Constraining regional paleo peak ground acceleration from back analysis of prehistoric landslides: Example from Sea of Galilee, Dead Sea transform

Tectonophysics 490 (2010) 81–92

Contents lists available at ScienceDirect

Tectonophysics j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t e c t o

Constraining regional paleo peak ground acceleration from back analysis of prehistoric landslides: Example from Sea of Galilee, Dead Sea transform Gony Yagoda-Biran a, Yossef H. Hatzor a,⁎, Rivka Amit b, Oded Katz b a b

Ben-Gurion University of the Negev, Department of Geological and Environmental Sciences, Be'er Sheva 84105, Israel Geological Survey of Israel, 30 Malkhe Israel St., Jerusalem 95501, Israel

a r t i c l e

i n f o

Article history: Received 8 November 2009 Received in revised form 24 March 2010 Accepted 19 April 2010 Available online 24 April 2010 Keywords: Earthquakes Landslides PGA Back analysis Limit equilibrium Seismic hazard

a b s t r a c t Accurate estimation of expected peak ground acceleration (PGA) in seismically active regions is a challenging task. The best way to estimate, quantitatively, expected PGA is by investigating instrumental data of past strong earthquakes in a given area. In some regions of the world however recorded data are scarce, and if they exist, they are typically available only since the late 19th century. As such they are hardly representative of the true seismicity in the studied region. We propose here an analytical approach to constrain the lower threshold of paleoseismic PGA on the basis of back analysis of old landslides. To perform the analysis we need a mapped landslide with geomorphic features that have been preserved in the field, the slip surface, a good reconstruction of the slope geometry and ground water level prior to failure, and the mechanical properties of the sheared material. We perform static and pseudo-static limit equilibrium analyses using standard solution procedures to obtain lower bounds of paleoseismic PGA. Back analyses of three different landslides around the Sea of Galilee (SOG) return similar results that range between 0.15 and 0.5 g, thus constraining the threshold paleoseismic PGA range for this region. The analytically inferred regional PGA is supported by results of an independent numerical analysis of toppled columns in a nearby Byzantine church. Using results from a recent paleoseismic trenching study performed on one of the studied landslides and a modified attenuation relationship for the study area we localize the loci of moment magnitude Mw = 7.0 earthquakes that can explain the studied failures along the boundaries of the SOG, and find that they coincide with traces of the Eastern and Western Margin faults of the Dead Sea transform. The temporal relationships between the observed failures are discussed on the basis of dated colluvial sediments, geomorphologic constraints, and archeological evidence. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Accurate estimation of expected peak ground acceleration (PGA) in seismically active regions is crucial for risk preparedness and sound engineering design. The best way to estimate quantitatively expected PGA would be to investigate long term recorded data of past strong earthquakes in the studied region. In some regions of the world however recorded data are scarce due to lack of seismic network infrastructure, and in all regions the availability of recorded data is restricted to the late 19th century. Therefore, existing instrumental data are hardly representative of the true seismicity of a region. When recorded data are scarce or not available, alternative methods may be applied, for example adopting a quantitative paleoseismic approach. In this paper we suggest a new and relatively simple paleoseismic approach for estimating a characteristic paleo PGA for a specific ⁎ Corresponding author. Dept. of Geological and Environmental Sciences, Ben-Gurion University of the Negev, Be'er Sheva 84105, Israel. Tel.: +972 8 6472621; fax: +972 8 6472997. E-mail address: [email protected] (Y.H. Hatzor). 0040-1951/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.tecto.2010.04.029

region, through back analysis of seismically triggered landslides. By back analysis, in the context of this paper, we mean that we seek a PGA value that will be sufficient to bring the modeled slope to a state of limiting equilibrium. The obtained results, coupled with qualitative intensity scale classification for the region, for example using the ESI 2007 scale (Guerrieri and Vittori, 2007), provide a useful seismic hazard assessment for a given region. Using intensity levels alone does not provide the valuable information regarding expected ground motions, as scaled by the expected PGA, necessary for seismic engineering design. The approach is demonstrated through the study of three landslides triggered around the circumference of the Sea of Galilee (SOG) during the Pleistocene. The SOG is a rhomb-shaped graben formed due to left segmentation of the sinister Dead Sea transform (DST) fault (Garfunkel, 1981; Hurwitz et al., 2002). The eastern and south-western margins of the SOG are normal faults associated with the DST system, whereas the northern and north-western margins exhibit a more complicated structure (Ben-Avraham et al., 1996). The stratigraphy of the SOG region largely consists of Miocene to Pleistocene lacustrine and fluvial

82

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

sediments with episodic events of lava flows, the entire sequence of which rests on an Eocene basement. The lithology consists of clay, marl, chalk, limestone, sandstones and conglomerates, capped by the Plio-Pleistocene Cover Basalt formation (Mor and Sneh, 1996). Seismicity in the SOG region is moderate (Arieh and Rabinowitz, 1989; Shapira, 1983), with an estimated Mw = 6 earthquake reoccurrence time interval in the order of 102 years increasing to 103 years for Mw = 7 earthquakes (Begin, 2005). The last strong event (Ms = 6.2) was recorded at the northern Dead Sea area in 1927 (Enzel et al., 1997; Niemi and Ben-Avraham, 1997; Shapira et al., 1993), about 100 km south of SOG. In 1973 a ML = 4.5 earthquake occurred a few kilometers south of the SOG (Arieh et al., 1977). According to the national seismic building code (S.I.I., 2004) a horizontal PGA of 0.3 g is estimated for this region with 10% probability of exceedence for every 50 years. The seismicity of this region and the population density around the SOG warrant such an attempt. The three analyzed landslides presented in this paper are Ein-Gev, Berniki Beach, and the Fishing Dock (Fig. 1). All three landslides are analyzed here using the same approach, but with different qualities of input data, depending upon their availability and accessibility in each case. We argue that the fact that not all data are available in each case is not necessarily a weakness of our approach. It proves, rather, that a regional PGA may be reasonably estimated from back analysis of several landslides even when the required data set is incomplete. The PGA for the SOG region thus estimated can be used to improve the existing building code and consequently reduce the seismic hazard in the next large earthquake expected in this region. Finally, landslides near major faults like the ones existing in the study area are often induced by coseismic fracturing, and fractures that are kept open by surface faulting; the influence of late Pleistocene

