• Tell Saidiyeh Trenches in Google Earth
  Tell Saidiyeh Trenches
  
  

  • Yellow line - Trench T-3 (to the south - bottom of image)
  • Yellow line - Trench T-4 (to the north - top of image)

  
  click on image to explore this site on a new tab in Google Earth
• Ghor Kabed Trenches in Google Earth
  Ghor Kabed Trenches
  
  

  • Yellow line - Trench T-1 (to the south - bottom of image)
  • Yellow line - Trench T-2 (to the north - top of image)

  
  click on image to explore this site on a new tab in Google Earth

• from Ferry et al. (2011:39)

The continuous record of large surface-rupturing earthquakes along the Dead Sea fault brings unprecedented insights for paleoseismic and archaeoseismic research. In most recent studies, paleoseismic trenching documents the late Holocene faulting activity, while tectonic geomorphology addresses the long-term behavior (>10 ka), with a tendency to smooth the effect of individual earthquake rupture events (M_w >7). Here, we combine historical, archaeological, and paleoseismic investigations to build a consolidated catalog of destructive surface-rupturing earthquakes for the last 14 ka along the left-lateral Jordan Valley fault segment. The 120-km-long fault segment limited to the north and the south by major pull-apart basins (the Hula and the Dead Sea, respectively) is mapped in detail and shows five subsegments with narrow stepovers (width < 3 km). We conducted quantitative geomorphology along the fault, measured more than 20 offset drainages, excavated four trenches at two sites, and investigated archaeological sites with seismic damage in the Jordan Valley. Our results in paleoseismic trenching with 28 radiocarbon datings and the archaeoseismology at Tell Saydiyeh, supplemented with a rich historical seismic record, document 12 surface-rupturing events along the fault segment with a mean interval of ∼1160 yr and an average 5 mm yr^-1 slip rate for the last 25 ka. The most complete part of the catalog indicates recurrence intervals that vary from 280 yr to 1500 yr, with a median value of 790 yr, and suggests an episodic behavior for the Jordan Valley fault. Our study allows a better constraint of the seismic cycle and related short-term variations (late Holocene) versus long-term behavior (Holocene and late Pleistocene) of a major continental transform fault.

• from Ferry et al. (2011:39-41)

The occurrence of large earthquakes on continental faults holds crucial questions on their physical and mechanical characteristics, their size, and their time distribution in terms of magnitude and frequency. Recent field investigations in paleoseismology and archaeoseismology along the Dead Sea fault (DSF) show evidence of historical coseismic surface rupturing at the Sicantarla Tell in Turkey (Amik basin; Altunel et al., 2009), the Al-Harif Roman aqueduct in Syria (Meghraoui et al., 2003), the Lebanese restraining bend (Gomez et al., 2003; Daëron et al., 2004; Nemer and Meghraoui, 2006; Daëron et al., 2007; Nemer et al., 2008), the Jordan Valley gorge and Hula basin (Ellenblum et al., 1998; Marco et al., 2003; Marco et al., 2005), the Jordan Valley (Reches and Hoexter, 1981), and the Wadi Araba (Zilberman et al., 2005; Haynes et al., 2006). Paleoseismic studies along the Jordan Valley fault (JVF) revealed the episodic activity from a long-term earthquake record (50 ka) in the Lisan lacustrine deposits (El-Isa and Mustafa, 1986; Marco et al., 1996; Migowski et al., 2004) and from the correlation between cumulative stream offsets and 48-ka-long paleoclimatic fluctuations (Ferry et al., 2007). This episodic activity is expressed not only by periods of earthquake clusters affecting a single segment but also by a sequence of earthquakes on different segments during a short period of time (Ambraseys, 2004; Sbeinati et al., 2005). The long-term record of past earthquakes contributes to better understand the faulting behavior and stability of segment boundaries (Sieh, 1996), as well as fault interactions during earthquake sequences (Stein et al., 1997). Although earthquake-induced soft-sediment deformations were largely studied in the Lisan formation, earthquake surface ruptures associated with the JVF needed a detailed paleoseismic and archaeoseismic study in order to document the earthquake sequence and related seismic cycle on a single fault segment.

The DSF forms the boundary between the African and Arabian plates and accommodates ∼1 cm yr^-1 of relative left-lateral strike-slip motion (Quennell, 1959; Fig. 1). The fault system exhibits a relatively simple geometry with large pull-apart basins distributed along strike (the Red Sea, the Dead Sea, the Hula basin, the Ghab basin, and the Amik basin) and a single major restraining bend at its center (Mount Lebanon and Anti-Lebanon). It is composed of eight major segments (Fig. 1a), all of which are capable of producing large destructive earthquakes and surface faulting as documented by an extended historical record (e.g., Guidoboni et al., 1994; Ambraseys and Jackson, 1998; Sbeinati et al., 2005; Ambraseys, 2009). However, other than the 1995 M_w 7.3 Aqaba earthquake, none of the main segments has released a large earthquake in the last eight to nine centuries (Fig. 2). The lack of seismicity and elapsed time since the most recent historical earthquakes suggest that a tectonic loading has been accumulating along most of the DSF. Global Positioning System plate velocities along the fault system suggest 2.5–6 mm yr^-1 (Fig. 1a; also, McClusky et al., 2003; Wdowinski et al., 2004; Reilinger et al., 2006; Gomez et al., 2007; Le Beon et al., 2008; Alchalbi et al., 2010) that are comparable to 4–7 mm yr^-1 geological slip rates (Garfunkel et al., 1981; Ginat et al., 1998; Klinger et al., 2000; Niemi et al., 2001; Meghraoui et al., 2003; Gomez et al., 2003; Daëron et al., 2004; Akyüz et al., 2006; Ferry et al., 2007; Karabacak et al., 2010; Sbeinati et al., 2010) measured at 2-to-100-ka time scales. It implies 3–5 m of slip deficit for the different segments and suggests an increasing potential for destructive events in the near future.

