• from Amit et al. (2002)

• from Zilberman et al. (2005)

• from Zilberman et al. (2005)

• from Amit et al. (1999)

• from Porat et al. (2009)

• from Amit et al. (2002)

• from Amit et al. (2002)

• from Amit et al. (1999)

• from Amit et al. (1999)

• from Amit et al. (1999)

• from Amit et al. (1999)

• from Amit et al. (2002:193)

Faults and lineaments were mapped through the use of 1:7000-scale aerial photographs and satellite images. In addition to aerial photographs taken in 1995 and 1996, aerial photographs from the 1950s and 1960s were used in order to avoid loss of information due to intensive human activity, which has damaged large parts of the area in recent years. In addition, information from the 1:50 000 geologic map of Garfunkel (1970) was incorporated. The resulting map is shown in Figure 2.

• from Amit et al. (2002:193-194)

To characterize the structural pattern of the shallow subsurface fault system, a high- resolution, shallow seismic reflection survey was conducted across the shear zone. Four geophysical lines were shot: one across the marginal area and three across the central area (Fig. 3). The lines followed a previous deep seismic reflection survey at the same locations (Ben Gai et al., 1993). They were shot in a generally east-west orientation by using a conventional common-midpoint (CMP) technique. The spacing between receiver stations on all lines was 2.5 m. The source was applied at every station by using split-spread geometry (Shtivelman et al., 1995). In most cases, a single blow of a hammer was sufficient to generate P and S waves. The data were processed with the PROMAX package. Because only the shallow information was needed, no static corrections were applied to the data (Shtivelman et al., 1998).

• from Amit et al. (2002:194-195)

Information derived from mapping geomorphic features and from the shallow seismic reflection survey served as the basis for site selection for trenching across morphologic and tectonic elements. Trenches were opened and logged at a 1:20 scale. The trenches are 10–70 m long, up to 5 m deep and 1.5 m wide, with vertical walls. One wall in each trench was selected for mapping. A metric horizontal datum line was established along the trench wall and vertical rod sections were taken to delineate contacts of alluvial and colluvial units and pedogenic horizons. Alluvial and colluvial stratigraphic units were differentiated on the basis of texture, structure, type of boundary, color, degree of cohesion, and pedogenic features (McCalpin, 1996). The soils and paleosols exposed in the sedimentary sequences were described, sampled, and analyzed following procedures of the Soil Survey Staff (1975) and Birkeland (1984).

• from Amit et al. (2002:195)

Two numerical dating methods were applied in this study: Conventional ¹⁴C radiometric dating and luminescence dating. Most sediments were dated by using luminescence methods; unless stated otherwise, all ages presented in this study are OSL (optically stimulated luminescence) ages. The gypsic and salic soils and associated sediments in the study sites rarely contain organic material and in general do not contain authigenic CaCO3 that can be dated by ¹⁴C. Therefore, only the playa sediments, in which a few charcoal samples were found, were dated by using ¹⁴C.

The luminescence technique dates the last sunlight exposure of mineral grains in the sediment, which resets the luminescence signature (Aitken, 1998). Samples were collected from holes drilled into natural outcrops or trench walls, at a depth of ~30 cm. To prevent exposure to daylight, the samples were immediately put in light-tight bags, and all subsequent sample preparation was made under subdued orange light. A sample was also collected from the present-day stream bed of Wadi Avrona, to test for the degree of resetting of the IRSL (infrared stimulated luminescence) signature in modern sediments. Laboratory procedures follow those outlined in Porat et al. (1999). Alkali feldspars with densities of <2.58 g/cm3 were extracted from the modal fine-sand size fraction of each sediment sample. Equivalent doses (De) were determined by using the IRSL signature and either the single-aliquot or multiple-aliquot added dose procedure (Aitken, 1998). Most of the dates mentioned here were previously published in detail (Porat et al., 1996, 1997; Amit et al., 1999). Semiquantitative ages were obtained from morphologic age dating by using the slope-age software (Enzel et al., 1996; Nash, 1987).

For relative dating, a chronosequence of Reg (Torrifluvent, Typic Haplogypsid, or Gypsic Haplosalid) soils was calibrated against dated alluvial surfaces, whose ages were known from ¹⁴C, OSL, and prehistoric artifact dating (Amit and Gerson, 1986; Amit et al., 1993; Avni et al., 2000). Reg soils can be used to differentiate among four different time periods: late Holocene (2–4 ka), early Holocene (10 ka), early Holocene to late Pleistocene (10–14 ka), and Pleistocene (older than 50 ka) (Amit and Gerson, 1986; Amit et al., 1995; Avni et al., 2000).

• from Amit et al. (2002:195)

To induce a surface rupture, an earthquake of at least M 6 is needed (Bonilla et al., 1984; Slemmons et al., 1989; Wells and Coppersmith, 1994). This statement implies that the tectonic events that occurred in the study area and that were detected in the trenches were at least of this magnitude. The approach used in this study for estimating paleoearthquake magnitude M is the maximum-displacement (Dmax) method (Bonilla et al., 1984; Wells and Coppersmith, 1994). The equation used for calculations of magnitude M, for normal faults, is M = 6.61 + (0.71 log Dmax). We applied this equation to provide a regression of M on log Dmax for 80 earthquakes in the historical data set (Wells and Coppersmith, 1994). Most of the studied faults are of normal multiple-event type with moderate levels of spatial preservation. On such multiple-event faults, the method based on integration of several displacement samples collected along a surface rupture with globally derived sampling statistics such as proposed by Hemphill-Haley and Weldon (1999) is not applicable. The surface-rupture length method for paleomagnitude estimation is not appropriate either for the study site because erosional obscuring, burial by younger deposits, and fragmentation of the continuous fault line by alluvial-fan incisions make the length measurements uncertain. In relation to the field evidence and the degree of preservation of the faults, our assumption was that the displacements observed in the trenches were maximal, on the basis of the morphologic evidence that led us to excavate the trenches at sites with the largest and clearest apparent offsets along each detected fault line. This assumption might not be true for all the events in a single studied point. But in the light of the problematic correlation between events that occurred simultaneously on an entire surface rupture in combination with dating uncertainties, there is no better way to tackle this issue. As defined by Wells and Coppersmith (1994), the maximum-displacement regression provides only the expected largest slip at a point along a rupture.

• from Amit et al. (2002:195)

A high-resolution seismic reflection survey was conducted at the study site. One line was shot across the alluvial fan of Wadi Shehoret and the Avrona playa sites (Fig. 2, line 1) and three others across the Avrona playa site (Shtivelman et al., 1998) (Fig. 2, lines 2, 3, 4). The interpreted seismic sections reveal that the Elat fault is composed of several distinctive subparallel branches. The subsurface fault pattern is complex, with several converging and diverging north-northeast–trending subparallel fault splays. The fault zone, interpreted as consisting of flower structures, is several hundred meters wide. Its surface expression changes along its trace, alternating between compressional and extensional structures along the central part of the fault zone; normal faults form the fault-zone margins. Generally, the subsurface structures are consistent with the surface expressions of the faults across the alluvial-fan surfaces and the playa (Amit et al., 1999; Shtivelman et al., 1998).

• from Amit et al. (2002:202-203)

Although various types of primary evidence have been used to infer magnitudes of paleoearthquakes, the length of surface rupture and the maximum displacement on continental fault traces are by far the most commonly used parameters (Bonilla et al., 1984; McCalpin, 1996; Wells and Coppersmith, 1994). In the Elat fault zone, the amounts of displacement during the late Pleistocene were higher, more consistent, and less variable than those occurring during the Holocene. The maximum-displacement method for evaluating paleoearthquake magnitudes shows that during the past 40 k.y., the earthquake magnitudes ranged between M 5.9 and M 7 (Fig. 11; Table 1); they showed a narrow range of M 6.7 to M 7 between 40 and 18 ka and a wider range of M 5.9 to M 6.7 since 18 ka. In addition, a decrease in earthquake magnitudes is evident. The highest-magnitude earthquake occurred between 40 ka and 18 ka. From 18 ka until the present time, smaller-magnitude events were common; the smallest-magnitude events occurred between 10 ka and 2.5 ka.

The use of the surface-rupture length method for evaluating paleoearthquake magnitudes is limited at the study site because of intense erosion and deposition processes at the central part of the Dead Sea rift. However, the two length values obtained, one from the marginal fault zone (3.5 km) and the other from the central part of the fault zone (35 km), showed a magnitude range of M 6 to M 6.9. These values are in agreement with the values derived from the maximum-displacement method.

Historical evidence of seismicity during the past 2500 yr along the Dead Sea rift documents ~110 earthquakes of local magnitudes in the range of 6.7 < Ml < 8.3 (Ambraseys et al., 1994; Ben-Menahem, 1991). In the southern Arava Valley segment, only two large events are inferred to have occurred during this period: the earthquake of A.D. 1068 (M >7) and the earthquake of A.D. 1212 (M 6.7) (Ambraseys et al., 1994; Amiran, 1994; Zilberman et al., 1998). However, earthquakes larger than M 7 have not been detected in the southern Arava Valley area. The fluctuations of earthquake activity appear to reflect changes in the seismic response of the Elat fault system (see subsequent discussion on this aspect).

