Structure of Tendaho Graben and Manda Hararo Rift: Implications for the evolution of the southern Red Sea propagator in Central Afar

Transcription

1 TECTONICS, VOL. 27,, doi: /2007tc002236, 2008 Structure of Tendaho Graben and Manda Hararo Rift: Implications for the evolution of the southern Red Sea propagator in Central Afar V. Acocella, 1 B. Abebe, 2 T. Korme, 2 and F. Barberi 1 Received 22 November 2007; revised 1 April 2008; accepted 21 April 2008; published 27 August [1] The Red Sea and Aden rifts (or propagators) meet in Afar. Here we use remote sensing and field analyses to define the geology and structure of the southern part of the Red Sea propagator in Central Afar. This consists of the NW-SE trending Tendaho Graben (TG) and the younger and active NW-SE trending Manda Hararo Rift (MHR), partly within TG. Tectonic and volcanic activity within TG developed mostly between 1.8 and 0.6 Ma, with a stretching factor b 1.1, an extension rate 3.6 mm/yr and the fissural eruption of part (7000 km 3 /Ma) of the Afar Stratoid sequence (mainly basaltic lava flows and ignimbrites). MHR, before terminating southward, has a b 1.04 and an extension rate 1.2 mm/yr, and is associated with the emission of 600 km 3 /Ma of basalts in the last 0.2 Ma. These data suggest that after the exceptional amount of magma erupted between 1.8 and 0.6 Ma, magmatic and tectonic activity significantly decreased along the southern part of the Red Sea propagator in the last 0.2 Ma. This decrease coincides with the onland development and migration of the more active (inferred extension rate in the order of 10 mm/yr, as proposed in previous studies) Aden propagator, suggesting that spreading in Central Afar mainly occurred along one active propagator at any one time. Citation: Acocella, V., B. Abebe, T. Korme, and F. Barberi (2008), Structure of Tendaho Graben and Manda Hararo Rift: Implications for the evolution of the southern Red Sea propagator in Central Afar, Tectonics, 27,, doi: / 2007TC Introduction [2] Divergent plate boundaries provide the most suitable site to study how extension occurs in the Earth s crust, including the development of a rift zone and its relationships with volcanic activity. Investigations in Iceland and along submerged oceanic ridges have revealed how extension and volcanism may occur along oceanic divergent 1 Dipartimento Scienze Geologiche Roma Tre, Largo S.L. Murialdo, Rome, Italy. 2 Department of Geology and Geophysics, Addis Ababa University, Addis Ababa, Ethiopia. Copyright 2008 by the American Geophysical Union /08/2007TC boundaries [Gudmundsson, 1995, 1998; Macdonald, 1998]. On continents, the most suitable place to investigate divergent plate boundaries is the East African Rift System. In a very similar fashion to oceanic boundaries, this continental rift consists of spaced spreading segments associated with volcanism and tectonic activity [Mohr and Wood, 1976; Ebinger and Casey, 2001, and references therein]. The composition of volcanism varies, from basaltic to rhyolitic, as a function of the crustal thickness [Trua et al., 1999; Lahitte et al., 2003a; Peccerillo et al., 2003]. Nevertheless, the overall deformation pattern, characterized by large extension fractures and open normal faults with tilted hanging-walls, is remarkably similar to that observed along oceanic boundaries [Kazmin and Byakov, 2000; Le Gall et al., 2000; Acocella et al., 2003; Williams et al., 2004]. Minor differences do exist in the size, aspect ratio and spacing of the faults, as a function the crustal thickness [Ebinger and Hayward, 1996]. [3] The northern portion of the East African Rift System, together with the Red Sea and Aden rifts, forms the Afar triple junction (Figure 1a). This marks the transition from continental (southern part) to oceanic crust (northern part [Hayward and Ebinger, 1996]). Most of the Afar tectonic and volcanic activity is due to the mechanical interaction of the on-land southern continuation of the Red Sea Rift [Beyth, 1991; Ghebreab, 1998], here referred as to the Red Sea propagator (or propagating rift [Manighetti et al., 2001, and references therein]), with the partly overlapping on-land western continuation of the Aden Rift (here referred as to the Aden propagator [Manighetti et al., 2001, and references therein]). This interaction forms an overlapping spreading center, resulting in intense faulting and diffuse block rotations in between [Mohr, 1972; Tazieff et al., 1972; Tapponnier et al., 1990]. The geology and structural features of Aden propagator into Afar have been widely studied [e.g., Needham et al., 1976; Stein et al., 1991; Manighetti et al., 1998, 2001; Audin et al., 2004, and references therein]. Conversely, our poorer knowledge of the geology and the structure of the Red Sea propagator is based on general studies, on average several decades old [e.g., Barberi and Varet, 1970; Schaefer, 1975; Abbate et al., 1995], or confined to sporadic episodes of activity [Ogubazghi et al., 2004; Wright et al., 2006]. Of particular interest is the definition of the geology and structural features of the NW-SE trending southern portion of the Red Sea propagator, overlapping with the Aden propagator in Central Afar (Figure 1a [Tapponnier et al., 1990]). The most important structural features of the former include the NW-SE trending Tendaho Graben, the largest depression of Central Afar, and 1of17

2 Figure 1. Regional setting of Afar region (bottom right). (a) General structure of the Afar triple junction. RSP, Red Sea propagator; SRSP, southern Red Sea propagator; AP, Aden Propagator; EAR, Erta Ale Range; MHR, Manda Hararo Rift; TG, Tendaho Graben. (b) Geology of Tendaho Graben area. (c) Projection of the age (in Ma) of the Afar Stratoid deposits along a NE-SW section perpendicular to the present MHR axis (age data from Barberi et al. [1975], Lahitte et al. [2001, 2003a, 2003b], and Kidane et al. [2003]). the younger and active Manda Hararo Rift, partly located within Tendaho Graben (Figure 1a). The knowledge of the structural features of the southern portion of the Red Sea propagator appears crucial to understand better the tectonic evolution of Afar, and, in general, how rift systems propagate and interact in a triple junction characterized by overlapping spreading centers. [4] The goal of this study is to understand how extension takes place at the southern portion of the Red Sea propagator, in the area of Tendaho Graben, in order to define its main structural features and to place these in the context of Afar tectonics. For this purpose, remote sensing and field structural analyses were carried out. The collected data permit to: 1) define the main geometric and kinematic features of the southern portion of the Red Sea propagator; 2) consider these in the frame of the recent tectonic and magmatic evolution of Central Afar, with particular attention given to the development of the overlapping zone formed by the Red Sea and Aden propagators. 2. Geologic Setting [5] In Central Afar, the Red Sea and Aden propagators meet with the northern portion of the Main Ethiopian Rift (MER; Figure 1 [McKenzie et al., 1970]), deforming a broad area and developing microplates (Danakil microplate 2of17