After a geology, tectonic geomorphology, and seismicity setting, we present the paleoseismic investigations with four trenches across the fault and the archaeoseismic studies at 11 different sites, with a specific focus on the Tell Saydiyeh damages. Our results yield an integrated catalog of faulting events and related large earthquakes for the last 14 ka. The earthquake catalog of faulting events that includes both clustering and quiescence periods sheds light on the distribution of interseismic periods and the associated tectonic-loading process. The long-term faulting behavior of the JVF and its relationship to neighboring segments is also discussed.

• from Ferry et al. (2011:41-42)

The north–south-trending DSF transform (Fig. 1) is made of a transtensional system to the south (including the Hula, Dead Sea, and Gulf of Aqaba pull-apart basins), the Lebanese restraining bend (the Yammouneh, Rachaya, Serghaya, and Roum faults) in the middle, and a strike-slip system to the north (the Missyaf fault and the Ghab pull-apart basin). The seismicity described by Aldersons et al. (2003) suggests that the lower crust has a brittle behavior below 20 km and possibly as deep as 32 km. This is supported by thermomechanical modeling (Petrunin and Sobolev, 2006), which suggests the brittle part of the cold lithosphere beneath the Dead Sea basin may be locally as thick as 27 km. Similarly, heat-flow measurements indicate relatively low values not typical of a rifting region (Ben-Avraham et al., 1978). Furthermore, Ryberg et al. (2007) image subvertical major faults and deep sedimentary basins along the Wadi Araba segment. Although the level of background seismicity is low (M<5), the seismotectonic characteristics along the plate boundary show a systematic pattern of strike-slip focal mechanisms and some normal faulting solutions (Salamon et al., 2003). The DSF is a typical continental transform fault and exhibits a narrow deformation zone dominated by strike-slip faulting that affects a thick cold crust. The post-Miocene left-lateral offset along the fault is estimated to be ∼45 km, which is consistent with the accumulation of more than 8 km of clastic, carbonatic, and evaporitic sediments in the Dead Sea basin (Quennell, 1984; ten Brink, 1993; Ginzburg and Ben-Avraham, 1997; Bartov et al., 2006). At a large scale, the DSF is highly segmented, and the JVF section is limited by the Lebanese restraining bend to the north and the large Dead Sea pull-apart basins to the south.

• from Ferry et al. (2011:42)

The active JVF is made of five 15-to-30-km-long subsegments (Fig. 1b; Al-Taj, 2000; Malkawi and Alawneh, 2000; Ferry et al., 2007) limited by relatively small (2-to-3-km-wide) transpressive and transtensive relay zones. Using aerial (1:25000 scale) and satellite photographs (SPOT-5, Landsat 7, Google Earth), field investigations, and offset measurements, we mapped in detail a total of 120 km of the active fault trace from the Hula basin to the Dead Sea. The active fault trace is visible within the valley and cuts through the former Lake Lisan (65–18 ka B.P.) and clastic Damya (18 ka B.P. to present) deposits. The geomorphology of the valley shows regionally flat Lisan lacustrine terraces displaying an average slope of ∼0.1° toward the present-day Dead Sea. The very fine-grained Lisan sediments (mostly varve-like detritus and aragonite), combined with a semiarid climate, produce classical badland morphology, where material is eroded away through a dense dendritic gully network. The hydrographic system provides clear markers for the study of left-lateral cumulative offsets along the fault trace (Ferry et al., 2007; Ferry and Meghraoui, 2008; Klein, 2008).

The JVF displays complex transtensional features with numerous 100-to-300-m-long and 50-m-wide pull-apart basins (Fig. 3) separated by right-stepping en echelon ruptures (fig. 3 in Ferry et al., 2007). Located in the southern section of the JVF, the impressive Ghor Katar badland area is made of numerous stream incisions that expose ∼50-m-thick Damya and Lisan lacustrine units in remarkable cliffs (Abed and Yaghan, 2000). The fault is well visible in the many incisions of Ghor Katar before it enters the flat lacustrine terrace.

Located 2.5 km south of Ghor Katar, the Ghor Kabed pull-apart system (Fig. 3c) affects the Damya and Lisan terraces and illustrates the pattern of active faulting in the valley. The accumulation of late Pleistocene and Holocene deposits in the Ghor Kabed pull-apart depressions constitutes a good record of past faulting events. Indeed, a detailed microtopographic survey (1–2-m resolution) illustrates the basin morphology and related north–south-trending 6-m-high fault scarps with 4–15° slope cutting through the middle of the depocenter. The clear fault scarp morphology and active alluvial and lacustrine sedimentary processes present a good potential for paleoseismic trenching at this site (see the Paleoseismology section).

