A top to bottom lithospheric study of Africa and Arabia

Transcription

1 Available online at Tectonophysics 444 (2007) A top to bottom lithospheric study of Africa and Arabia Michael E. Pasyanos a,, Andrew A. Nyblade b a Earth Science Division, Lawrence Livermore National Laboratory, Livermore, California, USA b Department of Geosciences, Pennsylvania State University, University Park, Pennsylvania, USA Received 8 November 2006; received in revised form 14 June 2007; accepted 11 July 2007 Available online 22 August 2007 Abstract We study the lithospheric structure of Africa, Arabia and adjacent oceanic regions with fundamental-mode surface waves over a broad period range. Including group velocities with periods shorter than 35 s allows us to examine shallower features than previous studies of the whole continent. In the process, we have developed a crustal thickness map of Africa. Main features include crustal thickness increases under the West African, Congo, and Kalahari cratons. We find crustal thinning under Mesozoic and Cenozoic rifts, including the Benue Trough, Red Sea, and East, Central, and West African rift systems, along with less abrupt crustal thickness changes at passive continental margins. We also find crustal thickness differences in North Africa between the West African Craton and East Saharan Shield. Crustal shear wave velocities are generally faster in oceanic regions and cratons, and slower in more recent crust and in active and remnant orogenic regions. Deeper structure, related to the thickness of cratons and modern rifting, is generally consistent with previous work. Under cratons we find thick lithosphere and fast upper mantle velocities, while under rifts we find thinned lithosphere and slower upper mantle velocities. However, we also find the lack of a thick cratonic keel beneath the central portion of the Congo Craton. There are no consistent effects in areas classified as hotspots, indicating that there seem to be numerous origins for these features. Finally, it appears that the African Superswell, which is responsible for high elevation and uplift over large portions of Africa, has had a significantly different impact (as indicated by features such as temperature, time of influence, etc.) in the north and the south. This is consistent with episodic activity at shallow depths, which is well-expressed in northeastern Africa and Arabia today Elsevier B.V. All rights reserved. Keywords: Crust; Moho; Upper mantle; Group velocity; Lithosphere; Congo Craton 1. Introduction In general, the seismic structure of Africa and Arabia remains poorly studied, owing largely to the sparse distribution of seismic stations and to the existence of large aseismic areas on the African and Arabian plates. Corresponding author. Lawrence Livermore National Laboratory, P.O. Box 808, L-205, Livermore, CA , USA. Tel.: ; fax: address: pasyanos1@llnl.gov (M.E. Pasyanos). Studies using temporary deployments of seismometers have been carried out in a number of geographic localities, for example around the Cenozoic rifts and plateaus in Ethiopia, Kenya and Tanzania, and across the shields in Arabia and southern Africa, providing detailed seismic images of the crust and upper mantle. However, for many parts of Africa and Arabia little is known about the seismic structure of the lithosphere and, for these regions, what little we know about structure comes primarily from global and continentalscale tomography studies /$ - see front matter 2007 Elsevier B.V. All rights reserved. doi: /j.tecto

2 28 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) The earliest tomography studies of mantle structure (Dziewonski, 1984) revealed thick, seismically fast keels under cratonic regions of Africa, and slow structure elsewhere. More recent studies have provided refined images of the cratonic keels (e.g. Ritsema et al., 1999; Ritsema and van Heijst, 2000; King and Ritsema, 2000; Grand, 2002; Priestley et al., 2006), as well as of anomalously slow upper mantle structure under eastern Africa and parts of the Arabian Peninsula (Debayle et al., 2001; Simmons et al., 2007). Many of these studies have relied on surface wave data to provide spatial coverage because of the sparse seismic station coverage mentioned above, and have variable resolution depending mainly on the density of crossing ray paths and the frequency content of the surface waves used. For crustal structure, continentalscale models for Africa have been provided as a part of global models (e.g. CRUST5.1; Mooney et al., 1998). This is in contrast to crustal thickness maps developed for

