• from Kagan et al. (2011) Correction
• these have been incorporated into the Master Seismic Events Tables for all sites

• from Kagan et al. (2011)
• these have been incorporated into the Master Seismic Events Tables for all sites

Image	Description	Source
All 3 Western Sections at Site ZA-4 Photo by Jefferson Williams 06 Feb. 2023	All 3 Western Sections	Jefferson Williams
Entire Western Section Photo by Jefferson Williams 06 Feb. 2023	Entire Western Section	Jefferson Williams
Entire Western Section at Site ZA-4 (view from below) Photo by Jefferson Williams 05 Feb. 2023	Entire Western Section (view from below)	Jefferson Williams
Top Left of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Top Left of Western Section	Jefferson Williams
Top Left and Top Right of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Top Left and Top Right of Western Section	Jefferson Williams
Abovet Top Middle of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Above Top Middle of Western Section	Jefferson Williams
Top Middle of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Top Middle of Western Section	Jefferson Williams
Mid Middle of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Mid Middle of Western Section	Jefferson Williams
Bottom Middle of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Bottom Middle of Western Section	Jefferson Williams
Bottom of Western Section at Site ZA-4 Photo by Jefferson Williams 05 Feb. 2023	Bottom of Western Section	Jefferson Williams

Image	Description	Source
All 3 Western Sections at Site ZA-4 Photo by Jefferson Williams 06 Feb. 2023	All 3 Western Sections	Jefferson Williams
Entire Western Section Connector at Site ZA-4 Photo by Jefferson Williams 06 Feb. 2023	Entire Western Section Connector	Jefferson Williams

Image	Description	Source
All 3 Western Sections at Site ZA-4 Photo by Jefferson Williams 06 Feb. 2023	All 3 Western Sections	Jefferson Williams
Far Western Section at Site ZA-4 Photo by Jefferson Williams 10 Feb. 2023	Far Western Section	Jefferson Williams

Seismite Samples

RC Depth (cm)	Sample Depth (cm)	Thickness (cm)	Quake	Sample  Name	InSitu Image	Extracted Image	Notes
195	192	~44	419 CE	NZ423K	NZ423K in situ Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023 NZ423K in situ Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023 193 cm. seismite Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023	NZ423K after extraction and "shaving" Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023
195	192	~44	419 CE	NZ423K - resample	NZ423K resample in situ Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023 NZ423K resample in situ Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023 193 cm. seismite Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023	Top Top of NZ423K resample after extraction and "shaving" Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023 Bottom Bottom of NZ423K resample after extraction and "shaving" Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023
240	251 (2018) 249 (2023)	16 (2018) 16 (2023)	363 CE	NZ423I	NZ423I in situ Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023	NZ423I after extraction and "shaving" (rightside up) Click on photo to open 3D Lidar Scan in a new tab (Sample is upside down in new tab) Photo by Jefferson Williams 2023
380	400 (2023) 389 (2018)	7.5 (2018) 7.5 (2023)	~31 CE	NZ423A	NZ423A in situ Click on photo to open high resolution magnifiable image in a new tab Photo by Jefferson Williams 2023	NZ423A after extraction and "shaving" Click on photo to open 3D Lidar Scan in a new tab Photo by Jefferson Williams 2023

Seismite Samples

RC Depth (cm)	Sample Depth (cm)	Thickness (cm)	Quake	Sample  Name	InSitu Image	Extracted Image	Notes
	387	12		NZ415C	NZ415C and NZ415D in situ Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2015	Hardened Sample NZ415C Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Thin Section of Sample NZ415C Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams
	406	8		NZ415D	NZ415C and NZ415D in situ Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2015	Hardened Sample NZ415D Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Slide Billet Sample NZ415D Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Thin Section Slide Sample NZ415D Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams
	405	7		NZ418F	NZ418F in situ Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2018 NZ418F in situ Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2018	Slide Billet NZ418F Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Thin Section of NZ418F with ruler Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Thin Section of NZ418F with ruler Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams
	466 in 2015 464 in 2018	10 in 2015 12-14 in 2018	31 BCE ?	NZ415G	NZ15G and other samples in situ Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2105 Seismite at 466 cm depth (Middle Outcrop 2015) Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2018 Seismite at 464 cm depth (Middle Outcrop 2018) Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams 2018	Hardened Sample NZ415G Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Hardened Sample NZ415G with ruler Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Slide Billet of Sample NZ415G Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams Thin Section Slide of Sample NZ415G Click on photo to open magnifiable image in a new tab Photo by Jefferson Williams

• from Ken-Tor et al. (2001a:2226-2227)

The ages of the deformed layers were constrained by ¹⁴C dates derived from the layers themselves and by extrapolating between ages of layers directly beneath and above a particular mixed layer (Table 2). The radiocarbon ages of the mixed layers were calibrated to calendar years according to Stuiver et al. [1998]. The calibrated ages lie in ranges that are defined by the 2σ envelope error of the measured data. For example, the measured radiocarbon age of event A (2120±40 years B.P., Table 1) is calibrated to the range of 360 to 40 years B.C. Applying the superposition principle and rate of sedimentation further narrowed these ranges. Samples which are stratigraphically lower in the section, and thus older, reduce the calendar range of the samples above. This procedure allows us to resolve the calendar ages of samples that yielded analytical data that are statistically indistinguishable. The use of sedimentation rates allows determining ages of mixed layers that were not directly dated by radiocarbon because of lack of organic debris. The sedimentation rate is sensitive to depositional hiatuses in the section. Thus, different rates are used for different parts of the section (see varying slopes in Figure 3b).

The radiocarbon sample collected from a mixed layer is dating the deposition within the layer before it was disturbed. Hence the seismic event that caused the deformation is younger. The time elapsed between the deposition and the deformation by the earthquake was short relative to the uncertainty in the dating. Two lines of evidence support this assumption:

1. Deformation of the mixed layer occurred at the sediment-water interface. This is indicated by graded bedding and by the juxtaposition of mixed layers with syndepositional faults in the Lisan Formation. Thicker accumulation above the mixed layer on the downthrown block is considered as evidence for a fault scarp. Some local mass transport near the scarp produced thicker mixed layers in the lower block [Marco and Agnon, 1995].


2. In the Ze'elim section, two of the mixed layers, A and B, are separated by only about 75 cm and approximately 200 ¹⁴C years (Figure 2). Thus, the time elapsed between the deposition and disturbance of unit A must have been shorter than 200 years.

Liquefaction is often attributed to subsurface deformation. In the Ze'elim section liquefaction seems to have occurred at the surface. This surface deformation is consistent with the absence of sand volcanoes and dikes.

A final assessment of the mixed layer ages and their identification as seismites is achieved by correlation with the chronology of the historical earthquakes. The concept used here is to assess whether all individual pieces of information (the structures, the ages, and the correlation to known chronology) fit an internally consistent framework.

• from Ken-Tor et al. (2001a:2231-2232)

Several earthquakes from the Dead Sea area that are reported in the historical catalogues have no correlative deformed units in the Ze'elim record (Figures 3b and 5). The absence of these earthquakes from the Ze'elim record may be due to a remote epicenter and/or small magnitude (<5.5), which was not sufficient to induce a disturbance in the Dead Sea area. Another obvious reason is the incompleteness of the geological record in the Ze'elim Terrace. During times of climatic change, the lake level dropped and the Ze'elim Terrace was exposed, resulting in depositional hiatuses.

Two major unconformities are identified in the section (Figures 3b and 5). To constrain the ages of the hiatuses we determined three ranges of calibrated ages for each unconformity. The first range is obtained from the closest samples below an unconformity; the second range represents samples obviously related to the unconformity (samples recovered from sand and aragonite crust); the third range is of samples collected from layers above the unconformity. The older unconformity has yielded the respective ranges of 340-540 A.D., 1030-1210 A.D., and 1220-1390 A.D. (Table 1). The younger unconformity has yielded the ranges 1270-1400 A.D., 1290-1420 A.D., and 1520-1800 A.D. (Table 1). The age ranges presented in Figure 3 are the samples obviously related to the unconformities (i.e., the second age range).

Nine earthquakes reported in historical documentation from the Dead Sea area do not appear in the Ze'elim seismite record and they all fall in the unconformities and depositional hiatuses discovered in the section (Figure 3b). This is consistent with the interpretation of the mixed layers as earthquake- induced structures in the Ze'elim record. If a strong earthquake that was reported broadly in the Dead Sea area were missing from a complete sedimentary record, it would have been difficult to explain the interpretation of the mixed layers as seismites.

The 419, 551, 659/660, 749, 1033/1034, 1068, and 1160 A.D. dates are correlative to the first major unconformity, and the 1458 and 1546 A.D. earthquakes are correlative to the younger unconformity (descriptions of the earthquakes are provided in Table 3). It is impossible to determine from the Ze'elim record whether these events were generated in the Dead Sea area. Probably, a few of the events were too remote or of moderate magnitude. However, the 749 A.D. earthquake caused regional damage along the Dead Sea Transform, destroyed the Hisham Palace in Jericho, and was identified as a fault rupture on the Jericho Fault [Ambraseys et al., 1994; Amiran et al., 1994; Ben-Menahem, 1991; Guidoboni et al., 1994; Reches and Hoexter, 1981]. This earthquake probably did affect the Ze'elim area, but the evidence was lost. We predict that this event will be identified in cores taken from deeper parts of the lake, where erosion due to lowering of the lake level is less frequent.

• from Agnon et al. (2006:197-198)

The following terms have been used in the literature to describe various types of deformed unconsolidated sedimentary layers associated with earthquakes:

• mixed layers (Marco and Agnon, 1995; Marco et al., 1996b): this term may cause confusion with a number of unrelated uses in the earth sciences;


• mixtites (Jackson and Bates, 1997): this term describes any clastic layer regardless of composition or origin; any flood deposit may fall in this category;


• homogenites (Kastens and Cita, 1981): the term stresses uniform composition of the deposit, yet does not account for systematic variations across the layer;


• seismites (Seilacher, 1969): this term is interpretative; and


• intraclast breccias (Marco and Agnon, 2005): we use this descriptive term to separate observations from interpretations: "intraclast" refers to the origin of the clasts being reworked from within the sedimentary section (Jackson and Bates, 1997), and "breccia" refers to the texture of the deposit.

• from Agnon et al. (2006:198)

Intraclast breccias are distinctive in sequences that are otherwise well-bedded or, better yet, laminated. Lamination is typical in the lacustrine facies of the Dead Sea deposits and makes recognition of intraclast breccias practical because of the conspicuous alternation between chemically precipitated white aragonite and darker detritus (Fig. 2) (Katz et al., 1977).

In the present context, intraclast breccias within a laminated sequence can be distinguished by the following criteria:

1. The primary mineralogical composition of an intraclast breccia is identical to the underlying strata (and typically the overlying layers also). Thus, the intraclast breccia appears similar to the enclosing deposits. The absence of fine-scale lamination observed from a short distance helps to recognize the intraclast breccia in the field.
2. Fragments of laminae may vary in distributions of size. Tabular fragments of competent laminae (with the long dimension commonly 1-5 mm) float in a fine-grained matrix. Graded bedding is common, either fining or coarsening upward.
3. Intraclast breccia layers are typically several centimeters thick, but can be as thin as a few laminae (viewed under a microscope; cf. Fig. 2B).
4. The upper contact of an intraclast breccia is invariably sharp and is typically overlain by laminated beds.
5. Basal contacts can be gradual, but occasionally are sharp. In the former case, folded and torn packets of laminae are abundant (Figs. 2A, 2C).
6. The verified lateral extent of individual intraclast breccias is on the order of 100 m. Over lateral distances of several tens of meters, the layers vary little in thickness, except where they onlap local paleorelief that formed during earthquakes (Fig. 3).

Jones and Omoto (2000) suggest the following criteria for the identification of seismic triggering of soft sediment deformation:

1. geological setting,
2. extent of the deformed units,
3. absence of evidence indicating other potential trigger mechanisms, and
4. presence of evidence of other potential trigger mechanisms elsewhere in the stratigraphic section associated with undeformed sediment.

Intraclast breccia layers in the sediments studied satisfy all these criteria.

• from Agnon et al. (2006:198)

The Dead Sea intraclast breccias have many similarities to seismites described from the lacustrine environment as well as from glacial deposits and volcanic terrains. Breccia layers associated with microfaults and intraformational folds in glacial deposits were reported in Scotland (Davenport and Ringrose, 1987). Earthquake-induced soft sediment deformation in Late Pleistocene lacustrine beds associated with activity at the Narugo Volcano, Japan, has been reported by Jones and Omoto (2000), who suggest the above-mentioned criteria to identify seismic triggering agents.

Most of the paleoseismic studies focus on Pleistocene to Recent deposits, but seismites have been reported in significantly older rocks, including Silurian strata (Kahle, 2002) and laminated Neogene deposits in Spain (Rodriguez-Pascua et al., 2003). In the area of the Dead Sea Rift, intraclast breccia layers are also present in Senonian chert (Mishash Formation), which crops out near the Lisan Formation (Fig. 2B). Similar fabrics in the breccias in these formations prompted Kolodny et al. (2005) to suggest a similar mechanism of formation.

Interpreting the origin of ancient seismites often relies on intuition and on understanding models of the mechanism of their formation. Features interpreted as ancient seismites should resemble those formed by modern earthquake deformation. Soft sediment deformation associated with strong earthquakes is documented by observations of recent seismic events (Allen, 1974; Sims, 1973). Earthquakes have caused silting and resuspension of sediments in Canadian lakes in association with earthquakes and in turbid water observed in lakes <10 km from the epicenter of the 1935 Temiskaming, Canada, M 6.3 earthquake (Doig, 1990, 1991). Piston cores from the bottom of that lake recovered a 20-cm-thick chaotic layer composed of tabular fragments derived from a preexisting silt layer. Graded bedding has been suggested as a criterion for subaqueous liquefaction based on observations in Kobe, Japan, following the 1995 earthquake (Kitamura et al., 2002).

Earthquake-induced historical homogenites are reported in Lake Lucerne, Switzerland (Siegenthaler et al., 1987). Lake Le Bourget, France, has homogenites that correlate with the A.D. 1822 earthquake (local intensity VII-VIII), the strongest known historical earthquake of the French outer Alps (Chapron et al., 1999). Historical accounts of this earthquake report violent lake water oscillations, which were probably a seiche, and an earthquake-induced subaqueous slide may have formed the homogenite layer.

• from Agnon et al. (2006:198-201)

The formation of intraclast breccias involves five stages (Fig. 3). First, layered deposits at the lakebed (Fig. 3A) are disrupted and deformed by ground shaking, motion of the water column, and water escape from the underlying uncompacted sediment (Fig. 3B). During this stage, the pressure of pore fluids in the sediment exceeds the confining pressure of the overlying lake brine, resulting in liquefaction of the sediment. Subsequently, the top of the sedimentary succession becomes fluidized and suspended at the bottom of the water body; fault ruptures can create topographic steps at the lake bottom (Fig. 3C). Seismic waves can trigger mechanical instability in the sediment, expelling pore fluid into the overlying suspension (Hamiel, 1999; Heifetz et al., 2005). Long water waves that oscillate the entire lake (seiche) carry significant momentum at the bottom of the lake, keeping the sediment suspended. After the waves have dispersed and attenuated, an intraclast breccia is deposited from the suspension by grain settling and water escape (Fig. 3D). After settling, the intraclast breccia is capped by the continuing deposition of laminated sediments that gradually bury any fault-related topography (Fig. 3E).

