Crustal Structure of Active Deformation Zones in Africa: Implications for Global Crustal Processes

The Cenozoic East African rift (EAR), Cameroon Volcanic Line (CVL), and Atlas Mountains formed on the slow‐moving African continent, which last experienced orogeny during the Pan‐African. We synthesize primarily geophysical data to evaluate the role of magmatism in shaping Africa's crust. In young magmatic rift zones, melt and volatiles migrate from the asthenosphere to gas‐rich magma reservoirs at the Moho, altering crustal composition and reducing strength. Within the southernmost Eastern rift, the crust comprises ~20% new magmatic material ponded in the lower crust and intruded as sills and dikes at shallower depths. In the Main Ethiopian Rift, intrusions comprise 30% of the crust below axial zones of dike‐dominated extension. In the incipient rupture zones of the Afar rift, magma intrusions fed from crustal magma chambers beneath segment centers create new columns of mafic crust, as along slow‐spreading ridges. Our comparisons suggest that transitional crust, including seaward dipping sequences, is created as progressively smaller screens of continental crust are heated and weakened by magma intrusion into 15–20 km thick crust. In the 30 Ma Recent CVL, which lacks a hot spot age progression, extensional forces are small, inhibiting the creation and rise of magma into the crust. In the Atlas orogen, localized magmatism follows the strike of the Atlas Mountains from the Canary Islands hot spot toward the Alboran Sea. CVL and Atlas magmatism has had minimal impact on crustal structure. Our syntheses show that magma and volatiles are migrating from the asthenosphere through the plates, modifying rheology, and contributing significantly to global carbon and water fluxes.


Overview
The geological record of the African continent spans three quarters of Earth history. Key events occurred in the Archaean when the West African, Congo, Kaapvaal, Zimbabwe, and Tanzanian cratons formed, followed by Palaeoproterozoic accretion of cratons. Unusually, only the northern and northwestern margins of the African plate have been involved in orogeny since the end Pan-African (~500 Ma), when a Himalayan-scale collision formed along eastern Africa and orogenies ringed the Zimbabwe-Kaapvaal-Congo cratons of southern and western Africa (e.g., Muhongo & Lenoir, 1994;Shackleton, 1986;Van Hinsbergen et al., 2011). Each of these processes has left an indelible signature on Africa's crust and contributed to the formation and destruction of continental topography.
Excluding the collisional belts of NW Africa (e.g., , the predominant tectonic processes since the Pan-African have been magmatism and extension that have increased the surface area and volume of African crust, as outlined in this paper. Southeastern Africa was affected by rifting and flood magmatism at 184-179 Ma (Ferrar-Karroo flood basalts) prior to the separation of Africa and Antarctica (e.g., Duncan et al., 1997), and large sectors of western Africa experienced flood magmatism at 137-127 Ma (Etendeka-Parana flood basalts) as the south Atlantic opened (e.g., Mohriak et al., 2012;Turner et al., 1994). Central, southern, and eastern Africa were affected by the Mesozoic rifting and breakup of Gondwana that left scars such as the Central African rift system and the widespread extensional basins from Somalia to S. Africa (e.g., Burke, 1996;Veevers et al., 1994). Over the past~45 Ma, Africa again has experienced kimberlitic to flood magmatism and plateau uplift, followed by diachronous rifting from Egypt south to South Africa and southwest through Botswana (e.g., Ebinger & Scholz, 2012) (Figure 1). The Tertiary rift basins and magmatic provinces include all stages and styles in the evolution of cratonic rift zones: incipient rifting in the Okavango to incipient seafloor spreading in the Afar depression, amagmatic sectors in the Western and Southwestern rift to the large igneous province in Ethiopia-Yemen, and sectors that transect Paleozoic lithosphere of variable thickness and composition (e.g., Corti, 2009;Ebinger & Scholz, 2012;Rooney et al., 2017) (Figures 1-4). Oligocene-Recent magmatism with little extension modifies crustal structure beneath West-Central Africa along the Cameroon Volcanic Line (CVL) (Figure 3). Perhaps owing to the very slow movement of the African plate since collision with Europe initiated at~40 Ma, domal uplift and magmatism without faulting also characterize northern Africa (e.g., Hoggar-Tibesti; Bond, 1978;Brown & Girdler, 1980;Rosenbaum & Lister, 2004) (Figure 1).

10.1002/2017TC004526
properties, including resistivity versus depth and azimuthal anisotropy of electrical conductivity (e.g., Selway et al., 2014;Whaler & Hautot, 2006). Magnetotelluric data are more sensitive to the presence of melt than seismic methods. Although models of gravity anomalies are highly nonunique, when constrained or jointly inverted with independent 2-D and 3-D data sets, the spatial patterns of gravity anomalies provide additional information on 3-D density variations (e.g., Roecker et al., 2017;Tiberi et al., 2005). The combination of magnetotelluric, seismic, and gravity data in separate and joint inversions enables tighter constraints on material properties and, in some instances, provides new insights into the distinction between magmatic fluids, aqueous fluids, and volatiles within the crust and their role in crustal deformation.
This contribution synthesizes our current understanding of crustal modification in tectonically active cratonic rift zones and mountain belts and lays out a road map for future studies of African rift and orogenic zones. Several large-scale crust and mantle imaging experiments have acquired promising data sets in incipient and weakly extended, magma-poor rift sectors (e.g., Gao et al., 2013;Hodgson et al., 2017;Shillington et al., 2016), but a synthesis of crustal structure of early stage rifts is premature. Our focus is to characterize crustal structure in zones of active deformation in the East African rift, Cameroon Volcanic Line (CVL), and Atlas Mountains. The review of along-axis variations in the East African rift affords the opportunity to compare and contrast magma-rich rift sectors, providing insights on the evolution of the continental crust in response to stretching and magmatism. This synthesis and comparison offers new insights into the nature of "transitional" crust beneath late-stage rifts (e.g., Fuis & Mooney, 1990;Persaud et al., 2016) and passive margins worldwide (e.g., Skogseid et al., 2000;Van Avendonk et al., 2009). Comparison and contrast with the magmatically modified crust beneath the Atlas Mountains and CVL add to our understanding of fluid and volatile migration through the African crust.
Section 2 provides a broad-brush summary of the geodynamical context of active deformation in Africa. Section 3 outlines the role of magma intrusion during the first 5-7 Myr of rifting in cratonic lithosphere, using examples from the southern sector of the Eastern (Gregory) rift. Sections 4 and 5 address the role of magmatism and thinning in the formation of crust transitional between continental and oceanic. The Horn of Africa is one of few areas worldwide where this transition is occurring, and comparisons with the well-studied Salton Trough and northern Gulf of California (e.g., Fuis & Mooney, 1990;Persaud et al., 2016) inform our understanding of magmatic modification along passive margins worldwide. In section 6, we journey across the continent, where warm asthenosphere rises beneath the eastern edge of the Congo craton and the southern margin of a preexisting Mesozoic rift zone. This linear belt of Cenozoic eruptive volcanic centers, neither the CVL, exhibits the age progression, mechanical stretching, magmatic modification displayed in parts of East Africa nor along linear volcanic tracks in oceanic plates. Here too, volatiles have likely played a central role in magmatic modification of Africa's crust. Section 7 presents a review of crustal structure and its relation to mantle dynamics beneath the Atlas Mountains where compressional tectonics may inhibit upward migration of melt, leading to the localized, linear trend of magmatism linked instead to the Canary Island hot spot. This integration of constraints on crustal structure beneath zones of active rifting and orogenesis in Africa shows that magma and volatiles are migrating from the asthenosphere through the plates, modifying lithospheric rheology and significantly contributing to global carbon and water fluxes.

Geodynamic Settings
The slow-moving African continent moves at 3-7 mm y À1 NW to WNW and collides with the Eurasian plate (e.g., Serpelloni et al., 2007). Africa has unusually high elevation, in large part owing to the broad plateau uplifts associated with mantle plumes and volcanic construction in the large igneous provinces of the East African Rift System (EARS) and the CVL and the unextended Hoggar and Tibesti volcanic uplifts ( Figure 1). Extension effectively increases the surface area of continental landmasses and creates continental platforms. Magmatism during rifting creates new continental crust through dike and sill intrusion and surface construction, but the rates and volumes of crustal accretion are poorly constrained Thybo & Artemieva, 2013), in part motivating this comparative study.

