Seismicity and Fault Interaction in the Gulf of Aqaba

The Gulf of Aqaba is considered one of the most seismotectonically active regions in the Middle East. This study analyzes the seismic activity from 1983 to 2018 to investigate the spatio-temporal distribution of the events in the Gulf’s three basins. The seismic activity is characterized by conventional mainshock–aftershock sequences and swarm activity. Most activity is located in the Aragonese basin, while the Dakar basin shows less seismic activity. The b-value was calculated for the three basins from the declustered catalogue, representing the fingerprint of the transform fault. Within the Aragonese, three significant earthquakes occurred in 1993, 1995, and 2015 with moment magnitude (Mw) of 5.8, 7.2, and 5.2, respectively. An intermediate seismic quiescence preceded the three earthquakes. The quiescence time for the 1995 earthquake began at the tail of the aftershocks of the 1993 mainshock, suggesting that the two earthquakes were not entirely independent. At the tail of the 1995 earthquake aftershocks, the seismic activity spread across the Gulf, increasing in frequency and decreasing in magnitude. A similar pattern was observed for the seismic quiescence preceding the 2015 mainshock. The occurrence of the three significant and proximal earthquakes in the Aragonese region over a 22-year interval, with the described similarities, suggests a link between these earthquakes. Finally, the Coulomb stress transfer shows that the 1993 source fault seems to increase the stress loading on the 1995 left-lateral strike-slip fault, and the latter would also increase the stress transfer on the 2015 left-lateral strike-slip fault.


Introduction
The Gulf of Aqaba is a tectonically active region that forms an en echelon strike-slip fault system along the plate boundary between Nubia-Sinai and Arabia. It occupies the southern end of the Dead Sea Transform fault (DSTF), with its main pull-apart basins which developed along this fault ( Fig. 1A) (Ben-Avraham et al., 1979;Freund et al., 1970;Garfunkel, 1981;Ben-Avraham & Tibor, 1993).
The Gulf of Aqaba is considered the most seismically active area in Egypt and the Middle East region, as it is the site of the strongest recorded earthquake in the region, the 1995 Nuweiba earthquake of Mw 7.2 ( Fig. 1B; Ambraseys et al., 1994;Arieh et al., 1994;Salamon et al., 1996). Studying the seismic activity pattern in a specific seismogenic zone may provide key evidence for the active tectonic processes. Therefore, in our study, the earthquake activity in the Gulf of Aqaba is classified into mainshock-aftershocks and swarms. Previous studies of the seismic activity in the Gulf of Aqaba showed isolated seismic sequences classified into three main swarms: 1983, 1990(El-Isa et al., 1984Al-Amri et al., 1991;Abou Karaki et al., 1993;Shamir & Shapira, 1994;Abdel-Fattah et al., 1997;Klinger et al., 1999;Hofstetter, 2003;Al-Arifi et al., 2012;El-Isa, 2012;Abd El-aal et al., 2019). Recent studies showed a correlation between these swarms by subsequent investigations that include the 2015 and 2016 sequences (Almadani, 2017;Abd El-aal & Badreldin, 2016;Abd El-aal et al., 2019).
The b-value, when adequately calculated, may be used to delineate the stress concentration, crack density, degree of heterogeneity, and thermal gradient (Molchan et al., 1997). Generally, any changes in the b-value are inversely related to changes in the Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s00024-023-03279-x.

Figure 1
A Regional tectonic and structural setting of the Dead Sea fault (DSF) that shows the Carmel fault (CA), Roum fault (RO), ElGhab (GH), and Yammouneh fault (YA) and the epicentres of historical events. B Structural setting of the Gulf of Aqaba. The bathymetric data was extracted from the General Bathymetric Chart of the Oceans (GEBCO)_08 global bathymetric data set (Becker et al., 2009), while the faults are from Ribot et al. (2021). EF (Eilat fault), HF (Haql fault), AF (Aragonese fault), ArF (Arnona fault), DF (Dakar fault), and TF (Tiran fault). The map also displays the epicentres of the three largest instrumental earthquakes of 1993, 1995, and 2015 and the fault plane solutions for these events 2046 M. Abdelazim et al. Pure Appl. Geophys. stress level (Bufe, 1970;Gibowicz, 1973;Scholz, 1968;Schorlemmer et al., 2004Urbancic et al., 1992;Wyss, 1973). Increasing material heterogeneity may also increase the b-value (Mogi, 1962). Changes in the b-value from one region to another can also reflect the degree of crustal deformation in that region. An increase in the b-value indicates a low degree of crustal deformation (El-Sayed et al., 1994). In addition, an increase in the thermal gradient causes an increase in the b-value (Warren & Latham, 1970;Wiens & Gilbert, 1996). Aftershocks and posterior mainshocks are frequent in locations where the previous mainshock has increased the Coulomb stress. In contrast, earthquakes are less common than before the mainshock in areas where the Coulomb stress decreased (Ma et al., 2005;Asayesh et al., 2020). Although several studies have found a strong link between Coulomb stress changes and aftershock distribution, the presence of aftershocks in stress shadows in various case studies has remained unexplored (Asayesh et al., 2018(Asayesh et al., , 2020Parsons et al., 1999;Toda et al., 2011a).
This work aims to identify spatio-temporal variations in seismic activity patterns and their relationship with tectonic settings in the Gulf of Aqaba. The nature of regional seismicity in the Eilat, Aragonese, and Dakar basins as well as the whole Gulf is investigated through an in-depth examination of the b-value. Another main objective of this work is to clarify whether there is a relation between the three largest earthquakes (1993, 1995, and 2015 mainshocks) that occurred in the Aragonese Basin during the 22-year period using the Coulomb stress approach.

