The geological excavation of Vadum Iacob Castle was funded by the Earth Science Administration of the Ministry of Infrastructure and the Geophysical Institute.
The excavation of the Dead See Rift Valley: The castle was torn apart by a movement on a geological fault which crosses the site from south to north. This unique situation will enable us, for the first time in the history of archaeology to excavate along the line of a major fault and to raise a major contribution to the study of Seismology of major earthquakes.
The chances of studying medieval construction methods and the behavior of a recent fault are added to the other scientific opportunities that the site offers and the other opportunities that are still to come. For example we still make efforts in order to discover and excavate the deep well into which, according to historical sources, the corpses of the garrison were thrown.
Historical Earthquakes Revealed at Vadum Jacob
The Crusader castle of Vadum Jacob, an outpost overlooking the Jordan River, was deformed during a destructive earthquake triggered by motion along the Dead Sea Transform. The M>7 earthquake occurred at dawn, 20 May 1202 and offset the castle walls by 1.6 m. This exceptional precision in dating and estimating displacement was achieved by combining accounts from primary historical sources, excavating the Dead Sea Transform where it bisects the castle, and by dating faulted archaeological strata. The earthquakes of October 1759 most probably account for the remaining 0.5 m out of a total 2.1 m of offset. Our study exploits the potential embodied in interdisciplinary historical-archeological-geological research, and illustrates how detailed histories of seismogenic faults can be reconstructed.
In spite of many large earthquakes known to have occurred in the Middle East throughout history (e.g., Amiran et al., 1994) , neither earthquake-induced surface ruptures, nor displacements of human-made structures by fault-slip have been recorded until the archaeological excavations Vadum Jacob (‘Ateret) Crusader Fortress in 1994. The excavations, conducted by Ronnie Ellenblum and Adrian Boas, revealed that E-W trending Crusader walls in the fortress are displaced sinistrally up to 2.1 m by a N-S fault. Younger Muslim structures are displaced by 0.5 m.
The history of the Middle East is punctuated by occasional strong destructive earthquakes that inflicted heavy losses of property and life. Historical accounts describe their effects but they do not relate them to slip on geological faults. At Ateret we observed the slip but we did not know which of the historical earthquake(s) was/were responsible to its formation. The research which followed the discovery at Ateret was aimed at matching the historical accounts of earthquakes and the observed deformations in the fortress area (Marco et al., 1997; Ellenblum et al., 1998) . Exploration of past earthquake activity can illuminate important aspects of the earthquake process as well as improve our prepardness for future ones.
The Dead Sea Transform in the study area
Most of the strong earthquakes in historical times are thought to have been triggered by motion on the Dead Sea Transform (DST) fault. This fault transfers opening at the Red Sea and Gulf of Aden to the collision between Arabia and Eurasia along the Tauros-Zagros mountain belt. Hence there is a sinistral motion (left-lateral) between the Arabia and the Sinai tectonic plates (Figure 1 ).
The paradigm of left-lateral shear along the DST is consistent with earthquake focal plane solutions (Salamon et al., 1996) and explains the systematic offset of pre Miocene geologic features by 105 km (Quennell, 1956; Freund et al., 1968; Garfunkel, 1981) .
The site of Ateret provides a unique opportunity to learn about past earthquakes because it was built right on the plate boundary, at the northern end of the Jordan Gorge.
The Jordan Gorge Fault is a segment of the DST between the Sea of Galilee and the Hula pull-apart basins (Figure 2 ). The Jordan River forms a deep and narrow gorge where it crosses the Rosh-Pina Saddle, the structural and topographic high between the Hula and the Sea of Galilee. The termination of rock units and the different geometry of secondary faults on both sides, along with morphologic indicators, suggest that the gorge is incised along the main fault (Garfunkel et al., 1981; Harash and Bar, 1988; Heimann and Ron, 1993) . A deep seismic reflection survey shows that the fault zone is made of several strands with the main active fault coinciding with the Jordan Gorge (Rotstein and Bartov, 1989) . A shallow reflection at the Bet Zayda (Beteiha) Valley also reveals faulted reflectors where the Jordan River formes a flat delta at its entrance into the Kinneret (Sea of Galilee) (Frieslander, 1997) .
North of the Hula, the displacement is distributed among the Roum, Rashaya, and Yammuneh faults; the latter links the Dead Sea Transform to the collision zone.
