Tectonophysics Constraints on fault and crustal strength of the Main Ethiopian Rift from formal inversion of earthquake focal mechanism data

We evaluate the frictional strength of seismogenic faults in the Main Ethiopian Rift (MER) by inverting the available, well-constrained earthquake focal mechanisms. The regional stress ﬁ eld is given by − 119.6°/77.2°, 6.2°/7.6°, and 97.5°/10.2° for trend/plunge of σ 1 , σ 2 and σ 3 , respectively agrees well with previous fault kine- matic and focal mechanism inversions. We determine the coe ﬃ cient of friction, μ , for 44 seismogenic faults by assuming the pore pressure to be at hydrostatic conditions. Slip on 36 seismogenic faults occurs with μ ≥ 0.4. Slip on the remaining eight faults is possible with low μ . In general, the coe ﬃ cient of friction in the MER is compatible with a value of μ of 0.59±0.16 (2 σ standard deviation). The shear stresses range from 16 to 129MPa, is similar to crustal shear stress observed in extensional tectonic regimes and global compilations of shear stresses from major fault zones. The maximum shear stress is observed in the ductile crust, below the seismologically determined brittle-ductile transition (BDT) zone. Below the BDT, the crust is assumed to be weak due to thermal modi ﬁ cation and/or high pore ﬂ uid pressure. Our results indicate linearly increasing μ and shear stress with depth. We argue that in the MER upper crust is strong and deforms according to Coulomb frictional-failure criterion.


Introduction
Thorough understanding of the regional stress field is of paramount importance in constraining the strength of faults and the crust along actively deforming plate boundaries. In general, fault zones are relatively weaker than the surrounding stable crust (Zoback et al., 1987). However, there are still questions regarding whether the strength of individual faults in a deformed region is the same or varies, and also whether the strength of the fault varies along strike on the same fault. In a recent study, Floyd et al. (2016) showed that the frictional parameters of a fault may vary over only a few kilometers distance depending on lithological controls. Fault strength varies not only spatially but also temporally. By reviewing about 300 experimental results, Di Toro et al. (2011) suggested that the strength of faults could be reduced during earthquakes.
Coefficient of friction, μ, can be used as a proxy to model the strength of faults and crust. Laboratory estimates of μ are usually high, ranging from 0.6 to 0.8 (Byerlee, 1978) although lower μ values are reported (using rotary shear apparatus) from the San Andreas fault (e.g., Carpenter et al., 2015). Some researchers argue that if these μ values really exist during seismic slip, we should have found high heat flow and extensive melting (e.g. pseudotachilite) along exhumed faults (e.g., Mulargia and Bizzarri, 2016). The apparent absence of these rock types is used to argue that coefficient of friction is lower than laboratory estimates. In line with this, numerical modeling studies in active tectonic areas, e.g. in East African Rift (EAR) (Bird et al., 2006), argue that the friction coefficient is much lower than the laboratory estimates. This agrees with recent laboratory experiment that found μ to be below 0.4 under pressure condition equivalent to a depth of ∼15 km (Di Toro et al., 2011). Mulargia and Bizzarri (2016) proposed a multi-stage earthquake process which involves a high friction coefficient with μ ∼ 0.7 during the first stage. The high friction stage induces high temperature that almost immediately induces fluid pressurization and reduces the permeability of fault gouges and subsequently the pore pressure reaches lithostatic state. In this case, μ drops to ∼0.2. Other, equally important fault lubrication mechanisms include melt lubrication (Di Toro et al., 2006), gelification (Di Toro et al., 2004) and decarbonation (Han et  2010), which are also proposed for the reduction of μ during seismic slip. Stamps et al. (2015) argue that buoyancy forces, weak continental faults and Couette-type mantle flow in the asthenosphere explain extension across the EAR. In that study, weak fault friction results in better fit between geodynamic model and present-day geodetic observations in Africa (Saria et al., 2014).