tectonic deformation and fault displacement on the slope stability near the Sea of Galilee is discussed by (Katz et al., in press). 2. Qualitative and quantitative paleoseismic approaches Various geological features have been used recently to deduce paleoseismic information for the DST. Seismites, defined as disturbed sedimentary structures, have been used to estimate earthquake recurrence intervals in the DST (Ken-Tor et al., 2001; Marco and Agnon, 1995, 2005; Migowski et al., 2004). Such methods provide reasonably accurate chronology of seismic events but no quantitative measure of the ground motion at the site that must have caused the mapped damage in the field. Trenching faults that are displacing the ground surface has been used as a methodology to obtain both earthquake chronology as well as quantitative measures of fault displacement during strong earthquakes (Amit et al., 2002; Marco et al., 2005). Stalactites and stalagmites can also be used as paleoseismic indicators (Kagan et al., 2005). Using this approach regional earthquake chronology may be obtained as well as recurrence interval, and in some cases upper constrains on paleo PGA can be evaluated (Szeidovitz et al., 2008). Another paleoseismic approach suggested by Matmon et al. (2005), utilizes soil dating below toppled rock blocks with cosmogenic isotopes and optically stimulated luminescence (OSL) to deduce recurrence intervals of earthquakes in the Timna valley, southern DST. The paleoseismic methods discussed thus far enable direct determination of earthquake chronology and indirect estimate of earthquake intensity, yet a quantitative assessment of paleo PGA is difficult

Fig. 1. (a) Location map of the study area based on a DEM of the Sea of Galilee region (shaded relief from Hall (1994)). Active faults are yellow (EMF = Eastern Margin Fault, WMF = Western Margin Fault, JGF = Jordan Gorge Fault). Studied landslides marked by grey circles (EGLS = Ein-Gev landslide; BBLS = Berniki Beach landslide; FDLS = Fishing Dock landslide). Also shown in grey triangles are the landslides of Umm El Qanatir and Almagor. Inset (b) shows the plate tectonic setting of the Dead Sea Transform (DST).

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

to obtain accurately from these paleoseismic indicators. To overcome this shortcoming, Brune (1996) introduced a method to constrain paleo PGA by performing a two dimensional pseudo-static stability analysis of precariously balanced rocks. The method introduced by Brune (1996) does provide an upper boundary of paleo PGA for the studied regions. Ancient earthquakes often leave evidence of damage in archeological sites and historic masonry structures (Ellenblum et al., 1998; Wechsler et al., 2009). Such observations provide strong chronological constraints on the causative earthquakes due to accurate archeological dating, but very limited information on the magnitude of the resulting ground motion. With the advent of numerical tools over the past two decades it has become possible to perform a fully dynamic analysis in the time domain both for continuous as well as discontinuous media using the finite element (e.g. Zienkiewicz and Taylor, 2000), finite difference (e.g. Itasca, 1993), and the distinct element (Cundall, 1988) methods. Recently Kamai and Hatzor (2008) introduced a novel approach to assess paleo PGA from inversion of mapped key block displacements in Roman and Byzantine arches embedded in historic masonry structures. Yagoda-Biran and Hatzor (2010) deduced paleo PGA from back analysis of collapsed columns in a Byzantine church at the Susita archeological site, located along the DST on the eastern margins of the SOG using a fully dynamic numerical approach. Landslides can also be used as paleoseismic indicators because the paleo PGA that triggered the slide may be constrained with application of an appropriate limit equilibrium analysis (LEA). Typically, a pseudo-static approach is adopted for back analysis to obtain the magnitude of horizontal PGA required for limiting equilibrium, provided that the original geometry of the sliding mass and the shear strength of the sliding surface are known. This approach was applied in the past for estimating threshold PGA of earthquakes suspected to induce landslides in the vicinity of the SOG: Harash and Bar (1988) and Wechsler et al. (2009) performed back analyses of seismically induced landslides a few kilometers north and east of the SOG, respectively. Longpre et al. (2008) used a similar approach in order to study the stability of a possibly recent landslide in Spain. In addition to constraining PGA, back analysis of ancient landslides may be used to obtain other paleoseismic parameters. Strasser et al. (2006) identified three large earthquakes using temporal and spatial correlation of multiple subaqueous landslide deposits through highresolution seismic surveys, radiocarbon-dating of cores, and application of empirical seismic attenuation models to obtain earthquake chronology, magnitudes, and epicenters. Shou and Wang (2003) conducted a back analysis for the Chiufengershan landslide triggered by the 1999 Chi-Chi earthquake in Taiwan to determine the failure mechanism, and then applied forward analysis of the residual slope to obtain its failure risk. Urgeles et al. (2006) performed a back analysis of a large submarine landslide for mechanism determination. Finally, landslides as a paleoseismic tool are thoroughly discussed by Jibson (1996). 3. Research methods 3.1. Limit equilibrium analysis The analytical approach for back analysis of slope failure adopted in this study is the limit equilibrium analysis (LEA) method. Specifically, for determination of slope stability we use the method of slices, as explained below. The LEA approach, a well known and commonly used method in geomechanics, introduces the concept of Factor of Safety (FS), namely the ratio between the stabilizing and mobilizing forces or moments. A FS of unity indicates limiting equilibrium, whereas a FS greater or smaller than 1.0 indicates stability or instability of the slope, respectively. In the case of a landslide the stabilizing forces and moments are provided by the frictional forces

