Geophysical Journal InternationalGeophys. J. Int. (2015) 202, 533–547 doi: 10.1093/gji/ggv136

GJI Seismology

Crustal structure of Precambrian terranes in the southern Africansubcontinent with implications for secular variation in crustal genesis

Marsella Kachingwe,1 Andrew Nyblade1,2 and Jordi Julia3

1Department of Geosciences, Penn State University, University Park, PA 16802, USA. E-mail: nyblade@psu.edu2School of Geosciences, The University of Witwatersrand, Johannesburg, South Africa3Departmento de Geofisica & Programa de Pos-Granducao em Geodinamica e Geofisica, Universidade Federal do Rio Grande do Norte, Natal, Brazil

Accepted 2015 March 19. Received 2015 March 18; in original form 2014 September 30

S U M M A R YNew estimates of crustal thickness, Poisson’s ratio and crustal shear wave velocity havebeen obtained for 39 stations in Angola, Botswana, the Democratic Republic of Congo,Malawi, Mozambique, Namibia, Rwanda, Tanzania and Zambia by modelling P-wave receiverfunctions using the H–κ stacking method and jointly inverting the receiver functions withRayleigh-wave phase and group velocities. These estimates, combined with similar resultsfrom previous studies, have been examined for secular trends in Precambrian crustal structurewithin the southern African subcontinent. In both Archean and Proterozoic terranes we findsimilar Moho depths [38–39 ± 3 km SD (standard deviation)], crustal Poisson’s ratio (0.26± 0.01 SD), mean crustal shear wave velocity (3.7 ± 0.1 km s−1 SD), and amounts ofheterogeneity in the thickness of the mafic lower crust, as defined by shear wave velocities≥4.0 km s−1. In addition, the amount of variability in these crustal parameters is similarwithin each individual age grouping as between age groupings. Thus, the results provide littleevidence for secular variation in Precambrian crustal structure, including between Meso- andNeoarchean crust. This finding suggests that (1) continental crustal has been generated bysimilar processes since the Mesoarchean or (2) plate tectonic processes have reworked andmodified the crust through time, erasing variations in structure resulting from crustal genesis.

Key words: Cratons; Crustal structure; Africa.

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

Whether present-day processes of crustal formation were also themost dominant processes in the first 1–2 billion years of Earth’shistory remains an unresolved question. Knowledge about the na-ture of Precambrian crust obtained from seismic data provides keyconstraints on geodynamic models of crustal genesis and evolution,and is therefore important for addressing questions about secu-lar variation in crustal formation. Some studies going back to theearly 1990s using compilations of seismic refraction data (e.g. Dur-rheim & Mooney 1991, 1994), have suggested that there may besignificant differences in crustal structure between Archean andProterozoic terranes, indicating secular changes in crustal genesis,while others (e.g. Rudnick & Fountain 1995; Rudnick & Gao 2003)have argued for little change in crustal structure throughout the Pre-cambrian. This debate has continued with more recent studies. Forexample, Thompson et al. (2010) using seismic images of crustalstructure in the Canadian Shield, and Abbott et al. (2013) usingpreviously reported seismic models of crustal structure in southernAfrica, Australia and Canada, have argued for secular evolutionin Precambrian crustal genesis. On the other hand, Tugume et al.(2013), using seismic images of crustal structure in eastern Africa

and Stankiewicz & De Wit (2013), using swaths of seismic dataacross southern Africa, find no discernible differences in Archeanand Proterozoic crustal structure.

In this study, we report new estimates of crustal structure (crustalthickness, shear wave velocity and Poisson’s ratio) across south-ern and eastern Africa, and use them in combination with otherpublished results for the southern Africa subcontinent to investigatewhether the conclusions reached by Tugume et al. (2013) for Africancrustal structure hold for a much larger dataset. Not only does ourstudy expand on the number of Precambrian terranes examined inAfrica, but it also allows for the comparison of continental crustpre- and post-3.0 Ga (i.e. Mesoarchean versus Neoarchean crust).

The new estimates of crustal structure come from Angola,Botswana, the Democratic Republic of Congo, Malawi, Mozam-bique, Namibia, Rwanda, Tanzania and Zambia. Estimates of Mohodepth, crustal shear wave velocities and Poisson’s ratio for 39new broad-band stations in those countries have been determinedby modelling P-wave receiver functions using the H–κ stackingmethod (Zhu & Kanamori 2000) and a joint inversion of P-wave re-ceiver functions and Rayleigh-wave dispersion measurements (Juliaet al. 2000, 2003). Results from the previous studies were obtainedby modelling P-wave receiver functions using similar methods,

C© The Authors 2015. Published by Oxford University Press on behalf of The Royal Astronomical Society. 533

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

534 M. Kachingwe, A. Nyblade and J. Julia

including the H–κ stacking method (Midzi & Ottemoller 2001;Nguuri et al. 2001; Dugda et al. 2005; Nair et al. 2006; Gallacher &Bastow 2012; Tugume et al. 2012; Youssof et al. 2013), the slant-stacking method (Last et al. 1997) and joint inversion methodsinvolving P-wave receiver functions and Rayleigh-wave dispersioncurves (Julia et al. 2005; Kgaswane et al. 2009; Tokam et al. 2010;Tugume et al. 2013). The data ensemble used in this study includesestimates for many Archean and Proterozoic terranes, allowing fora more comprehensive examination of Precambrian crustal struc-ture in the southern African subcontinent than was done by Tugumeet al. (2012, 2013).

As a point of clarification, throughout this paper Moho depth andcrustal thickness are used interchangeably in reference to the totalcrustal thickness.

2 B A C KG RO U N D G E O L O G Y

The southern African subcontinent is an amalgamation of manyPrecambrian terranes (Fig. 1), including several Archean cratonsand Proterozoic orogenic belts that have not been affected by anysignificant Phanerozoic tectonic activity. In this section, the geologyof each terrane for which there are seismic estimates of crustal struc-ture is briefly reviewed. A short summary of the crustal structurefor each terrane from previous studies is provided in the SupportingInformation.

2.1 Eastern Africa

The Archean Tanzania Craton forms the nucleus of the Precambriantectonic framework in eastern Africa (Cahen et al. 1984). The cratonis made up of two blocks separated by the Dodoma Schist Belt. Thenorthern Nyanzian Block is comprised of greenstone belts and gran-ites ranging in age from 2.80 to 2.66 Ga, overlain by a molasse that isintruded by 2.60 Ga granitoids. The southern Dodoman Block con-sists of 2.93–2.85 Ga granodiorites, granitic gneisses, migmatitesand other high-grade metamorphic rocks (Cahen et al. 1984; Schul-ter 1997; Manya & Maboko 2003; Begg et al. 2009 and referencestherein).

The Tanzania Craton is surrounded by many Proterozoic mobilebelts. The Palaeoproterozoic Usagaran Belt along the southeast-ern margin of the Tanzania Craton contains mostly supracrustalrocks metamorphosed to granulite facies at 1.92–1.89 Ga and1.87–1.83 Ga granitoids that were partially derived from reworkingand recycling of the Tanzania Craton (Cahen et al. 1984; Schul-ter 1997; De Waele et al. 2008; Begg et al. 2009 and referencestherein). To the southwest of the Tanzania Craton is the Palaeo-proterozoic Ubendian Belt. It consists of granulite and amphibolitefacies gneisses and metasedimentary rocks that formed during twoorogenic events (∼2.03–1.97 and ∼1.93–1.86 Ga, respectively, Ca-hen et al. 1984; Lenoir et al. 1994; Schulter 1997; Begg et al. 2009and references therein). Along the western side of the Tanzania Cra-ton is the Mesoproterozoic Kibaran Belt. The north–south orientedterrane is comprised of amphibolite grade rocks formed during theKibaran Orogeny (1.4 Ga), with intrusions of granites, ultramaficbodies and granitoids at 1.39, 1.3 and 1.0 Ga, respectively (Klerkx1987; Begg et al. 2009 and references therein). The Mesoprotero-zoic Rwenzori Belt is a deformed volcano-sedimentary sequence(1.86–1.78 Ga) located to the northeast of the Kibaran Belt (Begget al. 2009 and references therein). East of the Tanzania Craton isthe Neoproterozoic Mozambique Belt. The Mozambique Belt is thelongest terrane in Africa, extending along the east African coast

from Mozambique to southern Egypt (Schulter 1997). The north-ern part of the belt consists of reworked Archean continental crust,while the southern part (most of Malawi and northern Mozambique)consists of Palaeoproterozoic to Neoproterozoic basement gneisses(Begg et al. 2009 and references therein). The Mozambique Beltis believed to be a continental collision zone formed by multiplecollision events between 1200 and 450 Ma (Cahen et al. 1984;Shackleton 1986).