In the Main Ethiopian Rift (MER) (Fig. 1), the crust and faults are assumed to be weakened by hotspot tectonics over the last ∼30 Ma (e.g., Pik et al., 2006), including ongoing dyke intrusion (e.g., Keranen et al., 2009;Wright et al., 2006;Beutel et al., 2010). We constrain the coefficient of friction and stress magnitudes in the actively deforming MER. Our approach is that first we invert the focal mechanisms and follow basic assumptions of finding the shear and normal stresses using tensor transformation, acting on optimally oriented seismogenic faults and subsequently determine the coefficient of friction, μ. Our results show spatial variation of μ along the strike of the MER and the shear stress variation in the crust. The results presented will contribute to the understanding of the stress magnitude at the earthquake focal depths where in situ measurements are totally absent and the frictional strength of MER faults and crust.
2. Tectonic setting, focal mechanism data and regional stress field

Tectonics of the Main Ethiopian Rift
The Main Ethiopian Rift forms an active plate boundary between the Africa (Nubia) and Somalia plates in the northern EAR. Starting from ∼18 Myrs, the MER is thought to have initiated asynchronously along its length (Wolfenden et al., 2004). The asynchronous development of the different sectors of the rift potentially influences the melt production and strain accommodation mechanisms (Keir et al., 2015;Muluneh et al., 2017). The MER orientation is influenced by the Neoproterozoic, Pan-African suture zone that runs through Ethiopia, and significantly influences the orientation of faults (e.g., Agostini et al., 2011), seismic anisotropy (e.g., Gashawbeza et al., 2004) and also crustal thickness (e.g., Keranen and Klemperer, 2008). During the past ∼2 Myrs, deformation in the northern MER focused along 20 km-wide, 60 km-long magmatic segments arranged in an en echelon manner within mid-Miocene half-graben basin (Ebinger and Casey, 2001). The transfer of strain from border faults to the magmatic segments could be due to rift obliquity (Corti, 2008), weakening of the crust by release of magmatic fluids , or by localizing extensional stresses in strong crust within the rift in which the increase in the rock strength is due to metamorphic reactions in rocks intruded by ascending magmas (Lavecchia et al., 2016), and/or by solidification of new mafic material (e.g., Beutel et al., 2010).

Earthquake focal mechanism data
Earthquake data from the EAGLE catalogue  for the period 2001 to 2003 (Fig. 1a & b) shows that seismicity continues to the depth of 28 km with magnitude M L 0.0 to 4 ( Fig. 1a & b). We assume that the depth to the brittle-ductile transition (BDT) occurs at a depth of 16 km, above which 90% of seismicity occurs (Fig. 1b). This depth also coincides with the average depth to the base of the upper crust in the MER (Maguire et al., 2006).
We compiled 55 well determined earthquake focal mechanisms from CMT catalogues and published sources ( Fig. 2a) (Ayele, 2000;Hofstetter and Beyth, 2003;Ayele et al., 2006;Keir et al., 2006;Delvaux and Barth, 2010;Wilks et al., 2016) in the MER. Fig. 2b shows hypocentral depth of the focal mechanisms compiled and only two mechanisms are located in the lower crust. The focal mechanism data are strike, dip and rake of the fault planes. Most of the studies reported the fault planes except for Hofstetter and Beyth (2003) who reported both the auxiliary and interpreted fault planes. Distinction between fault and auxiliary planes is based on orientation of fault traces on the surface. Planar faults tend to have linear fault traces on the Earth's surface whereas listric fault planes tend to be arcuate in plan view and listric in 3-D with alternating and overlapping half grabens that change polarity along strike (Rosendahl, 1987). In the MER, the observed faults on the surface are linear segments with steep fault plane that led us to conclude that they continue to be planner at depth. In cases when multiple mechanisms for a single earthquake is reported, we use the most recent publication. The largest focal mechanism dataset comes from Keir et al. (2006) (33 out of 55 mechanisms) who assign a 2 sigma uncertainty of ± 20°to the focal parameters and ± 2 km for the hypocentral depth. Most focal mechanisms show dip-slip movements along the faults and slip with dips > 50°similar to the dip of recently active faults in the rift (e.g., Agostini et al., 2011).