In the middle section of the JVF, at the Tell Saidiyeh area, small stream offsets and abandoned beheaded channels provide an ideal site for faulting event characterization and slip-rate calculations (Fig. 3d). We performed a detailed microtopographic survey (Fig. 3d) of the site in order to study the cumulative offsets along the fault revealed by the drainage system. These stream channels expose the fault and exhibit a conspicuous shear zone with evidence of recent faulting (Fig. 4b). On the eastern block, the drainage system consists of a small catchment area to the north and a single linear stream (E1) flowing from the east with a 60°N–80°N direction that cuts into an abandoned alluvial terrace (Qt0). Because of the catchment area, the northern bank of that stream has been eroded and modified, and only the southern bank is adequately preserved. Against the fault, stream E1 flows into a marshy water hole that may be partly man-made based on an existing depression. On the western block, two strongly incising gullies flow westward to the valley. The northern one (W1) continues west of the water pond (E1) along a 60°N direction and displays a subtle offset of 7 ± 0.5 m. The southern one (W2), while being very well expressed, has no counterpart east of the fault where a small catchment area has been formed by regressive erosion. Hence, the only possible source for W2 is E1, making W2 a beheaded remnant left-laterally offset by 114 ± 5 m. Because W1 and W2 cut into the upper surface of the Lisan formation, they necessarily are younger than its ultimate deposits; that is, they are younger than 25 ka B.P. (Abed and Yaghan, 2000). Furthermore, a trench exposure (see the Paleoseismology section, Fig. 4a, and Fig. 5) reveals channel deposits related to Qt0, which have been subsequently radiocarbon dated at 19,700–16,800 B.C. (see sample L-02 AkR fraction in Table 1). Considering that sample L-02 originates from the middle of the stratigraphic section (see Fig. 5), the associated age is a minimum value for the emplacement of the channel, which suggests a minimum age of 22 ka for W1 and W2. Additionally, because the drainage source is limited to the surface of the late Lisan terrace, we may adopt the approach developed by Ferry et al. (2007) in the same region and assume the inception of that drainage was triggered by an abrupt lake-level drop of Lake Lisan. The present elevation of the terrace at Tell Saidiyeh is ∼255 m below sea level, which, from the lake-level curve of Bartov et al. (2002) and inferences by Ferry et al. (2007), yields an age of 21–25 ka B.P. for the latest level drop below that elevation and independently confirms our age inference. Taking into account the dating of channels and terraces, we infer that the total left-lateral offset of 114 ± 0.5 m has been accumulated during the last 22–25 ka. The resulting long-term average slip rate is 4.9 ± 0.3 mm yr^-1 for that period, which is in good agreement with a previous geological slip rate obtained from 20 offset streams (Ferry et al., 2007).

• from Ferry et al. (2011:42,46-47)

The Dead Sea fault exhibits scarce instrumental seismicity with mostly low to moderate earthquakes (M<6, Fig. 2a) mainly concentrated along the Lebanese Bend and the Jordan Valley fault with three recorded moderate earthquakes (11 July 1927 M_w 6.2; 23 April 1979 M_b 5.2; and 2 February 2004 M_L 5.2). However, the DSF is capable of producing large destructive events, as attested by the 22 November 1995 M_w 7.3 Aqaba earthquake (Hofstetter, 2003) and by large historical events (Fig. 2b; Ambraseys and Jackson, 1998; Sbeinati et al., 2005).

The 11 July 1927 event is the most destructive earthquake to strike the region in the last century, with a known toll reaching 285 people killed and ∼1000 injured. Widespread destruction was documented by Willis (1928) in Amman, Ramallah, Nablus, the Mount of Olives ("a mile east of Jerusalem"), Reineh (Nazareth), As Salt, and Jericho (Fig. 6 for location). The event was recorded at more than 100 seismological stations throughout the world, and its epicenter was located within the northern basin of the Dead Sea (Shapira et al., 1993). On the basis of reinterpreted historical documents, Avni et al. (2002) confirm these findings and describe a seiche wave in the Dead Sea that pleads for either a submarine landslide or a displacement of bathymetry along a surface rupture. Submersible images of the bottom of the Dead Sea (Lazar and Ben-Avraham, 2002) show an apparently fresh and sharp scarp continuing the Jordan Valley fault into the Dead Sea. It should be considered that the 1927 M_w 6.2 earthquake may have produced surface rupture locally with tenuous displacement, in the range of a few tens of centimeters (Wells and Coppersmith, 1994).

The 23 April 1979 event (epicenter 31.24° N, 35.46° E) was extensively instrumentally and macroseismically studied by Arieh et al. (1982). It reached I_o IV–V MSK (though isoseismal lines are available for Israel only) in the Jordan Valley and, according to Arieh et al. (1982), its focal mechanism points to a 20°N-striking border fault of the northeastern Dead Sea basin. The 2 February 2004 event (epicenter 31.69° N, 35.58° E) was felt strongly in Jordan, Israel, Palestine, and Syria and produced slight damage in Jordan and Israel, injuring a total of 20 persons (Jordan Seismological Observatory, 2004). The combined analysis of aftershock distribution and fault-plane solution suggests a normal fault branch perpendicular to the JVF, probably associated with the Dead Sea pull-apart (Al-Tarazi et al., 2006). One may notice that the instrumental seismicity does not reflect the level of active deformation of the JVF and its potential for large earthquakes.