2.2 Central Africa

Central African Precambrian geology is dominated by the CongoCraton. The Congo Craton is composed of Mesoarchean blocksand Palaeo- to Mesoproterozoic mobile belts amalgamated duringthe assembly of Gondwana (Goodwin 1996; De Waele et al. 2008),with regions of extensive Phanerozoic cover (De Waele et al. 2005).Archean basement rock is exposed in four shields around the edgesof the craton (Begg et al. 2009 and references therein) but onlythree are relevant to this study. The Kasai Block on the southeasternedge of the Congo Craton is a heterogeneous 3.01 Ga granulitecomplex, metamorphosed between 2.9 and 2.6 Ga, and overlain bysupracrustal rocks, which were later cut by 2.1 Ga granite plutons(Begg et al. 2009 and references therein). The Angolan Block onthe southwest side consists of gneisses, metasediments, a 2.82 Gagabbro-charnockite complex metamorphosed at 2.8–2.7 Ga and2.6 Ga granitic intrusions (Begg et al. 2009 and references therein).The Ntem Complex, the northern part of the Gabon-CameroonShield in the northwest corner of the craton, consists of 3.22–2.99Ga migmatites, gneisses and remnants of greenstone belts cut by2.94–2.54 Ga granites (Begg et al. 2009 and references therein).The Ntem Complex was later reworked in the Palaeoproterozoic,during the Eburnean Orogeny (Tchameni et al. 2001).

Bordering the Congo Craton are several Proterozoic terranes.On the western side of the Congo Craton is the PalaeoproterozoicWest Central African Belt. The terrane evolved between 2.5 and 2.0Ga during the Eburnean Orogeny and contains reworked Archeanrocks, metamorphosed and intruded by 2.1–1.92 Ga granitoids andoverlain by rift-related sequences that may be connected with thebreak-up of Rodinia (Begg et al. 2009 and references therein).North of the Congo Craton lies the Neoproterozoic OubanguidesBelt which resulted from the collision between the Congo Craton,the Sao Francisco Craton and the West African Craton during theformation of Gondwana (Castaing et al. 1994; Toteu et al. 2004).The mobile belt consists of 1.7–0.90 Ga sedimentary rocks over a2.0 Ga basement (Begg et al. 2009 and references therein).

The Neoarchean Bangweleu Block lies to the southeast of theCongo Craton and is composed of crystalline basement of schistbelts intruded by 2.73 Ga granitic and metavolcanic rocks (DeWaele et al. 2006; Begg et al. 2009 and references therein). Theterrane is considered to be a Neoarchean microcontinent that wasstrongly affected by tectonic events in the surrounding mobile belts(Begg et al. 2009 and references therein). To the southeast of theBangweleu Block is the Mesoproterozoic Irumide Belt. The ter-rane consists of reworked Archean and Palaeoproterozoic basementrocks, overlain by 1.88 Ga rhyolites and sediments (Begg et al. 2009and references therein). During the Irumide Orogeny at 1.02 Ga, theterrane was strongly deformed by isoclinal folding and thrusting,which resulted in crustal thickening without the addition of juvenilecrust (De Waele et al. 2005; De Waele et al. 2006). Further south ofthe Irumide Belt is the Southern Irumide Belt. Composed of Meso-proterozoic gneisses, this mobile belt is interpreted as a stack of

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 535

Figure 1. A geologic map of the southern African subcontinent showing the tectonic framework of the study region, modified from Begg et al. (2009). Terranesrelevant to this study are numbered and identified in the map key.

arc-related terranes imbricated during the Irumide Orogeny (Begget al. 2009 and references therein). The Zambezi Belt, south of theIrumide Belt, is a Neoproterozoic terrane that contains reworkedArchean and Proterozoic gneisses, overlain by strongly deformed

metasediments and granites formed during the amalgamation ofGondwana (Begg et al. 2009 and references therein). The mobilebelt was strongly deformed between 870 and 820 Ma during thePan-African Orogeny (Begg et al. 2009 and references therein).

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

536 M. Kachingwe, A. Nyblade and J. Julia

Located on the southwest side of the Bangweleu Block is theNeoproterozoic Lufilian Arc. The terrane is composed of reworkedArchean to Proterozoic basement overlain by Neoproterozoic vol-canics, all thrust over the edge of the Congo Craton at 560–550 Maduring the assembly of Gondwana (Begg et al. 2009 and referencestherein).

2.3 Southern Africa

Two cratons are found in southern Africa. The Zimbabwe Cratonconsists of several terranes that formed between 3.6 and 2.4 Ga(Dirks & Jelsma 2002). The central Tokwe Gneiss terrane (EastTokwe Block) formed at 3.6–3.3 Ga and contains mafic fragments.Clastic sediments accreted against the western part of the TokweGneiss terrane (West Tokwe Block) between 3.2 and 2.8 Ga andgreenstone belts formed between 2.7 and 2.6 Ga, marking the sta-bilization of the craton. The last major tectono-thermal event af-fecting the craton was the Great Dyke around 2.58 Ga (Jelsma &Dirks 2002). The Kaapvaal Craton is a conglomerate of several ter-ranes and predominantly consists of granitoids with gneisses andnarrow greenstone belts that formed between 3.7 and 2.7 Ga. (DeWit et al. 1992; Eglington & Armstrong 2004; Begg et al. 2009and references therein). The Swaziland terrane (>3.2 Ga) and theWitwatersrand terrane sutured together around 3.2 Ga, followed bythe joining of the Pietersburg and Kimberley terranes between 3.0and 2.8 Ga. The craton was later affected by several tectono-thermalevents, including the Dominion (3.1 Ga), Witwatersrand (3.0–2.8Ga), Ventersdorp (2.7 Ga), Transvaal (2.6–2.2 Ga) and Waterberg(2.0–1.8 Ga), and the emplacement of the Bushveld Complex (2.05Ga, Eglington & Armstrong 2004; Johnson et al. 2006).

The Archean Limpopo Belt is an east–west trending zone ofhigh-grade metamorphic rocks that separates the Zimbabwe Cratonfrom the Kaapvaal Craton and probably formed during a 2.7–2.6Ga collision between the two cratons (McCourt & Armstrong 1998;Kramers et al. 2006; Begg et al. 2009 and references therein).The terrane’s 3.3–3.1 Ga gneisses were affected by granulite faciesmetamorphism and granitoid magmatism at 2.7–2.57 Ga (Begget al. 2009 and references therein). Peak metamorphism was atca. 2.7 Ga with some metamorphism between 2.06 and 2.0 Galinked with the Bushveld Complex (Begg et al. 2009 and referencestherein).

Several Proterozoic mobile belts encircle the Zimbabwe andKaapvaal cratons. The Palaeoproterozoic Kheis-Okwa Belt runsalong the western edge of the Kaapvaal Craton (Cornell et al. 2006;Begg et al. 2009). The Kheis terrane is a thin-skinned, eastwardthrusting belt composed of 1.98 Ga basalts and clastic sediments(Cornell et al. 2006; Begg et al. 2009). They were folded and meta-morphosed at ca. 1.9 Ga and intruded by granitic and mafic rocksbetween 1.27 and 1.12. The Okwa terrane has ca. 2.1 metamor-phic basement rocks that may be underlain by Archean rocks (Cor-nell et al. 2006; Begg et al. 2009). Striking northeast–southwestto the east of the Kaapvaal Craton is the Palaeoproterozoic Re-hoboth Province. The province consists of 1.79–1.73 Ga gneissesand migmatites (Begg et al. 2009 and references therein). To thesouth of the Kaapvaal Craton is the Mesoproterozoic Namaqua-Natal Belt. The fold belt is comprised of igneous and supracrustalrocks that accreted against the craton during the Namaqua Orogeny(1.2–1.0 Ga, Cornell et al. 2006). The Damara Belt runs northeast-southwest between the Congo Craton and the Rehoboth Province.This Neoproterozoic terrane is highly complex. The northern partconsists of Neoarchean basement inliers overlain by 750 Ma rift se-

quences, while the southern part represents a passive margin foldedand metamorphosed from 690 to 485 Ma during the assembly ofGondwana (Begg et al. 2009 and references therein).

3 DATA A N D M E T H O D S

3.1 Data

Data recorded on a number of networks were used for the receiverfunction analysis (Fig. 2). Data were used from 15 seismic sta-tions in the AfricaArray East Africa Seismic Experiment phase 3network in Zambia (ZP network code), which operated between Au-gust 2010 and July 2011, four stations in the AfricaArray Mozam-bique Network (XV network code), which operated between August2011 and August 2013, one station each in Angola, Namibia andBotswana supported by an industry consortium between 2007 and2010 (Congo Craton network), and three stations in Angola that be-long to the National Meteorological Institute of Angola (INAMETnetwork). In addition, data from 11 permanent AfricaArray stations(AF network code) and two IRIS/GSN stations were used. Eachstation included a broad-band seismometer, a 24-bit data loggerand a Global Positioning System clock, and the data were recordedcontinuously at either 20 or 40 samples per second.

For the joint inversion of the receiver functions and surface wavedispersion measurements, fundamental-mode Rayleigh-wave phasevelocities at periods ranging from 20s to 182s were taken fromO’Donnell et al. (2013) and fundamental-mode Rayleigh-wavegroup velocities at periods ranging from 10s to 105s were takenfrom Raveloson et al. (2015).