• from Amit et al. (2002:203-204)

Recurrence intervals and slip rates are two fundamental descriptors of paleoseismicity, and both are critical components of seismic hazard assessments. In the present study, the recurrence interval involves dating each separate paleoearthquake directly (the geologic method) (McCalpin, 1996; Working Group on California Earthquake Probability, 1995). The purpose of dating almost each paleoearthquake is to bracket the time of faulting during the past 40 000 yr as tightly as possible. The analysis of six faults (seven trenches) and 27 dated tectonic events in the southern Arava Valley revealed that the long-term recurrence interval of large earthquakes in this area (M >6) is 4.4 ± 1.1 k.y. This value differs from the previously proposed value for this area of 1.5 k.y., which was determined on the basis of archaeologically determined seismicity and historical evidence from a time period only since 4 ka (Ben-Menahem, 1991).

One approach to interpreting these data is to subdivide the 40 k.y. time period into two subperiods (Fig. 11). One is the late Pleistocene (until 18 ka), which is characterized by large-magnitude earthquakes (M 6.7–M 7) and recurrence intervals of 2.8 ± 0.7 k.y. The second is the Holocene period, characterized by a higher frequency of tectonic events, a wider range of earthquake magnitudes (M 5.9–M 6.7), and a recurrence interval of 1.2 ± 0.3 k.y.—a value that is closer to the 1.5 ± 0.3 k.y. calculated by Ben-Menahem (1991) for the past 4 k.y. Events with magnitudes larger than M 6.7 were not detected during the past 18 000 yr. A similar trend is indicated in Figure 12, which shows that the elapsed time is longer on faults that were active mainly during the late Pleistocene than on those that were active also or only during the Holocene. The other more cautious way of interpreting the data is to accept the long-term recurrence interval of 4.4 ± 1.1 k.y. with a wide range of earthquake magnitudes. This more conservative value can be problematic when used for seismic hazard assessment, especially in the light of the currently nonactive fault system.

In the study site there are no measurable geomorphic or stratigraphic markers for determining lateral slip rates. As a result, only normal faults can be used for slip-rate evaluations. The average slip rate of the normal faults is 0.2 mm/yr. This value changes through time on different segments of the active, wide shear zone and among clusters of events related to the same segment (Table 1). The change is in the range of 0.1–0.35 mm/yr. This range indicates low slip rates in comparison to the worldwide slip-rate range for normal faults, which vary from <0.5 mm/yr to >3 mm/yr (Yeats et al., 1997).

Nonuniformity of large-slip events on seismogenic faults and variations of slip rates are normal characteristics of the faulting in the southern Arava Valley area. Considering that extension has characterized the Elat shear zone province during most of the past 40 k.y., it would be surprising if slips on individual faults were uniform for long periods of time or if local areas did not become active while others remained quiescent (Wallace, 1987; Yeats et al., 1997). Such a situation exists in most places, especially if the depth of the earthquakes is <20 km (Shapira and Shamir, 1994; Wallace, 1987). In addition, the cumulative displacement on faults is an increasing function of fault length (Scholz, 1982; Scholz et al., 1993; Sornette and Davy, 1991; Wesnousky, 1999). For example, the fault zone of the northern Arava Valley is longer and narrower than the fault zone in the southern Arava. With continued displacement, longer faults develop at the expense of shorter faults, and the longer faults accommodate an increasingly greater part of the displacement budget (Wesnousky, 1999). Thus, it may be speculated that the Elat fault zone is not subject to constant regional displacement rates. Instead this fault zone may be too young to have had enough cycles of large events (Ben-Zion et al., 1999).

• from Amit et al. (2002:204-205)

Comparative analysis of paleoseismic data from the different branches of the Elat fault zone shows that it consists of subzones in which several faults are active together. Each tectonically active subzone is at least 0.5 km wide. Taking into account the dating uncertainties, at least two neighboring branches of the irregular shear zone were tectonically active simultaneously during several of the total detected earthquakes (Fig. 6, Table 1). However, along one fault line, not all the tectonic events observed in one trench were detected in the other trenches. For example, trenches T-6, T-17, and T-16 are located along the same fault line (Fig. 2), but they differ from each other in the number of detected tectonic events, displacement amount per event, and total displacement amount. However, several events are correlative between these trenches (Table 1). Such changes could be due to high variability in the location of the basement beneath the alluvial fill. Indeed, a trend of deepening of the basement northward was detected by the geophysical survey (Frieslander, 2000). Furthermore, the Holocene has been a time of intense tectonic activity in the central part of the Arava Valley, followed by a decrease in tectonic activity at the valley’s western and eastern margins. This trend is depicted in a time and space diagram (Fig. 13), which shows that with time the large events migrate east toward the central part of the fault zone. However, the inference that most of the earthquakes during the Holocene were not as strong as those during the Pleistocene can be interpreted in two ways (Fig. 11): (1) During the Holocene, the magnitude range determined for the central zone (M 5.9–M 6.7) was likely not high enough to activate the marginal faults, or (2) the evidence for the larger magnitudes during the Pleistocene in the central zone is buried under the thick Holocene alluvial fill in the Avrona playa and cannot be reached and exposed by trenching. Nevertheless, both of these options support the hypothesis that the behavior of the Elat fault zone is time dependent and is characterized by changes in the faulting sequence.

Another important observation is that a particular fault or group of faults in the study area usually do not generate successive large and similar displacements per event, which is basic to the concept of the characteristic earthquake model (Schwartz and Coopersmith, 1984). For example, fault analysis in trenches T-6 and T-17 shows variable displacements at the same location and along the same fault line location on each of the different faults (Table 1).

As the southern Arava Valley fault system is distinguished by complex and nonhomogeneous fault traces, its frequency versus magnitude distribution can be much better supported by the Gutenberg-Richter (GR) distribution than by the characteristic earthquake model (CE) (Kagan, 1993; Wesnousky, 1999). The equation for the GR distribution is Log N = a + bM, where a and b are empirical constants, and N is the cumulative number of earthquakes of magnitude M or greater. The frequency versus magnitude equation for earthquakes in the local-magnitude range 1.5 < Ml < 8 established for the Elat fault yields a b-value of 0.86 (Ben-Menahem, 1981, 1991) and 0.9 (Shapira and Shamir, 1994). A frequency between 1000 and 5000 yr is derived from the GR model for earthquakes with magnitude M 6.5 to M 7. The long-term recurrence interval of 4.4 ± 1.1 k.y. calculated in this study for large events (M 6.5–M 7) over a time period of 40 000 yr is closer to the upper limit of the frequency value derived from the GR model. To conclude, although the GR model is much more consistent with the paleoseismic data presented in this study than the CE model, neither model results are in good agreement with the paleoseismic data.

These models also fail to explain the pattern whereby a decrease in the tectonic activity on one fault branch is accompanied by an increase in the activity on parallel branches during the same time period. To illustrate, trench T-6 shows two clusters of tectonic activity, a cluster of four events between 37 and 14 ka with magnitudes exceeding M ~6.8 and a cluster from 14 ka until the present during which three events occurred, all of magnitudes less than M 6.5. By itself, the fault behavior exhibited in this trench could be a classic example of a change with time from CE model behavior to a GR power-law relationship, as proposed by Leonard et al. (1998) on the basis of paleoseismic data from that trench (T-6). However, paleoseismic data from other trenches across the Elat fault zone show that during the same time period, tectonic activity and the amounts of displacement on faults exposed in trenches T-20 and T-18 increased (Fig. 7) and ranged from 0.2 to 1.3 m with evaluated magnitudes between M 6.1 and M 6.7 (Table 1). The clear trend detected in trench T-6 of a change with time from a CE model to a GR model was not observed across the entire Elat fault zone and cannot be easily applied to the southern Arava Valley area, as suggested by Leonard et al. (1998).

The model ‘‘mode switching’’ suggested by Ben-Zion et al. (1999) might better explain the seismic activity pattern of the southern Arava Valley area. According to this model, a relatively high ratio of characteristic times of damage healing to loading leads to the development of geometrically regular fault zones and frequency-size event statistics compatible with the CE earthquake distribution. Relatively low ratios of damage healing to loading lead to the development of highly disordered fault zones and frequency-size statistics compatible with the GR distribution. Intermediate ratios represent a fault zone with a seismic response that alternates between time intervals of intense seismic activity containing clusters of large events and lower-activity periods during which only small to moderate earthquakes occur. Such fluctuations of earthquake activity, which are typical in the Elat fault zone, appear to reflect mode switching of the seismic response of an individual, large fault system. An extreme manifestation of mode switching of seismic activity may produce a transition between brittle (seismic) and effectively creeping (or aseismic) responses to tectonic loading (Ben-Zion et al., 1999).

Yet another alternative is that the switching and the changes of seismic activity across the Elat fault zone with time represent a change in the mode of strain release on the underlying fault plane. Such changes can end with the formation of a locked fault zone. This possibility is supported by the observed change from large earthquakes to smaller ones with time in combination with evidence of quiescence during the past 1000 yr and no current microseismic activity (Shapira and Feldman, 1987). It is also consistent with the observation that the faults show a change in their directions with time (Gerson et al., 1984). The oldest and westernmost faults have a north-northwest trend, and the youngest faults, located in the central part of the Arava Valley, trend north-northeast.

• from Amit et al. (1999:78)

The geomorphology of the playa area was mapped using 1:7000 scale airphotos and satellite images. In addition to recent airphotos taken in 1995 and 1996, airphotos from the 1950s and 1960s were used to avoid loss of information due to intensive human activity. In addition, information from the 1:50,000 geologic map of Garfunkel (1970) was incorporated. The resulting map, shown in Figure 2, presents the playa morphology, drainage system, faults, inferred faults, and lineaments.