• from Ferry et al. (2011:48-49)

The historical seismicity relies on inscriptions and documents from Greek, Hebrew, Roman, Byzantine, Arabic, and Ottoman times (Guidoboni et al., 1994; Ambraseys, 2009). A wealth of testimonies, invoices, and reports are available for the last two millennia and document one of the most complete historical catalogs to date. However, it should be noted that the region was not evenly populated at any time in the past, with densely populated areas along the Mediterranean coast and the shores of the Sea of Galilee and the Dead Sea and barren areas in the Negev Desert and Wadi Araba. This situation induces a systematic bias into intensity maps by (a) shifting epicenters northward and (b) restricting the extent of felt testimonies. Furthermore, magnitudes derived from historical studies are usually very delicate to assess and carry a large (and undefined) uncertainty, especially for older events. Recent findings by Katz and Crouvi (2007) show that anthropogenic tali of archaeological origin have very poor geotechnical characteristics and may amplify seismic shaking. Previous works suggest that M_S >8 historical earthquakes possibly occurred in the Jordan Valley (Ambraseys and Jackson, 1998). However, the length of fault segments and thickness of the seismogenic crust suggest that M_w 7.2–7.4 is a reasonable maximum magnitude in this region.

Based on historical seismicity catalogs and recent literature, we list here the main earthquake events that occurred in the Jordan Valley (Fig. 2b):

• Event ZH (A.D. 1033, 10 Safar 425 A.H.): According to Ibn Al-Jawzi (A.D. 1113–1200), on that morning the walls of Jerusalem crumbled down during construction, and the cities of Jerusalem, Ramallah, Jericho, Nablus, and Tiberias were heavily damaged (Abou Karaki, 1987). This event was felt throughout Judea and possibly as far away as Egypt and Syria, and it produced a sea wave along the Mediterranean coast (Fig. 6 for locations; Poirier and Taher, 1980; Ambraseys et al., 1994).

• Event YH (A.D. 749): Theophanes (A.D. 760–818), a historian whose work constitutes one of the main sources for that period, narrates "a powerful earthquake in Palestine, along the river Jordan and throughout Syria, and countless thousands of people were killed, and churches and monasteries also collapsed, especially in the desert near the Holy City [Jerusalem]" (Ambraseys, 2009, p. 232). This event has been intensively studied by numerous authors, due in great part to inconsistencies between calendars (Abou Karaki, 1987; Tsafrir and Foerster, 1992).

• Event XH (759 B.C.): This earthquake produced great destruction and many casualties in Judea, Samaria, and Galilee (Guidoboni et al., 1994). A thorough reappraisal of this event by Ambraseys (2005) indicates that few contemporary accounts are available for this event, the earliest one being the Book of Amos. The first detailed description is given by Zachariah around 520 B.C. (i.e., ∼240 years later) and suggests that a large landslide developed on the Mount of Olives, southeast of Jerusalem, without a clear causative link.

Additionally, numerous other earthquakes have been felt in the Jordan Valley in historical times but may not be considered as candidates for surface-rupturing events along the JVF:

• One may consider a candidate earthquake in A.D. 418 (not A.D. 419, as justified by Ambraseys, 2009). However, evidence is very weak because contemporaneous chroniclers (Marcellinus Comes, Philostorgius; see also Ambraseys, 2009) and archaeological investigations (Meyers et al., 1976) describe earthquake damage in the north of the Galilee region. There is no mention of major damage or victims in Jerusalem, Jericho, and Palestinian villages located in the Dead Sea vicinity in A.D. 418.

• The A.D. 363 earthquake is better documented and consists of a sequence of two shocks on 18 and 19 May (Ambraseys, 2009). Although a large number of sites in Palestine, including Jerusalem, were damaged and a sea wave was observed in the Dead Sea, further south half of Petra was razed to the ground, and localities near the Red Sea were badly damaged (Niemi and Mansoor, 2002). The wide region of damage and the ∼250-km-long Wadi Araba fault from the Dead Sea to the Gulf of Aqaba to the south may well be the site of two major shocks of A.D. 363.

• Historian Flavius Josephus (A.D. 37–100) vividly describes an event in 31 B.C. in the Jewish Wars: "For in the early spring, an earthquake shock killed an infinite number of cattle and 30 thousand people; but the army was unharmed, because it was camped in the open" (Guidoboni et al., 1994, pp. 173–174). A critical and exhaustive reappraisal of that event by Ambraseys (2009) suggests that Flavius Josephus' account—the only coeval source—is greatly exaggerated, with a number of casualties larger than the actual population of the region. The author also concludes that previously reported damage to archaeological structures is actually spurious and not associated with earthquake faulting. In summary, a small to moderate earthquake probably occurred in 31 B.C. but did not produce significant damage to buildings or surface rupture. As mentioned by Ambraseys (2009, p. 101), "The reappraisal of the available data reveals nothing more than that the 31 B.C. earthquake in Judaea that caused damage and loss of life, which Josephus grossly exaggerates. There is no evidence that Jerusalem was affected and the destruction or damage of other historical sites in Judaea is conjectural and cannot be tested on archeological ground. The association of the earthquake with a fault break at Khirbet Qumran seems to me untenable and I can find no justification for the addition of Diospolis to towns affected and the dating of the event to A.D. 31 (Guidoboni et al., 1994; Guidoboni, 1989)."