East African Rift System
Active faulting and magmatism occur across a large part of the African continent between the Horn of Africa southwest to the Okavango region of Botswana and in a second diffuse arm that continues from southwestern Ethiopia to southeastern Mozambique, including offshore regions of the Indian Ocean, such as the Davie Ridge (Figures 1 and 2). The southern Red Sea, Main Ethiopian, and Eastern, Western, and Southwestern rift systems have developed atop broad topographic plateaux, whereas the Malawi rift and its southeastward continuation transect low elevation regions in Mozambique ( Figure 2).
The broad plateaus, their corresponding negative Bouguer gravity anomalies, low upper mantle seismic velocities, and the large volume and geochemistry of eruptive volcanic products have been cited as evidence for one or more mantle plumes beneath sections, or all, of the uplifted zones of Africa (e.g., Ebinger & Sleep, 1998;Marty & Yirgu, 1996;Nyblade & Robinson, 1994;Sengör & Burke, 1978). Global and local tomography and geochemical studies reveal low-velocity zones caused by one or a combination of elevated temperatures and the presence of melt in the asthenosphere (e.g., Adams et al., 2012;Debayle et al., 2001;Fishwick & Bastow, 2011;O'Donnell et al., 2013;Ritsema et al., 1999;Rooney et al., 2011;Simmons et al., 2007;Weeraratne et al., 2003). Although gaps remain in our knowledge of the Western rift and beneath the Indian Ocean, the lowest P and S wave velocity regions underlie the Main Ethiopian Rift (MER) and the Eastern rift zone and the isolated volcanic provinces of Tectonics 10.1002/2017TC004526 the Western rift (e.g., Adams et al., 2012;Bastow et al., 2008;Fishwick, 2010;O'Donnell et al., 2013). The Ethiopia-Yemen and East African Plateaux are separated by an~300 km wide topographic depression that is underlain by crust stretched during Mesozoic rifting, allowing the possibility that the two plateaux are actually one uplifted region extending from southern Africa to the Red Sea: the African superplume province (e.g., Kendall & Lithgow-Bertelloni, 2016;Nyblade & Robinson, 1994;Ritsema et al., 1999).
Thermal-mechanical modeling indicates that widespread Mesozoic rift zones spanning the breadth and length of Africa would have been actively subsiding and underlain by thinned lithosphere at the onset of flood magmatism. At~40 Ma the earliest flood basalts were erupted in southwestern Ethiopia and northern Kenya, suggesting that these thinned and heated regions may have been inherently weaker than surrounding regions (e.g., Hendrie et al., 1994;Morley et al., 1999) or may have ponded anomalously hot mantle material susceptible to decompression melting (e.g., Ebinger & Sleep, 1998). Edge-driven convection may enhance magma production and lithospheric heating at the edges of the deeply rooted Tanzania and Congo cratons where kimberlitic and carbonatitic magmatism has occurred since~40 Ma (e.g., Ebinger & Sleep, 1998;King & Anderson, 1998;King & Ritsema, 2000). The geochemistry of Eocene-Recent eruptive volcanic products points to a mantle plume origin for the Ethiopia-Yemen flood basalt sequences, but the East African plateau region south of Ethiopia shows more spatial variability (e.g., Chakrabarti et al., 2009;Furman, 2007, Furman et al., 2006Halldórsson et al., 2014;Pik et al., 2006;Rooney et al., 2011).

Cameroon Volcanic Line
The CVL formed between the northern edge of the deeply rooted Congo craton and the Benue trough, a highly extended Cretaceous rift system that connected to basins in eastern Sudan and Kenya (e.g., Benkhelil, 1989;Fairhead & Binks, 1991;Poudjom Djomani et al., 1997) (Figure 3). The CVL straddles the continent-ocean boundary ( Figure 3). Intriguingly, 42 Ma to Recent volcanism along the CVL displays no clear age progression (e.g., Fitton, 1980;Halliday et al., 1990;Marzoli et al., 2000;Nkouathio et al., 2008), an observation that has prompted numerous workers to explore beyond the classic plate-plume hypothesis for hot spot development to explain the CVL. Alternatives have included decompression melting beneath reactivated shear zones in the lithosphere (e.g., Fairhead, 1988;Fairhead & Binks, 1991;Freeth, 1979;Moreau et al., 1987), small-scale upper mantle convection that may advect mantle lithosphere (e.g., King & Anderson, 1998;King & Ritsema, 2000), and delamination (e.g., De Plaen et al., 2014;Fourel et al., 2013). It has also been suggested that lateral flow of buoyant asthenosphere, beneath continental lithosphere thinned extensively during Mesozoic rifting, may now be contributing to the younger volcanism along the line (Ebinger & Sleep, 1998). In support of this hypothesis, Perez-Gussinyé et al. (2009) constrained lithospheric strength via study of effective elastic plate thickness (Te) across the African continent using coherence analysis of topography and Bouguer anomaly data. Their study revealed corridors of relatively weak lithosphere that continue across the African continent from the Afar region to Cameroon, where Te is also depressed in comparison to the surrounding cratons.

Atlas Mountains
The Atlas Mountains of Morocco are a 2,000 km long intracontinental compressional belt that strikes ENE from Morocco into Algeria and Tunisia (Figures 1, 3, and 4). The structure was formed by reactivation of Triassic-Jurassic age normal faults that originally formed during the opening of the North Atlantic, followed by the Cenozoic collision of Africa with Eurasia (e.g., Gomez et al., 2000;Pique et al., 2002). Unlike some of the other orogenic belts around the Mediterranean, the origin of the topography of the Atlas is not directly linked to slab rollback (e.g., Dewey et al., 1989;Faccenna et al., 2004;Wortel & Spakman, 2000). The unusually high topography, modest tectonic shortening, thin lithosphere, and localized alkali volcanism may be explained by upwelling of a hot mantle anomaly linked to the Canary island hot spot (e.g., Duggen et al., 2009;Miller et al., 2015;.

Embryonic Rifting of Cratonic Lithosphere and the Role of Volatiles
A long-standing question in plate tectonics involves the initiation of rifting in strong cratonic lithosphere, where the inherent strength of the plate is greater than the far-field and gravitational potential energy forces available (e.g., Bott, 1990;Flesch & Bendick, 2012;Stamps et al., 2014). Where magma is generated, the buoyancy forces of magma add to the tectonic stress, and dike intrusion may accommodate extension at one eighth the force required to overcome friction along a fault (e.g., Bialas et al., 2010;Buck, 2004). Yet only very small melt volumes can be generated beneath thick lithosphere; magma-assisted rifting is unlikely in lithosphere greater than~100 km thick, unless the upper mantle is anomalously hot, carbonated, and hydrated (e.g., Dixon et al., 2008;Turner et al., 1994). The <7 Myr old Eastern rift in southern Kenya and northern Tanzania provides a unique opportunity to address this paradigm: faulting and magmatism occur in deeply rooted lithosphere, and crust and mantle xenoliths constrain compositions ( Figure 5). This region is one of the most seismically active regions in East Africa, and both teleseismically and locally detected earthquakes span the entire crustal thickness (e.g., Albaric et al., 2014;Craig et al., 2011;Lee et al., 2016;Yang & Chen, 2010) ( Figure 4).
The Eastern (Gregory) rift marks the divergent plate boundary between the slowly opening (≤ 6 mm/y) Nubia and Somalia plates (e.g., Birhanu et al., 2016;Saria et al., 2014) (Figures 4 and 5). The Eastern rift splays into a broad zone of~80 km long, seismically active faults, and isolated eruptive centers at its southern termination in northern Tanzania (Figures 4 and 5). The southern sector of the Eastern rift developed after~6 Ma, based on 40 Ar-39 Ar dating of volcanic samples and carbon dating of abundant fossil remains in the Olduvai and Natron basins (e.g., Ashley et al., 2009;Mana et al., 2012Mana et al., , 2015. Sedimentary basin thicknesses increase to almost 10 km where the East African rift overprints Mesozoic rift basins in northern Kenya (e.g., Hendrie et al., 1994).
Although heat flow data show little-to-no elevation of geotherms beneath the 1.5-2 km rift flanks, magmatic systems within basins have locally elevated geotherms (e.g., Nyblade et al., 1990;Simiyu, 2010;Wheildon et al., 1994). The geochemistry and petrology of mantle xenoliths indicate that the lithosphere beneath the craton (west), rift, and Pan-African belt (east) has been metasomatized in large part by magmatic fluids and volatiles (e.g., Aulbach et al., 2011;Mattsson et al., 2013;Rudnick et al., 1993). Uniquely, voluminous CO 2 emissions derived from magma degassing occur along the seismically active border fault systems bounding basins  at distances >20 km from the active carbonatitic volcano, Oldoinyo Lengai ( Figure 5). The volume of fault zone degassing along faults and from volcanoes may be 11% of the global CO 2 budget . The localized metasomatism, magma intrusion, and magma degassing may have weakened the cratonic lithosphere to enable rift initiation at the edge of the deeply rooted Tanzania craton. Crustal xenoliths from Archaean crust (west) and Pan-African crust (east) are mafic granulites (Jones et al., 1983;Mansur et al., 2014) (Figure 5). Yield strength envelopes for granulitic lower crust indicate that it is too weak to explain the unusual lower crustal seismicity in this rift sector (e.g., Wang et al., 2012;Weinstein et al., 2017).