3 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Europe (Meissner et al., 1987), Asia (Kunin et al., 1987), and North America (Das and Nolet, 1998). In this study, we have attempted to improve upon previous continental-scale models of crust and upper mantle structure by developing a new velocity model using group velocity measurements of fundamentalmodel Rayleigh and Love waves between periods of 7 and 100 s. Our model differs from previously published models in several ways, including the density of ray coverage in some localities, the period range of the group velocities, and the modeling algorithm. Importantly, our model is the first continental-scale model developed using dispersion measurements at periods less than 40 s, measurements that provide good constraints on crust and uppermost mantle structure. Because our data set does not contain dispersion measurements at periods greater than 100 s, our model provides limited resolution of sub-lithospheric mantle structure. Consequently, we focus the discussion of our model in this paper on structure at crustal and lithospheric mantle depths. Our model is broadly consistent with results of previous studies in many ways, but it also shows some new features. After describing the new model and comparing it with previous continental-scale models, we examine in more detail the structure of the lithospheric keels under the Archean cratons. 2. Background In this section, we provide a brief review of the major tectonic features of Africa and Arabia that are relevant to the discussion of our new model, in particular, Archean blocks, Mesozoic and Cenozoic rifts, and major sedimentary basins (Fig. 1). For comprehensive reviews of African geology and tectonics we refer the reader to Cahen et al. (1984) and Burke (1996). The tectonics of Africa (Fig. 1) are unusual in the sense that, in general, the structure of the continent is very old and that tectonic activity more recent than the Paleozoic occurs only in limited regions. In fact, of all the continents, Africa has the highest percentage of exposed Precambrian crust (Goodwin, 1996) Northern and Central Africa The Precambrian tectonic history of northern and central Africa, very briefly, includes major cratonforming events in the Archean, followed by a complex history of mountain building and rifting that culminated with the Pan-African Orogeny during the formation of Gondwanaland. Archean crust is exposed in the Man and Reguibat blocks within the West African Shield, within the Congo Shield in Gabon, Cameroon, and the Central African Republic, and in the Uweinat block in southern Egypt and northern Sudan. Cenozoic tectonic structure of northern Africa is dominated by convergence with Eurasia. In northwest Africa, this expresses itself in the Atlas Mts. orogeny. In the Western Mediterranean, the African Plate subducts along the Calabrian, Hellenic, and Cyprean arcs. North Africa is also dotted with several hotspots including the Hoggar in southern Algeria, Tibesti in northern Chad, and Darfour in western Sudan. Offshore, hotspots are found associated with the Canary and Cape Verde Islands. In addition, Mesozoic rift basins formed across parts of north-central and central Africa (the Central African Rift System) during the opening of the Atlantic in the Jurassic Eastern Africa and Arabia Eastern Africa is where much of the Cenozoic tectonic activity is found within Africa. The Red Sea Rift has split the Nubian and Arabian Shields; the Gulf of Aden separates the Arabian and Somali Plates; and the East African Rift is starting to separate the Somali Plate from the rest of Africa. These three rift zones form a triple junction in the Afar. This area also sits at the northern edge of the African Superswell, an uplifted region spanning much of eastern and southern Africa characterized by high residual topography (Nyblade and Fig. 1. Tectonic map of Africa. Platform and shield areas are indicated by the single and double hatched lines. Deep basins are noted by the 2.5 km sediment thickness contours (graylines). Plate boundaries are indicated by the thick blacklines. The African Superswell is outlined by the dashed lines. Af = Afar Triple Junction, Alp = Alps, AP = Arabian Platform, Ar = ArabianShield,AM=AtlasMts,Asc=AscensionI.,Atl = Atlantic Ocean, BT = Benue Trough, BS = Black Sea, Can = Canary Is., Car = Carpathians, Cas = Caspian Sea, C = Caucasus, CAB = C.African Belt, CAR = C.African Rift, CFB = Cape Fold Belt, Com = Comores I., Con = Congo Basin, CC = Congo Craton, D = Darfour, DKB = Damar-KatangaBelt,DSR=DeadSeaRift,E=Etna,EAR=E.AfricanRift,EM= E.Mediterranean, ES = E.Sharan Shield, Eth = Ethiopian Plateau, Eto = Etosha Basin, GA = Gulf of Aden, GG = Gulf of Guinea, Ho = Hoggar, Ill = Illizi Basin, Ion = Ionian Basin, Ind = Indian Ocean, IP = Iranian Plateau, IR = Indian Ridge, KB = Kheiss Belt, KV = Kaapvaal Craton, Ka = Kalahari Basin, Ke = Kenya Rift, Li = Limpopo Belt, LV = Lake Victoria, Mad = Madagascar, MC = Mt.Cameroon, M = Mesopotamian Foredeep, Moz = Mozambique Basin, NN = Namaqua-Natal Belt, Nu = Nubian Shield, Per = Persian Gulf, R = Reunion, RS = Red Sea Rift, Sey = Seychelles, Si = Sirt Basin, SH = St.Helena, So = Somali Basin, Tan = Tanzania Craton, Ta = Taoudenni Basin, Tib = Tibesti, Tri = Tristan da Cunha, TP = Turkish Plateau, Ukr = Ukranian Shield, Uw = Uweinat Block, V = Vema Seamount, Wal = Walvis Ridge, WAC = W.African Craton, WAR = W.African Rift, WM = W.Mediterranean, Zag = Zagros Mts., Zam = Zambia Craton, Zim = Zimbabwe Craton.

4 30 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Robinson, 1994). Rifting to the south continues along the Eastern and Western Branches of the East African Rift System into southern Africa. The Arabian Peninsula is separating from Africa along the Red Sea and Gulf of Aden rifts. Northwest movement of the Arabian Plate is expressed as convergence in the Zagros orogeny, elevation of the Turkish and Iranian Plateaus, and left-lateral movement along the Dead Sea Rift and East Anatolian Fault Zone. The large sedimentary basins of the Persian Gulf and Mesopotamian Foredeep lie at the base of the Zagros Southern Africa In southern Africa, two major Archean blocks are found, one in the southern Democratic Republic of Congo (D.R.C.) and extending into northern Angola, and the Kalahari Craton. The Kalahari Craton occupies much of southern Africa and is comprised of several major provinces. Trending from north to south, these are the Zimbabwe Craton, the Limpopo Belt, and the Kaapvaal Craton. The Archean blocks in the D.R.C. and Angola form the southern extent of the Congo Craton. Many Proterozoic mobile belts, which formed during multiple orogenic events, surround the Congo and Kalahari Cratons. At the southern tip of the continent, a younger (Proterozoic Paleozoic) orogenic event formed the Cape Fold Belt as part of the Gondwanide orogen with proto-south America and proto-antarctica. The Phanerozoic history of southern Africa is marked by the development of the Karoo rifts and basins, many of which are associated with the breakup of Gondwanaland. In the Mesozoic and Cenozoic, several of the hotspots formed offshore southern Africa, including the St. Helena and Ascension Island hotspots to the west, and the Comores and Reunion Island hotspots to the east. Starting during the Jurassic, Madagascar along with India, Antarctica and Australia, split off from the Somali Coast and traveled south before first separating from Antarctica and Australia ( 130 Ma), and then India ( 90 Ma) during the Cretaceous Rifts and basins The Mesozoic and Cenozoic rifts within Africa have largely formed in Proterozoic mobile belts surrounding the Archean blocks. For example, the Benue Trough and the rest of the Central African Rift System formed to the north of Archean blocks in Cameroon and the Central African Republic that are part of the Congo Craton. The Mesozoic rifts in north-central Africa formed to the east of the West African Shield, the two branches of the East African Rift System developed around the margins of the Tanzania Craton, and many of the Karoo rifts in southern Africa formed around the eastern, northern, and western sides of the Zimbabwe Craton. Throughout Africa, there are a large number of sedimentary basins. Using the system of Clifford (1986), the basins can be classified as interior basins (large basins within stable continental shield) like the Taoudenni and Etosha Basins, marginal sag basins (basins located at the continental edge) such as the Somali and Mozambique Basins, rift basins like the Benue Trough and East African and Sirt Basins, and composite and complex basins, such as the Cuvette Centrale and Illizi Basins (Schlüter, 2006). The Cuvette Centrale is of particular interest, because it covers the interior of what is often referred to as the greater Congo Craton. Under the Cuvette Centrale, the Archean blocks in the Central African Republic, Cameroon, and Gabon to the north and west are thought to connect with the Archean blocks in the southern D.R.C. and northern Angola to form perhaps the largest Archean craton in Africa (e.g. Kampunzu and Popoff, 1991). However, a Neo-Proterozoic rift has been imaged under the Cuvette Central (Daly et al., 1991), suggesting that there might not be any Archean crust under the sedimentary cover African Superswell Another significant influence for the continent is the African Superswell (Nyblade and Robinson, 1994). This large feature, as noted previously, consists of elevated portions of the African continent along a swath running from south of Africa to the middle of the Arabian Peninsula, and whose origin may be connected to deep lower mantle processes (Ritsema et al., 1999). Besides the high dynamic topography, visible features of the superswell include major volcanic episodes, elevated heat flow in mobile belts of southern Africa (Nyblade and Robinson, 1994), and rifting and volcanism in eastern Africa and the Afar region. Global tomography models (e.g. van der Hilst et al., 1997; Grand, 2002; Simmons et al., 2007) have persistently found low velocities in the lower mantle beneath southern Africa, and both global and regional tomography models show low velocities in the upper mantle beneath eastern Africa and the Arabian Shield (e.g., Achauer et al., 1994; Slack et al., 1994; Ritsema et al., 1999; Bastow et al., 2005; Benoit et al., 2006b; Park and Nyblade, 2006; Park et al., 2007). 3. Data and measurements Our main source for station coverage of the region comes from stations in the Global Seismic Network (GSN),