• Previous studies consider an event in 64 B.C. that would have damaged the Temple in Jerusalem and been felt throughout the region and as far away as Antioch (southeast Turkey). However, the critical reinterpretation by Karcz (2004) strongly suggests an opposite situation, with an event source near Antioch (probably along the Hacipasa fault or the Karasu fault) and related seismic shaking felt as far as Jerusalem. This is supported by the fact that only minor damage was reported in the Jordan Valley. Indeed, this event is most likely the 65 B.C. Antioch earthquake that claimed some 170,000 lives (Guidoboni et al., 1994; Sbeinati et al., 2005).

• from Ferry et al. (2011:55-56,61)

Rift valleys throughout the world are common passageways for human populations and are thus generally rich with the archaeological heritage of former civilizations. The evolution and migration of human groups with respect to active faults has hence long been established (King et al., 1994). Here, we first provide detailed evidence for earthquake-induced destruction at Tell Saidiyeh, an archaeological site that sits a few tens of meters from the active trace of the JVF (Fig. 3) and which has been studied by means of paleoseismic excavations (trenches T3 and T4 in section 5). Additionally, we present a critical reappraisal of published data about 20 archaeological sites scattered over the Jordan Valley and neighboring regions (Fig. 3 and Fig. 6), which show varying degrees of evidence for earthquake-related damage and possibly surface rupture. The archaeological evidence for paleoearthquakes takes the form of destruction to buildings with frequent fires and consequent signs of site abandonment. Thus, a widespread burnt layer with a noticeable content of rubble, pottery shards, and ashes may be a good candidate for earthquake evidence. However, indication for an earthquake is generally reduced to signs of destruction that can be instead related to regional wars, local raids, or even accidents (e.g., accidental fire caused by an unattended oil lamp). In a recent review, Ambraseys (2006) underlines the different caveats pertaining to the use of archaeological evidence to identify past earthquakes and particularly to the interpretation of toppled structures. We follow Ambraseys’s guidelines in making our interpretations of existing archaeological data from Tubb (1988 and 1998), Savage et al. (2001, 2002, and 2003), and Franken (1989) and retain the most robust evidence to identify destruction events that may be attributed to past earthquakes.

For example, Tell Saidiyeh (Fig. 8) is located almost exactly at the center of the Jordan Valley (Fig. 6), which suggests the site would probably experience intense ground-shaking in the case of a major earthquake on the JVF. Tell structures are common throughout the Middle East and are composed of layers of successive settlement (spreading over a few centuries to several millennia), which may pile up to reach 30–50 m elevation above the base level. Tell Saidiyeh has been studied for the last 60 years and been the object of systematic excavations since 1964 (Tubb, 1988). The site has produced a wealth of artifacts that document rather continuous occupation from the Chalcolithic (fourth millennium B.C.) up to the Roman period (Tubb, 1998). The tell itself is composed of two distinct mounds (Fig. 8), with noticeably different histories: an upper main tell where elaborate structures were discovered (e.g., an olive-pressing complex and a water staircase; see Fig. 8d) and a lower tell that was mostly used as a burial ground during the Omayyad period. A compilation of indications for strong perturbations to Tell Saidiyeh (Table 2) point to two specific strata for which destruction is significant: stratum L2, dated ∼2900 B.C., and stratum XII, dated 1150–1120 B.C. For both strata, the author mentions widespread damage with intense burning, collapsed walls, and broken potteries (Fig. 8c). An additional event may be linked to stratum VI, dated to the middle of the eighth century B.C., and may correspond to the 759 B.C. Jericho earthquake (Nur and Cline, 2000). Indeed, Tubb (1998, p. 126) mentions that “houses of Stratum VI were knocked down and leveled in preparation for another major building programme.” The reason for such a drastic solution may be the prior intense destruction of the tell by an earthquake with no possibility to re-use damaged buildings. Other strata show evidence for destruction or abandonment but could not be related to seismic shaking.

In parallel, we compile existing data for 11 archaeological sites in the vicinity of the JVF showing a potential for past earthquake damage (Fig. 6 and Table 2). However, we consider earthquake-induced damage in archaeological sites only if it is attested by a minimum of two sufficiently distant sites and sites with evidence for surface rupture. The analysis of damage to archaeological sites allows us to identify four events likely associated with large earthquakes along the JVF:

• Event ZA: This event is attested at Tell Deir’ Alla (Franken, 1989) and Tell Saidiyeh (Tubb, 1988) for the middle of the eighth century B.C. Available descriptions lack details, and it is probable that the corresponding destruction has been indirectly associated with the well-known 759 B.C. Zechariah’s earthquake (Nur and Ron, 1996) without further age determination. At Tell Saidiyeh, damage is not directly associated with an earthquake, but it is rather the subsequent massive leveling of the site that suggests a catastrophic event. Proposed date: 759 B.C.

• Event YA: This event is attested at the neighboring sites of Tell Saidiyeh and Tell Deir’ Alla, where the occurrence of an earthquake in the early twelfth century B.C. leaves no doubt for Franken (1989) and J. N. Tubb (personal comm., 2005). At Tell Al’ Umayri, ∼30 km east of the JVF, contemporary destruction associated with a burn layer rich in broken vessels is documented by Savage et al. (2001). It should be noted that this event may be part of the well-documented twelfth century B.C. “earthquake storm” studied by many authors and summarized by Nur and Cline (2000). Proposed date: ∼1150 B.C.

