shear wave velocity structure of the southern african...

Geophys. J. Int. (2011) 186, 808–824 doi: 10.1111/j.1365-246X.2011.05072.x

Sei

smol

ogy

Shear wave velocity structure of the southern African upper mantlewith implications for the uplift of southern Africa

Aubreya Adams1,∗ and Andrew Nyblade1,2

1Department of Geosciences, The Pennsylvania State University, University Park, PA 16802, USA. E-mail: [email protected] of Geosciences, The University of the Witwatersrand, Johannesburg, South Africa

Accepted 2011 May 5. Received 2011 May 5; in original form 2010 October 9

S U M M A R YBroad-band seismic data from the southern African seismic experiment and the AfricaArraynetwork are used to investigate the seismic velocity structure of the upper mantle beneathsouthern Africa, and in particular beneath the Kaapvaal Craton. A two-plane approximationmethod that includes a finite frequency sensitivity kernel is employed to measure Rayleighwave phase velocities, which are inverted to obtain a quasi-3-D shear wave velocity model ofthe upper mantle. We find phase velocities for the Kaapvaal Craton and surrounding mobilebelts that are comparable to those reported by previous studies, and we find little evidencefor variation from east to west across the Namaqua–Natal Belt, a region not well imagedin previous studies. A high-velocity upper-mantle lid is found beneath the Kaapvaal Cratonand most of southern Africa. For the Kaapvaal Craton, the thickness of the lid (∼150–200km) is consistent with the lid thicknesses reported in many previous studies. The cratonic lidis underlain by a ∼100-km thick low-velocity zone with a 3.9 per cent maximum velocityreduction. By comparing the velocity model to those published for other Archean cratons,we find few differences, and therefore conclude that there is little evidence in the shear wavevelocity structure of the mantle to indicate that the southern African plateau is supported byan upper-mantle thermal anomaly.

Key words: Tomography; Surface waves and free oscillations; Seismic tomography; Cratons;Dynamics of lithosphere and mantle; Africa.

1 I N T RO D U C T I O N

A first-order topographic feature of the African continent is the ∼1km high elevations across the southern and eastern African plateaus.This feature, termed the African Superswell (Nyblade & Robin-son 1994), also extends into the southeastern Atlantic Ocean basinwhere bathymetry is shallower than expected. In eastern Africa,anomalous topography is supported in large part by buoyancy fromthermally perturbed upper mantle associated with the East AfricanRift System (e.g. Nyblade 2011 and references therein). The upper-mantle thermal anomaly there is manifest as a low-velocity zone(LVZ) with a large decrease in velocity (i.e. ∼10–12 per cent inthe S-wave velocity at depths of 150–350 km; Ritsema et al. 1998;Weeraratne et al. 2003; Nyblade 2011). In southern Africa, how-ever, it is unclear whether support for the anomalous topographyis caused by a thermal perturbation in the lithospheric mantle (Ny-blade & Sleep 2003), the sublithospheric upper mantle (e.g. Deenet al. 2006; Li & Burke 2006; Burke & Gunnell 2008), or the lowermantle (e.g. Lithogow-Bertelloni & Silver 1998; Gurnis et al. 2000;

∗Now at Chevron, Covington, LA 70433, USA.

Conrad & Gurnis 2003; Simmons et al. 2007) or some combinationof all three.

Previous studies, described in greater detail in Section 2, haveemployed a range of seismic imaging methods to characterize upper-mantle structure beneath southern Africa and determine if seismicvelocities indicate the presence of a thermal anomaly sufficientlylarge to support the high plateau topography observed there. Manystudies find only a small decrease in sublithospheric velocities,providing little support for the suggestion that the topography ofsouthern Africa results from an upper-mantle thermal anomaly (e.g.Zhao et al. 1999; Freybourger et al. 2001; James et al. 2001; Saltzer2002; Fouch et al. 2004b; Larson et al. 2006; Chevrot & Zhao2007; Hansen et al. 2009). Other studies, however, suggest thatsouthern Africa is characterized by a sublithospheric mantle with apronounced LVZ (e.g. Priestley 1999; Li & Burke 2006; Priestleyet al. 2006; Wang et al. 2008), indicative of a substantial thermalanomaly in the upper mantle.

In this study, we investigate further the seismic velocity struc-ture of the upper mantle under southern Africa and reevaluatethe evidence for a thermal anomaly. We focus our investigationon the Kaapvaal Craton, where data coverage is best, but alsopresent results for some surrounding geologic areas, including theNamaqua–Natal Belt and the Kheis Province (Fig. 1). We use a two

808 C© 2011 The Authors

Geophysical Journal International C© 2011 RAS

Geophysical Journal International

Upper mantle velocities in Southern Africa 809

Figure 1. Topographic map of southern Africa showing the locations of the major Precambrian terrains (TML—Thabazimbi-Murchison Lineament). Terrainboundaries are taken from Eglington & Armstrong (2004). Political boundaries are shown by dashed lines. Inset map shows the location of the larger map withreference to Africa. The white box indicates the boundary of the study area.

plane wave approximation method (Forsyth & Li 2005) to measurephase velocities for periods between 17 and 167 s, which we in-vert for shear wave velocity structure following the method of Juliaet al. (2000). We employ a data set combining new AfricaArray datatogether with data from the Southern Africa Seismic Experiment(SASE) and Global Seismic Network (GSN) stations. In additionto using an expanded data set, our study differs from previous stud-ies in that in our modelling we account for finite frequency effectson Rayleigh wave propogation and use new estimates of crustalstructure as a priori model constraints.

2 B A C KG RO U N D

2.1 The geology of southern Africa

The Southern African Shield, consisting of Archean cratons and Pro-terozoic mobile belts, is characterized topographically by anoma-lously high elevations averaging more than 500 m higher than theglobal continental average elevation (Nyblade & Robinson 1994).This study focuses on upper-mantle structure beneath the centreportion of the Shield, which is composed of the Archean KaapvaalCraton, the Paleoproterozoic Kheis Province and the Neoprotero-zoic Namaqua–Natal Belt (Fig. 1).

The Kaapvaal Craton, which covers an area of 1.2 million squarekilometres and formed between 2.7 and 3.7 Ga (de Wit et al. 1992),can be divided into the Pietersburg, Witwatersrand, Kimberley andSwaziland terrains (de Wit et al. 1992; Eglington & Armstrong2004). The Swaziland terrain is the oldest part of the KaapvaalCraton and consists of the Ancient Gneiss complex (3.55–3.68 Ga)

and the southern Barberton Greenstone Belts (3.4–3.55 Ga; Fig. 1;Eglington & Armstrong 2004). The basement of the younger Wit-watersrand terrain to the north and west includes granitoid plutonsand the northern Barberton Greenstone Belts. The Swaziland andWitwatersand terrains are separated by the Inyoka Fault, marking thesuture between the two terrains, which occurred at approximately3.2 Ga (de Wit et al. 1992; Eglington & Armstrong 2004).

The basement of the Kimberley terrain (Fig. 1) includes green-stone belts, gneisses and deformed volcanosedimentary sequences(e.g. Anhaeusser 1999; Schmitz et al. 2004). The Kimberley ter-rain was accreted to the western margin of the combined Swazilandand Witwatersrand terrains approximately 2.88–2.93 Ga ago. TheColesberg Magnetic Lineament demarcates the location of the su-ture (Schmitz et al. 2004). To the north, the Pietersburg terrain wasaccreted between 2.8 and 3.1 Ga, and the Thabazimbi-MurchisonLineament forms the southern boundary of this terrain (Fig. 1;Eglington & Armstrong 2004).

The assembly of the Kaapvaal Craton was completed by 2.75 Ga(e.g. Tinker et al. 2002; Eglington & Armstrong 2004). Followingits formation, additional crustal material was added to the cratonthrough sedimentation and magmatism. Between 2.06 and 2.1 Ga,one of the largest layered mafic intrusions anywhere, the BushveldComplex (Fig. 1), was emplaced into the northern part of the craton.And across the craton, basement rocks have been covered by a seriesof sedimentary basins ranging in age from 1.8 to 3.1 Ga.

The Kaapvaal Craton is bordered on three sides by mobile belts.To the north, the Archean Limpopo Belt joins the Kaapvaal Cratonto the Archean Zimbabwe Craton (Fig. 1). The Limpopo Belt consistof highly metamorphosed granite-greenstone and granulite terrains,

C© 2011 The Authors, GJI, 186, 808–824


810 A. Adams and A. Nyblade

Figure 2. Locations of stations used in this study. Circles indicate SASE stations. Triangles indicate GSN stations. Squares indicate AfricaArray stations.Terrain boundaries are the same as shown in Fig. 1.

which underwent a series of orogenic events between 2.0 and 3.0Ga during the suturing of the Kaapvaal and Zimbabwe Cratons(McCourt & Armstrong 1998; Kramers et al. 2006).