5 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) which provides coverage both in Africa and Arabia, as well as offshore. Other contributing FDSN stations include those from the GEOSCOPE network, the MEDNET network, and other U.S. networks. Coverage of several regions has been significantly improved with PASSCAL deployments in Tanzania (Nyblade et al., 1996), Saudi Arabia (Vernon et al., 1996), Ethiopia and Kenya (Nyblade and Langston, 2002), South Africa (Nguuri et al., 2001), and is continuing with a deployment in Cameroon (Tibi et al., 2005). We have included data from the one open station (CM18) in the Cameroon deployment. In the case of other PASSCAL deployments (i.e. eastern Turkey, Tanzania, South Africa) sometimes only subsets of the complete station list were used. This is because, with surface waves, resources are best spent making measurements at widely dispersed stations rather than at closely spaced stations in a small region. Data from several stations in the MIDSEA deployment (van der Lee et al., 2001) improved coverage in North Africa. In some instances, particularly in the Middle East, we made use of data from local seismic networks including those in Jordan, Kuwait, Iraq, Libya, and the United Arab Emirates. A map showing stations where we have made at least 10 Rayleigh wave dispersion measurements at 50 s period is shown in Fig. 2a. Using these stations, we have made surface wave dispersion measurements for tens of thousands of paths across Africa and Arabia. The measurements employed here represent a five-fold increase in the number of paths than an earlier study which covered North Africa and portions of the Middle East (Pasyanos and Walter, 2002). Surface wave path coverage is shown in Fig. 2b. Seismicity along the mid-atlantic rift, Indian ridge, East African Rift, Red Sea, Gulf of Aden, and along the Tethys collision zone provides the bulk of events for dispersion analysis. Overall, we have excellent path coverage of the region. Clearly, however, we have better coverage (greater path density and more crossing paths) in North Africa and Arabia than for southern Africa and the south Atlantic. Path coverage is generally poorer for Love waves than Rayleigh waves, and the number of paths in the region drops off significantly at the shortest periods ( 10 s). Surface wave group velocities for the region are determined using seismic tomography. We use the latest updates of the group velocity maps of Pasyanos (2005). The inversion uses a 1 by 1 equal area grid, and a conjugate gradient method is used to solve for lateral variations in group velocity. A variable smoothness technique is used to improve the resolution of the model. Fig. 2. a) Station map for Africa and Arabia showing stations (marked by red triangles) which have been used to make dispersion measurements. b) Path map of surface wave dispersion measurements, shown for 50 s Rayleigh waves. Yellow circles indicate events, red triangles indicate stations, and blue lines indicate paths.