3 [Le Pichon and Franchteau, 1978; Stein et al., 1991; Eagles et al., 2002; Redfield et al., 2003]). The mean spreading rates of the Aden propagator and the northern portion of the Red Sea propagator are 1.1 cm/yr [Cattin et al., 2005, and references therein] and 2 cm/yr [Jestin et al., 1994] respectively, significantly higher than the 2.5 mm/yr of MER (Figure 1a [Wolfenden et al., 2004, and references therein]). This suggests that most of the strain in Afar results from the activity and interaction of the Aden and Red Sea propagators [Tazieff et al., 1972; Schaefer, 1975; Tapponnier et al., 1990; Bosworth et al., 2006]. Volcanic activity has accompanied the development of the triple junction; its evolution can be summarized through 3 main stages: a) the emplacement of the widespread and thick Stratoid sequence (for terminology, see Kidane et al. [2003, and references therein]), made up of flood basalts and ignimbrites, marking the transition to an oceanic crust, from 4 to 1 Ma; b) the development of central silicic volcanoes, as precursors to rift propagation, in the last 2 Ma; c) the current oceanic-type basaltic volcanism, along the active rift zones [Barberi et al., 1972, 1975; Lahitte et al., 2003a, 2003b; Beyene and Abdelsalam, 2005]. [6] The structure of the Aden propagator has been widely studied. It mainly consists of two NW-SE trending rift segments: these are the Asal-Ghoubbet Rift, to the south, and the Manda Inakir Rift, to the north. The two segments are connected by an oblique transfer zone, characterized by the rotation about a vertical axis of small rigid blocks (Figure 1a [Manighetti et al., 2001, and references therein]). Each segment essentially consists of subvertical to riftwarddipping active normal faults and an axial portion of active tensile fracturing and volcanic activity [Needham et al., 1976; Manighetti et al., 1998]. The long-term vertical deformation is not steady state; rather, it involves repeated cycles [Stein et al., 1991]. [7] The less studied Red Sea propagator is characterized, to the north, by the Erta Ale Range, which separates the Ethiopian Plateau from the Danakil Block. The Erta Ale Range consists of a NW-SE alignment of active volcanoes, largely shield volcanoes with basaltic composition, associated with fracturing [Barberi and Varet, 1970; Acocella, 2006]. To the South, the Red Sea propagator branches into an eastern segment, the Tat Ali Rift, and a western segment, which propagates into Central Afar and we name southern Red Sea propagator. This includes the NW-SE trending Tendaho Graben (TG), the largest basin of Central Afar (Figures 1a and 1b), and the active NW-SE trending Manda Hararo Rift (MHR; Figure 1a [Tapponnier et al., 1990; Manighetti et al., 2001, and references therein]). TG marks the northern limit of the MER structures (Figure 1b). TG, formed in the last 1.8 Ma [Acton et al., 1991, 2000], is several tens of km wide and a few hundred kilometres long and hosts eruptive fissures and central volcanoes. Its bedrock mainly comprises Stratoid deposits, which generally young toward the central part of the graben (Figure 1b [Lahitte et al., 2003b; Kidane et al., 2003]). The infill of TG, overlying the Stratoids, consists of volcanic and sedimentary deposits, among which are, to the NW, the youngest basaltic deposits ( Ma [Lahitte et al., 2001]) of the axial part of MHR (Figure 1b). The northernmost part of MHR, outside the study area, underwent a major rifting episode in 2005, associated with the emplacement of a dike 60 km long and up to 8 m wide. Dike intrusion was accompanied by seismic activity. The 2005 episode shows the importance of catastrophic magma emplacement in the development of the rift [Wright et al., 2006; Ayele et al., 2007]. To the south, MHR grades into the northern TG (Figure 1b). [8] The area between the Asal-Ghoubbet-Manda Inakir rifts and Manda Hararo Rift shows pervasive faulting, with the development of tens of rigid blocks separated by NW- SE trending faults. Their kinematics is not well defined, as the deformation may be consistent with (1) an overall extension on rift parallel normal faults [Souriot and Brun, 1992; Rouby et al., 1996; Tesfaye, 2005] and (2) a reactivation of the NW-SE trending normal faults as left-lateral structures, due to the interaction between the two overlapping propagators; the left-lateral component is consistent with the measured clockwise rotations of crustal blocks about a horizontal axis in the last 1.8 Ma, resulting in a bookshelf faulting mechanism involving Central Afar [Courtillot et al., 1984; Tapponnier et al., 1990; Acton et al., 1991]. However, as block rotation in western Afar has not reached the same evolution as in eastern Afar [Acton et al., 2000], the NW-SE faults may display both normal and strike-slip motions, as suggested by seismic activity, associated with surface fracturing [Hofstetter and Beyth, 2003, and references therein; Hagos et al., 2006]. [9] MER reached Afar propagating northward in the last 11 Ma [Wolfenden et al., 2004]. NNE-SSW trending MER is characterized by a progressive widening northwards, decreasing the length of its rift segments, the separation of the magmatic centers and the effective elastic thickness of the crust [Ebinger and Hayward, 1996; Hayward and Ebinger, 1996]. The structure and volcanism of the southern portion of northern MER has been studied [i.e., Chorowicz et al., 1994; Korme et al., 1997; Boccaletti et al., 1998; Mahatsente et al., 1999; Ebinger and Casey, 2001; Acocella and Korme, 2002; Bonini et al., 2005; Kendall et al., 2005] in more detail than the northernmost portion [i.e., Christiansen et al., 1975; Chernet et al., 1998; Chernet and Hart, 1999; Bastow et al., 2005; Keir et al., 2006]. 3. Methods [10] Satellite and DEM images of Central Afar have been used to define the location and pattern of the major lineaments, which are inferred to be Quaternary faults. The subsequent field analysis has been focused along the most significant lineaments, to verify that they are faults and to define the fault geometry and kinematics. [11] Fault kinematics has been obtained using two methods. One is the measurement of fault pitch (angle which a slickenline on the fault plane makes to the strike direction), when the slickenlines where present on the fault plane. However, as in most cases the slickenlines were not present, the pitch was calculated measuring the vertical and horizontal topographic offset of dissected NNE-SSW structures 3of17