The intraclast breccia's texture attests to the interplay between forces in the sediment; namely, the pressure of the pore fluids, the contact forces between solid particles, and gravity. The clasts were originally part of laminae, and rupture of the laminae is a precursor of a liquefied state, where pore pressure exceeds cohesive forces and drives cracks through the sediment. As long as the pore pressure exceeds the lake pressure head, fluids that escape the liquefied bed exert drag stress on particles in the top part of the layer. When this drag exceeds gravitational forces, particles are suspended and the sediment is fluidized.

The formation energy of seismites in general is supplied by seismic shaking, but gravitational energy contributes on slopes, where the disturbed bed slides downhill, expending potential energy. Gravitational energy can also contribute to the formation of seismites where the density profile of the undisturbed sediment is inverted (dense on top): overturning the sediment releases the gravitational potential for overcoming resistance. There is no evidence that gravitational energy was a factor in the formation of common Dead Sea intraclast breccias. The seismites were deposited on flat surfaces, and no evidence for density inversion was found. Some gravitational energy is involved when pore fluid is injected upward, but most of the formation energy is of seismic origin. Three agents of seismic energy for disruption can be considered: the shaking of the ground below, the motion of the water above, and the injection of pore water from below.

Fragments of laminae in upward-fining intraclast breccias indicate that the lake-bottom sediment was compacted and cohesive before the earthquake. During the event, laminae shattered, and fragments were suspended into the fluid. Liquefaction of sediments under earthquake shaking is well documented and is traditionally related to the passage of shear waves (e.g., Allen, 1982). Yet observations of liquefaction features from recent large earthquakes highlight the role of P-waves (Lin, 1997), and engineering design based on resistance to cyclic shear loading has occasionally failed (Hatanaka et al., 1997). Observations of intraclast breccias in the Dead Sea basin have stimulated new theoretical and experimental studies of liquefaction (Bachrach et al., 2001; Hamiel, 1999; Lioubashevski et al., 1996).

An alternative mechanism is the Kelvin-Helmholtz Instability (KHI) mechanism, in which stably stratified layers undergo a shear instability during relative sliding, which is set off by earthquake shaking (Heifetz et al., 2005). Analysis suggests a threshold for ground acceleration increasing with the thickness of the folded layers. The maximum thickness of folded layers, on the order of decimeters, corresponds to ground accelerations of up to 1 g. The application of the KHI model to earthquakes is based on a translation of the instrumentally measurable ground accelerations to pressure gradients. The KHI model is at a preliminary stage and does not provide precise correspondence between field observations and the actual driving ground accelerations. Moreover, it does not rule out alternative sources for pressure gradients, such as surface and internal waves in the depositing water body. Since water depth of Lake Lisan above the investigated area was several tens of meters (Bartov et al., 2002), ground acceleration waves might have dominated over water waves.

• from Agnon et al. (2006:201)

Several authors have cited criteria to distinguish between the seismic and nonseismic origin of soft sediment deformation features (for reviews, see Jones and Omoto, 2000; Obermeier, 1996). Marco and Agnon's (1995) studies of seismites were originally motivated by intraclast breccias juxtaposed to intraformational faults in the vicinity of Masada (Fig. 1), where a terrace capped by laminated sediments is present between the Dead Sea and the western fault escarpment (Agnon, 1983; Sagy et al., 2003). Similar exposures of fault zones juxtaposed to intraclast breccias are also present in the Lisan Peninsula. Analysis of the microstratigraphy at these sites shows simultaneity between two processes acting at the lake bottom; namely, faulting and homogenization of the lake bed (Marco and Agnon, 1995, 2005). The time interval between intraclast breccia formation and lake bottom faulting is shorter than the time it took to deposit a lamina, which is likely less than one year (see evidence below for varve-like lamination). This association is perhaps the strongest evidence for attributing datable sedimentary structures to earthquakes, hence naming them seismites. The geological observation that deformation occurred at the water-sediment interface makes intraclast breccias excellent markers to determine the times of past earthquakes, if the time of sedimentation can be determined.

• from Agnon et al. (2006:201)

The Masada fault zone has repeatedly ruptured the surface along several km (Fig. 1). In order to examine the subsurface continuity of the faults, Marco et al. (1996a) carried out ground penetrating radar (GPR) and high-resolution seismic reflection surveys to image the fault zone (Figs. 4-7).

The 275-m-long GPR survey shows fault planes extending several meters below the surface (Fig. 5). A parallel calibration profile runs at the top surface of the Lisan Formation above a buried fault whose location and dip are known from exposures (Fig. 4).

The 450-m-long seismic reflection line across the Masada fault zone overlaps the GPR profile but extends farther east and west (Fig. 6). In addition to conventional reflection data, we present diffraction data analyzed using the method proposed by Landa et al. (1987) and Kanasewich and Phadke (1988). The diffracted waves are sensitive to the discontinuities in beds due to faulting, providing an independent support for the interpretation of the reflection profile.

Three zones of discontinuous reflectors on the processed profile represent faults that extend from <0.05 s down to 0.25 s (two-way traveltime [TWTT]) at shot points 50-60, 80-95, and 125-135. The faults are also the sources of diffractions that are shown in the diffraction section from 0.05 s to 0.15 s (TWTT) at shot points ~60, 95, and 125-135 (Fig. 6).

The coincidence of the faults in the geological, radar, and seismic reflection sections shows that every outcropping fault can be traced down to ~250-300 m (Fig. 7). Eyal et al. (2002) reported similar results in another study of a fault zone in an alluvial fan. Several faults evident in the seismic section are not expressed in outcrops, which may be evidence for syndepositional faulting, or alternatively, faults that did not rupture the surface.

• from Agnon et al. (2006:201,203)

Based on correlations with historical earthquakes, Migowski et al. (2004) concluded that local intensity is a critical factor in the formation of intraclast breccia layers. We review this work below and define the conditions required to form intraclast breccias.

The Dead Sea region has not experienced strong earthquakes during the instrumentally recorded twentieth century (Fig. 1). Therefore, to assess the intensity required to brecciate laminated sediments, we must rely on historical records of earthquakes. The Dead Sea region has a long, rich record of historical earthquakes, which consists of information on shaking in settlements nearby (Ambraseys et al., 1994; Amiran et al., 1994; Guidoboni, 1994). Recent retreat of the Dead Sea shorelines has exposed sediments deposited in the past millennia, where the effects of known historical earthquakes can be seen in the sediments. Exposures revealing historical deposits are limited to where the lake level was higher than the level during the past decade. Even for these intervals, the sediments may lack fine bedding or may have been eroded during times of low water levels. Drill cores in lacustrine laminated facies overcome these limitations.

Another analysis, which considers both the thickness of the breccia beds and the lithology of beds directly overlying them, is applied in order to identify the stronger (M > 7) earthquakes within the record recovered from the Lisan Formation (Begin et al., 2005). The analysis is based on the occurrence of gypsum immediately above 11 breccia layers between 54 and 16 ka, a coincidence that is explained by the triggering of a strong seiche, which mixed the stratified waters of Lake Lisan. Mixing of the sulfate-rich upper water layer with the calcium-rich lower water layer could trigger the deposition of gypsum (Stein et al., 1997a). The resulting time series of earthquake recurrence interval is similar to the M ≥ 7.2 recurrence interval in the Dead Sea basin, as extrapolated from present seismicity; therefore, Begin et al. (2005) suggest that the present seismic regime in the Dead Sea basin, as reflected in its magnitude-frequency relationship, has been stationary for the past ~40 k.y.

• from Agnon et al. (2006:203-205)

Ken-Tor et al. (2001a) studied Holocene outcrops of Dead Sea sediments in Ze'elim fan, east of Masada (Fig. 1), where eight late Holocene seismites are exposed due to the accelerated recess of the Dead Sea during the past decades (Fig. 8). Six intraclast breccia layers (A-F) are identified in the lacustrine laminated facies. The uppermost part of the section exposes only the near-shore sandy facies, showing two liquefied sand units (G and H). Some intraclast breccias in the lower section grade laterally into liquefied beds showing flame structures as these beds change facies into beach sands. Twenty-four radiocarbon ages of plant debris are largely consistent with the stratigraphic order in the section sampled. Ken-Tor et al. (2001a) were able to fit a model of moderately varying deposition rates (3-9 mm/yr) between hiatuses by assuming that all intraclast breccias were formed during historically recorded earthquakes (Fig. 8). Significant uncertainty remained only regarding event D (~4 m above the base of the section, Fig. 8) that could correspond to either of two historical earthquakes: 363 or 419 A.D. Ken-Tor et al. (2001a) have considered both events (compare their Figures 3 and 4).

By combining radiocarbon dating with the precise dates of historical earthquakes and with field evidence suggesting subaerial exposures, Ken-Tor et al. (2001a) determined that two unconformities exist in the section between 400 (or 420) and 1200 A.D., and between 1300 and 1750 A.D. (Fig. 9). This interpretation is remarkable in that the eight major historical earthquakes that are missing from the Ze'elim fan section have all happened during the hiatuses thus dated (551, 559, 749, 1033, 1068, 1160, 1456, 1546 A.D.). This suggests that the Ze'elim earthquake record is complete for the periods of deposition, with the exception of either 363 or 419 A.D., for which only a single intraclast breccia is found. We return to this dilemma after reanalyzing the outcrop data, and again after presenting the results from drill cores.

Subsequently, Ken-Tor et al. (2001b) used the historical earthquakes to refine calibration of their radiocarbon dates and to infer that the time between death of the plant and burial in the sediment is 50 yr or less.

Figure 8 offers a revised correlation scheme between the Ze'elim stratigraphic record and the historical earthquake record that satisfies two conditions:

1. All model dates match event horizons with historical earthquakes.
2. A uniform sedimentation rate between successive event horizons and a slowly varying sedimentation rate between hiatuses. This allows interpolation between event horizons and determination of ages based on historical dates exceeding the precision of radiocarbon.

We used 24 calibrated radiocarbon ages (Bookman et al., 2004; Ken-Tor et al., 2001a) to guide the matching of the sediment height-age model. Nineteen ages are compatible with the model in that the higher bound on the age is older than the model age. For most cases, the model line goes through the calibrated age range. Five calibrated ages are younger than the model by several decades. Some incongruence can be resolved by considering low-probability ranges in the calibrated date distributions. Discrepancy may also result from sampling of roots debris that might have deteriorated in situ. Alternatively, our uniform deposition rate may be an oversimplification for this arid climate featuring irregular flash floods.

Our model indicates a strikingly uniform mean sedimentation rate during the three periods of continuous sedimentation: 5-6 mm/yr. Where data is ample, we note that the deposition rate may fluctuate by 50% around that mean rate: at the short period before the Christian era (between the events of 140 and 31 B.C.), we see an anomalously high rate of deposition (7 mm/yr). Subsequently, between 31 B.C. and 33 A.D., the rate declines to 3 mm/yr, maintaining an average of ~5 mm/yr.

This uniform deposition rate model results in breccia layer A in the section of Ken-Tor et al. (2001a) correlating with the 140 B.C. earthquake. The 64 B.C. earthquake, which was originally correlated with breccia layer A, cannot be distinguished from the 31 B.C. breccia layer. We prefer this correlation to the correlation of breccia layer A with the 64 B.C. earthquake because the latter correlation implies an excessive deposition rate of 24 mm/yr.

Our model correlates breccia layer D with the 419 A.D. earthquake, noting that this results in the 363 A.D. earthquake being uncorrelated. Our reconstruction agrees with lamina counting data (Migowski et al., 2004) reviewed in the next section.

• from Agnon et al. (2006:205-206)

Continuous cores from three sites along the Dead Sea shore were drilled during the fall of 1997 (Migowski et al., 2004), including cores from Ze'elim fan and Ein Gedi Spa (Fig. 1). The staggered-pair drilling technique recovered a continuous record of the subsurface sediments. We chose the Ze'elim fan site to correlate subsurface strata with the outcrops (Ken-Tor et al., 2001a, 2001b) and verify that both surface and drill core methods agree. The two other boreholes were drilled very close to the contemporary shoreline to avoid hiatuses due to lake level drops beyond the current level (Bookman et al., 2004) and to obtain data from very recent sediments. The 20-m-long Ein-Gedi core provides a continuous sedimentation record that spans the past 10,000 yr (Migowski et al., 2004). Above a 10 ka salt layer, the core contains two alternating principal facies: laminated fine-grained chalk (laminites) and bedded to massive silt. The laminated chalk contains aragonite laminae, resembling the lacustrine facies of Lisan Formation (Bartov et al., this volume).

The Ein Gedi core has penetrated 53 deformed intervals (Migowski et al., 2004, their Table 2). Many of the 53 deformed intervals in the core resemble the intraclast breccia beds of Marco and Agnon (1995). Migowski et al. (2004) focused on a 2.25-m-long interval of aragonite-rich laminites for a detailed inspection under an optical microscope. They counted couplets of detritus and chemically precipitated laminae as single depositional cycles. In some cycles, laminae of gypsum added to form triplets. They identified 1500 deposition cycles and suggested that a cycle represented one year of sedimentation. One way to test the annual depositional cycle hypothesis is to evaluate the lamina chronology with intraclast breccia events to see whether the time intervals match the historical record. Within that microscopically analyzed interval, Migowski et al. (2004) found 22 intraclast breccia layers and developed a chronological model for the sequence in which each cycle represents one year.

Migowski et al. (2004) constrained their chronological model to minimize the number of breccia beds for which no historical earthquake is known and found only one model that matched as many as 20 out of 22 breccia layers with historical earthquakes since ca. 150 B.C. (e.g., Fig. 10). They found that four additional earthquakes correspond to periods in which the record was destroyed by brecciation associated with subsequent earthquakes, postdating the missing events by several years (and, in a single case, by 33 yr). Figure 10 shows a unique situation where the contact between two breccia layers is preserved, so two earthquakes separated by 10 yr may be resolved.

The chosen matching, leaving out two subcentimeter breccia layers at 90 A.D. and 175 A.D., is significantly better than any other chronology model. The chosen model also minimizes the number of historical earthquakes for which no disturbance is shown in the sediment: six historical earthquakes from the entire region are missing. Clearly, some of these missing historical earthquakes were too distant or too weak to generate significant shaking in Ein Gedi.