6 32 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Fig. 3. Group velocities (in km/s) for 20 s and 60 s Rayleigh waves. Slow velocities are indicated in red and fast velocities in blue. Model resolution is also significantly improved over the 2002 study that covered North Africa. Resolutions in well-sampled regions at intermediate periods (20 40 s) approach 1, while the resolution of periods sensitive to mantle depths can be as good as 2. The maps provide group velocities of Love and Rayleigh waves across the region for periods from 7 to 100 s. Surface wave tomography maps for two periods are shown in Fig. 3. The first is for 20 s Rayleigh waves (Fig. 3a), which are generally sensitive to shallow structure. Fast group velocities at this period correspond to thin, oceanic crust, not only in the Atlantic and Indian Oceans, but also in the Western Mediterranean Sea, Red Sea, and Gulf of Aden. Slow group velocities correspond to large sedimentary basins as illustrated with the Eastern Mediterranean, Black Sea, North and South Caspian Basins, and Persian Gulf. Within Africa itself, slow velocities are associated with the Illizi basin in Algeria and Tunis, the Taoudenni Basin in Mauritania and Mali, the basins in Somalia, and the Cuvette Centrale (Central Basin) of the Congo Basin. Offshore basins do not appear as slow features on this map because of the counter-effect of thin underlying crust. At 60 s, the Rayleigh waves are sensitive to deep crust and upper mantle structure. In Fig. 3b, slow velocities are associated with either slow upper mantle velocities or thick crust (i.e. Zagros Mts.). Plate boundaries, in particular, stand out and include both divergent boundaries, such as the mid-atlantic ridge, Indian ridge, Red Sea rift, and Benue Trough, and convergent boundaries, such as the Tethys Belt. Fast velocities stand out under continental cratons and old oceanic crust. In East Africa, the two branches of the East African Rift System can be seen encircling the fast Tanzania Craton. We assess the reliability of the surface wave group velocities with uncertainty maps. Uncertainties are calculated using a bootstrapping approach described in Pasyanos et al. (2001) where multiple inversions are performed on several realizations of the data and a randomized starting model. The uncertainty map for 20 s Rayleigh waves is shown in Fig. 4. As expected from path coverage, group velocities across southern Europe, the

7 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) these are probably the best resolutions we can hope to achieve for deeper mantle structure. 4. Methodology Fig. 4. Uncertainties (in km/s) of group velocities for 20 s Rayleigh waves. Low uncertainties are indicated in red and high uncertainties in blue. Uncertainties are generally low in southern Europe, the Middle East, and East Africa, and higher in oceanic regions. Middle East, and Eastern Africa are very reliably determined. Uncertainties are slightly higher, but still low, in the Mediterranean Sea, within most of continental Africa, and along the mid-atlantic and mid-indian rifts. Results are generally less reliable in the ocean away from the rifts and in Madagascar (where measurements from a new GSN station are not yet included), and the uncertainties are high in regions to the south and southwest of Africa. The resolution of the surface waves is plainly sensitive to path coverage. Resolution will also be limited by the wavelengths of the surface waves. At short periods (b30 s), the wavelengths are less than 1, so we should have high resolution for shallow structure (e.g. sediments, upper crust), at least where we have fairly good path coverage. For intermediate periods (30 60 s), the wavelengths range from 1 to 2 and we will have resolution on this order for deeper structure like crustal thickness and upper mantle velocity. Wavelengths are longer than 2 at long periods (N60 s), indicating that We use the group velocities determined from seismic tomography to estimate crustal structure. At each point in our model, we collect the group velocities (and associated uncertainties) of both Love and Rayleigh waves for periods from 7 to 100 s. We then use a modification of the Pasyanos and Walter (2002) grid search to invert for the best 1-D isotropic velocity model that fits the scaled dispersion data (dispersion scaled by the uncertainty) at each point. The final model is the one that minimizes the misfit of the scaled group velocities. Like the previous grid search methodology, we solve for crustal thickness (initial range of 5 45 km), average velocity of the crystalline crust (range of km/s), and the velocity of the uppermost mantle (range of km/s). All parameters are allowed to expand beyond their initial range in the cascading grid search. The sediment profile is fixed based on the sediment model of Laske and Masters (1997). The modification to the search that we make here is the introduction of a lithospheric lid layer overlying an asthenospheric layer with a high Poisson's ratio (σ =0.29) and accordingly lower shear-wave velocity. This is in contrast to the Poisson's ratio of 0.26 that is used for the crust and mantle lid. Having the asthenospheric layer allows us to fit the long period surface waves that we are unable to fit with a single layer upper mantle. In the grid search, we solve for the thickness of this layer. Starting values for lithospheric thickness are derived from the group velocity lithospheric thickness relationship (Eq. 5) developed in Pasyanos (2005). Layered mantle structure below the lithosphere is represented as perturbations to the ak135 model (Kennett et al., 1995). We emphasize that the primary objective of the modeling is not to determine the exact layered structure of the mantle (which is difficult to derive from surface waves), but rather the overall thickness of high-velocity lithospheric material required to fit the long period group velocities. Uncertainties are calculated by examining the range of models that can fit the data within the uncertainties of the group velocity points. Where uncertainties on the group velocity data are large, a wide range of models can fit the data, whereas when the uncertainties are small, only a relatively narrow range of models can. In regions with good path coverage (discussed earlier) the uncertainties on crustal thickness are about 5 km and uncertainties on velocities are about 0.10 km/s (Pasyanos and Walter, 2002). Uncertainties are significantly higher in regions with poor path coverage.

8 34 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Fig. 5. Examples of grid search for crustal structure from a) West Africa and b) East Africa. In each figure, the panels to the left show the fit of the model (solid lines indicating Rayleigh waves and dashed lines indicating Love waves) to the tomographic models (symbols; triangles for Rayleigh and circles for Love), while the panels to the right indicate the resulting 1-D model. Fig. 5 shows examples of the grid search for seismic structure at two locations in North Africa. In each figure, the panel to the left shows the fit of dispersion predicted by the model to the group velocities derived from seismic tomography, while the panel to the right shows the resulting model. Uncertainties in the group velocities are highly dependent on period and wavetype. The uncertainties are higher for Love waves and at the longest and