4 Figure 2. Sketch illustrating the hard-linkage type of interaction between MER and TG structures (see Figure 1b for location). A NNE-SSW (MER) fault is offset by a NW-SE (named Tendaho) fault, even though the former slightly cuts the latter (see inset b). Roll-over structures accompany the deformation. The resulting predominant left-lateral displacement along the TG fault can be directly appreciated. Inset c shows the general geology of the site (dashed square), also indicated in Figure 1. (see example in Figure 2 and details in section 4.1). The extension direction of a fault with known pitch is then obtained using the procedure of Marrett and Allmendinger [1990], where the extension direction is at 45 from the slip vector along the plane which passes through the pole to the fault and the slip vector. Where the fractures had a significant opening, the extension directions have been obtained matching the asperities on the sides of faults (with moderate throw, in map view) and extension fractures [e.g., Acocella and Korme, 2002]. In this case, to quantify any amount of horizontal shear, we defined the angular difference (angle c) between the direction perpendicular to fracture strike (theoretical direction of pure extension) and the measured opening direction. The value of c is proportional to the amount of horizontal shear of the fracture. [12] Field analysis was focused on the identification of crustal domains with consistent attitude of the volcanic deposits. As the area of study mainly comprises domino blocks rotated about horizontal axes, the attitude measurements permitted estimation of the mean dip of the faults before (a 1 ) and after (a 2 ) rotation and, subsequently, the amount of stretching. In fact, when there is no evidence that the rotation of the faults is controlled by younger faults, the dip of the fault after rotation a 2 can be directly estimated [Jackson and White, 1989] from the amount of tilt d of the block, where d ¼ a 1 a 2 ð1þ [13] This allows the evaluation of the stretching factor b associated with the tilted blocks [Jackson and White, 1989], given by b ¼ a 1 =a 2 [14] The amount of across-rift extension b d is estimated summing the stretching factors b n of each block with consistent tilt, weighted to the width L n of the block relative to the entire width of the graben L tot, so that: b d ¼ S b n L n =L tot [15] An additional stretching factor b t within the study area was estimated from the initial thickness (pre-extension) t i and the current thickness (post-extension) t f of the crust, using the relation: b t ¼ t i =t f [16] For this crustal thickness b value estimate, we have assumed symmetric extension and neglected any depth variation in the Moho. The current crustal thickness within (t f ) and outside (t i ) the stretched area were estimated with a significant error (30% [Redfield et al., 2003, and references therein]). Therefore the estimate of the stretching factor ð2þ ð3þ ð4þ 4of17

5 given by (2) and (3) is considered more accurate, at least for the upper crust, than that given by (4). 4. Results 4.1. Relationships Between Tendaho Graben and Main Ethiopian Rift [17] TG is 200 km long and its width varies from 30 km near its northern and southern tips to >50 km in its center. TG interrupts the continuity of MER, as suggested by the abrupt termination of the structures forming the latter (Figure 1b). The northernmost MER structures in the TG area are found within the Afar Stratoids on the SW shoulder of TG (Figure 1b). On this shoulder, the NNE-SSW trending MER structures show two distinct modes of interaction with the NW-SE trending TG structures. In some cases, they curve, changing orientation northward and progressively merging with the TG structures. This change in strike of MER faults occurs, with similar patterns, at very different scales, from several tens of km to a few km, on the SW and E side of TG (Figure 1b). This type of interaction, with the progressive variation of the extension direction, is scale-independent. In the other common interaction geometry, observed only on the SW shoulder of TG, discrete cross-cutting relationships between MER and TG structures occur (Figure 1b). These cross-cutting structures dissect each other almost perpendicularly, forming a characteristic rhomboid pattern. The frequency of this type of interaction decreases to the north, disappearing in correspondence with the southern border of TG. Both the merging of MER structures against TG structures and the lack of MER structures within and to the N of central TG indicate a drastic change in the tectonic conditions, probably controlled by the higher spreading rate (almost 1 order of magnitude) of the southern Red Sea propagator relative to the Main Ethiopian Rift. [18] More detailed field measurements of these interactions show that both faults systems are active and crosscutting each other, but have different displacement rates. An example of cross-cutting relationship is shown in Figures 2a 2c: the NNE-SSW trending normal fault in the northern MER is displayed by a NW-SE trending fault on the TG shoulder. However, the foot-wall of the latter is slightly dissected by the MER fault, suggesting an almost contemporaneous, even though not comparable, activity of the two faults. As the Afar Stratoid sequence is not typically eroded in this desert climate, topography can be used as a reference to measure the offset of faults [Rouby et al., 1996]. While the northern MER faults have a predominant extensional component [Keir et al., 2006, and references therein], the NW-SE fault has a predominant strike-slip activity, with a left-lateral component 300 m and a vertical component 20 m (Figure 2). Very similar kinematic results have been obtained at other sites on both sides of TG and show significant left-lateral slip along the NW-SE faults (Table 1) Structure of Tendaho Graben [19] TG is bordered by NW-SE trending faults, at all scales (Figure 3). Remote sensing and field data show that the overall architecture of TG is characterized by the following features (Figure 3): [20] 1. The NE and SW shoulders consist of subhorizontal Stratoid layers, often dissected by NW-SE trending faults. [21] 2. At the NE and SW graben margins, marked by the highest visible scarps, the subhorizontal Stratoid layers end, forming inward dipping Stratoid blocks arranged in a domino configuration inside the margins. The dominos, usually 1 km wide, are bordered by NW-SE trending faults which dip away from the graben axis; the displacement of these faults decreases toward the graben axis, in a similar fashion to the amount of tilt of the blocks that they bound (between 30 and 5 ). [22] 3. Stratoids in the inner portion of TG are deformed into a broad NW-SE trending syncline-like structure (Figure 3), formed by the tilted blocks separated by the NE-SW trending faults. Therefore the syncline is faultcontrolled and its axis represents the axis of TG. d) Thick (1000 m) volcanic and sedimentary sequences are present within the TG; the youngest basalts (300 to 30 ka; Figure 1b [Lahitte et al., 2001]) outcropping in its NE part mark the present active axis of MHR (Figure 3), described in section 4.3. [23] The spatial coincidence between the sharp topographic variations at the sides of TG (scarps >200 m high), the locations of the tilted blocks, and sporadic hydrothermal activity suggest the presence of master faults bordering Tendaho Graben (Figure 3). The foot-wall of the master faults is made up of Stratoids with subhorizontal attitude, whereas the hanging-wall is made up of tilted Stratoid blocks. Their vertical displacement does not only result from topographic variations (>200 m), but also from the tilt of the blocks on the hanging-wall. This accounts for a significant part of the 1600 m of vertical displacement calculated by means of drill data in the Dubti area (Figure 1 [Battistelli et al., 2002]). [24] The amount of rift-perpendicular extension across TG is estimated using the two methods described in section 3. In the one case, the 1600 m of vertical displacement of the Stratoid sequence, over a crust with a mean thickness of 15 ± 5 km [Redfield et al., 2003, and references therein], gives a stretching factor (equation (4)) b t =t i /t f 1.11 (where t i 17 km and t f 15 km). In the other case, the amount of extension is estimated from the tilt (and width) of the domino blocks, along two sections (northern and southern TG; Figure 3). In fact, knowing the original dip of the faults (a 1 ), the amount of tilt allows determining (through equation (1)) the post-rotation dip of the faults (a 2 ). The active normal faults in TG are characterized by a significant dilational component, with dragged hanging-wall. These features are identical to those of the normal faults observed along the axial part of rifts at divergent margins, including the northern portion of the Red Sea Propagator [Acocella, 2006], MER [Acocella et al., 2003] and Iceland [Gudmundsson, 1992]. As their mechanism of formation implies the growth from subvertical extension fractures in the axial zone of the rift and a progressive moderate shallowing with depth (1 degree 5of17