3.2 P-wave receiver functions

Receiver functions are a time series that show the response of Earth’sstructure beneath the recording station (Langston 1979) and arecommonly used for investigating crustal structure. The main phasesin the waveform are the direct P wave, the P-to-S conversion atthe Moho (Ps) and its reverberations between the Moho and thefree surface (PpPs and PsPs + PpSs, Fig. 3). The amplitudes andarrival times of the phases place valuable constraints on crustalstructure beneath the receiver (Langston 1979; Ligorria & Ammon1999) and can be used to estimate the thickness of the crust and thebulk crustal P-wave velocity/shear wave velocity (Vp/Vs) ratio (e.g.Zandt & Ammon 1995; Zhu & Kanamori 2000).

P-wave receiver functions were computed for each station usingteleseismic events with magnitudes equal to 5.5 and greater and atepicentral distances between 30◦ and 90◦ (Fig. 4, see SupportingInformation for list of events). For calculating the P-wave receiverfunctions, the originally recorded seismograms were de-trended, ta-pered, high pass filtered above 0.05 Hz to remove low frequencynoise from the instrument response, and low pass filtered below8 Hz. Afterwards, the waveforms were decimated to 10 samplesper second to avoid aliasing, and cut to 10s before and 110s af-ter the first P-wave arrival. Next, the horizontal components wererotated to the great circle path to obtain radial and transverse com-ponents. The vertical component was then deconvolved from theradial and tangential components using the iterative time-domaindeconvolution method (Ligorria & Ammon 1999). A maximum of500 iterations were used in the deconvolution method (Ligorria &Ammon 1999).

For each teleseismic event, radial and tangential receiver func-tions were computed for Gaussian width factors of 1.0 (f ≤ 0.5 Hz)

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 537

Figure 2. Topographic and geologic map of the southern African subcontinent showing the locations of the temporary and permanent broad-band seismicstations used in this study. Precambrian terrane boundaries are the same as in Fig. 1. Descriptions of the networks not provided in this paper can be found inTugume et al. (2013), Tokam et al. (2010), Kgaswane et al. (2009), Nair et al. (2006), Dugda et al. (2005) and Last et al. (1997). Seismic stations are labelledfor which new estimates of crustal structure are provided in this study.

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

538 M. Kachingwe, A. Nyblade and J. Julia

Figure 3. A diagram illustrating the P-wave receiver function that would begenerated in response to a simple layer over a half-space model. Redrawnfrom Ammon (1990). (a) Ray paths. (b) The receiver function waveform.

and 2.5 (f ≤ 1.25), typical for receiver function analysis. Lowerfrequencies result in longer wavelength receiver functions, whichare better for observing long period phases from the lower crust andmantle, while higher frequency receiver functions can reveal de-tailed phases from shallow crustal structure (Owens & Zandt 1985;Ligorria & Ammon 1999; Julia 2007).

Next, the receiver functions were assessed for quality using aleast squares minimization of the difference between the originalradial component and the predicted radial component. The predictedradial component was generated by the convolution of the originalvertical component and the radial receiver function. A similarityof 85 per cent or greater was used to select receiver functionsfor further analysis. In addition, events that resulted in transversereceiver functions with large amplitudes were not considered forfurther processing, even if they passed the 85 per cent criterion.Fig. 5 illustrates the quality of receiver functions at one station.

3.3 H–κ stacking method

Crustal thickness estimated solely from the delay time betweenthe Moho Ps phase and direct P-wave trades off strongly with thecrustal Vp/Vs ratio. Therefore, the H–κ stacking method of Zhu& Kanamori (2000) was applied to the receiver functions to obtainestimates for crustal thickness (H) and bulk Vp/Vs ratio (κ). The H–κ

stacking method reduces the ambiguity in H and κ by incorporatingthe later multiple converted phases from the Moho (PpPs and PsPs

Figure 4. Map showing the distribution of the earthquakes used in this studycentred on the middle of the study region (star). The circles are the locationsof events used in the computation of receiver functions. Black circles areevents recorded by permanent stations and open circles are events recordedby temporary stations. The numbers on the circles give the epicentral dis-tance in degrees.

Figure 5. A plot of the radial (left-hand panel) and tangential (right-handpanel) receiver functions versus ray parameter for station TEZI estimatedusing a Gaussian filter of 1.0.

+ PpSs). In this method, the receiver functions are transformed tothe H–κ parameter space by an objective function:

s (H, κ) =N∑

j=1

w1r j (t1) + w2r j (t2) − w3r j (t3), (1)

where ti are the traveltimes of the three main P-to-S converted phasesfrom the Moho (Ps, PpPs and PsPs + PpSs), wi are weights assignedto each phase (sum of wi = 1), rj is the receiver function amplitudefor the jth receiver function and N is the number of receiver func-tions. S(H,κ) reaches its maximum when optimal values for H andκ are determined, satisfying a simple layer over a half space crustalmodel. The values of H and κ are taken as estimates for the Mohodepth and Vp/Vs ratio near the receiver (Zhu & Kanamori 2000).

In order to apply the H–κ stacking method to the receiver func-tions, weights for the converted phases (eq. 1) and an average crustal

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 539

Vp must be selected. A Vp of 6.5 km s−1 was chosen because it isconsistent with average crustal P-wave velocities determined fromprevious studies of Precambrian crust in the area (Fuchs et al. 1997;Julia et al. 2005; Tugume et al. 2012 and references therein). Aweighting system of w1 = 0.4, w2 = 0.3 and w3 = 0.3 was em-ployed to place close to equal weights on all the Moho phases.For a few stations (CHAM, GETA, KTWE, LSZ, MTVE, MWEN,MZM and PORTQ) the third phase (PsPs + PpSs) was not clearlyobserved and so the weights were readjusted to w1 = 0.5, w2 = 0.5and w3 = 0.0.

Results from the H–κ stacking method for crustal thicknessand Poisson’s ratio obtained from the Vp/Vs ratio are tabulated inTable 1, and an example for one station is illustrated in Fig. 6(see Supporting Information for results from all stations). Receiverfunctions with a Gaussian filter of 1.0 were used for all stationsexcept MTVE, MWEN, MKUS and MUFZ, where a Gaussian of2.5 was used. For these stations the Moho converted phases weremore clearly observed at higher frequencies than at lower frequen-cies. The H–κ stacking method was performed on both 1.0 and 2.5Gaussian filters for all stations, and similar estimates were obtained.

Formal uncertainties for H and κ were estimated using a boot-strapping method to repeat the summation procedure 200 timeswith random, resampled data from the original dataset (Efron &Tibshirani 1991). Additional uncertainties in H and κ arise fromthe chosen mean crustal Vp. Therefore, the H–κ stacks were recom-puted for P-wave velocities of 6.3 and 6.7 km s−1 to place lowerand upper error bounds on H and κ for a range of crustal P-wavevelocities. The combination of the two methods yields an overalluncertainty in the Moho depth at each station that is ∼3–4 km.The formal uncertainties obtained from using a mean crustal Vp of6.5 km s−1 are given in Table 1.

For two stations (KISZ and SENA), the crustal multiples (PpPsand PsPs + PpSs) were not clearly seen and the H–κ stackingmethod could not constrain the crustal thickness and Vp/Vs ratio.Therefore, the Moho depth was estimated using the Moho Ps arrivaltime, an assumed crustal Vp of 6.5 km s−1, an assumed Poisson’sratio of 0.26 and eq. (2) from Zandt & Ammon (1995). The re-ceiver functions for these stations can be found in the SupportingInformation and the estimated depths are reported in Table 1.

3.4 Joint inversion of Rayleigh-wave dispersion curvesand receiver functions

Crustal shear wave velocity models were obtained for each station byjointly inverting the receiver functions with Rayleigh-wave phaseand/or group velocities. Receiver functions constrain shear wavevelocity contrasts at interfaces between mediums, and single modeRayleigh-wave dispersion curves place constraints on averages ofthe absolute shear wave velocity at depth. Therefore, the combi-nation of the two datasets provides tighter constraints on the shearwave velocity structure at depth and bridges the resolution gaps in-herent in each dataset individually. The joint inversion method usedis a linearized, damped least squares method developed by Juliaet al. (2003) that incorporates a priori constraints.

To perform the inversion, receiver functions and dispersion mea-surements that have sampled the same area near a station are firstselected. The receiver functions are binned and stacked by similarray parameter in the groups 0.04–0.049, 0.05–0.059, 0.06–0.069and 0.07–0.079, to account for moveout from differing incident an-gles. The groups used for a particular station depend on the spreadof ray parameters associated with the receiver functions for that sta-

tion, and so the ray parameter groups vary from station to station.For each station, receiver functions were binned and stacked fortwo sets of overlapping frequency bands corresponding to Gaussianbandwidths of 1.0 (f ≤ 0.5 Hz) and 2.5 (f ≤ 1.25 Hz). Invertingreceiver functions at several frequency bandwidths help distinguishsharp discontinuities from gradational ones in the velocity models(Cassidy 1992; Julia et al. 2005; Julia 2007).