9 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) shortest periods, primarily due to poorer path coverage. As illustrated in Fig. 5a for a profile in West Africa, even simple lithospheric models (shown in the right panel) are able to fit the surface waves, although many of the details of the profile such as the sharpness of the Moho discontinuity cannot be resolved by the surface waves. In this case, we find the surface waves fit by a velocity profile of 35 km thick crust with thin sediments (1.5 km), very fast upper mantle velocities (4.67 km/s), and a thick lid (which extends in this case down to 280 km). The velocity model that is found for a region in East Africa looks very different (Fig. 5b). The dispersion curves that we see for this region are slower than those for West Africa, particularly at the longest periods where we see a reduction in the group velocities of the Rayleigh waves at periods longer than about 50 s. Once again, though, we are able to fit the dispersion data with a simple model, although there appears to be some misfit between the input sediment profile and the short period Love waves. Here, we see a thin (25 km thick) crust with 4 km sediments and slower upper mantle velocities. In addition, we find that a thinner lid thickness is necessary in order to fit the observed long period surface wave data. It also appears that some transverse isotropy with v SH Nv SV (not modeled here) might be necessary in order to simultaneously fit both the Love and Rayleigh wave dispersion. 5. Results By assembling the individual 1-D inversions, we have created a 3-D model for Africa and Arabia and use the results to map regional lithospheric structure. Fig. 6a shows a crustal thickness map for the region. The most outstanding features are the significantly thinner crust in oceanic regions, not only the Atlantic and Indian Oceans, but also the Red Sea and Mediterranean Basins (i.e. Western Mediterranean and Ionian Basin). More interesting are the variations within oceanic regions and the African continent. In the oceans, we find several instances of increases in crustal thickness associated with islands such as Ascension, Tristan du Cunha, and Reunion, as well as increases in and around Madagascar and the Mozambique Channel. Within Africa, we find thicker crust (N35 km) in the West African, Congo, and Kalahari cratons, in contrast to moderate thicknesses in the East Saharan Shield (25 35 km). This is primarily controlled by intermediate period ( 30 s) group velocities which indicate faster group velocities in the East Saharan Shield than nearby cratons. In general, we find a relatively gradual increase in crustal thickness from the oceanic edges of the continent to the continental interior. Another notable feature is the crustal thinning (b25 km) in the Benue Trough and its northern and eastern extensions. To the northeast, the model shows a gradual increase along the Arabian Peninsula from the Red Sea (b10 km thickness) across the shield and platform to the Zagros Mts. and Iranian Platform (N45 km). It also shows thick crust in the Turkish Plateau and Eastern European Platform, but thin crust in the extended crust of Western Europe. Along the subduction zone in the eastern Mediterranean, the slow shear wave velocities due to the presence of volatiles and partial melting in the mantle wedge were being interpreted as crustal velocities and the crustal thickness was vastly overestimated. In the case of several nodes from this region, we needed to alter the range of crustal thickness to be less than 45 km. Another interesting indicator of regional tectonics is the map of average crustal shear velocity (Fig. 6b). Here we typically find faster crust in oceanic regions (Atlantic, Mediterranean, Red Sea, etc.) at least in the well-resolved oceanic regions at the edges of the continents. In general, we see a correlation between crustal velocities and the age of the latest thermo-tectonic event, with more recently affected crust (like the extended crust in western Europe) having slower velocities than less-recently affected crust. In particular, we can see slower crust in Cenozoic orogenic regions (Atlas Mts., Zagros Mts., Turkish Iranian Plateau), and Madagascar. It also appears that the Benue Trough has slow crustal velocities. In contrast, cratons have high average crustal velocities. In oceanic regions farther away from the continents, we don't see any coherent pattern, probably due to the poorer path coverage at short periods necessary to robustly estimate this parameter. Shear wave velocities in the upper mantle directly beneath the Moho (Fig. 6c) show dramatic contrasts between the cratons (West African, Congo, Ukraine, Kalahari Craton) that are associated with fast velocities, and the rifts (Dead Sea Rift, Red Sea, Gulf of Aden, East African, mid-atlantic) associated with slow velocities. In comparison to these very slow regions, the Benue Trough and its northern and eastern extensions are only marginally slow. Upper mantle velocities are quite slow near Madagascar and below Reunion Island and the Comores. Another interesting feature are some of the linear marks in Arabia and the south Atlantic. In both cases, these correspond to known volcanic features such as Holocene to recent Arabian volcanism and the Tristan hotspot track. In southern Africa, we find fast upper mantle velocities in the Kalahari Craton, but slower velocities at the southern edge of the continent. There does

10 36 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Fig. 6. Crustal and upper mantle structure of Africa. a) Crustal thickness map (in km), b) Average crustal shear velocity map (in km/s), c) Upper mantle shear velocity (Sn) map (in km/s), and d) Input sedimentary thickness map (in km).