• Event XA: This event is documented at Tell Abu en-Ni’aj by Savage et al. (2003) as a series of ash layers offset by a fault splay. According to our mapping of the JVF (Fig. 6; Al-Taj, 2000; Ferry et al., 2007), Tell Abu en-Ni’aj appears to be west and off the main active trace by ∼1 km. Savage et al. (2003) do not provide details about the amount or intensity of deformation along that fault, and it is presently not possible to decide whether the observed offsets are related to coseismic fault slip. At Khirbet Iskander, located ∼30 km southeast of the JVF (Fig. 6), Savage et al. (2003) describe a destruction layer in great detail (Table 2), where fire was attested by burnt stones, ash layers, and charred grain. Well-preserved wooden beams and human remains suggest a sudden possibly earthquake-related catastrophe. Proposed date: ∼2300 B.C.

• Event WA: At Tell Saidiyeh, Tubb (1988) documents dense destruction debris associated with fragmentary and disturbed architecture and indicates these are consistent with ground-shaking-related damage (J. N. Tubb, personal comm., 2005). At the nearby Tell el-Fukhar site, Savage et al. (2003) mention evidence of intense destruction to walls and infer a possible earthquake-related origin. Proposed date: ∼2900 B.C.

Overall, our compilation of archaeological studies on tells in the vicinity of the JVF strongly suggests the occurrence of four events: in 759 B.C., ∼1150 B.C., ∼2300 B.C., and ∼2900 B.C. Here, the archaeological record intersects the historical record, as attested by the 759 B.C. earthquake described in written documents. Furthermore, the archaeoseismic event XA can be correlated with the paleoseismic event YT4 and compensates for the sedimentary hiatus observed in paleoseismic trenches (period ∼1500 B.C. to ∼4500 B.C.). It should be noted that indications for surface breaks affecting tells may not necessarily be considered as faulting evidence. Indeed, tells are artificial mounds made of heterogeneous material that includes rubble, dirt, and architectural remains (Fig. 8a). As such, they are very sensitive to gravitational collapse, especially under seismic shaking. Hence, for sites that are not directly located across the main active trace of the JVF, we consider surface deformation as secondary evidence in the sense of McCalpin (1998). Nonetheless, such indications may provide a relevant chronological constraint for a seismic event. Thus, the mention by Franken (1989, p. 203) of “a victim […] found completely squashed in a crack in the earth” at Tell Deir’ Alla (∼3 km east of the JVF) and Savage et al.’s (2003) previously mentioned “earthquake slip fault” at Tell Abu en-Ni’aj are not considered as strong evidence for surface rupture but may document off-fault effects of a large earthquake.

• from Ferry et al. (2011:61)

We determined 3 seismic events from historical studies (ZH–XH), 4 damaging events from archaeological studies (ZA–WA), and 12 faulting events from paleoseismic investigations (ZT2 and YT2, ZT3, ZT4–ST4, and CT4–AT4). The consistency in the correlation between historical, damaging, and faulting events constrains the occurrence of the earthquake events. Table 3 presents the resulting catalog of 15 large earthquakes (Z–O and C–A) produced by the JVF between A.D. 1033 and 12,060 B.C. and between 15,000 B.C. and 18,000 B.C. (Fig. 9). It should be noted that an additional event was identified in trench T2 (WT2) for which we could not estimate an age. The fusion of the different datasets reinforces the identification of past earthquakes. Indeed, of the 13 paleoseismic events, two are also present in the historical record and one (possibly two) in the archaeological record. One event is present in both historical and archaeological datasets.

• from Ferry et al. (2011:61-62)

With major barriers (the Dead Sea and the Hula basin) to the rupture propagation toward nearby fault segments and very weak structural relays (stepovers) within the JVF, it is likely that large earthquake ruptures are characteristic in length and associated with M_w 7.2–7.4 and a ∼3.3 m average coseismic displacement. An estimate of the total seismic moment release for the 12 faulting events (Z–O in Fig. 9) during the last 14 ka yields an ∼3.3 mm yr^-1 slip rate, which probably indicates a lack in the paleoseismic record (possibly due to the sedimentary hiatus). By contrast, using the total seismic moment for the admittedly most complete part of the catalog (events Z–U) over a period of ∼3.9 ka, we obtain an ∼4.2 mm yr^-1 slip rate (Fig. 10), which is comparable to the slip rate obtained from offset streams and paleoclimatic fluctuations (Ferry et al., 2007). The slight difference is easily explained by distributed deformation around the main trace (possibly ∼0.5 m per event).

Taken as a whole, our integrated catalog of past seismic events yields a mean recurrence interval of 1165 yr (standard deviation 243 yr). A close examination reveals highly variable recurrence intervals with values ranging from 284 yr to 2700 yr (Table 3). To some extent, we agree that very high values may reflect the incompleteness of our catalog, especially for its purely paleoseismic part (events T–O). However, long intervals are also present in the assumedly complete historical and archaeological parts, between events Y and X (1508 yr) and events W and V (1150 yr), which suggests very long quiescence periods are real. In parallel, short interval values are clearly observed through the whole catalog: between events Y and Z (284 yr), between events W and X (391 yr), and across events O, P, and Q (190–520 yr). We propose that these short intervals reflect clustering and a generally episodic behavior of the JVF over the last 14 ka. This independently confirms the observation by Ferry et al. (2007) of slip rate variations along the JVF from a long-term value of 4.9 mm yr^-1 to a peak value of 11 mm yr^-1 during a 2-ka-long period.

