Knowledge of the seismic activity associated with the MER and adjacent NW plateau comes from both instrumental and historical records. Limited earthquake monitoring equipment in the early stages and sparse distribution of monitoring stations have resulted in an incomplete earthquake catalog for the region (Ayele et al., 2004; Ayele and Kulhánek, 1997; Hofstetter and Beyth, 2003; Kebede and Kulhánek, 1994). Regional and local instrumental studies reveal that seismicity is predominantly concentrated in the rift, and in particular most closely associated with major border faults, the Wonji Fault Belt (WFB) and volcanic centres (Kebede and Kulhánek, 1994; Ayele, 2005; Keir et al., 2006a; Asefa and Ayele, 2020; Ogden et al., 2021; Alemayehu and Asefa, 2023). Earthquakes occurrence in the region are characterized by frequent moderate to small-magnitude (Kebede and Kulhánek, 1994; Ayele and Kulhánek, 1997), and commonly occur in episodic swarms (Keir et al., 2006a; Greenfield et al., 2019; Raggiunti et al., 2023). Earthquake depths are mostly within the shallowest 20-km of the crust, with the exception of rare lower crustal microseismic swarms interpretted to be induced by motion of deep magma or volatile fluids (e.g., Ayele et al., 2024; Muluneh et al., 2021). Focal mechanisms from previous studies (Brazier et al., 2008; Hofstetter and Beyth, 2003; Keir et al., 2006a) reveal normal faulting in the MER, which is consistent with the extensional tectonic regime.

In contrast to the seismicity of the MER, the seismicity beneath the NW plateau is less understood. Keir et al. (2009) investigated the seismicity of the MER during 2001–2003 (project EAGLE–Ethiopia Afar Geoscientific Lithospheric Experiment), but also recorded earthquakes outside of the rift. The study showed clusters of seismicity beneath the Yerer-Tullu Wellel Volcanic Lineament (YTVL) which is on the rift margin between the MER and NW plateau. Some of the earthquakes are reliably located in the lower crust and interpreted as induced by magma migration (Keir et al., 2009). In addition, Brazier et al. (2008) analyzed earthquake activity that occurred during the 2000–2002 project period as part of the Ethiopia Broadband Seismic Experiment (EBSE) and identified minor seismic activity along the northwestern edges of the NW plateau.

Pre-instrumental earthquake observations from various sources spanning over 500 years have been compiled, revealing a significant history of seismic activity (Gouin, 1979). Notable events included the 6.8 (M_W) Langano earthquake in August 1906, which was felt in Addis Ababa and its surroundings. Ayele and Kulhánek (2000) relocated this event to the eastern shoulder of the MER with a magnitude of M_W = 6.6. The NW plateau experiences notable seismic activity, particularly along its eastern escarpment. Historical earthquakes, such as the 1938 (M 5.6, M 4.5) and 1942 (M 5.3–5.9) events, occurred near Debre Berhan and the plateau-Afar escarpment, indicating active faulting. Durham University deployed seismic network in 1974 in the Afar region and recorded 106 low-magnitude earthquakes in the NW plateau (39.5°E−40°E, 10°N-13°N), with seismic clustering in the Guf graben and northwest of Debre Sina. Additionally, a 1971 swarm of 151 microearthquakes (M ∼3.8) near Debre Berhan revealed ongoing fault movement. This seismicity is linked to extensional tectonics along the Plateau-Afar escarpment, where crustal adjustments continue to shape the region.

Previous geophysical studies imaged mantle and crustal seismic structure beneath the MER and adjacent NW plateau, focusing mainly on seismic wavespeeds and crustal thickness. Based on receiver function evaluations, the Moho depth below the MER is between 27 and 38 km, and Poisson’s ratios are between 0.27 and 0.35 (Dugda, 2005). According to surface and body wave research, the greatest Moho depth beneath the Addis Ababa region is 48 km (Searle and Gouin, 1971). Studies of seismic velocity, such as ambient noise tomography (ANT) reveal especially slow S-wave velocities (Vs.) in the lower crust beneath the MER. The observed Vs. (∼3.20 km/s) is suggestive of partial melt influence (Chambers et al., 2019; Ogden et al., 2019).