11 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) not seem to be any systematic difference in the uppermost mantle velocities of Archean and post-archean terranes. Among areas designated as hotspots, we don't find any consistent changes in seismic structure in our model. In regions where the hotspots are associated with plate boundaries (such as Afar and Ascension) the effects appear to be strong. In a continental setting, the Afar hotspot results in thinned crust, but in an oceanic setting, the Ascension hotspot has thickened crust. In both instances, we see very slow upper mantle velocities and thinned lithosphere. The Mt. Cameroon hotspot seems to have a lot in common (thinned crust, slow upper mantle velocities, and thinner lithosphere) with the continental Afar hotspot. In hotspots away from plate boundaries (Hoggar, Tibesti, Darfour, Canary, Reunion, Cape Verde), the effect on upper mantle structure appears to be weaker and less consistent. Under most, we find slow Sn velocities, but this effect is much stronger under some hotspots than others. In some regions, the depth extent of faster lid velocities appears to be thinner; in other regions the opposite is seen. In Arabia, we find thin lithosphere in the extension zones of the Red Sea and Gulf of Aden. The thinnest lithosphere and slowest mantle velocities are found near the Afar Triple Junction. Farther to the northeast, we also find thin lithosphere under the Turkish Iranian Plateau. The Arabian Platform has lithospheric thickness in excess of 150 km, but thins toward the Zagros Mts. For completeness, we have also included a plot (Fig. 6d) of the input sediment model of Laske and Masters (1997). 6. Comparisons In this section, we compare our model to other models in the open literature. We focus mainly on continental-scale models, but also make some limited comparisons with models from regional or local studies. We begin with the crust, and then move to the lithospheric mantle. Sublithospheric mantle structure is not discussed, because, as noted previously, our model does not have sufficient resolution to image below the lithosphere. For this study, we define the lithosphere as the seismic lid. Given the simple layer parameterization of our model, the velocity of the lid can vary throughout the model, but everywhere in the model beneath the lid is a lower velocity region that we associate with the asthenosphere Crustal structure Crustal structure in our model overall is consistent with a number of profiles averaged by crustal types (Fig. 2 in Rudnick and Fountain, 1995; Fig. 2 in Mooney et al., 1998; Fig. 12 in Pasyanos and Walter, 2002) which show fast crustal velocities in oceanic regions and slow crustal velocities in orogenic zones (including Paleozoic orogens) and extensional regions (but not active rifts). We compare our crustal thickness map to the velocities from two a priori geophysical models: CRUST2.0 (Bassin et al., 2000) and 3SMAC (Nataf and Ricard, 1996). Like the CRUST5.1 model (Mooney et al., 1998), CRUST2.0 is based on regional studies where available, and on geophysical inference where they are not. Since the regional studies are not available for large portions of Africa, the crustal thickness in the CRUST2.0 model in many parts of Africa is inferred. We find several major differences between the crustal thickness estimates of CRUST2.0 and our model. First, while the crustal thickness in the West African, Congo, and Kalahari Cratons is thick in both models, there is a significant difference in the East Saharan Shield. Our model has crustal thicknesses ranging from km for this region, while CRUST2.0 shows thicknesses between 35 and 45 km, leading to significant differences in the southern Sahara (Niger, Chad, Sudan). Farther to the south, our model shows crustal thinning associated with the Benue Trough and continuing through southern Chad and Sudan into Kenya. While the CRUST2.0 model does not have any crustal thinning associated with these features, a study of crustal structure in southern Sudan and northern Kenya using surface waves also finds thinned crust (Benoit et al., 2006a). Another region showing a large difference is the island of Madagascar. While CRUST2.0 has a crustal thickness of about 40 km for this region, our model finds crustal thicknesses of km. Reviewing a number of studies, de Wit (2003) suggests a continental crust of km below the center of Madagascar that rapidly thins to about km along the east coast, along with a lithospheric thickness ranging from km. As mentioned earlier, there seems to be a contrast in the crustal thickness variations along continental margins. While CRUST2.0 has abrupt boundaries, the crustal thickness model presented here has more gradual crustal thickness variations between oceanic crust and the continental margins. This finding is well-illustrated in Libya, Morocco, and Angola. A comparison of our crustal thickness map to the 3SMAC model yields a better match. Like our study, the 3SMAC Moho model does not have sharp boundaries right at the continental margins. The model finds crustal thickness increases in North Africa and the Congo Craton, but doesn't have any thickening associated with the

12 38 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) Fig. 7. Horizontal cross sections showing the shear-wave velocity variation (in %) from ak135 at a) 100, b) 150, c) 200, and d) 250 km depth. The figures correspond well to the upper mantle values from Ritsema and van Heijst (2000).

13 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) cratons in southern Africa. There is also some thinning associated with the Benue Trough, although the thinning does not extend into the West and Central African rift systems. Finally, the crustal thickness for Madagascar in 3SMAC (25 35 km) is more in line with our study than CRUST2.0. Instead of comparing the crustal structure of our model to the numerous high-resolution models that exist for several regions in our study area (i.e. West Central Africa, Tanzania Craton, Ethiopia/Kenya), we have selected one of the regions (southern Africa) and compare the results for this region. In southern Africa, Nguuri et al. (2001) estimated crustal thickness from the southern Africa seismic experiment using receiver functions. The authors find moderate thickness crust ( km) along the undisturbed Kaapvaal Craton and Zimbabwe Craton interspersed with thicker crust ( km) along the Bushveld Complex and the Limpopo and Namaqua-Natal mobile belts. In our study, we find similar moderate thicknesses in the cratonic nuclei, along with thick crust under the Bushveld Complex and surrounding Okwa Belt, but find very different crustal thickness estimates in the Namaqua-Natal Mobile Belt. This discrepancy can perhaps be explained by the fact that the crustal thicknesses in the Nguuri et al. (2001) model were constructed by using a cratonic velocity structure which, the authors admit, might not be valid for off-craton structure or modified cratonic regions such as these regions Lithospheric mantle structure We also compare our uppermost mantle velocity results to the velocities from the CRUST2.0 and 3SMAC models. Again, the fit to the CRUST2.0 model is rather poor. The model has uniform, extremely fast ( 4.7 km/s) upper mantle velocities throughout the African continent, except along the East African Rift system. In contrast, the fit to the 3SMAC model is rather good. Like our results, this model also predicts moderately slower velocities outside of the extremely fast cratons, although the velocities are even slower in our model. Overall, it seems that the 3SMAC model is the better predictor of seismic structure in Africa than CRUST2.0. In the upper mantle beneath southern Africa, we find fast velocities in the Kalahari Craton, but slower velocities at the southern edge of the continent. While there is some suggestion that upper mantle velocities in the region are uniformly high, at least for P-waves (Wright et al., 2003), there is evidence on variations in the craton. Reporting on a study of Cichowicz and Green (1992), Durrheim and Mooney (1994) state that the Kaapvaal Craton has relatively low velocities (mean 4.3 km/s) for the crust and uppermost mantle down to 80 km depth, consistent with many results from the Kaapvaal seismic project (e.g., de Wit et al., 2004 and references therein). The fast velocities are consistent with previous studies of the region (Zhao et al., 1999; James et al., 2001), with the latter also finding slower velocities in the Cape Fold Belt. A number of global and continental-scale tomographic models have been published, and most of them show similar lithospheric mantle structure beneath Africa and Arabia. Thus, for comparison, we focus on just one model, that from Ritsema and van Heijst (2000), hereafter referred to as RV&H. Fig. 7 shows horizontal slices through our model at 100 km, 150 km, 200 km, and 250 km depth. The depths were chosen to allow a direct comparison to the results of RV&H. At all depths, there is excellent agreement between the two studies, with both having high velocity structures beneath the West African, Congo and Kalahari cratons down to about 250 km depth, and low velocity structures observed beneath the East African, Red Sea, Gulf of Aden, mid- Atlantic and northwestern Indian Ridge. However, there are some differences between the models, particularly in central Africa beneath the Congo Craton, in southern Africa, and in the Zagros Mts. of Iran. In our model, there is no pronounced lithospheric keel under the center part of the Cuvette Centrale and hence Congo Craton. Our model shows thick lithosphere beneath the Archean blocks surrounding the Cuvette Centrale but not directly beneath it, whereas the RV&H model show the thickest lithosphere beneath the Cuvette Centrale. In southern Africa, RV&H find more or less a continuous zone of thick fast lithosphere that extends from the northern edge of the Congo Craton to the southern continental edge. Our model shows the same feature, but the nature of it changes in the vicinity of Namibia, Botswana, and Zimbabwe, where the Kaapvaal Craton abuts with the Congo and Zimbabwe Cratons along the Kheiss and Limpopo Belts. The models also differ in the Zagros Mts. of Iran where RV&H find thick lithosphere extending down to 200 km. This is in contrast to the relatively thin lithosphere found in our model. We have also compared our lithospheric thickness estimates with thermal studies. In Artemieva and Mooney (2001), the authors use heat flow data to estimate temperatures in the upper mantle and use the projected temperatures to derive lithospheric thickness. While heat flow coverage in Africa is sparse, where they have been able to make estimates, the patterns are very similar. Both the seismic and heat flow data suggest thick