Crustal Velocities, Thinning, and Evidence for Intrusions
Seismic refraction/wide-angle reflection, arrival time and mantle tomography, MT, and gravity data acquired as part of the KRISP94 (Kenya Rift International Seismic Project) with profile locations shown in Figure 5 provide details of Archaean and Late Proterozoic crust, as well as rift structure. Data from two temporary seismic arrays spanning three rift segments of the Eastern rift, including the active Oldoinyo Lengai volcano, provide Tectonics 10.1002/2017TC004526 critical insights into magmatic modification and crustal stretching processes within the faulted basins and beneath the numerous eruptive volcanic centers (Ibs-von Seht et al., 2001;Plasman et al., 2017;Roecker et al., 2017) (Figures 2b and 5).

10.1002/2017TC004526
between upper and middle crust is marked by a velocity increase to~6.5 km/s (Birt et al., 1997;Keller et al., 1994). Lower crustal velocities are variable but 6.7-6.9 km/s on average. Last et al. (1997) and Dugda et al. (2005) analyzed receiver functions from widely spaced stations spanning the Archaean-Pan-African suture in the southernmost (Tanzania) and central (Kenya) sectors of the Eastern rift (Figures 2 and 5). Last et al. (1997) find no change in crustal thickness crossing the Archaean-Proterozoic boundary through the central Manyara rift. Dugda et al. (2005) find slightly thicker crust in the Pan-African belt east of the rift (39-42 km) than west of the rift beneath the Tanzania craton (37-38 km), and normal Poisson's ratios of 0.24-0.27, indicating little or no crustal thinning or magmatic modification outside the rift zones.
Geophysical data from the Magadi-Natron-Manyara sector of the Eastern rift reveal crustal thinning and magma intrusion across the rift zone and low-velocity, high V P /V S bodies beneath some volcanoes. Moho depth determined from receiver functions and the KRISP94 profile varies between 28 and 41 km beneath the rift valley and rift shoulders, respectively (Figures 2b, 6a, and 6b). The thickest crust (41 km) underlies the Crater Highlands, a wedge of uplifted basement, and <6 Myr volcanic constructs (Plasman et al., 2017). The V P /V S ratio depicts a lateral variability, with zones of high V P /V S strongly correlated with the location of volcanic edifices. Unusually high lower crustal velocities of 7.1-7.2 km/s in the lowermost crust beneath the Magadi basin have been interpreted as evidence for cooled gabbroic crustal intrusions (e.g., Thybo et al., 2000) or cumulates (Christensen & Mooney, 1995). Beneath the Natron basin, Plasman et al. (2017) find local low-velocity zones with a strong reflective signature, and high V P /V S (1.83-1.98) in the lower crust, interpreted as magma bodies. Based on wide-angle reflection-refraction studies, the mantle reflection, PmP, is high amplitude but only at low frequencies. At high frequencies, reverberative phases are observed and modeled as an approximately 4 km thick series of mafic sills (Thybo et al., 2000).
At a greater depth of~60 km, an intramantle interface roughly mimics the step-like and symmetrically outward dipping geometry of the Moho (Plasman et al., 2017). Intramantle reflectors were also detected in refraction/wide-angle reflection profiling in the northern part of the Eastern rift Keller et al., 1994), and in the Main Ethiopian Rift, as outlined in section 4. A narrow zone of thinned mantle lithosphere is imaged directly beneath the fault bounded rift basins:~90 km thick versus >125 km outside the rift zone (Green et al., 1991;Ritsema et al., 1999), consistent with P-T conditions of eruptive volcanic products (Macdonald, 1994;Mechie, Fuchs, et al., 1994). Along the length of the rift, velocities in the mantle just below the Moho from Pn analyses are 7.5-7.7 km/s, significantly lower than values beneath the flanks and stable interior: 8.0-8.1 km/s (Maguire et al., 1994;Mechie, Keller, et al., 1994). The magnitude of the velocity decrease is 0.3 km/s larger than predicted from higher geothermal gradients, suggesting that partial melt within the upper mantle also contributes to the observed velocity reduction beneath the rift (Mechie, Fuchs, et al., 1994).
If we assume an initial crustal thickness of 40 km, the~28 km thick crust beneath the Natron and Magadi basins crustal thickness estimated from wide-angle refraction/reflection studies and receiver function studies indicates~30% extension achieved in <7 Myr. The volume increase from magma intrusion may bias crustal stretching estimates to smaller values.

Crustal Magmatic Plumbing Systems
KRISP-94 MT experiments crossing the Proterozoic suture, the Magadi rift basin and flanks, and the off-rift Quaternary-Recent Chyulu Hills eruptive centers reveal not only the shallow basin structure but also several high conductivity zones interpreted as active and ancient intrusives. Using a 2-D inversion, Simpson (2000) report high conductivity lower crust, and Meju and Sakkas (2007) use a 3-D inversion to image narrow high conductivity zones rising to~8 km subsurface beneath the central rift and beneath the Chyulu hills. Seismic tomography models reveal similar patterns. Tomography models of local earthquake P and S wave arrival times indicate a~10 km wide, linear low-velocity zone beneath the central Magadi basin, which may be a zone of fractured or intruded rock (Ibs-von Seht et al., 2001) (Figure 6c). For the Magadi-Natron-Manyara sector, Roecker et al. (2017) jointly inverted ambient noise, arrival time, and gravity data to image V P , V S , and V P /V S in the crust shallower than~25 km. The most significant features are (1) low P and S wave speeds underlying the rift zone; (2) a relatively high wave speed,~10 km wide tabular feature located along the western edge of the Natron and Manyara rifts; and (3)

10.1002/2017TC004526
V P /V S to pore fluid compressibility, Roecker et al. (2017) suggest that the low V P /V S values that correlate with zones of active CO 2 degassing at the surface are caused by the volcanic plumbing in the upper crust being suffused by a gaseous CO 2 above a deeper, crystalline mush with high V P /V S , and high reflectivity in some areas (Plasman et al., 2017).
In a detailed study of the off-axis Chyulu hills shield cone province ( Figure 5), Sakkas et al. (2002) find steeply dipping high conductivity zones at depths 1-8 km and 9-18 km, interpreted as magma intrusions. Similarly, in a MT study of the Baringo rift segment to the north, Hautot et al. (2000) suggest that Mesozoic sediments underlie a 4 km-thick African rift sequence and interpret a high-density magma intrusion with top no shallower than 6 km. Within the southern Natron basin, pancake-shaped zones of persistent M L ≤ 4.5 earthquakes above the inferred lower crustal crystal-rich mush zone at depths of 6-11 km below sea level may be stacked sills, with the variable fractionation depths feeding the wide range of magma compositions at Oldoinyo Lengai and the adjacent monogenetic cone complex Weinstein et al., 2017). Guth (2016) provides coarse estimates of eruptive magma volumes both from map data and from seismically imaged basin fill, which combine to 310,000 km 3 . Because her work includes only basins in Kenya and ignores the Eastern rift in northern Tanzania and southwestern Ethiopia, we extrapolate estimates of intrusive volumes to the same region. For a conservative estimate of intrusive volumes, we assume 20% new magmatic material (underplate and intrusions into the central rift zone, 8 km thick) for a 70 km wide rift zone that is 600 km long; intrusive volumes are at least 336,000 km 3 . The total volume is at least 43,000 km 3 per million years, assuming all sectors are 15 Myr old (average age of Eastern rift).

Constraints on Lower Crustal Deformation in Early Stage Rifts
Although lower crustal earthquakes occur in magmatic and amagmatic rift basins in Africa, we focus on lower crustal earthquakes in magmatic rift sectors. The Magadi, Natron, and Manyara basins are all associated with earthquakes in the lower crust, as relocated from local seismic networks (Albaric et al., 2014;Ibs-von Seht et al., 2001;Mulibo & Nyblade, 2009;Roecker et al., 2017) and from teleseismic studies (Craig et al., 2011;Shudofsky et al., 1987;Yang & Chen, 2010). Within the Natron basin, precisely relocated earthquakes occur at depths of 20-27 km along a projection of the border fault to the base of the crust, consistent with thermomechanical models of basin and flank morphology and gravity anomalies (e.g., Sippel et al., 2017). Volumetrically significant gas fluxes measured near these deeply penetrating fault zones bounding basins indicate that the border faults serve as conduits for CO 2 exsolved from active magma intrusions . Lower crustal seismicity also occurs around lower crustal mush zones, and the inferred sill complexes where high fluid pressures may induce hydraulic fracture (Weinstein et al., 2017).
Laboratory experiments simulating lower crustal conditions demonstrate that granulite is weak, particularly if the rock contains a small weight percent of water (Wang et al., 2012). Although cooled magmatic underplate or cumulates may locally strengthen the lower crust, the high V P /V S ratios in the lower crust and CO 2 degassing along faults that penetrate to the lower crust indicate active intrusion processes and locally hot, weak lower crust (Plasman et al., 2017;Roecker et al., 2017). A strong, mafic lower crust, therefore, seems an unlikely explanation for widespread lower crustal seismicity in this sector of the Eastern rift. Instead, slip along border faults that penetrate to lower crustal levels, locally high pore fluid pressures, as indicated by volatile degassing, and active magma intrusion are explanations for the unusual lower crustal earthquakes in magmatically active early stage rift zones.