6 Table 1. Geometric and Kinematic Features of Extension Fractures and Faults in TG Area Site Fault Direction ( ) Extension Direction ( ) c ( ) DV (m) DH (m) DW (m) Pitch ( ) s d s d s s s s s s s s s s d s d s s s s s d see Figure 5b d see Figure 5b s see Figure 5b s see Figure 5b d see Figure 5b d see Figure 5b d see Figure 5b s see Figure 5b d see Figure 5b d see Figure 5b s see Figure 5b d see Figure 5b s see Figure 5b d see Figure 5b d see Figure 5b s s s s Site, measure site (for the location, see Figures 3 and 5); c, angle between the direction of extension and the direction orthogonal to the fracture (d, s, dextral and sinistral components of shear, respectively); DV, DH, DW, components of vertical, horizontal, and dilational displacement along normal faults; pitch, angle which a fabric on the fault plane makes to the strike direction. Values from sites 36 and 37 are reported in detail in Figure 5. for every ca. 150 m [Gudmundsson, 1992]), it is expected that the normal faults in TG have an original subvertical dip a 1 in the order of Knowing a 1 and a 2 and applying equations (2) and (3) to the northern part of TG (profile 1, Figure 3), for a 1 =80, one obtains a b d = 1.05 and, for a 1 =70, one obtains a b d = On the southern part of the graben (profile 2, Figure 3), one obtains, for a 1 = 80 a b d = 1.04 and, for a 1 =70 a b d = 1.1. Therefore the 6of17

7 Figure 3. (a) Structure of the central part of Tendaho Graben, reporting the largest faults (smallest-scale faults are neglected), as well as the attitude of the Stratoid and MHR basalt (darker grey) layers. Both define domino-like structures, with blocks characterized by a uniform tilt angle separated by faults. The obtained extension directions in the TG area and along MHR are reported (double arrows refer only to the minimum and maximum value for each site). A weak component of the left-lateral shear is observed in the axis of MHR. (b) Section drawings summarizing the attitude of the deposits in TG along the profiles of Figure 3a. Dark grey, Afar Stratoids; light grey, MHR basalts; double arrows, areas characterized by bookshelf faulting; r.a., rift axis. estimates of stretching in the two profiles across TG are similar and consistent with a mean value of b d =1.07± Adding the horizontal extension of 0.3 km, due to the vertical displacement of 200 m along two 70 -dipping border faults, gives a final b tot This is similar to the extension estimate calculated using equation (1) (b t 1.11). [25] Extension rates for the TG can be obtained from the mean value of the two calculated stretching factors (b tot 7of17