To the south, the Kaapvaal Craton is bordered by theNamaqua–Natal Belt (Fig. 1). The Namaqua–Natal Belt is a bandof igneous and metamorphic rocks from the Namaqua or KibaranOrogeny (1.0–1.2 Ga). Much of the belt has been covered byphanerozoic sediments, but surficial exposures are found in theeastern Natal sector and the western Namaqua sector (Thomaset al. 1994; Cornell et al. 2006). The Namaqua sector is sepa-rated on its northeastern boundary from the Kaapvaal Craton bythe Kheis Province (Fig. 1), a fold-thrust belt of low-grade meta-morphosed supracrustal rocks, with structures thought to pre-datethe Namaqua Orogeny (approx. 2.0 Ga; Moen 1999; Eglington &Armstrong 2004; Cornell et al. 2006). The boundaries between theNamaqua–Natal Belt and the Kheis Province and between the KheisProvince and the Kaapvaal Craton are not well defined due to a lim-ited number of dates and difficulties with defining boundaries basedon tectonic structures alone (Moen 1999).

2.2 Previous studies of southern Africa

Southern Africa has been the subject of numerous seismic studiesover the past several decades. Some of the earliest seismic studiesused body wave traveltimes from tremors in the Witwatersrandgold mines to determine average crustal velocities and thickness(e.g. Gane et al. 1949, 1956; Willmore et al. 1952). These studiesconcluded that there was little variation in crustal structure, andthat the crust was underlain by high-velocity mantle rock. Blochet al. (1969) used seismograms from regional events recorded onworldwide standard seismograph network and other local seismic

stations to measure phase and group velocities for surface waves.From their measurements, they obtained shear wave models of thecrust and upper mantle that confirmed the existence of a high-velocity upper mantle. But their results suggested greater variabilityin crustal structure than reported in previous studies, as well as thepresence of several thin low-velocity layers in the upper mantle(Bloch et al. 1969).

The deployment of the Southern African Seismic Experiment(SASE) stations (Carlson et al. 1996) and, more recently, AfricaAr-ray stations (www.africaarray.org), has allowed for renewed andmore detailed studies of the crust and upper mantle beneath south-ern Africa (e.g. Nguuri et al. 2001; Niu & James 2002; Li & Burke2006; Nair et al. 2006; Kgaswane et al. 2009).

Several studies have used SASE data and receiver function analy-sis to estimate crustal thickness and VP/VS ratios in southern Africa(e.g. Nguuri et al. 2001; James et al. 2003; Nair et al. 2006). Thesestudies found that crustal thickness beneath the Kaapvaal Cratonvaries from 35 to 45 km, and is slightly thicker (40- to 53-kmthick) beneath the Bushveld Complex. For the Kheis Province andthe Namaqua–Natal Belt, they obtained crustal thickness estimatesof 40 and 40–50 km, respectively. Nguuri et al. (2001) and Nairet al. (2006) also report estimates of VP/VS ratios greater than 1.78for these two terrains, values which indicate the presence of maficlithologies within the crust.

More recently, Kgaswane et al. (2009) conducted a study ofcrustal and uppermost mantle structure in southern Africa by jointlyinverting receiver functions and Rayleigh wave group velocities forboth SASE and AfricaArray stations. They found that the averagecrustal thickness within the four main terrains of the KaapvaalCraton varied between 37 and 39 km, with a few kilometres ofthickening beneath the Bushveld Complex. They report a similar

C© 2011 The Authors, GJI, 186, 808–824



Figure 3. Event locations. Grey circles indicate events recorded by the SASE array and used in Li & Burke (2006). Black circles indicate events recorded byAfricaArray stations. Large grey circles indicate distance intervals of 30◦ from our study area.

crustal thickness of 39 km for the Kheis Province, and within theNamaqua–Natal Belt, crustal thickness that ranges from 26 km nearthe Cape Fold Belt to 46 km near the Kaapvaal Craton. These resultsare generally consistent with those reported by previous studiesusing SASE data.

As mentioned in Section 1, many seismic studies have attemptedto characterize the seismic structure of the lithospheric mantle (i.e.seismic lid) in southern Africa and determine whether or not thereis a significant LVZ in the upper mantle beneath it. Every study hasfound that southern Africa is underlain by a high-velocity upper-mantle lid that is at least 150-km thick, but the thickness of the lidand the amount of velocity reduction beneath the lid varies con-siderably between studies. A brief review of these studies follows,and where the studies present results relevant to just the KaapvaalCraton, we highlight those results.

Priestley (1999), building on the model of Qui et al. (1996),developed a shear wave model for southern Africa by matchingregional S waveforms and previously reported phase velocity mea-surements (Bloch & Hales 1968; Bloch et al. 1969). The Priestley(1999) model included a high-velocity lid extending to a depth of160 km, underlain by an LVZ with minimum velocities between4.35 and 4.53 km s−1 and a thickness of ∼180 km. These findingsare consistent with the more recent studies by Preistley et al. (2006,2008). Priestley et al. (2006) constructed an SV model for south-ern Africa using fundamental and higher mode surface waves, and

found a lithospheric thickness of 175 km. They also used kimber-lite nodules to estimate a geotherm for southern Africa, and from itestimated a lithospheric thickness of 204 km. Priestley et al. (2008)constructed an SV model for the entire African continent usingfundamental and higher mode Rayleigh waves and report high ve-locities to a depth of 170 km beneath the Kalahari Craton, consistentwith Priestley et al. (2006).

Zhao et al. (1999) used P waveforms and traveltimes from a SouthAfrican mine event recorded on a network of stations in Tanzaniato model the P and SV velocity structure beneath southern Africa.They found that P and SV velocities in the upper 300 km smoothlyincrease with depth, providing little evidence for an LVZ beneaththe lithosphere.

Freybourger et al. (2001) measured Rayleigh and Love wavephase delays for 10 earthquakes recorded on SASE stations, andthen inverted for S velocities in the mantle. They did not find anyevidence for a significant velocity decrease within the upper 200km of the mantle, but did find that the upper 100 km of the mantlewas radially anisotropic (∼4 per cent). Saltzer (2002) also obtainedsimilar results using phase delays from Rayleigh and Love wavesmeasured at SASE stations.

Simon et al. (2002) used regional P-wave traveltimes recordedon SASE stations to estimate the seismic structure of the upper 800km of the mantle beneath southern Africa. They did not observeany significant decrease in velocities within the upper 400 km of

C© 2011 The Authors, GJI, 186, 808–824



Figure 4. Ray path coverage for 50 s Rayleigh waves. Station symbols aredescribed in Fig. 2. Terrain boundaries are the same as shown in Fig. 1.Political boundaries are shown by thick grey lines.

the mantle, and they also found little evidence for thinning of thetransition zone, indicating the absence of a thermal anomaly at thosedepths.

A P and S wave tomography study by Fouch et al. (2004b) usingSASE data found a high-velocity upper-mantle lid (∼0.6 per centfaster than their reference model for P and SV velocities) extend-ing to depths between 250 and 300 km, with no evidence for alow-velocity asthenosphere. They found reduced velocities in theupper mantle beneath the Bushveld Complex and suggested that thereduction in velocity may be due to alteration during the emplace-ment of the Bushveld Complex or during the Karoo magmatic event(∼180 Ma). They also found reduced velocities in the upper mantlebeneath the Cape Fold Belt and the Namaqua–Natal Belt.

Silver et al. (2004) analysed azimuthal anisotropy in southernAfrica obtained from S-wave splitting measurements (Silver et al.2001; Fouch et al. 2004a). The fast direction of anisotropy is pri-marily in an east to northeast direction, and is most well developedalong the western and northern boundaries of the Kaapvaal Craton.Silver et al. (2004) suggested that most of the anisotropy is withinthe lithospheric mantle and that it developed during orogenic eventsprior to 1.8 Ga. Others have attributed the pattern of anisotropy toflow in the sublithospheric mantle (e.g. Vinnik et al. 1995).

Using a combination of xenolith data and Rayleigh wave phasevelocities measured between SASE station pairs in the KaapvaalCraton and the Kheis Province, Larson et al. (2006) developed ashear wave model for southern Africa that does not include a sig-nificant LVZ in the upper mantle. Furthermore, Larson et al. (2006)argued that their measured phase velocities cannot be matched by ashear wave model with a significant LVZ such as that proposed byPriestley (1999).