14 40 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) lithosphere under the West Africa, Congo, and Kalahari Cratons and thin lithosphere in the East African Rift system, extending to Madagascar. Moreover, both studies find the thickest lithosphere beneath the craton in West Africa. Where the studies differ is in the lithospheric thickness of the Benue Trough and the

15 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) eastern portion of the Saharan Shield. In both cases, the seismic thicknesses are significantly thinner than estimates of the thermal thicknesses, which approach 200 km depth. In each instance, however, the regions are far from the location of any heat flow measurements used in the analysis. 7. Discussion In summary, we have used surface wave group velocities to invert for the lithospheric structure of Africa and Arabia, producing data-driven maps of crustal and lithospheric thickness that, in contrast to other maps of the whole region, are not derived using geophysical inference. Deeper features in the model, as characterized by horizontal slices through the upper mantle, correspond well with other studies of the region. Shallower features like crustal thickness, however, differ significantly from inference-based models. In particular, we find significant increases in crustal thickness associated with the West African Craton, the Congo Craton, Tanzania Craton, and Kalahari Craton. We also find thinning associated with the East, Central and West African Rift Systems. No significant variations in crustal thickness appear to be related to hotspots located away from plate boundaries. Crustal velocities are notably slower in regions of recent tectonism and in orogenic zones both older and more recent. Shallow upper mantle structure shows velocity variations consistent with tectonic structure. Slow upper mantle velocities are found under the Red Sea, Dead Sea Rift, Gulf of Aden, East African Rift, and mid- Atlantic rift. Very fast upper mantle velocities are found beneath the West African Craton, Congo Craton and cratons in eastern and southern Africa (Zimbabwe, Tanzania, Kaapvaal). The Benue Trough appears to be a deeper mantle feature only, with thinned crust and thinned lithosphere, but not particularly anomalous velocities in the upper mantle directly beneath the Moho discontinuity. Lithospheric structure is also well correlated to cratonic structure and is very consistent with the results of RV&H. Our model does not resolve well sublithospheric mantle structure, and is therefore of limited use for addressing questions such as the origin of hotspots. Next we describe variations in lithospheric structure across Africa and Arabia with three vertical depth slices through our model (Fig. 8), and use them to comment in some detail on the structure of the lithosphere under the Archean cratons. Layering in the cross-sections is due to layers of the ak135 model (Kennett et al., 1995), which has discontinuities in the mantle at 150, 190, 225, and 270 km depth. The first cross-section, A A' (Fig. 8b), extends from the north Atlantic to the Indian Ocean, and cuts across the West African, Congo, and Tanzanian cratons. All three cratons have both fast upper mantle velocities and thick lithosphere. However, the Tanzania Craton is noticeably thinned on both ends where the cross-section crosses the two branches of the East African Rift, and it has the shallowest lithospheric keel of the three. The slowest upper mantle velocities along A A' are found at the southeastern portion of the crosssection and are associated with the Comores Island hotspot. The second cross-section (B B', Fig. 8b) extends from the south Atlantic to the Arabian Plate crossing the African Superswell. In this cross-section, thick lithosphere associated with the cratons in southern Africa can be seen, as well as thinned lithosphere and slow upper mantle velocities along the East African Rift. Even away from the rift zone, however, the mantle appears to be slower under the northern portion of the superswell than neighboring regions. At the northeast end, the lithosphere thickens again under the Arabian Platform. The third cross-section (C C', Fig. 8b) runs parallel to cross-section B B', extending from the Atlantic to Eurasia, offset to the northwest of profile B B'. Here, away from the African Superswell, we find lithosphere in the Atlantic that thickens with increasing age towards the coast, and thick lithosphere under the Congo Craton and East Saharan Shield, with significantly thicker lithosphere under the western and northwestern portions of the Congo Craton. Thin lithosphere and slow mantle velocities can be seen under the Red Sea, Dead Sea Rift, and northern portion of the Arabian Plate and Turkish Plateau. The outlines of exposed Archean crust in Africa, which can be used to define the nominal extent of the Archean cratons, are shown in Fig. 8a, superimposed on a slice through our velocity model at 200 km depth. For the most part, the spatial extent of the cratonic keels in our model, identified as regions with thick ( km) Fig. 8. Cross-sections traversing the model. Cross-sections show shear-wave velocity variations (in %) from the ak135 model. a) Map showing the location of cross-sections shown below, along with a depth slice through the model at 200 km. Dashed lines outline the African Superswell (Nyblade and Robinson, 1994). The pink regions show the Archean cratons and the green regions the locations of Cenozoic volcanic fields (Kampunzu and Popoff, 1991). b) Cross-section A A' extending from the north Atlantic Ocean to the Indian Ocean. Values along the profile indicate the shear wave velocities at the top of the mantle. Cross-section B B' extending from the south Atlantic Ocean to the Iranian Plateau. Cross-section C C' extending from the mid Atlantic Ocean to the Caspian Sea. Abbreviations are the same as in Fig. 1.