8 Figure 4. (a) Variation of the extension direction of fractures across TG with regard to the Manda Hararo Rift axis. The direction of pure extension (perpendicular to the rift axis) is reported for reference. (b) Variation of the angle c (between the extension direction of the fracture and the direction orthogonal to its strike, see inset) across TG with regard to the Manda Hararo Rift axis. Positive values of c imply a component of sinistral shear; negative values imply a component of dextral shear. 1.10), the mean width of the graben (50 km) and the mean age of the Stratoids in the TG area (1.4 ± 0.8 Ma, K-Ar determinations; Figure 1b [Kidane et al., 2003, and references therein]), roughly consistent with the inferred onset of development of TG (1.8 Ma [Acton et al., 1991, 2000; and references therein]). The total extension, as the difference between the width of TG after and before extension (the latter is given by the present width of TG divided by b tot ), is 5 km. This value of total extension is similar to the estimate of Schaefer [1975] and produces an extension rate of 3.6 mm/yr for the last 1.4 Ma, one to two orders of magnitude larger than that of the nearby grabens [Acton et al., 1991, 2000; Tesfaye, 2005]. [26] The directions of extension on faults and extension fractures in the area of TG are shown in Figure 3 and listed in Table 1. The extension directions in the axial zone of MHR are described in section 4.3. All of the extension directions for MHR and TG have been projected along a NE-SW trending section across the graben (Figure 4a). There is a general variation from a cluster between N30 - N60 E in the axial zone to values N-S at the graben sides. For the mean trend of faults in the axial region (N325, in MHR), the opening directions are here consistent with a pure extension (N55 E ± 5 ), with the possibility of a small component of left-lateral shear (N30 E to N50 E). However, the left-lateral component increases toward both margins of the TG. The angular difference c between the direction perpendicular to that of the structures (direction of pure extension) and the measured opening direction (inset of Figure 4b) is also reported as a function of the across-rift distance (Figure 4b). The cluster of values 5 < c <5 in the axial part confirms the overall pure opening, though often (5 < c <25 ) associated with a minor component of left-lateral shear. The values of c >40 at the TG margins 8of17

9 Figure 5. (a) Structure of the southern portion of MHR (for location, see Figure 3). (b) Map view transect across the rift axis (shown in Figure 5a by a thick line in the upper left) with structures (faults, solid lines; extension fractures, dashed lines) and opening directions. confirm that here most of the strain (70% of total) is accommodated by left-lateral faulting. [27] Widespread E-W right-lateral faults have been often observed throughout TG. These structures are considered as conjugate systems of the NW-SE left-lateral faults Structure of Manda Hararo Rift [28] The Manda Hararo Rift constitutes the currently active axial zone of the Tendaho Graben. The part of MHR studied here is located along the southern termination of the basaltic plateau surrounded by sedimentary and subordinate volcanic deposits (Figure 5). While the plateau is highly fractured, with fault scarps reaching several meters of vertical displacement, the surrounding deposits do not appear to be faulted or fractured. [29] The basaltic plateau can be divided in two portions. To the west, despite the evident faulting, the lavas have an overall horizontal attitude (Figures 6a and 6b). If the basaltic layers are tilted, this typically occurs in the immediate hanging-wall of open normal faults, and decreases in 9of17

10 Figure 6. (a) Extension fracture within subhorizontal basalts in the axial part of southern MHR (site 10, Figure 5). (b) Graben-like structure, bordered by open faults with tilted hanging-wall within subhorizontal basalts in the axial part of MHR (site 37 in Figure 5). (c) Bookshelf faulting and SWdipping normal fault scarps in the eastern part of MHR (east of site 26 in Figure 5). (d) NW-SE trending open fault separating the subhorizontal basalts of the axial side of MHR (right side) from the tilted blocks, interrupted by antithetic fault scarps of the rift side (site 26 in Figure 5). The tilted hanging-wall of the open fault coincides with the first tilted block; the car (within the circle) provides a scale. (e) Sketch along a section perpendicular to the MHR axis, illustrating the overall structure of the rift, characterized by graben-like structures, formed by open faults, and extension fractures within horizontal deposits (axial side), and tilted blocks with antithetic normal faults to the east. The open fault in Figure 6d separates the two domains. 10 of 17

11 Figure 7. Variation of the stretching factor b in the axial part of southern MHR with distance along the rift axis. The dashed line is the inferred continuation of the best-fit line, suggesting the termination of the axial zone 10 km to the south of the area in Figure 5 (Dubti area in Figure 1b). importance a few m from the fault (Figures 5b and 6b). The faults often form opposite-facing scarps, between 1 to 10 m high, arranged in a graben-like fashion (Figure 6b). Extension fractures, up to a few hundreds of m long and a few m wide, are usually found in these graben (Figure 6a). The eastern part of the basaltic plateau is largely characterized by north-eastward dipping tilted blocks, decreasing in dip, over a distance of several km, between 25 and 5. These blocks are interrupted by regularly spaced SWdipping normal faults, forming scarps several tens of m high, whose displacement decreases north-eastward over a distance of a few km (Figures 6c and 6d). Tension fractures are lacking. Younger volcanic activity is sometimes present [Lahitte et al., 2001]. Subhorizontal and tilted domains are juxtaposed along a NW-SE trending and NE-dipping normal fault, with a dilational component of a few tens of m (Figure 6e). The overall architecture of the foot-wall (flat layers) and hanging-wall (domino decreasing with distance) of the normal fault separating the eastern and western plateau is remarkably similar to that observed at the borders of TG (section 4.2). [30] The opening direction of the normal faults and extension fractures in MHR is reported in Figures 3 and 4 and has been discussed, in general terms, at the end of section 4.2. The data from the MHR are consistent with an almost pure extension, even though these faults exhibit a moderate left-lateral component. Neglecting any strike-slip component, it is possible to estimate the total amount of extension along profiles perpendicular to the rift axis, in the subhorizontal western part of the plateau (Figure 5a). For this purpose, all of the horizontal components of displacement of the faults (using a mean dip of 80 in the axial zone) and amount of opening of extension fractures have been estimated along 4 profiles. Section 1 (northernmost) gives a total extension of 76.7 m (b 1 = ), section 2 (to the south) an extension of 33.2 m (b 2 = ), section 3 (further to the south) an extension of 27.4 m (b 3 = ) and section 4 (southernmost) an extension of 9.5 m (b 4 = ). Stretching factors decrease along this portion of MHR with the SE termination of MHR (i.e., b = 1 at 10 km to the south of section 4, immediately south of the study area) (Figure 7). [31] Because of this southward reduction of stretching factor, the approximate amount of extension measured along the northernmost section (section 1; Figure 5a) may be representative for MHR. This hypothesis is supported by the fact that the portion of the MHR to the north of the investigated area has a similar structure to that along section 1 (Figure 3). Given the mean age of the southernmost basaltic plateau = 169 ± 102 ka [Lahitte et al., 2001], we obtain a mean extension rate across the axis of southern MHR = 0.7 ± 0.4 mm/yr. To evaluate the amount of extension in the bookshelf faulting area to the east, we applied equation (3) to section 5 (south) and section 6 (north) (Figure 5). In section 5 one obtains a stretching b 1.05, equivalent to 190 m of extension. These, added to the extension of the nearby section 3, give a total extension of 217 m across MHR, equivalent to an extension rate (over 169 ± 102 ka) of 1.3 mm/yr. For section 6, one obtains a stretching b 1.04, equivalent to 110 m of extension, which when added to extension on the nearby section 1, gives a total extension of 187 m across MHR, at a rate (over 169 ± 102 ka) of 1.1 mm/yr. Therefore unlike the spreading rates in the axis of MHR, the two spreading rates along the southern part (axial + bookshelf faulting parts) of MHR are quite constant and consistent with a mean value of 1.2 ± 0.1 mm/yr. The fact that the structure and width of MHR rift to the north do not vary significantly (Figure 3), suggests that these values may be representative of a significant portion of MHR. The volume of basaltic magma erupted along the southern portion of MHR, associated with these spreading rates, is consistent with an extrusion rate of 600 km 3 /Ma. 5. Discussion 5.1. Geometry and Kinematics of Tendaho Graben and Manda Hararo Rift [32] Bookshelf faulting accommodates extension along the TG and MHR. The field data, matched with a previously 11 of 17