• from Amit et al. (1999:78)

To characterize the structural pattern of the shallow subsurface fault system of the DST and to evaluate its relation to the playa surface morphology, a high-resolution shallow seismic reflection survey was conducted across the playa at three locations 3 to 5 km apart (Fig. 3). These geophysical lines followed a previous deep seismic reflection survey at the same locations (Ben-Gai et al., 1993). The present survey included three P-wave seismic lines (GI-0061, GI-0065, and GI-0074) and one S-wave line (GI-0066), which overlapped line GI-0065 (Fig. 3). The lines were shot in a generally east–west direction using conventional common midpoint technique (Shtivelman et al., 1997).

• from Amit et al. (1999:78)

Information derived from geomorphic mapping and the shallow seismic reflection survey served as the basis for site selection and for trenching across morphological and tectonic elements. Three trenches were opened and logged at a 1:20 scale (Fig. 3). The trenches are 10 to 70 m long, up to 5 m deep, and 1.5 m wide and have vertical walls. Two trenches (T-21 and T-20) were excavated across a fault scarp. A third trench (T-22) was excavated in the easternmost part of the playa, across an elevated ridge composed of playa deposits. One wall in each trench was selected for mapping. Alluvial and colluvial stratigraphic units were differentiated on the basis of texture, structure, type of boundary, color, type of salic horizons, degree of cohesion, and pedogenic features (McCalpin, 1996). The sedimentary sequence was also analyzed in two outcrops: one in the central part of the playa (E-23, Fig. 3) and a second in its southern part (E-24, Fig. 3).

The soils and paleosols exposed in the sedimentary sequences were described, sampled, and analyzed according to the Soil Survey Staff (1975), Birkeland (1984), and Amit et al. (1995).

• from Amit et al. (1999:78,81)

Two dating methods were applied in this study: conventional ¹⁴C radiometric dating and luminescence dating. The gypsic and salic soils and the sediments in this extremely arid region only rarely contain organic material and in general do not contain authigenic CaCO₃ that can be dated using ¹⁴C. Therefore only a few charcoal samples were dated using ¹⁴C; most sediments were dated using the luminescence methods. Unless stated otherwise, all ages presented in this study are infrared stimulated luminescence (IRSL) ages.

The luminescence techniques date the last exposure of mineral grains in the sediment to sunlight, which resets the luminescence signal (Aitken, 1998). After burial the signal increases again with time as a result of the natural ionizing radiation emitted by radioactive isotopes (U, Th, and K) and by cosmic rays that penetrate the sediment. In this study the IRSL signal was used, because it requires only several minutes of exposure to sunlight for complete zeroing and resetting of the luminescence “clock” (Aitken, 1998). This signal is most suitable for dating desert fluvial and colluvial sediments such as those in the study area (Clarke, 1994; Porat et al., 1996; Tandon et al., 1997).

Alkali feldspars with densities less than 2.58 g/cm³ were extracted from the modal fine-sand size fraction of each sediment sample under subdued light, using routine laboratory procedures (Porat et al., 1996, 1997). The equivalent dose (D_e) was determined by the single-aliquot method (Duller, 1991), using the infrared emission at 880 nm for stimulation. External and internal dose rates were obtained as in Porat et al. (1997, 1999).

Present-day water contents in the sediment samples were measured in the laboratory. However, the values obtained do not reflect water content throughout the sediment’s burial history. Pedogenic analysis indicates that the water table at the Avrona Playa was at the surface several times during the past 10,000 yr. Laboratory experiments gave saturation water contents for several sediment types from Avrona in the range of 20–30%. This range represents past water content, and therefore ages were calculated using both the measured water contents and the saturation values of 25 ± 5%.

Table 1 summarizes the data used for calculating the ages of the sediments. Uncertainties in the D_e values were calculated from the standard deviation of the individual single-aliquot D_e determinations. Other error estimates are also at the ±1 s confidence level and include systematic as well as random errors. The ages presented on the trench logs and soil profiles are those calculated using saturation water contents.

• from Amit et al. (1999:81)

The seismic history of the Avrona Playa was reconstructed by analyzing the geomorphology of the playa and the fault pattern in the playa sediments.

• from Amit et al. (1999:81)

The Avrona Playa area now forms a broad elongated ridge. Its relief is composed of low longitudinal subridges (0.5–4 m high and up to 5 km long) and shallow braided channels flowing along and across these ridges (Figs. 2 and 3). Nahal Raham flows from the west, cuts the northern edge of the playa, and then flows southward at its eastern edge along a gradient of 0.1–0.15%, forming steep and deep playa margins. The western side of the playa moderately slopes toward the alluvial fans of Nahal Avrona, Shehoret, and Roded. Nahal Avrona, which drains the western margin of the rift, crosses the entire width of the playa and merges into Nahal Raham. The outlet of the Avrona Playa is a narrow channel that flows along the eastern margin of the steep (4–5%) alluvial fan of Nahal Roded, which separates the Avrona Playa from the Elat Sabkha (Fig. 2).

The ridges crossing the playa expose gypsic and salic units interbedded with fine sand and clay. In most playas such sterile fine-grained salic deposits occupy the lowest elevations. However, in the Avrona Playa they are found at the highest topographic position. In addition, the surface of the playa has a patchy appearance, with remnants of sterile nonvegetated areas situated adjacent to gypsic and nonsalic vegetated areas. This indicates that fluvial incision has affected the natural zonation of the playa. Thus, the Avrona Playa is at present externally drained and does not function as a closed basin with an endoreic fluvial system, as expected of playas in arid zones.

• from Amit et al. (1999:81,85)

The results and interpretations of the geophysical survey are shown on Figure 4. The three seismic sections show complicated fault patterns identified as several flower structures. Some of the interpreted faults coincide with surface tectonic morphologic elements, such as surface ruptures, lineaments, and small compression or extension structures that permit correlation among the lines. The interpretation revealed that the Avrona Fault is composed of several distinctive subparallel branches. These are labeled A–E in Figure 3. The subsurface fault pattern is complex, with converging and diverging NNE-trending faults extending for several hundred meters.

Zone A on lines GI-0074 and GI-0065 (Fig. 3) is associated with NNE-trending morphologic features that can be traced between the two lines. The eastern branch of zone A on line GI-0074 (Fig. 4), which shows a compressional structure, is expressed at the surface as an elevated ridge built of uplifted playa deposits. In some places outside the lines, the ridge forms a topographic step up to 4 m high. The western branch of zone A is expressed as a low, elongated depression occupied by braided shallow channels of Nahal Avrona. The northern extension of the western fault of this shear zone displaces the playa surface, forming a sharp, 2-km-long east-facing fault scarp.

Zone B, appearing on lines GI-0074 and GI-0065 (Fig. 4), is expressed by a row of low hills capped by gypsic unit of groundwater origin, which is elevated above the east-flowing alluvial fans. The eastern flank of the zone is related to an elongated ridge extending south of line GI-0065 and forming a straight lineament about 3 km long and 100 m wide. A shallow depression occurs between the western and the eastern branches of flower structure B. This depression serves as the local base level for the east-flowing streams blocked by the low ridge in the east.

Zone C on line GI-0065 (Figs. 3 and 4) seems to be related to zone C on lines GI-0061 and GI-0074. Zone C is situated along the most prominent lineament in the Avrona Playa. This lineament is about 10 km long. In the southern part of the playa, it forms a low ridge bounded in the west by a short west-facing scarp; in the northern part, it is expressed as an approximately 1-m-high east-facing scarp. A clear discontinuity in the reflectors coinciding with the surface scarp occurs in line GI-0061. A trench at this site (T-20, Fig. 3) exposed a normal fault.

Zones D and E (Figs. 3 and 4) are expressed as a series of low hills composed of alluvium cemented by gypsum. These hills represent uplifted alluvial-fan sediments deposited at the western playa margins.

The results of the high-resolution seismic reflection survey at the Avrona Playa area enable us to detect in the subsurface, and to map at the surface, several NNE- and NW-trending shear zones that extend continuously between the seismic lines. The seismic lines across the Avrona Playa provide evidence for several NNE-trending subparallel fault splays. The zones are interpreted as subsurface flower structures, which are typical of strike-slip faults (Sylvester, 1988). The surface expression of the flower structures changes along their traces and they exhibit alternating compression and extension features.

• from Amit et al. (1999:85)

The field data, laboratory results, and calculated ages of the trench and outcrop samples are listed in Table 1 and shown on Figures 3, 5, and 7. The main source of error for all samples is the scatter in individual single aliquot D_e determinations, ranging from 4 to 24%, depending on the degree of homogeneity of the sediments. For most samples, errors of less than 5% indicate that they were well bleached at the time of deposition (Clarke, 1994). Sediments collected from the active channel of Nahal Avrona in the Avrona Playa yield residual D_e values of 0.28 ± 0.09 Gy, equivalent to about 110 yr when using an average annual dose rate of 2500 mGy/a (Table 1). Therefore the ages shown on the figures and in Table 1 are probably overestimated by about 100 yr.

IRSL ages in all sections and trenches are in stratigraphic order (Figs. 5 and 7) and show approximately uniform rates of sedimentation for the upper parts of the sections (Fig. 3). The ages obtained by IRSL and ¹⁴C are compared in Table 2. Sample A-8931 from trench T-21 yields a calibrated ¹⁴C age in the range of 1519–1283 yr which agrees with an estimated IRSL age of ;1500 yr, calculated by extrapolating between samples T22-1 and T22-2.