83

between the sliding mass and the undeformed slope, whereas the mobilizing forces and moments are provided by the weight of the sliding mass, water forces on the boundaries of the sliding mass, and in the case of earthquakes — inertia forces. In a pseudo-static analysis the inertia forces are taken as static forces applied at the centroid of the sliding mass. The main advantage of the LEA method is that it is a straight forward approach, easy to formulate and solve. Yet, for LEA the sliding surface geometry must be known in advanced. Furthermore, the solution is relevant only for the stage of incipient failure and does not provide information on the rate and amount of displacement in case of failure. The method of slices is typically used in slope stability analysis for obtaining higher accuracy by discretizing the sliding mass into a finite number of vertical slices, where each is considered as a separate sliding mass. The forces acting on each individual slice are its weight, friction at the base of the slice, water pressures on the boundaries of the slide (if exist), inertia forces (if exist), and inter-slice normal and shear forces. Several methods of slices were developed over the years, differing in the way inter-slice forces are considered, and the equilibrium criteria they satisfy. Some methods satisfy only force equilibrium (Corps of Engineers, 1982; Janbu, 1954; Lowe and Karafiath, 1960), moment equilibrium (e.g. Bishop, 1955; Fellinius, 1936) or both (e.g. Bishop, 1955; Morgenstern and Price, 1965; Sarma, 1973; Spencer, 1967). We employ the commercial software package GEO-STUDIO with the slope stability software SLOPE/W 2004 (Krahn, 2004) for back analysis of the landslides discussed in this paper, where the geometry of the slope profile, ground water table (GWT) level, material properties, and field based determination of the location and geometry of the failure surface are used as input parameters. The method of solution we use is the Morgenstern and Price (1965) method. 3.2. Determination of shear strength parameters The three studied landslides are all located around the SOG (see Fig. 1). The Ein-Gev landslide (EGLS; Fig. 1) is the one studied most comprehensively since the identification, geological mapping, and experimental determination of mechanical parameters were all performed by the authors. Six drained direct shear tests were performed on prismatic, undisturbed, samples from the Ein-Gev landslide slide at their natural water content. The tested samples were cast into a shear box with inside dimensions of 15 × 15 × 30 cm. Direct shear tests were performed in a hydraulic, closed loop servo controlled load frame under an imposed constant normal stress level and under imposed constant shear displacement rate, utilizing the two independent servo control systems for the normal and horizontal shear pistons, respectively. The six levels of constant normal stress in each of the six tests were: 210, 275, 415, 550, 690 and 830 kPa, and the sliding velocity was 0.0254 mm s− 1 in all tests. The tests were performed on initially intact samples so that both peak and residual shear strength values could be obtained for each test. The Fishing Dock landslide (FDLS; Fig. 1) was initially analyzed by Keisar and Peferbaum (2001) and later more laboratory test results were reported by Saltzman (2005). In the analysis we use Keisar and Peferbaum's (2001) geological mapping and the material properties provided by Saltzman (2005). The Berniki Beach landslide (BBLS; Fig. 1) is the least studied, and no laboratory test data are available. Therefore the back analysis performed here is coupled with sensitivity analyses for mechanical parameters. 4. Results The three landslides are back analyzed to deduce a threshold of paleo PGA for each, and to arrive at some conclusions regarding a possible paleo PGA range for the SOG region.

84

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

4.1. The Ein-Gev landslide The Ein-Gev landslide is located along the south-eastern margins of the SOG (EGLS; Fig. 1). The landslide is 1500 m long, 200 m wide, and the elevation difference from scarp to toe is 500 m (Fig. 2). The landslide must have been triggered after the current topography of the SOG marginal slopes was defined, namely post the PlioPleistocene Cover Basalt formation. The slide area was mapped

geologically and morphologically to a scale of 1:10,000. The scarp and toe were identified on the basis of field mapping and aerial photos. The geological unit through which the slip surface developed, the Miocene Ein-Gev Sandstone formation, was identified in the field, and undisturbed samples were taken for physical and mechanical testing at the BGU rock mechanics laboratory. The average density of the sheared material is 2200 kg m− 3. Direct shear test results obey the Coulomb Mohr shear strength criterion

Fig. 2. Top — Aerial photo of the Ein-Gev landslide (encircled); the bounding intact slopes to the north and south of the landslide, and the shoreline of the SOG to the west. Middle — superposition of the slumped slope geometry (dashed line) on the southern, undeformed, slope (solid line). Bottom — cross section of the modeled landslide. Legend: sliding mass — diagonal pattern, GWT — dashed line, material defined by Ein-Gev sandstone properties — light grey, Cover Basalt — dark grey.

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

with cohesion and friction angle of c = 376, 0 kPa, ϕ = 43°, 38°, for peak and residual shear strength, respectively. Comprehensive summary of shear test results is provided by Yagoda-Biran (2008). Back analysis is performed for undeformed slope geometry, as inferred from a virgin profile of a slope bounding the landslide in the south (see Fig. 2). The location and geometry of the sliding surface are inferred from field mapping and from superposition of the slumped slope geometry on the southern, undeformed, slope (see middle panel in Fig. 2). Note that while some material is missing at the central area of the slide profile, there is excess material at the toe area. This observation led to the reconstruction of the landslide as shown in the bottom panel of Fig. 2. GWT level is inferred from the location of springs at the top of the slope and from the location of the shoreline at its toe. Results of static analysis indicate that the slope is safe against sliding under gravitational loading both for a relatively dry slope with GWT location as in Fig. 2, as well as for a fully saturated slope with GWT level coinciding with the ground surface. The results of a sensitivity analysis are presented in Fig. 3 for the analyzed slope for FS vs. different values of shear strength parameters, where as mentioned above, FS = 1.0 indicates limiting equilibrium. Inspection of Fig. 3 clearly reveals that the slope is safe against failure as long as the friction angle is greater than 18°. Since the sheared geological unit is comprised of sandstones, that typically exhibit a minimum value of friction angle of 32°, it may be safely assumed that the slope is safe under static conditions. The analytical results, obtained with the Morgenstern–Price method of slices, are supported by laboratory test results. The peak and residual shear strength parameters obtained experimentally for the Ein-Gev Sandstone Fm. are plotted in Fig. 3 as solid and open squares, respectively. With peak shear strength the FS against failure is 4.8 and with residual shear strength the FS is 2.5. Interestingly, even in the case of a fully saturated slope with the GWT level coinciding with the ground surface, the slope is safe both with peak as well as residual shear strength parameters (solid and open triangles respectively in Fig. 3). The same strength values were used for both GWT location scenarios. The results of the static analyses imply that an external force had to act on the slope to trigger sliding. The most suitable candidate would be a pseudo-static earthquake inertia force, the analysis of which is discussed below.

Fig. 3. Results of static sensitivity analyses for Ein-Gev landslide. With cohesion of 100 kPa the landslide is statically stable for any friction angle value. Limiting equilibrium (FS = 1.0) is marked by dashed horizontal line. Legend: squares — results of static analysis with GWT as in Fig. 2, triangles — results of static analysis for saturated slope, solid symbols — experimentally obtained peak shear strength, open symbols — experimentally obtained residual shear strength.