Summary: Early Stage Magmatic Rift Zones
Half-graben basins bounded by steep border faults that penetrate to lower crustal levels characterize the <7 Myr southern sector of the Eastern (Gregory) rift. Seismic, gravity, MT, seismicity, and xenolith studies show widespread magmatic underplating, with high levels of magmatic CO 2 degassing along border faults, as well as through active volcanoes. Although much of the area was affected by magmatism between 6 and 1 Ma, during the past~1 Ma, eruptive centers have formed within the central Magadi and Natron basins. Tomographic imaging and seismicity indicate that the central volcanoes and monogenetic cones are sourced by one or more lower crustal magma chamber(s) overlain by stacked sills that may explain the wide range of lava compositions. Dike and sill intrusions have reached 8-10 km subsurface, in patterns similar to the Main Ethiopian Rift outlined below. Sharp contrasts in velocity and V P /V S between hanging wall and footwall suggest that magmatic modification has localized to the faulted basin area.

Magma-Rich Continental Rifting-The Main Ethiopian Rift
The Horn of Africa captures the combined processes of crustal stretching and intrusion at various stages of continental breakup. The Ethiopian rift is the northernmost sector of the EARS, which is the incipient boundary between the Nubia and Somalia plates (Figures 2a and 7). It encompasses early to midstage continental rift sectors, and it captures the transition from deformation along 50-100 km long border fault systems to zones of magma intrusion. Here we focus our synthesis on how geophysics and petrology are providing improved understanding of where and how magma migrates and ponds in the crust during progressive rift development. Some of these patterns mimic along-strike patterns noted in the <7 Myr old Natron-Magadi rift sectors (section 3) and highlight strain and structural changes associated with the rise of magma to upper crustal levels.
The Main Ethiopian Rift (MER) formed within the Precambrian metamorphic crustal basement of the Pan-African orogenic belt that exhibits N-S to NNE-SSW suture zones (Berhe, 1990;Vail, 1983) and NW-SE oriented strike-slip faults in Ethiopia (Brown, 1970;Purcell, 1976). The central and northern sectors of the MER developed between 18 and 10 Ma, respectively (e.g., Rooney et al., 2013;Woldegabriel et al., 1990;Wolfenden et al., 2004). Between 18 and 12 Ma, large offset border faults commonly marked by chains of eruptive centers formed along one or both sides of the rift (e.g., Chernet et al., 1998;Rooney et al., 2014) (Figure 7). South of~7.5°N, seismically active faults in a broad depression effectively link the~30 Myr Eastern rift to the MER, across a broad zone of subparallel and time-transgressive basins (e.g., Ebinger et al., 2000;Morley et al., 1992). Further north, within the basin bounded by Miocene border faults, the rift floor is segmented into a series of Quaternary-to-Recent aligned volcanic centers cut by small offset normal fault systems, referred to as magmatic segments (Corti, 2008(Corti, , 2009Ebinger & Casey, 2001;Mazzarini et al., 2013) (Figure 7). The magmatic segments are the locus of magma intrusion and are underlain by high-velocity zones interpreted as gabbroic intrusions, as outlined below ( Figure 8). GPS data show that the axial volcanic segments are the locus of current extension (Bilham et al., 1999), with the surface geology suggesting that this has been the case since the Quaternary (Ebinger & Casey, 2001). Plate kinematic models constrained by marine magnetic anomaly patterns and geodetic data require that some extension across the MER took place by about 16 Ma (DeMets & Merkouriev, 2016).
As outlined below, crustal structure in Ethiopia is constrained using receiver functions, wide-angle controlled source seismology, MT, and inversion of gravity data. Combined, these results provide important clues to extensional processes that modify the crust during rifting. Stuart et al. (2006) and Hammond et al. (2011) analyzed receiver functions from broadband stations in the Ethiopian rift, building on results of earlier studies of data from widely spaced networks (Dugda et al., 2005;Dugda & Nyblade, 2006) (Figures 2a and 5). The EAGLE controlled source wide-angle experiment constrained across-rift and along-axis variations in crustal structure, as well as the uplifted flanks of the Ethiopian flood basalt province (Figures 8, 9, and 10a). EAGLE comprised two~400 km long profiles: one cross rift ( Figure 9) and one rift axial profile . A dense network of stations was deployed over the intersection of the two profiles to provide 3-D images of the along-axis segmentation patterns (Keranen et al., 2004;Mackenzie et al., 2005;Maguire et al., 2006) (Figure 7). Figure 10 compares receiver function results projected onto the line of controlled source and MT profiles to facilitate comparisons of crustal thickness estimates from seismic tomography, reflection, and MT results. Crustal thicknesses are similar to those determined using receiver functions, but absolute determination of V P and variations thereof provides further insights.

Crustal Velocities, Thinning, and Evidence for Intrusion
The crust beneath the Somalian plate east of the rift is 38-40 km thick (Figures 2a and 7). The crust on the Nubian plate west of the rift is of similar thickness but thickens to 41-43 km between~9°N and 12°N. MT data show significant differences between the lowermost crust beneath the plateaux on either side of the rift (Figure 9): thicker and resistive beneath the Somalian Plateau, thinner with higher conductivity beneath the Ethiopian Plateau in agreement with the interpretation of underplate there (Mackenzie et al., 2005;Rooney et al., 2017). This region of thicker crust lies at the northern edge of an off-rift upper mantle lowvelocity structure imaged by traveltime tomography (e.g., Bastow et al., 2008Bastow et al., , 2005Hammond et al., 2013). At the surface, this upper mantle low-velocity zone and thicker crust on the northwestern rift shoulder is   (Figure 7). Magmatic underplating during the early stages of rifting may have thickened and strengthened crust to the north and localized later strain and magmatism (e.g., Keranen & Klemperer, 2008;Kim et al., 2012). It should be borne in mind that receiver function Moho depth estimates for the Ethiopian Plateau are shallower than those inferred by wide-angle reflection analysis: receiver functions appear sensitive primarily to the top, not the bottom of the 8-12 km thick mafic layer(s) in the lower crust (e.g., Stuart et al., 2006).
Along the rift axis, crustal thickness varies from around 38 km in the south to 30 km in the north, with most of the change in Moho depth occurring just south of the Boset-Kone magmatic segment where the rift floor drops in elevation into the Afar depression (Figures 2a and 7). The change in crustal thickness from~40 km beneath the rift shoulders to 30 km in the rift implies a minimum stretching factor of 1.5 (28 km) or considerably less than the stretching factor~2 estimated by plate kinematic models (42 km;DeMets & Merkouriev, 2016). Stretching factors estimated from crustal thickness alone will underestimate total extension, as outlined below.
The crust is interpreted to be slightly more mafic (V P /V S > 1.85) beneath the Ethiopian Plateau than the Somalian Plateau (V P /V S ≈1.80) (Figures 2a and  7c). This could either be due to Oligocene to Recent magmatic activity or different prerift crustal compositions. Regions of volcanism inside and on the side of the rift (e.g., Silti Debre Zeit Fault zone, SDFZ) are characterized by thinned crust and V P /V S > 2.0, indicative of partial melt within the crust (Figure 7c). Seismic, MT, gravity, and petrological data image or require significant magma volumes, particularly in the lower crust of these magmatic rift zones (Cornwell et al., 2006;Hammond et al., 2011;Keir et al., 2011aKeir et al., , 2011bWhaler & Hautot, 2006). For example, elevated V P /V S ratios in regions of thinning likely indicate that both solidified and still partially molten intrusions contribute to extension .
Using spatial variations in V P imaged by wide-angle controlled source seismology, it has been possible to infer spatial variations in amount of intrusion. Along the rift, V P in the lower crust ranges from 6.6 to 7.2 km/s, peaking at 7.1-7.2 km/s in the lowermost crust between Gedemsa and Fentale volcanoes in the central part of the MER (Figures 7 and 9). Elevated wave speeds of >6.8 km/s have been interpreted as evidence for gabbroic crustal intrusions (e.g., Keranen et al., 2004;Mackenzie et al., 2005;Maguire et al., 2006). In the upper crust, 3-D controlled source tomography shows significant variation in V P , with discrete high V P (>6.4 km/s) zones beneath the magmatic segments (Figures 8 and 9). The >7 km/s velocities and high densities inferred from forward and inverse models of gravity data indicate that the magmatic segments are underlain by gabbroic material (e.g., Cornwell et al., 2006;Mahatsente et al., 1999;Tiberi et al., 2005). The observed pattern of segmented high-velocity zones was reproduced by gabbroic bodies that rise to 10 km below the surface, indicating that 30% of the crust beneath the magmatic segments is new igneous material (Keranen et al., 2004).
The observations corroborate the surface geology evidence that axial magma intrusion, likely in the form of gabbroic cumulates from crustal magma chambers, fed sheeted dykes. The dikes, as well as faults above the dikes, accommodate most of the deformation across the rift, based on time-averaged and active deformation patterns (Casey et al., 2006;Keir et al., 2015). New GPS data confirm that active deformation occurs across the Ethiopian Plateau (Birhanu et al., 2016), which hosts aligned chains of Quaternary eruptive centers, highconductivity lower crust, and lower crustal seismicity (Keir et al., 2011a(Keir et al., , 2011b) (e.g., Figure 10).