• from Ferry et al. (2011:62-64)

The field investigations and collected data, their analysis, and results provide a remarkable succession of past earthquakes and related rupture parameters along a single segment of the DSF. We have identified several issues that can be summarized in:

1. The complex alluvial stratigraphy (hiatuses and channeling) and related uncertainties in radiocarbon dating led to only partial identification of some palaeoseismic events (e.g., A, B, C, R, and T).
2. The possibility that the palaeoseismic record is missing a small number of events.
3. Uncertainty in the accuracy of identifying seismic clusters, episodes of quiescence, and the definition of interseismic periods.

However, the combination of paleoseismic trenching, historical studies, archaeoseismic results, and their relationships allowed a most favorable characterization of paleoseismic events Z–V.

The analysis of paleoseismic results revealed nine faulting events that we attempted to correlate (Table 3) with three historical events (A.D. 1033, A.D. 749, and 759 B.C.) and four archaeoseismic events (759 B.C., 1150 B.C., 2300 B.C., and 2900 B.C.). The oldest part of our catalog is based on paleoseismic trenching alone, which we strengthened by exploring any alternative interpretation. In trench T4, event VT4 is interpreted on the basis of increasing displacement with depth along FZ3 and related splays, suggesting that unit i1 has been affected by at least one extra event with respect to unit h. An alternative explanation might be that unit h is isopach, while unit i1 is an erosive channel fill with a noticeable dip to the west–southwest (direction of flow). In this case, any horizontal movement would produce apparent vertical displacement: all offsets described along FZ3 could have occurred during event ZT4 only. However, our interpretation is supported by a second observation close to the main shear zone (see labels in Fig. 5d). Besides, the upper unit c appears to have been deposited over a preexisting scarp across the main shear zone (FZ1) or to have been warped with ∼10 cm vertical displacement (Fig. 5d). In the former case, the most recent displacement would have occurred between A.D. 87 and A.D. 1920, thus indicating either of the historical events of A.D. 749 and A.D. 1033. In the latter case, warping would have taken place after the deposition of unit c and before the deposition of the wedge unit b (i.e., after A.D. 1660) at a time when no large earthquake is attested for the region, which is very unlikely. Over the four trench exposures, we collected and dated 28 radiocarbon samples, most of which yielded large (>1 mg) amounts of carbon (AoC) and adequate uncertainties (20–50 yr) and appeared to be in stratigraphic order (Table 1 and Fig. 5). Some samples present uncertainties that should be clarified: Samples Ka-T3-S03 (T1), Ka-T3-6N (T2), L-14 (T4), and Tbc-04 (T4) present low to insufficient AoC (0.2–0.8 mg) and were not used for event determination. Although alkali residue (AkR) fractions of samples L-02 (T4) and L-07 (T4) yield low AoC, their respective carbonatic and humic acid fractions yield similar ages that confirm the quality of the results. However, because sample L-02 is older than any other sample of the section by some 5 ka, including the stratigraphically older samples L-07 and L-06, we consider it is redeposited and chose to reject it. Sample Tbc-18 presents an extremely low AoC and needed to be normalized to specially prepared targets. Because it is used to define event R (paleoseismic event VT4) and rejuvenation is very likely, we relied on nearby samples and stratigraphy to correct its age (see description, Trench 4). We used the same approach for samples Tbc-23 and Tbc-24, which bracket event T (paleoseismic event XT4) and show a relatively low AoC (0.6 mg each), similar ages, and age inversion. Although we applied compensation for rejuvenation of the samples, it is possible that events R and T are actually slightly older. This would move event T closer to event S and event R closer to events O, P, and Q, while extending the empty period between events R and S.

Although we combined three independent datasets into a uniquely long catalog of large earthquakes for the JVF, it is possible that some events are still missing. The mean recurrence interval for our paleoseismic record is 1480 yr, almost double the 787 yr mean recurrence interval obtained for the assumedly complete historical and archaeological periods (i.e., the last 14,000 years). This may point either to the incompleteness of the long-term paleoseismic record (one extra event has been identified but not dated) or to different faulting behaviors over the two periods. Considering the situation of alluvial deposits in channels and the relatively long sedimentary hiatus observed in trenches (Fig. 5), it is likely that part of the sedimentary record of paleoseismic events has been truncated. Furthermore, plotting age versus cumulative displacement for our catalog (Fig. 10) reveals that the paleoseismic period does not fit cumulative offsets measured on streams. This may be only partly explained by distributed off-fault deformation (Griffith et al., 2010).