For the samples from trench T-20, ¹⁴C ages are older than IRSL ages. This discrepancy may be due to sampling restrictions for radiometric dating in hyperarid environment. Only minuscule charcoal fragments were available for sampling, and they were collected from an extended horizon 0.5 m thick and 2 m long. In addition, the source of organic matter found in Avrona is mainly from Acacia trees growing upstream, which can reach an age of several hundred years before being washed down and incorporated in the sediment. Thus, the ¹⁴C dates represent the maximum age of sediment deposition. On the other hand, the IRSL ages may represent a minimum age due to possible pedoturbation of the upper 1 m of sediment.

Previous experience shows that colluvial sediments attached to a fault plane are not well bleached at the time of deposition (Porat et al., 1997). Therefore the colluvial material in trench T-20, which is preserved only close to the fault plane, is not suitable for luminescence dating.

• from Amit et al. (1999:85)

Several high longitudinal subridges composed of sterile playa deposits were trenched and their sedimentary sequences were studied (Fig. 3; fault zone A, trench T-22 and outcrop E-23; and fault zone C, trench T-21). The exposed sedimentary sequence of the easternmost and highest elevated ridge shows alternating fluvial and playa deposits (Fig. 5). A silty–clayey unit occurs at the bottom of the sequence at a depth of 6.5 m from the uppermost ridge surface (Fig. 8, unit 1). It has a weak laminar structure with many non-pedogenic gypsum lenses scattered through the sediment. The age of this unit is 8700 ± 1600 yr. It is overlain by a thick fluvial unit (2–3 m) composed of fine gravel, coarse sand, fine sand, silt, reworked gypsic aggregates, and mud crusts (Fig. 5, unit 2). Cut-and-fill and well-developed cross-bedding structures, which are defined by the reworked gypsic aggregates, indicate that this unit was deposited in wide and shallow channels oriented NW–SE. The middle section of this unit was dated 4100 ± 300 yr. The upper part of the sequence is composed of barren, sterile playa deposits in which a 1.20-m-thick saline soil (Salorthids, Soil Survey Staff, 1975; or Solontchak, Duchaufour, 1977) developed (Fig. 5, unit 3, Fig. 6). Their age is 1400 ± 100 yr. The soil is characterized by in situ pedogenic gypsic nodules, salt efflorescences, and a well-cemented petrosalic horizon at a depth of 1 m (Fig. 5). Reworked playa deposits composed mainly of silt, on which a saline soil has also developed (Fig. 5, unit 6), cap the lower erosion surface. As in unit 3, the soil profile contains large amounts of halite and gypsum scattered throughout; however, it does not have a petrosalic horizon.

The westernmost longitudinal subridge, T-22, shows that playa sediments were deposited during two periods; between these two periods (IRSL age of 3600 ± 300 yr) the playa area was fluvially drained. Prior to 8700 yr and after ;2000 yr, deposition of playa sediments followed by soil development were the dominant processes. Playa sediment deposition and soil development mark stages of stability during which the basin was probably closed or semiclosed and had a high water table.

In summary, the sterile playa deposits at the Avrona Playa are at the highest topographic position and are at present under intensive erosion. The sedimentary sequence of the elevated deposits is composed of alternating playa and fluvial sediments, with playa deposits at the top. At all the studied sites, a Solonchack soil with a petrosalic horizon at a depth of 1 m is present in the upper playa unit (Fig. 3). In a closed basin such soil can develop only if the water table is high.

• from Amit et al. (1999:86-88)

A trench (T-20) was excavated perpendicular to one of the east-facing fault scarps related to Zone C (Fig. 3). The fault scarp, 2 km long and 1 m high, crosses a transitional, nonsterile, vegetated playa zone composed of sand to sandy-loam sediment. On the down-faulted block along the scarp, an active channel flows southward, along it are coarse sand, coarse gypsic reworked aggregates, and granules.

The trench exposed a normal fault of multiple-event type with one main fault plane (Figs. 7 and 8, Table 3). The exposed sedimentary sequence exhibits three different depositional environments that are not always consistently represented on both sides of the fault plane. The first environment is a playa, in which mainly clay and silt are deposited (Fig. 7, unit 1a). This unit occurs only on the uplifted block; its absence from the down-faulted block probably results from intense fluvial activity on this side of the fault. The second environment is a fluvial system that deposited coarser sediments (gravel, sand, and reworked gypsum aggregates) on both sides of the fault plane in broad shallow channels flowing southward (Fig. 7, units A, 1b, 2, 3). The third depositional environment is represented by a combination of eolian and fluvial–eolian sheet-flow deposits of playa margin type composed of fine-grained sediments (fine sand and silt). This unit was deposited first on the down-faulted block and later on both sides of the fault (Fig. 7, units 4, 5, 5a, and 5b). A soil was detected in this unit (units 5–5a,b). The soil is a typical Salorthid (Solontchak) similar to the profiles detected on top of the elevated longitudinal ridges (Fig. 3; T-21, T-22, and E-23). The soil has weak horizonation, salt crystals (mainly halite and gypsum) scattered in the sediment, and a well-cemented 5-cm-thick petrosalic horizon composed mainly of halite that formed at a depth of 50 cm. The soil profile developed in units 5a and 5b and reached the same degree of development on both sides of the fault plane.

Five partially preserved colluvial units were identified adjacent to the fault plane (Fig. 7, units I–V; Fig. 8). This succession of vertically stacked colluvial wedges maintains a constant thickness up to the fault contact and terminates abruptly at the fault plane. Each wedge merges with a fluvial unit on the down-faulted block (units 3 and A). The colluvium is gravely to sandy depending on the composition of the sedimentary units from which it was derived. Fluvial erosion on the down-faulted block removed most of the material forming the colluvial wedges, leaving only the uppermost part of each wedge. Pedogenic features such as biotubation, gypsic nodules, and horizonation were detected in several of the colluvial wedges. Unit I is composed of a mixture of nonsorted gravel, with sand and clay intraclasts and collapsed clay blocks. The colluvium is stratified parallel to the slope. Unit II is composed of coarse gravel-rich, grain-supported colluvium with a sandy matrix. The gravel ranges from 1 to 4 cm in size and has a rubble-like texture, with no preferred orientation. Unit III is composed of grain-supported gravel-rich colluvium with a coarse sandy matrix. The gravel fines upward, ranges in size from 2 to 10 cm, and was deposited parallel to the slope. These three units do not contain soils. Unit IV is composed of a mixture of sand and granules with no clear sedimentary structure. A soil profile in the upper part of the unit contains shattered gravel, gypsic nodules, and root casts. The sediment is porous and reddish. Unit V is compose of sand with some lithoclasts. This unit did not display any sedimentary structure due to intensive pedogenic processes. Abundant pedogenic gypsic nodules are scattered in the colluvium.

The variations in pedogenic features in units IV and V indicate that each wedge formed during a different time period. The shape of the wedges, the different sedimentological character of each colluvial wedge, the absence of a fining-upward trend in the whole colluvium section, and the absence of correlative colluvial units on the up-thrown block indicate that they are tectonically induced and not a result of nontectonic slope processes (Amit et al., 1995; McCalpin, 1996). The absence of soils in units I–III can be interpreted as an indication of a short quiescence period.

Based on the ages of unit 3, which overlies the colluvial units, and unit 1a at the base of the section, five tectonic events occurred between 14,200 ± 1300 and 3700 ± 300 yr ago. These are summarized in Table 3. Each event was accompanied by a displacement of at least 0.5 m, as calculated and measured along the fault plane (McCalpin, 1996), and each triggered the deposition of a colluvial unit (represented by wedges I–V) with a characteristic sedimentary composition, structure, and the degree of pedogenic alteration.

The deposition of unit 3 marks the beginning of at least 2000 yr of quiescence. During this period, fluvial sediments were deposited mainly on the down-faulted block (Fig. 7, units 3, 5). Scour-and-fill fluvial structures with fluvial crossbedding in these units indicate a north–south flow direction subparallel to the fault trace (Fig. 9).

Fluvial activity along the fault scarp decreased over the last 3000 yr. Toward the end of this period, fluvial–eolian sand sheets were deposited over the entire playa area (unit 5b) and not only on the uplifted block (unit 5a). Unit 5a covered the fluvial units (units 3, 5) on the down-faulted block and the fault trace. At this stage, a soil developed continuously across the inactive faulted area (unit 5b).

Tectonic activity along this fault was renewed sometime between 1000 and 700 yr ago, with a vertical surface rupture of about 1 m (the petrosalic horizon in unit 5b was used as a stratigraphic marker for the measurement of the displacement). In addition to the surface rupture, this event caused reactivation of fluvial activity on the down-faulted block, a rapid retreat of the fault scarp, and severe erosion of the upper soil profile of unit 5b. In the down-faulted block, only a small remnant of the soil profile was left; furthermore, the petrosalic horizon was exposed by slope erosion and partly washed away by fluvial erosion.