While direct historic information on local shaking intensity at Ein Gedi is rare, empirical formulas can evaluate the historic data and estimate the magnitude of a given earthquake, as well as its location (Ambraseys, 1988). In a recent compilation of Middle East intensity data, Ambraseys and Jackson (1998) develop a formula that relates mean earthquake magnitudes to the intensity of shaking at given source distances:

M = -1.74 + 0.66I + 0.0015R + 2.26 log R,   (1)

where R is the distance from the earthquake focus (assumed at 7.4 km depth), I is the local intensity, and M is the magnitude. Formulas of this form, known as attenuation relations, estimate the local intensity at a site if both the earthquake magnitude and its location are known. A sufficiently close earthquake source with sufficiently large magnitude should generate intraclast breccia layers; sources that are too weak or too distant would not generate intraclast breccia layers. Figure 11 shows a compilation of all historical and instrumental earthquake sources in terms of magnitude and distance from Ein Gedi. The diagonal bold curves separate three domains: strong and close sources on the lower right, all matched with breccia layers; weak and far sources on the upper left, all unmatched; a median domain where about half of the events are matched by intraclast breccia layers. These lines can be fitted to equations of the form of equation (1). The straight line is given by:

M = 1.9 log R + 2.8.   (2)

The dotted and dashed lines in Figure 11 mark intensities V and VI, respectively. Note that for distances R > 50 km and magnitudes M > 5.5, we can consider the isoseismal I = V as a domain boundary. At lower distances and magnitudes, a higher intensity seems to be required for generating breccia layers.

As shown in Figure 11, all historical earthquakes with calculated local intensity at Ein Gedi I > V are matched with intraclast breccia. The earthquake of 363 A.D. may constitute an exception to that rule.

The unique match of the two independent records, namely the historical and the one derived from the core, supports three assumptions used to develop the chronological model:

1. breccia layers form by seismic shaking,
2. strong shaking results in breccia layers, and
3. the lamination is seasonal with a detectable annual cycle.

• from Agnon et al. (2006:212)

The systematic approach to accumulating observations on intraclast breccia layers permits their analysis as recorders of paleo-earthquakes. Such breccia layers, previously called "mixed layers," are abundant in sedimentary sections of Quaternary lakes from the Dead Sea basin. The finding of intraclast breccia layers juxtaposed against surface faults has driven a wide range of studies focused on these long-term seismic recorders. The fault scarps form a fault zone, traced to the subsurface by high-resolution geophysical surveys. This fault zone is a subsidiary structure to the master fault bounding the Dead Sea basin from the west and extending north to the transform plate boundary. While the record of displacement of the fault zone is limited to particular slip events that activate this secondary structure, the record of breccia layers may be complete for events that rupture the plate boundary.

We define field criteria for the identification of intraclast breccias, focusing on features that can indicate a seismic origin. The wealth of data for such earthquake indicators collected in natural outcrops within the Dead Sea basin offers insights into the phenomenology and systematics of earthquakes on time scales that are not obtainable elsewhere.

Cores from the receding shore of the Dead Sea contain continuous sedimentary records of the past 10 k.y., undisturbed by lowstands. The Ein Gedi core features 3 m of alternating seasonal laminae. The two independent earthquake records—historical and sedimentary—offer a simultaneous test regarding two hypotheses: the earthquake origin of breccia and the annual cycle of laminae. The likelihood of matching historical earthquakes with an arbitrary time series that corresponds to the breccia layers in the core is negligible. The observation that the breccia layers match earthquakes from the historical catalogues that are strong and/or close to the coring site supports both hypotheses (Fig. 11). Changes in the rate of recurrence of earthquakes in the Dead Sea record during the historical period seem to correlate with changes in the Anatolian Fault system. If the rates of recurrence could be taken as indicators of activity of the plate boundaries, then these plate boundaries might be coupled on the time scale of 500 yr.

The brecciation from an earthquake that succeeds another strong earthquake might obliterate the breccia layer of the predecessor. This hampers the potential that lies in the laminated sediment to resolve pairs of earthquakes. The resolution of an interseismic interval is no better than the ratio of the thickness of a breccia layer to the rate of deposition. But the resolution limit of individual earthquakes does not affect the observation of clustering in the record, which is evident in the long periods of quiescence alternating with periods of recurrence of earthquakes. During most of the period recorded, we find that the apparent recurrence interval is significantly longer than the resolution limit. During a cluster of earthquakes ca. 52 ka, the interseismic interval becomes shorter than 200 yr, which is close to the resolution limit for the Lisan outcrops.

• from Leroy et al. (2010:191-192)

The Dead Sea is the lowest water body in the world (e.g., Niemi and Ben-Avraham, 1997), 423 m bsl in 2009. This water body is the Holocene surviving trace of a much larger late Pleistocene lake, Lake Lisan (Bookman et al., 2006). Until 1931 (when a dam on the outlet of Lake Kinneret was constructed), lake levels were changing according to rainfall on the watershed (approximately 43,000 km²). Now with the added withdrawal of water for agriculture, the level drop is ca. 1 m per year (e.g. Dayan and Morin, 2006). Anati et al. (1995) have analysed the evolution of the DS water column stratification over the period of 1977-1995. Until 1978, the DS was mostly meromictic. However, from February 1979, it became holomictic with a December overturn and is only stratified and meromictic when there is a higher freshwater input due to either higher rainfall or a large release of water by the Degania dam (Gertman and Hecht, 2002). Moreover, a drastic change in the sedimentation in the DS took place in 1983 when the annual lamination was replaced by massive halite precipitation; this phenomenon therefore followed the first complete mixing by only a few years (Anati, 1993).

The average annual rainfall over the DS itself is ~90 mm, but most of the water entering the lake comes from a zone of relatively high rainfall in the upper Jordan River watershed, with an annual precipitation in excess of 600 mm (Dayan and Morin, 2006). Rain occurs between October and May, often with more rainfall at the transitional seasons (autumn and spring) and mostly in the form of rainstorms. The mean annual temperature east of Jerusalem (at Ma'ale Adumim, 330 m asl) is 19 °C, whereas near the DS's northern tip it rises to 23 °C (at Kalia, 60 m bsl) (Kutiel et al., 1995). Winds, a consequence of depressions during winter and spring, are mostly westerlies.

The main tributary to the DS is the Jordan River. A few perennial streams like Wadi el Mujib (Arnon), opposite Ein Gedi, discharge into the lake (El-Naqa 1993). Most of the wadis draining the hills are dry with the exception of winter floods from October to May. Springs such as Ein Feshkha, fed by aquifers from the Judean Mountains, emerge along the flanks of the DS (Ben-Itzhak and Gvirtzman, 2005).

• from Leroy et al. (2010:192)

The Dead Sea is surrounded by three altitudinal phytogeographic belts (Zohary and Orshansky, 1949; Rossignol, 1969). First, along the DS itself, the Saharo-Sindian (= Saharo-Arabic) vegetation (dry hamada dominated by Chenopodiaceae) reaches north up to 50 km from the DS, with rare enclaves of Sudano-Deccanian vegetation (with tropical elements such as Acacia) linked to freshwater springs. Higher up, Irano-Turanian vegetation is found on the slopes: a steppe with dwarf shrubs where Artemisia herba-alba then dominates. At higher elevations, the Mediterranean vegetation is developed to the west on the Judean hills (1020 m asl) and to the east on the Moab hills (1065 m). The summits of these hills are more than 1400 m higher than the present DS level. Where the vegetation climax is reached, the semi-steppe is then replaced by evergreen maquis with Quercus calliprinos (evergreen oak) and Juniperus. Neither oaks nor olive trees now grow in the DS trough.

Over the millennia, the cultivated surface has gone through periods of extension and contraction. During the Roman-Byzantine times (from 63 BC to AD 638), large farmsteads and intensive production for exportation of olive oil, dates, opobalsam (an unspecified tree growing in Gilead) and spices have developed in addition to production for local use (Harland, 2002).

The strongly contrasted rainfall pattern is reflected by the period of flowering of many plants (Zohary and Orshansky, 1949). The vegetation around the DS therefore produces a pollen rain, which has a strong seasonal character (Feinbrun et al., 1959; Kantor et al., 1966; Horowitz 1979; Al-Eisawi and Dajani, 1988). This influences the pollen assemblages found in the sediment's sublaminae (Leroy, 2010). The strongest signal, however, remains the pollen concentration, which is high in autumn-winter-spring (dark laminae) when most plants bloom and low in summer (light laminae) when most plants are suffering from lack of humidity (Leroy, 2010).

• from Leroy et al. (2010:192)

The Ze'elim sampling site is located on the southwestern shore of the present-day DS on a flat plain at an elevation of ca. 400 m bsl (Bookman et al., 2004). The Ze'elim River flows into the southern part of the DS north basin from the west and has a catchment area (250 km²) that is located mainly in the southern Judea desert. Presently, water flows through the canyon only several days a year, associated with storm events in the high-elevation Judea Desert (Magaritz et al., 1991). The Nahal Ze'elim delta is the largest of the active deltas located along the western margin of the depression. In terms of the morphometry of fan deltas worldwide, Nahal Ze'elim is relatively small (~6 km²), its slope is steep (2.5°) and its catchment is quite large for the size of the sediment lobe (Warren, 2006).

• from Leroy et al. (2010:192)

One of the best records so far for sea level changes over the last 2500 yr has been obtained from exposed sedimentary sequences along the western shores of the DS. The sea level change curve is based on identification of buried shoreline deposits within their sequences and their associated radiocarbon ages on organic debris (Enzel et al., 2003; Bookman et al., 2004, 2006). The Hellenistic and Byzantine high stand at ≥395 m bsl (ca. 250 BC to AD 500) is interrupted by a brief low level at ca. AD 270 down to 404 m bsl. The final end of this high stand, however, comes after the 5th century AD (Bookman et al., 2004, 2006).

Six pollen diagrams from sites along the west shore of the Dead Sea north and south basins have been compiled by Neumann et al. (2009b) in order to provide the history of vegetation change over the last 3500 yr. Reference will be made here especially to the deep-sea core taken offshore Ein Gedi (Heim et al., 1997; Leroy, 2010). Vegetation and agriculture changed roughly synchronously with lake levels, with the exception of the brief lake level drop around AD 270, which is not well recorded in the six pollen diagrams so far.

• from Leroy et al. (2010:192-193)

A large part of the Pleistocene sediment of the Lisan formation and most of the non-halite facies of the Holocene sediment of the Dead Sea are laminated. The usual, but not mandatory, succession of three facies in a year is the following: (1) clastics from river inflow in winter or runoff floods in the transition seasons (allochthonous origin), (2) gypsum and (3) aragonite, the last two by chemical precipitation (autochthonous origin) (Bentor and Vroman, 1960; Reid and Frostick, 1993; Migowski et al., 2004). More complicated laminations have been observed too: for example, light evaporitic ones with one or more thin detritic beds that could perhaps represent more than a year (Heim, 1998), or thick massive clastic units that may represent high frequency flashflood event(s) during relatively wet years.

The Dead Sea lies in a pull-apart basin within a continental transform plate boundary (Garfunkel, 1981; Niemi and Ben-Avraham, 1997). The distribution of earthquake epicentres suggests that the DS is one of the most active sections, north of Red Sea. Earthquakes caused by fault motion on the transform fault and associated structures have been felt throughout historical times. Several catalogues are available (e.g., Amiran et al., 1994; Guidoboni et al., 1994; Guidoboni et al., 2004) as well as surveys of earthquake-damaged structures (e.g., Reches and Hoexter, 1981; Ben-Menahem, 1991; Ellenblum et al., 1998; Zilbermann et al., 2005; Haynes et al., 2006; Thomas et al., 2007; Marco 2008).

Breccia layers composed of broken and mixed lacustrine seasonal laminae in the DS basin were interpreted as seismites (layers that exhibit earthquake-triggered deformation) (Marco and Agnon, 1995; Ken-Tor et al., 2001a; Agnon et al., 2006). The DS sediment in the Ze'elim gully is unique as it is extremely well dated owing to a detailed chronology and correlation of seismites to historical earthquakes (Fig. 2) (Ken-Tor et al., 2001a, 2001b). The organic debris used for dating comes from vegetation growing along streams and springs around the DS. Ken-Tor et al. (2001a) show a remarkable agreement between similar breccia layers, ages and historical earthquakes in the last 2000 yr in the Ze'elim fan delta. Two seismites from this record were investigated in detail; seismite B correlated to the 31 BC earthquake and seismite D correlated to the AD 363 earthquake. These seismites were chosen for the following reasons. Both can be traced along the outcrops continuously and were dated previously in high accuracy. They were correlated to seismic events that left significant information in historical documentation and occurred during periods with relatively high population compared to following centuries.

For the two Roman-Byzantine earthquakes, physical characteristics followed by societal consequences are now briefly summarised. The earthquake of 31 BC has been described by the contemporary historian, Josephus Flavius, in a very detailed account (Flavius, 1982). According to him, 30,000 persons were killed. There was severe damage in the DS area, Galilee and Judea. In Jerusalem, the earthquake damaged the second temple (see references in table 3 of Ken-Tor et al., 2001a). The earthquake caused confusion and fear. This paved the way for the expansion of the kingdom of Herod (Ben-Avraham et al., 2005). The earthquake was strong, with an estimated magnitude of 6.7 (Migowski et al., 2004, and reference therein) with an epicentre located on the main Jericho fault, north of the Dead Sea.

For the earthquake of AD 363, reports are available from Baniyas in the north through to the Red Sea in the south and from the coastal littoral through the Jordan Valley and beyond (Thomas et al., 2007). In Jerusalem, the Temple area was damaged. A seiche was reported in the DS (see references in table 3 of Ken-Tor et al. 2001a). More specifically for the earthquake of AD 363, Russell (1980) suggested that major demographic shifts occurred in the Jewish populations of Galilee. Ken-Tor et al. (2001a) estimated that this was a strong earthquake whose epicentre was, however, not close to Ze'elim. This earthquake was part of an "earthquake storm," which may have caused the destructions of the middle 4th century AD when a series of significant earthquakes during a 30-yr period between AD 350 and 380 resulted in much damage at sites in Israel, Cyprus, NW Turkey, Crete, Corinth, Reggio Calabria, Sicily and northern Libya (Nur and Cline, 2000).

• from Leroy et al. (2010:193)

Coseismic uplift or subsidence may alter vegetation formation, which in turn may be registered in pollen assemblages and concentrations within sediments (Mathewes and Clague, 1994; Mirecki, 1996). Exposure of new land has also been detected in pollen diagrams by the presence of pioneer plants (Cowan and McGlone, 1991). In the case of Lake Sapanca (NW Turkey), located on the North Anatolian Fault (Schwab et al., 2009; Leroy et al., 2009), palynology was used as a taphonomic indicator of changes in sedimentation type into the lake centre, especially sources (local vegetation versus altitudinal, shore versus deep lake) and accumulation rates (rapid inwash of pollen-depleted soils and river sediment following earthquakes). An assessment of the impact of earthquake activity on regional vegetation has been attempted in the Dead Sea's oasis of Ein Feshka (Neumann et al., 2009a). The investigation focused on blocks of sediment (including 1 to 3 yr of sedimentation according to a laminite count) taken in three Holocene seismites and above them. The conclusions indicate that after the seismite in the Roman period there was no fluctuation. For the seismite at the Hellenistic to Roman transition, a weak signal is found in a decrease of Olea but not in the other cultivated plants. Finally, in the Byzantine period, a possible weak signal was found but only perhaps aggravating a trend started before the earthquake.