85

Pseudo-static analyses are performed with peak and residual shear strength parameters as above for GWT configuration as in Fig. 2 and the results are presented in Fig. 4 showing FS vs. threshold paleo PGA required for sliding for both peak and residual shear strength. The results suggest that a horizontal paleo PGA of 0.95 g is required for failure when the material possesses peak shear strength (solid squares), yet with residual shear strength parameters (open squares) a paleo PGA of 0.37 g would be sufficient to trigger the studied landslide. 4.2. The Berniki Beach landslide The Berniki Beach landslide (Fig. 5) is located at the western side of the SOG just opposite the Ein-Gev landslide (BBLS; Fig. 1). The length of the slide is 1000 m, and the elevation difference between the scarp and toe is 200 m. The rock slope comprised the Miocene Hordos Formation, a very inhomogeneous rock formation consisting mainly of sandstones, marls, and conglomerates. We assume the landslide was triggered after the current topography of the SOG margin slopes have been defined, i.e. post the Plio-Pleistocene Cover Basalt Formation. As in the case of the Ein-Gev landslide the original slope geometry prior to failure, is assumed to be identical to an adjacent, undeformed slope, in this case immediately to the south of the landslide (see top panel in Fig. 5). Again, superposition of the slumped slope on the southern undeformed slope topography (middle panel, Fig. 5) led to the reconstruction of the landslide for back analysis (bottom panel, Fig. 5). Finally, the GWT level is not relevant to the analysis as it is positioned below the reconstructed failure surface. Due to the inhomogeneity of the Hordos Fm. composing most of the section it is very difficult to obtain reliable laboratory test data. Nevertheless, we attempt to constrain paleo PGA from back analysis of the Berniki Beach landslide using sensitivity analyses with respect to the unknown shear strength parameters of the material, namely cohesion and friction angle. We begin with a static analysis using the modeled profile presented in Fig. 5 and the Morgenstern–Price method of slices, as before. The results of static analyses for a reasonable range of cohesion and friction angles for the sliding rock mass are presented in Fig. 6. The slope is obviously stable statically for a reasonable range of shear strength values; therefore an additional external force required to trigger the landslide is assumed below. We perform a pseudo-static analysis for the modeled slide geometry with additional inertia force acting horizontally on the sliding mass. Since the material properties are not known we seek a solution for a FS = 1.0 for three different levels of paleo PGA between

Fig. 4. Results of pseudo-static analysis of Ein-Gev landslide. Limiting equilibrium marked by horizontal dashed line. Legend: required threshold PGA for sliding with peak (solid squares) and residual (open squares) shear strength values.

86

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Fig. 5. The Berniki Beach landslide: Top — aerial photo of the landslide, middle — superposition of the slumped slope geometry (dashed line) on the southern, undeformed, slope (solid line). Bottom — the modeled landslide. Legend: Shaded — Hordos Fm., Diagonal pattern — sliding mass, dashed line — recent GWT.

0.3 and 0.4 g. The results are plotted in Fig. 7 in cohesion–friction angle space. The three values of PGA selected for the analysis return failure for a material with cohesion ranging from 0 to 140 kPa, and friction angle between 15° and 35°.

The experimentally obtained residual shear strength parameters for the similar material at the opposite eastern margin of the SOG (the Ein-Gev sandstone) tested in drained direct shear at natural water content are zero cohesion and 38° friction angle. These values are well

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Fig. 6. Static sensitivity stability analysis of the Berniki Beach landslide to different c–ϕ combinations. Horizontal dashed line — stability border.

within the range of the constrained shear strength parameters obtained here for the Berniki Beach landslide, and are considered by us as representative of Miocene fluvial sediments in the region. 4.3. The Fishing Dock landslide The ancient Fishing Dock landslide is located 2 km north to the city of Tiberius (FDLS; Fig. 1). Recent rejuvenation of the slide prompted the current research of this landslide. The Fishing Dock landslide is small compared to the Ein-Gev and Berniki Beach landslides, with length of 250 m and an elevation difference of 80 m between scarp and toe. The sliding rock mass is comprised of the Pliocene Bira Formation consisting largely of marls, chalks, and to a lesser extent clays, overlain unconformably by the Cover Basalt Formation (Saltzman, 1964). The geometry of the modeled slope as presented in Fig. 8 is based on the profile suggested by Keisar and Peferbaum (2001) who performed a preliminary analysis of the slide after it was rejuvenated. The mechanical parameters for the sheared material, mostly marls and chalks, used in our analysis are based on laboratory test results

87

performed by various commercial laboratories in Israel, as reported by Saltzman (2005). Both consolidated drained (CD) direct shear and consolidated undrained (CU) triaxial shear tests were performed. CD tests at natural water content yield cohesion of 400 kPa and friction angle of 46°. CD tests for saturated samples yield cohesion of 35 kPa and friction angle of 42°. CU tests of saturated samples yield cohesion of 86 kPa and friction angle of 33°. Shear strength parameters obtained from CD tests provide the ‘effective stress’ failure envelope for the material and represent shear strength under a very slow loading rate during which excess pore pressure cannot be generated as it dissipates during shear. These loading conditions correspond to long term, static loading of natural slopes. The CU tests provide the ‘total stress’ failure envelope for the material and represent shear strength that would be obtained under rapid loading during which excess pore pressure may not have sufficient time to dissipate. These shear strength parameters are therefore relevant to rapid shear loading conditions that may develop during strong earthquakes. Since the groundwater regime and condition of the material at time of failure are not known, the slope is analyzed for different possible scenarios to constrain threshold horizontal PGA required for failure, as follows: Scenario A — shear strength for samples at natural water content (CD tests). Scenario B — shear strength for saturated samples (CD tests): 1) Recent GWT level (as in Fig. 5), 2) GWT level at the surface. Scenario C — shear strength for saturated samples (CU tests): 1) Recent GWT level (as in Fig. 5), 2) GWT level at the surface. Results of both the static and pseudo-static analyses are presented in Table 1. Results of static LEA indicate that the slope is stable for every GWT scenario; therefore failure of the slope can only be explained by considering an additional external load as in the two previous cases. Assuming a pseudo-static, horizontal, inertia force as the additional external load required to trigger sliding, we find that in most scenarios analyzed a relatively low level of horizontal PGA is sufficient to induce sliding. Since CD tests are relevant for slow shearing as explained above, we only consider the results obtained with CU tests where the loading conditions at the lab simulate rapid shear loading expected to develop during earthquakes. The LEA based on CU test results suggest a paleo PGA range between 0.15 and 0.5 g for GWT position at ground surface or as in recent times, respectively. 5. Discussion 5.1. Inferred PGA for the SOG region from paleoseismic studies

Fig. 7. Sensitivity pseudo-static limiting equilibrium analysis of the Berniki Beach landslide modeled slope. All symbols indicate FS of unity under different c–ϕ combinations, at different PGA levels.