Tectonic and Geological Setting of the Gulf of Aqaba
The Gulf of Aqaba occupies the southernmost part of the DSTF, a north-south-trending left-lateral major strike-slip fault and plate boundary between Arabia and Africa-Sinai. Quennell (1959) was among the first to document the fault zone and the related sense of motion. The DSTF extends for about 1100 km from the Gulf of Aqaba, and represents extensional tectonics in the Red Sea to the Zagros-Bitlis convergence zone in eastern Turkey (Bartov et al., 1980;Ben-Avraham, 1985;Ben-Avraham et al., 1979;Eyal et al., 1981;Garfunkel, 1981), where it forms the northern part of the Syrian-African rift system. The DSTF plate boundary is composed of three segments: the southern segment, which runs from the end of the Gulf of Aqaba through the Dead Sea and Jordan River valleys and exhibits leaky transform characteristics; the central segment of the NNE-SSE-striking bend, which runs in Lebanon and southwestern Syria; and the northern segment, which is characterized by two N-S-striking faults and runs through northwestern Syria and southern Turkey (Ben-Avraham et al., 1979;Garfunkel, 1981;Quennell, 1984;Bosworth & Strecker, 1997;Ben-Avraham & Zoback, 1992). The DSTF separates the Arabian plate to the east and the African plate to the west, as shown in (Fig. 1A). Several plate tectonic models (DeMets et al., 1990;Jestin et al., 1994;McClusky et al., 2003;McClusky et al., 2003) which are based on the analysis of global seafloor spreading, fault systems, earthquake slip vectors, and GPS data have indicated that the Arabian plate is moving at about 18 mm/year north-northwest relative to Eurasia, while the African plate is moving at about 6 mm/year relative to Eurasia (McClusky et al., 2003). The differential rates of motion between the African and Arabian plates are thought to be taken up predominantly by left-lateral motion along the DSTF.
The activity along the DSTF is thought to have started in the Middle Miocene (* 14-11 Ma) (Quennell, 1959). The cumulative displacement along the DSTF was 105 km (Freund et al., 1968(Freund et al., , 1970Bartov et al., 1980;Garfunkel et al., 1981), giving an average slip rate of 0.76-0.97 cm/year (Quennell, 1959). Some geodetic studies have deduced that the present-day slip rate for the southern Dead Sea Rift is 0.44 ± 0.03 cm/year, which is only about half the rate deduced from the cumulative displacement (Reilinger et al., 2006), while others have suggested a geological rate of 0.5-0.7 cm/year for the past 5 Myr Garfunkel et al., 1981). The DSTF's extensional regime and lateral motion identified five basins, with the Gulf of Aqaba as the largest one (Ben-Avraham et al., 2012). Vol. 180, (2023) Seismicity and Fault Interaction in the Gulf of Aqaba 2047 The Gulf of Aqaba is a narrow-elongated depression with an average width of around 17 km, decreasing to 6 km in the north, and about 195 km long between latitudes 28.0°and 29.6°N, bordered by Sinai and Arabia. The bathymetry of the Gulf shows that the water depth in the central basin reached 1770 m (Fig. 1B) (Moustafa & Khalil, 2020), which was formed in the Late Oligocene to Early Miocene during the Red Sea rifting (Ben-Avraham & Garfunkel, 1986). Although the Gulf of Aqaba belongs to the left-lateral strike-slip plate boundary, the tectonic structure displays NNE-SSWtrending pull-apart basins predominantly with dipslip extensional movement and accompanying footwall uplift (Bosworth et al., 2017). The uplift was evidenced by measuring the elevated Pleistocene coral terraces in the central basin. The maximum vertical cumulative movement reaches * 19 m and provides an average rate of * 0.015 cm/year. Several geophysical studies have been carried out to investigate the structural setting of the Gulf as a means of detecting faults and sedimentary patterns. These methods include marine magnetic survey (Ginzburg & Ben-Avraham, 1986), the Bouguer gravity field , bathymetry and seismic reflection (Ben-Avraham et al., 1979;Ben-Avraham, 1985;Ben Avraham et al., 1987;Reches et al., 1987;Ben-Avraham, 1996;Ben-Avraham & Tibor, 1993;Ehrhardt et al., 2005;Makovsky et al., 2008;Tibor et al., 2010;Hartman, 2012;Ribot et al., 2021), and ground-penetrating radar (Hartman et al., 2014). The complex structural setting of the Gulf of Aqaba is revealed by geophysical investigations, which raises the possibility of crustal thinning beneath the Gulf of Aqaba. A crustal thinning has been attributed to the intrusion of the upper mantle into the lower crust as a result of the Red Sea spreading activity. Seismic profiles, Moho discontinuity, and crustal thickness studies showed that the crustal thickness is 35 km in the northern part of the Gulf and decreases to 27 km to the south (Ginzburg et al., 1981;Mart, 1982;Hartman et al., 2014;El-Khrepy et al., 2016;Bosworth et al., 2017). Aeromagnetic and free-air gravity data confirmed these results, especially in the northern and southern parts of the Gulf (Almalki & Mahmud, 2018;Hamouda & El-Gharabawy, 2019;Khattab, 1989).
Classically, the active motion along the DSTF led to the formation of a series of three connected pullapart basins in the Gulf of Aqaba: the Eilat, Aragonese, and Dakar basins (Ben-Avraham & Tibor, 1993). Recently, Ribot et al. (2021) identified six basins in the Gulf of Aqaba, including Dakar Deep, Tiran Deep, Hume Deep, Aragonese Deep, Arnona Deep, and Eilat Deep (Fig. 1B). These basins are structured by three major strike-slip faults, namely, the *N24°-trending Eilat fault, *N17°-trending Aragonese fault, and *N20°-trending Arnona fault, separated by pull-apart basins that exhibit a slight normal dip-slip motion. Normal faults border each basin, with a N20°fault trend parallel to the Dead Sea Transform and an N160°fault trend oblique to the Gulf of Aqaba general direction. The second trend bounds the northern and southern basins (Fig. 1B). We describe the three basins in the following subsections.

The Eilat Basin
The Eilat Basin is the largest and shallowest in the Gulf of Aqaba (Ben-Avraham et al., 2012) (Fig. 1B). Morphological studies showed that the Gulf head could be divided into at least two separate deeps, with the Eilat sub-basin in the west and the Aqaba subbasin in the east, separated by the Ayla high (Tibor et al., 2010). According to Hamouda and El-Gharabawy (2019), bathymetric studies showed that the average depth of the Eilat Basin is about 900 m, extending to a length of about 50 km with a width of 3-8 km. The Eilat Basin is bounded by strike-slip and normal fault systems, where the first set represents the Eilat and Aragonese faults that border the basin from west and east, respectively. The second set corresponds to normal faults that run along the northern and southern boundaries of the basin. The Eilat fault (EF) is a sub-vertical continuous strike-slip fault (Ben-Avraham et al., 1979). According to seismic reflection data, the Aragonese fault split into two branches at the northern end of the Eilat Basin: the eastern and western branches. The eastern branch bends to the right and becomes the Aqaba fault, while the western branch extends diagonally to become the EF zone (Ehrhardt et al., 2005).