Inversion method and regional stress field
The stress field can be quantified by formally inverting a group of earthquake focal mechanisms or active fault data. A least square inversion technique is developed to invert a group of diverse earthquake mechanisms by assuming the regional stress field is uniform (Michael, 1984). The inversion technique minimizes the difference between the slip vector and the resolved shear stress which can be used to assess the success of the inversion result. The resulting stress field quantifies the relative magnitude, ϕ, trend and plunge of principal stresses (σ 1 , σ 2 and σ 3 ; where σ 1 ≥ σ 2 ≥ σ 3 ). We used the inversion method of Michael (1984,1987) in order to estimate the regional stress field of the MER. The confidence region for the best fit stress tensor is calculated using the bootstrap resampling method. To estimate the 95% confidence limit we used 2000 repetitions (Michael, 1987). During the inversion process, we used the reported fault planes (Ayele, 2000;Ayele et al., 2006;Keir et al., 2006;Wilks et al., 2016) as they are. In cases when we encounter large angular misfit during the inversion process, we change the nodal plane until the misfit is reduced to an acceptable level. The nodal planes that reduce the angular misfit are the preferred fault planes (Michael, 1987). However, the individual misfit of each plane must be kept below 25° (Michael, 1991). Only two focal mechanisms show individual misfit of higher than 25°but below 30°. The observed regional stress field is given by − 119.6°/77.2°, 6.2°/7.6°, 97.5°/10.2°f or trend/plunge of σ 1 , σ 2 and σ 3 ( Fig. 3; Table 1). The quality of the inversion result is measured by the β value; a misfit measure that defines the angle between the observed and predicted rake angles for the mechanisms (Michael, 1984). Our inversion results an average misfit, β , of 10.3°± 7.6°. Although the inversion process suffers from the absence of diverse focal mechanisms, it results in similar trend and plunge of principal stress axes to other studies in the rift (e.g., Delvaux and Barth, 2010;Keir et al., 2006) and extension direction inferred from GPS observations (e.g., Saria et al., 2013). In addition, the focal mechanism dataset is a complete compilation of literature.
Several studies have been conducted in order to compare different inversion techniques (e.g., Kastrup, 2003;Delvaux and Barth, 2010). Irrespective of the methods used, the optimal solutions for the stress inversion are consistent and similar. Using 7 earthquake focal mechanism solutions from CMT catalogue in the MER, Delvaux and Barth (2010) conducted stress inversions using the method of Michael (1984Michael ( , 1987 and Delvaux and Sperner (2003). They showed that the orientation of the principal stress axes is almost the same with significant difference in stress ratio.

Absolute stress magnitudes
In areas where in-situ stress measurements are lacking, a number of  Result of stress tensor inversion using earthquake focal mechanisms in the MER. Circle, triangle and diamond symbols show σ 1 (largest compressive stress), σ 2 (intermediate), and σ 3 (least compressive stress), respectively. The 95% confidence level is estimated with bootstrap resampling (Michael, 1987). Inversion results in our study are shown by open symbols. Principal stresses from Delvaux and Barth (2010) using Delvaux and Sperner (2003) are shown by blue and using Michael (1984) by red filed symbols. Black diamond symbol shows σ 3 orientation from Keir et al. (2006). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) assumptions can be used to find the absolute magnitude of principal stresses at the earthquake focal depths and later used to quantify the strength of the crust. In order to quantify the shear and normal stresses and later the coefficient of friction, μ, on optimally oriented faults, information about the absolute magnitude of the principal stresses is required (Zoback, 1992). Earthquake focal mechanism inversion allows the determination of direction and relative magnitudes of principal stresses, σ 1 , σ 2 , and σ 3 . When one principal stress axis is oriented vertically (σ 1 in extensional regime), its magnitude is given by the overburden weight (e.g., Zoback and Zoback, 2002) where ρ is the density of crustal material (taken here as 2800 kg/m 3 ), g is acceleration due to gravity [∼10 m/s 2 ] and z is the focal depth [m]. Eq.