Li & Burke (2006) used a two plane-wave approximation method(Forsyth & Li 2005) to measure Rayleigh wave phase velocities forthe SASE network, and then inverted them for a model of mantleSV velocities. Their models show a high-velocity upper-mantle lid,with thickness varying from 80 and 180 km between the mobile beltsand the cratons, underlain by an LVZ between the depths of 160 and260 km, where velocities are on average 4 per cent slower than lidvelocities. In the southern Kaapvaal Craton, Li & Burke (2006) re-

ported that the minimum LVZ velocity is 5 per cent slower than themaximum lid velocity. They interpreted this velocity reduction to beindicative of a thermally perturbed upper mantle that distinguishesthe sublithospheric mantle under the Kaapvaal Craton from thesublithospheric mantle under Archean cratons in other parts ofthe world. Their models also show that at depths less than 100km, the northern Kaapvaal Craton is slow relative to the southernKaapvaal Craton. In a study following the same methodology butaccounting for finite frequency effects, Li (2011) reached a sim-ilar conclusion that the lithosphere beneath the Kaapvaal Cratonextends to depths of 160–200 km.

Deen et al. (2006) used a combination of geochemical datafrom xenoliths and seismic velocities from a global velocity model(Grand 2002) to model density and geotherms beneath Africa andareas surrounding other cratons. From these models, they made in-ferences about the composition of the lithospheric mantle, reportingthat the composition of the lithospheric mantle beneath the Kaap-vaal Craton is a depleted ‘Archon-type’.

Using phase delays from fundamental mode Rayleigh wavesrecorded on SASE stations, Chevrot & Zhao (2007) inverted fora 3-D shear wave model of southern Africa that included finitefrequency effects on wave propagation. Their results show littleevidence for an LVZ under the high-velocity lid of the KaapvaalCraton, with fast velocities extending to a depth of at least 250 kmbeneath most of the Kaapvaal Craton. They also suggested that theColesberg Magnetic Lineament coincides with a velocity contrastbetween the Witwatersrand and Kimberley terrains at depths be-tween 100 and 200 km, and that the lower velocities they image inthe southeastern portion of the Kaapvaal Craton could have resultedfrom metasomatism and lithospheric heating during the initial for-mation of the Atlantic Ocean ca. 180 Ma.

Wang et al. (2008) used triplicated P and S phases from anevent in East Africa recorded on the SASE stations to model P andSH velocities in the upper mantle beneath southern Africa. Theyfind evidence for a 150-km thick, high-velocity upper-mantle lidunderlain by a 200-km thick LVZ, where SH velocities are reducedby 5 per cent and P velocities are reduced by 2 per cent.

In a study of S-wave receiver functions and Rayleigh wave phasevelocities for SASE, AfricaArray and GSN stations, Hansen et al.(2009) estimated that the lithosphere under the Kaapvaal Cratonextends to a depth of 160 km, and that an LVZ of no more than 90-km thick with an approximate 4 per cent decrease in velocity existsunder the lithosphere. They argued that the depth and thicknessof the LVZ is not necessarily anomalous when compared to theupper-mantle structure beneath other Archean cratons.

Pedersen et al. (2009) inverted surface wave dispersion curvesfor shear wave velocity structure beneath the Kaapvaal Craton, theSlave Province, the cratonic region of South-Central Finland and theYilgarn Craton. For the inversion for structure beneath the Kaap-vaal Craton, Pedersen et al. (2009) used a combination of dispersioncurves published by Larson et al. (2006) and Li & Burke (2006).They used dispersion curves for both the southern Kaapvaal Cratonand the northern Kaapvaal Craton, which includes the BushveldComplex. Pederson et al. (2009) concluded that lithospheric shearwave velocities were similar beneath the Kaapvaal Craton, SlaveProvince and South-Central Finland, but that velocities in the litho-sphere of the Yilgarn Craton were faster. They also concluded thatthere is an LVZ beneath the Kaapvaal Craton, but that its depthextent and velocity reduction are poorly resolved.

Begg et al. (2009) used a combination of compositional mea-surements from xenoliths with the seismic tomography model fromGrand (2002) to characterize the upper mantle beneath Africa. Their

C© 2011 The Authors, GJI, 186, 808–824



Figure 5. Inversion node locations. Small circles indicate background nodes. Stars represent nodes in the northern Kaapvaal Craton. Nodes marked by squaresare in the southern Kaapvaal Craton, and triangles indicate the Kheis Province. Large circles represent nodes in the Namaqua–Natal Belt. Terrain boundariesare the same as in Fig. 1.

Figure 6. Average one dimensional phase velocity dispersion curve for southern Africa. For comparison, results from Li & Burke (2006) are shown by adashed line. Error bars indicate one standard deviation.

C© 2011 The Authors, GJI, 186, 808–824



Figure 7. Average phase velocity dispersion curves for areas shown inFig. 5. Error bars indicate one standard deviation.

model shows a high-velocity lid extending to at least 250 km, with noevidence for reduced velocities beneath this lid when compared tothe mantle beneath the rest of Africa at comparable depths. Based onxenolith geochemistry, they estimate that the subcontinental litho-spheric mantle extends to a depth of approximately 200 km beneaththe Kaapvaal Craton.

In addition to the growing number of seismic studies of theupper mantle beneath the Kaapvaal Craton, a number of studieshave also employed xenolith composition and thermo-barometrymeasurements to characterize the cratonic upper mantle of the re-gion (e.g. Eaton et al. 2009; Rudnick & Nyblade 1999 and refer-ences therein). For example, Eaton et al. (2009) study xenolith andxenocryst thermo-barometry and report that the lithospheric thick-ness beneath the Kimberley region of southern Africa is between195- and 215-km thick, and do not find evidence of asthenosphericmaterial. James et al. (2004) used xenolith composition to estimateP- and S-wave velocities and densities, and showed that althoughthere is a decrease in velocity with depth in the lithospheric mantle,the minimum velocity obtained is 4.55 km s−1, which is greater thanwhat is typically considered to be an asthenospheric velocity.

3 P H A S E V E L O C I T Y I N V E R S I O N

3.1 Data and processing

Our study area is 1650 km × 1442 km and includes the KaapvaalCraton and the surrounding Kheis Province and Namaqua–NatalBelt, but not the Limpopo Belt, Cape-Fold Belt or Zimbabwe Craton(Fig. 1). This area was selected to allow for investigation of man-tle structure beneath the interior portion of the Southern AfricanShield, where our data set provides the best ray coverage, whilesimultaneously minimizing the total aperture between stations. Theadditional constraint of limiting the aperture between stations isnecessary because of the assumption of planar waves on a spheri-cal Earth when employing the two plane-wave method to measurephase velocities (Donald Forsyth, personal communication, 2009).

In this study, we use data from a large number of broad-band sta-tions spread across our study area (Fig. 2). The stations belong to theGSN (stations LBTB, BOSA and SUR), the AfricaArray network(www.africaarray.org), and the IRIS/PASSCAL SASE (Carlsonet al. 1996).

The SASE experiment consisted of 78 broad-band stations de-ployed from April 1997 to July 1999. The stations were installedin a linear swath extending from the Cape Fold Belt in the south-west to the Zimbabwe Craton in the northeast (Fig. 2). Thirty-two stations remained in the same position during the entire ex-periment, whereas 23 stations were redeployed after 1 yr. Streck-eisen STS-2 seismometers were used in these stations; other detailsabout this experiment can be found in Carlson et al. (1996) or at(www.dtm.ciw.edu/mantle/kaapvaal). For this study, we use datafrom 69 SASE stations, located south of 22◦S latitude.

The AfricaArray permanent network consists of broad-band sta-tions throughout the continent. We use data from nine stations inSouth Africa in operation since 2006. Seismometers at these sta-tions include KS2000, Gulralp 3T and Gulrap 40T models. Thelocations of the AfricaArray stations improve data coverage in thesoutheastern part of the Southern African Shield where no SASEstations were located.

For events recorded by SASE stations within our study area,we used windowed and filtered data that was processed by Li &Burke (2006), and have followed a similar processing techniquefor events recorded on AfricaArray and GSN stations. Events wereselected to add to the data set only if they had a high signal-to-noiseratio, and were well recorded on at least five stations. Individualrecords were selected for inclusion in the inversion if the unfiltereddata had no clear interference patterns or timing problems beforefiltering. Because it is important to model both constructive anddestructive interference of the two plane waves within our studyarea, low signal-to-noise ratio at a single station for a generallywell-recorded event was not a sufficient reason to eliminate therecord. By including data from both SASE stations (Li & Burke2006) and AfricaArray and GSN stations, we use a total of 101events with ms > 5.8, providing good azimuthal coverage (Fig. 3).To illustrate the data coverage, in Fig. 4 we show ray paths for eventstation pairs for 50 s Rayleigh waves.