Microseismicity in the Jordan Gorge is subdued relative to the adjacent Sea of Galilee and Hula basins. The strongest event recorded instrumentally in the Hula Valley was an ML 4.3 earthquake about 3 km north of Ateret; a focal plane solution shows mainly sinistral motion and a minor normal component (van Eck and Hofstetter, 1990) . A geodetic survey of the area has not detected any creep movement across the fault since the establishment of the benchmarks in 1988 (Karcz, 1995) .
DFORMED ARCHAEOLOGICAL STRUCTURES AT ATERET
We discovered offset archaeological remains at four excavated locations within the castle (Figure 3 and Figure 4 ). Three of these locations are Crusader structures and the fourth is a late mediaeval-early modern Muslim structure. The offset, fully expressed in the southern and northern defense walls, reaches 2.1 m in sinistral displacement with less than 5 cm of vertical slip (Marco et al., 1997) . The 50-m-long walls crossing the hill are ideal baselines, having meticulously laid masonry and a fault-perpendicular (east-west) orientation. The displacement measured on the walls record the cumulative slip on the fault since 1179, the year the castle was conquered by Saladin. Displacement is distributed over about a 10-m-wide zone, and the deformation is accommodated primarily by small offsets and rotations of the carved limestone blocks. All the displacements on the southern wall are purely horizontal (all the blocks retain their original level), and all the rotations are about vertical axes. A minor vertical component of slip, up to 10 cm, is observed in the northern wall. North of the southern main gate, the fault trace bends westward and a Crusader floor is torn forming a 2-m-wide graben (Figure 4 ). The geometry of the graben is compatible with a small left bend in the trace of the fault.
In the northern part of the castle we also unearthed a Muslim mosque whose northern wall is displaced sinistrally by 0.5 m. A mikhrab (the Muslim praying apse) is well preserved in the southern wall. According to the study of the pottery, the mosque was built, destroyed, and rebuilt at least twice: the initial structure was built in the Muslim period (12th century) and later rebuilt once or twice during the Turkish Ottoman period (1517-1917). The 0.5 m displacement is observed in the northern wall of the latest building phase. The repetitive building of this site might be due to earthquakes.
DATING OF EARTHQUAKE-RELATED DISPLACEMENTS
Four post A.D. 1179 major earthquakes are recorded by historical sources in northern Israel. These are the earthquakes of A.D 1202, 1546, 1759, and 1837 (Amiran et al., 1994) .
The estimated zone of damage to buildings (meisoseismal zone) of the 20 May 1202 earthquake extends from 100 km south to 150 km north of Vadum Jacob. It was felt in the entire eastern Mediterranean region and throughout the Levant. Ambraseys and Melville (1988) estimated the magnitude at 7.6, with maximum displacement of about 2.5 m.
Damage from the 14 January 1546 event extended from 75 km to 150 km south of Vadum Jacob (Amiran et al., 1994) . Ambraseys and Karcz (1992) argue for a moderate magnitude (Ms ~6.0) for this event, and infer the epicenter to have been located in Judea, ~100 km south-southwest of Vadum Jacob. Thus this event is unlikely to have ruptured the Jordan Gorge segment of the Dead Sea Transform.
Two close events occurred on 30 October and 25 November 1759. Sieberg (1932) located the maximum damage zone of the October earthquake between the Sea of Galilee and the Hula Valley, and that of the November event some 150 km farther north in northeast Lebanon. Ambraseys and Barazangi (1989) , quote a letter dated 1760 in which the French ambassador to Beiruth reports surface ruptures along 100 km of the Yammuneh segment of the Dead Sea Transform and attributes them to the November 1759 earthquake. Ambraseys and Barazangi (1989) estimate the magnitude of the 25 November 1759 earthquake at ~7.4. The October 1759 M ~6.6 foreshock, determined on the basis of isoseismals that center at the Jordan Gorge (Sieberg, 1932; Ambraseys and Barazangi, 1989) , could be related to faulting at Vadum Jacob.
The most recent destructive earthquake to strike the study area was the M ~6.3 1 January 1837 Safed earthquake. The center of the highest damage zone, IX -X Mercalli intensity, coincides with Vadum Jacob (Vered and Striem, 1977) .
In order to date the slip events at Vadum Jacob, we excavated a trench adjacent to the offset southern wall of the fortress (Figure 4 ). Sediments exposed in the trench provide constraints on the history of surface ruptures, as recorded in the castle construction debris and overlying colluvium (Figure 5 ).