16 42 M.E. Pasyanos, A.A. Nyblade / Tectonophysics 444 (2007) lithosphere, is consistent with location of mapped Archean crust. There are a few notable differences, however. Firstly, our model shows thick, craton-like lithosphere extending to the southwest of the Kaapvaal Craton offshore. There are few crossing ray paths in this part of our model, and therefore the fast lithospheric structure of the Kaapvaal Craton (on-shore) is most likely being smeared to the southwest off-shore. Thus, the fast lithospheric structure under the southeastern Atlantic Ocean is probably a modeling artifact. Interestingly, our model also shows the fast lithospheric structure of the Zimbabwe Craton extending to the northeast beneath central and northern Mozambique. A similar northeastward extension of the Zimbabwe Craton can be seen in the RV&H model. The Proterozoic mobile belts flanking the northeast side of the Zimbabwe Craton may represent reworked Archean crust, some of which could have been thrust to the west over the margin of the Zimbabwe Craton. Thus, it is not geologically unrealistic to find thick, fast lithosphere under central and northern Mozambique along the edge of the Zimbabwe Craton. A third difference, and perhaps the most important one, between published maps of Archean crust in Africa (Fig. 8a) and the location of thick craton-like lithosphere in our model is the Congo Craton. The extent of the Congo Craton shown in Fig. 8a is inferred. Archean crust is exposed around the periphery of the Congo Craton in the Central African Republic and Cameroon to the north and northwest, in Gabon to the west, and in the southern D.R.C. and Angola to the south and southwest (Cahen et al., 1984). The middle of the so-called Congo Craton is covered by up to 9 km of sediments in the Cuvette Centrale. Most previous tomographic images of the African upper mantle show one large lithospheric keel centered on the Cuvette Centrale, reinforcing the notion of a single, large craton comprising the Archean blocks surrounding the Cuvette Centrale that join together under the Cuvette Centrale (e.g., Grand et al., 1997; Ritsema et al., 1999; Ritsema and van Heijst, 2000; Debayle et al., 2001; Grand, 2002). However, the origin of basin subsidence in the Cuvette Centrale has been linked to a Neoproterozoic failed rift and subsequent thermal relaxation of the lithosphere (Daly et al., 1991; Daly et al., 1992). A Neoproterozoic rift origin for the Cuvette Centrale suggests that the lithosphere under the basin, and consequently the region that is commonly assumed to be the center of the Congo Craton, may not be typical thick, cold, strong Archean lithosphere. Because the Cuvette Centrale is surrounded by Archean blocks, it is quite possible that in many published tomographic images the deep lithospheric keels under these blocks are being smeared together, making the spatial extent of thick lithosphere appear much greater than it really is. The limited resolution of published tomographic images is clearly illustrated by the fact that these images do not distinguish between the Tanzania Craton and the Congo Craton, which are separated by the western branch of the East African rift system. Our model is different in that it does not show a region of thick craton-like lithosphere under the center of the Cuvette Centrale. This can be seen in the depth slice through the model at 200 km (Fig. 8a), or in the vertical section A A' in Fig. 8b. On the other hand, our model does show thick ( km) lithosphere under the Archean blocks to the south, west and northwest of the Cuvette Centrale (C C', Fig. 8b). Consequently, our model suggests that the Congo Craton might not be one large Archean block, but rather a number of smaller Archean blocks that do not join under the sedimentary cover of the Cuvette Centrale. In comparison to estimates of lithospheric thickness for other Archean cratons (West Africa, Tanzania, Kaapvaal and Zimbabwe) from other studies, our results are similar. The thickness of the lithosphere under the cratons in West Africa and southern Africa (Kaapvaal and Zimbabwe) is about km. Beneath the Tanzania Craton, the lithosphere is only km thick. As a final comparison, we point out that our model shows a small region with lithosphere about 200 km thick under the Archean Uweinat block in southern Egypt and northern Sudan extending to the southwest into the region referred to as the East Saharan Shield (ES) on Fig. 1. This region of thick, fast lithosphere can be seen in both the depth slices (Fig. 8a) and the vertical profile C C' in Fig. 8b. This region of thick lithosphere is not apparent on many other published seismic models of African upper mantle. It appears that the African Superswell, which is responsible for the uplift of eastern and southern Africa, is a shallow seismic feature only beneath the northern portion of the superswell. Seismically, we do not see this feature in the south at depths for which we have surface wave sensitivity. Earlier studies have established that no such evidence exists at least to lower mantle depths from both P (Zhao et al., 1999; Simon et al., 2002) and S wave studies (Simon et al., 2003). This is consistent with the lava lamp model (Ladbury, 1999) advocated by Doucouré and de Wit (2003) and supported by other seismic data (Ritsema et al., 1999; Behn et al., 2004). In this model, the superswell has been only episodically active at shallow depths, coinciding with progressing magmatic events in southern Africa during the Jurassic,