The Aragonese Basin
With a depth of 1777 m, the Aragonese Basin comprises the Gulf's deepest section, and the area is bounded by normal and strike-slip systems (Ben Avraham, 1985). The Aragonese and Arnona faults, which surround the region from the west and the east, are examples of the strike-slip faults. (Fig. 1B). The normal faulting comprises the NW-trending lines surrounding the NW-trending lines that surround the valley on the north and south. The southern normal fractures are currently more active, straight, and morphologically distinct than the northern ones (Ribot et al., 2021). The existence of an uninterrupted morphology stretching 53 km along the Aragonese Basin is limited to its central segment. The Arnona Basin is a smaller subsidiary basin located to the southwest and at the southern termination of the Aragonese fault. This basin is bordered to the west by oblique-slip faults isolated from the Aragonese Basin by an elevated sea floor (1394 ± 1.3 m depth) surrounded by normal faults on each side (Ribot et al., 2021).

The Dakar and Tiran Basins
The eastern boundary of the Dakar fault defines the southernmost basins, the Dakar and Tiran basins. The structural limit separating the two basins, even though they are different, is unclear (Ben-Avraham, 1985;Ben-Avraham, et al., 1979). The Dakar fault exhibits a double scarp with considerable vertical motion at its southern end. The normal faulting mechanism of the 1993 earthquake ( Fig. 1B) is in good agreement with this type of motion distinguishing this part of the Gulf . Based on multi-beam bathymetric data, there is no evidence for strike-slip motion in this basin (Ribot et al., 2021). The southern part of the Tiran Basin is bounded by a series of NW-trending short parallel normal faults dipping northward (Fig. 1B). Between the Aragonese and Dakar basins, the Gulf of Aqaba suffers from the uplifting of margins. According to Bosworth et al. (2017), the uplift is strongest in the Arabian margin, as evidenced by the elevated Pleistocene coral terraces. Most of the uplift is thought to be generated by the margin parallel N-S to NNE-SSW young extensional faults. The Hume Basin is located south of the Straits of Tiran, linking the Gulf of Aqaba with the axial depression of the Red Sea, and is bounded by the strike-slip fault at its northern side.

Data Sources and Unified Earthquake Catalogues
The earthquake catalogue is an essential database for any study of seismic activity. In this work, we compiled an earthquake catalogue for the Gulf of Aqaba and its vicinity that covers the period from 1983 to 2018, with magnitudes ranging from 0 to 7.2 Mw, from the following sources: (1) Egyptian National Seismological Network (ENSN), (2) International Data Center (IDC), (3) International Seismological Centre (ISC), (4) Incorporated Research Institutions for Seismology (IRIS), (5) Observatories & Research Facilities for European Seismology (ORFEUS), and (6) European-Mediterranean Seismological Centre (EMSC). Our catalogue includes the main parametric information-date, time, surface location, depth, and magnitude-for each event.
The duplicate events resulting from different sources were removed from the catalogue using CompiCat software (Kossobokov et al., 2011). The merging process was carried out according to the following criteria: (1) difference in location (latitude and longitude) not exceeding 0.1°, (2) difference in time not exceeding 30 s, and (3) difference in magnitude not exceeding 0.2. A manual quality inspection of this process was implemented to ensure that the automated procedure was running well. The depth is not considered due to inaccuracies caused by the shortage in station coverage.
The locations from the ISC bulletin are given priority because the earthquake data have been assembled from various stations worldwide. The locations of the small local events which are not listed in the ISC bulletin are picked from the ENSN bulletin. Different types of magnitudes are used in the assembled catalogue, e.g., Mw, M L , Ms, Mb, and M d . The empirical relations constructed by Hussein et al. (2008) are used to convert different types of magnitudes to moment magnitude (Mw) to obtain a Vol. 180, (2023) Seismicity and Fault Interaction in the Gulf of Aqaba 2049 homogeneous and unified catalogue. The final compiled catalogue contains 4364 events, with Mw values ranging from 0 to 7.2.