12 Figure 8. Possible model for the development of dominos at the sides of TG and MHR. (a) Initial undeformed stage. (b) An open fault, with tilted hanging-wall, forms as a result of crustal tension and subsidence. (c) Increased strain forming an antithetic normal fault, tilting the hanging-wall of the open fault further. (d) Prolonged strain inducing the migration of the antithetic normal faults and block tilting to the right. Displacements are not to scale; dashed lines indicate the topography of the previous stage. proposed mechanism of formation of normal faults along divergent plate boundaries [Gudmundsson, 1992; Acocella et al., 2003], suggest a model for domino development in MHR and TG (Figure 8). The fact that the highest tilt of the domino blocks coincides with the drag in the hanging-wall of the open normal fault (Figure 6e) suggests that these dominos result from the development of an open normal fault. This occurs when growing extension fractures, formed at the surface, reach a critical depth of a few hundreds of m, becoming shear fractures, or normal faults [Gudmundsson, 1992]. These faults are associated with the tensile stresses induced at the surface by the subsidence of a portion of the crust. The drag in their hanging-wall results from the flexural response to subsidence (Figure 8b). If the strain increases, an antithetic normal fault (without opening) may form in the hanging-wall of the open fault, locally increasing its tilt (Figure 8c). A further increase in the strain may form an additional antithetic normal fault, repeating the process (Figure 8d). In synthesis, subsidence induces normal faulting and widens the domino area. Subsidence may result from the crustal thinning at depth and/or emptying of shallow reservoirs due to magma extrusion. The migration in the development of the antithetic faults toward the more subsided area is inferred from the progressively lower vertical displacements observed (Figures 3 and 6); however, a migration in the opposite direction is also possible. Accordingly with this mechanism, the open normal fault may be considered the master fault, whose displacement includes its original hanging-wall tilt, at the margin of the extending zone. [33] The kinematics of the faults that form the margins of TG appears complex. An extensional component is evident from the topographic variations (200 m) across the border faults and the 1600 m of downthrown of the Afar Stratoids within TG [Battistelli et al., 2002]; this is confirmed by the widespread dominos, with blocks tilted about horizontal axes. Nevertheless, the kinematic data at both sides of TG are consistent with some left-lateral motion of the NW-SE trending faults (Figure 4); these data are also in agreement with focal mechanisms of earthquakes, including the M = 5.9 Serdo event in 1969 (Figure 3 [Hofstetter and Beyth, 2003, and references therein]). The apparent discrepancy between the geometry of the hanging-wall and footwall of the NW-SE trending faults (implying significant extension) together with their kinematics observed in the field and from seismicity (implying a significant left-lateral component), may be reconciled by postulating that the NW- SE systems were generated as normal faults during the opening of TG, and subsequently reactivated as strike-slip faults. Similar observations have been made to the north in the southern Red Sea [Ghebreab et al., 2002]. This hypothesis is also similar to that previously proposed for bookshelf faulting and rotations along NW-SE trending left-lateral faults within the interacting Manda Hararo and Manda Inakir rifts [Tapponnier et al., 1990]. These data do not support the inferred main extension along the NW-SE trending faults on the NE side of TG [Souriot and Brun, 1992]. However, conversely to Tapponnier et al. [1990], our data suggest that the strike-slip reactivation of the NW-SE trending faults included also the SW margin of TG. These results are consistent with previous structural data [Abbate et al., 1995], confirming the left-lateral slip to the SW of TG, as well as the possible importance of strike-slip motions at triple rift junctions [Abbate et al., 1995; Mouslopoulou et al., 2007]. In fact, the left-lateral faults on the SW margin of TG may result, in addition to the overlapping mechanism [Tapponnier et al., 1990], from the progressive variation in extension in the area between MHR and MER [Mouslopoulou et al., 2007]. Here the extension direction may rotate from NE-SW (MHR), to N-S (south of TG), to WNW-ESE (MER), being locally (south of TG) consistent with the left-lateral motions on NW-SE faults. [34] In addition to any strike-slip component, both TG and MHR display a clear extension, estimated through a mean b 1.1 (corresponding to 3.6 mm/yr; section 4.2) 12 of 17