(1) shows that σ 1 increases in an approximately a linear fashion with depth. Based on limiting frictional strength of optimally oriented faults in the crust (e.g., Sibson, 1974;Zoback and Townend, 2001), the least compressive stress, σ 3 , can be estimated by where p 0 is the pore fluid pressure which is assumed to be hydrostatic and is given by λ × σ v where λ = 0.3737 (Albaric et al., 2009;Zoback, 1992;Fadaie and Ranalli, 1990); μ c is a static frictional coefficient ∼0.75 (Sibson, 1974;Jager and Cook, 1979) and is considered to represent the regional value for absolute stress determination (Kastrup, 2003). From Eq. (2), it is evident that p 0 can not exceed σ 3 without the occurrence of hydraulic fracturing, although λ values greater than 1 are possible without causing hydraulic fracturing if total principal stresses within the fault zone are magnified due to contrasts in rheological properties (e.g., Rice, 1992).
where θ is the angle σ 1 makes with the fault plane, gives the dip angle of ∼ 63°for optimally oriented normal faults (Sibson, 1974). This angle is the typical dip of normal faults (Agostini et al., 2011) and the interpreted fault planes of earthquake focal mechanisms  ( Fig. 4a & b) in the MER. Fig. 3 shows that σ 1 is near-vertical and fulfills the assumption of vertical maximum stress in order to determine absolute stress magnitude.
Finally an additional constraint to the determination of absolute magnitudes is given by in which ϕ value is independently determined from the inversion of earthquake focal mechanisms. Eqs.
(1)-(3) are adequate to estimate the absolute magnitude of principal stresses.
To get the normal (σ n ) and shear (τ) stresses on individual fault plane, we need to transform the stress tensor through tensor transformation (e.g., Allmendinger et al., 2012). To quantify the frictional strength of the faults, the ratio of τ to σ̇n must be determined using the modified linear frictional sliding equation (Eq. (4)) where the cohesion is assumed to be close to zero (Zoback, 1992, and references therein) where μ is considered to represent the frictional strength of the earthquake faults. We assume that the pore pressure to be hydrostatic and we vary μ to match the normal and shear stresses.

Coefficient of friction, μ
The ratio of shear to effective normal stresses give μ values on each fault (Fig. 5). Fig. 6a & b shows μ values plotted in the MER. We estimated the frictional parameters (μ, normal and shear stresses) for 44 well constrained earthquake focal mechanisms in the MER. Earthquakes used in the this study together with calculated σ n , τ and μ are included as a supplementary material to this paper.
Most faults in the MER fail under high frictional stress with μ ≥0.3. A regression of all the μ values (Fig. 5) indicates the rift deforms according to laboratory-determined friction coefficients (Byerlee, 1978).
The average value for the crust is computed using An exception to the relatively high μ values for most of the data, is that six focal mechanisms have μ ranging from 0.21 to 0.3 (Fig. 7a). There are potentially several explanations for the low μ results. We observed a cluster of 3 earthquakes near the axis of the northernmost MER with μ < 0.3 (Fig. 7a). These earthquakes are spatially and temporally coincident with the intrusion of a dyke near Amoissa volcano during May 2000, and interpreted to be induced by stress change above the new intrusion (Keir et al., 2011). Two strike-slip faults of M L 1.4 and 1.55 beneath Fentale volcano make angles of 64°and 56°to σ H , higher than the angle expected for reactivated strike-slip faults, and as a result slip with μ of 0.23 and 0.3, respectively. Weak faults such as the San Andreas Fault fail by creeping with small magnitude earthquakes (M w ∼ 2) (Nadeau and Guilhem, 2009). The low frictional strength and low magnitude earthquakes might explain failure by creeping along the two strike-slip faults.
An additional focal mechanism  with a selected nodal plane with a dip of 22°(selected during the inversion since this lowers the individual misfit to 6.8°) shows a μ of 0.21. The new dip angle deviates by ∼40°from optimally oriented normal faults (Fig. 4b) so that a lower μ is evident. Traditionally, normal faults with dip angle of < 30°were not thought to be able to lock and subsequently slip as earthquakes (Collettini et al., 2011), but emerging evidence of slip on low angle normal faults, e.g. in West Salton detachment faults (Prante et al., 2014) argues that low angle fault might slip through earthquakes. The geometry of this fault can be explained by very low μ, near # of fms = number of focal mechanism solutions; tr/pl -trend/plunge of the principal stress axes; ϕ measures the relative size of principal stresses. ± τ s.d. -the average angular misfit with standard deviation. High standard deviation might be due to a non-diverse range of focal mechanisms . fms = focal mechanism solutions, DB10* = Delvaux and Barth (2010) using Michael (1984) method, DB10 = Delvaux and Barth (2010) using Delvaux and Sperner (2003) method; D06 = Keir et al. (2006). lithostatic pore pressure or some combination of the two. Middleton and Copley (2013) noted that slip on such low angle normal faults possibly occurs due to the presence of weak materials along the preexisting fault surface. Lower strength faults at shallower depth could be due to the presence of clay minerals along the fault surfaces and high temperatures near active magmatic centers, such as beneath Fentale and Amoissa volcanoes (e.g., Keir et al., 2011).