The trench was excavated into an artificial slope. Sections were made into the slope at several locations, and the pottery was studied and dated. The sediments exposed in the trench compose four primary stratigraphic units (Figure 5 ). The lowest, unit 1, is composed of the sediments that were dumped by the builders at the time of castle construction; these bury the first six stone layers of the wall. At two distinct horizons the fill was leveled out and covered with a layer of lime. These limy layers are continuous at the same level around the perimeter of the wall. Unit 2 is the upper limy layer. We found building tools, mason stone blocks, arrowheads, and other remains of the siege immediately on top of the layer indicating that unit 2 was the surface level as of August 1179. This therefore, serves as an accurate time marker for this month.
One of the Muslim chroniclers of the siege (Abu-Shama) stated that Saladin, the conqueror of the castle, "tore away the stones of the castle, by his own hands, destroying it like one effaced letters of a parchment." The conquest of the castle and the massacre of its defenders brought about a plague that started within a few days of its capture, causing the evacuation of the site (Lyons and Jackson, 1982) . We therefore conclude that the massive destruction and dismantling of the wall started and terminated within a very short period of the conquest and that the two ashlars on the top of unit 2 fell or tumbled there from the wall immediately after the castle was captured by the Muslims, or possibly during the siege.
Overlying unit 2 is a wedge of colluvium, unit 3, that was shed off the wall infill after the exterior wall face was removed. Presumably, the colluvial unit 3 piled up as the weakly cemented interior wall deteriorated and decomposed over the ensuing centuries. Thus unit 3 started to accumulate after the Muslim conquest, and it ranges in age from about the end of A.D. 1179 to the present. Unit 4 is the modern bioturbated A-soil horizon.
The trench also exposed a system of faults and cracks that are coincident with the displacement of the Crusader wall. The faults extend to two different stratigraphic levels: One group of faults displaces the alluvium of unit 1 and the limy level of unit 2, but extends only a few centimeters into post-A.D. 1179 unit 3; the second group of faults breaks much higher into the colluvial wedge, up to the base of the modern soil horizon, and possibly to the surface. These observations suggest that at least two earthquakes produced the 2.1 m offset of the southern wall that is now observed. One event occurred soon after the outer ashlar wall was removed, i.e., very soon after 1179. The second post-1179 earthquake also produced rupture at Vadum Jacob, but well after removal of the wall and the accumulation of the colluvium, probably much closer to the present.
The A.D. 1202 earthquake occurred fewer than 23 years after the capture of the castle, and is almost certainly the source of the first set of interpreted ruptures breaking the lowermost part of unit 3. The later rupture(s) may be associated with the 1759 and/or the 1837 earthquakes. We adopt Ambraseys and Karcz’s (1992) view that the 1546 earthquake was considerably farther south, closer to Judea, and its magnitude was too small to produce surface rupture at Vadum Jacob.
Our findings support the historical records of large and destructive earthquakes in the Dead Sea region and demonstrate that the Dead Sea Transform poses substantial seismic hazard. The low seismicity in recent years and the absence of significant creep along the fault, evident from repeated geodetic surveys, suggest that the fault is currently locked. Projecting the destructive force of past earthquakes to likely future events warns for the potential of widespread damage in large population centers of Jordan, Syria and Israel. Future excavation of additional archaeological and prehistoric sites along the Dead Sea Transform should help to characterize the long-term behavior of rupture segments along the fault.
The abundant historical and archaeological data available from the Dead Sea Transform fault region provides an opportunity for detailed studies of destructive earthquakes. This study illustrates the use of geoarcheology to assess history of faulting in active tectonic zones. On-fault archeoseismology offers the potential for identifying and accurately dating past earthquakes over several earthquake cycles, thus providing fundamental information on earthquake recurrence patterns and mechanisms.
- Abu-Shama, Le livre des deux jardins. Histoire des deux regnes, celui de Nour ed-Din et celui de Salah ed-Din, in Recueil des historiens des croisades, Historiens orientaux, ed. Acadטmie des Inscriptions et Belles-Lettres, Paris (1872-1906), vol. IV, p. 194-208.
- Ambraseys, N., and Karcz, I., 1992. The earthquake of 1546 in the Holy Land. Terra Nova, 4: 253-262.
- Ambraseys, N. N., and Barazangi, M., 1989. The 1759 earthquake in the Bekaa valley: Implications for earthquake hazard assessment in the eastern Mediterranean region. J. Geophys. Res., 94: 4007-4013.