The starting model used in the joint inversion is an isotropicmedium with a 38-km-thick crust and a linear shear wave velocityincrease in the crust from 3.4 to 4.0 km s−1. The crust lies overa flattened Preliminary Reference Earth Model (PREM) for themantle (Dziewonski & Anderson 1981). The mantle shear wavevelocity structure is modelled to 290 km and then constrained to beequal to 5 per cent lower than PREM to better fit the longer perioddispersion velocities, for stations in Malawi, Rwanda, Tanzania andZambia within or near to the East African rift system (Julia et al.2005). For stations in Angola, Botswana, the Democratic Republicof Congo, Mozambique and Namibia away from the East African riftsystem, the shear wave velocity below 290 km depth is constrainedto PREM (Julia et al. 2005). The model is parametrized with layersof constant velocity that increase with depth. Layer thicknesses are1 and 2 km at the top of the model, 2.5 km between 3 and 65 kmdepth, 5 km between 65 and 265 km depth and 10 km below a depthof 265 km. Poisson’s ratio was fixed at 0.25 for crustal layers and atPREM values for mantle layers.

Uncertainties in the velocity models were estimated using theapproach in Julia et al. (2005) by repeating inversions for a range ofparameters, constraints and Poisson’s ratios. We found uncertaintiesin the crustal shear wave velocity to be about 0.1 km s−1 and about0.2 km s−1 in the upper mantle. These uncertainties in the velocitytranslate to uncertainties of ±2.5 km in crustal thickness.

Results from the joint inversion technique are tabulated in Table 1and the results for one station is illustrated in Fig. 7. Results for allstations can be found in the Supporting Information. Shear wavevelocities for typical lower crustal lithologies derived from exper-imentally determined Vp/Vs ratios have shown that the shear wavevelocities in the lower crust do not likely exceed 4.3 km s−1 and thatvelocities greater than 4.3 km s−1 are typical for mantle lithologies(Christensen & Mooney 1995; Christensen 1996). Therefore, fol-lowing the approach of Kgaswane et al. (2009) for southern Africa,Tokam et al. (2010) for western Africa and Tugume et al. (2013) foreastern Africa, in the joint inversion models the Moho is defined asthe depth at which the shear wave velocity exceeds 4.3 km s−1. Formost stations there is a significant increase in velocity or a velocitydiscontinuity at the depth where the velocity exceeds 4.3 km s−1,except for stations MKUS, MPIK, SERJ, MZM, KTWE and LSZ,where the shear wave velocity increase is gradational from the lowercrust to the upper mantle.

We also obtain estimates of the thickness of the mafic layeringin the lower crust from the joint inversion models. Previous stud-ies that have examined continental crustal structure (e.g. Holbrooket al. 1992; Christensen & Mooney 1995; Rudnick & Fountain1995; Rudnick & Gao 2003) have reported that common lowercrustal mafic lithologies, such as amphibolites, garnet-bearing andgarnet-free mafic granulite and mafic gneiss, have high shear wavevelocities (>3.9 km s−1) while intermediate-to-felsic lithologieshave lower shear wave velocities (<3.9 km s−1). Therefore wedefine the mafic lower crust as layers with shear wave velocitiesbetween 4.0 and 4.3 km s−1.

Within our reported uncertainties, we find a 1-to-1 correlationbetween our Moho estimates from the H–κ stack method and ourMoho estimates from the joint inversion method, for all but a few

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

540 M. Kachingwe, A. Nyblade and J. Julia

Table 1. Crustal structure results for the stations used in this study.

Tectonic region Station N Lat (◦) Long (◦)Elevation

(m) WDepth (1)

(km) Poisson’s ratioDepth (2)

(km)

Maficlower crust

(km)Avg Vs

(km s−1)

Kibaran Belt KIG 1 –1.96 30.06 1545 4 39.6 ± 0.9 0.20 ± 0.01 38 0 3.8PWET 1 –8.28 28.53 948 12 41.8 ± 1.0 0.25 ± 0.01 43 5 3.8

Average 40.7 0.23 41 3 3.8Nyanzian Block GETA 1 –2.87 33.24 1267 3 33.9 ± 1.3 0.28 ± 0.02 35 2 3.7MozambiqueOrogenic Belt

MOCU 4 –16.86 36.83 232 13 36.3 ±1.3 0.24 ± 0.01 35 0 3.7

NAPU 4 –15.08 39.25 373 11 42.8 ± 2.7 0.22 ± 0.01 40 10 3.7MTVE 1 –10.25 40.17 21 2 31.1 ± 1.8 0.29 ± 0.02SENA 4 –17.44 35.03 50 9 38.1a 0.26ZOMB 1 –15.38 35.35 885 12 38.3 ± 0.7 0.27 ± 0.02 38 0 3.6Average 37.3 0.26 38 3 3.7

BangweleuBlock

LWNG 3 –10.25 29.92 1455 25 42.5 ± 0.7 0.25 ± 0.02 43 0 3.7

MWEN 3 –10.06 28.70 995 10 45.2 ± 3.0 0.31 ± 0.01 38 0 3.7Average 43.9 0.28 41 0 3.7

Irumide Belt KASM 3 –10.22 31.14 1386 19 43.0 ± 2.8 0.20 ± 0.01 40 2 3.7CHAM 3 –10.95 31.07 1211 6 41.1 ± 0.9 0.24 ± 0.01 38 0 3.7ISOK 3 –10.17 32.65 1301 16 44.2 ± 1.1 0.22 ± 0.02 40 0 3.7KISH 3 –12.02 29.61 1132 6 37.8 ± 0.8 0.24 ± 0.01 38 3 3.7

MANS 3 –11.14 28.87 1259 10 41.8 ± 1.6 0.27 ± 0.01 43 13 3.8MKUS 3 –13.60 29.38 1250 5 43.5 ± 0.9 0.20 ± 0.01 40 5 3.7MPIK 3 –11.82 31.45 1392 11 48.2 ± 1.7 0.22 ± 0.02 45 7 3.8SERJ 3 –13.23 30.21 1388 10 45.2 ± 1.6 0.24 ± 0.02 43 5 3.8

SHWG 3 –11.19 31.74 1211 5 45.1 ± 2.6 0.23 ± 0.02 43 5 3.7Average 43.3 0.23 41 4 3.7

SouthernIrumide Belt

TETE 4 –16.13 33.57 159 20 39.2 ± 0.8 0.27 ± 0.01 38 0 3.7

MZM 1 –11.43 34.03 1258 10 34.7 ± 0.7 0.28 ± 0.01 40 5 3.7Average 37 0.28 39 3 3.7

Damara Belt TSUM 2 –19.20 17.58 1260 21 37.5 ± 0.4 0.24 ± 0.02 35 0 3.6RUND 5 –17.91 19.76 1073 9 41.0 ± 1.3 0.25 ± 0.01 38 0 3.6WIN 1 –22.56 17.10 1728 14 40.6 ± 0.9 0.29 ± 0.02 40 2 3.7

KAMZ 3 –14.79 24.80 1161 9 41.5 ± 1.5 0.22 ± 0.01 40 2 3.7KMPZ 3 –13.46 25.83 1251 13 40.7 ± 1.7 0.27 ± 0.01 38 0 3.7MONG 1 –15.25 23.15 1046 13 40.3 ± 1.1 0.27 ± 0.02 40 0 3.7GABZ 3 –12.17 26.37 1393 12 45.2 ± 1.8 0.25 ± 0.01 45 7 3.8KISZ 3 –12.11 25.50 1358 3 42.3a 0.26

MUFZ 3 –13.14 25.02 1189 8 49.3 ± 2.8 0.25 ± 0.02 38 8 3.7Average 42 0.26 39 2 3.7

Lufilian Arc KTWE 1 –12.81 28.21 1227 7 35.0 ± 1.1 0.28 ± 0.01 45 7 3.8LBB 1 –11.63 27.48 1283 7 34.4 ± 1.2 0.27 ± 0.02 35 2 3.6

Average 34.7 0.28 40 5 3.7Zambezi Belt TEZI 1 –15.75 26.02 1115 37 37.7 ± 0.5 0.26 ± 0.01 38 2 3.7

LSZ 2 –15.28 28.19 1200 26 34.7 ± 0.6 0.27 ± 0.01 40 5 3.7Average 36.2 0.27 39 4 3.7

West CentralAfrican Belt

POTQ 6 –8.64 13.55 108 6 33.9 ± 0.9 0.25 ± 0.01 33 8 3.7

Kasai block LUCA 5 –8.45 20.72 919 9 38.6 ± 1.9 0.26 ± 0.02 38 3 3.6DUNDO 6 –7.41 20.84 722 14 33.6 ± 0.9 0.19 ± 0.01 35 5 3.6Average 36.1 0.23 37 4 3.6

Angolan Block LUBAN 6 –14.91 13.45 1921 6 37.1 ± 1.6 0.30 ± 0.02 38 13 3.7RehobothProvince

MAUN 5 –19.90 23.53 949 5 44.2 ± 2.4 0.27 ± 0.01 38 2 3.7

Notes: N, network: 1 = AF, 2 = GSN, 3 = ZP, 4 = XV, 5 = Congo Craton, 6 = INAMET. W, number of waveforms.Depth (1) Crustal thickness from stacking P-wave receiver functions.Depth (2) Crustal thickness from jointly inverting P-wave receiver functions and surface wave dispersion curves.aCrustal thickness from the Moho Ps arrival time, an assumed crustal Vp of 6.5 km s−1, an assumed Poisson’s ratio of 0.26 and eq. (2) from Zandt &Ammon (1995).