The estimated μ for one mechanism at a depth of ∼19 km is ∼0.67 (Fig. 7a). The shear stress for the same mechanism is ∼70 MPa (Fig. 7b), lower than the maximum shear stress observed at the BDT (∼100 MPa) (Fig. 7b). In order for a slip to occur at higher μ and lower shear stress, the effective normal stress should be low which in turn implies higher pore pressure at this depth. This hints that the pore pressure in the lower crust might be in super hydrostatic or lithostatic condition.
Plotting the distribution of μ with depth ( Fig. 7a) indicates that μ generally increases upwards from the BDT. Below the BDT, μ decreases from the maximum value but is still high (0.65-0.7). Fig. 7a shows μ values cluster in two zones. At 0-10 km depth, the crust is mainly characterized by μ of 0.2-0.4. Below 10 km up to 16 km, the crust is generally characterized by higher μ (∼0.6-0.75). The high μ (∼0.6) observed for MER faults is also supported by the agreement between P-, B-, and T-axes of the earthquake focal mechanisms and the regional stress field (e.g., Keir et al., 2006). Zoback and Zoback (2002) argue that in regions for which μ ranges from 0.6 to 1, P-, B-, and T-axes approximate the average principal stress orientations.
Our estimate of μ for the MER crust deviates significantly from the frictional parameter inferred using numerical modeling studies in the EAR (Bird et al., 2006;Stamps et al., 2015). To produce a better fit with observed plate scale separation of Somalia from Africa, low fault friction is required (Bird et al., 2006;Stamps et al., 2015). We argue that discrepancy between our results and numerical modeling studies might be due to the input and model parameters used in the numerical modeling studies. Maggi et al. (2000) questioned the popular view of continental strength profiles in which a weaker lower crust resides between stronger upper crust and mantle. Spatial variation in continental strength of the lithosphere is mainly controlled by the presence or absence of smaller amounts of water (Maggi et al., 2000). Topography and gravity field analysis indicate that lithospheric stress is supported by upper crust overlying a weaker lower crust (Thatcher and Pollitz, 2008). Albaric et al. (2009) noted that the strength of lithosphere/crust can vary within the same tectonic setting. Even on a finer scale, Floyd et al. (2016) showed how the along strike variation of rheological properties controls the frictional strength of a fault. Here we discuss the frictional strength of the MER crust based on shear stress magnitudes estimated at hypocentral depth. The shear stress magnitude varies from 16.2 to 129 MPa with an average value of 60 MPa similar to crustal-average shear stress of 56 MPa (e.g., Bird, 1999). Similar to μ values, the shear  Tectonophysics 731-732 (2018) 172-180 stress also show a lower value (∼20 MPa) above 10 km and ranges from 40 to 100 MPa between 10 and 16 km. One earthquake fault fails with a shear stress of 129 MPa (Fig. 7b), at ∼20 km depth, higher than friction controlled shear stress for the upper crust. The shear stress variation with depth seems inconsistent with models in which significant decrease in strength occurs at the BDT (e.g., Chester, 1995) but rather shows an increase to a depth of ∼20 km, within the ductile, lower crust. This argument, however, is based on only one focal mechanism. A recent thermo-mechanical modeling study in the MER (Lavecchia et al., 2016) shows a transition from brittle to ductile rheology at a depth between 20 km and 25 km. On the other hand, geophysical studies show that the lower crust is weak (Keranen et al., 2009) and therefore the highest shear stress observed here might be due to the hydrostatic pore pressure imposed in our calculation. Fig. 7b shows that faults at depths shallower than the 10 km appear to be weaker than their equivalent at depth below 10 km. This agrees with the global compilation of stress-depth data (Behr and Platt, 2014) that middle crust sustains higher stress than the brittle faults above. Furthermore, the process that allows slip at lower shear stress (∼16-50 MPa, Fig. 7b) in the brittle crust terminates at 10 km depth probably due to higher temperature in the lower part of the upper crust.