Seismicity of the Gulf of Aqaba
Seismicity is a very useful and essential tool for determining the activity in a particular zone over time. Previous studies investigating the seismic activity in the Gulf of Aqaba showed isolated seismic sequences, including 1983, 1990, 1993, 2015, and 2016 sequences, as mentioned in the introduction. A review of these seismic sequences will be explained in the following paragraphs: The 1983 swarm struck the Eilat Basin in the Gulf of Aqaba between January 21 and April 20, 1983, with magnitudes ranging from 1.7 to 4.9 without a distinguishable mainshock (Abou karaki et al., 1993;El-Isa et al., 1984;Hofstetter, 2003). The activity level oscillates over time, peaking on February 3 when a few major shocks occurred, including the swarm's largest magnitude event. The 1983 swarm consists of 94 recorded earthquakes with local magnitude (ML) [ 3 (Hofstetter, 2003). The activity is concentrated in a narrow zone extending about 20 km in length with an N-S trend offshore Haql city at the eastern edge of the Gulf (Abou karaki et al., 1993). The 1983 swarm concentrated in the northern part of the Gulf, around the following coordinates (28.8°-29.4°N and 34.3-35.1°E) (Al-Arifi et al., 2012).
The 1990 swarm, with an epicentre mainly between Lat 28.8°-29.3°and Long 34.3°-34.85°(El-Isa, 2012), was a relatively small swarm that occurred in the central part of the Gulf, between the Eilat Deep and the Aragonese Deep (Klinger et al., 1999). The swarm occurred during the period from April 1990 to May 1991 (El-Isa, 2012), and the largest event had an ML of 4.3 (Klinger et al., 1999), with only three earthquakes with ML C 3.0 (El-Isa, 2012).
The 1993 activity represents the third swarm. It is located south of the 1983 sequence, close to the western edge of the Gulf, around the Aragonese and Dakar basins, i.e., along the Aragonese/Arnona fault (Hofstetter, 2003). This sequence continued for about 5 months and consisted of 600 events with magnitudes ML C 3; the most significant events had ML = 5.8 and 5.6. The strongest one happened on August 3, 1993. The 1993 activity cannot be categorized as a conventional mainshock-aftershock sequence because it is not spatially correlated with a mainshock segment (Shamir, 2008). The magnitude difference between the mainshock and the most significant subsequent event is 0.2, comparable to magnitude standard error estimates. Consequently, it can be classified as swarm-type activity due to the lack of a single large earthquake and the concentration of seismic activity in a small area. Abdel Fattah et al. (1997) showed that the distribution of 1993 activity followed a NW-SE trend, which is consistent with transverse fault trends in the Gulf of Aqaba.
The 1995 sequence occurred in the Aragonese Basin at 28.76°N, 34.66°E. The main event occurred on November 22, 1995, at 04:15 GMT, with Mw = 7.2 . It was the most significant event in the whole DSTF in nearly 175 years (Shamir, 2008). This sequence is more energetic than any prior seismic activity in the Gulf of Aqaba (Salamon et al., 1996). The 1995 sequence took place to the south of the 1993 sequence. The hypocentral space distribution of this sequence extends along a NNE-SSW trend (Marzouk et al., 1996). During the first 100 days that followed the 1995 mainshock, more than 7000 aftershocks with ML [ 1.5 occurred (AL-Qaryouti, 2002).
The 2015 sequence can be classified as mainshock-aftershocks activity. The mainshock of the 2015 earthquake sequence occurred on June 27, 2015, at 15:34 UTC, at longitude 34.74°E and latitude 28.90°N with ML = 5.2. This shock was followed by 95 aftershocks with ML ranging from 0.7 to 4.2 (Abd El-aal et al., 2019).
The 2016 sequence was classified as foreshocksmainshock-aftershocks type of activity (Abd El-aal et al., 2019). The mainshock of the 2016 earthquake sequence occurred on May 16, 2016, at 01:45 UTC at longitude of 34.7°E and latitude of 28.49°N with ML = 5.5 in the Gulf of Aqaba. This shock was followed by 93 aftershocks with ML ranging from 1.6 to 3.8 and preceded by a single foreshock (ML = 4.2) (Abd El-aal et al., 2019).
The occurrence of these sequences led some authors to classify the seismicity in the Gulf of Aqaba into two types: swarm and foreshocks-aftershocks (El-Isa et al., 1984) or foreshocks-mainshock-aftershocks, mainshock-aftershocks and swarm type activity (Abdel-Fattah et al., 1997) or foreshockmainshock-aftershocks and mainshock-aftershocks sequences (Abd El-aal et al., 2019). The swarm activity defines distinct active segments around the different pull-apart basins forming the Gulf of Aqaba (Shamir & Shapira, 1994). This activity is concentrated in the Aragonese Basin, the location of the largest earthquake, followed by the Eilat Basin in the ranking (El-Isa, 2012).

Historical Seismicity
The Gulf of Aqaba has a long historical seismic record, as it has suffered from historical earthquakes, whether occurring in its vicinity or affecting it (Fig. 1A). The historical seismicity and seismotectonics of Egypt and Arabian regions have been studied extensively (Poirier & Taher, 1980;Maamoon et al., 1984;Abou Karaki, 1987;Ghawanmeh, 1992;Ambraseys et al., 1994;Guidoboni et al., 1994;Badawy, 1999;Klinger et al., 2000;Zilberman et al., 2005;Marzouk, 2008;Ambraseys, 2009;Ferry et al., 2011;El-Isa, 2012;Al-Arifi et al., 2012;Klinger et al., 2015;Bektas, 2021) based on field investigations of earthquake geology in addition to the existence of many written sources, historical archives, press reports and archaeological inscriptions that recorded the effects of earthquakes. Historical catalogues mentioned that only three to four earthquakes occurred near the Gulf of Aqaba. Bektas (2021) tried to locate the historical earthquakes using the sedimentary traces observed from the turbidity and chronology of the deep sea sediment cores in the Gulf of Aqaba. According to Bektas (2021), the historical earthquakes occurred in 980, 1068, 1212, 1588, 1800, and 1839, but no historical documents exist for the 980 and 1800 events. Therefore, we will focus on the most famous earthquakes in the vicinity of the Gulf of Aqaba, including the earthquakes of March 18, 1068, May 1, 1212, January 4, 1588, and 1839 (Fig. 1B). The details regarding these earthquakes are explained in the supplementary material. According to Bektas (2021), the 1068 event may have broken both the Aragonese fault (AF) and the Arnona fault (ArF), and the 1212 event occurred due to the rupture of the EF, while Kilenger et al. (2015) and Lefevre et al. (2018) deduced that the 1068 and 1212 events occurred on the EF based on palaeoseismology trenching studies. This study also showed that the 1588 event occurred along the Haql fault (HF) and that the AD 1839 earthquake was followed by another earthquake in 1840 on the Tiran fault (TF).

Instrumental Seismicity
The instrumental seismicity in Egypt in general passed through different phases of evolution up until 1997 when the ENSN stations were fully installed. This is reflected and explained in Fig. 2. This figure shows the frequency of earthquakes in the vicinity of the Gulf of Aqaba from 1983 to 2018, not only the number of earthquakes but also their magnitude, where the events with low magnitude were recorded. Before 1982, the number of stations surrounding the Gulf of Aqaba was insufficient to detect small-magnitude earthquakes. After 1982, the number of earthquakes increased as a result of an increased number of regional stations throughout the Gulf. After 1997, the installation of ENSN improved the detectability of microearthquakes. This led to a significant increase in the number of the recorded earthquakes.
The seismicity map for the period from 1983 to 2018 with Mw ranging from 2.5 to 7.2 shows that the activity is distributed throughout the Gulf (Fig. 3), with a concentration of activity in the Aragonese Basin compared with the other two basins. In the Aragonese Basin, we observe the well-located three largest earthquakes: the Mw = 5.8 1993 earthquake, the Mw = 7.2 1995 earthquake, and the Mw = 5.2 2015 earthquake. The activity in the Dakar Basin represents the lowest activity in the Gulf.
The seismicity in the Gulf of Aqaba shows distinct spatio-temporal behaviour (Fig. 3). We suggest that this activity can be classified into the 15 phases based on earthquake density and level of magnitude. Also, we tried to differentiate between swarms and sequences according to the following criteria (Fischer et al., 2003;Hill, 1977;Ogata et al., 1995;Scholz, 2002;Utsu, 1971): (1) Swarms are defined as groups of earthquakes clustering in time and space, none of which may be classified as the Vol. 180, (2023) Seismicity and Fault Interaction in the Gulf of Aqaba 2051 mainshock; (2) origin times of earthquakes of a swarm are usually similar and differ from the origin times of other earthquakes in the area of interest; (3) clustering in time is the most salient (prominent) feature and probably displays a common origin of the swarm events; (4) a decrease in the inter-event (waiting) time is observed for swarms, while it is still significant for background activity; and (5) the largest aftershock in the sequence is at least one unit smaller than the mainshock (Utsu, 1971), but some researchers have decreased this difference to 0.45; i.e., if the largest aftershock is lower than the mainshock by less than 0.45, this will be considered a swarm (e.g., Ogata et al., 1995). The summary of these phases is shown in Table 1. A detailed description of these phases is provided in the supplementary material.

