without modifying the oceanic crust formation of the adjacent basins (Keßling, 2008; O’Connor et al., 2012). Moreover, the transition between the Walvis Ridge and the Angola Basin is more gradual than in the Cape Basin, which might be explained by proximity to an active mid-ocean ridge system.

These various observations are consistent with buoyant plume material plume flowing under the lithosphere (Sleep, 2002, 2006) and using lines of weakness to reach the surface, such as existing fracture zones in ca. 20 Ma or younger oceanic crust that was adjacent to a slowly migrating spreading ridge (Fig. DR1c) (Keßling, 2008; O’Connor et al., 2012).


An age-distance relation for volcanism along the Tristan-Gough hotspot track was first inferred using 40Ar/39Ar ages measured using 1980s methods (O’Connor and Duncan; 1990, O’Connor and le Roex, 1992). Remeasured ages using modern 40Ar/39Ar methods for some of the samples analyzed in these earlier studies confirm the notion of an age-progressive hotspot track (Rohde et al., 2013a). The geochemistry of the samples used in this study was reported in detail by Rohde et al. (2013b), Hoernle et al. (2013), and J. Rohde (2015, personal commun.).

We report here 15 40Ar/39Ar incremental heating isotopic ages, measured on acid-leached mineral (plagioclase) and one groundmass separate, for 11 sample locations, 7 of which have not previously been dated, or redated using modern 40Ar/39Ar methods. We have also calibrated our results with those reported by Rohde et al. (2013a) by replicating ages at three additional sites using the same samples (Fig. 3). The locations and sample numbers for all of the dated samples used in this study are shown on a gravity map of the Walvis Ridge in Figure DR1; sample information, 40Ar/39Ar results, plateau age and K/Ca spectra, analytical methods, and data files are provided in the Data Repository.

Figure 3 shows the along-chain age-distance relation using new (Table DR2) and published 40Ar/39Ar ages (Rohde et al., 2013a; O’Connor and le Roex, 1992; Renne, 2011; Hicks et al., 2012; Maund et al., 1988). Distances are measured from the Etendeka continental flood basalts (northwestern Namibia; Fig. 3) in order to avoid introducing uncertainty by assuming that the hotspot is at Tristan da Cunha Island or Gough Island.

The age-distance relation based on increased age coverage points to new trends.

1. The oldest along-chain ages predict a linear age-distance trend, which shows a first-order correlation with narrow high-relief, chain-like intraplate structures, aligned along the eastern side of the Walvis Ridge (Fig. 3).

2. Sample ages younger than predicted by this linear age-distance trend show a relation with low-relief, plateau-like morphology, which is characteristic of a mid-ocean ridge setting (Fig. 3). A sample from a ridge-like structure located in the southeast corner of this region (sample AII-93–19–4) is a 55 Ma trachyte dredged at a depth of 2450 m (Rohde et al., 2013a). Samples dredged ∼40 km away at a depth of 3480 m from the surrounding plateau (samples CIR139D-2 and CIR139D-3; Table DR2) are 79 Ma tholeiites; this suggests a clear association with a nearby mid-ocean ridge (Fig. 3).

3. The linear age-distance trend is not disrupted by the formation of this non-age-progressive plateau region between ca. 100 Ma and 70 Ma, which implies a connection to a stable or a constantly moving deep mantle source (Fig. 3).

4. Volcanism seems to have been propagating at ∼0.23° ± 0.01°/m.y. (26 ± 1 km/m.y.) until ca. 44–40 Ma, when it slowed thereafter to ∼0.17° ± 0.02°/m.y. (19 ± 2 km/m.y.) (Fig. 3).

This implies a more precise timing for slowdown of the African plate compared to the current estimate of between ca. 44 and 30 Ma (or occurring gradually over ∼14 m.y.; O’Connor et al., 1999, 2012). An alternative interpretation is that there was no slowdown, implying that the present-day location of the hotspot is ∼100 km closer to the present-day mid-ocean ridge (Fig. 3). This latter interpretation is supported by seafloor spreading models (e.g., Cande and Kent, 1992; Cande and Stegman, 2011; Colli et al., 2014) showing that ca. 44 Ma spreading rates in the South Atlantic did not change significantly, but were increasing from a pronounced minimum at the Cretaceous-Paleocene boundary.


The gravity field map derived from satellite altimetry (Sandwell and Smith, 2009) shows significant variations in basement structure along the Walvis Ridge. We use ship multibeam bathymetry combined with seismic profiles to define distinct types of morphology and crustal structure that can distinguish mid-ocean ridge from plume segments. We calibrate these observations with new geochronology for volcanic rock samples to develop an understanding that high-relief, eastern flanking volcanic point sources represent the track of the Tristan-Gough plume.

Our interpretation assumes that plume location is recorded where buoyant material in the immediate vicinity of the plume orifice (Sleep 2002) reached the plate surface via zones weakness in oceanic crust that was as much as 20 m.y. old, located in close proximity to a slowly spreading mid-ocean ridge. This tectonic setting resulted in the restriction of age-progressive volcanism to a few separated narrow chains of high-relief seamounts and discrete volcanic centers aligned along the eastern side of the Walvis Ridge. In contrast, broad low-relief plateau-like regions reflect the motion of the mid-ocean ridge relative to the source of plume material, and possibly overprinting by a widespread volcanic-tectonic event in the Eocene.

Evidence for a correlation between age-distance relation and plate morphology and structure provides new support for the concept that hotspot trails scattered across the 2000-km-wide South Atlantic bathymetric swell (i.e., Tristan-Gough, Discovery, Shona, Bouvet) arise where buoyant plume magma could find a route through the African plate, such as spreading boundaries or lines of weakness in the ocean crust or where it was sufficiently young (i.e., thin) and maybe also moving more slowly (O’Connor et al., 2012).

The inferred continuity of the age-distance relation of volcanism between widely separated plume segments points to a connection to a deep region in the mantle. This in turn has implications for using scattered hotspot tracks across the 2000-km-wide southeast Atlantic bathymetric swell to reconstruct relative motion between the African plate and the underlying large low shear velocity province (Burke et al., 2008; Torsvik et al., 2010; O’Connor et al., 2012). Discovering buried or disrupted plume tracks in other primary hotspot trails could improve our understanding of the relationship between plates and the deep mantle.

We thank Vincent Courtillot and Norman Sleep and an anonymous reviewer for comments that significantly improved this manuscript. We thank also Captain Schwarze and the crew of the Polarstern cruise ANT XXIII/5 and Captain Schmidt and the crew of the RV Maria S. Merian cruise MSM 17/2. Samples from repositories were provided as follows: Susan Humphries (Woods Hole Oceanographic Institution); Bobbie Conrad (Oregon State University); Louis Géli, Gilbert Floch, and Roger Hekinian (IFREMER—Institut Français de Recherche pour l’Exploitation de la Mer); Georg Lozefski and Rusty Lotti Bond (Lamont Doherty Earth Observatory). We thank Jan Wijbrans, Klaudia Kuiper, and Christel Bontje (VU University Amsterdam) for help with analyses. Financial support was provided by the German Federal Ministry of Education and Research and the German Research Foundation (DFG grant Jo-191/15-1).

1GSA Data Repository item 2015246, Figure DR1 (sample locations), Figure DR2 (age and K/Ca spectra), Table DR1 (sample information), Table DR3 (40Ar/39Ar results), analytical methods, and ArArCalc Excel data files, is available online at www.geosociety.org/pubs/ft2015.htm, or on request from editing@geosociety.org or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.