• from Amit et al. (1999:88-90)

The Avrona Basin is now not internally drained, as expected of playas. Usually playas in arid zones occupy topographic lows and serve as receptacles of fine sediments delivered by episodic fluvial or eolian inflows (Shaw and Thomas, 1989; Cooke et al., 1993). At present, the Avrona Playa drains externally and displays no zonation; furthermore, its positive relief contrasts with other playas. Despite its inverted morphology, the Avrona Basin preserved the sedimentary and pedogenic characteristics of an endoreic basin. At all studied sites, the elevated playa deposits contain saline Solonchack soils with well-developed petrosalic horizon. Such saline soils form as a result of complex processes involving salt and water transfer at the groundwater–surface interface (Yaalon, 1961; Dan, 1981; Dan and Yaalon, 1982). This leads to the precipitation of evaporite deposits at the surface and within near-surface deposits on the central and lowest parts of a playa. In this zone, episodic flooding, saline soil formation, and salt accumulation discourage vegetation growth and encourage the formation of sterile areas (Yaalon, 1961; Amiel and Friedman, 1971; Zohary, 1973).

In light of the above description, the wide distribution of the Solontchak soil with almost identical profiles on each of the elevated Avrona Playa remnants, including along the trenched fault trace (Figs. 3, 5, and 7), indicates that the soil formed throughout a closed basin with a shallow water table. At present, these soils occur at the highest topographic position, 1 to 4 m above the modern fluvial system that dissects and erodes the playa. The soils are totally disconnected from the water table, which is now ca. 27 m below the surface. The fact that the soil profiles throughout the basin have a similar degree of maturity implies that they were all abandoned at the same time. The similar age range of the host sediments 700–1500 yr (Fig. 3) supports this conclusion. It is suggested that such extreme and synchronous change of the playa surface is related to the last large earthquake that occurred in this area.

An alternative interpretation—that the current configuration of the elevated playa sediments is solely the result of fluvial down-cutting of a filled basin—is unlikely. Under such circumstances, the drainage pattern would form a dendritic network with random distribution of very shallow channels and a high drainage density (Morisawa, 1968). However, in the Avrona Playa, the pattern is a system of straight, deep channels bounded by fault scarps and longitudinal elevated ridges several meters high (Figs. 2 and 3). These features are the surface expression of extensional and compressional structures detected by the high-resolution shallow reflection survey. Such a structurally controlled drainage network indicates recent tectonic activity (Morisawa, 1964).

It is equally unlikely that climatic change triggered the relief inversion. The sediments of the playa are of Holocene age (8000 yr), a time of arid to extremely arid climatic conditions in the southern Arava, as indicated by the widespread occurrence of salic–gypsic soils (Amit et al., 1993). Climatic change in the Avrona region would have been manifested in a playa environment by the formation of either an ephemeral or a perennial body of water or by changes in the frequency of floods entering the playa (Currey, 1994; Enzel and Wells, 1997). Sedimentological or pedological evidence for such hydrological changes are not seen in the sedimentary section. Therefore, it is highly unlikely that the intensive incision, draining, and inversion of the basin resulted from a change in its hydrologic regime. The only way the playa surface can be incised is by tectonic uplift followed by fluvial erosion.

Geomorphic observations further support the tectonic interpretation for the origin of the topographic inversion. The vertical differences between the upper playa sediments and Nahal Raham channel are greatest near the center of the playa (T-22, E-23), where it is 2–4 m. At the margins of the playa (e.g., E-24), this difference decreases to 1 m (Zilberman et al., 1998). This geometry indicates differential uplift rather than natural fluvial erosion of a closed basin (Bull and Knuepfer, 1987; Rockwell et al., 1988).

The transition from a playa to fluvial sedimentary units as recorded in the Holocene sequence of the Avrona Playa is also interpreted to be related to tectonic deformation. According to the paleoseismic analysis, the playa was deformed intensively by tectonic activity several times during the Holocene. Certain tectonic episodes triggered fluvial incision on the down-faulted block and along the fault scarp, resulting in the deposition of fluvial sediments in wide braided channels on top of the previous playa sequence (trench T-20). It is suggested that such alterations, especially in the central part of a playa (e.g., trench T-22), mark a change from a closed basin configuration to an open one that resulted from tectonic deformation.

• from Amit et al. (1999:90)

Based on the analysis of trench T-20, the most recent tectonic event was associated with a vertical displacement of 1 m. This event affected the entire playa area, as inferred from the correlation between the soils and sedimentary sequence in trench T-20 and those of the elevated playa remnant in sites T-21, T-22, E-23, and E-24 (Fig. 3). The age of the faulted sedimentary sequence in trench T-20 ranges between 700 and 1500 yr (Table 1 and 3; Fig. 3). Therefore, the sedimentary sequence was uplifted and abandoned after this date. Historical evidence supports this result, for only one damaging earthquake (1212 AD) was reported during the period from 184 BC to 1992 AD (Amiran, 1996). The earliest account states that it was strongest in the region of the ancient city of Aila (Fig. 1), suggesting that its epicenter was in the southern Arava Valley (Ambraseys et al., 1994).

Using the average displacement parameter (Bonilla et al., 1984; Wells and Coppersmith, 1994), a vertical displacement of 1 m, such as in trench T-20, can be produced by an earthquake with a magnitude greater than 6.5. Another indication of the earthquake magnitude is the rupture length of the fault (Wells and Coppersmith, 1994). The uplifted ridge formed by playa deposits is about 35 km long [the exact rupture length is uncertain due to erosion, burial by younger deposits, and difficulties in distinguishing the end points (Bonilla et al., 1984; McCalpin, 1996)]. This is a minimum estimate of the original rupture length and it is consistent with an earthquake having a magnitude of about 7 (Wells and Coppersmith, 1994). However, the width of the deformed zone is up to 1.5 km, indicating that the branches of the shear zone were reactivated in one event. In this case, the magnitude estimate should take into account the seismic energy released by all the segments of the shear zone (McCalpin, 1996). Therefore, the estimated magnitude of 7 should be considered a minimum for the earthquake that triggered the latest relief inversion in the Avrona Playa.

• from Zilberman et al. (2005:83)

The geomorphological and morphostructural features of the Avrona playa area were mapped using 1:7000 airphotos and satellite images (Fig. 2). In addition to recent airphotos taken in 1995 and 1996, airphotos from the 1950s and 1960s were used to obtain information since lost due to intensive human activity.

Detailed stream profiles and topographical cross sections (vertical error of ±20 cm) across the playa area were drawn based on photogrametric maps. The present low gradient (0.1–0.15%) of the stream channels of Wadi Avrona and Wadi Raham, which drain the Avrona basin, serves as a reference line for estimates of the playa area deformation.

• from Zilberman et al. (2005:83-84)

To characterize the structural pattern of the shallow subsurface fault system of the DST and to evaluate its relation to the playa surface morphology, a high resolution seismic reflection survey was conducted across the Avrona playa at three locations 3 to 5 km apart (GI-0061; GI-0065; GI-0074; Fig. 3). These geophysical lines followed a previous deep seismic reflection survey at the same locations (Ben Gai et al., 1993).

The present survey included three P-wave seismic lines (GI-0061; GI-0065; GI-0074) and one S-wave line (GI-0065). The lines were shot in a general east–west direction using the conventional common midpoint technique (CMP) with 2.5 m spacing between receiver stations. The source was applied at every station using split-spread geometry (Shtivelman et al., 1998). The energy source was an 8 kg sledge hammer. In most cases a single blow of the hammer was sufficient to generate P- and S-waves. We used a 48-channel EG and G seismograph, model ES-2401X. PROMAX package was used for data processing. Since only the shallow information was needed, no static corrections were applied to the data (for more details see Shtivelman et al., 1998).

• from Zilberman et al. (2005:84)

Information derived from geomorphic mapping and from the shallow seismic reflection subsurface data served as a basis for site selection for trenching across morphological elements. Three trenches were opened and logged at a 1:20 scale (Fig. 3). The trenches are 10 to 70 m long, up to 5 m deep and 1.5 m wide. Trench T-20 was excavated across a fault scarp. Trench T-21 was excavated perpendicular to a morphological scarp in the northern part of the playa. Trench T-22 was excavated in the easternmost part of the playa across a structural ridge composed of playa deposits. One wall in each trench was selected for mapping. A metric horizontal datum line was established along the trench wall and vertical rod sections were taken to delineate contacts of alluvial and colluvial units and pedogenic horizons. Two exposures of the sedimentary sequence were also analysed: one at the central part of the playa (E-23) and a second (E-24) at the southern part (Fig. 3). Alluvial and colluvial units were differentiated on the basis of texture, structure, type of boundary, colour, type of salic horizons, degree of cohesion and pedogenic features (McCalpin, 1996). The soils and paleosols exposed in the sedimentary sequences were described, sampled and analysed after the Soil Survey Staff (1975).

• from Zilberman et al. (2005:84)

Two dating methods were applied in this study: conventional ¹⁴C and luminescence dating. The gypsic and salic soils and the sediments in this extremely arid region only rarely contain organic material and in general do not contain authigenic CaCO₃ that can be dated by ¹⁴C. Therefore, only a few charcoal samples were ¹⁴C dated. Most sediments were dated using luminescence methods and thus unless stated otherwise, all ages presented in this study are luminescence ages.

The luminescence technique dates the last exposure of mineral grains in the sediment to sunlight, which resets the luminescence signal (Aitken, 1998). After burial, the signals increase again with time as a result of the natural ionizing radiation emitted by radioactive isotopes (U, Th and K) and by cosmic rays that penetrate the sediment. Minerals in the sediment, such as quartz and feldspar, have dosimetric properties; thus, the intensity of the accumulated signal reflects both the time passed since solar resetting and the amount of environmental radiation. Two values are required for age determination: the amount of radiation that the mineral has absorbed since it was last reset (the equivalent dose De), and the environmental dose rate.