- Ambraseys, N. N., and Melville, C. P., 1988. An analysis of the eastern Mediterranean earthquake of 20 May 1202. In: Lee, W. K. H., Meyers, H. and Shimazaki, K. (Editors), Historical Seismograms and Earthquakes of the World. Academic Press, San Diego, California, 181-200.
- Amiran, D. H. K., Arieh, E., and Turcotte, T., 1994. Earthquakes in Israel and adjacent areas: Macroseismic observations since 100 B.C.E. Israel Exploration Journal, 44: 260-305.
- Bartov, Y., 1979. Israel - Geological Map 1:500,000.
- Belitzky, S., 1987. Tectonics of the Korazim Saddle (M.Sc. thesis). Hebrew University, Jerusalem (in Hebrew, English abstract).
- Ellenblum, R., Marco, S., Agnon, A., Rockwell, T., and Boas, A., 1998. Crusader castle torn apart by earthquake at dawn, 20 May 1202. Geology, 26: 303-306.
- Freund, R., Zak, I., and Garfunkel, Z., 1968. Age and rate of the sinistral movement along the Dead Sea Rift. Nature, 220: 253-255.
- Frieslander, U., 1997. High-resolution seismic surveys in the Hula and Beteiha Valleys. The Geophysical Institute of Israel, report #647/162/96.
- Garfunkel, Z., 1981. Internal structure of the Dead Sea leaky transform (rift) in relation to plate kinematics. Tectonophysics, 80: 81-108.
- Garfunkel, Z., Zak, I., and Freund, R., 1981. Active faulting in the Dead Sea rift. Tectonophysics, 80: 1-26.
- Goren-Inbar, N., and Belitzky, S., 1989. Structural position of the Pleistocene Gesher Benot Ya’akov site in the Dead Sea Rift zone. Quaternary Research, 31: 371-376.
- Harash, A., and Bar, Y., 1988. Faults, landslides and seismic hazards along the Jordan River Gorge, northern Israel. Eng. Geol., 25: 1-15.
- Heimann, A., and Ron, H., 1993. Geometric changes of plate boundaries along part of the northern Dead Sea Transform: Geochronologic and paleomagnetic evidence. Tectonics, 12: 477-491.
- Karcz, I., 1995. Development of a geodetic system for monitoring of recent crustal movements along the Dead Sea rift. Earth Sciences Administration, Ministry of Energy and Infrastructure, Israel, report #TR-GSI/7/95.
- Lyons, M. C., and Jackson, D. E. P., 1982. Saladin: The Politics of the Holy War: Cambridge, London, 454 pp.
- Marco, S., Agnon, A., Ellenblum, R., Eidelman, A., Basson, U., and Boas, A., 1997. 817-year-old walls offset sinistrally 2.1 m by the Dead Sea Transform, Israel. Journal of Geodynamics, 24: 11-20.
- Quennell, A. M., 1956. Tectonics of the Dead Sea rift. Congreso Geologico Internacional, 20th sesion, Asociacion de Servicios Geologicos Africanos, Mexico, p. 385-405.
- Rotstein, Y., and Bartov, Y., 1989. Seismic reflection across a continental transform: an example from a convergent segment of the Dead Sea rift. J. Geophys. Res., 94: 2902-2912.
- Salamon, A., Hofstetter, A., Garfunkel, Z., and Ron, H., 1996. Seismicity of the eastern Mediterranean region: Perspective from the Sinai subplate. Tectonophysics, 263: 293-305.
- Sieberg, A., 1932. Erdbebengeographie. In: (Editors), Handbuch der Geophysik, Band IV. Borntraeger, Berlin, 527-1005.
- van-Eck, T., and Hofstetter, A., 1990. Fault geometry and spatial clustering of microearthquakes along the Dead Sea-Jordan rift fault zone. Tectonophysics, 180: 15-27.
- Vered, M., and Striem, H. L., 1977. A macroseismic study and the implications of structural damage of two recent major earthquakes in the Jordan rift. Bull. Seism. Soc. Am., 67: 1607-1613.
We are most grateful to all the people who worked in the excavation in the soaring heat of summer in the Jordan Valley. We also thank the Jewish National Fund, the Atomic Energy Commision, The National Center for Cooperation Between Science and Archaeology, and the U.S.-Israel Binational Science Foundation (grant to T. R., R. E., and A. A.). Thanks to Shmuel Belitzky for elucidating the local geology.