The Ethiopian plateau is characterized by thicker and relatively stable crust, and shows distinct geophysical features compared to the MER. Receiver function analyses estimate Moho depths of 36–44 km beneath the NW plateau (Kibret et al., 2022). S-wave velocity studies reveal a heterogeneous crustal structure, with low velocities (∼3.6 km/s) in the lower crust of some regions such as Lake Tana (Alemayehu et al., 2023). The uppermost mantle beneath the plateau exhibits Vs. values of 3.9–4.3 km/s, reflecting presence of a thermal anomaly and partial melt (Eshetu et al., 2021).

Previous studies highlight a lack of geophysical constraints due to absence of data on the patterns of seismicity and the mode of deformation in the NW plateau and its surroundings. To address these gaps a new passive seismic network consisting of thirty broadband seismograph stations was deployed in the Ethiopian plateau. This study aims to provide the first observation for understanding of the seismicity, focal mechanisms and the local stress condition covering a larger area and greater depth than previous attempts. We also performed focal mechanisms for selected earthquakes. The majority of the stations were fairly distributed across the NW plateau in order to acquire high resolution coverage. Here we investigate, for the first time, the seismicity, focal mechanisms and related local stress condition of NW plateau and surrounding rift flanks with the rare data set acquired during the aforementioned project period.

To investigate the seismicity and focal mechanisms in the NW plateau we utilized data from 30 temporary Guralp CMG-3T seismometers deployed during 2014–2016 by the University of Cornell (Keranen, 2013) in collaboration with the Institute of Geophysics, Space Science, and Astronomy (IGSSA) of Addis Ababa University. These instruments were equipped with three component seismometers and recorded data at a sampling rate of 100 Hz. Additionally, seven permanent Ethiopian seismic stations were included in the analysis (Figure 1).

Earthquake records were identified manually from continuous seismic data recorded at both permanent and temporary seismic stations in the region. To enhance the clarity of the seismic arrivals, we band-pass filtered the data within the range of 0.5 and 15 Hz, utilizing a Butterworth band-pass filtered seismogram as described by Havskov and Ottemoller (1999). We employed manual phase picking to determine the seismic phase arrival times. P-waves were measured on the vertical component due to their sharp onset and S-waves detected on either of the clearer horizontal component. A sample earthquake waveform is plotted on Figure 2. The earthquake location determination was conducted using the HYPO71 routine, which is integrated within the SEISAN software package (Figure 3a). Earthquake epicentral error ellipse are shown in Figure 3b and the distribution of P and S-wave travel time residuals are shown in Figure 3c. The 1D velocity model of Keranen et al. (2009) is used for the location, which was constructed using controlled-source seismic experiment conducted along the rift and across the Ethiopian Plateau. This model gives a high-resolution representation of the shallow crustal structure (less than 15 km deep), which reduces the uncertainty in the hypocenter depths. The 1D models of Alemayehu et al. (2023) and Baranov and Morelli (2014) were used to compute focal mechanism solutions. These include regional and global constraints (receiver functions, surface wave tomography) and best suited for resolving the structure of the deeper crust and upper mantle (>15 km). This makes it possible to do accurate full-waveform modeling needed for moment tensor inversion (Dreger and Helmberger, 1993; Herrmann, 2013). A minimum of three seismic stations was used to calculate the epicenter and depth of the detected earthquakes. In total we located 800 earthquakes.

In this study, earthquake magnitudes were determined using the local magnitude (M_L) scale of Keir et al., 2006b. M_L is a measure of the energy released during an earthquake and is calculated based on the logarithm of the maximum amplitude of the earthquake signal observed on the seismogram taking into account the distance between the earthquake focus and the seismometer. The calculated magnitudes are displayed in Figure 3a and Supplementary Figure S1. The scaling equation (Keir et al., 2006b) read as: M L = log A W A + 1.196997 * log r / 17 + 0.001066 * r − 17 + 2 (1)

In Equation 1, AWA represents the maximum zero-to-peak amplitude recorded on the horizontal components on a Wood-Anderson seismograph in mm and r is the hypocentral distance in km.