Crustal Magma Plumbing Systems
The interpretation of gravity and electrical conductivity anomalies has been used to construct an ever more detailed picture of the crustal magma plumbing system accommodating extension during breakup and  (Figure 9). Major features of the gravity study are an axial low-density upper mantle or high-density lower crustal zone which is modeled as a~50 km wide body with a density of 3,190 kg/m 3 , supporting interpretations of a mafic underplate layer beneath the northwestern rift flank (Cornwell et al., 2006). Two high-density (3,000 kg/m 3 ) upper crustal bodies underlie the MER: a 20 km wide axial body and a 12 km wide off-axis body, both of which are likely gabbroic in composition.
MT data provide some of the best evidence for the emplacement and migration of melt during breakup. Major features of the MT survey are a conductive body at 20-25 km depth beneath the rift axis interpreted as a middle-to-lower crustal magma reservoir (Figure 9). A second highly conductive region at 25-35 km depth is beneath the rift flank near the SDFZ and YTVL (Figures 6 and 8). The lower crustal high conductivity zone coincides with seismicity, leading to its interpretation as a pressurized magma reservoir beneath the rift flank volcanoes   (Figure 9). Ambient noise tomography reveals low V S (<3.2 km/s) beneath the magmatic segments, suggesting that the zones of solidified axial intrusions still include a small fraction of partial melt (Kim et al., 2012). Lower crustal melt retention is also consistent with high V P /V S ratios imaged in the upper crust beneath the rift axis using local earthquake tomography (Daly et al., 2008). Petrological analyses along the SDFZ indicate that erupted lavas fractionated at a range of crustal depths, in line with the geophysical observations (Rooney et al., 2011(Rooney et al., , 2014(Rooney et al., , 2007. A shallow conductive lens at <1 km depth underlying Boset Volcano is most likely indicative of either a shallow hydrothermal or a magmatic system (Whaler & Hautot, 2006).
Farther south in the MER, Samrock et al. (2015) collected broadband MT data over the Aluto volcanic center (Figure 7a). They found no middle or lower crustal conductor there, despite models of geodetic data indicating melt storage and migration at <10 km depths (Biggs et al., 2011). These observations have subsequently been interpreted as resulting from a locked, crystal-rich mush which lacks the connectivity to provide high bulk electrical conductivity (Hutchison et al., 2016). Samrock et al.'s (2015) data did, however, indicate a conductor beneath the SDFZ to the northwest of Aluto, with top surface at~12 km depth. Both observations have subsequently been confirmed as melt storage beneath rift flank volcanoes by a 110 km long MT   (Keranen et al., 2004). (c) Incipient seafloor spreading in the Asal rift at the westernmost end of the Gulf of Aden rift based on velocity data from Makris and Ginzburg (1987) and Ruegg (1975) and structure from Doubre et al. (2007). (d) Shorter profile from 3-D V P tomography model shows detail across rift. Greek letters refer to specific faults with measured slip rates and colors from Doubre et al. (2007).

10.1002/2017TC004526
profile centered on Aluto and orthogonal to the rift axis, which included deeper-probing long-period MT measurements (Hübert et al., 2016).

Summary of Mature Rift Crustal Structure
Geophysical and petrological studies indicate that intrusion and fluid release into the lower crust occur beneath the rift and parts of the uplifted plateau region, 40 Myr after the onset of flood magmatism. This long-lived magmatism produces a high-velocity, high P wave lower crust with or without low S wave velocities. Within the fault-bounded basins, border faults are largely inactive, and strain is localized to <20 km wide magmatic segments: zones of localized magma intrusion, volcanic construction, and dense faulting. At least 30% of the crust beneath the magmatic segments is new igneous material. Comparison of refraction/wideangle reflection Moho and receiver function Moho reveals a discrepancy: the receiver function analyses image the top of the high-velocity lower crust in zones of active magma intrusion and provide minimum constraints on intrusive magma volumes.

Late Stage Continental Breakup and Initiation of Seafloor Spreading: The Afar Depression
The Afar depression encompasses the rift-rift-rift triple junction between the Nubian, Somalian, and Arabian plates: The Red Sea (Arabia-Nubia) and Gulf of Aden (Arabia-Somalia) rifts that bound the Arabian plate and the East African rift (Nubia-Somalia) (e.g., Beyene & Abdelsalam, 2005;Manighetti et al., 1998;McKenzie & Davies, 1970;Mohr, 1972) (Figure 7). The cumulative extension in Afar is much higher than in the MER, and it therefore is the ideal setting to understand the final stages of continental breakup (e.g., Doubre et al., 2007;Stab et al., 2016;Wolfenden et al., 2005). Well-established seafloor spreading with seafloor spreading anomalies is observed in some sectors of the Red Sea (Ligi et al., 2011) and in the Gulf of Aden (e.g., Leroy et al., 2010).
Rifting of Arabia from Africa to form the Red Sea and Gulf of Aden rift zones preceded rifting in the Main Ethiopian Rift by >10 Myr, based on structural, stratigraphic, and plate kinematic data (e.g., Chernet et al., 1998;Corti, 2009;DeMets & Merkouriev, 2016;Wolfenden et al., 2004). Stratal relations of dated volcanic sequences and aligned eruptive centers indicate that some border faults along the western margin of Afar began at 29-31 Ma, approximately coeval with the highest magma production rates in the Ethiopia-Yemen flood basalt province (e.g., Ayalew et al., 2006;Wolfenden et al., 2005). Faulting and basin subsidence had begun by about 35 Ma along large portions of the Gulf of Aden but without magmatism (Leroy et al., 2010). As in the Ethiopian rift, structural and geomorphological analyses indicate that extensional strain was initially focused on large offset border faults, and subsequently, extension has become more localized to axial volcanic segments (Hayward & Ebinger, 1996;Sembroni et al., 2016;Wolfenden et al., 2005) (Figure 7).
The comparatively long history of flood magmatism and faulting has produced unusual crust; the structure and composition of the 18-25 km thick crust beneath Afar are transitional between continental and oceanic (e.g., Makris & Ginzburg, 1987;Mohr, 1989). Its structure informs magmatic passive margin studies in the Red Sea and worldwide. Specifically, debate continues regarding the nature of crust beneath thick sequences of seaward dipping volcanic layers that mask basement structure (e.g., Franke et al., 2007;Van Avendonk et al., 2009). Is the Afar crust, which is slightly thicker than normal oceanic crust, completely new igneous material intruded as sheeted dikes fed from lower crustal magma chamber(s), as along mid-ocean ridges with slow opening rates, or is it a combination of highly stretched continental crust with magmatic intrusions and metamorphosed sediments?