• from Leroy et al. (2010:193-194)

In the dry eastern Mediterranean region, earthquakes may damage the infrastructure necessary for agriculture, such as irrigation structures, and therefore cause a more or less long-lasting change in the pollen indicators of such activity. In the Near East, archaeological records show that ancient earthquakes have affected farm animals, water resources and agricultural infrastructure. In 31 BC, "an earthquake shock killed an infinite number of cattle" (Flavius, 1982). After the AD 1202 earthquake, many sources mention abandonment of villages and deadly pestilence that affected livestock animals (Guidoboni et al., 1994). After the earthquake of AD 749, "the spring of water near Jericho was moved six miles from its place" (Michael the Syrian in Guidoboni et al., 1994). In the city of Sagalassos, Turkey, during the first half of the 6th century, an earthquake caused a lot of damage, amongst others destroying the aqueduct and causing severe water shortage to the town (Waelkens et al., 2000; Similox-Tohon et al., 2006). After reconstruction, a further earthquake in the 7th century may have been the cause of the final abandonment of the city. Water conduits in Iran, locally called "qanats," were also displaced by earthquakes, probably causing disruption in agricultural work (Ambraseys and Jackson, 1998). A 2000-yr-old aqueduct in Syria was found to have been repeatedly damaged by movements on the DS fault (Meghraoui et al., 2003). Haynes et al. (2006) described historical earthquake damage to a water reservoir and some aqueducts, one of them feeding an irrigation system in the archaeological site of Qasr Tilah in the Wadi Araba basin (Jordan).

Archaeological excavations at the site of Umm el Kanatir, east of the Sea of Galilee, revealed that the ancient village was abandoned after the AD 749 earthquake, which triggered a landslide that damaged most of the houses, the main water reservoir and the synagogue (Wechsler et al., 2008). The qanat system north of the Red Sea has been damaged by earthquakes on the Dead Sea Transform (Zilberman et al., 2005).

• from Leroy et al. (2010:194)

The Ze'elim canyon was already inhabited in the Neolithic period. Hirschfeld (2004) has proposed that the spring of Ein Aneva as known from hagiographical literature was located in the Ze'elim Wadi. Although this spring is now dry, it flowed abundantly in the Byzantine period. Byzantine terraces and a farmhouse indicate plentiful supplies of water. No signs of settlement were found dating of the period after the Byzantine one. In the oasis of Ein Gedi, a few kilometres to the north, irrigated agriculture with palm trees and barley was practised (Hadas, 2008). This collapsed in the middle of the 6th century.

• from Leroy et al. (2010:194)

During our fieldwork in the Ze'elim fan, samples were taken in gully A unit III of Bookman et al. (2004) (Fig. 2). The same sampling location, however, does not exist anymore because the walls of the gullies erode very rapidly with each flash flood (Ben Moshe et al., 2008) in response to the drop of lake level. Two seismite layers, B and D, dated respectively to 31 BC and AD 363, have been selected (Ken-Tor et al., 2001a) (Fig. 2). An alternative age has been suggested for event D in Ein Gedi, AD 419 (Agnon et al. 2006), which is unlikely because fewer earthquake damages are known and based on a less likely chronology. These two seismites have been subsampled at a quasi-seasonal resolution by hand in the field directly from the gully wall (Fig. 3). The subsampling focused on the dark layers, i.e. autumn to spring, avoiding the white aragonitic and orange oxidised laminae in order to maximise the pollen concentration and to obtain in each sample most of the annual pollen influx. A few samples were taken in the seismite itself, representing a mix of the few years before the earthquake. Then, following the seismites, 9 to 10 consecutive samples were selected, representing most likely 9 to 10 yr of sedimentation. The influx of pollen per ml per winter (roughly equivalent to year) is also in this present case taken as being identical to the annual pollen concentration.

• from Leroy et al. (2010:194)

The ages of the seismites were constrained by ¹⁴C dates on organic debris derived from the deformed unit itself (seismite B) or by extrapolating between ages directly beneath and above the deformed unit (seismite D) (Ken-Tor et al., 2001a). The ¹⁴C ages were calibrated to calendar years defined by the 2σ envelope error (Stuiver et al., 1998). The time elapsed between the deposition of the organic debris and the correlated historical earthquake was short relative to the uncertainty in the dating (Ken-Tor et al., 2001b).

• from Leroy et al. (2010:194-195)

The samples were treated at the University of Bangor with the following sequence: pyrophosphate, 10% HCl, 40% hot HF, 10% HCl and sieving at 125 and 10 µm. The concentration in number of pollen grains and spores per ml of wet sediment was obtained with the initial addition of Lycopodium spore tablets. The counts reach a base sum of at least 600 grains per sample for the calculation of reliable percentages for the less abundant taxa. Outside of the base sum are the rare aquatic elements, the spores and the varia (unknown and indeterminable grains). Pollen diagrams are plotted using the program Psimpoll 2.27 (Bennett 2007). The pollen zones (pz) of the selected taxa diagrams are based on lithology and on a CONISS analysis after square-root transformation. The detailed diagram of the two seismites is presented in the Supplementary material.

• from Leroy et al. (2010:195)

Samples for geochemical analysis were taken at exactly the same levels as for the pollen analysis. The geochemistry data (10 oxides and two trace elements) were obtained at Gloucester University with inductively coupled plasma atomic emission spectroscopy (ICP-AES). Dried powder (0.25 g) mixed intimately with 0.75 g of lithium metaborate was fused at 1050 °C for 30 min, then dissolved in 10% nitric acid and made up to volume to 250 ml with distilled water. The ICP-AES data were calibrated against 10 international reference rock materials. The magnetic susceptibility was obtained by measurement by mass on the Bartington MS2 magnetometre. The dry samples were measured with the MS2B Dual Frequency Sensor for discrete samples. For a limited number of samples just after seismites B and D, mineralogy was tested by X-ray. At the Geological Survey of Israel, a Philips X-ray diffractometer (17130/1710) was operated with 40 kV and 30 mA, CuKα radiation, 1°, 0.1 mm, 1° slits and sealed proportional detector. The APD software controlled the running, calculated and printed peak locations and relative intensities, and controlled the diffractogram printout. Bulk mineralogical composition was determined by the highest reflections of each mineral.

• from Leroy et al. (2010:195)

The field observation of the thickness and colour of lamination following seismites provide clues to the environmental changes caused by the earthquake events. After the 31 BC earthquake, alternating dark and light lamination did not return until 5 yr afterwards. Instead of alternating dark detrital and aragonitic laminae, we observe four grey clay-rich layers separated by thin dark layers; whereas after the AD 363 earthquake, a visually massive deposition of 1.7 cm occurs that could have been deposited over several years (Fig. 3). The mineralogy of the first cm above seismite D is dominated by calcite, whereas above seismite B it is co-dominated by calcite and quartz. The secondary minerals were respectively quartz and clays or clays only. Minor traces of dolomite and aragonite were found. This conforms with a river inflow- or runoff-flood type of deposition.

Whereas the number of years after seismite B is obtained by counting the number of black (or winter) layers, a field estimation for seismite D was taken as 3 yr, following on a common rule that an unclearly varved segment is equal to the average sedimentation rate of that above and below (Lotter, 1991). Moreover, for both seismites, a simple exercise has been done with the pollen concentration which should be roughly similar from year to year. If the massive deposition (divided in three adjoining samples of similar volumes in the field) is taken as formed in 1 yr (by adding the concentration values of the three samples together), then the influx of pollen becomes extremely high (Fig. 4, top). If it is taken as forming in increasingly more years, the influx decreases. Finally, the best fit (alignment of the values with those earlier and later) is 4 yr for seismite D, suggesting one more year than the number of samples. For seismite B, the best fit is with 5 yr, confirming these laminites as yearly formed (Fig. 4, bottom).

The stable geochemical and magnetic susceptibility data indicate very little change in soil in-wash at the sampled locality (Figs. 5 and 6). Two samples after the 31 BC seismite, however, contain a clear peak in both Sr and Ba. This could reflect a brief and late (after 6 and 7 yr) influence of hydrothermal activity. The magnetic susceptibility indicates that in both cases the earthquakes had no effect on the values measured. There are, however, distinctively lower overall values for the seismite D samples than for the seismite B ones (Figs. 5 and 6).

The lack of clear variation in geochemistry and in the magnetic susceptibility indicates that the changes of pollen taxa percentages are not due to a different taphonomy such as a dilution of the lacustrine sediment by additional soil, nor to an accretion by unusual flash floods. The laminites just after the earthquakes therefore were probably deposited under the same processes than before the earthquake. Hence, the high-resolution pollen diagrams may be interpreted in terms of changes in vegetation cover rather than taphonomic change (Figs. 5 and 6).

• from Leroy et al. (2010:195,198)

The pollen zonation closely follows the lithology in both seismites. In the 31 BC diagram (seismite B), the pre-earthquake and post-earthquake pollen assemblages (pz ZB1 and 3) show signs of intense agriculture such as Cerealia-t. pollen grains and arboriculture with pollen of Olea (olive) and curve of other potentially cultivated trees/bushes: Pistacia-Juglans-Vitis (pistachio, walnut and grape vine) (Fig. 5). However, the laminites following the earthquake (pz ZB2) show a significant decline of human activities and an increase of desert plants (Amarantaceae-Chenopodiaceae) during the 5-yr interval. The olive groves would have produced less for a while (less blooming, fewer pollen grains) because of a temporary lack of management (ploughing and weeding of inter-tree soil, fertilisation and pruning) (Makhzoumi, 1997; Terral, 2000), whereas many cereal fields were not planted or suffered from the lack of maintenance of irrigation canals.

In the AD 363 diagram (seismite D), the vegetation (pz ZD1 and 3) is already clearly less modified by human activities than for the time of the 31 BC earthquakes: cereal fields are still there but the olive groves have been partially replaced by a wooded steppe and an open woodland (evergreen forest with oaks and pistachios with steppic elements such as Sarcopoterium (thorny burnet)), showing already in the area at that time a decline of human activities or a shift to other activities in that century (Fig. 6). About 4 yr after the seismite (pz ZD2), a further temporary decline of human activities is nevertheless noticed: less Olea, less cereals and perhaps also less other cultivated trees/shrubs. The regional vegetation is affected with clearly more desertic/ruderal plants colonising the fallow lands.

• from Leroy et al. (2010:198)

Other pollen analyses elsewhere in lake sequence with seismites show characteristic soil inwash or river inflow leading to widely different pollen concentrations. In Lake Sapanca (Turkey), pollen assemblages are different in turbidites and reworked horizons caused by earthquakes and, even more clearly, show much lower pollen concentrations due to dilution of the lake sediment by pollen-barren soils (Leroy et al., 2009). In other settings, it has been demonstrated that lakes with river inflow receive more pollen grains (higher influx) than lakes without rivers (Pennington, 1979). Therefore, here the atypical sediment above the seismites is most likely not attributable to single flood event(s) with an instantaneous duration(s). In conclusion, the impact of these earthquakes is estimated to have lasted for about 4-5 yr.

• from Leroy et al. (2010:198)

The pollen spectra of the 31 BC (seismite B) and the AD 363 (seismite D) earthquakes (Figs. 5 and 6) indicate that the two seismites most likely belong respectively to the beginning of pz DS2 (high Olea, some Juglans and Vitis) and to the end of pz DS2 (presence of Sarcopoterium and Pistacia but still low Pinus (pine) and reworked pollen) in core DS7-1SC (Leroy, 2010). The concentration values in the two seismite diagrams are much higher than those of the same period in core DS7-1SC (pz DS2 with only 6000 to 18,000 grains per ml; Leroy, 2010) and in the Ze'elim outcrop ZA-2 (Neumann et al., 2007). This is explained by the origin of pollen grains in a near-shore site, which is closer to plants and more affected by river sediment. In the ZA-2 pollen diagram (Neumann et al., 2007), seismite B is towards the beginning of the zone of maximal olive cultivation related to the Roman period (LPAZ 5), and seismite D is at the beginning of the increase of Pinus related to the transition from the Byzantine to the early Islamic period (LPAZ 6). Bookman et al. (2004) placed seismite B clearly in a period of high lake levels, whereas seismite D is at it the end of one.

During the time of the two studied earthquakes, there were only isolated areas of agriculture along the shores of the Dead Sea. This was taking place in oases surrounded by desert. Hirschfeld (2004) described farmhouses, terraces and vegetable gardens in Ein Aneva and Ein Gedi. Archaeological digging in Ein Gedi showed barley growing (Hordeum vulgare) in irrigated fields but the clear absence of olive trees (Hadas, 2008). Moreover, it is likely that owing to the higher rainfall of that period, the boundaries between the Mediterranean vegetation, the steppe and the desert would have shifted southwards and downwards. More land was devoted to agriculture than nowadays and a larger portion of land with olive groves (Leroy, 2010).

The length of the impact on agriculture on local society implies that people had to move away temporarily and that agriculture was deeply disrupted both in the oases along the coast and higher on the slopes of the rift valley (the two main sources of pollen indicators of human activities). The lack of maintenance decreased the production of the Olea trees. The Olea biennial alternation of pollen production cannot be the cause as the observed changes are >2 yr (Fernandez-Mensaque et al., 1998). More decisively for the cereals, the disruption of irrigation and the absence of people would have stopped plantation in the fields. It is also not impossible that the earthquake caused a drop of the level of the water table and a modification (decrease?) of the spring flow. In 31 BC, Palestine was under the reign of King Herod. This was a relatively stable period; although in that year, Herod won a battle against the Nabateans but after a brief war only (about one year long). The 4th century AD Palestine knew a period of political stability (Hirschfeld, 2004). It was then still under Roman rule. Therefore, it is not unlikely that rural life was restored quickly, perhaps more easily than after the 31 BC earthquake.

The AD 363 earthquake was followed in the succeeding centuries by a period with an accumulation of damaging factors to the society. There was not only a progressive aridification of climate (falling DS levels) from AD 550 onwards, a radical change in modes of food production (agriculture to pastoralism) in the late 6th and early 7th centuries, deep social/religious changes, the Justinian plague in AD 541 and invasions into the Roman Empire, but also an earthquake storm causing devastation to a large part of the east Mediterranean world in the 4th to 6th centuries (Stiros, 2001; Hirschfeld, 2006). Under those conditions, past societies may have suffered from a "ratchet effect" of vulnerability. This occurs when each succeeding event reduces the resources that a group, or individual, has to resist and recover before the next environmental shock or stress (Ford et al., 2006). In extreme cases, it may lead to societal collapse (Leroy, 2006). Earthquakes must be seen as part of a larger ensemble of hazards that affected the population leading to collapse at or close to the transition between the Byzantine period and the Islamic period, a time of the dramatic decrease of agriculture in the DS basin and its surroundings (Hirschfeld, 2006; Neumann et al., 2009b; Leroy, 2010).