In this study the infrared stimulated luminescence (IRSL) signal in alkali feldspars was used because it requires only several minutes of exposure to sunlight for complete zeroing of the signal and resetting of the luminescence “clock” (Aitken, 1998; Clarke, 1994). Samples were collected from holes drilled into natural outcrops or trench walls, at a depth of ~30 cm. To prevent exposure to daylight, the samples were immediately put in light-tight bags and all subsequent sample preparation was made under subdued orange light. Field data, laboratory results and calculated ages for the samples from the trenches in Avrona are listed in Table 1.

Fine-sand size alkali feldspars with densities less than 2.58 g/cm³ were extracted from the sediment using routine laboratory procedures (Porat et al., 1997). About 10 mg of the extracted alkali feldspars was deposited on aluminum discs using a silicon spray as adhesive and measurements were carried out on a Risø reader (Bøtter-Jensen et al., 1991). De was determined by the single aliquot added dose protocol (Duller, 1994) roughly following the routine outlined in Porat et al. (1996).

• from Zilberman et al. (2005:84-85)

External γ dose rates were measured in the field using a portable calibrated γ scintillator while external α and β dose rates were calculated from the concentrations of the radioelements. Present-day water contents in the sediment range from 1% to 7% (Table 1). However, these values do not reflect water contents throughout the sediments’ burial history. Pedogenic analysis indicates that the water table at the Avrona playa was at the surface several times in the past 10,000 years and that in general the sediments were water-saturated for most of their history. Therefore laboratory experiments were conducted on several sediment types to estimate the saturation values, which are in the range of 20–30%. This range is assumed to represent past water content and ages were calculated using both the current water contents and the laboratory saturation value of 25 ± 5% (Table 1). The ages presented on the trench logs and soil profiles are those calculated using the saturation water contents.

• from Zilberman et al. (2005:85)

A sediment sample was collected from the present-day streambed of Wadi Avrona, to assess the degree of solar resetting of the IRSL signal in a modern environment. This sample had a low De value of 0.28 ± 0.01 Gy. Age calculations using the average dose rate for the Avrona sediments, ~2400 μGy/year, give 115 years. This residual age implies that the ages presented below could be about 115 years too old.

• from Zilberman et al. (2005:85)

The Avrona basin is at present externally drained, its relief is inverted and it does not function as a closed basin with an endoreic fluvial system, as expected in playas. A detailed analysis of the tectonic history–during which the Avrona playa was deformed and changed from a closed basin with an internal drainage system to an open basin–is presented in Amit et al. (1999).

The playa sediments are exposed along a broad elongated deformation belt composed of low, longitudinal, sub-parallel ridges (0.5–4 m high and up to 5 km long). Shallow braided channels flow along and across these ridges (Figs. 2 and 3). Wadi Raham flows from the west, cuts the northern edge of the playa, and then flows southward at its eastern edge along a gradient of 0.1–0.15%, forming steep and deep playa margins. Wadi Avrona, which drains the western margin of the rift, crosses the entire width of the playa and merges into Wadi Raham. The outlet of the Avrona playa is a narrow channel that flows along the eastern margin of the steep (4–5%) alluvial fan of Wadi Roded, which separates the Avrona playa from the Elat sabkha.

Several low ridges composed of sterile playa deposits were trenched and their sedimentary sequences were studied (for details see Amit et al., 1999). The upper part of the sequence is composed of barren, sterile silty–clayey playa deposits, which include a 1.2 m thick saline soil profile (Salorthids, Soil Survey Staff, 1975; or Solontchak, Duchaufour, 1977). This soil is characterized by in-situ pedogenic gypsic nodules, salt efflorescence, and a well-cemented petrosalic horizon at a depth of 1 m. This horizon marks a stage of stability during which the basin was probably closed and had a high water table. Playa sediments at the base of the exposed sections date to the Early Holocene, whereas the upper 2 m of the sequence dates between 690 ± 50 and 1400 ± 100 years BP (Table 1). As a result of tectonic deformation the saline soils and the playa deposits are at the highest topographic position and at present are undergoing intensive erosion (Amit et al., 1999).

• from Zilberman et al. (2005:85-87)

The results and interpretations of the geophysical survey are presented in Fig. 4. The three seismic sections show a complex fault pattern identified as a flower structure, with converging and diverging NNE-trending faults extending for several hundred meters.

Some of the interpreted faults coincide with surface morphostructural elements, such as surface ruptures, lineaments, and small compression structures that permit correlation between the lines. The geophysical survey reveals that the Avrona fault is composed of several distinct sub-parallel branches; these are marked A–E in Fig. 4.

Zone A on line GI-0074 (Figs. 3 and 4) is associated with a NNE-trending morphological feature, which shows a compressional structure that is expressed as an elevated ridge built of uplifted playa deposits. In some places the ridge forms a topographical step up to 4 m high. The western branch of zone A is expressed as a low, elongated depression occupied by braided shallow channels of Wadi Avrona. The northern extension of the western fault of this shear zone displaces the playa surface, forming a sharp, 2 km long, east-facing fault scarp.

Zone B appearing on lines GI-0074 and GI-0065 (Figs. 3 and 4) is expressed by a row of low hills capped by a gypsic unit of groundwater origin, which is now elevated above the western alluvial fans. The eastern flank of the zone is related to an elongated ridge extending south of line GI-0065 and forming a straight lineament about 3 km long and 100 m wide. A shallow depression occurs between the western and the eastern branches of flower structure B. This depression serves as the local base level of the east-flowing streams blocked by the low ridge in the east.

Zone C on line GI-0065 (Figs. 3 and 4) is related to zone C of line GI-0061 and of line GI-0074. Zone C is situated along the most prominent lineament in the Avrona playa. This lineament is about 10 km long. In the southern part of the playa, it forms a low ridge bounded in the west by a short west-facing scarp; in the northern part, it is expressed as an approximately 1-m high east-facing scarp. A clear discontinuity in the reflectors coinciding with the surface scarp occurs in line GI-0061. A trench at this site (T-20, Fig. 5) exposed a normal fault.

Zones D and E (Fig. 3) are expressed as series of low hills composed of alluvial material cemented by gypsum. These hills represent uplifted alluvial fan sediments deposited at the western playa margins.

The results of the high-resolution seismic reflection survey in the Avrona playa area enable us to relate surface structures to subsurface faults, hence confirming the presence of several NNE-trending sub-parallel fault splays. These observations are confirmed also by Ground Penetrating Radar (GPR) imaging, which has been conducted across the Avrona playa to a depth of 25 m below the surface (Basson, 2000). It is possible to consider the deformed belt of the Avrona playa as a positive flower structure, which is typical of strike-slip fault systems (Sylvester, 1988). The surface expression of the positive flower structure changes across the shear zone of the Elat fault as a result of frequent transitions from normal to reverse displacements (Shtivelman et al., 1998; Basson, 2000; Amit et al., 2002).

• from Zilberman et al. (2005:87-89)

• Fig. 3 Aerial photograph showing
  Figure 3
  
  Air photo of the Avrona playa showing the traces of the photogrametric topographic profiles (circles) and the sites of the trenches and exposures (rectangles) described in the text. The straight lineament that bounds and crosses the playa (dark area)represents a NNE-trending fault of the Avrona Shear Zone. This fault is trenched in T-20. The dark area bounded by straight lineaments corresponds to the uplifted sterile playa sediment. The light-tone areas mark the wide braided Wadi Avrona stream, which occupies a structural depression in the north and splits into several narrow and shallow streams within the playa area. The main faults are designated A–E; the profiles are labelled GI-0061, GI-0065 andGI-0074.
  
  click on image to open in a new tab
  
  Zilberman et al (2005)
  locations of Trenches T-20-22 and seismic lines from Zilberman et al. (2005)

The most recent seismic history of the Elat fault was examined by the paleoseismic analysis of the most prominent branch of this shear zone in the Avrona playa (fault C). This fault trace, which is 10 km long and about 1 m high, dissects a transitional vegetated playa zone composed of sand to sandy-loam sediment (Fig. 3; fault C). Based on its clear morphology and the high-resolution seismic reflection analysis, this fault is assumed to be the youngest in the playa area. A trench (trench T-20) excavated at this site exposes a multiple-event fault with normal displacement and one main fault plane (Fig. 5). No lateral displacement was detected on this fault.



Five partially preserved colluvial units were identified adjacent to the fault plane (Fig. 5, units 1-5). This succession of vertically stacked colluvial wedges maintains a constant thickness up to the fault contact and terminates abruptly at the fault plane. The colluvium is gravelly to sandy, depending on the composition of the sedimentary units from which it derived. Pedogenic features such as bioturbation, gypsic nodules, and horizonation were detected in several of the colluvial wedges. The variation in the pedogenic features indicates that each wedge formed during a different time period. The shape of the wedge, the different sedimentological character of each colluvial wedge, the absence of a fining upward trend in the whole colluvial section, and the absence of correlative colluvial units on the upthrown block indicate that they were tectonically induced and not a result of nontectonic slope processes (Amit et al., 1995; McCalpin, 1996).