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 541

Figure 6. H–κ stack results for station TEZI with weights of w1 = 0.4, w2 = 0.3, w3 = 0.3. On the right are the receiver functions labelled by the event’sback azimuth (top number) and epicentral distance in degrees (bottom number). On the left is the H–κ parameter space with the optimal estimates for H and κ .The contours map out the percentage values of the objective function given in the text. The centre bull’s-eye is the 95 per cent confidence bound, and the othercontours are incremented by 10 per cent. The optimal values for H and κ obtained are summarized along the top with their formal uncertainties.

stations (Fig. 8). Because of this, we use the crustal thickness esti-mates from the joint inversion method for our interpretation.

4 R E S U LT S

In this section, we review the crustal thickness, mean crustal shearwave velocity, thickness of the mafic lower crust and the meancrustal Poisson’s ratio for terranes of similar age. To do this, es-timates obtained in this study are combined with previously pub-lished results obtained using similar techniques, as described andreferenced in the Supporting Information and summarized andreferenced in Table 2. Estimates for a terrane’s crustal structureare averages of estimates from all the stations within that terrane.Table 2 gives the estimates for each terrane ordered by Precambrianage and in Fig. 9 the estimates are shown graphically.

4.1 Mesoarchean terranes

The estimates for average crustal thickness are very similar betweenthe Mesoarchean terranes and range from 36 to 39 km. The Swazi-land and Pietersburg terranes in the Kaapvaal Craton both have anaverage crustal thickness of 39 km and a fairly thick mafic lowercrust (Swaziland Terrane = 14 km and Pietersburg Terrane = 12

km). The Witswatersrand Terrane and the Kimberley Terrane, alsoin the Kaapvaal Craton, have average crustal thicknesses of 37 km,but the mafic lower crust in the Witswatersrand Terrane is 7 kmand only 2 km in the Kimberley Terrane. The East Tokwe Blockin the Zimbabwe Craton has an average crustal thickness of 36 kmand a mafic lower crustal thickness of 12 km. Poisson’s ratio forthe terranes range from 0.25 to 0.28. The estimates for the averagecrustal shear wave velocity range from 3.6 to 3.7 km s−1.

4.2 Neoarchean terranes

The average crustal thickness in the Neoarchean terranes is quitevariable, ranging from 36 to 45 km. In the Zimbabwe Craton, theWest Tokwe Block has an average crustal thickness of 36 km and amafic lower crustal thickness of 4 km. The Limpopo Belt is slightlythicker and has an average thickness of 41 km, and a mafic lowercrustal thickness of 14 km. In the Tanzania Craton, the Dodomanand Nyanzian blocks have an average crustal thickness of 38 km,and mafic lower crustal thicknesses of 3 and 4 km, respectively.The Bangweleu Block has an average crustal thickness of 41 kmand no mafic lower crust. The Angola Block has a 38 km thickcrust with a mafic lower crustal thicknesses of 13 km, and the KasaiBlock has a 37 km thick crust with a 3 km mafic lower crust. Incontrast, the Ntem Complex in Cameroon is much thicker than any

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

542 M. Kachingwe, A. Nyblade and J. Julia

Figure 7. Results from the joint inversion for station TEZI. (a) Receiverfunction stacks with low Gaussian frequency (1.0) stacks on top and highGaussian frequency (2.5) stacks on the bottom. Each group is ascending inray parameter bin. (b) Rayleigh-wave phase (top curve) and group velocities(lower curve). (c) The shear wave velocity model.

other Neoarchean terrane in the southern African subcontinent. Theaverage crustal thickness is 45 km and the mafic lower crustal thick-ness is 23 km. Most of the Neoarchean terranes have a Poisson’sratio of 0.25. However, in the Kasai Block it is 0.23, in the WestTokwe Block and Ntem Complex it is 0.26, in the Nyanzian Blockit is 0.27, in the Angola Block it is 0.30. The estimates for averageshear wave velocity are similar between the Neoarchean terranes,averaging between 3.6 and 3.7 km s−1. The exception is again theNtem Complex, with a fast average crustal shear wave velocity of3.9 km s−1.

4.3 Palaeoproterozoic terranes

Average crustal thickness in the Palaeoproterozoic terranes rangefrom 33 to 43 km. In southern Africa, the Okwa Terrane has anaverage crustal thickness of 43 km and an average mafic lowercrustal thickness of 13 km. Nearby, the Kheis Province has a 39-km-thick crust and mafic lower crustal thickness of 7 km. Comparatively,the Bushveld Complex crust is 41 km thick with a mafic lowercrustal thickness of 10 km. Also in southern Africa, the Rehoboth

Province has an average crustal thickness of 38 km and a maficlower crustal thickness of 2 km. The thinnest terrane, with a crustalthickness of 33 km and a mafic lower crustal thickness of 8 km, is theWest Central African Belt. The estimate for the West Central AfricaBelt comes from a single station on the west coast of Africa, and sothe crustal thickness may simply reflect the overall thinning of thecrust towards the ocean basin. In eastern Africa, the Ubendian Beltis 43 km thick with a mafic lower crustal thickness of 4 km, whilethe average Usagaran crust is 36 km thick with a 5-km-thick maficlower crust. Poisson’s ratio for the Palaeoproterozoic terranes variesslightly from 0.25 to 0.27. The estimates for the average shear wavevelocity are similar for the Palaeoproterozoic terranes. All have anaverage crustal shear wave velocity of 3.7 km s−1 except for theUsagaran Belt, which has an average velocity of 3.6 km s−1.

4.4 Mesoproterozoic terranes

The average crustal thickness in the Mesoproterozoic terranesranges from 33 to 41 km. The Namaqua-Natal Fold Belt has anaverage crustal thickness of 33 km and a mafic lower crustal thick-ness of 12 km, however the reported crustal thickness for this terranehas a large standard deviation of ±6 km. The Irumide Belt has anaverage crustal thickness of 41 km and a mafic lower crustal thick-ness of 4 km, while the average Southern Irumide crust is 39 kmthick with a 3-km-thick mafic lower crust. In eastern Africa, theKibaran Belt has a 40-km-thick crust and a 4-km-thick mafic lowercrust. Comparatively, the average crustal thickness of the RwenzoriBelt is 39 km with a 2-km-thick mafic lower crust. Poisson’s ratiofor all the Mesoproterozoic terranes is 0.25 except for in the Iru-mide and Southern Irumide Belt. The Irumide Belt has a Poisson’sratio of 0.23, while in the Southern Irumide it is 0.28. The aver-age crustal shear wave velocity is similar for all Mesoproterozoicterranes, averaging between 3.7 and 3.8 km s−1.

4.5 Neoproterozoic terranes

The Neoproterozoic terranes have similar average crustal thick-nesses, ranging from 38 to 40 km. The Damara Belt has an averagecrustal thickness of 38 km and a 2-km-thick mafic lower crust. Insoutheastern Africa, the Zambezi Belt has an average crustal thick-ness of 39 km and a mafic lower crust of 4 km. Comparatively, thecrustal thickness in the Lufilian Arc is 40 km with a 5-km-thickmafic lower crust. The Mozambique Belt has an average crustalthickness of 38 km and a mafic lower crustal layer 2 km thick.In western Africa, the crustal thickness of the Oubanguides Beltis 39 km and the mafic lower crust is 7 km thick. Estimates forPoisson’s ratio ranges from 0.26 to 0.28. The average crustal shearwave velocities in the Neoproterozoic terranes range from 3.6 to3.8 km s−1.

5 D I S C U S S I O N

Are there patterns in mean crustal thickness, crustal Poisson’s ratio,mean crustal velocity and lower crustal velocity that show age-dependent trends through the Archean and Proterozoic? Here weaddress that question, determining if the conclusion reached byTugume et al. (2013) that there are no secular trends is consistentwith our expanded dataset. We also compare our estimates to severalprevious studies that have examined the structure of Precambriancrust and discuss the implications of our estimates for understandingcrustal genesis from the Mesoarchean and onwards.

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 543

Figure 8. A comparison of Moho depth estimates determined from the H–κ stack method for receiver functions with Moho depth estimates determinedfrom the joint inversion method. The solid line shows the 1-to-1 correlation between to the two estimates, and the dashed lines show ±4 km from the 1-to-1correlation line.

The average crustal thickness for Archean terranes is 38 ± 3 km(SD), the average crustal thickness for Proterozoic terranes is 39± 3 km (SD) and the average crustal thickness for all Precambrianterranes is 39 ± 3 km (SD, Table 2). While some terranes are thickeror thinner than the average 39 km, most Precambrian terranes havethicknesses that fall within 1 SD of the mean thickness. The WestCentral African Belt (33 km) and the Okwa Terrane (43 km), forwhich there is only one data point, are exceptions, as is the NtemComplex.

Granitic rocks with a felsic composition have a Poisson’s ratio of0.24, while intermediate lithologies such as diorite have a Poisson’sratio around 0.27. Rocks with mafic lithologies like gabbro havevalues around 0.30 (Tokav & Vavakin 1982; Christensen 1996).The average Poisson’s ratio for many of the Precambrian terranes isbetween 0.25 and 0.27 with an overall average of 0.26 ± 0.01 (SD),which is indicative of bulk felsic to intermediate lithologies for thecontinental crust. A few terranes are more felsic (i.e. Irumide Belt,0.23), or more mafic (i.e. Pietersburg Terrane 0.28, Angola Block0.30, Southern Irumide Belt 0.28 and Lufilian Arc 0.28) (Table 2).