Each record was processed separately for 20 periods ranging from17 to 167 s. At each period, records were filtered using a zero-phase-shift 10-mHz wide bandpass filter centred around each frequencyof interest. A single window length was manually selected for eachperiod-event combination such that it contained the Rayleigh wavearrivals at all stations. For each period, the same window length wasused for all stations recording a particular event. This windowingprocess limits noise and interference from body wave arrivals andthe surface wave coda. The phase and amplitude of the Rayleighwave in each seismogram were then calculated using Fourieranalysis.

3.2 Inversion methodology

Many surface tomography studies assume that the incoming sur-face wavefield from an earthquake can be represented by a singleplane wave propagating along a great-circle path from a sourceto a receiver. Although this assumption is accurate for a radiallysymmetric Earth, it neglects the multipathing effects caused by het-erogeneities a wavefield may encounter along a real path. To betteraccount for the non-great circle propagation of an incident wave, wemodel the wavefield for each event in this study using the methodof Forsyth & Li (2005), by estimating the phase, amplitude and az-imuth of two hypothetical incoming plane waves before solving forvelocity parameters. In the initial approximation of the incomingwavefield, velocity is held constant and iterative simulated down-hill simplex annealing is used to approximate the wave parameters

C© 2011 The Authors, GJI, 186, 808–824



Figure 8. Phase velocity maps for selected periods.

(Press et al. 1992). This process is run for each event, without de-pendancies on other events, regardless of the source data set (i.e.SASE or AfricaArray). After solving for the initial estimates ofwave parameters for each event, a linearized inversion (Tarantola& Valette 1982) is used to solve simultaneously for corrections tothe combined data set of modelled wave parameters from all eventsand to the current velocity model by minimizing errors, in a leastsquares sense.

To implement this methodology, the study region was dividedinto a set of 667 nodes, with an interior node spacing of 0.5◦ andan exterior node spacing of 1◦ (Fig. 5). An a priori error estimateof 0.15 km s−1 was assigned to each of the nodes to determinedamping values, whereas the error estimate for the exterior nodeswas assigned a value of 1.5 km s−1. This differential assignmentof estimated error allows any deviations of the incoming wavefieldfrom a single great circle path outside of the study area, whichare not accounted for by the two plane wave approximation, to bepreferentially mapped into the exterior nodes, outside of our regionof primary interest.

To obtain phase velocity maps for each period, a series of inver-sions were preformed, starting with a simple inversion for an average1-D velocity dispersion curve for all of the nodes in the study area.We chose to use the average southern Africa phase velocities fromLi & Burke (2006) as our starting model for the average 1-D in-version, but tests that we performed using different starting modelsindicated that the inversion is sensitive to the starting model onlyfor periods of 20 s or less. The average 1-D curve we obtained from

this initial inversion was then used as a starting model to invertfor average 1-D velocity curves within the craton and mobile beltsparameterized in Fig. 5. Velocity curves for the craton and mobilebelts were in turn used as starting models for an inversion for 2-Dvelocity maps. This approach minimizes the bias imposed by thestarting model on the 2-D inversions for phase velocity maps.

The 1-D models in our study used a Gaussian sensitivity functionto represent sensitivity to structures off of the great-circle pathsof the two plane waves. Many studies, however, have shown thatbecause the frequencies of surface waves are finite, this Gaussianapproximation is only valid for structures larger than the character-istic wavelength of the period being measured (e.g. Zhou et al. 2004;Yang & Forsyth 2006a). Zhou et al. (2004) developed a method ofcalculating finite frequency sensitivity kernels for both the phaseand amplitudes of surface waves from source to receiver. Theydemonstrated that the sensitivity of surface waves to off path struc-ture varies along the assumed great-circle travel path, as well asaway from the path in both polarity and amplitude. Yang & Forsyth(2006a) applied the kernels developed by Zhou et al. (2004) to re-gional surface wave tomography studies using both one and twoplane waves, and demonstrated that the resolution of small featurescan be improved by accounting for the finite frequency effects ofsurface wave propagation. Therefore, in our 2-D models, we ac-count for the finite frequencies effects of surface wave propagationfor our two plane wave tomography following the method describedand applied by Yang and Forsyth (2006a,b). We also solve for anaverage azimuthal anisotropy at each period.

C© 2011 The Authors, GJI, 186, 808–824



Figure 9. Plot showing the fast direction for azimuthal anisotropy as afunction of period and per cent change in velocity. At each period, the fastpolarization direction is indicated by the direction of the line relative to thenorth arrow. The position of the line along the y-axis indicates the per centchange in velocity at that period.

3.3 Phase velocity results

Average 1-D phase velocities for the entire study area are shownin Fig. 6, and are compared with phase velocities reported by Li& Burke (2006). Good agreement between the dispersion curvesis found. Fig. 7 shows the 1-D phase velocity dispersion curve foreach of the areas within the study region. The Namaqua–Natal Belthas significantly lower velocities than the other geologic regions at

all periods. The velocities of the Kheis Belt are comparable to theKaapvaal Craton at periods of 100 s and less. The phase velocitiesof the northern and southern Kaapvaal Craton are similar, but atperiods less than 100 s, the southern part of the craton is slightlyfaster, whereas the northern part of the craton is faster at periodsgreater than 100 s.

Fig. 8 shows 2-D velocity maps for selected periods. At all pe-riods, the Kaapvaal Craton is faster than the Namaqua–Natal Beltand the Kheis Belt. Velocities are highest in the central region of theKaapvaal Craton, and we do not observe systematic differences be-tween the Kimberley, Witwatersrand and Swaziland terrains acrossthe southern Kaapvaal Craton at any period. However, variations onthis scale, if present, might not be well resolved at all periods. In thenorthern Kaapvaal Craton, low velocities are observed around theBushveld Complex for periods of 50 s and less. At longer periods,however, velocities are comparable throughout the Kaapvaal Cra-ton. When compared to the Kaapvaal Craton, velocities are lower atall periods for the Namaqua–Natal Belt, but there are no significantdifferences in velocity between the Namaqua and Natal Provinces.

As above-mentioned, the 2-D velocity models not only includefinite-frequency wave propagation, but also include a solution foran average azimuthal anisotropy at each period. Fig. 9 shows thefast direction and per cent velocity change at each period. The fastdirection is predominantly east-northeast, which is consistent withfindings from previous studies of shear wave splitting (e.g. Silveret al. 2004). Variations in the fast direction and per cent velocitychange at different periods are small, indicating similar anisotropiccharacteristics at different depths.

The error bars shown in Figs 6 and 7 indicate a formal errorestimate of one standard deviation calculated during the 1-D in-versions. These error estimates are probably unrealistically small,and a more reasonable estimate of error is obtained for the 2-Dinversions based upon the diagonals of the covariance matrices.Fig. 10 illustrates model uncertainty at 50 s for this study when onlythe data from Li & Burke (2006) is used, without accounting for

Figure 10. Model uncertainty for phase velocities at a period of 50 s for (a) this study and (b) the data set used by Li & Burke (2006) without includinganisotropy or finite frequency kernels in the inversion. The solid line indicates the boundary where uncertainty is less than 0.06 km s−1 for our data set. Thispolygon is used in Figs 8 and 11 to crop the region displayed.

C© 2011 The Authors, GJI, 186, 808–824



Figure 11. Resolution tests for selected periods. Input checkers for (a)–(d) are 2◦ squares, whose locations are shown by the dotted lines in (d). For (e) and (f),input checkers are 3◦ squares. For (g), the input was a single 2◦ square.

anisotropy or finite frequency effects. The figure illustrates the im-proved resolution achieved by the addition of these features in thisstudy. These plots are based, in part, on the a priori uncertainties,which were chosen to be conservative and consistent with previousstudies. Given this dependency, the error estimates in Fig. 10 arebest interpreted as relative errors within the study area. We choose

to crop our 2-D maps of phase velocity to show only areas withuncertainty less than 0.06 km s−1, which is an intermediate valuecompared to those used in similar studies (e.g. Weeraratne et al.2003; Li & Burke 2006).