13 Figure 9. Evolution of the TG area on the southern Red Sea propagator (same area shown in both maps). (a) A spreading center (here named Tendaho Rift), with extension rates of 3.6 mm/yr, possibly caused the fissural emplacement of the Stratoids (age of Stratoids indicated by the shades of grey) and was active in the NE part of the present TG between 2 and 0.2 Ma. (b) In the last 0.2 Ma, tectonic activity along the Tendaho Rift vanished, with sporadic eruptions at selected points; the southern MHR formed, with spreading rates of 1.2 mm/yr. and b 1.05 (corresponding to 1.2 mm/yr; section 4.3) respectively. During activity of MHR, no evidence of extension occurred within TG, characterized by strike-slip faulting. Therefore a discrepancy in the amount and rate of extension in the area of TG is evident, with higher rates during the development of TG and lower rates during the development of MHR. The boundary between the two extensional episodes cannot be easily detected. In fact, MHR formed in the last 169 ± 102 ka, or roughly 0.2 Ma. The above-mentioned TG spreading rate is a mean value obtained in the last 1.4 Ma (section 4.2). Given the significantly lower rates of extension in the last 0.2 Ma, this spreading rate is probably underestimated and may refer more realistically to the period 1.4 to 0.2 Ma Tectonic and Magmatic Evolution of Tendaho Graben [35] The onset of development of TG (1.8 Ma [Acton et al., 1991, 2000, and references therein; Audin et al., 2004]) is roughly coeval to the mean age of the Stratoid deposits in the area (1.4 ± 0.8 Ma). This suggests a relationship between the emplacement of the relevant volumes of Stratoid magma and moderate extension (>3 mm/yr). Moreover, there is an overall progression in the age of the Stratoid deposits, becoming younger toward the NE margin of TG (Figure 1c [Kidane et al., 2003, and references therein]). This, together with the lack of significant erosion [Rouby et al., 1996], suggests that the Stratoid in the TG area may have been erupted from within the NE portion of TG. The fact that the variation of the Stratoid ages is broadly consistent throughout the length of TG (Figure 1b) suggests a fissure-like mechanism of eruption, rather than a central or aerial distribution of vents (Figure 9a). A presentday analogue, even though restricted to a single episode, may be the 2005 fissure eruption in the northern MHR, resulting from the emplacement of a 8 m-thick dike along 60 km of rift [Wright et al., 2006]. On the basis of this analogy, we speculate that the relevant rifting episode forming TG may have been induced by the generation and rise of a significant amount of magma, leading to the repeated fissural emission of significant volumes of Stratoids, between 1.8 and 0.6 Ma. These Stratoid volumes, roughly estimated considering their aerial extent in the TG area ( km) and thickness in drillings (1 km [Battistelli et al., 2002]), are consistent with 7000 km 3 /Ma of magma in the TG area. This period of development of the proto-tendaho Graben, or Tendaho Rift, marks therefore an important step in the tectonic and magmatic build-up of Central Afar (Figure 9a): an exceptional amount of magma (several thousands of km 3 /Ma) was extruded from a moderately extending crust (i.e., stretching of b 1.1 and extension rate of 3.6 mm/yr). Similar results, highlighting the predominant role of magmatic processes over tectonic processes in the development of continental rifts, have been obtained for MER [Kendall et al., 2005], Red Sea [Buck, 2006] and the northern part of Manda Hararo Rift, several 13 of 17

14 tens of km to the north of the studied area [Wright et al., 2006]. [36] The location of the youngest Stratoid deposits, in the NE part of TG (Figure 1b), broadly coincides with the axis of Tendaho Rift, defined by the attitude of the same deposits (Figure 5). Moreover, the rift axis is also coincident with the younger (<0.22 Ma) volcanic deposits of Kurub, Manda Gargori and Dama Ale (Figure 2). This suggests a spatial relationship between the inferred area of emission of the Stratoids, a part of the more recent volcanism and the structure of TG (Figure 9b). We speculate that the rift axis of TG may be the surface expression of a subsided area related to a previous phase of magma withdrawal. The Tendaho rift axis may have been the focus of subsidence, spreading and volcanism from 1.8 to 0.6 Ma (Figure 9a). From 0.2 Ma, volcanic activity was restricted to selected spots (Kurub, Manda Gargori and Dama Ale; Figure 1b), while the new focus of spreading, with moderate extension rates and eruptive volumes, migrated to MHR. In this setting, the Kurub, Manda Gargori and Dama Ale volcanic system is interpreted to result from the gradual termination of the former Tendaho Rift, rather than being the southern continuation of the Manda Hararo Rift. Therefore it appears that 2 main offset and non-overlapping spreading axes have been active, in part simultaneously, within TG in the last 1.8 Ma, in a context of decreasing volcanic and tectonic activity (Figure 9) Evolution of the Southern Red Sea Propagator of Central Afar [37] Tectonic and magmatic activity along the southern Red Sea propagator in Central Afar occurred (in the last 1.8 Ma) in the Tendaho Graben and, subsequently, Manda Hararo Rift. In particular, our data on MHR and TG suggest that the activity of the southern Red Sea propagator was characterized, between 1.8 and 0.2 Ma, by moderate extension and significant magmatism. However, both decreased significantly, in the last 0.2 Ma, along MHR. Such an evolution of the southern Red Sea propagator in Central Afar, in the last 2 Ma, may be related to that of the Aden propagator, in eastern Afar. In fact, the on-land propagation of the Aden Rift was characterized by the development of the Asal-Ghoubbet Rift, between 0.9 and 0.2 Ma, and subsequently, in the last 0.2 Ma, of the Manda Inakir Rift (Figure 1a [Manighetti et al., 2001, and references therein]). The only available estimate of the extension rate of the Manda Inakir Rift, of 6 km, was given by Schaefer [1975]. Since the estimates of Schaefer for the amount of extension across TG are consistent with ours, the estimate on the Manda Inakir Rift may also be reliable. Considering the age of Manda Inakir Rift (0.2 ka [Manighetti et al., 2001, and references therein]), its extension rate is 3 cm/yr. This value is larger than the Somalia-Arabia spreading velocity (1.1 cm/yr [Stein et al., 1991; Manighetti et al., 1998]) and is probably an overestimate. Nevertheless, as a similar spreading rate (17 29 mm/yr) has been calculated for the Asal-Ghoubbet Rift in the last 150 ka [De Chabalier and Avouac, 1994, and references therein], we maintain that a significant component (cm/yr or more), may characterize the spreading of Manda Inakir and Asal Ghoubbet Rifts in the last 200 ka. If this holds true, the on-land Aden Rift has had spreading rates nearly one order of magnitude higher than those of Tendaho Rift, and even more than MHR. [38] These considerations suggest the following evolutionary model for rift development in Central Afar (Figure 10). Between 2 and 0.6 Ma, most of the strain in Central Afar was accommodated by the Red Sea propagator. This branched to the south of Erta Ale in the incipient Tat Ali Rift [Lahitte et al., 2001] and the already active Tendaho Rift, in continuity with the northern part of the propagator (Paleo MHR segment in Figure 10a). At around Ma the Asal-Ghoubbet Rift was at an incipient stage (Figure 10a [Manighetti et al., 2001]). Between 0.6 and 0.2 Ma, the developed Asal-Ghoubbet Rift started to propagate northwards, at rates of 1 cm/yr (Figure 10a [Manighetti et al., 2001]). Contemporaneously, the emplacement of the Afar Stratoid sequence ended in the area of TG. The Tat Ali Rift developed. Between 0.2 Ma and Present, activity along the Tat Ali Rift decreased, while the present MHR formed and the Tendaho Rift showed sporadic minor volcanic and tectonic activity; the Asal- Ghoubbet Rift propagated further, forming the more active Manda Inakir Rift (Figure 10c). Significant tectonic activity (extension rate 1.1 cm/yr [Jestin et al., 1994]) along the Red Sea propagator is currently limited to northern Afar, outside the overlapping zone. [39] Therefore it appears that the general decrease in volcanic and tectonic activity of the southern Red Sea Propagator from 0.6 Ma (final emplacement of the Stratoids) is related to the on-land propagation of the more active Aden Rift. It is not possible to determine whether the latter has been a stronger spreading system, capable of accommodating most of the extension in Central Afar, or its higher activity has been enhanced by the decrease in activity of the southern Red Sea Propagator. The general lines of our evolutionary model are consistent with those of Audin et al. [2004], suggesting that spreading in Central Afar mainly took place along one active propagator at any one time, first with the southern Red Sea (approx Ma) and subsequently with the Aden propagator (approx. 0.6 Ma-Present). [40] One implication of this scenario is that the Red Sea and Aden propagators never really overlapped, at least with comparable tectonic rates, in Central Afar in the last 2 Ma. In fact, when the Red Sea propagator reached its southern end, the Asal-Ghoubbet rifting was incipient (Figure 10b), whereas when the Aden rift finally propagated on-land, the extension on the southern Red Sea propagator started to decrease (Figure 10c). Therefore the overlap between spreading centers and the related bookshelf faulting may be due to the alternation of extension between rift segments, rather than resulting from consistent spreading rates along both segments. 6. Conclusions [41] The tectonic and magmatic evolution of the southern Red Sea propagator in Central Afar in the last 1.8 Ma 14 of 17