Frictional strength of the crust
Our argument for strong crust and faults in the MER agrees well with the modeling studies in the region (Lavecchia et al., 2016;Beutel et al., 2010). Lavecchia et al. (2016) showed that the variations in mineral assemblages due to temperature increase during dyke intrusion and subsequent metamorphism locally increases the strength of the crust. This mainly occurs by changing weak minerals to strong minerals (Lavecchia et al., 2016). This agrees with the modeling results of Beutel et al. (2010) who noted an increase in the strength of the crust due to solidified mafic intrusions beneath magmatic segments. In regions of active magmatism and dyke intrusion where magma input is likely to be particularly high (i.e. near active volcanic centers), the crust appears to  (Wolfenden et al., 2004). The magmatic segments are characterized by higher μ (≥0.4) which could be due to melt-induced thermal metamorphism as shown by Lavecchia et al. (2016). However, low μ values are observed at or near active volcanic centers (red triangles). YTVL = Yerrer-Tullu Wellel Volcano-tectonic Lineament. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) A.A. Muluneh et al. Tectonophysics 731-732 (2018) 172-180 be weak as shown by lower μ values (Fig. 6b) in agreement with modeling studies in the EAR (e.g., Bialas et al., 2010;Daniels et al., 2014). The strong faults and crust inferred in our study directly influence kinematics of crustal blocks in the MER. Recent GPS study from EAR (e.g., Saria et al., 2014) shows significant deviation in Nubia-Somalia motion from the stable parts of the plates and crustal blocks in the MER. This points to argument that strength of the faults and crust controls the surface kinematics of the MER.
Strength of the faults and the crust are very sensitive to the pore fluid pressure considered (Sibson, 2000). Elevated pore pressures change the state of effective stress and reduce the force required for deformation to occur (Hubbert and Rubey, 1959). An experiment on fluid-rock interaction (Reynolds and Lister, 1987) showed existence of high fluid pressure in the ductile part of the crust. For the EAR, high pore pressure due to dehydration of metamorphic minerals in the lower crust is invoked (Seno and Saito, 1994). However, Keir et al. (2009) argued that in the absence of any accumulated fluid in the lower crust, earthquake activities are controlled by emplacement of melt supplied from the upper mantle into the lower crust. This argument is in line with geochemical evidence of water-poor magmas in the lower crust (Rooney et al., 2005). High pore fluid pressure and hence low strength of faults and of the crust makes a number of predictions (e.g., Scholz, 2000) including a low magnitude of shear stress and high angle between σ 1 and the fault plane. Although, there are no reported measurements on the magnitude of shear stress in MER, structural mapping on active faults and studying earthquake focal mechanisms show that σ 1 forms a low angle with the fault planes. The latter lends support to the interpretation that the faults in the rift are strong.

Hydrostatic vs lithostatic pore pressure
Pore fluid pressure is the most uncertain parameter in the calculation of the strength profile of the crust (e.g., Brace and Kohlstedt, 1980). Constraints on pore pressure in the MER are scarce. In order to assume the pore pressure is near lithostatic state, either the rock permeability must be very low (Nur and Walder, 1990) or an active source of overpressured fluids must exist at depth (Rice, 1992). Townend and Zoback (2000) and Barton et al. (1995) showed that critically stressed faults are hydraulically conductive and act like fluid conduits. These faults control the permeability of the crust and results in short diffusion time (10-1000 years over distances of 1-10 km). In extensional tectonic setting of the Taupo volcanic zone, high permeability (k > 10 −16 ) causes earthquake ruptures (Sibson and Rowland, 2003). This implies that fluid pressures in the crust equilibrate over relatively shorter time and hydrostatic pore pressure develops within the crust. Note that long diffusion time (> 10 5 years) leads to near-lithostatic pore pressure to be maintained (Nur and Walder, 1990).