Crustal Thinning and Intrusive Bodies
Over the past 30 Myr, the zone of stretched and subsiding crust has broadened to about 300 km in central Afar, whereas the zone of active deformation has narrowed to intensely faulted, <20 km wide and~50 km long magmatic segments fed from central magma chambers. Similar to the shoulders of the Ethiopian rift, receiver function analysis shows that the crust beneath the Ethiopia-Yemen Plateau is 40-45 km thick. It is 35 km thick beneath the Somalian Plateau (Dugda et al., 2005;Hammond et al., 2011;Keranen et al., 2009;Stuart et al., 2006), similar to average crustal thickness of 35 km beneath the plateau of western Yemen (Ahmed et al., 2013) (Figures 2a and 9). The Danakil horst, separating the two subparallel arms of the Red Sea rift in the offshore and in Afar, is underlain by~30 km thick crust. In general, the Ethiopian, Somalian, and Yemen Plateaux have V P /V S of 1.7-1.9, whereas localized regions of thicker (~30 km thick) crust within Afar have V P /V S of 1.8-1.9 (Dugda et al., 2005;Hammond et al., 2011;Reed et al., 2014;Stuart et al., 2006) (Figure 7c). V P /V S ≤ 1.85 can be explained compositionally (Christensen, 1996;Hacker et al., 2015), but observations of V P /V S > 1.85 strongly suggest present-day melt in rift-parallel fractures and cracks, with anisotropy contributing to the high V P /V S values (Hammond, 2014). Ambient noise tomography images low Vs in the middle to upper crust (Korostelev et al., 2015), consistent with the high V P /V S (>1.9) imaged with receiver functions , with both indicating melt-filled pore spaces/fractures.
Within the fault-bounded depression that includes the MER, southern Red Sea, and Gulf of Aden rifts, distinct differences occur. The <10 Myr old northern sector of the MER in Afar cuts through crust originally stretched and intruded as Arabia and Africa separated to form the third arm of the triple junction. The superposition of two rift systems (Red Sea, MER) causes a striking northward decrease in crustal thickness from~35 km in the central MER to~25 km in southern Afar (e.g., Figure 7b). What remains unclear, however, is the ratio of intrusive to extrusive volumes and total mechanical stretching.
In central Afar, the crust is 20-25 km thick beneath the current locus of volcanism and strain in the Manda-Harraro rift (MHR) and Tendaho Graben (TG), which are the southernmost segments of the Red Sea rift arm (Reed et al., 2014) (Figure 7). Currently active and abandoned magmatic segments are the locus of the greatest crustal thinning (<20 km), and they are separated by bands of~30 km thick, intruded crust . This "ridge-jump" pattern was created as the magma storage systems supplying magmatic segments jumped laterally across the broad Afar triple junction zone . Further suggestion of ridge migration comes from thermal modeling (Daniels et al., 2014) and Medynski et al.'s (2015) inference of chemically distinct, alternately active magma chambers. The spatial arrangement of subparallel, migrating magma accretion zones is similar to patterns observed on passive margins worldwide, and predicted in ridge-jump seaward-dipping accretion models (Buck, 2017).
As outlined in section 5. 3 and Van Ngoc et al. (1981), MT results indicate at least two of the magmatic segments: the MHR segment in the southern Red Sea and the Asal-Ghoubbet segment in the Gulf of Aden rifts extend in a fashion similar to a mid-ocean ridge (Figure 7). During intense faulting and fissural eruption episodes, dikes sourced from magma chambers near the center of magmatic segments achieved meters of rift opening and created up to 60 km long columns of mafic crust in periods of days (e.g., Abdallah et al., 1979;Ayele et al., 2009;Barnie et al., 2016;Keir et al., 2009). Gravity and magnetic data from a 50 km long,~1 m station density profile across the Tendaho basin provide indirect evidence that the intense dike intrusion events produce narrow columns of new igneous crust, similar to processes on slow-spreading ridges (Bridges et al., 2012). Here the magnetic data show near-symmetrical anomalies across the center of the rift, whose scale, shape, and magnitude are similar to those observed in nearby seafloor spreading centers in the Gulf of Aden. These data, together with the Bouguer gravity field, are consistent with symmetrical sheeted mafic dike complex intruded into the upper~10 km of the crust and capped by 2-3 km thick basin fill of fissural basalts and sediments (Bridges et al., 2012). The observations provide evidence that most crustal extension is facilitated by magma intrusion, mimicking the extensional processes at slow-spreading mid-ocean ridges. Thus, it is possible that the first magnetic anomalies recorded at volcanic rifted margins form in comparatively thick (~15 km thick) crust originally classified as "transitional crust" (e.g., Bridges et al., 2012;Bronner et al., 2011;Rooney et al., 2014).

Magmatic Plumbing System
Constraints on the magma plumbing system come from receiver functions, seismicity, MT, gravity, and magnetic studies. MT data provide unprecedented constraints on the locus and scale of magma plumbing systems that feed dyke intrusion and volcanism in the MHR segment (Desissa et al., 2013;Johnson et al., 2016) (Figure 11). A small-scale, but high melt fraction (≤22%), conductor lies slightly deeper than the inferred source of magma supplying the recent dyking events, determined from geodesy and seismology (e.g., Ebinger et al., 2008;Grandin et al., 2011;Hamling et al., 2009). A much more substantial magma reservoir straddling the Moho (~22 km deep in this area; Hammond et al., 2011) lies~25 km to the west of the rift axis ( Figure 11) where Ferguson et al. (2013) find geochemical evidence for ponding and reequilibration of magma at the top of the mantle, most likely stored in stacked sills. Although the regularized MT inversion model of Figure 11 gives a single conductor, Johnson et al. (2016) show that a series of sills, with a similar total magma volume, fit the data equally well.
20° phase split (Johnson, 2012) 5% anisotropy  5% anisotropy (Hammond, 2014) b) a) d) c) Figure 11. Comparison of MT and seismic models of crustal structure and anisotropy: (a) Magnetotelluric results for Profile 1 (Desissa et al., 2013). (b) Magnetotelluric results for Profile 3 (Johnson et al., 2016), with instrument locations shown in Figure 11c. Location of study region shown in Figure 7a. Seismic traces on both plots show receiver function migrations from Hammond et al. (2011). (c) Comparison of seismic anisotropy in the upper crust from shear wave splitting (green arrows; Keir et al., 2011aKeir et al., , 2011b and geoelectric strike from MT (blue arrows) for periods related to the low conductive (blue) areas shown in Figures 11a and 11b (1-10 s) (Johnson, 2012). (d) Comparison of seismic anisotropy in the whole crust from receiver functions (red arrows) (Hammond, 2014) and geoelectric strike from MT (blue arrows) for periods related to the highly conductive (red) areas shown in Figures 11a and 11b (100-1,000 s) (Johnson, 2012).

10.1002/2017TC004526
A second MT profile in the northwestern MHR was acquired on a trend oblique to the rift and approaching within 10 km of Dabbahu volcano at the northern end of the magmatic segment (Johnson et al., 2016). At its closest point to the volcano, the model has very high (≥12 S/m) conductivity from depths of~4 km, peaking at 6.5 km depth. Interpretation of these data is a challenge since model conductivities exceed that of basaltic melt at the appropriate temperatures and pressures, even with high water content (Pommier & Le Trong, 2011). However, Field et al. (2012Field et al. ( , 2013 found a wide range of erupted products from the Dabbahu volcano, including very evolved, wet rhyolites, which could have crystallized in the top 6 km of the crust. Recent estimates of rhyolite melt conductivity (Guo et al., 2016) are 3-4 times higher than conductivities of basaltic melt at the appropriate temperature, pressure, Na 2 O, and water content but still less than the maximum observed in our model. Very high conductivities and hence substantial melt volumes (even using the higher melt conductivities of rhyolitic magmas deduced by Guo et al., 2016) are also detected beneath the Dabbahu volcano, where Field et al. (2012) inferred protracted fractional crystallization of parent basaltic magma in a substantial, long-lived storage region, preferably in sills, between 1.5 km and 5.5 km depth, consistent with the pressured zone imaged with seismicity data ; however, MT data cannot be satisfied by a conductor confined to such shallow depths (Johnson et al., 2016). Jointly, the data support the view of sill complexes beneath axial and flank volcanic systems. However, petrological constraints suggest that fractionation starts in the lower crust beneath flank volcanoes and in the upper crust beneath axial volcanoes.
The two most recent eruptions at the northernmost magmatic segment, Erta'Ale, were sourced from dikes and sills in the upper 3 km (Magee et al., 2017;Pagli et al., 2012;Xu et al., 2017) (Figure 7b). The persistent lava lake in the Erta'Ale crater and the frequent eruptions suggest a semipermanent melt-rich sill, as along the magma-rich slow-spreading Reykjanes ridge. Farther south where the MHR segment merges with the Tendaho Graben, high conductivity is again imaged off axis at lower crustal depths (Johnson et al., 2016) ( Figure 11) but with smaller inferred volumes than beneath the recently active northern MHR. This coincides with an inferred partially molten gabbroic body that has been intruded into a thinned lower crust deduced from microgravity data (Lewi et al., 2015) (cf. models of MT and gravity data through the Boset volcano in the northern MER; Whaler & Hautot, 2006;Cornwell et al., (2006), section 4.2). Didana et al. (2014Didana et al. ( , 2015 imaged an upper crustal zone of extensive partial melt fraction (up to 0.13) beneath the Tendaho Graben geothermal prospect to the east of the southern MHR axis, connected to a large intrusive magmatic body underlying the TG and MHR, fed by a mantle source.
Farther south across the currently inactive part of the MHR, lying within the TG, high conductivity is again imaged at lower crustal depths (Profile 3), in this case~10 km to the east of the rift axis (Johnson et al., 2016) ( Figure 11). The maximum conductivity and lateral extent of the conductor, and hence inferred partial melt fraction and total melt volume, are lower than in the off-axis reservoir beneath the recently active northern MHR. In a similar location, Lewi et al. (2015) modeled microgravity data to infer that a partially molten gabbroic body has been intruded into a thinned lower crust. This agreement between MT and gravity results is reminiscent of that found for the location of dense subsurface bodies, interpreted as gabbroic, and conductors across the northern MER through Boset volcano. Didana et al. (2014Didana et al. ( , 2015 have modeled several MT profiles across the TG geothermal prospect to the southeast of the profile of Johnson et al. (2016) and to the east of the southern MHR axis. Seismic surface wave (Guidarelli et al., 2011), Pn (Stork et al., 2013), and body wave tomography  studies image slow upper mantle velocities throughout the MHR rift segment, with the most extreme reductions centered on the recently active northern part.