15 Figure 10. Evolutionary model of rift propagation in Central Afar in the last 2 Ma. (a) Between 2 and 0.6 Ma the Tendaho Rift, on the southern Red Sea propagator, accommodated most of the extension in Central Afar. Toward the end of this period, the on-land portion of the Aden propagator was incipient. (b) Between 0.6 and 0.2 Ma the Tendaho Rift decreased its volcanic (and probably tectonic) activity, whereas the developed Asal-Ghoubbet Rift, the on-land propagation of the Aden Rift, migrated northwards. (c) Between 0.2 and Present MHR replaced the spreading, even though with significantly lower rates, of Tendaho Rift, whereas the Aden Rift further propagated, forming the Manda Inakir Rift, characterized by higher spreading rates. The significant activity along the Red Sea propagator is currently limited to northern Afar [Jestin et al., 1994]. mainly results from the activity of Tendaho Graben and, subsequently, Manda Hararo Rift. [42] The Tendaho Graben has a stretching factor b 1.1 and an extension rate of 3.6 mm/yr. It mainly developed between Ma, contemporaneously to the Afar Stratoids in the area, which become older toward the TG sides. This suggests that TG may have been the site of fissural eruption of part of the Stratoids, with output rates of 7000 km 3 /Ma. [43] The Manda Hararo Rift marks the active site of extension within TG in the last 0.2 Ma. Before terminating in the Dubti area, MHR has a stretching factor b 1.04 and extension rate 1.2 mm/yr, with a magmatic production in the order of 600 km 3 /Ma. [44] While MHR undergoes almost pure extension, with a weak left-lateral component, the TG sides are characterized by significant left-lateral slip. This kinematic variation is interpreted to result both from the bookshelf faulting in Central Afar and from the progressive variation of the extension direction from MER to MHR. [45] An exceptional amount of magma extruded from a moderately extending crust along Tendaho Graben between 1.8 to 0.6 Ma. However, in the last 0.2 Ma, magmatic and tectonic activity, confined to MHR, significantly decreased. The general decrease in activity of the southern Red Sea propagator in the last 0.6 Ma coincides with the on-land appearance of the more active Aden propagator, suggesting that spreading in Central Afar took mainly place along one active propagator at any one time. [46] Acknowledgments. Melkamu Taddese participated in the field work. The Ethiopian Army and Police provided security during the part of the field work. Marco Bonini, Agust Gudmundsson, Andy Nicol and two anonymous reviewers provided constructive and helpful reviews, improving the paper substantially. M. Mattei and R. Bialas provided helpful suggestions. Funded by Miur funds (Progetto Ethiopia, F. Barberi responsible). References Abbate, E., P. Passerini, and L. Zan (1995), Strike-slip faults in a rift area: A transect in the Afar Triangle, East Africa, Tectonophysics, 241, Acocella, V., and T. Korme (2002), Holocene extension direction along the Main Ethiopian Rift, East Africa, Terra Nova, 14, Acocella, V., T. Korme, and F. Salvini (2003), Formation of normal faults along the axial zone of the Ethiopian Rift, J. Struct. Geol., 25, Acocella, V. (2006), Regional and local tectonics at Erta Ale caldera, Afar (Ethiopia), J. Struct. Geol., 28, Acton, G. D., S. Stein, and J. F. Engeln (1991), Block rotation and continental extension in Afar: A comparison to oceanic microplate systems, Tectonics, 10, Acton, G. D., A. Tessema, M. Jackson, and R. Bilham (2000), The tectonic and geomagnetic significance 15 of 17