One of the most crucial earthquake source parameters is the focal mechanism solution, which serves as a starting point for understanding the state of stress of a region. Our objective is to determine the focal mechanisms, depth, and size of earthquakes in the study area that are relatively minor in magnitude. To achieve this, we utilized a software package ISOLA, which has been developed and widely used for focal mechanism analysis (Singh et al., 2024; Sokos and Zahradník, 2013; Sokos and Zahradnik, 2008; Zahradník and Sokos, 2018; Zahradník and Sokos, 2025). ISOLA provides advanced capabilities for analyzing earthquake focal mechanisms from the waveform records and extracting valuable information regarding the earthquake source characteristics. By employing ISOLA, we aimed to obtain reasonable estimate of the focal mechanism solutions for selected earthquakes under investigation, allowing us to gain insights into the stress conditions prevailing in the study area.

Assuming a point source, the deviatoric moment tensor was obtained by inverting the low-frequency full waveforms. A fourth-order Butterworth filter was used to band-pass filter the data between 0.02 and 0.09 Hz and cut from the event origin time for about 300 s after removing the instrument response, mean, and trend. The discrete wavenumber (DWN) approach was used in ISOLA to calculate the Green’s functions for the trial point source positions below the epicenter (Bouchon, 2003). The correlation (corr) or variance reduction (VR) parameter of the ISOLA code is used to quantify the waveform agreement between the synthetic and observed seismograms.

The double-couple (DC) percentage of the moment tensor solution of tectonic earthquakes may indicate the quality of the inversion result. The validity of the focal mechanism is significantly compromised when the double-couple percentage falls below 0.5 (DC < 0.5) (Carvalho et al., 2016).

ISOLA employs a grid search technique to determine the optimal moment tensor. Preliminary analyses of a horizontal grid search indicated optimal waveform fit beneath the epicenter. We generated a database of Green’s functions for depths ranging from 1.0 to 36 km in 2.5 km increments, calculated using ISOLA’s default maximum frequency of 1.6667 Hz, the Nyquist frequency. The inversion simultaneously searched for the optimal source depth and centroid time shift relative to the origin time (t = 0). The temporal grid search spanned from −5.0 to 5.0 s, using a step of 0.3 s. Stability of the solution was verified by jack-knifing of stations and components, implemented in ISOLA (Supplementary Figure S2).

Significant seismic activity in the rift valley is highlighted in this study, particularly near the Fentale Volcano and the Guraghe Border Fault (GBF). There were about 600 located earthquakes in the rift, mostly with magnitudes between 2.0 and 3.0 M_L and most of the depths are less than 20 km. Of these, 497 were concentrated around the Fentale Volcano, while 27 occurred close to the GBF. We measured earthquake swarms in the Fentale region in March and April 2015 (Figures 4a, b). These were further documented by Ayele et al. (2024). With magnitudes of Mw 3.5 and 4.3, moment tensor inversion for two events in GBF revealed normal faulting (Table 1; Table 2; Supplementary Figures S3-S6). The Afar and Ankober Border Faults (ABF) have seen less seismic activity. A single event that happened on 13 December 2014, at 23:37:19, close to Asayita, had a moment tensor inversion that showed an orientation of NE-SW and a depth of 11 km (Table 1; Table 2; Supplementary Figure S7). The same event was located by the USGS with a body magnitude of 4.4 (Mb).

We recorded seismic activity across the NW plateau with magnitudes as low as M_L 1.1. There were 115 earthquakes in this region, with magnitude ranging from 1.1 to 3.7 M_L. Most of these earthquakes are located at depths less than 25 km (Figure 4a; Supplementary Figure S1) while clusters of earthquakes are also observed close to Lake Tana. In contrast, there was no seismic activity reported beneath the southeastern plateau in this monitoring period (Figure 3a).

A normal faulting mechanism with a minor strike-slip component is obtained from the moment tensor inversion for the earthquake that occurred on 18 November 2015, at 14:49:55 GMT (Figures 5, 6a–c; Supplementary Figure S1).