The average crustal shear wave velocity for all Archean andProterozoic terranes is between 3.6 and 3.8 km s−1 with an overallaverage of 3.7 km s−1 (Table 2). Given the uncertainties, all terranesfall within 1 SD of the mean velocity.

Most Precambrian terranes have a high shear wave velocity (be-tween 4.0 and 4.3 km s−1) layer in the lower crust that is between 2and 23 km thick. The thickness of this mafic layer is quite variable.In Archean terranes, the average thickness is 8 ± 6 km (SD) andin Proterozoic terranes the average thickness is 6 ± 3 km (SD).All Precambrian terranes exhibit a mafic lower crust except for theBangweleu Block, where a high velocity layer at the base of thecrust was not found.

In summary, we find little evidence for secular trends in crustalthickness, Poisson’s ratio, mean crustal velocity and lower crustalthickness for the southern African subcontinent. Importantly, we

find as much variability in crustal structure within an age range(i.e. Mesoarchean, Neoarchean, Palaeoproterozoic, Mesoprotero-zoic and Neoproterozoic) as there exists across all the Precambrianterranes. In addition, we find little difference in crustal structurebetween Meso- and Neoarchean terranes. Thus our findings areconsistent with the conclusion reached by Tugume et al. (2012,2013), which was based on a more limited dataset for Africa.

5.1 Comparison to other studies

Durrheim & Mooney (1991, 1994), Thompson et al. (2010), andAbbott et al. (2013) concluded that average Archean crust is differ-ent from post Archean crust, arguing for a secular trend in crustalgenesis. In contrast to these studies, our findings do not indicate achange in crustal structure as a function of age. Rudnick & Foun-tain (1995), Zandt & Ammon (1995), Tugume et al. (2012, 2013)and Stankiewicz & De Wit (2013), on the other hand, found nosignificant difference between the mean thickness, mean velocity,or composition of Archean and Proterozoic terranes. Our estimatesare consistent with these studies. However, our estimates indicatea larger variability in the lower crustal structure within terranes ofa similar age, as well as between terranes of different ages, thansuggested by Rudnick & Gao (2003). This finding is illustrated inFig. 10, which shows a plot of the thickness of the crust versus thethickness of the high shear wave velocity layer in the lower crust.The data are grouped by terrane age and show that for Precambrianterranes of all ages, there is significant variability in the thicknessof the high velocity layer in the lower crust and that there is nocorrelation with crustal thickness.

5.2 Implications for crustal genesis

The estimates from this study show little evidence for secular trendsin crustal structure for terranes in the southern African subcontinent

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

544 M. Kachingwe, A. Nyblade and J. Julia

Table 2. Summary of crustal structure for terranes in the southern African subcontinent.

Age Terrain N

Avg depth(1) (km)± 1 SD

Avg depth(2) (km)± 1 SD

Source(a)

AvgPoisson’s

ratioSource

(b)

Avg Maficlower crust(km) ± 1

SD

Avgcrustal Vs(km s−1)± 1 SD

Source(c)

Mesoarchean Swaziland Terrane (Kaapvaal Craton) 2 39 39 6,7 0.26 6 14 3.7 7Mesoarchean Witswatersrand Terrane (Kaapvaal Craton) 9 38 ± 2 37 ± 2 6,7 0.26 6 7 ± 2 3.7 7Mesoarchean E. Tokwe Terrane (Zimbabwe Craton) 3 38 ± 1 36 ± 1 6,7 0.25 6 12 ± 5 3.6 7Mesoarchean Pietersburg Terrane (Kaapvaal Craton) 2 39 39 6,7 0.28 6 12 3.6 7Mesoarchean Kimberley Terrane (Kaapvaal Craton) 14 37 ± 2 37 ± 2 6,7 0.25 6 2 ± 1 3.7 7Neoarchean Dodoman Block (Tanzania Craton) 7 39 ± 2 38 ± 2 2,3 0.25 2 3 ± 3 3.7 ± 0.1 3Neoarchean W. Tokwe Terrane (Zimbabwe Craton) 3 37 ± 0 36 ± 1 6,7 0.26 6 4 ± 1 3.6 7Neoarchean Limpopo Belt 11 41 ± 3 41 ± 3 6,7 0.25 6 14 ± 6 3.7 7Neoarchean Nyanzian Block (Tanzania Craton) 6 38 ± 4 38 ± 4 1,2,3 0.27 1,2 4 ± 4 3.7 ± 0 1,3Neoarchean Angola Block (Congo Craton) 1 37 38 1 0.30 1 13 3.7 1Neoarchean Ntem Complex (Congo Craton) 5 45 ± 2 45 ± 2 4,5 0.26 4 23 ± 4 3.9 ± 0.1 5Neoarchean Kasai Block (Congo Craton) 2 36 37 1 0.23 1 4 3.6 ± 0 1Neoarchean Bangweleu Block 2 44 41 1 0.25 1 0 3.7 1

Palaeoproterozoic Okwa Terrane 1 43 43 6,7 0.27 6 13 3.7 7Palaeoproterozoic Bushveld Complex (Kaapvaal Craton) 15 42 ± 3 41 ± 3 6,7 0.27 6 10 ± 4 3.7 7Palaeoproterozoic Kheis Province 6 40 ± 3 39 ± 3 6,7 0.25 6 7 ± 3 3.7 7Palaeoproterozoic West Central African Belt 1 34 33 1 0.25 1 8 3.7 1Palaeoproterozoic Ubendian Belt 12 43 ± 4 43 ± 4 2,3 0.26 2 4 ± 5 3.7 ± 0.1 3Palaeoproterozoic Usagaran Belt 9 38 ± 2 36 ± 3 2,3 0.26 2 5 ± 5 3.6 ± 0.1 3Palaeoproterozoic Rehoboth Province 1 44 38 1 0.27 1 2 3.7 1Mesoproterozoic Rwenzori Belt 5 37 ± 2 39 ± 2 2,3 0.25 2 2 ± 1 3.7 ± 0.1 3Mesoproterosoic Namaqua-Natal Fold Belt 8 40 ± 9 33 ± 6 6,7 0.25 6 12 ± 5 3.8 7Mesoproterosoic Irumide Belt 9 43 ± 3 41 ± 2 1 0.23 1 4 ± 4 3.7 ± 0.1 1Mesoproterosoic Southern Irumide Belt 2 37 39 1 0.28 1 3 3.7 1Mesoproterosoic Kibaran Belt 10 39 ± 3 40 ± 2 1,2,3 0.25 1,2 4 ± 3 3.8 ± 0.1 1,3Neoproterozoic Oubanguides Belt 4 39 ± 3 39 ± 3 4,5 0.23 4 7 ± 2 3.8 ± 0.1 5Neoproterozoic Zambezi Belt 2 36 39 1 0.27 1 4 3.7 1Neoproterozoic Damara Belt 9 42 ± 3 38 ± 3 1 0.26 1 2 ± 3 3.7 ± 0.1 1Neoproterozoic Mozambique Orogenic Belt 20 37 ± 3 38 ± 2 1,2,3 0.26 1,2 2 ± 3 3.6 ± 0.1 1,3Neoproterozoic Lufilian Arc 2 35 40 1 0.28 1 5 3.7 1

Average 39 ± 3 39 ± 3 0.26 ± 0.01 7 ± 5 3.7 ± 0.1

Notes: N, number of stations. Avg depth (1): Results from H–κ stacking of P-wave receiver functions. Avg depth (2): Results from jointly inverting P-wavereceiver functions and surface wave dispersion curves.Source (a): Crustal thickness from H–κ stacking and jointly inverting P-wave receiver functions and surface wave dispersion curves.Source (b): Poisson’s ratio from H–κ stacking of P-wave receiver functions.Source (c): Mafic lower crustal thickness and velocity from jointly inverting P-wave receiver functions and surface wave dispersion curves.1 = this study, 2 = Tugume et al. (2012), 3 = Tugume et al. (2013), 4 = Gallacher and Bastow (2012), 5 = Tokam et al. (2010), 6 = Nair et al. (2006),7 = Kgaswane et al. (2009).

spanning some 3.6 Ga of Earth’s history. Evaluated in the contextof secular variation and the models initially presented by Durrheim& Mooney (1991) and Rudnick & Fountain (1995), this findingsuggests one of two things. Either there have been few changes overmuch of Earth’s history in the processes that formed the crust of thesouthern African subcontinent, or crustal structure has been mod-ified to the point where secular trends are no longer observed. Ifsimilar modes of crustal genesis have been active throughout muchof Earth’s history, then modern day plate tectonic processes thatform continental crust, such as island-arc accretion and arc mag-matism, would appear to be the dominant crustal forming processesfrom at least the Mesoarchean.