We have shown in this paper how paleo PGA values may be constrained from back analysis of ancient landslides, triggered in three different rock slopes around the Sea of Galilee. The value obtained from back analysis of the Ein-Gev landslide in the south-eastern slopes of the SOG is 0.95 g and 0.37 g for peak and residual shear strength, respectively. The obtained threshold PGA with peak shear strength parameters is very high and cannot be supported by current ground motion measurements (Hofstetter et al., 2003; Zaslavsky and Shapira, 2000) or predictions (Shapira, 1983, 2002). It is reasonable to assume, however, that the material has reached residual conditions by the time of the major sliding event, perhaps due to several ancient ground motions which must have been sufficiently strong to overcome the peak shear strength of the material. This assumption is supported by other studies (e.g. Palmer and Rice, 1973; Skempton and Petley, 1967), suggesting that residual strength in large landslides may be reached after very small displacements. It is plausible to assume therefore that a single or a

88

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Fig. 8. Cross section through the Fishing Dock landslide. Legend: sliding mass —diagonal pattern, sheared material — light grey, Cover Basalt — dark grey, recent GWT — dashed line.

Table 1 Results of both static and pseudo-static limit equilibrium analyses for the Fishing Dock landslide. GWT level scenario

Material conditions

Test type

Static FS

Pseudo-static PGA (g)

Recent Recent At ground surface Recent At ground surface

Natural water content Saturated Saturated Saturated Saturated

CD CD CD CU CU

4.9 2.8 1.52 2.3 1.35

1.61 b 0.85 0.28 0.5 0.15

sequence of strong tremors caused small displacements which were sufficient to reduce the available shear strength to residual values, yet keeping the body of the slide in place. Following this stage, of unknown date and duration, a tremor causing horizontal PGA of 0.37 g at the site would be sufficient to trigger sliding of the entire slope. Therefore the value adopted from back analysis of the Ein-Gev landslide is 0.37 g. The results obtained from back analysis of the Berniki Beach landslide in the south-west slope of the SOG suggest that with a material possessing characteristic shear strength parameters for

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Miocene fluvial sediments, the slope would fail if subjected to a PGA of 0.3–0.4 g. The results obtained from back analysis of the Fishing Dock landslide in the western slopes of the SOG return a range of PGA values between 0.15 and 0.5 g, for GWT position at ground surface or as in recent times, respectively. These values nicely envelope the single value obtained for the Ein-Gev landslide, 0.37 g. Note however that the scenario of GWT coinciding with the ground surface is very severe, as it implies a rainstorm that has saturated the entire rock section all the way up to the surface during the assumed earthquake. We therefore consider this scenario less likely than assuming GWT position as in recent times. In summary, the results obtained from back analysis of three different landslides, located in three different rock slopes around the SOG, suggest a relatively narrow range of possible threshold values for the triggering PGA, between 0.15 and 0.5 g. A similar study based on pseudo-static back analysis of landslides was performed in the region by Harash and Bar (1988). They determined a paleo PGA range of 0.2–0.3 g for triggering two landslides near Alamagor, located 10 km north of the SOG inside the Jordan River Gorge (Fig. 1). Their results fall within our obtained range of paleo PGA for landslides that were triggered on slopes immediately on the boundaries of the SOG. Wechsler et al. (2009) performed a Newmark back analysis (Newmark, 1965) of a landslide in Umm el Qanatir, an archeological site located 10 km east of the SOG (Fig. 1). Their work returns critical acceleration values that range between 0.36 and 0.78 g, with the high value representing a low GWT scenario. All studies mentioned thus far focused on pseudo-static back analysis of landslides using LEA approach to constrain regional paleo PGA. Supporting evidence is obtained from a different study performed nearby by Yagoda-Biran and Hatzor (2010) that utilizes a fully dynamic numerical discrete element approach to investigate the level of horizontal paleo PGA required to topple collapsed columns in a Byzantine church at the archeological site of Susita, located immediately on the eastern slopes of the SOG, 3 km north of the Ein-Gev slide (see Fig. 1). Historical and archeological evidence indicate that Susita was destroyed by an earthquake (Amiran, 1996), and the series of toppled columns found today lying parallel to one another on the ground surface support the historical and archeological accounts. Using the numerical Discontinuous Deformation Analysis method (Shi, 1993) a model of a typical Susita column was subjected to a suite of real earthquake accelerograms. The resulting PGA range obtained this way for the Susita site is 0.2 to 0.4 g, well within the range obtained here for landslides around the SOG with pseudo-static LEA. 5.2. Inferring seismic intensity from mapped landslides The three seismically induced landslides mapped in the study area scale with ESI 2007 (Guerrieri and Vittori, 2007) intensity IX labeled: DESTRUCTIVE, qualitatively described there as: “Landsliding is widespread in prone areas, also on gentle slopes; where equilibrium is unstable (steep slopes of loose/saturated soils; rock falls on steep gorges, coastal cliffs) their size is frequently large (105 m3), sometimes very large (106 m3)”. Both the Ein-Gev and Berniki Beach landslides are of the order of 106 m3, and the Fishing Dock landslide is of the order of 104 m3. Therefore assigning an ESI 2007 intensity of IX for the driving earthquakes seems to be reasonable, suggesting that the earthquakes causing those landslides were rather strong. 5.3. Localization of paleo earthquake epicenters A recent paleoseismic study performed in the eastern margins of the SOG by Katz et al. (in press) reveals several normal faults that display multiple slip events in three trenches excavated several kilometers south of the Ein-Gev landslide site. The maximum detected offset for a single event was 1.5 meters, correlated by Katz et al.