• from López-Merino et al. (2016:50-52)

The Dead Sea (Fig. 1A) is a closed, inland hypersaline lake (e.g. Niemi and Ben-Avraham, 1997) and the descendant of the larger Late Pleistocene Lake Lisan (e.g. Bookman et al., 2006; Stein, 2014). Its main fresh water tributaries are the Jordan River and flash-floods that flow west from the Jordanian Mountains and east from the Judean Mountains. Long-term fluctuations of the Dead Sea lake level are caused by rainfall fluctuations over the watershed (Enzel et al., 2003; Bookman (Ken-Tor) et al., 2004). Rain occurs between autumn and spring and it can be either spatially localised or widespread (Dayan and Sharon, 1980). The main synoptic conditions triggering rain and dust transport to the Dead Sea area are the east Mediterranean cyclones and the Red Sea trough. The former is responsible for most winter rain and dust transport, while the latter is more related to autumn and spring dust storms (Ganor and Foner, 1996; Dayan et al., 2007). Because the Dead Sea is located on the border between semiarid and arid climates, rainfall varies seasonally and annually and is often concentrated in intense showers that cause flash-flood events and erosion (Dayan and Morin, 2006; Greenbaum et al., 2006).

The Ze'elim Formation consists of the lacustrine Holocene deposits in the DSB (Fig. 1B). Extensive outcrops of the formation were described at the Ze'elim Plain, one of the largest fan deltas along the western margin of the Dead Sea (Fig. 1C). Outcrops of the Ze'elim Formation were used for palaeoclimate and palaeoseismicity reconstructions (Ken-Tor et al., 2001a, 2001b; Bookman (Ken-Tor) et al., 2004; Kagan et al., 2011). The reconstructed lake-level curve revealed two high-stand periods characterised by relatively continuous stable high water-levels of at least a few decades. These periods of high-stand enabled the formation of alternating detrital and aragonite laminae sequences. The two periods correspond to the Hellenistic-early Roman and the late 19th–early 20th centuries high-stands (Bookman (Ken-Tor) et al., 2004; Fig. 1D). In order to eliminate the long-term climatic influence on the pollen record and to have regularly laminated sediments, these two relatively stable high-stand periods were chosen to be the focus of this study.

• from López-Merino et al. (2016:52-53)

Although several aerobiological studies are available for the Israel-Jordan area, most of them do not provide data regarding the Dead Sea region. Thus, we focussed on pollen calendars prepared in areas under semiarid to arid conditions close to the Dead Sea. Surveys of allergenic airborne pollen data collected on the roof of Tsell Harim hotel (Ein Bokek) at the western coast of the southern Dead Sea and on the roof of the Megilot Regional Council building (Kalya) at the northern part of the Dead Sea (Fig. 1A; Waisel, unpublished report 1, unpublished report 2) were chosen for comparison purposes, as the Ze'elim Plain is located in between and shares semi-arid features with both. Additionally, a flowering calendar (Supplementary Table S1) was composed using the information contained in the Handbook of Wildflowers of Israel (Shmida and Darom, 2000, 2002), paying special attention to the desert and steppe flora, along with the more frequent Mediterranean types in the pollen data.

Reported Ein Bokek and Kalya air-borne data present low total pollen concentrations during the entire year although with a well-defined spring peak. A dominance of Amaranthaceae pollen during the whole year is obvious. Poaceae pollen is also present throughout the year, peaking in spring (March-June). Another characteristic pollen type with large abundance is Olea, recorded in April-May at Kalya and in April-May (-June) at Ein Bokek. Pinus peaks in March-April in both stations. Plantago, another spring bloomer, peaks in March at Kalya and in April at Ein Bokek. An interesting difference between Kalya and Ein Bokek is related to Artemisia, which is abundant in Kalya in October-November but almost absent in Ein Bokek, likely because Artemisia is a steppe plant and although both Ein Bokek and Kalya are located in the desert, the latter is closer to the steppe area.

Other interesting pollen types that give information on specific flowering periods are infrequent or absent in the aerobiology studies performed in stations close to the Dead Sea. However, they are common in the palaeopalynological studies on Dead Sea Holocene sediments (Baruch, 1993; Heim et al., 1997; Neumann et al., 2007, 2009, 2010; Leroy, 2010; Leroy et al., 2010; Litt et al., 2012; Langgut et al., 2014, 2015a). Brassicaceae and Asteraceae have been identified in March-April in Arad, a town located in the southern Judean Desert (Kantor et al., 1966). Additionally, Brassicaceae species start their flowering period in February in Jerusalem, while Asteraceae species extend their blossom up to May and a few species also bloom in autumn (September-December) (Feinbrun et al., 1959). In fact, the number of Asteraceae and Brassicaceae species is large in Israel, and although many bloom in spring, some of them also flower in autumn (Supplementary Table S1). Arboreal pollen types are representative of long-distance transport to the Dead Sea region. Nonetheless, they are equally informative. Quercus is recorded during spring, in March-April mainly (Keynan et al., 1989). Finally, Pinus blooming peaks during spring (March-April) as well (Feinbrun et al., 1959).

The use of flowering calendars and palynological analysis to resolve the season in which an event occurred, i.e. a historical earthquake at central Israel (Langgut et al., 2015b), has proven to be reliable in a similar Dead Sea climate context.

• from López-Merino et al. (2016:53-54)

The level of the modern Dead Sea is the result of a large human induced retreat linked to the damming of its northern tributary (Sea of Galilee) and water pumping for industrial purposes. This retreat has triggered the formation of several meter deep gullies in the Ze'elim Plain (Ben Moshe et al., 2008). Late Holocene laminated sequences are exposed in these gullies (Fig. 1B). The two aforementioned high-stand periods were identified in the radiocarbon-dated outcrop (Bookman (Ken-Tor) et al., 2004; Fig. 1D) in February 2011. One block of laminated sediments from the late 19th–early 20th centuries high-stand (ZA11B2), and three blocks of laminated sediments from the Hellenistic-early Roman high-stand (ZA11B3R, ZA11B4L, ZA11B5L) were collected (Fig. 1D). Sampling of individual, mm-scale laminae was done at the sedimentology laboratory of the University of Haifa using a scalpel. Only consecutive detrital and aragonite laminae with no suspicion of contamination or mixing with the upper and lower layers were considered. Overall, 65 detrital-aragonite couplets (130 samples) were analysed: 13 in block ZA11B2 (couplets 1 to 13), 11 in block ZA11B3R (couplets 14 to 24), 17 in block ZA11B4L (couplets 25 to 41) and 24 in block ZA11B5L (couplets 42 to 65) (Supplementary Table S2).



Additionally, fine mud carried in suspension in flash-floods that occurred between February 2009 and June 2012, was collected from drying puddles immediately after the events. A total of 28 flash-flood samples were analysed (Fig. 1C; Supplementary Table S3).

• from López-Merino et al. (2016:54)

Grain-size analysis of the Ze'elim detrital laminae was performed at the University of Haifa using a Beckman-Coulter LS 230 laser particle size analyser over the particle size range of 0.02–2000 μm. Five replicate samples were measured for each detrital lamina. Grain-size was determined after dissolution of carbonate minerals. In block ZA11B4L and in a few samples of block ZA11B2 it was not determined due to low sample mass available (Supplementary Table S2). Additionally, for two thicker detrital laminae, grain-size distribution was measured in sub-samples from the top and bottom parts in order to identify settling patterns.

• from López-Merino et al. (2016:54-55)

Palynological extraction was carried out at Brunel University London. Around 0.5–4 ml (usually ~2 ml) of detrital material (detrital laminae and flash-flood samples) and 0.5–12 ml (usually ~5 ml) of aragonite laminae were deflocculated with a Na₄O₇P₂ solution (10%). Carbonate dissolution was done with concentrated HCl (35%). Elimination of silicates was obtained by HF (48%) followed by HCl. The residue was sieved through 125 and 10 μm nylon meshes and mounted on slides with glycerol. Concentration estimates (number of palynomorphs/ml of sediment) were calculated based on the addition of Lycopodium tablets at the beginning of the chemical treatment (Stockmarr, 1971). Palynological identification and counting was completed on light microscopes at ×400, and at ×1000 using immersion oil for more delicate identifications, and complemented by the Brunel University London pollen reference collection and atlases (Reille, 1992, 1995, 1998). Where possible, a minimum of 200 well-preserved (air-borne), non-reworked, pollen grains were counted per sample (average = 242). Palynological diagrams were elaborated using Psimpoll 4.27 (Bennett, 2009).

As pollen found in southern Dead Sea sediments could be transported by both wind and water run-off (Horowitz et al., 1975; Baruch, 1993), it is necessary to be sure about the air-borne component to be able to detect the timing of laminae deposition. Hence, reworked pollen grains (water-borne) were counted separately from well-preserved (air-borne) pollen grains and classified into different categories: broken, corroded, crumpled, degraded and other reworked (when two or more features were present) (Supplementary Table S1). Broken stamina were grouped with the reworked material. On average, 618 reworked pollen grains were counted per sample. Fungal spores had a significant presence and thus were also quantified. On average, 195 fungal spores were counted per sample. Full counts of reworked pollen grains and fungal spores are given in Supplementary Figs. S2–S6.

The concentration of Total well-preserved pollen grains, Total reworked pollen grains and Total fungal spores were calculated separately. The first includes the well-preserved pollen grains only (number of well-preserved pollen grains per ml of sediment), the second one includes the reworked pollen grains only (number of reworked pollen grains per ml of sediment), and the third includes the fungal spores only (number of fungal spores per ml of sediment).

• from López-Merino et al. (2016:55-56)

In order to unravel the seasons in which the laminae could have been deposited, Principal Components Analysis (PCA) was applied to the well-preserved palynological dataset. This approach has been confirmed as valid for detecting seasonality in palaeoenvironmental studies (i.e. Festi et al., 2015). In this case, the PCA has been performed on the transposed well-preserved pollen data matrices (samples in columns as variables, and taxa in rows as cases). This approach enables us to summarise the main palynological assemblages of co-existing taxa and their importance in each sample. Thus, the palynological composition of the samples can be compared based on co-variation patterns (Lopez-Merino et al., 2012). Taxa showing large factor scores (i.e. larger abundances) in a given principal component explain most of the variation of the palynological dataset in samples with large factor loadings. This enables quantitative expression of the proportion of variance explained by each principal component for each sample (Lopez-Merino et al., 2012).

Baruch (1993) presented the palynological analysis of 35 surface samples distributed along four vegetation zones of Israel: Mediterranean, transitional and steppe, desert, and the Dead Sea shore. The Mediterranean zone was dominated by arboreal taxa (Quercus calliprinos, Pinus and Sarcopoterium spinosum), the transitional zone was characterised by Sarcopoterium spinosum, Brassicaceae and Artemisia, while the desert area was dominated by Brassicaceae and Asteraceae. In the pollen rain in the Dead Sea shore, Amaranthaceae pollen was overrepresented. Therefore, in order to reduce background noise and detect the main characteristic palynological associations, Amaranthaceae pollen was excluded from the analysis. Many species belonging to this pollen type bloom all year long and very locally. Therefore, its pollen is overrepresented and dominates the pollen assemblages regardless of the season (Baruch, 1993; Waisel, unpublished report 1, unpublished report 2). After the exclusion of Amaranthaceae and prior to the statistical analysis, data-set proportions were recalculated. Correlation matrices and varimax rotation solutions were applied to constrain the co-variation in the components. PCA was done using the IBM SPSS Statistics 20 software.

• from López-Merino et al. (2016:56-57)

The siliciclastic fraction of the detrital laminae showed at least three different grain-size modes. Most samples had two grain-size modes, a first fine mode of 3–10 μm and a second coarser mode of 20–90 μm (Fig. 2). A ~10 μm mode was obtained on samples from collectors installed on a buoy on the Dead Sea surface during three years (1997–1999) (Singer et al., 2003). This grain-size distribution represents dust deposited over the Dead Sea similar to that of long-range transported Harmattan dust (Stahr et al., 1994) and is comparable to the grain-size distribution of air-borne dust in the Sde Boker area (Negev) (Offer et al., 1992). Thus, the fine grain-size mode in the detrital laminae siliciclastic fraction is consistent with fine dust largely wind-transported from medium to long range. Although a bimodal distribution is in accordance with Haliva-Cohen et al. (2012), the average grain-size measured for the second mode in this study is coarser, possibly due to the carbonate dissolution stage. Few samples included a minor third mode of 100–200 μm (Fig. 2) consistent with the contribution of loess (Haliva-Cohen et al., 2012), likely representing a local source.



The general grain-size distribution measured from detrital laminae suggests it is composed of recycling of local and distal sources as demonstrated in previous studies (e.g. Belmaker et al., 2014). Additionally, grain-size distribution within thicker detrital laminae shows graded bedding that suggests sediment deposition during flash-flood events. This phenomenon was shown with upward fining of the grain-size distribution in specific laminae (Fig. 2D) and was also described in petrographic thin-slides (Haliva-Cohen et al., 2012). The understanding that the detrital laminae consist of both air-borne particles and recycled local (mostly Quaternary sequences) material leads to the conclusion that the palynological analysis should separate air-borne from reworked pollen. This approach is practiced in the Dead Sea sediments at a very fine detail here for the first time.

• from López-Merino et al. (2016:60-63)

If Dead Sea laminae were varves in the way assumed by some authors, pollen spectra in aragonite laminae would correspond always to summer bloomers, while in detrital layers would have to correspond to autumn-winter-spring pollen assemblages. However, our palynological analysis shows that the air-borne component presents a more complex picture for the timing of aragonite deposition. Furthermore, the detailed palynological examination of the deposition season of detrital-aragonite couplets suggests three possible scenarios (Fig. 10):

• Scenario (i): both detrital and the following aragonite lamina have autumn palynological assemblages (27 out of 65 couplets, 41.5%),


• Scenario (ii): both detrital and the following aragonite lamina have spring palynological assemblages (9 out of 65 couplets, 13.8%), and


• Scenario (iii): the detrital lamina has an autumn palynological assemblage and the following aragonite lamina a spring one (21 out of 65 couplets, 32.3%).

These scenarios show no summer pollen assemblages in the aragonite laminae. One may argue that the lack of summer pollen results from the relatively low dust flux during this season (Singer et al., 2003). However, considering the rapid deposition of aragonite laminae along with the abundance of autumn pollen assemblages this argument does not stand. Thus, we conclude that the deposition of aragonite did not depend exclusively on evaporation and warming of surface waters (Neev and Emery, 1967). This indicates that aragonite precipitation occurred always during the rainy season, either at the same season of the detrital laminae deposition (scenarios i and ii) or during the following spring (scenario iii). These scenarios suggest that the deposition of aragonite laminae requires higher carbonate alkalinity in the flash-flood, hence enabling aragonite saturation without evaporation (Barkan et al., 2001).

Several of the consecutive couplets analysed showed a deposition within the same season (Fig. 10). On the one hand, consecutive autumn pollen assemblages were observed in the following couplets of the late 19th-early 20th century high-stand (block ZA11B2): couplets 13, 12 and 11; couplet 9 and the detrital layer of couplet 8; couplets 6, 5 and 4 and the detrital layer of couplet 3; and couplets 2 and 1. In the Hellenistic-early Roman high-stand consecutive autumn pollen assemblages were observed in a) couplets 19, 18 and 17 in block ZA11B3R; b) couplets 39 and 38; couplet 36 and the detrital layer of couplet 35; and couplets 30 and 29 and the detrital layer of couplet 28 in block ZA11B4L; and c) couplet 50 and the detrital layer of couplet 49; and couplet 43 and the detrital layer of couplet 42 in block ZA11B5L (Fig. 10). On the other hand, consecutive spring pollen assemblages were observed in the aragonite layer of couplet 35 and couplet 34 in block ZA11B4L; and in the aragonite layer of couplet 48 and couplet 47 in block ZA11B5L (Fig. 10).