To further examine the resolution of the phase velocity maps, wehave preformed checkerboard tests. The resolution matrix at each

C© 2011 The Authors, GJI, 186, 808–824



period was used to map how well velocity perturbations in a patternsquare checkers of a variety of sizes could be resolved. The shape,location and amplitudes of the 2-degree checkers are well resolvedfor periods less than 125 s and 3-degree checkers are well resolvedat greater periods (Fig. 11). As expected, resolution is best wherewe have the highest concentration of stations in the centre of thestudy area. There is slight smearing of the checkerboard structure tothe northwest and southeast, but the general shape of the checkersis still recovered. Resolution is best for periods of 25–87 s. Theresolution of a single 2-degree square anomaly among an otherwisehomogeneous model was also tested for 100 s period (Fig. 11),further indicating that model resolution is good at the 2 degreelevel, at least within the centre of the study area. Results fromthe checkerboard resolution tests for other periods can be found inAdams (2010).

4 I N V E R S I O N F O R S H E A R V E L O C I T YS T RU C T U R E

4.1 Method of inversion for shear wave velocity

We invert our phase velocity measurements for shear wave veloc-ities to obtain a quasi-3-D velocity structure of the upper mantlebeneath southern Africa following the method described by Parket al. (2008). In this method, at each of the interior nodes (Fig. 5),we first invert for a 1-D shear wave profile, with crustal structureconstrained by a priori information, and then apply a lateral smooth-ing operator to the 1-D models. Lateral smoothing is applied usinga Gaussian weight with a characteristic length of 150 km away fromeach node. Adams (2010) shows results from tests using a range oflateral smoothing lengths.

To invert phase velocity dispersion curves for shear wave velocityprofiles at each grid node, we use the inversion methodology of Juliaet al. (2000). This method was designed to jointly model dispersionand receiver function data using an iterative damped generalizedlinearized inversion by minimizing a weighted least squares normof the data. To apply the method of Julia et al. (2000) to our inversionusing only phase velocities, the weight assigned to receiver functiondata is set to zero.

For the inversions, we use 1-D models constructed with 6-km-thick layers extending to a depth of 260 km, and 10-km-thick layersextending to a depth of 810 km. Below 810 km depth, three 30-kmthick layers are included, followed by one 50-km-thick layer andthen a half-space. To account for the sensitivity of phase velocitiesto crustal structure, we employ a different starting model for theupper 50 km (crust and top-mantle layer) for the nodes in each geo-logic region, taken from Kgaswane et al. (2009; Table 1). Below 50km, the velocities for each starting model were linearly tapered toa velocity model at 410-km depth obtained from inverting the aver-aged 1-D phase velocity curve for the entire region shown in Fig. 6.

During each inversion, a minimal smoothing parameter was appliedto the crustal layers to prevent large and unrealistic variations invelocity between layers. Also, to constrain Moho depths to thosereported by Kgaswane et al. (2009), while simultaneously allowingfor a large velocity change across the Moho, a 1-km-thick layer wasincluded above and below the Moho. Velocities in the 1-km-thicklayers were weighted, but not fixed or smoothed.

4.2 Shear wave velocity results

Our quasi-3-D shear wave model is illustrated with several depthslices and cross-sections in Fig. 12. In the upper 200 km of themodel, similar velocities are found within the Witwatersrand, Swazi-land and Kimberley terrains, which comprise the southern portionof the Kaapvaal Craton, and there is little evidence for system-atic variations between terrains. Beneath this part of the KaapvaalCraton, the uppermost mantle has maximum velocities of approxi-mately 4.65 km s−1. At depths between 100 and 250 km, shear wavevelocities are reduced and reach a minimum velocity of approxi-mately 4.5 km s−1 at 175 km depth (i.e. a 3.9 per cent reductionfrom the highest lid velocity).

Velocities in the northern Kaapvaal Craton are similar beneaththe Bushveld Complex and the Pietersburg Terrain. In these regions,velocities do not vary significantly with depth throughout the up-permost mantle, and remain between 4.5 and 4.6 km s−1 to a depthof at least 150 km. Velocities reach a minimum of 4.45 km s−1 ata depth of 175 km, which is comparable, within our uncertainties,to velocities at that depth beneath the southern part of the KaapvaalCraton.

Velocities in the upper 150 km of the Kheis Province are ashigh as 4.65 km s−1. At depths greater than 150 km, velocities arereduced, reaching a minimum velocity of 4.4 km s−1 at 250 kmdepth. This represents a 5 per cent reduction in velocity from thehighest velocities in the lid.

Velocities in the uppermost mantle beneath the Namaqua–NatalBelt vary between 4.5 and 4.6 km s−1, but these variations do notappear to be correlated from east to west. At depths between 125and 300 km, velocities are reduced by up to 5 per cent, reaching aminimum velocity of 4.35 km s−1 at 200 km depth. Velocities at alldepths are comparable in the eastern Natal and western Namaquaregions.

Fig. 13 shows average 1-D models for each geologic region,obtained by averaging the velocity at each depth for each node inthe region, along with one standard deviation of the mean value. Forcomparison, the AK135 model is also shown (Kennett et al. 1995).All regions exhibit a high-velocity lid at shallow mantle depths,underlain by a reduction in velocity, but the absolute velocities andthe depth at which the velocity reduction occurs varies between thecraton and mobile belts. At depths less than 125 km, all regions

Table 1. Starting crustal models for shear wave velocity inversion.

Crustal VS < 20 km VS 20–30 km VS > 30 km Uppermost mantle SGeologic terrain thickness (km) (km s−1) (km s−1) (km s−1) velocity (km s−1)

Average for S. Africa 38 3.47 3.89 4.06 4.57Witwatersrand Terrain 37 3.53 3.83 4.00 4.60Swaziland Terrain 39 3.41 3.91 4.10 4.60Kimberley Terrain 37 3.62 3.73 3.90 4.60Bushveld Complex 41 3.38 4.00 4.00 4.50Kheis Province 42 3.66 3.77 4.10 4.65Namaqua–Natal Belt 45 3.56 4.03 4.30 4.60

C© 2011 The Authors, GJI, 186, 808–824



Figure 12. Shear wave velocities for selected depths (a–d) and cross-sections (e–h). Locations for the cross-sections are shown on the 300-km-depth slice (d).NNB, Natal–Namaqua Belt; KC, Kaapvaal Craton; KB, Kheis Belt.

within our study area are faster than AK135, but at greater depths,all regions are comparable to or slightly slower than AK135.

5 D I S C U S S I O N

In this section, we first compare our model results to models fromseveral other studies, and then investigate to what extent the upper-

mantle structure beneath southern Africa may be thermally per-turbed by comparing our model results to the upper-mantle shearvelocity structure of other cratonic regions. Because there is ev-idence of radial anisotropy in the upper mantle beneath southernAfrica (Saltzer 2002), we only use SV models in our comparisons.And we also limit our comparisons to models for the terrains of thesame age as those in our study area.

C© 2011 The Authors, GJI, 186, 808–824



Figure 13. Comparison of shear wave velocity profiles from this study (solid line), Li & Burke (2006) (dotted line) and AK135 (dashed line). One standarddeviation of the mean shear velocity at each depth is shown with a grey swath, indicating the variability of velocities within a terrain.

Other published SV velocity models for the Kaapvaal Craton areshown in Fig. 14. Relative to our shear wave model, the modelby Hansen et al. (2009) is faster at depths greater than 240 kmand slower at shallower depths. The Hansen et al. model fits thephase velocity measurements published by Li & Burke (2006),illustrating the trade-off between velocities at different depths whenfitting dispersion curves. Nevertheless, when taking into accountuncertainties in the models, the models are generally similar. The

model from Larson et al. (2006) has a faster lid than our model, andshows little reduction in velocity beneath the lid. Li & Burke (2006)show a model with a lid that is faster than our lid, but which iscomparable to our model at depths greater than 225 km. A study ofshear wave velocity in southern Africa by Chevrot & Zhao (2007)does not report average shear wave velocities, but by comparing ourshear wave velocity maps to their velocity perturbation maps, weobserve similar patterns in velocity variation at depths greater than

C© 2011 The Authors, GJI, 186, 808–824



Figure 14. Velocity models for the Kaapvaal Craton from several studies.

200 km. At depths of 100–150 km, Chevrot & Zhao (2007) showlower velocities in the southeastern portion of the craton comparedto other parts of the craton, which we do not observe.

The shear wave models of Li & Burke (2006) and Larson et al.(2006) were calculated from the inversion of Rayleigh wave phasevelocities, and it is instructive to compare the dispersion curves tosee why our models of mantle structure differ from theirs. Fig. 15compares our observed and predicted phase velocities for the south-ern Kaapvaal Craton to those reported by Li & Burke (2006) andLarson et al. (2006). At all periods, phase velocities from Larsonet al. (2006) are equal to or faster than our measured velocities.Phase velocities from Larson et al. (2006) were measured using thetwo-station method along ray paths that were primarily oriented in anortheast–southwest direction. This direction is consistent with thefast direction of anisotropy found in our study (Fig. 10) and previousstudies (Silver et al. 2001, 2004; Fouch et al. 2004a), which mightexplain why their model is faster than ours.