89

(in press) with magnitude of Mw = 6.6 ± 0.4 on the basis of the well established empirical relationship of Wells and Coppersmith (1994). A correlation between moment magnitude and PGA may further be obtained with the empirical attenuation relationship proposed by Boore et al. (1997). An adapted Boore et al. (1997) attenuation relationship for the DST (strike–slip mechanism and shear wave velocity Vs = 620 m/s), is presented as inset in Fig. 9. Inspection of the adapted attenuation model reveals that the maximum PGA that can be generated by Mw = 6.6 earthquake, the value obtained from paleoseismic trenching, is 0.34 g at zero distance from the epicenter. Back analysis of the Ein-Gev landslide however indicates a threshold PGA value of 0.37 g, too high for a Mw = 6.6 event; rather, a Mw = 6.75 event at least is required to generate a PGA of 0.37 g at zero distance from the epicenter. We therefore need an event of magnitude greater than Mw = 6.75 to explain the Ein-Gev landslide, and will assume here a Mw = 7.0 event for the sake of this discussion. Moreover, since the Ein-Gev landslide is the only case that provides a single paleo PGA value from back analysis, we use this case for paleo epicenter localization, and then test the implications concerning the other landslides analyzed here. To localize the epicenter that could generate the Ein-Gev landslide we seek the intersection points between a circle representing the locus of Mw = 7.0 events that would generate a PGA of 0.37 g at EinGev, and the trace line of a major tectonic fault in the vicinity, the Eastern Margin Fault (EMF) (see Hurwitz et al., 2002) being the most likely candidate (points A and A′ in Fig. 9). Consider a theoretical epicenter location at A. An event of Mw = 7.0 would generate a PGA of 0.37 g at Ein-Gev, 0.3 g at Berniki, 0.22 g at the Fishing Dock, and 0.3 g at Susita. Similarly, a theoretical epicenter location at A′ would generate a PGA of 0.37 g at Ein-Gev, 0.3 g at Berniki, 0.26 g at the Fishing Dock, and 0.38 g at Susita. Therefore, by virtue of geometry alone, and ignoring chronological constraints for the moment, a single Mw = 7.0 event either at A or at A′ could generate all failures observed around the SOG circumference. The same localization approach can be applied to the Western Margin Fault (WMF). Back analysis of the Fishing Dock landslide provides a range of paleo-PGA values, between 0.15 and 0.5 g. For epicenter localization that calls for a single PGA value we choose 0.35 g as the representative PGA for this case study. Assuming a Mw = 7.0 event as before, we seek the intersection between a circle representing the locus of Mw = 7.0 events that would generate a PGA of 0.35 g at the Fishing Dock, and the trace line of a major tectonic fault in the vicinity, the WMF (see Ben-Avraham et al., 1996) being the most likely candidate (point B in Fig. 9). A single Mw = 7.0 event at point B will generate a PGA of 0.41 g at Berniki and 0.24 g at Susita thus explaining the failures observed and analyzed at those sites. Nevertheless, at Ein-Gev this event will only generate a PGA of 0.27 g that is not sufficient to cause the landslide. In fact, Triggering the EinGev landslide requires a minimum Mw = 7.6 event on the WMF, exceeding current seismological estimates for the DST, and paleoseismic and archaeoseismic observations (Amit et al., 2002; Ellenblum et al., 1998). 5.4. Chronological constraints Paleoseismic trenching was performed by Yagoda-Biran (2008) near the toe of the Ein-Gev landslide to try and resolve temporal relationships between the landslide with its sedimentary and pedological characteristics, and the slip and ground surfaces. Three identified paleosols overlaying the sliding mass were dated using OSL with the lowermost layer dated to 60 k years BP, the intermediate to 6 k years BP, and the uppermost to 4 k years BP (Yagoda-Biran, 2008). The Ein-Gev landslide reflects therefore a strong earthquake event along the EMF of the DST that occurred at least 60,000 years before present. Indeed, the current morphology of the slumped mass supports this finding with channel incisions through the rock and marked erosion

90

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Fig. 9. Paleo epicenter localization based on back analysis results for the studied landslides and a Mw = 7.0 event. Legend: Circles — the loci of Mw = 7.0 earthquakes that will generate a PGA of 0.37 g at Ein-Gev (East) and 0.35 g at the Fishing Dock (West). A and A′ — intersection of the loci of Mw = 7.0 earthquakes that will generate a PGA of 0.37 g at Ein-Gev, and the EMF. B — intersection of the loci of a Mw = 7.0 earthquake that will generate a PGA of 0.35 g at the Fishing Dock and the WMF. EMF trace after Hurwitz et al. (2002), WMF trace after Ben-Avraham et al. (1996). Inset: Modified Boore et al. (1997) attenuation relationship for the DST.

of the scarp and ground surface, leaving relatively little material between the slip and ground surfaces. OSL dating was not performed in other landslide sites, yet we suggest that the Berniki Beach landslide could have been triggered at the same event that generated the Ein-Gev landslide because of: a) the location of the Berniki Beach landslide immediately across the Ein-Gev slide on the other bank of the SOG, b) similar morphological appearance of the two slopes with similar erosion patterns, and c) the agreement between the analytically constrained paleo PGA levels and the location of the inferred fault epicenter along the EMF segment of the DST (Fig. 9). As in the case of the Berniki Beach slide, the Fishing Dock landslide has no firm time constraints because no dating was performed; yet

field observations imply a younger age as the scarp of the landslide can still be detected today in the field, whereas in both the Ein-Gev and Berniki Beach slides the scarp has been eroded. Therefore the Fishing Dock landslide implies a younger strong earthquake that ruptured on either the EMF or the WMF. 6. Summary and conclusions We show in this paper how regional paleo PGA values may be constrained from back analysis of landslides triggered by strong earthquakes. Several slope parameters must be known in order to perform this analysis: the slope geometry prior to failure, the