The b-Value in the Gulf of Aqaba
The b-value is simply the slope of the frequencymagnitude distribution for a specific group of earthquakes occurring in a particular region over a given period (Gutenberg & Richter, 1944). In this study, we applied the Gutenberg-Richter (GR) relationship of Peresan et al. (2005) in the form:  El-Sayed and Wahlström (1996) found that the b-value for the Gulf of Aqaba could be set at 1.2 ± 0.3, while Hussein et al. (2008) reported a significantly smaller b-value of about 0.81 ± 0.06. According to Hofstetter (2003), b-values for the whole Gulf, the Eilat Basin, and the Aragonese Basin were 1.01 ± 0.08, 1.15 ± 0.11, and 0.89 ± 0.09, respectively. Al-Tarazi (2005), after eliminating the swarm activity from the Gulf of Aqaba catalogue, showed that the central value of the b-slope dropped drastically from 0.9 ± 0.14 to 0.66 ± 0.04. The values of the second value agreed with those calculated by Ali (2016), showing that the b-value was 0.66 ± 0.01. Korrat et al. (2006) reported that the central values of b-slopes calculated for earthquakes that occurred before November 22, 1995, including the mainshock and its aftershocks, were 1.13 and 1.25, respectively.
In this section, we apply the GR relation, where the b-value represents the frequency as a function of magnitude. Provided the size of the study area is sufficiently large (Molchan et al., 1997), the b-value for a specific group of earthquakes occurring in a particular region over a given time interval (Gutenberg & Richter, 1944) is a power-law expression to consider the size of the most significant events: where N (m) is the number of events having a magnitude, and b (size distribution) and a (seismic activity level) are constants. Accordingly, the calculation of the b-value for any region allows the identification of earthquake occurrence rate and thus contributes to the definition of the tectonic regime in that region (El-Sayed et al., 1994). The b-value in active tectonic areas is typically about 1.0 (Båth, 1979;Frohlich & Davis, 1993); however, in volcanic areas and regions of earthquake swarms, the b-value may be higher (Frohlich & Davis, 1993;McNutt, 2005;Molchan et al., 1997;Wyss et al., 1997).
In this work, we used the maximum likelihood estimation approach, which is widely used to estimate the b-value based on the greater curvature (Aki, 1965) and fulfilling the condition posed by Molchan et al. (1997) that makes the relation (1) a generally valid law. To calculate the b-value, it is necessary to decluster the catalogue, i.e., the earthquake catalogue is subdivided into mainshocks (assumed independent events), foreshocks, and aftershocks (earthquake rates dependent only on the main tremors). This approach is widely used in seismology, particularly for seismic hazard assessment and earthquake prediction models (Keilis-Borok & Soloviev, 2002;Panza et al., 2021). Several methods can be used in declustering (Gardner & Knopoff, 1974;Reasenberg, 1985). Using these approaches, the earthquakes in a space-time window around a large event called the mainshock are declustered. The declustering technique of Gardner and Knopoff (1974) is applied in this study. The obtained declustered catalogue comprises 2641 events and the cumulative number of earthquakes in the original catalogue with magnitude M [ 1.0 (Fig. 32S). The curve of the declustered data set is naturally smoother than the original catalogue curve. The catalogue will be regarded as complete in terms of events with M [ Mc when the data settle down to define a log-linear straight line (e.g., Yadav et al., 2009). The magnitude of completeness (Mc) can be defined as the minimum magnitude above which all earthquakes can be detected during the catalogue period. This magnitude is not unique temporally and spatially, mainly due to the change in network detectability and coverage. Figure 33S shows the temporal change in Mc over the whole Gulf area. Applying statistical analysis to the catalogue without considering the strong perturbations in the Mc will lead to a biased b-value. Data analysis has two ways to address such a challenge: the classical method that chooses the lower Mc, which gives flexibility and expands the range of magnitude, or the multi-scale seismicity model technique (Molchan et al., 1997), which divides the catalogue into sub-catalogues with different Mc values.
In the present study, we applied two techniques to calculate the b-value for the whole Gulf and each of the three Gulf basins and compared our findings. For the classical method, we divided our catalogue into two different periods, the first one up to 2002 with Mc = 3.5, and the other period from the beginning of 2002 with Mc = 2.5 (Figs. 34S, 35S, 36S, 37S). The results are shown in Table 1 of the supplementary material. Ultimately, we need a single value that expresses the value of b for the Gulf and each basin. Therefore, we apply the second approach (multi-scale seismicity model). After all, it is a more accurate technique for inhomogeneous catalogues containing different levels of Mc that require a b-value at each Mc to obtain an average b-value.
In addition, the catalogue is divided into two subcatalogues with different Mc values (Molchan et al., 1997). The maximum likelihood technique is used to estimate the GR b-values for the whole Gulf of Aqaba and each basin from 1983 to 2018. The results with this technique are more reliable than those using the traditional least-squares regression method and give accurate estimations of the b-value. This method is preferable because the estimated variance is smaller and the calculation depends on the number of input parameters. The results for b-values obtained using the magnitude-frequency relation (2) for the whole Gulf of Aqaba and each basin are shown in Table 2. According to the multi-scale seismicity model reported by Molchan et al. (1997), b-values for the three basins and the Gulf as a whole are nearly identical. This means that the estimated value in the Gulf of Aqaba represents regional tectonics, and the distinctive b-value in the Gulf of Aqaba reflects the transform fault nature of the Gulf. The MATLAB code of Kijko and Smit (2012) was used in the b-value calculations.
To show the changes in the b-value along the Gulf of Aqaba, b-value maps were constructed for the periods from 1983 to November 21, 1995, before the occurrence of the 1995 earthquake (Fig. 4A) and from November 23, 1995, to June 26, 2015, before the occurrence of the 2015 earthquake (Fig. 4B). It is clear that higher b-values were observed in the northern and southern part of the Gulf, ranging from 0.34 to 0.81 (Fig. 4A) and from 0.7 to 1.1 (Fig. 4B) for the two periods, respectively, while the lowest b-values were found in the central part of the Gulf, where the 1995 and 2015 events took place. According to Fig. 4A and B, the lower b-values were 0.11 and between 0.6 and 0.7 in the area of the two events, respectively. The areas with low b-values were directly proportional to the earthquake rupture 2054 M. Abdelazim et al. Pure Appl. Geophys. area. The lower b-values indicated that the area is subjected to significantly more shear stress.