Overall, our phase velocities are similar to those measured by Li& Burke (2006), with Li & Burke’s (2006) being slightly faster atperiods less than 100 s and slightly slower at periods greater than100 s (Fig. 15). Li & Burke (2006) also used a slightly thicker (3km) crust in their model than the one we used. Modelling tests showthat the differences in the shear wave models shown in Fig. 14 resultfrom the variations in the phase velocity curves (Fig. 15) and fromthe different crustal models used, rather than from a difference ininversion methods.

Velocities in our shear wave model for the craton are also con-sistent with shear wave velocities estimated from xenoliths (Jameset al. 2004). James et al. (2004) find a maximum velocity of 4.71km s−1 in the upper-mantle lid at a depth of 50 km, and a minimumvelocity of 4.49 km s−1 at a depth of 200 km. Within our modeluncertainties, the xenolith-derived velocities are similar to the max-imum (4.67 km s−1) and minimum (4.49 km s−1) velocities in ourmodel within comparable depth intervals.

Fig. 13 shows a comparison between our average models andthose from Li & Burke (2006) for regions where our study areasoverlap. For the northern Kaapvaal Craton, Li & Burke (2006)

find lid velocities that are faster than ours, but similar minimumvelocities beneath the lid. For the Western Namaqua–Natal Belt, wefind mantle lid velocities that are comparable to those from Li &Burke (2006), within the variability of our model. At depths between125 and 275 km, we find a greater reduction in velocity relative toLi and Burke’s model, but the variability in our model indicates thatthe differences between our studies may not be resolvable. Upper-mantle lid velocities for the Kheis Province are similar between ourstudy and Li & Burke (2006), but at depths greater than 150 km,our velocities are up to 0.15 km s−1 slower than those reported byLi & Burke (2006).

In the northern section of the Kaapvaal Craton, we observe thatvelocities in the upper 150 km are 0.05–0.1 km s−1 slower than inthe southern section of the craton. Our model uncertainties may beinsufficient to confidently interpret these velocity variations. How-ever, in their P- and S-wave tomography images of the KaapvaalCraton, Fouch et al. (2004b) also find lower velocities beneath theBushveld Complex, as well as to the west of the Bushveld Com-plex, compared to velocities at similar depths beneath the southernKaapvaal Craton, which is consistent with our results. Fouch et al.(2004b) suggest that the lower velocities in the northern KaapvaalCraton may be due to refertilization and metasomatism during theemplacement of the Bushveld Complex ∼2.1 Ga.

To address the question of whether or not the upper mantle be-neath southern Africa is thermally perturbed, we now compare ourmodel for the Kaapvaal Craton to SV velocity models for otherArchean cratons. We also discuss results from previous studies ofxenoliths in the Kaapvaal Craton and other Archean cratons thathave relevance to upper-mantle thermal structure. We focus on thesouthern Kaapvaal Craton because the resolution is best in this partof our model and to also avoid complications from the possible alter-ation of the northern Kaapvaal Craton lithosphere by the BushveldComplex.

Fig. 16 shows our average model for the southern KaapvaalCraton with SV models for other Archean cratons. The modelsfor the Slave (Chen et al. 2007) and Tanzania (Weeraratne et al.(2003) Cratons were obtained from phase velocities measured us-ing the two plane wave approximation method employed in thisstudy (Forsyth & Li 2005). Pedersen et al. (2009) inverted phasevelocities for Yilgarn Craton of western Australia from Fishwicket al. (2005). Darbyshire et al. (2007) used a Monte Carlo method(Shapiro et al. 1997) to model phase velocities measured usinga two-station method, and presented shear wave velocity profilesfor many two-station paths across the Superior Province. We showthe average of their profiles from the central part of the SuperiorProvince (Fig. 16).

Considerable variation exists between the cratons. The Slave Cra-ton, Superior Province and Yilgarn Craton have average velocityreductions of 2.5, 4.1 and 1.9 per cent, respectively, from the maxi-mum lid velocity to the minimum velocity in the LVZ, whereas theaverage velocity reduction for the Tanzania Craton is 12.1 per cent.Li & Burke (2006) and Li (2011) find an average velocity reduc-tion of 5.3 and 4.7 per cent, respectively, for the Kaapvaal Craton,whereas our results indicate a slightly lower velocity reduction of3.9 per cent. These results indicate that changes in velocity in theupper mantle beneath the Kaapvaal Craton fall within the range ofvelocity variations observed beneath most other cratons and there-fore are not necessarily anomalous. Given this result, we find littleconclusive evidence for a thermal anomaly within the upper mantlebeneath the Kaapvaal Craton. The velocities beneath the TanzaniaCraton, however, are much lower than beneath the other cratons andare likely indicative of thermally perturbed upper mantle.

C© 2011 The Authors, GJI, 186, 808–824



Figure 15. Observed phase velocity curves and predicted curves for the southern Kaapvaal Craton.

The similarity between the upper-mantle structure of the Kaap-vaal Craton and other Archean cratons is also reflected inxenolith data. Rudnick & Nyblade (1999) used a large set ofpressure–temperature (P–T) estimates for mantle xenoliths to char-acterize the average geotherm for the Kalahari Craton (includingboth the Kaapvaal and Zimbabwe Cratons) and to estimate crustaland mantle heat production. By comparing the Kalahari P–T profileto those from other Archean cratons, they were able to draw com-parisons between the thermal structures of the cratons. They foundthat P–T estimates for the Superior Craton are similar to those forthe Kalahari Craton, indicating a similar geotherm for the two cra-tons. The P–T estimates for the Siberian Craton are, on average,also similar to the Kalahari Craton, but exhibit a large amount ofscatter. For the Slave Craton, however, P–T estimates indicate acooler geotherm than for other Archean cratons (Rudnick & Ny-blade 1999). These findings are consistent the velocity models inFig. 16, showing that the upper mantle beneath the Slave Cratonis somewhat faster than for the Kaapvaal Craton and the SuperiorProvince.

The thickness of the lithospheric mantle lid beneath Archean cra-tons is difficult to determine because of uncertainty in how to definethe base of the lithosphere seismically. Some studies estimate litho-spheric thickness based on absolute velocity comparisons, whereasother studies use the depth of the velocity minimum in the LVZ orelse the depth of the maximum negative velocity gradient. If wechoose the depth at which the maximum negative velocity gradientoccurs as the base of the lithosphere, we obtain a lithospheric thick-ness of ∼170 km for the Kaapvaal Craton. Regardless of how thebase of the lithosphere is chosen, Fig. 16 illustrates that the base ofthe lithosphere beneath the Kaapvaal Craton probably lies between150- and 200-km depth, and is comparable to lid thicknesses ofother Archean cratons.

6 C O N C LU S I O N S

In this study, we use data from SASE, AfricaArray and GSN stationsto measure Rayleigh wave phase velocities using the two plane waveapproximation method of Forsyth & Li (2005). The method that weuse also solves for azimuthal anisotropy and accounts for finitefrequency effects of Rayleigh wave propogation. We find phasevelocities that are generally similar to previous studies, with phasevelocities that are faster in the southern Kaapvaal Craton and slower

for the Bushveld Complex and the surrounding mobile belts. We alsodo not observe evidence of systematic velocity variations betweenthe Namaqua and Natal sectors of the Namaqua–Natal Belt.

We have inverted the phase velocity measurements for a quasi-3-D shear wave velocity model using the inversion technique ofJulia et al. (2000), and compared the shear wave model to modelsfrom other studies. The models are generally similar and the smalldifferences in the models that we find result from differences inphase velocities for periods of 40–100 s and a priori constraintsused for crustal structure.

We find a high-velocity upper-mantle lid beneath the KaapvaalCraton and a reduction in shear wave velocities of 3.9 per cent withinan LVZ below the lid. When compared to the velocity structureof other cratons, the upper-mantle velocity structure beneath theKaapvaal Craton does not appear to be anomalous. This conclusion

Figure 16. Velocity models for several cratons. Models are shown for theSlave Craton (Chen et al. 2007), the central Superior Craton (Darbyshireet al. 2007), the Tanzania Craton (Weeraratne et al. 2003), an average ofmodels of the Yilgran Craton from Pedersen et al. (2009), and the averagevelocity model for the southern Kaapvaal Craton (this study).

C© 2011 The Authors, GJI, 186, 808–824



is supported by seismic velocity estimates derived from xenolithcompositions, as well as P–T estimates from xenoliths.