These short deposition sequences do not allow us to determine whether they were deposited during the same year (i.e. a year with more flash-flood events) or during consecutive years in the same season. The compilation of the annual distribution of flash-floods in five Dead Sea stations (Supplementary Fig. S7) shows the occurrence of flash-flood events from October to May mainly, with few events in June and September. Thus, the modern flash-floods record shows that both situations are possible. However, more than one event a season is highly probable in the large drainages (Supplementary Fig. S7). In fact, the palynological assemblages obtained in this study for both detrital and aragonite laminae are those of the rainy season, i.e. when flash-floods are more likely to occur.

Scenario (iii), when a detrital lamina is deposited in autumn and the following aragonite one in spring, agrees with an aragonite precipitation due to mixing of flood freshwater with Dead Sea brine during the rainy season. However, in contrast to scenarios (i) and (ii), the laminae from the same couplet are deposited in different seasons within the rainy period. This scenario may also indicate a longer lag period between the freshwater arrival to the Dead Sea and the chemical deposition. The lacustrine conditions that can lead to this situation are only estimated at this stage. However, the need for sufficient accumulation of dissolved bicarbonate in the lake surface water or the effect of turbulence on chemical deposition is hypothesised as possibly critical in this interpretation.

The analyses performed in this study present for the first time direct evidence for the timing of laminae deposition in the DSB. The abovementioned scenarios suggest that both detrital and aragonite laminae were deposited in varying seasonal patterns rather than in a strict annual cycle. Detrital layers present grain-size distributions composed of sediment recycling of rock and sediment outcrops in the Dead Sea watershed and distal sources by aeolian transport. These recycled sediments reach the Dead Sea via run-off or as direct dust deposition from the atmosphere. Run-off sediment transport as the main process is confirmed by the large presence of reworked pollen grains and fungal spores (water-borne component) in the detrital laminae compared to the aragonite laminae (Fig. 12). Erosion of soils delivered reworked pollen and fungal spores with the sediments that deposited within the detrital laminae. Aragonite precipitates from the upper surface water layer due to mixing of flood-water with Dead Sea brine (Stein et al., 1997; Barkan et al., 2001).

This study indicates that, in contrast to previous assumptions, detrital-aragonite couplets were deposited during the rainy season as a result of flash-flood events, hence Dead Sea laminated sediments cannot be considered as varves. This result is of great importance for the accurate use of the DSB laminated sequences as palaeoenvironmental and paleoclimate archives (e.g. the ICDP Dead Sea Deep Drilling Project), and it delivers exciting new information to the Dead Sea scientists and the palaeoclimate community alike.

• from López-Merino et al. (2016:63-64)

During a flood event, water entering the Dead Sea is dispersed in a plume that floats over the dense saline waters (Nehorai et al., 2013). The suspended sediment in the plume ultimately deposits on the lake bottom. Plume sediment dispersal is a function of flood discharge, wave energy and distance from the shore. Consequently, the central deep part of the basin will record fewer flash-flood events as compared to the margin. In addition, the potential problem of reworking of deposited sediment destroying depositional laminations has to be taken into account. This may prevent, with the uncertainty of having more than one couplet deposited per year, the use of the detrital-aragonite couplets for laminated-based chronologies, i.e. if the laminated-based chronology of a studied sediment core is of long duration the number of couplets per time-unit may vary due to variations in the shore distance or/and flash-flood discharge. In addition, larger discharge carrying more sediment will deposit thicker and coarser detrital laminae. On the other hand, the higher the lake level the further the deposition location from the shore. This yields thinner and finer grain-size laminae. In brief, both parameters are climatically controlled, but result in a reverse effect on the deposition character of the detrital laminae, preventing the use of the thickness and grain-size of the laminae as a direct climatic indicator.

Barkan et al. (2001) pointed out that the aragonite precipitation rate calculated by Stein et al. (1997) for the laminated sediments of the Lisan Formation required a run-off influx six-fold higher than the contemporary run-off influx. Thus, indicating that the Lisan period was more humid than nowadays and that flash-flood events could have been more pronounced. Stein et al. (1997), however, based on previous reports (Neev and Emery, 1967; Begin et al., 1974), interpreted the aragonite formation as precipitation due to increased evaporation in summer. Therefore, understanding the deposition processes of the laminated sediments is of considerable importance for palaeoenvironmental reconstructions. This work has provided the first palynological evidence for unravelling the timing of aragonite laminae deposition, which has been identified as occurring during the rainy season rather than summer, but further research is necessary.

• from Kagan et al. (2011:3)

[10] The sediments of the Holocene Dead Sea comprise the Ze’elim Fm. of the Dead Sea Group. The sediments represent various depositional environments: fluvial, fan deltas, shores, and lacustrine (see detailed description given by Bookman (Ken‐Tor) et al. [2004]). The current (2009) lake level is 422 m below sea level (mbsl), reflecting mainly anthropogenic diversion of freshwater inflow, while during the Holocene the natural (climate related) level varied from ∼370 mbsl (e.g., ∼6000 years ago) to lower than ∼430 mbsl around 8000 years ago [Bookman (Ken‐Tor) et al., 2004; Bookman et al., 2006; Migowski et al., 2006; Bartov et al., 2007]. The drop of the current lake level (12 m from 1980 to 2000 [Bookman (Ken‐Tor) et al., 2004]) has caused the formation of deep gullies along the retreating shores. These gullies provided an excellent opportunity to study the late Holocene sedimentary sections in detail. The paleoseismic data in the current study was derived from the outcrops of the Ze’elim Gully and the Ein Feshkha Gully. Another site used for comparison, the Ein Gedi core site, was studied by Migowski et al. [2004].

[11] The study site of Ein Feshkha Gully (at the Ein Feshkha Nature Reserve) is located at the northwestern shore of the Dead Sea, 60 km north of the Ze’elim Gully (Figure 1). Ein Feshkha is an oasis of brackish streams and pools. Nearly exclusively lacustrine sediments are exposed in the Ein Feshkha site by a ∼6.5 m deep gully (as of 2008). The site is close to the Jordan Valley segment of the DST and may be located on the WNW continuation of the Kalia transverse fault mentioned in section 2.

[12] The Ze’elim gullies (site ZA) are dissected into the Ze’elim Plain east of the ancient fortress of Massada (Figure 1). Currently (as of 2009) the ZA Gully is ∼11 m deep (approximately at lake level, to slightly above). The gully exposes a stratigraphic sequence of lacustrine, shore environment, and fluvial sediments. The ZA site is closer to the Arava segment of the DST, about 50 km away, than to the Jericho fault. The active eastern normal boundary fault of the DST is at a similar distance to all sites, less than 5 km away (Figure 1).

• from Kagan et al. (2011:3-4)

[13] El‐Isa and Mustafa [1986] used intraformational folds to generate an earthquake archive. Subsequent studies presented more complete archives recognizing that folds might present the weakest events. “Mixed layers” were renamed “intraclast breccia layers” [Agnon et al., 2006].

[14] Agnon et al. [2006] define field criteria for the recognition of intraclast breccias, focusing on features diagnostic of a seismic origin. The field criteria reflect the mechanisms of breccia formation, which include ground acceleration, shearing, liquefaction, water escape, fluidization, and resuspension of the originally laminated mud.

[15] In the current study we recognize deformed structures such as intraclast breccias, liquefied sands, folded laminae, and small faults (centimeter scale) (Table 1). Figure 2 displays photographs and photo tracings of seismites from the study sites. In addition we recognize another type of deformation termed microbreccia or homogenite. This type of mid‐gray‐color sedimentary layer ranges in thickness from a few mm to 1–2 cm and appears homogenous in the field. Thin‐section investigation under a polarizing microscope shows that these are actually brecciated laminae, and include a mixture of detritus, aragonite, and in places gypsum fragments.

[16] In the more fluvial Ze’elim section there are instances of seismites with a combination of lacustrine breccia and sand liquefaction. For example (see Figure 2c), ZA seismite III is the product of the deformation of a lower sandy layer and an upper laminated marl layer, resulting in brecciated marl laminae (near top of Figure 2c) with injection of sand fingers (near bottom of Figure 2c) from below.

• from Kagan et al. (2011:5)

[17] Fieldwork included the detailed description and sampling of subvertical to vertical outcrops in the Ein Feshkha and Ze’elim gullies. Columnar sections were prepared with emphasis on measurement and description of the deformations. Adjacent outcrops were examined in order to describe spatial variations in lithology and seismites. Sediment blocks (∼10 × 10 × 10 cm in size) were collected for further analysis in the lab. At Ein Feshkha, 58 continuous blocks of wet sediment were retrieved from the gully wall at the columnar section site, from the surface plain down to 40 cm below the autumn 2005 water level of the spring outflow (see our previous paper, Neumann et al. [2007]). At the Ze’elim Gully sediment blocks were retrieved from the various lithological units. In addition, organic debris (typically short‐lived leaves or twigs), found in the two outcrops, were sampled for radiocarbon dating.

• from Kagan et al. (2011:7-9)

[20] The chronologies of the Ein Feshkha and Ze’elim sections were constructed by radiocarbon dating of terrestrial organic debris (mainly small pieces of wood and twigs). All the recovered wood in Dead Sea sections can be considered driftwood, however their transport time is relatively short. We made an effort, where possible, to date only short‐lived organic debris. Nine samples from EFE and twelve samples from ZA2 were prepared for radiocarbon dating at the Radiocarbon laboratory, Weizmann Institute, Rehovot, Israel. The samples were then measured by accelerator mass spectrometry (AMS) or liquid scintillation counting (LSC) at the NSF radiocarbon facility in Arizona. Eight additional organic debris samples from EFE were taken from a core drilled a mere few meters away, on the cliff bounding the gully, and analyzed at the AMS facility in Kiel. The core was correlated with the outcrop by Marcus Schwab at GFZ‐Potsdam. Radiocarbon ages are reported (Table 2) in conventional radiocarbon years (BP = before present; present defined as 1950 A.D.) in accordance with international convention [Stuiver and Polach, 1977]. Calibrated ages (=cal BP) were calculated by applying the INTCAL04 calibration scheme of Reimer et al. [2004] by means of the OxCal v4.1 program of Bronk Ramsey [1995, 2001, 2008]. Age‐depth models (Figures 3 and 4) and seismite model ages (see Table 3) were also created with OxCal (v4.1) [Bronk Ramsey, 1995, 2001, 2008].

[21] Radiocarbon data are listed in Table 2. Table 2 presents the measured ages, calibrated ages, and deposition model ages applying the Bayesian statistics of the OxCal v4.1 program. The depositional model ages were used to establish an age‐depth chronological model for the seismites. The fundamental assumption in Bayesian modeling of stratigraphic sequences is that age increases with depth. This requires use of a function usually termed “Boundaries” in OxCal. The boundaries separate different sedimentary units that may have different sedimentation rates, grain sizes, and facies. They are also placed on the top and bottom of the entire series to constrain the model to a specific time interval. With no other information, this would be treated by what is usually termed the “Sequence” model by OxCal. A uniform sedimentation rate would be treated with the “U_Sequence” type model. Depth and other dating information can be included in a less rigid way using Poisson distribution priors, termed “P_Sequence” models, where the time gap between deposition of grains varies, and the events are basically random but deposition is given approximate proportionality to depth. This requires the estimation of the uniformity of the deposition (given as the k parameter), which signifies the increment size (conceptually the grain size, or size of deposition events) and indicates the relation between the events and the stratigraphic process [Bronk Ramsey, 2008].

[22] In this study a P_sequence (Poisson distribution) Bayesian depositional model was used, with a k factor value of 1 (see Bronk Ramsey [2008] and Kagan et al. [2010] for details of Bayesian factors used). In the work of Kagan et al. [2010] the main objective was to test the Bayesian model with and without historical earthquake anchor points. The conclusion of the work was that the “known‐earthquake‐anchors” do not significantly improve the age model. For that reason, and due to the complexity in choosing definite historical anchors, in this study no anchors are used and the models are based solely on radiocarbon data, stratigraphic data, and the P_sequence and k factor constraints discussed in this section and by Kagan et al. [2010].

• from Kagan et al. (2011:9-10)

[23] For the EFE section the chronological model is based on the treatment of seventeen radiocarbon ages of which five were excluded as outliers (Table 2). In the last 2500 years, the period with historic earthquake correlations and implications, there were only two outliers, both of which were too old and probably represent long‐lived organic debris from the shores. One of these two outlier samples also appeared in the work of Neumann et al. [2007] (169 cm depth) and was considered an outlier. One interval, from 230 to 390 cm, is slightly anomalous: the sediment is much darker than the rest of the section and has less aragonite layers. Within this interval, between 230 and 330 cm, we did not recognize any deformed layers (Table 3). No organic debris was found from 220 to 410 cm depth (Table 2 and Figure 3).

[24] Several different models were run: (1) No internal boundaries from 0 to 500 m depth; 500 cm to base modeled separately. (2) Two internal boundaries in the 0–537 cm interval, at 230 and at 500 cm depth, which allow, but do not force, the model to have sediment rate changes. (3) The 0–230 cm and 390–500 cm deep segments run separately. (4) Various other options with different boundaries and various outliers.

[25] We choose option 2 from the above list (Figure 3). This curve yields the best “agreement indexes” for the Bayesian model, with one index value under 60% (at 17%) while the other models have lower agreement indexes. Alternative models give seismite ages (from at least the 5th century B.C. and on) that are very similar to the chosen model and do not change the paleoseismic conclusions (for an example, see option 1 in Figure S1).1 The slight facies change at 230 cm depth is allowed a degree of freedom to coincide with sedimentation rate change, but in the resulting model shows no significant rate change (see Figure 3).

[26] The chronology of the top 537 cm of the section is Bayesian‐modeled as one space with two internal boundaries at 230 cm and 500 cm, implying continuous sedimentation and allowing, but not forcing, sedimentation rate change at these boundaries. Agreement values are found to be well above 60% at most depths of the model. The resulting model ages of the section indicate a maximum range of 1261 B.C. to 1383 A.D., but more likely from ∼1100 B.C. to 1312 A.D. (Table 2 and Figure 3). The top unit, from 0 cm (surface) to 500 cm, shows ages that range from the 5th century B.C. to the 14th century A.D., with a 0.27 ± 0.03 cm/yr sedimentation rate (based on 2s age ranges). The age range of the lower unit (500 to 537 cm) is from approximately 11th–5th century B.C. (0.07 ± 0.03 cm/yr sedimentation). The base of the seismite‐bearing investigated section is at 590 cm, however in the bottom 53 cm no organic matter was found and therefore the age was not modeled. The sedimentation rate of the top 500 cm calculated here (0.27 cm/yr) is approximately constant, in comparison to that stated for the same section by Neumann et al. [2007] (0.14, 0.51, and 0.11 cm/yr for three stratigraphic units within the same depth interval). The rates presented here, based on the new Bayesian model, are more similar to published Holocene rates (e.g., Migowski et al. [2004]: ∼0.15 cm/yr for the entire Holocene Ein Gedi core) and more congruous with homogeneous pollen concentrations [Neumann et al., 2007], which are normally closely linked to sedimentation rate [Horowitz, 1992].