The Shape and Distribution of Partial Melt
Seismic anisotropy in the upper crust is constrained from shear wave splitting from local earthquakes (Keir et al., 2011a(Keir et al., , 2011b and surface waves (Bastow et al., 2010) (Figure 11d). Large amounts of fault-parallel (NNW) anisotropy are seen only in the area of 2005-2012 diking activity. Outside of this region, anisotropy in the upper crust is lower but remains aligned with major faults. These patterns suggest that the presence of melt enhances a pervasive fracture-parallel anisotropy in this region. Seismic anisotropy in the whole crust can be estimated by looking at azimuthal variations in V P /V S from receiver functions. V P /V S throughout Afar have values >2, implying that melts are also present (Dugda et al., 2005;Hammond et al., 2011;Stuart et al., 2006). However, Hammond (2014) shows that such high values are an effect of seismic anisotropy, suggesting that if melt is the cause, then it must be preferentially aligned. Inverting the azimuthal variations in Tectonics 10.1002/2017TC004526 V P /V S suggests that melt in the lower crust must be stored in interconnected sills, consistent with reflectivity studies in rift zones (e.g., Thybo et al., 2000). Estimates of azimuthal anisotropy for the whole crust have the same orientation as that in the upper crust but with much stronger anisotropy.
Further support for aligned crustal melt pockets comes from measurements of geoelectrical strike orientations determined from the MT profiles outlined in section 5.2 (Johnson, 2012). The presence of electrical anisotropy implies that the aligned melt is interconnected. For periods synonymous with the mostly low-conductivity upper crust (1-10 s), the electrical anisotropy strikes are similar to those from upper crustal earthquake shear wave splitting, and the magnitude of electrical anisotropy (measured as the difference in MT phase for currents flowing along and across geoelectrical strike; see, e.g., Hamilton et al., 2006;Padilha et al., 2006) is relatively low. For periods synonymous with the conductive lower crust (100-1,000 s) the strike of the electrical anisotropy matches well with fast directions from receiver functions and teleseismic shear wave splitting, and the amount of electrical anisotropy is higher ( Figure 11). These similarities support the idea that seismic and electrical anisotropy are caused by the preferential alignment of melt, primarily in the lower crust. These patterns suggest that melt can flow within these interconnected networks in the lower crust and feed the relatively localized zones of deformation in the upper crust (Desissa et al., 2013;Hammond, 2014). Note that there are areas with little or no electrical anisotropy, but high conductivities indicating abundant melt, in the upper crust (e.g., close to Dabbahu volcano; Figure 11), suggesting no particular geometry in its arrangement. Here the penetration depth is severely limited, and the 100-1,000 s periods are probably not penetrating the deeper crust.

Crustal Thinning and Subsidence at Plate Rupture
Wide-angle seismic experiments show that the crust throughout Ethiopia has a consistent layering with lower crustal V P = 6.7-7.0 km/s, an upper crustal V P = 6.0-6.3 km/s, and cover rocks of lava flows and sediments with V P = 2.2-4.5 km/s (Makris & Ginzburg, 1987;Prodehl & Mechie, 1991a, 1991b (Figures 8, 10, and 12). Excluding the regularly spaced volcanoes marking the centers and tips of magmatic segments, the cover rocks are generally thickest where the crust is thinnest suggesting a strong link between crustal thinning, rift valley subsidence, and resultant accumulation of basin infill ( Figure 12). The Danakil basin, which hosts the Erta'Ale magmatic segment, has the thinnest crust in Afar at 15 km thick and is also characterized by the thickest sedimentary basins:~5 km as opposed to 2-3 km elsewhere (Berckhemer et al., 1975;Makris & Ginzburg, 1987) (Figure 12). The change in crustal thickness along rift from south to north is primarily accounted for by markedly thinned lower crust, although the upper crust is also appreciably thinner (Makris & Ginzburg, 1987). Existing controlled source data do not traverse the rough relief of the currently active magmatic segments where intrusion volumes may be largest.
Spatial variations in crustal thickness in Ethiopia correlate with variations in effective elastic thickness, with the strongest plate (Te ≈ 60 km) beneath the plateaus and the weakest plate (Te ≈ 6 km) beneath the Danakil depression (Perez-Gussinyé et al., 2009). Both crustal thickness and effective elastic thickness, a measure of strength over time periods much longer than the earthquake cycle, also correlate to variations in seismogenic layer thickness estimated from local seismicity and teleseisms; seismogenic layer thickness is 25-30 km beneath the Ethiopian Plateau ) but decreases to~5 km beneath the Danakil depression (Craig et al., 2011;Nobile et al., 2012).
There is debate regarding whether the regions of thinnest crust (~15-20 km) in Afar, such as the Danakil depression, are fully oceanic in nature. In the Danakil depression, the basin floor is at~120 m below sea level and the Holocene stratigraphy is dominated by intercalated basalts and evaporites (Atnafu et al., 2015). The basin stratigraphy and environment of deposition are similar to that interpreted to form the seaward dipping reflector sequences (at fully rifted margins), and therefore, the Danakil depression could be a modern analog for their formation Buck, 2017;Corti et al., 2015). There is at least 50% more melt available Figure 12. (a) Variation in Quaternary-Recent basaltic volcanism exposed every 0.1°latitude along strike in the Red Sea rift in Afar, with crustal velocity structure largely based on the EAGLE along-axis seismic profile  and location shown in Figure   . SDR = seaward dipping reflector; δV P indicates regional velocity reduction relative to global norm.

10.1002/2017TC004526
beneath the northern MHR than in the TG segment in the south and likely up to an order of magnitude more. This change is consistent with regional northward increases in the number of Holocene volcanoes as well as the volume of erupted material (Barberi & Varet, 1977), and it is likely a result of a northward increase in thinning of the plate and resultant decompression melting of asthenosphere .
The most highly extended crust beneath the Danakil depression is 5-10 km thicker than normal oceanic crust. There are two end-member hypotheses to explain the structure and composition of the Danakil crust.
In one model, heavily intruded and extended crust deforms by brittle failure at the surface and by magma intrusion and ductile flow in the lower crust . Unlike thin crust at oceanic rifts, the intrusive material may stall and fractionate to produce peralkaline and rhyolitic lavas, and felsic intrusive material, contributing to the unusual structure. The second model relates to high magma production rates in earlystage seafloor spreading caused by the steep lithosphere-asthenosphere boundary on either side of the rift (e.g., Boutilier & Keen, 1999;Korenaga et al., 2000;Ligi et al., 2011;White & McKenzie, 1989). The high extrusion rates at the weak spreading ridge cause bending above the ridge and may produce seaward dipping reflectors. Our syntheses suggest that both processes are important.

Summary: Defining Transitional Crust and Plate Rupture Along Magmatic Margins
Passive and controlled source seismic imaging shows that the crust in Afar varies from 15 to 25 km thick, and the crust is heavily modified by solidified mafic intrusion and partial melt currently residing in magma reservoirs through the crust. MT combined with modeling seismic and electrical anisotropy supports the interpretation that the lower crust is an important region of melt ponding. The melt in the lower crust resides in horizontal sills near segment centers, and it is then transported laterally and vertically in dikes that accommodate most of the plate boundary deformation. The central magma chamber that maintains the along-axis segmentation over multiple episodes and the extension via dike intrusion and faulting above the dikes produce columns of new crust with corresponding magnetic stripes. The unusually thick crust (~15 km) matches the thick crust beneath magmatic rifted margins with seaward dipping reflector sequences, pointing to a common mode of emplacement. In contrast, the observed seismic velocity and structure of the crust and the low-silica composition of basalts suggests that seafloor spreading processes are still evolving.
Seismic imaging, gravity, and MT all suggest that the whole crust beneath the rift flanks and the rift axis is heavily modified by magma intrusion. Seismic and electrical anisotropy support the view that magma accumulates in the lower crust as networks of interconnected stacked sills, whereas magma in the upper crust may be both in sill networks and subvertical dikes. Petrological constraints suggest that fractionation starts in the lower crust beneath flank volcanoes during the early stages of rifting, whereas most fractionation occurs in the upper crust beneath axial volcanoes as plate rupture initiates (Rooney et al., 2011(Rooney et al., , 2014.
Gravity studies in Cameroon (e.g., Fairhead & Okereke, 1987 ;Poudjom Djomani et al., 1992  Recent work by Milelli et al. (2012) and Fourel et al. (2013) has used scaled laboratory models and analytical solutions to investigate the effects of large lateral variations in lithospheric thickness on lithospheric stability. Their work demonstrates that lithospheric instabilities can develop over long timescales with small rates of