Consequently, our results do not support the presence of a signif-icant thermal anomaly in the upper mantle beneath southern Africa,and therefore the high topography of southern Africa is likely sup-ported by some other mechanism. Two alternative hypotheses havebeen proposed by other studies, which would be consistent with ourfindings. Nyblade & Sleep (2003) proposed that the uplift observedin southern Africa might be due to heating from the tail of a Meso-zoic plume that was stationary beneath southern Africa for ∼26Ma or longer. They argue that the plume tail could have heated thelithosphere enough to create ∼1 km of plateau uplift, but that theperturbation to the thermal structure of the lithosphere might not besufficient to be detected seismically. Metasomatic alteration of themantle lithosphere could have also occurred.

Alternatively, some studies have suggested that uplift in southernAfrica may be due to buoyancy from the anomalous structure inthe lower mantle beneath southern Africa that forms the AfricanSuperplume. For example, Lithogow-Bertelloni & Silver (1998)suggested that the low-velocity anomaly in the lower mantle beneathsouthern Africa may represent a large-scale low-density mantleupwelling which can provide dynamic uplift sufficient to supportthe high topography observed in southern Africa. This suggestionis supported by numerical modelling of mantle flow over the past75 Ma by Conrad & Gurnis (2003), as well as by Simmons et al.(2007), who jointly inverted seismic, gravity and topography datato obtain a 3-D model of the superplume.

A C K N OW L E D G E M E N T S

We thank Aibing Li for providing phase velocity measurementsfrom the SASE stations, Dayanthie Weereratne for help with usingthe two plane wave method for obtaining phase velocities and twoanonymous reviewers for constructive comments. This work hasbeen supported by the National Science Foundation (grants OISE0530062 and EAR 0440032).

R E F E R E N C E S

Adams, A., 2010. A seismic investigation of crust and upper mantle structurebeneath the Zagros Mountains and the southern and east African plateaus,PhD thesis, Pennsylvania State University, University Park, PA.

Anhaeusser, C., 1999. Archean crustal evolution of the central KaapvaalCraton, South Africa; evidence from the Johannesburg Dome, S. Afr. J.Geol., 102, 303–322.

Begg, G., et al. 2009. The lithosphere architecture of Africa: seismic tomog-raphy, mantle petrology, and tectonic evolution, Geosphere, 5, 23–50.

Bloch, S. & A. Hales, 1968. New technique for the determination of surfacewave phase velocities, Bull. seism. Soc. Am., 58, 1021–1034.

Bloch, S., A. Hales & M. Landisman, 1969. Velocities in the crust and uppermantle of southern Africa from multi-mode surface wave dispersion, Bull.seism. Soc. Am., 59, 1599–1629.

Burke, K. & Y. Gunnell, 2008. The African erosion surface: a continental-scale synthesis of geomorphology, tectonics, and environmental changeover the past 180 million years, Memoir. Geol. Soc. Am., 201, 1–66.

Carlson, R., T. Grove, M. de Wit & J. Gurney, 1996. Anatomy of an Archeancraton: a program for interdisciplinary studies of the Kaapvaal craton,southern Africa, EOS, Trans. Am. geophys. Un., 77, 273–277.

Chen, C., S. Rondenay, D. Weeraratne & D. Snyder, 2007. New constraintson the upper mantle structure of the Slave craton from Rayleigh waveinversion, Geophys. Res. Lett., 34, 1–5.

Chevrot, S. & L. Zhao, 2007. Multiscale finite-frequency Rayleigh wavetomography of the Kaapvaal craton, Geophys. J. Int., 169, 201–215.

Conrad, C. & M. Gurnis, 2003. Seismic tomography, surface uplift, and thebreakup of Gondwanaland: integrating mantle convection backwards intime, Geochem. Geophys. Geosyst., 4, 1–16.

Cornell, D., R. Thomas, H. Moen, D. Reid, J. Moore & R. Gibson, 2006. TheNamaqua–Natal Province, in The Geology of South Africa, pp. 325–379,eds Johnson, M., Anhaeusser, C. & Thomas, R., Geological Society ofSouth Africa, Johannesburg.

Darbyshire, F., D. Eaton, A. Frederiksen & L. Ertolahti, 2007. Newinsights into the lithosphere beneath the Superior Province fromRayleigh wave dispersion and receiver function analysis, Geophys.J. Int., 169, 1043–1068, doi:10.1111/j.1365-246X.2006.03259.x.

Deen, T., W. Griffin, G. Begg, S. O’Reilly, L. Natapov & J. Hronsky, 2006.Thermal and compositional structure of the subcontinental lithosphericmantle: Derivation from shear wave seismic tomography, Geochem. Geo-phys. Geosyst, 7, 1–20.

Eaton, D., F. Darbyshire, R. Evans, H. Grutter, A. Jones & X. Yuan, 2009.The elusive lithosphere-asthenosphere boundary (LAB) beneath cratons,Lithos, 109, 1–22.

Eglington, B. & R. Armstrong, 2004. The Kaapvaal Craton and adjacentorogens, southern Africa: a geochronological database and overview ofthe geological development of the craton, S. Afr. J. Geol., 107, 13–32.

Fishwick, S., Kennett, B. & Reading, A., 2005. Contrasts in lithosphericstructure within the Australian Craton – insights from surface wave to-mography, Earth planet. Sci. Lett., 231, 163–176.

Forsyth, D. & A. Li, 2005. Array analysis of two-dimensional variations insurface wave phase velocity and azimuthal anisotropy in the presence ofmultipathing interference, in Seismic Earth: Array Analysis of BroadbandSeismograms, pp. 81–97, eds Levander, A. & Nolet, G., GeophysicalMonograph Series Volume 157, AGU, Washington, DC.

Fouch, M., P. Silver, D. Bell & J. Lee, 2004a. Small-scale variationsin seismic anisotropy near Kimberley, S. Afr. Geophys. J. Int., 157,764–774.

Fouch, M., D. James, J. VanDecar, S. Van Der Lee & Kaapvaal SeismicGroup, 2004b. Mantle seismic structure beneath the Kaapvaal and Zim-babwe Cratons, S. Afr. J. Geol., 107, 33–44.

Freybourger, M., J. Gaherty, T. Jordan & the Kaapvaal Seismic Group, 2001.Structure of the Kaapvaal craton from surface waves, Geophys. Res. Lett.,28, 2489–2492.

Gane, P., H. Logie & J. Stephen, 1949. Triggered telerecording seismicequipment, Bull. seism. Soc. Am., 39, 117–143.

Grand, S., 2002. Mantle shear-wave tomography and the fate of subductedslabs, Phil. Trans. R. Soc. Lond., A360, 2475–2491.

Gurnis, M., J. Mitrovica, J. Ritsema & H. van Heijst, 2000. Constrain-ing mantle density structure using geological evidence of surface upliftrates: the case of the African Superplume, Geochem. Geophy. Geosyst.,1, 127–142.

Hansen, S., A. Nyblade, J. Julia, P. Dirks & R. Durrheim, 2009. Upper-mantle low-velocity zone structure beneath the Kaapvaal craton fromS-wav receiver functions, Geophys. J. Int., 178, 1021–1027.

James, D., M. Fouch, J. VanDecar, S Van Der Lee & the Kaapvaal SeismicGroup, 2001. Tectospheric structure beneath southern Africa, Geophys.Res. Lett., 28, 2485–2488.

James, D., F. Niu & J. Rokosky, 2003. Crustal structure of the Kaap-vaal craton and its significance for early crustal evolution, Lithos, 71,413–429.

James, D., F. Boyd, D. Schutt, D. Bell & R. Carlson, 2004. Xenolith con-straints on seismic velocities in the upper mantle beneath southern Africa,Geochem. Geophys. Geosyst., 5, 1–32.

Julia, J., C. Ammon, R. Herrmann & A. Correig, 2000. Joint inversion ofreceiver function and surface wave dispersion observations, Geophys. J.Int., 143, 99–112.

Kennett, B. E., Engdahl & R. Bulard, 1995. Constraints on seismic velocitiesin the Earth from travel times, Geophys. J. Int., 122, 108–124.

Kgaswane, E., A. Nyblade, J. Julia, P. Dirks, R. Durrheim & M.Pasyanos, 2009. Shear wave velocity structure of the lower crustin southern Africa: evidence for compositional heterogeneity withinArchaean and Proterozoic Terrains, J. geophys. Res., 114, B12304,doi:10.1029/2008JB006217.

C© 2011 The Authors, GJI, 186, 808–824



Kramers, J., S. McCourt & D. van Reenen, 2006. The Limpopo Belt, in TheGeology of South Africa, pp. 209–236, eds Johnson, M., Anhaeusser, C.& Thomas, R., Geological Society of South Africa, Johannesburg.