[27] The truncation of the last six centuries from the studied EFE section eliminates recording the key instrumental earthquake M6.2, 7 November 1927, the source zone of which spans the site (Figure 1). Macroseismic evidence for the 1927 A.D. instrumentally recorded earthquake was reported along the Jordan River [Hough and Avni, 2011]. Niemi and Ben‐Avraham [1994] interpreted large submarine slumps in the northern Dead Sea basin to have been caused by this earthquake. For the purpose of the discussion (section 5), this event will be considered recorded in the northern Dead Sea basin.

• from Kagan et al. (2011:10)

[28] Twelve organic debris samples from the 10.7 m deep Ze’elim (ZA2) outcrop were measured. Their calibrated radiocarbon ages range from 1056 to 1276 A.D. to 4843–4583 B.C. (95% probability). A deposition model is calculated for the top 8 m of this section. Model ages of samples are given in Table 2. The more western ZA1 section (∼100 m away) was dated by Ken‐Tor et al. [2001a, 2001b] and revised by Agnon et al. [2006]. In the Ze’elim Gully previous studies infer the sedimentation rate to range between 0.28 to up to ∼1.3 cm/yr [Ken‐Tor et al., 2001a; Agnon et al., 2006; Neumann et al., 2007] reflecting the additional detrital‐clastic sediments that are more abundant in the fan delta environment. The lower sedimentation rate (0.3 ± 0.03 cm/yr) at the ZA2 section of Ze’elim (current study) reflects the proximity of this section to the lacustrine environment. The ZA2 outcrop is interpreted to show continuous deposition according to the age‐depth model (Figure 4), as opposed to the numerous unconformities at the more landward ZA1 outcrop. However, at ZA2 there is the possibility of short hiatuses compensated by additional sediments at sandy facies which therefore are not manifested in the age‐depth diagram.

• from Kagan et al. (2011:10-11)

[29] Ages of seismites (Table 3) are interpolated from the radiocarbon age‐depth data using Bayesian stratigraphic constraints. The ages and their uncertainties are interpolated using the OxCal program and take into consideration the asymmetrical probability distribution of radiocarbon ages. Each seismite is assigned a probability distribution histogram with a 68% (∼1s) and 95% (∼2s) probability age range (Figures 3 and 4). Model ages are presented (Table 3) with a nominal precision of a single year, however due to the Bayesian statistical modeling each model run produces slightly different age ranges and therefore ages could be rounded off by 10 years. Although the annual dates are shown, they are dealt with as if rounded off; for example, when giving the historical “fit” in Table 3, the age ranges are considered in decades.

[30] Seismite ages have been compared to historical catalogs as a major component in the assessment of the validity of the interpretation of the breccia layers as seismites [e.g., Ken‐Tor et al., 2001a; Migowski et al., 2004]. At the same time, seismites can be used for the corroboration of individual earthquakes in the historical record. Ken‐Tor et al. [2001a, 2001b] used the radiocarbon ages of the individual breccia layers or liquefied sands for direct comparison with the historical records and noted that notorious historic earthquakes unrepresented in the geological record lie within sedimentary hiatuses in the western Ze’elim Gully section (termed here ZA1). Migowski et al. [2004] positively identified these “missing” earthquakes in the continuous lacustrine section of the Ein Gedi core, supporting the hiatus‐hypothesis. Moreover, by counting the laminae in the intervals between seismites they were able to correlate almost the entire historical and Ein Gedi core records.

[31] Table A1 presents the historic earthquakes in the region with information regarding damage, casualties, sources of historical data, and, in the footnotes, selected archeological and paleoseismic data. Table A1 is based on earthquake catalogs, whereas the information in the catalogs is derived from historical sources. Table A1 is reliable mostly during the past two millennia (from the Roman period and onward), but less information is available for the time interval 750–1100 A.D. (when the Muslim empire center moved from Damascus to Baghdad). The historical accounts in the pre‐Christian era are rare and if they do exist tend to be vague [Karcz, 2004]. A mid‐8th century event and its paleoseismological and historical implications are discussed in detail in Appendix C. Local source moderate earthquakes are probably missing in the historical catalogs. For seismite ages where only very distant correlative historic earthquakes exist, we propose small local source events as possible sources of seismite genesis (marked LS on Table 3). A map of historical locations is given in Appendix B (Figure B1).

[32] For the past two millennia we correlated almost all of the seismites in the Ze’elim and Ein Feshkha records to historic earthquakes (details in Tables 3 and 4). All historic earthquake dates that correspond to the 95% probability range of each seismite age are given in Table 3 (right column). Those that correspond to the 68% probability range are in bold.

[33] The protocol for assigning a particular historic earthquake to a seismite in the sedimentary section is the following: (1) We consider all known earthquakes within a time segment of the age‐depth model pertaining to the seismite depth (segment = within 1–2s uncertainty of the radiocarbon model age); this step is given in column titled “all possible events” in Table 3. (2) Among the earthquakes within this time segment, we select the one that is most consistent with age‐depth models of Figures 3 and 4 (preserving the sedimentation rate); see the correlation in Figure 5. We also considered the local intensity for the earthquakes estimated for the study area when deliberating, in certain cases, between the various earthquakes.

[34] Table 4 and Figure 5 present the results of the correlation of the paleoseismic evidence (Table 3) with the historical record (Appendix A) and the comparison of these results from four sections: EFE, EG, ZA1, and ZA2. In Figure 5 the historical dates of seismites are superimposed on the age‐depth models to display the matching of the two models; the deposition model and the historical correlation model.

[35] There are two possible sources of errors in a comparison between two archives, such as the historic earthquakes and the radiocarbon dated seismites. As noted in Table 3 the uncertainty in the age‐depth model is variable but typically less than 100 years (2s). This reflects the errors derived from the age‐depth Bayesian model. The uncertainty in the “historic ages” of specific seismites reflects the spread of all historic earthquakes that lie within the 2s model age range of the specific seismite depth (the right‐hand column of Table 3). Thus, the errors on the Bayesian curve are the reasonable estimate of errors in the historical ages–seismite comparison. In other words, we say that the maximum error in our comparison is less than 100 years and, as Table 3 shows, typically lower than 50 years.

[36] A special case is the couplet of earthquakes at 1202 and 1212 A.D. that, with the typical temporal resolution in Dead Sea sediments, are not resolvable. We chose to present them as a pair of events as 1202/1212 A.D. The seismite at 28 cm depth at EFE has a 1s model age of 1199–1240 A.D. Both the 1202 and 1212 A.D. events are large M >7 earthquakes that ruptured far from the Dead Sea (north of the Sea of Galilee to Lebanon, minimum 130 km [Marco et al., 2005] and south of the Arava, minimum 250 km, respectively). Agnon et al. [2006] show two adjacent seismites at this time in the EG core record and interpret these to represent both the 1202 and 1212 A.D. events. Both are candidates for this EFE seismite.

[37] The Ein Gedi core was dated by 20 radiocarbon ages and by laminae‐counting of ∼1500 years, from 200 B.C. to 1300 A.D. [Migowski, 2001; Migowski et al., 2004]. The laminae‐counted floating chronology of the seismites was matched with the historic earthquake catalog. The best‐fit history of Migowski et al. [2004] gave ages younger than their radiocarbon ages by 50–200 years, consistent with reworking of organic debris (e.g., wood) in the nearshore environment before settling to the bottom of the dense saline lake. In our analysis, the chronologies of the Ze’elim and Ein Feshkha section indicate no long reworking time of the organic debris before settling in the sediment. When referring to the seismites from the Ein Gedi (EG) core only we use the shifted laminae‐counted chronology of Migowski et al. [2004] for the EG section.

[38] Note in Table 3 that the type B seismites, “homogenites,” clearly correlate with important historic earthquakes, which supports their interpretation as seismites.

[39] The recording of earthquakes by seismites, as well as by historical documents, requires intensity above respective thresholds. In this study our data suggest that these two thresholds are similar. Quiescence intervals are more robust than specific earthquakes, because they are less sensitive to individual date correlation. Specifically there is a quiescence interval in the seismite archive of the three sites from the end of the 2nd to the beginning of the 4th century A.D. (Figure 7). This correlates to an historical earthquake quiescence period noted without a single historically documented earthquake in the region from 127 to 306 A.D. (Appendix A).

• from Kagan et al. (2011:11-13)

[40] Here we summarize the occurrence of seismites at the three sites presented in section 5.1 and in Table 4: Ein Feshkha (EFE), this study; Ein Gedi (EG core; after Migowski et al. [2004]); and the two Ze’elim Gully subsites: ZA1 (western, landward; after Ken‐Tor et al. [2001a] and Agnon et al. [2006]) and ZA2 (eastern, lakeward; this study), considered henceforth as one site. We limit the comparison to the historical period starting at the 2nd century B.C.

[41] 1. Seismites that appear in all three sites (termed here intrabasin seismites (IBS)): Mid‐2nd century and 31 B.C. and 33, 419, 551, 749, 1202/1212, 1293, and 1927 A.D.

[42] 2. Seismites that appear only in Ein Gedi: 76, 90, 112, 500/502, 1042, 1546, 1588, 1656, 1712, 1759, and 1822 A.D.

[43] 3. Seismites that appear only in Ein Feshkha: 64 B.C., 349, 363, 634, 847, 859, 956, 1063, 1170, and 1312 A.D., and numerous older prehistoric seismites.

[44] 4. Seismites that appear in Ein Gedi and Ein Feshkha but not in Ze’elim: 92 B.C. and 660, 757, 873, 991*, 1033*, 1114/1117*, and 1068* A.D. Stars indicate dates at which time there is no archive at Ze’elim.

[45] 5. There is one quiescence interval at ZA and EG ∼500–150 B.C. and another at all three sites from the end of the 2nd to the middle of the 4th century A.D.

[46] The new chronologies of the seismites in the Ze’elim (ZA) and Ein Feshkha (EFE) sedimentary sections are integrated with the high‐resolution seismite chronology of the Ein Gedi (EG) core to produce a comprehensive archive of late Holocene paleoseismic earthquakes from the entire Dead Sea basin. The paleoseismic archives also provide an opportunity to reevaluate a number of earthquake histories with timing and patterns that were not well established (e.g., single or several episodes).

• from Kagan et al. (2011:13-16)

[47] The chronologies that were established for the Ein Feshkha and Ze’elim sections combined with that of the Ein Gedi core [Migowski et al., 2004] allow us to compare the recurrence time of the seismites in these sites and to produce an integrated picture for the appearance of seismites in the northern Dead Sea basin (Table 4 and Figure 6). The number of seismites in the Ze’elim Gully sections is significantly smaller than at Ein Feshkha and Ein Gedi for the same time interval. Ken‐Tor et al. [2001a] and Agnon et al. [2006] recognized that the missing seismites at ZA1 (explained in section 1) relate to sedimentary hiatuses in the section. The new section we described at ZA2 yielded an apparently continuous age‐depth profile, and the hiatuses in the ZA1 section can be correlated with clastic‐sandy sequences in the ZA2 section. One of the missing (sedimentary hiatus) earthquakes (in the landward ZA1 section) does appear in the continuous ZA2 section as liquefaction in a sandy unit (correlative to the historical earthquake of 551 A.D.). In two instances the situation is reversed, where two seismites, correlated to 1293 and 1212 A.D. appear in the more landward ZA1 outcrop, and do not appear in the more lakeward ZA2 section. This specific period is characterized by a sandy facies at ZA2 (Figure 4) and detailed detection is also inhibited by difficult access at this part of the section.

[48] The EFE section has 52 seismites, while for the same time period the EG section shows ∼30 seismites. A quiescence period at EFE at around mid‐1st to 3rd century A.D. is concurrent to a period in EG with microscopic seismites (Type III of Migowski et al. [2004]). This could reflect the higher detection resolution of the Ein Gedi study. Despite this resolution difference, the situation is reversed in the pre‐2nd century B.C. period where EFE has 25 seismites (<1 cm to >9 cm) compared to 7 at EG. The recurrence of earthquakes in each one of these sections is illustrated as a cumulative function in Figure 7.

[49] A quiescence at ZA and EG during a period of enhanced seismicity in the north (EFE) at ∼500–150 B.C. (Figure 7) may suggest a period of moderate earthquakes concentrated north of the Dead Sea (i.e., Kalia fault). Additionally, there is a quiescence interval in the seismite archive of the three sites from the end of the 2nd to the middle of the 4th century A.D., which correlates to an historical earthquake quiescence period 127–306 A.D. (Appendix A). This is in line with the low‐seismicity interval during this period along the DST, the high‐seismicity period on the North Anatolian Fault, and the mechanical coupling and alternation of activity of the two faults suggested by Migowski et al. [2004] and Agnon et al. [2006].

[50] The comparison of EFE versus both EG and ZA clearly suggests higher activity in EFE. This can be explained by a difference in sensitivity between the sites, or the proximity of EFE to the Kalia transverse fault bounding the Dead Sea basin from the north (Figure 1). The EFE site is located on the continuation of this fault to the WNW, and has likely recorded local earthquakes of magnitude ∼5.5 that were too far to affect EG and ZA. Also, several seismites (during the time interval of the historical charts) were recorded only at the northern site of Ein Feshkha (e.g., 64 B.C. and 349, 363, 634, 847, 859, 956, 1063, 1170, and 1312 A.D.). Most of these events have destruction documented mainly in the northern Holy Land or further north (Antioch, Tyre, Turkey; see Appendix A), 1312 A.D. being the main exception. Since the work of Russell [1980], the 363 A.D. earthquake is often considered as one that ruptured from the north to the Arava. We suggest that this interpretation congeals two earthquakes, one northern and another southern (see Appendix A). The lack of documentation of earthquakes in the south can reflect bias due to population density, the south being more arid. However, the excess of recorded earthquakes at Ein Feshkha may corroborate higher seismic activity in the north. First let us consider the local setting of the Ein Feshkha Nature Reserve site: it is positioned at the edge of both the Jericho fault and the Kalia transverse fault (Figure 1). Ze’elim Gully, on the other hand, is several tens of kilometers from both Jericho and Arava faults, the likely sources of M > 6.5 events. Therefore, earthquakes rupturing the northern part of the Jericho segment will record at Ein Feshkha but not at the southern sites. Likewise, magnitudes 5.5–6 from the Kalia fault may be recorded locally but not at the southern sites.

[51] Our sites are located on the western shore of the lake, close to the western strand of the transform duplex. This observation may suggest an alternative explanation to the excess of earthquakes in the northern site EFE: The site is close to the Jordan (aka Jericho) fault that might act as a waveguide, a property documented for the plate boundary south and north of the Dead Sea [Haberland et al., 2003; Shtivelman et al., 2005]. Guided earthquake waves have been invoked to explain anomalous accelerations and damage in instrumentally recorded Dead Sea events [Wust‐Bloch, 2002]. The seismite sites in the south (EG, ZA) are farther from the Jordan fault, and disconnected from the Araba/Arava fault. This explanation can be tested by a similar research on the eastern shore: it would predict that the southern sites there will show more frequent events.