According to data from unit 3, which overlies the colluvial units and unit 1a at the base of the section, five tectonic events occurred between 14,200 ± 300 and 3700 ± 300 years BP. Each of these events was accompanied by displacements of at least 0.5 m. The deposition of unit 3 marks the beginning of at least 2000 years of quiescence. During this period, playa sediments were deposited on the downthrown block and later also over the up-thrown block (Fig. 5; units 3 and 5). Scour-and-fill sedimentary structures with fine cross-bedding, especially at the bottom of unit 5, indicate a north–south flow direction sub-parallel to the fault trace for a limited period. The deposition of fluvial sediments decreased during the last 2000 years and changed into fluvial-eolian deposition of sand sheets (Fig. 5; unit 5b). A saline soil with a petrosalic horizon developed in the fluvial-eolian deposits in the entire playa area (Fig. 5; unit 5). The soil marks a period of stability with a low rate of sediment deposition and tectonic quiescence. Tectonic activity of this fault was renewed later than 140–800 AD with vertical surface rupture of about 1 m (Fig. 5; Table 2). A small remnant of unit 5b in which a soil with a petrosalic horizon was preserved was left on the down-faulted block, enabling us to measure the amount of displacement of the last tectonic event that occurred on this fault (Fig. 5; unit 5b on the down-faulted block).

• from Zilberman et al. (2005:89)

• Fig. 3 Aerial photograph showing
  Figure 3
  
  Air photo of the Avrona playa showing the traces of the photogrametric topographic profiles (circles) and the sites of the trenches and exposures (rectangles) described in the text. The straight lineament that bounds and crosses the playa (dark area)represents a NNE-trending fault of the Avrona Shear Zone. This fault is trenched in T-20. The dark area bounded by straight lineaments corresponds to the uplifted sterile playa sediment. The light-tone areas mark the wide braided Wadi Avrona stream, which occupies a structural depression in the north and splits into several narrow and shallow streams within the playa area. The main faults are designated A–E; the profiles are labelled GI-0061, GI-0065 andGI-0074.
  
  click on image to open in a new tab
  
  Zilberman et al (2005)
  locations of Trenches T-20-22 and seismic lines from Zilberman et al. (2005)

Two topographic profiles were drawn across the Avrona playa area. One follows the ridges scattered across the elevated playa and the other follows the stream channel of Wadi Avrona (Fig. 6). The profile along the stream channel of Wadi Avrona, which flows sub-parallel to the trace of the Elat fault, can be used as a reference datum for estimating the amount of deformation of the playa surface. The ridges whose topographical elevations were measured are those on which a Salorthid soil was detected. This soil represents the un-eroded playa surface which reached the same degree of development over the whole playa area while the playa was a closed basin with a shallow water table (Amit et al., 1999). The age range of five soil profiles in the playa area located on elevated sub-ridges (Fig. 6) is between 690 ± 50 years BP and 1400 ± 100 years BP, with an average age of about 1000 years (Fig. 6; Tables 1 and 2). With a known age and uniform degree of development this soil can be used as a chronostratigraphic marker in reconstructing the previous undeformed playa surface.

Vertical changes along the ridge topographic profile reflect differential vertical uplift in relation to the Avrona stream channel (Fig. 6). It reaches a height of 3.5–4 m in the central part of the Avrona basin and about 1 m at the basin outlet (segment 9 in Fig. 6) close to the alluvial fan of Wadi Roded. The vertical topographical differences between the stream-channel and the elevated playa ridge decrease from the central part of the ridge northward and southward toward the present outlet of the playa. Such a topographical situation could not have been achieved by fluvial incision but only by the amount and pattern of a tectonic uplift. The maximum incision that can be attributed to later fluvial erosion under such conditions is 1 m, which is measured at the Avrona playa outlet. This leaves a maximum tectonic uplift of about 2.5–3 m.

In summary, the topographic profiles of the channels and the sub-ridges across the Avrona playa show that the maximum amount of vertical uplift along the entire Avrona shear zone ranges between 2.5 and 3 m. The uplift of the playa area is reflected by the present location of the saline soil, which must have formed during a former phase when the playa was a closed basin. As this soil is about 1000 years old, the uplift, which changed the soil location, happened about 1000 years ago.

• from Zilberman et al. (2005:89,92)

• Fig. 3 Aerial photograph showing
  Figure 3
  
  Air photo of the Avrona playa showing the traces of the photogrametric topographic profiles (circles) and the sites of the trenches and exposures (rectangles) described in the text. The straight lineament that bounds and crosses the playa (dark area)represents a NNE-trending fault of the Avrona Shear Zone. This fault is trenched in T-20. The dark area bounded by straight lineaments corresponds to the uplifted sterile playa sediment. The light-tone areas mark the wide braided Wadi Avrona stream, which occupies a structural depression in the north and splits into several narrow and shallow streams within the playa area. The main faults are designated A–E; the profiles are labelled GI-0061, GI-0065 andGI-0074.
  
  click on image to open in a new tab
  
  Zilberman et al (2005)
  locations of Trenches T-20-22 and seismic lines from Zilberman et al. (2005)

A well-preserved Early Islamic ranch dated to the 11th century AD was discovered at the western part of the Avrona playa (Avner, 1993). This ranch, which provided fresh agricultural products to the town of Aila, was irrigated by a network of water canals connected to a central water pool. The water pool was fed by a Qanat system (subsurface tunnel, marked on the surface by a row of vertical shafts), which collected shallow groundwater deriving from alluvial fans located along the western rift margins (Fig. 7A). A roofed water canal (aqueduct) connected the outlet of the Qanat system to the central irrigation pool. The farm was abandoned in the 11th century at the same time that Aila was destroyed by an earthquake because of the destruction of the irrigation system by surface deformation (Avner, 1993).

Topographic profiles of the roofed water canal measured using Electro-Optical Distance Measurements (EDM) revealed that today the inlet of the canal to the pool is about 1 m higher than the Qanat outlet (Fig. 7B). Three points of deformation were detected along the canal: near the outlet of the Qanat, at the middle part of the canal and near its inlet to the water pool. The inversion of the original southward gradient of the roofed canal is clearly related to an uplift of the water pool area by more than 1 m in relation to the Qanat system.

The deformed water system is located a few hundred meters west of the main lineament underlain by the subsurface fault C (Fig. 3) and presumably also underlain by fault D, which was detected further to the northwest of fault C (Fig. 3). The deformation of the irrigation system can be directly related to reactivation of these two branches of the Elat fault during the historic earthquakes that occurred in the 11th century.

• from Zilberman et al. (2005:92-94)

The catalogues of historic earthquakes in the Middle East document only two large seismic events in the southern region of the eastern Mediterranean in the last 1500 years: the strong 1068 AD earthquake and the moderate less destructive 1212 AD earthquake (Amiran, 1951; Ben-Menahem, 1981, 1991; Turcotte and Arieh, 1988; Ambraseys et al., 1994; Amiran et al., 1994). According to the earliest account of the 1212 AD earthquake it was strongest in the part of Aila that was by the sea (Ambraseys et al., 1994; Amiran et al., 1994; Avner, 1993), suggesting that it originated on the eastern fault of the Elat depression extending between Aqaba in the north and Nuwieba in the south (Ben Avraham et al., 1979; Ben Avraham, 1985; Abou Karaki et al., 1983). The 1068 earthquake was more damaging and caused the total destruction of the city of Aila and widespread damage throughout the Middle East, resulting in thousands of casualties within a radius of more than 500 km from the Gulf of Elat–Aqaba including north Hijaz, Sinai, Egypt, and the Mediterranean coast (Fig. 8) (Ambraseys et al., 1994). Reports indicate that it was felt all the way to Baghdad (Ambraseys et al., 1994). This earthquake was also associated with a tsunami at several sites along the Mediterranean coast (Willis, 1928; Amiran, 1951; Amiran et al., 1994; Ben-Menahem, 1977, 1981, 1991; Poirier and Taher, 1980; Arieh et al., 1985; Arieh and Feldman, 1985; Turcotte and Arieh, 1988; Ambraseys et al., 1994). Such an event seems more likely to have caused the severe damage and surface rupture of the playa area than the weaker 1212 AD earthquake.

It was previously demonstrated that a relief inversion of the Avrona playa was triggered by tectonism (Amit et al., 1999). The paleoseismic analysis reveals that at least six M >6 tectonic events have affected the playa during the last 14,000 years. The limiting age of the sedimentary sequence, the extent of the soil development and the longitudinal profiles of Wadi Avrona and Wadi Raham indicate that one of the faulting events of the playa, which was accompanied by compression and uplift, occurred in the last 1000 years. Supportive data for such an event were also obtained from a paleoseismic study in the nearby faulted Wadi Yutim fan complex at the head of the Gulf of Elat–Aqaba (Mansoor and Niemi, 1999). The age of the tectonic event that deformed the playa is more accurately constrained to the 11th century by the archaeological evidence forthcoming from the deformed Islamic irrigation system, which was damaged by an earthquake at the same time that Aila was destroyed. This tectonic event displaced the surface by at least 1 m and changed the morphology of the Avrona playa from a closed system with internal drainage to an open basin, resulting in relief inversion (Amit et al., 1999, 2002). Such amount of displacement is consistent with an earthquake magnitude of M 6.6 (Wells and Coppersmith, 1994).