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

properties of the material comprising the slope, and location of the slip surface and GWT level. We show that even when not all parameters are available, educated assumptions can be made regarding the missing data, and an analysis can still be carried out. Furthermore, the limit equilibrium pseudo-static back analysis approach demonstrated here is rather simple and straight forward, lending further applicability to the suggested procedure. The paleo PGA levels constrained by back analysis of the three landslides around the SOG fall within a relatively narrow range: between 0.15 and 0.5 g. This result is supported by an earlier analytical landslide study and a recent numerical study of collapsed columns in the study area. Considering intensities, both the Ein-Gev and Berniki Beach landslides are of the order of 106 m3, and the Fishing Dock landslide is of the order of 104 m3 suggesting an ESI 2007 (Guerrieri and Vittori, 2007) intensity of IX for the driving earthquakes, classified there as “Destructive”. Using paleoseismic trenching data and a modified Boore et al. (1997) attenuation relationship for the DST, the loci of earthquakes of Mw = 7.0 that can account for the cases studied in this paper are determined, assuming that the source for strong seismic events in the area coincides with the Eastern and Western Margin Fault segments of the DST. We suggest that the EinGev and the Berniki Beach landslides could have been triggered by the same Mw = 7.0 at least 60,000 years before present. The Fishing Dock landslide was triggered, most likely, after these two landslides. According to our analysis it could have been triggered by another Mw = 7.0 event along the same DST segments in the 749 AD event that severely damaged the Susita site, or possibly before that date but not earlier than 60,000 BP. A robust chronological determination of this is beyond the resolution of this study. Results of our study imply that the SOG region experienced repeated earthquakes of Mw ∼ 7 with local PGA values of 0.15 to 0.5 g. These PGA values must be considered in any attempt to quantify the seismic hazard and to develop future hazard reduction programs for this region. Acknowledgements The authors would like to thank the Ministry of National Infrastructure of Israel for partial support of this research through a grant from the National Steering Committee for Earthquake Preparedness for the study of the seismic risk in the vicinity of the Sea of Galilee through contract 27-06-020. Naomi Porat of Israel Geological Survey is thanked for OSL dating. References Amiran, D.H.K., 1996. Earthquakes in the land of Israel. Qadmoniot 29, 53–61. Amit, R., Zilberman, E., Enzel, Y., Porat, N., 2002. Paleoseismic evidence for time dependency of seismic response on a fault system in the southern Arava Valley, Dead Sea rift, Israel. Geological Society of America Bulletin 114, 192–206. Arieh, E., Rabinowitz, N., 1989. Probabilistic assessment of earthquake hazard in Israel. Tectonophysics 167, 223–233. Arieh, E., Peled, U., Kafri, U., Shaal, B., 1977. The Jordan Valley Earthquake of September 2, 1973. Israel Journal of Earth Science 26, 112–118. Begin, Z.B., 2005. Destructive earthquakes in the Jordan Valley and the Dead Sea — their reoccurrence interval and the probability of their occurrence. Geological Survey of Israel Report GSI/12/2005. Ben-Avraham, Z., Tenbrink, U., Bell, R., Reznikov, M., 1996. Gravity field over the Sea of Galilee: evidence for a composite basin along a transform fault. Journal of Geophysical Research. Solid Earth 101, 533–544. Bishop, A.W., 1955. The use of the slip circle in the stability analysis of slopes. Geotechnique 5, 7–17. Boore, D.M., Joyner, W.B., Fumal, T.E., 1997. Equations for estimating horizontal response spectra and peak acceleration from western North American earthquakes: a summary of recent work. Seismological Research Letters 68, 127–153. Brune, J.N., 1996. Precariously balanced rocks and ground-motion maps for southern California. Bulletin. Seismological Society of America 86, 43–54. Corps of Engineers, 1982. Slope Stability Manual EM-1110-2-1902. Department of the army, Office of the Chief of Engineers, Washington D.C. Cundall, P.A., 1988. Formulation of a 3-dimensional distinct element model. 1. A scheme to detect and represent contacts in a system composed of many polyhedral blocks.

91

International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts 25, 107–116. Ellenblum, R., Marco, S., Agnon, A., Rockwell, T., Boas, A., 1998. Crusader castle torn apart by earthquake at dawn, 20 May 1202. Geology 26, 303–306. Enzel, Y., Kadan, G., Eyal, Y., 1997. Holocene earthquakes inferred from a fan-delta sequence in the Dead Sea graben. Quaternary Research 53, 34–48. Fellinius, W., 1936. Calculations of stability of earth dams. Transactions, 2nd Congress of Large Dams, Washington D.C. Garfunkel, Z., 1981. Internal structure of the Dead Sea leaky transform (rift) in relation to plate kinematics. Tectonophysics 80, 81–108. Guerrieri, L., Vittori, E., 2007. Intensity scale ESI 2007. Mem. Descr. Carta Geologica d'Italia, 74. Servizio Geologico d'Italia - Dipartimento Difesa del Suolo, APAT, Rome, Italy. Hall, J.K., 1994. Digital shaded-relief map of Israel and environs 1:500,000. I.G. Survey. Harash, A., Bar, Y., 1988. Faults, landslides and seismic hazard along the Jordan River Gorge, northern Israel. Engineering Geology 25, 1–15. Hofstetter, A., Thio, H.K., Shamir, G., 2003. Source mechanism of the 22/11/1995 Gulf of Aqaba earthquake and its aftershock sequence. Journal of Seismology 7, 99–114. Hurwitz, S., Garfunkel, Z., Ben-Gai, Y., Reznikov, M., Rotstein, Y., Gvirtzman, H., 2002. The tectonic framework of a complex pull-apart basin: seismic reflection observations in the Sea of Galilee, Dead Sea transform. Tectonophysics 359, 289–306. Itasca, 1993. FLAC — Fast Lagrangian analysis of continua. ITASCA Consulting Group Inc., Minneapolis, Minn. Janbu, N., 1954. Application of composite slip surface for stability analysis. European Conference on Stability of Earth Slopes, Stockholm, Sweden. Jibson, R.W., 1996. Use of landslides for paleoseismic analysis. Engineering Geology 43, 291–323. Kagan, E.J., Agnon, A., Bar-Matthews, M., Ayalon, A., 2005. Dating large infrequent earthquakes by damaged cave deposits. Geology 33, 261–264. Kamai, R., Hatzor, Y.H., 2008. Numerical analysis of block stone displacements in ancient masonry structures: a new method to estimate historic ground motions. International Journal for Numerical and Analytical Methods in Geomechanics 32, 1321–1340. doi:10.1002/nag.671. Katz, O., Amit, R., Yagoda-Biran, G., Hatzor, Y.H., Porat, N., Medvedev, B., in press. Quaternary earthquakes and landslides in the Sea of Galilee area, the Dead Sea Transform; paleoseismic analysis and evaluation of current hazard. Israel Journal of Earth Science. Keisar, I., Peferbaum, S., 2001. Landslide on road 90, km 416–417: definition of problem and stability analysis. 2nd national conference for construction and infrastructure, Tel Aviv. Ken-Tor, R., Agnon, A., Enzel, I., Stein, M., Marco, S., Negendank, J.F.W., 2001. Highresolution geological record of historic earthquakes in the Dead Sea basin. Journal of Geophysical Research-Solid Earth 106 (B2), 2221–2234. Krahn, J., 2004. Stability modeling with SLOPE/W, an engineering methodology. GEOSLOPE International, Calgary. Longpre, M.A., Del Potro, R., Troll, V.R., Nicoll, G.R., 2008. Engineering geology and future stability of the El Risco landslide, NW-Gran Canaria, Spain. Bulletin of Engineering Geology and the Environment 67, 165–172. doi:10.1007/s10064007-0119-9. Lowe, J., Karafiath, L., 1960. Stability of earth dams upon drawdown. Proceedings of the 1st Pan American Conference on Soil Mechanics and Foundation Engineering, Mexico City. Marco, S., Agnon, A., 1995. Prehistoric earthquake deformations near Masada, Dead-Sea graben. Geology 23 (8), 695–698. Marco, S., Agnon, A., 2005. High-resolution stratigraphy reveals repeated earthquake faulting in the Masada Fault Zone, Dead Sea Transform. Tectonophysics 408 (1–4), 101–112. Marco, S., Rockwell, T.K., Heimann, A., Frieslander, U., Agnon, A., 2005. Late Holocene activity of the Dead Sea Transform revealed in 3D palaeoseismic trenches on the Jordan Gorge segment. Earth and Planetary Science Letters 234 (1–2), 189–205. Matmon, A., Shaked, Y., Porat, N., Enzel, Y., Finkel, R., Lifton, N., Boaretto, E., Agnon, A., 2005. Landscape development in an hyperarid sandstone environment along the margins of the Dead Sea fault: implications from dated rock falls. Earth and Planetary Science Letters 240, 803–817. doi:10.1016/j.epsl.2005.06.059. Migowski, C., Agnon, A., Bookman, R., Negendank, J.F.W., Stein, M., 2004. Recurrence pattern of Holocene earthquakes along the Dead Sea transform revealed by varvecounting and radiocarbon dating of lacustrine sediments. Earth and Planetary Science Letters 222 (1), 301–314. Mor, D., Sneh, A., 1996. Complementary mapping of Plio-Pleistocene basalt units and landslides in the southern Golan and the Yarmouk Valley and its bearing on the Geomorphological history of the area. Geological Survey of Israel, Current Research 10, 28–31. Morgenstern, N.R., Price, V.E., 1965. Analysis of stability of general slip surfaces. Geotechnique 15, 79. Newmark, N., 1965. Effects of earthquakes on dams and embankments. Geotechnique 15, 139–160. Niemi, T.M., Ben-Avraham, Z., 1997. Active tectonics in the Dead Sea basin. In: Niemi, T.M., Ben-Avraham, Z., Gat, J.R. (Eds.), The Dead Sea – The Lake and its Setting. Oxford Monographs of Geology and Geophysics, pp. 73–81. Palmer, A.C., Rice, J.R., 1973. Growth of slip surfaces in progressive failure of overconsolidated clay. Proceedings of the Royal Society of London. Series A: Mathematical Physical and Engineering Sciences 332, 527–548. S.I.I., 2004. Israel Building Code #413, 2nd amendment. The Standards Institution of Israel. Saltzman, U., 1964. The geology of the Tabcha-Chokuk-Migdal region. Hebrew University of Jerusalem.