The Atlas Orogen
In contrast to the Mediterranean arcs (Alps, Apennine, Carpathian, Hellenic, and Betic-Rif), the Atlas is subparallel to and south of the convergent margin and on the African plate (Figures 1 and 3a). The mountain belt has high topography (>4,100 m), yet there is no apparent deep crustal root to isostatically support the high elevations of the Atlas (e.g., Ayarza et al., , 2014Jessell et al., 2016;Missenard et al., 2006;Sandvol et al., 1998;Zeyen et al., 2005) (Figure 3a). Furthermore, only a modest amount of tectonic shortening has been estimated and has been suggested to be achieved through thick-skinned thrusting and folding (e.g., Gomez et al., 2000;Teixell et al., 2003).
Many recent efforts have focused on understanding the crustal-and lithospheric-scale structure of the Atlas to understand the orogenesis. Structural seismological imaging suggests that the lithosphere beneath the Atlas is particularly thin or perhaps the uppermost mantle is abnormally warm with low seismic velocities found at depths of~65-160 km (e.g., Bezada et al., 2014;Fullea et al., 2007;Miller et al., 2015;Palomeras et al., 2014;Sun et al., 2014). Figure 14 shows S receiver function estimates of the Moho and lithosphereasthenosphere depths along a profile across the Atlas from Miller and Becker (2014). Geophysical modeling indicates that the lithosphere is thin (~65 km) compared with the topographically high, thick lithosphere of the Saharan platform (≥150 km) and the Morocco Atlantic margin along a corridor that is aligned with the highest topography in the Atlas (Fullea et al., 2007;Miller et al., 2015;Missenard et al., 2006;Missenard & Cadoux, 2012;Teixell et al., 2003Teixell et al., , 2005) (see Figure 3a). The combination of thinned lithosphere and mantle upwelling has been invoked to explain the unusual high topography, low seismic velocities in the uppermost mantle, and lack of a significant crustal root (Arboleya et al., 2004;Frizon de Lamotte et al., 2009;Fullea et al., 2007;Missenard et al., 2006;Teixell et al., 2005;Zlotnik et al., 2014). The source of the upwelling has been suggested to be part of the Canary plume from geochemical analyses of Quaternary alkali basalts in the Atlas and their spatial location above the low velocity material beneath the mountain range (Anguita & Hernán, 2000;Duggen et al., 2009).
Shear wave splitting analyses (Díaz & Gallart, 2014;Miller et al., 2013) also indicate a strong change in azimuthal anisotropy strength at the northern edge of the Middle Atlas, as well as an alignment of fast polarization orientations parallel to the strike of the Atlas Mountains, suggesting shearing in a mantle channel guided by lithospheric topography (Miller et al., 2013; and deflected by the high-velocity slab   Miller et al. (2015) interpret the localized volcanism and low velocity anomalies in the sub-lithospheric mantle to be the result of the Canary Island plume flowing sub-horizontally beneath the Atlas. This seismologically based interpretation is supported by numerical experiments of mantle flow that incorporate the effects of the stiff, deep West African cratonic keel and nearly vertical, narrow slab beneath the Alboran (Alpert et al., 2013), numerical experiments of strongly tilted plumes and resulting instabilities (Mériaux et al., 2011), and by scaled analogue models of the Alboran slab and Canary plume interaction (Mériaux et al., 2015). Other hypotheses explaining the Cenozoic uplift of the Atlas and the links to recent volcanism and to upper-mantle structure include: lithospheric delamination Duggen et al., 2009;Levander et al., 2014) and edge-driven convection (Kaislaniemi & van Hunen, 2014;Missenard & Cadoux, 2012). The two edge-driven convection studies (Kaislaniemi & van Hunen, 2014;Missenard & Cadoux, 2012) are based on numerical experiments, which are designed to evaluate the dynamics of the thinned lithosphere and mantle melting processes in the Moroccan Atlas. They also find good agreement with observations of lithospheric thickness and volcanism, but neither incorporate the presence of the subducted slab beneath the Alboran Sea.
Despite the lack of consensus on the origin of the Atlas topography, various recent studies using a range of methods, from the SIMA active source seismic experiment (Ayarza et al., 2014), broadband waveform analysis (Sun et al., 2014), MT (Anahnah et al., 2011;Ledo et al., 2011), body wave (e.g., Bezada et al., 2014), and surface wave  tomography, for example, have suggested the presence of anomalous lowvelocity material beneath the lithosphere and in the crust. The temperature of this material is also inferred to be warm, due to the presence of partial melt (Anahnah et al., 2011;Ayarza et al., 2014;Ledo et al., 2011;Sun et al., 2014).

Summary
Our synthesis of geophysical studies that aim to image the crust beneath zones of active rifting and orogenesis in Africa shows that magma and volatiles are migrating from the asthenosphere through the plate, leading to changes in rheology and making large contributions to H 2 O and CO 2 fluxes globally. The addition of magmatic fluids permanently alters the composition of continental crust accreted during the Proterozoic, and it increases the continental crustal volume. The magma intrusion accommodates plate boundary deformation and masks crustal stretching. In young, weakly extended sectors of the East African rift, volatiles released from upwelling asthenosphere, heated mantle lithosphere, and magma intrusions alter crustal composition and rheology. Even in rifts that formed within the last 7 Myr,~20% of the crust is new magmatic material intruded as sills in the lower crust (underplate) and as more localized zones of dike intrusion in the middle and upper crust as imaged with seismic, MT, and gravity methods. In the mature Main Ethiopian Rift (MER), the surface area of the rift has doubled in the past~10 Myr, and intrusive volumes are 30%, increasing into the central Afar rift zone where dike intrusion accommodates much of the active extension, and symmetric magnetic anomalies mimic seafloor spreading. Intrusive to extrusive volumes (surface flows plus basin fill) are roughly equivalent but increase with increasing age of rifting and amount of extension and proximity to prerift flood magmatism. In the older, more evolved Horn of Africa, intense episodes of magma intrusion fed from crustal magma chambers beneath segment centers create new columns of mafic crust up to 8 m wide in a process akin to that along slow-spreading ridges. The magma intrusion and faulting above the dikes accommodate centuries of plate boundary opening and suggest that plate rupture is achieved through catastrophic magma intrusion events in 15-20 km thick transitional crust. Regional analyses suggest that progressively smaller screens of continental crust are heated and weakened by repeated episodes of magma intrusion immediately prior to seafloor spreading.
On the western side of the African continent, magmatism with very minor extension over the past 30 Myr has formed the Cameroon Volcanic Line (CVL)-a linear chain of basaltic volcanoes that formed on the broad Adamawa Plateau that lacks the age progression predicted by the traditional hot spot track hypothesis. Intrusive volumes are much smaller than in the Eastern, MER, and Afar rifts, and magmas largely fractionate at subcrustal depths. The comparison with East African crustal structure suggests that extensional forces are much smaller in the CVL, inhibiting the creation and rise of magma into the crust and preventing full-scale rifting. In the Atlas orogen of northwestern Africa, recent magmatism is also primarily along a linear trend from the Canary Islands to the Alboran Sea and linked to the Canary Island hot spot. In contrast to the Tectonics 10.1002/2017TC004526 magma-rich Horn of Africa, however, magmatism in Cameroon and Morocco has had relatively little impact on crustal structure, but the mantle flow and lithospheric heating cause dynamic uplift supporting the high Atlas mountain range.
Our work highlights consistent patterns and new insights into continental lithospheric deformation processes. Based on our results and complementary crust and mantle xenolith studies, future studies would benefit from constraints on physical properties across a broader range of crustal and magmatic rock compositions, and in particular, relatively low silica, and gas-rich melts. Our work focuses on areas with sometimes dense data coverage, but large sectors of the African continent remain virtually unexplored in terms of crust and upper mantle structure through seismic, MT, heat flow, and detailed structural analyses. Gap-filling studies will tighten the context for studies of active deformation, and they promise new insights into continental lithospheric behavior, given the relative stability of the African continent. Critical areas include the largely amagmatic but remote Western and Southwestern rift zones, the Turkana gap between the Ethiopian and East African Plateaux, the seismically active Indian Ocean margin, and the boundaries of thick Archaean cratons, which have been the locus of plate deformation through multiple Wilson Cycles. Tanzania Development Corporation. C. T. and S. H. were supported by ANR grant 12-JS06-000. C. E., C. T., and D. K. conducted studies with approval of the Commission for Science and Technology (Tanzania) and the National Council for Science and Technology (Kenya