Our catalog of seismic events allows us to identify earthquake clusters and quiescence periods and suggests episodicity for the JVF over the last 14 ka. It appears that the majority of events from our catalog (9 events out of 12) is grouped into clusters of two or three earthquakes: events Y and Z (cluster YZ, interval of 284 yr), events W and X (WX, interval of 391 yr), events U and V (UV, interval of 600 yr), and events O, P, and Q (OPQ, mean interval of 330 yr). These clusters are generally preceded and followed by long periods of quiescence: up to 1800 yr after cluster OPQ, 2335 yr before cluster UV, 1150 yr between clusters UV and WX, 1508 yr between clusters WX and YZ, and at least 977 yr after cluster YZ (time since the A.D. 1033 historical earthquake). Although one may also consider events A, B, and C as a cluster, the large uncertainty associated with radiocarbon dating (±1500 yr) implies that the three events may have occurred any time within a 3-ka period.

Considering that the inland 120-km-long JVF may extend southward inside the Dead Sea basin for a further ∼20 km (Lazar and Ben-Avraham, 2002; Ben-Avraham and Schubert, 2006) and northward in the Hula basin for ∼10 km (Marco et al., 2005), this yields a maximum ∼150 km surface-rupture length. Hence, some of our inferences rely on a maximum M_w 7.2–7.4 magnitude and a related average ∼3.3 m coseismic left-lateral slip estimated from a maximum 120–150-km surface-rupture length (Kanamori and Anderson, 1975; Wells and Coppersmith, 1994). Although the 22 November 1995 M_w 7.3 Aqaba event is the only large modern earthquake observed along the DSF until today, its rupture parameters may not serve as a base of comparison for earthquakes along the JVF. Indeed, it was composed of two or three subevents along offshore faults (Hofstetter, 2003), and no surface rupture could be observed so far. Unfortunately, the only coseismic slip values available were modeled and may not be adequately compared to surface-rupture values (Hofstetter, 2003). However, several coseismic or cumulative displacements were actually measured along other segments of the DSF and typically show coseismic values around 3 m (e.g., Klinger et al., 2000; Gomez et al., 2003; Meghraoui et al., 2003; Marco et al., 2005; Haynes et al., 2006), which suggests that a similar value may be considered for the JVF. Furthermore, results of recent earthquakes, such as the 1999 M_w 7.3–7.4 Izmit earthquake (Turkey) and related right-lateral strike-slip faulting, yield a comparable estimate with ∼140 km rupture length and an average 3 m coseismic slip (Barka et al., 2002).

• from Ferry et al. (2011:64)

The JVF is the main source of destructive earthquakes for the Jordan Valley region. Its characterization has crucial implications for the seismic hazard assessment of large urban areas such as Jerusalem, Amman, and Irbid, as well as to numerous historical and archaeological heritage sites such as Pella, Jerash, Madaba, Qumran, Jericho, and Megiddo (Fig. 6).

Through an integrated approach involving (1) the compilation of historical seismicity, (2) the careful reappraisal of archaeological data, and (3) detailed fault mapping, tectonic geomorphology, and paleoseismic trenching, we produce an original catalog of at least 12 surface-rupturing events in the last 14 ka and at least 16 events in the last 17 ka along the Jordan Valley segment of the Dead Sea fault. The mean recurrence interval (787 yr) indicates that the historical and archaeological record might be complete, while lower and upper bounds of the extreme recurrence intervals (284–1508 yr) imply a generally time-episodic behavior. The plausibility of this episodic model should be compared to apparent episodicity of conventional renewal models (Fitzenz et al., 2010). In contrast, the paleoseismological dataset shows incompleteness with a mean recurrence interval of 1480 yr, suggesting sedimentary hiatuses and a gap in the geological record.

Taking into account that the historical and archaeological datasets provide a mostly complete catalog of surface-rupturing events, we obtain a 787-yr mean recurrence interval, ∼3.3 m of slip per event (derived from Wells and Coppersmith, 1994), and a mean 5 mm yr^-1 slip rate for the last 5 ka (or 4.8 mm yr^-1, using a regression curve; see Fig. 10), in agreement with the mean value obtained by Ferry et al. (2007) for the last 48 ka. This is also confirmed by the direct measurement of a stream offset at Tell Saydiyeh that yields 4.9 ± 0.3 mm yr^-1 for the last 25 ka (see Tectonic Geomorphology along the Jordan Valley Fault). Finally, the last millennium of seismic quiescence along the JVF indicates up to 5 m of slip deficit and points to either an imminent earthquake and/or a future earthquake cluster similar to the A.D. 749/A.D. 1033 sequence. Neighboring segments being at a slightly lower, if not similar, level of tectonic loading (i.e., no large events since the twelfth-century sequence), it is plausible that a present-day large earthquake on the JVF may trigger earthquake ruptures on nearby segments.

• from Ferry et al. (2011:64-65)

Digital elevation model data used to produce maps are from Shuttle Radar Topography Mission (SRTM) 3 topography from the National Aeronautics and Space Administration and are available from the Consultative Group for International Agriculture Research–Consortium for Spatial Information (CGIAR–CSI) Consortium for Spatial Information at https://srtm.csi.cgiar.org/ (last accessed March 2009). Bathymetry used for maps in Figure 1 and Figure 2 is SRTM 30 produced by the University of California, San Diego https://topex.ucsd.edu/ Instrumental seismicity in Figure 2 was obtained from the Incorporated Research Institutions for Seismology Data Management Center at https://www.iris.edu/ (last accessed April 2007). Calibration of radiocarbon ages was performed using the OxCal software (Bronk Ramsey, 1995), available at https://c14.arch.ox.ac.uk/oxcal.html with the IntCal04 calibration curve (Reimer et al., 2004).