Alternatively, during the Mesoarchean, hotter mantle tempera-tures may have favoured mantle plumes as the primary generator ofcontinental crust via intracontinental rifting, magmatism and under-plating (Arndt & Davaille 2013). If plumes were the primary driverof continental crustal genesis prior to ∼3.0 Ga, then secular trendsin crustal structure may be evident only between Mesoarchean andyounger crust. But, after 3.0 Ga, the rate of crustal growth may have

decreased substantially coeval with the onset of crustal recyclingand modification of the crust through subduction (Cawood et al.2013; Hawkesworth et al. 2013), and so finding Mesoarchean crustthat has not been reworked can be difficult. Indeed, the five terraneswhich formed in the Mesoarchean (Fig. 9) are overlain by the 3.1–2.87 Ga Pongola Basin, the 3.07–2.7 Ga Witwatersrand Basin andthe 2.7–2.8 Ga Ventersdorf Basin, possibly indicating some degreeof crustal modification in the Neoarchean (Begg et al. 2009 and ref-erences therein). Therefore, from our data ensemble it is difficult tomake definitive statements about differences in Mesoarchean versusyounger crustal structure or the lack thereof.

Consequently, while our findings do not show secular variationin crustal structure from the Mesoarchean and onwards, becauseof crustal recycling and modification by plate tectonic processes,which may have homogenized crustal structure across many ter-ranes, drawing firm conclusions about secular variations in crustalgenesis is not straightforward. The lack of secular changes in crustalstructure may simply reflect crustal reworking and not the processesby which the crust was initially extracted from the mantle.

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 545

Figure 9. A plot of average crustal structure. The terranes are ordered from oldest (left-hand panel) to youngest (right-hand panel) and are shaded with the fillpatterns by age (Mesoarchean, Neoarchean, Palaeoproterozoic, Mesoproterozoic and Neoproterozoic), which is the same as in Fig. 1 and shown in the key. Theerror bars on Moho depths are ±1 SD of the average for each terrane for which there are 3 or more stations. The grey shading represents the average thicknessof layers in the lower crust with Vs ≥ 4.0 km s−1. The bold, dashed lines at 36 and 42 km show the average crustal thickness range (i.e. 39 km ± 3 km). Theblack bars indicate the number of stations in each terrane.

Figure 10. Plot showing average crustal thickness for each terrane versus the thickness of crustal layers with Vs ≥ 4.0 km s−1 (i.e. mafic lower crust).

6 S U M M A RY

In this paper, we have obtained estimates of crustal structure for 39new stations using the H–κ stacking method and jointly invertingP-wave receiver functions with surface wave dispersion measure-

ments. This includes estimates for several Precambrian terranes forwhich there were few previously published results of crustal struc-ture (Bangweleu Block, West Central African Belt, Irumide Belt,Southern Irumide Belt, Zambezi Belt and Lufilian Arc).

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

546 M. Kachingwe, A. Nyblade and J. Julia

After combining our new estimates with published results forthe southern African subcontinent and examining crustal thickness,Poisson’s ratio and the shear wave velocity for some 30 terranesranging in age from ∼3.6 to 0.5 Ga, we find little evidence forsecular trends in Precambrian crustal structure. For both Archeanand Proterozoic crust, we find similar ranges in crustal thickness(38–39 ± 3 km SD), Poisson’s ratio (0.26 ± 0.01 SD), and shearwave velocity (3.7 ± 0.1 km s−1 SD), as well as similar amounts ofheterogeneity in the lower crustal structure.

This finding can be interpreted in one of two ways, (1) either thatcontinental crust has been formed by similar processes since theMesoarchean, or (2) crustal reworking by plate tectonic processeshas erased any secular trends in crustal structure that would indicatea change in crustal genesis at some point in Earth’s history.

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

We gratefully acknowledge the field support from IRIS-PASSCAL,the Geological Survey of Zambia and the Direccao Nacional DeGeologia in Mozambique, and many AfricaArray station operatorsin Angola, Botswana, the Democratic Republic of Congo, Malawi,Mozambique, Namibia, Rwanda, Tanzania and Zambia. We alsothank three anonymous reviewers for their helpful comments. Thisstudy was funded by the National Science Foundation, grant num-bers EAR 1128936, 0838426, 0824781, 0440032; OISE 053006.Figures were prepared using GMT (Wessel & Smith 1998).

R E F E R E N C E S

Abbott, D.H., Mooney, W.D. & VanTongeren, J.A., 2013. The characterof the Moho and lower crust within Archean cratons and the tectonicimplications, Tectonophysics, 609, 690–705.

Ammon, C.J., 1990. The isolation of receiver effects from teleseismicP waveforms, Bull. seism. Soc. Am., 81(6), 2504–2510.

Arndt, N. & Davaille, A., 2013. Episodic Earth evolutions, Tectonophysics,609, 661–674.

Begg, G.C. et al., 2009. The lithospheric architecture of Africa: seismictomography, mantle petrology, and tectonic evolution, Geosphere, 5(1),23–50.

Cahen, L., Snelling, N., Delhal, J. & Vail, J., 1984. The Geochronology andEvolution of Africa, Clarendon Press.

Cassidy, J.F., 1992. Numerical experiments in broad-band receiver functionanalysis, Bull. seism. Soc. Am., 82, 1453–1474.

Castaing, C., Feybesse, J.L., Thieblemont, D., Triboulet, C. & Chevre-mont, P., 1994. Palaeogeographical reconstructions of the Pan-African/Brasiliano orogen: closure of an oceanic domain or intraconti-nental convergence between major blocks?, Precambrian Res., 67, 327–344.

Cawood, P.A., Hawkesworth, C.J. & Dhuime, B., 2013. The continentalrecord and the generation of continental crust, Bull. geol. Soc. Am., 125,14–32.

Christensen, N.I., 1996. Possion’s ratio and crustal seismology, J. geophys.Res., 101, 3139–3156.

Christensen, N.I. & Mooney, W.D., 1995. Seismic velocity structure andcomposition of the continental crust: a global view, J. geophys. Res., 100,9761–9788.

Cornell, D.H., Thomas, R.J., Moen, H.F.G., Reid, D.L. & Gibson, R.L.,2006. The Namaqua-Natal province, in The Geology of South Africa,pp. 325–379, eds Johnson, M.R., Anhaeusser, C.R. & Thomas, R.J. Ge-ology Society of South Africa.

De Waele, B., 2005. The Proterozoic geological history of Irumide belt,Zambia, PhD thesis, Perth, Australia, Curtin University of Technology,358 p.

De Waele, B., Kampunzu, A.B., Mapani, B.S.E. & Tembo, F., 2006. TheIrumide Belt of Zambia, J. Afr. Earth Sci., 46, 36–70.

De Waele, B., Johnson, S.P. & Pisarevsky, S.A., 2008. Paleoproterozoic toNeoproterozoic growth and evolution of the eastern Congo Craton: itsrole in the Rodinia puzzle, Precambrian Res., 160, 127–141.

De Wit, M.J. et al., 1992. Formation of an Archean continent, Nature, 357,553–563.

Dirks, P.H.G.M. & Jelsma, H.A., 2002. Crust-mantle decoupling and growthof the Archean Zimbabwe Craton, J. Afr. Earth Sci., 34, 157–166.

Dugda, M.T., Nyblade, A.A., Julia, J., Langston, C.A., Ammon, C.J. &Simiyu, S., 2005. Crustal structure in Ethiopia and Kenya from receiverfunction analysis: implications for Rift development in eastern Africa,J. geophys. Res., 110, B01303, doi:10.1029/2004JB003065.

Durrheim, R.J. & Mooney, W.D., 1991. Archean and Proterozoic crustalevolution: evidence from crustal seismology, Geology, 19, 606–609.

Durrheim, R.J. & Mooney, W.D., 1994. Evolution of the Precambrian litho-sphere: seismological and geochemical constraints, J. geophys. Res., 99,15 359–15 374.

Dziewonski, A.M. & Anderson, D.L., 1981. Preliminary reference earthmodel, Phys. Earth planet. Inter., 25, 297–356.

Efron, B. & Tibshirani, R., 1991. Statistical data analysis in computer age,Science, 253, 390–395.

Eglington, B.M. & Armstrong, R.A., 2004. The Kaapvaal Craton and adja-cent orogens, southern Africa: a geochronological database and overviewof the geological development of the craton, South Afr. J. Geol., 107,13–32.

Fuchs, K., Altherr, B., Muller, B. & Prodehl, C., 1997. Structure and dynamicprocesses in the lithosphere of the Afro-Arabian rift system, Tectono-physics, 278, 1–352.

Gallacher, R.J. & Bastow, I.D., 2012. The development of magmatism alongthe Cameroon Volcanic Line: evidence from teleseismic receiver func-tions, Tectonophysics, 31, TC3018, doi:10.1029/2011TC003028.

Goodwin, A.M., 1996. Principles of Precambrian Geology, Academic Press,327 pp.

Hawkesworth, C., Cawood, P. & Dhuime, B., 2013. Continental growth andthe crustal record, Tectonophysics, 609, 651–660.

Holbrook, W.S., Mooney, W.D. & Christensen, N.I., 1992. The seismicvelocity structure of the deep continental crust, in Continental LowerCrust, Chapter 1, pp. 1–43, eds Fountain, D.M., Arculus, R. & Kay, R.W.,Elsevier.

Jelsma, H.A. & Dirks, P.H.G.M., 2002. Neoarchean tectonic evolution of theZimbabwe Craton, in The Early Earth: Physical, Chemical, and Biolog-ical Development, Vol. 199, pp. 183–211, eds Fowler, C.M.R., Ebinger,C.J. & Hawkesworth, C.J., Geological Society of Special Publications.