The seismicity pattern in the Afar region aligns with ongoing extensional deformation. The focal mechanisms in this region predominantly exhibit normal faulting with strikes of more or less NE-SW orientations, aligning with the regional extensional stress regime in Afar. The shallow focal depths (about 11 km) suggest brittle deformation in the upper crust (Table 1; Table 2; Supplementary Figure S7). Focal mechanisms along the MER are predominantly normal faulting in style with a strike-slip component, suggesting complex fault interactions and potential reactivation of pre-existing structures, influenced by diking and regional stress fluctuations (Manighetti et al., 1998; Keir et al., 2006a). Substantial activity has been observed along the GBF and in proximity to the Fentale volcano, where earthquake swarms suggest connections between tectonic and magmatic processes. The Fentale-Dofan region notably displays considerable shallow seismicity and evidence of ongoing dike intrusions, emphasizing the importance of magmatism in strain accommodation (Acton et al., 1991; Temtime et al., 2020). The results show that the Wonji Fault Belt (WFB) traversing the rift floor, is a significant source of seismic activity and facilitates active deformation by fault slip and diking process (Keir et al., 2025; Ayele et al., 2024).

Numerous micoearthquakes recorded during the operation of extensive broadband seismic networks show that the NW plateau is more seismically active than previously expected. The majority of these earthquakes occur near Lake Tana, with Figure 3a showing are associated with a significant morphological rift valley, known as the Lake Tana Basin and thought to have been mostly active during the late Miocene to Quaternary (Chorowicz et al., 1998). Most of the earthquakes therefore appear to occur on existing lines of past extension, suggesting that pre-existing faults are being reactivated. The single focal mechanism for the NW plateau shows normal faulting with a strike-slip component and yielding a near E-W oriented T-axis, consistent with the extension direction across the MER.

While the dominant cluster of seismicity that occurs north-west of Lake Tana (LT) is spatially separated from the main concentration of hot springs and volcanic features located to the east, our results also reveal a secondary, yet significant, cluster of earthquakes occurring southeast of LT (Figures 1, 3a). This south-eastern cluster is spatially associated with the hot springs and volcanic features located specifically in that region, with the earthquakes also occurring down to lower crustal depths (Figure 4a; Supplementary Figure S1). Both observations point towards fluid involvement in triggering the seismicity. Seismic velocity models and receiver function analysis indicate that the crust beneath the NW plateau has a thickness of 20–35 km and exhibits locally elevated Vp/Vs ratios, suggesting the presence of partial melt zones or magmatic intrusions (Ogden et al., 2019). Low velocity zones in the lower crust suggest that melting and magmatic underplating significantly influence the mechanical properties of the crust (Eilon, 2025; Alemayehu et al., 2023; Chambers et al., 2019; Chambers et al., 2021). These magmatic alteration can locally cause high strain rates and facilitate the onset of micoseismicity in the deep crust.

Seismic activity in the rift valley is predominantly localized along the major fault zones in the rift valley, while it is more dispersed beneath the NW plateau, suggesting a diffuse extension regime. Our observations of seismicity beneath the NW plateau is consistent with recent GNSS findings that show ∼2 mm/yr of near E-W extension across the plateau (Birhanu et al., 2016). The findings are important since they show that deformation is not only localized to the rift valley, but that non insignificant amounts of extension occurs across a broad region outside of the rift. Seismic images of the mantle suggest the broad extension is associated with some limited partial melting (e.g., Asefa and Ayele, 2021; Chambers et al., 2022), linking the distributed extension to fluid involvement in the seismicity. The unique faulting process in the region illustrate how the divergence of the Somalian and Nubian plates within the East African Rift System has led to regional extension (Corti, 2009). The Afar mantle plume induces localized uplift, thermal weakening, and extension, hence enhancing fault activity (Ebinger and Casey, 2001). Tectonic processes are corroborated by seismic and geodetic data regarding extensional strain rates and upper crustal fault slip (Jones et al., 2019). Active magmatism, thermal degradatiation and pre-existing crustal heterogeneity all contributes to seismicity, particularly in the NW plateau, where the hypothesis of early-stage rifting is potentially occurring is supported by the cumulative observations.