Larson, A., J. Snoke & D. James, 2006. S-wave velocity structure, mantlexenoliths and the upper mantle beneath the Kaapvaal craton, Geophys. J.Int., 167, 171–186.

Li, A., 2011. Shear wave model of southern Africa from regional Rayleighwave tomography with 2-D sensitivity kernels, Geophys. J. Int., 185,832–844.

Li, A. & K. Burke, 2006. Upper mantle structure of southern Africafrom Rayleigh wave tomography, J. geophys. Res., 111, B10303,doi:10.1029/2006JB004321.

Lithogow-Bertelloni, C. & P. Silver, 1998. Dynamic topography, plate driv-ing forces and the African superswell, Nature, 395, 269–272.

McCourt, S. & R. Armstrong, 1998. SIMS U-Pb zircon geochronologyof granites from the Central Zone, Limpopo Belt, southern Africa: im-plications for the age of the Limpopo Orogeny, S. Afr. J. Geol., 101,329–338.

Moen, H., 1999. The Kheis tectonic subprovince, southern Africa: a lithos-tratigraphic perspective, S. Afr. J. Geol., 102, 27–42.

Nair, S., S. Gao, K. Lin & P. Silver, 2006. Southern African crustal evolutionand composition: constraints from receiver function studies, J. geophys.Res., 111, B02304, doi:10.1029/2005JB003802.

Nguuri, T., J. Gore, D. James, S. Webb, C. Wright, T. Zengeni, O. Gwavava,J. Snoke & Kaapvaal Seismic Group, 2001. Crustal structure beneathsouthern Africa and its implications for the formation and evolution ofthe Kaapvaal and Zimbabwe cratons, Geophys. Res. Lett., 28, 2501–2504.

Niu, F. & D. James, 2002. Fine structure of the lowermost crust beneath theKaapvaal craton and its implications for crustal formation and evolution,Earth planet. Sci. Lett., 200, 121–130.

Nyblade, A., 2011. The upper mantle low velocity anomaly beneath Ethiopia,Kenya and Tanzania: constraints on the origin of the African Superswellin eastern Africa and plate versus plume models of mantle dynamics,in Volcanism and the Evolution of the African Lithosphere, GeologicalSociety of America Special Publication 478, pp. 1–14, eds Beccaluva, L.,Bianchini, G. & Wilson, M., GSA, Boulder, CO.

Nyblade, A. & S. Robinson, 1994. The African superswell, Geophys. Res.Lett., 21, 765–768.

Nyblade, A. & N. Sleep, 2003. Long lasting epeirogenic uplift from man-tle plumes and the origin of the Southern African Plateau, Geochem.Geophys. Geosyst., 4(12), doi:10.1029/2003GC000573.

Park, Y., A. Nyblade, A. Rodgers & A. Al-Amri, 2008. S wave velocity struc-ture of the Arabian Shield upper mantle from Rayleigh wave tomography,Geochem. Geophys. Geosyst., 9, Q07020, doi:10.1029/2007GC001895.

Pedersen, H.A., S. Fishwick & D.B. Snyder, 2009. A comparison of cratonicroots through consistent analysis of seismic surface waves, Lithos, 109,81–95.

Press, W., S. Teukolsky, W. Vetterling & B. Flannery, 1992. NumericalRecipes in FORTRAN: The Art of Scientific Computing, 2nd edn, Cam-bridge University Press, Cambridge.

Priestley, K., 1999. Velocity structure of the continental upper mantle: evi-dence from southern Africa, Lithos, 48, 45–56.

Priestley, K., D. McKenzie & E. Debayle, 2006. The state of the uppermantle beneath southern Africa, Tectonophysics, 416, 101–112.

Priestley, K., D. McKenzie, E. Debayle & S. Pilidou, 2008. The African up-per mantle and its relationship to tectonics and surface geology, Geophys.J. Int., 175, 1108–1126.

Qui, X., K. Priestley & D. McKenzie, 1996. Average lithospheric structureof southern Africa, Geophys. J. Int., 127, 563–587.

Ritsema, J., S. Ni, D. Helmberger & H. Crotwell, 1998. Evidence for strongshear velocity reductions and velocity gradients in the lower mantle be-neath Africa, Geophys. Res. Lett., 25, 4245–4248.

Rudnick, R. & A. Nyblade, 1999. The thickness and heat production ofArchean lithosphere: constraints from xenolith thermobarometry and sur-face heat flow, in Mantle Petrol.: Field Observations and High PressureExperimentation, Special Publication of the Geochemical Society, Vol. 6,pp. 3–12, eds Fei, Y.-W., Bertka, C.M. & Mysen, B.O., GSA, Boulder,CO.

Saltzer, R., 2002. Upper mantle structure of the Kaapvaal craton from surfacewave analysis – a second look, Geophys. Res. Lett., 29, 1–4.

Schmitz, M., S. Bowring, M. de Wit & V. Gartz, 2004. Subduction andterrane collision stabilize the western Kaapvaal craton tectosphere 2.9billion years ago, Earth planet. Sci. Lett., 222, 363–376.

Shapiro, N., M. Campillo, A. Paul, S. Singh, D. Jongmans & F. Sanchez-Sesma, 1997. Surface-wave propagation across the Mexican Volcanic Beltand the origin of the long-period seismic-wave amplification in the Valleyof Mexico, Geophys. J. Int., 128, 151–166.

Silver, P., S. Gao, K. Liu & the Kaapvaal Sesmic Group, 2001. Mantledeformation beneath southern Africa, Geophys. Res. Lett., 28, 2493–2496.

Silver, P., M. Fouch, S. Gao, M. Schmitz & Kaapvaal Seismic Group, 2004.Seismic anisotropy, mantle fabric, and the magmatic evolution of Pre-cambrian southern Africa, S. Afr. J. Geol., 107, 45–58.

Simon, R., C. Wright, E. Kgaswane & M. Kwadiba, 2002. The P wavespeedstructure below and around the Kaapvaal craton to depths of 800 km,from traveltimes and waveforms of local and regional earthquakes andmining-induced tremors, Geophys. J. Int., 151, 132–145.

Simmons, N.A., A. Forte & S. Grand, 2007. Thermochemical structure anddynamics of the African superplume, Geophys. Res. Lett., 34, L02301.

Tarantola, A. & B. Valette, 1982. Generalized non-linear problems solvedusing the least-squares criterion, Rev. Geophys. Space Phys., 20, 219–223.

Thomas, R., D. Cornell, J. Moore & J. Jacobs, 1994. Crustal evolution of theNamaqua–Natal metamorphic province, Southern Africa, S. Afr. J. Geol.,97, 8–14.

Tinker, J., M. de Wit & J. Grotzinger, 2002. Seismic stratigraphic constraintson the Neoarchean-Paleoproterozoic evolution of the western margin ofthe Kaapvaal craton, South Africa, S. Afr. J. Geol., 105, 107–134.

Vinnik, L., Green, R., & Nicolaysen, L., 1995. Recent deformations of thedeep continental root beneath southern Africa, Nature, 375, 50–52.

Wang, Y., L. Wen & D. Weidner, 2008. Upper mantle SH- and P-velocitystructures and compositional models beneath southern Africa, Earthplanet. Sci. Lett., 267, 596–608.

Weeraratne, D., D. Forsyth, K. Fischer & A. Nyblade, 2003. Evidence foran upper mantle plume beneath the Tanzanian craton from Rayleigh wavetomography, J. geophys. Res., 108, 1–17.

Willmore, P., A. Hales & P. Gane, 1952. A seismic investigation of crustalstructure in the western Transvaal, Bull. seism. Soc. Am., 42, 53–80.

de Wit, M., et al. 1992. Formation of an Archaean continent, Nature, 357,553–562.

Yang, Y. & D. Forsyth, 2006a. Regional tomographic inversion of the ampli-tude and phase of Rayleigh waves with 2-D sensitivity kernels, Geophys.J. Int., 166, 1148–1160.

Yang, Y. & D. Forsyth, 2006b. Rayleigh wave phase velocities, small-scaleconvection, and azimuthal anisotropy beneath southern California, J. geo-phys. Res., 111, 1–20.

Zhao, M., C. Langston, A. Nyblade & T. Owens, 1999. Upper mantle velocitystructure beneath southern Africa from modeling regional seismic data,J. geophys. Res., 104, 4783–4794.

Zhou, Y., F. Dahlen & G. Nolet, 2004. Three dimensional sensitivity kernelsfor surface wave observables, Geophys. J. Int., 158, 142–168.

C© 2011 The Authors, GJI, 186, 808–824


shear wave velocity structure of the southern african...

Documents