Johnson, M.R., Anhaeusser, C.R. & Thomas, R.J., eds, 2006. The Geologyof South Africa, Vol. 691, Geological Society of South Africa.

Julia, J., 2007. Constraining velocity and density contrasts across the crust-mantle boundary with receiver function amplitudes, Geophys. J. Int., 171,286–301.

Julia, J., Ammon, C.J., Herrmann, R.B. & Correig, A.M., 2000. Joint in-version of receiver function and surface wave dispersion observations,Geophys. J. Int., 143, 99–112.

Julia, J., Ammon, C.J. & Herrmann, R.B., 2003. Lithosphere structure ofthe Arabian Shield from the joint inversion or receiver functions andsurface-wave group velocities, Tectonophysics, 371, 1–21.

Julia, J., Ammon, C.J. & Nyblade, A.A., 2005. Evidence for mafic lowercrust in Tanzania, East Africa, from joint inversion of receiver functionsand Rayleigh wave dispersion velocities, Geophys. J. Int., 162, 555–569.

Kgaswane, E.M., Nyblade, A.A., Julia, J., Dirks, P.H.G.M., Durrheim, R.J.& Pasyanos, M.E., 2009. Shear wave velocity structure of the lowercrust in Southern Africa: evidence composition heterogeneity withinthe Archean and Proterozoic terrains, J. geophys. Res., 114, B12304,doi:10.1029/2008JB006217.

Klerkx, J., 1987. Crustal evolution of northern Kibaran Belt, eastern andcentral Africa, Geodyn. Ser., 17, 217–233.

Kramers, J.D., McCourt, S. & van Reenen, D., 2006. The Limpopo Belt,in The Geology of South Africa, pp. 209–236, eds Johnson, M.R., An-haeusser, C.R. & Thomas, R.J., Geology Society of South Africa.

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from

Crustal structure of Precambrian terranes 547

Langston, C.A., 1979. Structure under Mount Rainier, Washington, inferredfrom teleseismic body waves, J. geophys. Res., 84, 4749–4762.

Last, R.J., Nyblade, A.A., Langston, C.A. & Owens, T.J., 1997. Crustalstructure of the East African Plateau from receiver functions and Rayleighwave phase velocities, J. geophys. Res., 102, 24 469–24 483.

Lenoir, J.L., Liegeois, J.P., Theunissen, K. & Klerkx, J., 1994. The paleopro-terozoic Ubendian shear belt in Tanzania; geochronology and structure,J. Afr. Earth Sci., 19, 160–184.

Ligorria, J.P. & Ammon, C.J., 1999. Iterative deconvolution and receiverfunction estimation, Bull. seism. Soc. Am., 89, 1395–1400.

Manya, S.M. & Maboko, M.A.H., 2003. Dating basaltic volcanism in theNeoarchean Sukumaland Greenstone Belt of the Tanzania Craton Us-ing the Sm-Nd method: implications for the geological evolution of theTanzania Craton, Precambrian Res., 121, 35–45.

McCourt, S. & Armstrong, R.A., 1998. SIMS U-Pb zircon geochronology ofgranites from Central Zone, Limpopo Belt, southern Africa: implicationsfor the age of the Limpopo Orogeny, S. Afr. J. Geol., 101, 329–338.

Midzi, V. & Ottemoller, L., 2001. Receiver function structure beneaththree Southern Africa seismic broadband stations, Tectonophysics, 339,443–454.

Nair, S.K., Gao, S.S., Liu, K.H. & Silver, P.G., 2006. Sothern Africa crustalevolution and composition: constraints from receiver function studies,J. geophys. Res., 111, B02304, doi:101029/2005JB003802.

Nguuri, T., Gore, J., James, D., Webb, S., Wright, C., Zengeni, T., Gwavava,O. & Snoke, J., 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.

O’Donnell, J.P., Adams, A., Nyblade, A.A., Mulibo, G.D. & Tugume, F.,2013. The upper most mantle shear wave velocity structure of easternAfrica from Rayleigh wave tomography: constraints on rift evolution,Geophys. J. Int., 194(2), 961–978.

Owens, T.J. & Zandt, G., 1985. The response of the continental crust-mantle boundary observed on broadband teleseismic receiver functions,Geophys. Res. Lett., 12, 705–708.

Raveloson, A., Nyblade, A., Fishwick, S., Mangongolo, A. & Master, S.,2015. The upper mantle seismic velocity structure of south-central Africaand the seismic architecture of the Precambrian lithosphere beneath theCongo Basin, in The Geology and Research Potential of the Congo Basin,Chapter 1, eds De Wit, M.J., Guillocheau, F. & De Wit, M.C.J., Springer.

Rudnick, R.L. & Fountain, D.M., 1995. Nature and composition of thecontinental crust the continental crust: a lower crustal perspective, Rev.Geophys., 33, 267–309.

Rudnick, R.L. & Gao, S., 2003. Composition of the continental crust, inTreatise on Geochemistry, Vol. 3, pp. 1–64, Elsevier.

Schulter, T., 1997. Geology of East Africa, Druckerei zu Altenburg GmbH.Shackleton, R.M., 1986. Precambrian collision tectonics in Africa, collision

tectonics, Geol. Soc. Spec. Publ., 19, 329–249.Stankiewicz, J. & De Wit, M., 2013. 3.5 billion years of reshaped Moho,

southern Africa, Tectonophysics, 609, 675–689.Tchameni, R., Mezger, K., Nsifa, N.E. & Pouclet, A., 2001. Neoarchean

crustal evolution in the Congo: evidence from K rich granitoids of theNtem Complex, southern Cameroon, J. Afr. Earth Sci., 30(1), 133–147.

Thompson, D.A., Bastow, I.D., Helffrich, G., Kendall, J.M., Wookey, J.,Snyder, D.B. & Eaton, D.W., 2010. Precambrian crustal evolution: seis-mic constraints from the Canadian Shield, Earth planet. Sci. Lett., 297,655–666.

Tokam, A-P.K., Tabod, C.T., Nyblade, A.A., Julia, J., Wiens, D.A. &Pasyanos, M.E., 2010. Structure of the crust beneath Cameroon, WestAfrica, from joint inversion of Rayleigh wave group velocities and re-ceiver functions, Geophys. J. Int., 183, 1061–1076.

Tokav, A.P. & Vavakin, V.V., 1982. Poisson’s ration behavior in crystallinerocks: application to study of the Earth’s interior, Phys. Earth planet.Inter., 29, 24–29.

Toteu, S.F., Penaye, J. & Poudjom Djomani, Y., 2004. Geodynamic evolu-tion of the Pan-African Belt in Central Africa with special reference toCameroon, Can. J. Earth Sci., 41, 73–85.

Tugume, F., Nyblade, A. & Julia, J., 2012. Moho depths and Poisson’s ratiosof Precambrian crust in East Africa: evidence for similarities in Archeanand Proterozoic crustal structure, Earth planet. Sci. Lett., 355–356, 73–81.

Tugume, F., Nyblade, A., Julia, J. & van de Meijde, M., 2013. Precam-brian crustal structure in Africa and Arabia: evidence lacking for secularvariation, Tectonophysics, 609, 250–266.

Wessel, P. & Smith, W., 1998. New, improved version of the Generic Map-ping Tools Released, EOS, Trans., Am. Geophys. Un., 79, 579.

Youssof, M., Thybo, H., Artemieva, I.M. & Levander, A., 2013. Mohodepth and crustal composition in Southern Africa, Tectonophysics, 609,267–287.

Zandt, G. & Ammon, C.J., 1995. Continental crust composition constrainedby measurements of crustal Poisson’s ratio, Nature, 374, 152–154.

Zhu, L. & Kanamori, H., 2000. Moho depth variation in SouthernCalifornia from teleseimic receiver functions, J. geophys. Res., 105,2969–2980.

S U P P O RT I N G I N F O R M AT I O N

Additional Supporting Information may be found in the onlineversion of this paper:

Supplemental File A. A summary of the seismic estimates ofcrustal structure from previous studies for the Precambrian terranesof the southern African subcontinent is provided in this supple-mental file. Many of the estimates reviewed have been include inTable 2.Supplemental File B. List of tele seismic events used.Supplemental File C. Results from the H–κ stacking method forall stations.Supplemental File D. Results from jointly inverting P-wave re-ceiver functions and Rayleigh wave phase and group velocities.Supplemental File E. Radial (left) and tangential (right) receiverfunctions versus ray parameter for stations KISZ and SENA wherethe Moho depth was estimated using the Moho Ps arrival time, an as-sumed crustal Vp of 6.5 km s−1, an assumed Poisson’s ratio of 0.26,and eq. (2) from Zandt et al. (1995). (http://gji.oxfordjournals.org/lookup/suppl/doi:10.1093/gji/ggv136/-/DC1)

Please note: Oxford University Press is not responsible for the con-tent or functionality of any supporting materials supplied by theauthors. Any queries (other than missing material) should be di-rected to the corresponding author for the paper.

by guest on May 19, 2015

http://gji.oxfordjournals.org/D

ownloaded from