• from Kagan et al. (2011:16-19)

[52] In this section we discuss the temporal distribution of seismites that are recorded at all of our sites (intrabasin seismites (IBS)). Eight seismic events are recorded in all three sections, north, center and south. In addition we add to this list the 1927 A.D. instrumentally recorded event that formed seismites at the EG and ZA1 sites for which macroseismic evidence was found along the Jordan River [Avni, 1999] and caused slumping under the Dead Sea waters (interpreted by Niemi and Ben‐Avraham [1994]) near the EFE site. The 1927 A.D. event also produced the most pronounced sedimentary structure (in the ZA Gully) with sand liquefaction reaching >1 m in thickness (Figure 8). In addition, two seismites that were recovered from the Ze’elim and Ein Gedi sections and correlated to the 1458 and 1834/1837 A.D. historical events are not represented in Ein Feshkha since this part of the section is missing. However, we predict that processing of the upper part of the section preserved east of our EFE study site will recover these events. Note that the age of the seismite at ZA2 correlated to 1458 A.D. is above the dated and modeled part of the section and its age is extrapolated from the deposition model (see Figures 4 and 5). Part of this group of IBS seismites (mid‐2nd century and 31 B.C. and 33, 419, 1212, and 1293 A.D.) appears in sedimentary sequences of the lacustrine facies indicating clearly offshore conditions of at least 10–20 m of water above the sediment. Other IBS seismites (551, 749, and 1927 A.D.) were at nearshore conditions (hiatus at ZA1, sand and lacustrine sediments at ZA2, lacustrine sediments at EFE and EG). Thus, we see no clear correlation between lacustrine conditions and the three‐site seismite appearance. This observation is corroborated by the lack of seismites in intervals of the lacustrine section at Ze’elim while they appear in Ein Feshkha and Ein Gedi (e.g., between 830 and 1200 A.D.; see Table 4). The conclusion that we can draw from these observations is that the temporal and spatial appearance of the seismites does not depend strongly on the limnological‐sedimentological conditions. Seismites appear in both sandy facies and clay‐evaporite (marly) sequences. The Ze’elim sections contain prominent sand layers that were clearly affected by earthquakes, producing liquefied structures. Significant earthquakes, such as 1927, do appear in all lithological units. This does not imply that low‐magnitude or remote earthquakes have no effect on sandy layers. The topic clearly requires more investigation. If sediments were deposited in the lake they are affected by the earthquakes whether they comprise sands or marls. Figure 10 indicates that soil liquefaction and lacustrine brecciation have apparently similar thresholds. Hence, the archives we documented provide a reasonable picture of the earthquake activity and its effects in the lake basin, not filtered by the lacustrine environment. This conclusion opens the way for using the seismite spatial and temporal distribution to evaluate basin effects and recurrence patterns.

[53] All seismites in the Dead Sea basin are marked on an epicentral distance versus magnitude diagram along with the field of instrumental earthquake data (Figure 9). This diagram highlights domains of intensity, which is a function of magnitude and distance of epicenter from the recording site. In Figure 9 the intensity lines are plotted according to the equation proposed by Ambraseys and Jackson [1998] (here termed A&J):

M_s = 1.54 + 0.65 I_i + 0.0029 R_i + 2.14 log R_i + 0.32p     (1)

where

R_i = (r_i² + 9.7²)^0.5

r_i, in kilometers, is the mean isoseismal radius of intensity I, and p is zero for mean values and one for 84 percentile values. This attenuation relationship is based on 123 instrumentally recorded shallow (depth <26 km) earthquakes from the eastern Mediterranean from a period of 85 years and ∼9000 intensity points. Different coefficients may be more appropriate for the magnitude‐distance field of the earthquakes associated specifically with Dead Sea Rift seismicity. The earthquakes plotted are mainly after the similar diagram by Migowski et al. [2004] and Agnon et al. [2006], where modified input data are explained below and in Appendix A. Each symbol represents a historical earthquake documented in the region, most matched to seismites (open squares), and some matched to seismites at all three sites in the study, the intrabasin seismites (IBS, solid squares). Distances are from the Ein Gedi site, for consistency with previous publications. A field corresponding to earthquakes not matched to seismites is demarcated by the thick gray curves (solid gray curve: earthquakes from historical catalogs; dashed gray curve: instrumentally recorded events). The magnitude‐distance data for each historic earthquake has significant uncertainties (for examples, see Figure 10); however this type of diagram has been shown to be useful [Migowski et al., 2004] for portraying a pattern in the presence of a large sample, barring any systematic bias. Figure 10 depicts only the IBS with estimated uncertainties. Each earthquake shows as a rectangle. We were especially cautious when estimating the upper left corner for each IBS rectangle. This corner, minimum magnitude and maximum distance, corresponds to the minimum intensity at the seismite site, which may be a threshold for intrabasin seismites. The considerations we applied when defining the IBS positions and uncertainties in Figure 10 are given here:

[54] Mid‐2nd century B.C.: Guidoboni et al. [1994] cite one event or more recorded at Antioch (for a summary of historic earthquakes in the region, see Appendix A; for locations of historical cities and towns, see Appendix B). The only traceable historical record for an earthquake comes from the cultural and political center at Antioch, where buildings were reportedly damaged, and Sbeinati et al. [2005] assign local intensity I = VII. For comparison, the 1202 A.D. event was only felt in Antioch, no damage reported [Ambraseys and Melville, 1988; Ambraseys, 2009; Guidoboni and Comastri, 2005]. Therefore if the magnitude of the mid‐2nd century B.C. event is smaller than M7.5 assigned for 1202 A.D., then the source was closer to Antioch and farther from the Dead Sea. Hence for the mid‐2nd century B.C. event we assign an uncertainty rectangle constrained by a bottom left corner coinciding with the 1202 A.D. position. The rectangle represents a range of local intensities spanning V–VII at Antioch, where the distance is calculated to the closer end of the respective rupture (consistent with the magnitude) along the DST. For this specific earthquake we cannot, at present, constrain the top left corner.

[55] 31 B.C.: The magnitudes of 31 B.C. and 749 A.D. are set at 7.2 assuming similarity in rupture length, both reported to have ruptured 110‐km‐long Jordan Valley segment [Reches and Hoexter, 1981]; the sites of damage attributed to the 31 B.C. event demarcate that segment. Ambraseys [2009] points out that 3.5 m dip‐slip displacement reported by Reches and Hoexter [1981] would correspond to an earthquake too large comparing with the historical reports. However, the displacement is measured locally on unconsolidated sediments. Reches and Hoexter [1981] explicitly avoid rejecting the possibility that a part of the slip occurred during several centuries following the event. Moreover they are aware of local complications in the strike of the fault that amplify dip‐slip. Hence we tentatively adopt the identification of the surface rupture with the 31 B.C. event. Gardosh et al. [1990] reevaluated the trench data in light of a newer geomorphic surface faulting study in the Dead Sea area. They conclude that slip accumulation reaches 1.2 m for two events in the past 2000 years on the trench strand. The uncertainty range of the magnitude of this event (Figure 10) is projected from a minimum given by Karcz [2004] and a maximum given by the rupture length discussed here.

[56] 419 A.D.: Damage from this event was reported for Jerusalem and “many cities and towns” and “all great cities” (sources in the works of Russell [1985] and Guidoboni et al. [1994]). Archaeological damage from Antipatris (central Holy Land) has been attributed to this earthquake [Karcz and Kafri, 1978] suggesting a Jordan Valley rupture. We think that it is feasible that the source of this event was similar to that of 1927 A.D. earthquake (see below). We assume a 6 ≤ M ≤ 6.5, with a maximum distance of 50 km.

[57] 551 A.D.: The event was updated to a larger magnitude offshore Lebanon earthquake, as is more widely accepted in the literature (Appendix A). Magnitude estimation is based on sonar imaging of seafloor morphology [Elias et al., 2007] and historical account compilation [Sbeinati et al., 2005].

[58] 749 A.D.: The historical sources are consistent with a rupture event or two in the Jordan Valley (between the Dead Sea and Sea of Galilee). The range of magnitude (M6.6–7.7) in Figure 10 reflects either a single event or a double event with a cumulative rupture of that 110‐km‐long segment (calculated using the results of Wells and Coppersmith [1994], Marco et al. [2003], and Karcz [2004]; see Appendix A).

[59] 1202/1212 A.D.: A single event brecciated the sediments in the EFE section in the early 13th century. Two events are recorded in EG. ZA recorded one or two events. Therefore only one of them is an IBS and the dating cannot rule which. For the 1202 event we use M7.4–7.6 based on historical analysis of Ambraseys and Melville [1988] and Ellenblum et al. [1998]. Paleoseismic and archaeoseismic trenching corroborate these assessments [Ellenblum et al., 1998; Marco et al., 2005; Daeron et al., 2007; Nemer et al., 2008]. The distance of the rupture edge from the farthest seismite site is 165 ± 10 km, based on rupture uncovered in trenching at the northern shore of the Sea of Galilee [Marco et al., 2005]. For the 1212 event, Ambraseys et al. [1994] suggest a rupture south of the Dead Sea or in the Gulf of Eilat (Red Sea). In severity of damage and aftershock occurrence it is seemingly similar to the 1995 modern event [Hofstetter, 2003], or could have been closer to the Dead Sea, according to the high level of damage at Aila and Karak. This similarity prompts us to give a best estimate of 7.2 magnitude and 300 km distance.

[60] 1293 A.D.: Based on evidence at an archeological site built on the Arava segment of the DST, the northern Arava did not rupture during this event [Haynes et al., 2006]. We consider the 12‐km‐long Amatzyahu fault (Figure 1) as the source for this event. This rupture length is consistent with a 6.2–6.7 magnitude earthquake. The maximum intensity recorded for this event was recorded at Karak (eastern Dead Sea), 45 km from the Amatzaya fault [Ambraseys et al., 1994; Guidoboni and Comastri, 2005], consistent with a magnitude of 6.7 according to the A&J equation (equation (1)). Taking into account poor construction and site effects this intensity could be achieved at a somewhat lower magnitude.

[61] 1927 A.D.: This event was recorded instrumentally [Shapira et al., 1993] (M6.2) and its distance uncertainty range is based on the distance from the ZA site to the Kalia transverse fault in the northern Dead Sea (Figure 1). It is also a possible scenario that the main fault of the DST ruptured along a limited length causing the 1927 earthquake.

[62] We have excluded the 33 A.D. IBS event from Figure 10 for lack of reliable historical evidence [see Ambraseys, 2009].

[63] In addition to the IBS magnitude‐distance discussion in this section (above), other modifications (Figure 9) made to the published magnitude‐distance diagrams are explained here. Regarding the 363 A.D. event, our review of the evidence indicates two or more separate earthquakes from ∼362 and 363 A.D., with damage in geographically disparate regions (see Appendix A). Also, symbols were added (in Figure 9) for 331 and 199 B.C. and 835 and 847 A.D. historic earthquakes, which are matched to seismites in this study, but not in previous studies at the Dead Sea basin. For the 331 B.C. event, Sbeinati et al. [2005] give intensity VI in the general region of “Syria.” For this ancient and not well‐covered event only a rough calculation is possible. An isoseismal distance of 70 km is consistent with a M6.5 earthquake using the attenuation relation of A&J. This is a relatively ancient event, population density was low, and a distance of ∼70 km from seismic source to historic source is reasonable. For the 199 B.C. event, assuming the intensities documented are from the same event (VII and VIII in “Syria,” probably Damascus, and Sidon, respectively [Sbeinati et al., 2005]), the magnitude is estimated in the same way to Ms6.8. For 847 A.D. the magnitude is taken from the analysis of Sbeinati et al. [2005]. The 873 and 956 A.D. events [Ambraseys et al., 1994; Guidoboni et al., 1994], matched to seismites in this study, are not on the distance‐magnitude diagram for lack of sufficient information.

[64] Second earthquakes were added at ∼mid‐2nd century B.C. and at 362/363 and 747/749 A.D. The location and magnitude of these added events are not known, each appearing in the diagram as a small circle on the symbol of the previously published single event.

[65] The intrabasin seismites that were recorded in all three sites (EFE, EG, ZA) define a well‐constrained field in the magnitude‐distance diagram, which cuts the A&J intensity lines plotted (Figures 9 and 10). Of the earthquakes matched to seismites on this diagram, 60% occupy the field of intensities larger or equal to IV. Eighty‐nine percent of the IBS seismites occupy the field of intensities larger or equal to V (or 100% if 1202 is chosen over 1212 A.D.; see above discussion), as opposed to 46% of all seismites.

[66] Figures 9 and 10 suggest that farther and stronger earthquakes require lower local intensities for being recorded in the entire basin (IBS). If we accept that 1212 A.D. is the IBS (as opposed to 1202 A.D.) at the beginning of the 13th century then it is the farthest (300 km) with M7 and I = IV. Otherwise the 551 A.D. earthquake and the mid‐2nd century B.C. earthquake are the farthest. The intensity threshold for magnitude 6.2 seems to be VII (419 and 1927 A.D.). A possible explanation for this observation is sensitivity to long‐period waves. The frequency content of the wave train is biased to long periods in earthquakes from large and remote sources. A Ms6 earthquake shows a corner frequency fc ∼ 0.1 Hz (period ∼10 s), whereas Ms7 shows fc ∼ 0.04 (period 25 s) [e.g., Geller, 1976]. Attenuation of the wave during travel, where the waves are damped according to the number of cycles between the source and the site in question, results in further bias toward lower frequencies.

[67] The sensitivity to low frequency may indicate that the critical condition for brecciation may depend on ground velocity rather than ground acceleration, where the frequency equals the ratio of the latter to the former. Heifetz et al. [2005] and Wetzler et al. [2010] suggest a Kelvin‐Helmholtz instability mechanism for the disturbances in the sediments (intraformational folding leading ultimately to brecciation). In this scenario the sediment bed is considered to have a gradient in the horizontal velocity (due to a density decrement). If the duration of the wave cycle is sufficiently long (or the frequency sufficiently low), a disturbance can be sustained: the growth rate of a disturbance must be larger than the driving frequency.

[68] The thick gray curve on Figure 10 represents the farthest epicentral distance of liquefaction of soil caused by modern earthquakes in the Aegean region [Papathanassiou et al., 2005]. Note that if 1202 A.D. is the date of the early 13th century IBS (as opposed to 1212 A.D.) then the threshold for intrabasin seismite genesis is very similar to this soil liquefaction curve.

[69] The average recurrence time of IBS is ∼200 years, which is significantly longer than the ∼50–95 years based on all seismites in the Ein Gedi core during the past 1600 years [Migowski et al., 2004] or ∼50 years at EFE since 525 B.C. The possibility to establish a high‐resolution comparison between distinct sedimentary sections located in different sites of the Dead Sea basin opens the way to further explore the response of the lacustrine system to various sources of seismic activity and thus extends the paleoseismic study to older sections such as those of the last glacial Lake Lisan. Such a comparison is currently under investigation.