The Hypothesis of Coulomb Stress Change
Part of the released accumulated stresses on the seismic fault rupture induce a ''permanent'' change in the earthquake area's stress field (also called static stress change). The earthquake location, the rupture geometry, and sense of fault motion control this change (Asayesh et al., 2020). One measure of this change is the so-called Coulomb failure function (DCFF), which depends on the changes in both shear stress (Dss) and normal stress (Dr n ), with l 0 as the effective coefficient of friction on the fault that includes the unknown effect of pore pressure change (King et al., 1994;Mouyen et al., 2010), where l 0 may range from 0.2 to 0.8 according to the fault mechanism. The change in failure stress Dr f can be calculated according to the following equation: Suppose the redistribution of static stress variations around the ruptured fault is identified. In that case, it may be possible to predict the location (not the time) of future mainshocks on neighbouring faults, relying on the increase or decrease in Coulomb stress generated by the ruptured fault (Mitsakaki et al., 2006). As a result, coseismic static Coulomb stress changes are thought to be a key element in earthquake triggering and clustering, as well as having a substantial influence on the generation of aftershocks and subsequent mainshocks (e.g., Deng & Sykes, 1997;Harris, 1998;Stein, 1999;Freed & Lin, 2001;King et al., 2001;Freed, 2005;Steacy et al., 2005a;Parsons et al., 2008;Toda et al., 2011a;Yadav et al., 2012;Mitsakaki et al., 2013;Kariche et al., 2017). A sudden slight increase in static stress of \ 0.01 MPa may reduce the failure time of an earthquake, while a similar decrease in static stress may delay earthquake occurrence (e.g., Kariche et al., 2018;Marsan, 2006;Parsons et al., 2008;Toda et al., 2012). A positive Drf means that the shear stress loading on the fault overtakes the shear strength, causing the rupture to be accelerated. In this instance, the first earthquake may accelerate the failure of a closer second fault. If Drf is negative, the normal stress increase increases the shear strength and decreases loading, bringing the second event away from failure and into what is known as a ''stress shadow''. In other words, a positive change in the Coulomb failure stress transfer in a specific area is evaluated as an indication that a fault plane existing in this area has been driven extremely close to failure; on the other hand, if it decreased, the interpretation is the exact reverse situation (i.e., relaxed) (Gràcia et al., 2019).

Coulomb Stress Change Calculations for the Largest Earthquakes in the Gulf of Aqaba
As previously stated, three significant earthquakes struck the Aragonese Basin in the Gulf of Aqaba within a period of 22 years: 1993 (Mw = 5.8), 1995 (Mw = 7.2), and 2015 (Mw = 5.2) earthquakes. The proximity of these earthquakes to one another suggests fault rupture interlinkage and stress transfer followed by subsequent fault interaction with large earthquake triggering. To examine the probable interaction among the three significant Gulf of Aqaba  (Toda et al. 2011a(Toda et al. , 2011b(Toda et al. , 2011c). In the current study, the static stress change is calculated on both fixed receiver earthquake ruptures and optimally oriented faults using the effective coefficient of friction (l0) of the fault and fault plane parameters with a given (strike/ dip/rake) combined with the hypocentral information for earthquakes. These parameters are summarized in Table 3. Several studies provide fault plane solutions for the 1993 and 1995 earthquakes. The 1993 earthquake shows that it could be described as a normal faulting mechanism with a slight strike component (Abdel-Fattah et al., 1997;Abdulaziz & Al-Arifi, 2001;Pinar & Türkelli, 1997). The calculated fault plane solutions for the 1995 earthquake show strike-slip motion with a minor normal component (Dziewonski et al., 1995;Pinar & Türkelli, 1997; or a pure left-lateral strike-slip motion (Kikuchi, 1995;Shamir, 1996;Hofstetter et al., 1998;Klinger et al., 1999;Klinger et al., 2000;Abdulaziz & Al-Arifi, 2001;Hofstetter et al., 2003;Baer et al., 2008). Regarding the 2015 earthquake, we constructed the focal mechanism solution for this event based on the polarity of the first onset (the P-phase), the polarities of the S H phase and S V phase, and the amplitude ratios of (S H /P), (S V /P), and (S V / S H ). Detailed explanations and results are presented in Table 2S, Table 3S, and Fig. 38S in the supplementary material. It is a strike-slip fault with a minor normal component.
In the present study, we have evaluated the Coulomb stress changes induced by the 1993 earthquake using the fault parameters (strike/dip/rake) proposed by Pinar and Türkelli (1997), which can be represented by a NNW-SSE-trending normal faulting mechanism with a minor strike-slip component, as shown in Table 3. We select this nodal plane based on the aftershock distribution (Abdel-Fattah et al., 1997). The results of Coulomb stress show that a positive Coulomb stress change is obtained close to both fault terminations (Fig. 5). Also, the aftershocks following the 1993 mainshock are located within the positive Coulomb stress change area (in regions of stress increase), which is the potential zone of future failure (Fig. 5). The specific receiver fault approach was utilized to determine the influence of static stress fluctuation caused by the 1993 earthquake on the 1995 earthquake fault. The fault parameters of the 1995 earthquake were extracted from both the InSAR (interferometric synthetic aperture radar)-based model and seismological studies and expressed by a NW-SE-trending strike-slip left-lateral fault plane (Baer et al., 2008) as shown in Table 3. This plane was used as a receiver plane for the Coulomb stress changes caused by the 1993 earthquake. Figure 6 shows the calculated Coulomb stress variations caused by the 1993 event as a source fault (the source fault is the fault plane that is slipped during the earthquake) and the 1995 event as a receiver fault (the fault plane which is obtained from the focal mechanism solution). The results of the Coulomb stress showed that there are four positive stress lobes and four negative stress lobes. The first type includes two lobes with NNW-and WNW-trending pulses and two lobes with SSE-and ESE-trending pulses, while the second type includes two lobes with NE-SWtrending pulses and two lobes with NW-SE-trending pulses. The DCFF modelling on the receiver plane suggests that the 1995 earthquake and its aftershocks were triggered by 0.2-0.4 bars. This means that the transferred Coulomb stress changes due to the 1993 event are appropriate to create failure of the 1995 mainshock.
The specific receiver fault approach was employed to assess the impact of static stress variation produced by the source of the 1995 mainshock on the receiver fault of the 2015 earthquake fault. The receiver plane parameters are shown Table 3 The hypocentral information, fault plane solution parameters, and rupture parameters for the three most significant earthquakes in the Gulf of Aqaba  in Table 3. We chose this fault plane depending on the distribution of the aftershocks, as shown in Fig. 7. The preferred fault plane is the NNW-SSE-trending left-lateral strike-slip plane from our study. The Coulomb stress shows four positive lobes and four negative lobes. The event of 2015 is located in one of the positive lobes of the 1995 earthquake. From the calculation of the Coulomb stress change (Fig. 8), we find that the 2015 receiver fault receives a positive DCFF stress change ranging between 0.1 and 0.3 bars as a result of the 1995 event, since it is in the positive Coulomb lobe. The Coulomb stress changes modelling suggests that the stress transfer caused by the 1995 earthquake facilitated the 2015 earthquake failure (Fig. 8). Some of the well-located significant aftershocks (Belferman et al., 2011) occupy the positive Coulomb stress region.