92

G. Yagoda-Biran et al. / Tectonophysics 490 (2010) 81–92

Saltzman, U., 2005. A geological engineering survey, Tiberius, road 348. Un-published report, submitted to Israel National Roads Company, in Hebrew. Sarma, S.K., 1973. Stability analysis of embankments and slopes. Geotechnique 23, 423–433. Shapira, A., 1983. Potential earthquake risk estimation by application of a simulation process. Tectonophysics 95, 75–89. Shapira, A., Avni, R., Nur, A., 1993. New estimate of the Jericho earthquake epicenter of July 11, 1927. Israel Journal of Earth Sciences 42, 93–96. Shapira, A., 2002. Explanations to the PGA map for the Israeli building code 413. http:// www.gii.co.il/2002. Shi, G., 1993. Block System Modeling by Discontinuous Deformation Analysis. Computational Mechanics Publication, Southhampton, UK. Shou, K.J., Wang, C.F., 2003. Analysis of the Chiufengershan landslide triggered by the 1999 Chi-Chi earthquake in Taiwan. Engineering Geology 68, 237–250. Skempton, A.W., Petley, D.J., 1967. The shear strength along structural discontinuities in stiff clays. Proceedings of the Geotechnical Conference Oslo. Spencer, E., 1967. A method of analysis of stability of embankments assuming parallel inter-slice forces. Geotechnique 17, 11. Strasser, M., Anselmetti, F.S., Fah, D., Giardini, D., Schnellmann, M., 2006. Magnitudes and source areas of large prehistoric northern Alpine earthquakes revealed by slope failures in lakes. Geology 34, 1005–1008. Szeidovitz, G., Suranyi, G., Gribovszki, K., Bus, Z., Leel-Ossy, S., Varga, Z., 2008. Estimation of an upper limit on prehistoric peak ground acceleration using the

parameters of intact speleothems in Hungarian caves. Journal of Seismology 12, 21–33. doi:10.1007/s10950-007-9068-9. Urgeles, R., Leynaud, D., Lastras, G., Canals, M., Mienert, J., 2006. Back-analysis and failure mechanisms of a large submarine slide on the Ebro slope, NW Mediterranean. Marine Geology 226, 185–206. doi:10.1016/j.margeo.2005.10.004. Wechsler, N., Katz, O., Dray, Y., Gonen, I., Marco, S., 2009. Estimating location and size of historical earthquake by combining archaeology and geology in Umm-El-Qanatir, Dead Sea Transform. Natural Hazards 50, 27–43. doi:10.1007/s11069-008-9315-6. Wells, D.L., Coppersmith, K.J., 1994. New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. Bulletin. Seismological Society of America 84, 974–1002. Yagoda-Biran, G., 2008. Seismic hazard estimation along eastern margins of Sea of Galilee by back analysis of seismically induced natural and structural failures. Ben Gurion University of the Negev. http://www.bgu.ac.il/geol/hatzor/theses/Gony_MSc_thesis.pdf. Yagoda-Biran, G., Hatzor, Y.H., 2010. Constraining paleo PGA values by numerical analysis of overturned columns. Earthquake Engineering and Structural Dynamics 39, 462–472. doi:10.1002/eqe.950. Zaslavsky, Y., Shapira, A., 2000. Experimental study of topographic amplification using the Israel seismic network. Journal of Earthquake Engineering 4, 43–65. Zienkiewicz, O.C., Taylor, R.L., 2000. Finite Element Method. Volume 1 — the basics. Butterworth Heinemann, Oxford.