In the previous sections, we showed that the MER faults are favorably oriented and hence facilitate the easy passage of fluids and supports the notion that the pore pressure in the region is near hydrostatic state. Recent CO 2 degassing study at Aluto volcano in the central MER estimated a total of 250-500 t d −1 CO 2 emitted along major faults and volcanic structures (Hutchison et al., 2015). Such emission rate is comparable to rates observed in the Eastern rift of the EARS . Furthermore, maximum CO 2 flux is observed adjacent to the faults implying that the permeability of faults in the MER is high and gives an easy access to the flow of fluids. Hutchison et al. (2015) also noted that the faults in the Aluto volcano penetrate deep and connect the reservoir to the surface. This provides a favorable condition for the fluid to reach hydrostatic state in a relatively short time period. Permeability values ranging 0.5-1.1 × 10 −12 m 2 are estimated from ∼2 km deep Aluto-Langano geothermal wells (Electroconsult, 1985). Deep penetrating faults in other parts of the EAR (e.g. in Tanzania) are known to act as pathways for CO 2 ascent (e.g., Lee et al., 2016). Seismicity recorded by CRAFTI array suggest that volatiles migrate from upper mantle or lower crust following extensional faults (e.g., Lee et al., 2016). However, it is questionable whether the permeability at upper crustal level is adequately representative of the deeper level or not (Sibson, 2000). Recent geophysical studies in the EAR (Lindenfeld et al., 2012;Weinstein et al., 2017) and Taupo rift (Reyners et al., 2007) showed that high pore fluid pressure induce faulting and seismic activities in the lower crust. Geophysical studies in MER (Keranen et al., 2009) and Tanzanian rift (Weinstein et al., 2017) indicated that the lower crust in these regions is weak and ductile and higher pore fluid pressure can provide a mechanism for brittle failure at lower shear stress. Our compilation of earthquake focal mechanisms in MER indicated that only two earthquake focal mechanisms occurred in the lower crust (below 16 km). For the two lower crustal earthquake focal mechanisms, we examine how increased pore pressure (from hydrostatic to super hydrostatic; i.e. 0.56 × σ v ) affects the strength of the faults. Using a mean μ of 0.59 and normal stress values (see the supplementary material), Eq. (4) results in the shear stress of 34 MPa and 3 MPa for focal mechanisms at depths of 20.29 km and 19.01 km, respectively. Increasing the pore pressure from hydrostatic to super hydrostatic conditions leads to significant reduction of shear strength of earthquake faults.
However, since most of focal mechanisms occur at depths above 16 km, the assumption of hydrostatic pore pressure can be considered as adequately representative of the state of pore fluid pressure in the upper crust and well depicts the strength of the crust and faults in the MER. We found both high and low μ values for earthquake faults in the MER implying that thermal modifications are more important than pore fluid pressure in modifying the strength of the faults in the upper crust.
Finally, further studies constraining the condition of pore fluid pressure in the crust are required in order to link fluid weakening and deep crustal earthquakes in MER.

Conclusions
Based on our findings, we reach the following conclusions: 1. Based on the orientation of optimally oriented faults controlling the flow of fluids in the rift and high CO 2 seepage rate, we argue that pore pressure in the upper crust is in near hydrostatic state. However the argument for hydrostatic pore pressure in the lower crust might be flawed as indicated by high μ, low shear strength fault and deep crustal seismicity. 2. Very low μ (≤0.3) values are observed near active volcanic centers, whereas high μ correspond to areas either at margins or in between magmatic segments implying that high pore fluid pressure might not be required for slip on weak faults. 3. Although data from the southern MER are scarce, there are no indications of weak faults, which could be in agreement with a less evolved rifting and lower magmatic modification in the area. 4. A best fit of μ values for MER faults indicate a regional value of 0.59 ± 0.16 implying the crust is strong under hydrostatic condition. 5. The strong upper crust contributes to the strength of the lithosphere in the MER, with the lower portion of the upper crust (10-16 km) being the strongest layer.
The strong faults and crust inferred in our study directly influence the kinematics of crustal blocks in the MER. Therefore, the conclusion we draw can be tested by future geodetic and modeling studies. also thank Prof. Jean-Philippe Avouac for his comments and editorial handling. The figures were drafted with Generic Mapping Tools (Wessel and Smith, 1998).