Discussion and Conclusion
The collected homogeneous seismicity data from several sources in the Gulf of Aqaba were investigated in terms of spatio-temporal patterns of seismic activity and b-values from 1983 to 2018. The continuous seismic activity along the Gulf of Aqaba from 1983 to 2018 demonstrated that all faults undergo continuous seismotectonic deformation. Before the August 3, 1993 earthquake, the activity occurred within the Eilat Basin (Figs. 1S, 7S in the supplementary material). A marked increase in seismic activity is observed with the transfer of activity towards the Aragonese Basin, particularly after the August 3, 1993 earthquake. All events with magnitudes larger than Mw = 5.2 are located in the Aragonese Basin, reflecting its active tectonics (Fig. 10S in the supplementary material). The central section of the Aragonese fault showed signs of recent deformation of about 53 km along its length, with a recorded displacement of about 1.9 m due to the Mw 7.2 1995 mainshock (Ribot et al., 2021). Compared to the other basins, the Dakar Basin shows low seismic activity, with an Mw of less than 5.0. The decrease in this basin's activity may be attributed to the interplay between the Gulf of Aqaba transform fault and the northern Red Sea incipient oceanic spreading (Bosworth et al., 2017). The Dakar Basin also reflects lower deformation and minor stress and strain. This conclusion is in accord with the results based on the multi-beam bathymetry in the Gulf of Aqaba, which showed no evidence of deformation on the Arnona fault, which limits the eastern edge of the Dakar Basin (Ribot et al., 2021). Based on a 5 mm/year slip rate along the Dead Sea Fault, the lack of recent large earthquakes along the 83 km-long Arnona fault suggests the potential for a future large and damaging earthquake with a moment magnitude ranging from 7.0 to 7.5 (Ribot et al., 2021). This may happen if this fault was continuous. Based on a detection level of Mc = 3.5 and Mc = 2.1 during two consecutive periods, the multiscale seismicity model technique is applied to the construct frequency-magnitude relation (b-value), which shows b-values of 0.80 ± 0.03, 0.79 ± 0.07, 0.77 ± 0.07, and 0.82 ± 0.08 for the whole Gulf and the Eilat, Aragonese, and Dakar basins, respectively. The estimated b-value from the declustered catalogue using the multi-scale seismicity model technique is more consistent than the values from previous studies in the Gulf, which were calculated by the classical method. All the values fall around 0.8 ± 0.02 for the whole Gulf region or each basin, which is nearly the same as the global average value. The Aragonese Basin is the most active region, as it generated three instrumentally recorded significant earthquakes (i.e., the Mw = 5.8 1993, Mw = 7.2 1995, and Mw = 5.2 2015. It is also characterized by the highest clustering degree of seismicity, as shown in the seismicity map (Fig. 3). The activity linked to the largest known events and swarms greatly amplifies the activity in this basin. Declustering the catalogue removed these effects. Therefore, the similarity of the calculated b-values for the three basins can be considered as the transform fault's fingerprint.
The mainshock-aftershock sequences, swarmtype activity, or a mix of the two make up the total seismicity of the Gulf. Our investigation revealed several low-magnitude swarm-type behaviours, including the swarms of 1983, 1990, and 2000, as well as four mainshock-aftershock sequences in 1993, 1995, 2015, and 2016. The first type is located within the Aragonese Basin, while the second type is located in the Eilat Basin. The distribution of the aftershocks, which occurred after the Mw = 5.8 1993 mainshock, reveals that the activity in the days after the mainshock is focused along a WNW-ESEtrending segment (Abdel Fattah et al., 1997). The activity of the 1993 earthquake is marked by the successive occurrence of two mainshock-aftershock sequences. The magnitudes of these mainshocks are 5.8 and 5.6, respectively. Twenty-seven months after the Mw = 5.8 August 3, 1993 event, the Mw = 7.2 November 22, 1995 mainshock occurred in the middle of the Aragonese Basin. Its aftershocks migrated northward to cover the Eilat Basin. This picture reflects the directivity of the rupture propagation (Figs. 12S and 14S in the supplementary material) (Pinar & Türkelli, 1997;Kilnger et al., 1999;Hofstetter et al., 2003;Kariche et al., 2018). Aftershocks of the 1995 event are entirely collocated within the Gulf of Aqaba. In 2015, another mainshock with Mw = 5.2 struck the Aragonese Basin again, around 20 years after the November 22, 1995 mainshock. Aftershocks of this event extend along the transverse NW-SE-trending normal fault in the northern part of the Aragonese Basin.
The seismicity patterns preceding the 1993, 1995, and 2015 mainshocks in the Aragonese Basin demonstrated an apparent seismic quiescence, which could signify an upcoming seismic event. From August 13, 1991, to August 2, 1993, the seismicity pattern before the 1993 mainshock showed a distinct intermediate period of seismic quiescence. This period was distinguished by limited seismic activity Vol. 180, (2023) Seismicity and Fault Interaction in the Gulf of Aqaba 2059 throughout the Gulf. Following that, the second type of swarm activity following the 1993 mainshock continued from August 3, 1993, to April 14, 1994 Consequently, another marked seismic quiescence was recorded along the whole Gulf before the 1995 mainshock, i.e., from April 15, 1994, until November 21, 1995. Aftershocks of the 1995 mainshock continued for about 1 year throughout the northern part of the Gulf, including the Aragonese and Eilat basins. The seismic quiescence durations observed in this study before the 1993 and 1995 main events matched those estimated by Hussein et