Using historical documents Ambraseys et al. (1994) located the epicenter of the 1068 earthquake near Tabuk in Saudi Arabia, about 250 km southeast of Aqaba. This places it at a distance of about 300 km from the study area, making it unlikely to have caused the observed surface rupture of the Avrona playa. Measured dimensions of surface faulting detected in worldwide historic and recent earthquakes of various magnitudes revealed that the maximum length of surface rupture in a M_w 7.5 earthquake is 80 km followed by maximum displacement of 3 m (Wells and Coppersmith, 1994). Taking this into account it is clear that Tabuk is too far from the Avrona playa area to have caused any surface ruptures. The only response to this earthquake in the Tabuk area was three new springs, which appeared at a place called al-Qur, and in Taima the ground was split open “to reveal buried treasures of gold and golden jewellery” (Ambraseys et al., 1994). Whereas the DST is a well-known source of strong earthquakes, in the Tabuk area in Saudi Arabia there is no tectonic feature that could have produced such a large magnitude earthquake (Bartov, 1990). The total destruction of Aila further suggests that it was situated in the near-field of the earthquake rupture. Moreover, the damaged sites (the Baniyas area in the north and the Taima area in the southeast) fall within the radius of about 500 km around the proposed surface rupture.

Taking into consideration the sparsely inhabited region one cannot obtain accurate data on natural and artificial damage and the epicentral area of this event must be sought, as proposed by Ambraseys et al. (1994, p. 31), between Aila and Taima. The results of this study suggest that it [JW: the 1068 CE earthquake] was located at the southern Arava area at the Avrona playa and not in the Tabuk area.

• from Zilberman et al. (2005:94)

Several considerations indicate that the earthquake that deformed the Avrona playa was of very large magnitude and that it was caused by the Elat fault. Using the maximum displacement parameter (Bonilla et al., 1984; Wells and Coppersmith, 1994), the vertical displacement of 1 m measured for the youngest fault in trench T-20 could have been produced by an earthquake with a magnitude of about M 6.6. A similar displacement was measured in the deformed irrigation canal in the Early Islamic ranch, which is underlain by fault D (Fig. 3) and was also supported by GPR measurements (Basson, 2000). However, this amount of displacement may represent a minimum in the area, as shown by the topographic profiles of the elevated playa ridges. These represent the entire deformation caused by the last earthquake in the southern Arava valley, and show an uplift of 2.5 to 3 m along a deformation belt, which is 10 km long and 0.5–0.2 km wide. Such uplift must correspond to an earthquake with a magnitude higher than M 6.6, which was measured at trench T-20. The surface rupture length cannot be determined in the playa area because of erosional obscuring, burial by younger deposits of the playa and fragmentation of the continuous fault line by incision of the alluvial fan. If the full 50 km length of the fault zone is assumed to have been active during the studied event, the maximum magnitude is expected to be M 7 (Wells and Coppersmith, 1994).

In the Elat fault zone there are no measurable geomorphic or stratigraphic markers for determining lateral slip rates. As a result, only vertical displacement can be used for slip rate evaluations. The average vertical slip rate of the normal faults is 0.2 mm/year (Amit et al., 2002). This value changes through time along different segments of the active shear zone and among clusters of events related to the same segment, with a range of 0.1 to 0.35 mm/year. The value calculated for the Avrona playa fits the higher value and reaches 0.3 mm/year (Amit et al., 2002). However, such average slip rate range is low in comparison to the worldwide vertical slip rate for normal faults, which vary from <0.5 mm/year to >3 mm/year (Yeats et al., 1997). The low vertical slip rate range can emphasize both the complexity of the Elat fault zone, which is composed of a large number of fault branches and their dominant strike-slip character. The fault pattern of this segment is different from the pattern of the northern Arava segment. The latter is characterized by a narrower fault zone with clear and measurable lateral components of motion. Estimates of its lateral slip rate range between 4 ± 2 mm/year and 4.7 ± 1.3 mm/year (Klinger et al., 2000b; Niemi et al., 2001).

• from Porat et al. (2009:37)

Quartz from the T-18 samples was extracted from the fraction remaining after heavy liquid separation previously used to isolate the KF. Heavy minerals and any remaining insoluble carbonates were removed by magnetic separation (Porat, 2006) and the non-magnetic fraction etched with 40% HF. Luminescence measurements were carried out on Risø DA-12 TL/OSL readers either at the Geological Survey of Israel (GSI) or at the luminescence research laboratory at the University of Wales, Aberystwyth (UWA). Table 1 gives the field and laboratory dosimetry data for the samples.

• from Porat et al. (2009:37)

The IRSL measurements presented in Amit et al. (2002) were made using standard aliquots which contain approximately 1000–2000 mineral grains. These had revealed some significant scatter in the De values obtained from replicate aliquots of certain samples (Fig. 3a). To investigate this scatter further, two samples were selected, one from the heel (18-10) and one from the toe (18-6) of the third colluvial wedge from the top, and 24 small aliquots were measured, each aliquot containing 150–200 grains. Measurements were at the GSI, and the SAR protocol was followed Wallinga et al. (2000). Preheats of 10 s ranged between 230 and 290 1C and a cut heat of 5 s at 210 1C was used after giving the test dose. The IRSL signal was measured at 50 1C and to reduce buildup of signal background, at the end of each measurement cycle the discs were bleached by IR for 100 s at 280 1C to background level.

An example of the IRSL signal and dose–response curve for an aliquot of sample 18-6 is shown in Fig. 4a and b. These small aliquot KF measurements have a recycling ratio within 10% of unity, indicating that any sensitivity change was being corrected for, and although there is large scatter in the De values, there is no trend in De as a function of preheat temperature (Fig. 4c). For both samples, a wide scatter in De values was observed, with a skewed distribution towards smaller De values (Fig. 3c). This pattern is what would be expected from a sediment that had not been completely bleached at deposition, and consisted of a mixture of bleached and unbleached grains (Porat et al., 1997, 2001; Olley et al., 1998; Prescott et al., 2007).

• from Porat et al. (2009:37-38)

In this sedimentary environment, where exposure to daylight may be too short for resetting the IRSL or OSL signals, it is important to identify the most light-sensitive signal. Bleaching experiments were carried out to determine how fast the IRSL signal from KF and the OSL signal from quartz are bleached by sunlight. Ten aliquots each of KF and quartz from two samples were placed in direct sunlight for periods ranging from 2 s to 10 min. IRSL was measured at 50 1C and OSL at 125 1C. Normalization was performed by measuring the natural IRSL and OSL signals for 0.1 and 0.2 s, respectively, prior to bleaching. After bleaching the KF IRSL signal was measured for 5 s, the discs were then preheated for 10 s at 280 1C and the remaining IRSL measured to its full depletion. Quartz OSL signal was measured for 3 s, the discs then preheated for 10 s at 260 1C and the remaining OSL signal measured for a further 130 s.

The results of the bleaching experiments (Fig. 5) show that the IRSL signal of KF bleaches at a slower rate than the OSL signal of quartz; the time required for the IRSL and OSL signals to decay to 10% of their original intensity is 50–80 s and 2–3 s, respectively. The signal measured after preheat was bleached at roughly the same rate as the non-preheated signal. During the short sediment transport along the fault scarp and into the colluvial wedge, quartz would be better bleached than the KF, and therefore further work focused on measurements of quartz.

• from Porat et al. (2009:38-39)

The samples from the colluvial wedges and three samples from the alluvial terraces were measured using the SAR protocol (Murray and Wintle, 2000). About 5 mg of the purified quartz was deposited on 10 mm diameter aluminum discs using silicone spray as an adhesive. Measurements were carried out either at the GSI using a green-filtered halogen bulb for stimulation or at UWA using blue LED for stimulation, and signal detection was through 7 mm U-340 filters. De values were determined on 12–24 aliquots, preheat temperatures ranged between 180 and 280 1C and the OSL signal was measured at 125 1C. Test dose was ca. 5 Gy and a cut heat of 5 s at 160 1C was used to remove unstable signals. Fig. 4d and e show a typical OSL shine down curve and growth curve for one aliquot. Recycling ratios were within 10% of unity, indicating that any sensitivity change was being appropriately corrected for, and all aliquots had an IR OSL depletion ratio (Duller, 2003) within 10% of unity, suggesting that the signal was dominated by quartz. Laboratory given doses could be recovered within 90–95%.

Given the wide scatter observed in the KF small aliquot data, it is not surprising that there was scatter in the De values observed from individual aliquots of a sample (e.g. Fig. 3b). Fig. 4f shows the variation in De as a function of preheat temperature, and in spite of the variability between aliquots, there is no trend with preheat. Thus, the luminescence behavior of the quartz appears to be satisfactory, and the variability in the De is thought to derive from incomplete bleaching. To investigate this further required analysis of single grains.

• from Porat et al. (2009:39-40)

Single grains from nine samples, six from the colluvial wedges and three from the terraces, were measured. De values were measured for 400–700 grains from each sample at the UWA on a DA-15 Single Grain Risø reader (Duller et al., 1999). The SAR protocol was used to measure the De, with measurement conditions following those of Jacobs et al. (2003). To reduce recuperation and buildup of signal background, at the end of each measurement cycle the samples were heated to 280 1C and the OSL signals were measured at this elevated temperature for 100 s to background level (Murray and Wintle, 2003). Only grains with monotonically growing dose–response curves, recycling ratios within 30% of unity, an IR OSL depletion ratio within 30% of unity (Duller, 2003) and errors on the De of less than 30% were selected. Ages were calculated using the minimum age model of Galbraith et al. (1999), as it is most appropriate for insufficiently bleached sediments (Bailey and Arnold, 2006).