al. (2008) from seismicity rate variations before the same events. It may be inferred that several seismicity phases mark the seismicity pattern before the Mw = 7.2 1995 mainshock. This includes the first seismic quiescence before the 1993 mainshock, the swarm of the second type, the second seismic quiescence before the 1995 mainshock, and finally, the mainshock and aftershocks. According to the model proposed by Kanamori (1981), the first seismic quiescence referred to the nature of asperity in which, following the stoppage of the swarm that came after the 1993 mainshock, stress accumulation developed due to complicated tectonics. After the 1995 aftershocks decayed, seismic activity in the Eilat and northern Aragonese basins increased from December 27, 1996, to June 27, 2003. Moreover, the seismic activity spread along the whole Gulf from December 27, 1996, to August 15, 2013. Following that, the activity dropped dramatically, forming seismic quiescence along the whole Gulf about 2 years before the 2015 mainshock. It is noted that the three significant earthquakes that occurred in the Aragonese Basin were preceded by seismic quiescence. This quiescence continued about 2 years before the 1993 mainshock, while the 1995 mainshock had shown a seismic quiescence of about 1.5 years. Although the 1995 main shock was the largest, the quiescent period was the smallest because the stress buildup began when the quiescent period preceded the 1993 main shock.
The occurrence of the three significant and proximal earthquakes in the Aragonese region over a 22-year interval, with the similarities as described, suggests a link between these earthquakes because the region is affected by several faults with an interplay of different mechanisms. Therefore, we decided to calculate the Coulomb stress changes for the three significant earthquakes that struck the Aragonese Basin on August 3, 1993. This step was achieved by compiling data such as the focal mechanism solutions and seismological information (the location, depth, and magnitude) following the mainshocks. The Coulomb failure function was modelled to investigate the interaction between the fault ruptures of the 1993 and 1995 events, and the ruptures of the 1995 and 2015 events (Figs. 5,6,7,8).
The Coulomb stress variation map created using the 1993 event as a source fault and the 1995 event as a receiver fault shows that the 1993 earthquake transferred stress which promoted the failure of the 1995 event (Fig. 6). The 1995 fault rupture is located in a loading zone which is distinguished by a positive stress change. This means that the slip on the 1993 fault increased stress at the 1995 receiver fault and consequently induced and triggered the 1995 earthquake.
The Coulomb stress change was also computed using the 1995 earthquake as a source fault and the 2015 earthquake as a receiver fault (Fig. 8). The 2015 fault rupture is situated in a loading zone with positive stress shift. This suggests that the slip on the 1995 fault, which increased stress at the 2015 receiver fault, was what caused and triggered the earthquake in 2015.
Coulomb stress modelling with a fixed receiver fault plane, the correlation between the loading lobes, and the distribution of the aftershocks for 1993, 1995, and 2015 suggests that there is close interaction between the 1993, 1995, and 2015 earthquakes. Stress changes due to the 1993 event promote the failure of the 1995 event while the 1995 event promotes the failure of the 2015 event. This means that transferred stress was enough to trigger the 1995 and 2015 events and their associated aftershocks. The link between the three earthquakes is mainly related to the existence of faults with different types of motion in the Aragonese Basin, including the N17E-trending Aragonese, the N20E-trending Arnona strike-slip faults which delimited the basin to the west and east, and the NNW-trending normal faults which delimited the basin to the north and south. The interaction of these faults with each other is the main reason for the interrelation between them.
Subsequently, the study of the seismic activity for each basin (namely the Eilat, Aragonese, and Dakar basins) from 1983 to 2018, in addition to the three significant earthquakes which struck the Aragonese Basin, indicates that the Gulf of Aqaba is one of the most seismic source zones in Egypt and the Middle East region. The Aragonese Basin is the most active of the three, and its activity is caused by the rate of active deformation and coseismic faulting behaviour of the Gulf. The seismic activity in the Dakar Basin is still ambiguous, according to Bektas (2021). This basin had been struck by the AD 1839 historical earthquake, which is apparently recorded in the sedimentary deposits observed in the cores taken from the Dakar Deep, according to Ribot et al. (2021). The Dakar Basin is bounded by the 83 km long Arnona fault which can generate earthquakes with magnitudes ranging from 7 to 7.5 (Wells & Coppersmith, 1994). Nevertheless, it has the lowest seismic activity, with a reported magnitude not exceeding about 5. The InSAR results and related deformation rate show that the Dakar basin has a significant slip rate measured along the eastern and western Gulf coastal edges that exceed nearly 10 mm/year (NRIAG, 2018). For this reason, this part of the Gulf needs intensive geological and geophysical investigations, as well as intensified seismic network coverage.

Funding
Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in cooperation with The Egyptian Knowledge Bank (EKB). The authors have not disclosed any funding.

Declarations
Competing interests The authors declare no competing interests.

Conflict of Interest
The authors have no relevant financial or non-financial interests to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.