• from Sbeinati et al. (2010:249-250)

The Al-Harif aqueduct is located ~4 km north of the city of Missyaf, immediately west of a limestone shutter ridge and related ~200 m left-lateral stream deflection. According to the remaining aqueduct walls and related mills in the region, the aqueduct was built during the Roman time (younger than 65 B.C. in the Middle East) to drain freshwater collected from springs of the western mountain range to the eastern semiarid plains. The remaining ruins of the aqueduct suggest an ~40-km-long construction that may have included several bridges over streams and landscape gorges.

The aqueduct building description and related age have not been reported so far in any archive, manuscript, or in the literature. There is, however, an interesting anecdotal story from the local tradition that it was built by a local prince to supply potable water to Apamea and/or Sheizar cities, located northeast of the aqueduct. Apamea during that time was the most famous and strategic city during the Hellenistic and Roman period, whereas Sheizar is known to have been an important political and military fortress during the Middle Ages.

In their description of the Dead Sea fault in Syria, Trifonov et al. (1991) mentioned the existence of a faulted aqueduct near the city of Missyaf, but neither the precise location nor the accurate amount of offset walls was given. However, this early tectonic observation was helpful and allowed us to discover the site and consider a detailed study (Meghraoui et al., 2003), which is extended here using combined methods in archaeoseismology, paleoseismology, and tufa investigations. In addition, a micro- topographic survey of measurements accompanied all field studies.

Previous investigations on the aqueduct (Meghraoui et al., 2003) established: (1) an evaluation of its age based on an account of the large size blocks, the dating of sedimentary units below the aqueduct wall foundation, and dating of early tufa deposits on the aqueduct wall, and (2) the identification of the seismic faulting origin of damage in nearby trench A. The building style, with typical bridge arch and large stone size disposition (Opus caementum), suggested a Roman age, which was confirmed by the radiocarbon dating of sedimentary layers below the walls and the early tufa deposits on the walls. The faulted aqueduct revealed 13.6 ± 0.20 m of total left-lateral offset and called for detailed investigations on the characteristics and history of successive fault movements.

The aqueduct design, with an open canal on top of the 4-m-high wall, allowed freshwater and carbonate-saturated water to overflow and induce significant tufa accumulation from 0.30 m to 0.83 m in section. The carbonate-rich and cool water collected from the nearby western range is associated with a semiarid and karstic area of the Mesozoic limestone that favors rapid carbonate precipitation and tufa accumulation. The tufa deposits show successive growths of lamination carbonate with high porosity, banded texture, and rich organic encrustations. Field observations show that tufa accumulation developed on both eastern and western sections (from the fault line), but only on the north-facing wall, likely due to a slight tilt of the damaged aqueduct wall, probably after the two first earthquakes.

The following paragraphs present the field investigations, which consisted of:

1. four archaeoseismic excavations near the aqueduct walls and remains
2. four paleoseismic trenches across the fault zone and the alluvial sediments
3. four cores (two cores were previously studied in Meghraoui et al., 2003) of tufa deposits collected from different sections of the aqueduct.

More than 200 samples of organic matter, charcoal fragments, and tufa core pieces were taken for radiocarbon analysis in order to characterize the timing of successive faulting and related damage of the aqueduct construction. All radiocarbon dating were calibrated (2σ range, 95.4% probability density) using Oxcal v4.0 and INTCAL04 calibration curve.

• from Sbeinati et al. (2010:251-256)

The remaining aqueduct construction forms an ~50-m-long, ~5-m-high, and 0.60-m-thick wall that includes an ~15-m-high arch bridge in its eastern section (Figs. 5 and 6A). The outer part is coated by a thick layer of tufa deposits, probably due to a long period of freshwater flow. The construction material that may vary with the successive building and repair ages is made of:

1. large-size limestone blocks (Opus quadratum, 1.0 m × 0.5 m × 0.5 m; see also https://www.romanaqueducts.info/aquasite), similar to the typical Roman archaeological constructions and visible at the lower bridge (pier section) and wall sections


2. medium-size limestone blocks (Opus incertum; 0.50 m × 0.30 m × 0.30 m), which form the foundation or the upper half wall section and show visible small portions of cement


3. small sizes of mixed stones of irregular shape with significant portions of mortar (cement), mostly visible in the apparently rebuilt part of the wall

Figures 5 and 6A also show a detached small piece of the aqueduct wall made of small-size stones and related cement ~3.5 m away from the eastern wall. Therefore, four areas (noted I to IV in Fig. 5) were excavated near the aqueduct using proper archaeological methods.

The large excavation I was dug on the fault zone near the dragged wall fragment, in the area between the eastern and western aqueduct walls (Figs. 5, 8A, and 8B). The purpose of excavation is here to study the relationships between the fault zone and aqueduct. The excavation that has ~4.5 × 4.5 m surface and ~0.6 m depth exposed missing parts of the aqueduct. A buried and fallen wall piece rotated and dragged parallel to the fault and a remaining wall piece in an oblique position between two shear zones were discovered. The buried wall fragments are not comparable to the Opus caementum (quadratum) of the original construction and suggest a rebuilding phase. The excavation floor displays oriented gravels and pebbles that mark the shear zones and related fault branches also visible in the inner trench section E (Figs. 8B and 8C).

We collected four samples in the fallen wall sections labeled A, B, and C of excavation I (Fig. 8B): Two cement samples (AQ-CS-1 and AQ-CS-4) found in between building stones are made of typical medieval rubble mortars (mainly mud, gypsum, and lime); the two other samples (AQ-CS-3-2 and AQ-CS-3-3) are tufa deposits preserved on building stones. All four samples contained enough organic matter to allow radiocarbon dating (Table 1). Two dates of cement yield A.D. 532–641 (section A, AQ-CS-4) for the large fallen wall in excavation I and A.D. 650–780 (section C, AQ-CS-1) for the wall fragment piece in between the walls (Fig. 8B). In addition, two tufa deposits on wall stones provide consistent ages A.D. 560–690 (section B, AQ-CS-3-2) and A.D. 639–883 (section C, AQ-CS-3-3) with cement ages. The two different cement dates of the fallen wall and dragged wall fragment can be correlated to the new tufa deposits that testify for two rebuilding phases. The dated buried fallen wall in section B (CS-3-2) obtained from a thin (~5 cm) tufa accumulation correlates with the similarly fallen wall in section A and related cement date of CS-4 (Fig. 8B; and Table 1). In section C, the tufa deposits and related dated sample CS-3-3 correlates with the cement age of CS-1. The type and size of stones (opus incertum) and thin tufa accumulation in sections A and B suggest an early rebuilding phase postdating the first damaging event that may have occurred between the first and sixth centuries A.D. The different building layout of section C made of small sizes of mixed stones of irregular shape, and dating of cement sample CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) indicate a repair and rebuilding period postdating a second damaging event at the end of the Byzantine time and beginning of the Islamic period (seventh to eighth century A.D.). The damaged and dragged most recent wall section C along the fault indicates the occurrence of a third event, after which the aqueduct was definitely abandoned.

Three small excavations II, III, and IV (1.5 m to 3.0 m long, 1.0 m wide and 1.50 m deep) were dug in the base layer of the western aqueduct wall in order to expose its foundation and related sedimentary units underneath that predate the early building phase (Figs. 5 and 9). Excavations II and III were dug under the wall section with maximum (> 0.80 m) and minimum (~0.30 m) thickness of tufa deposition, respectively (Figs. 9A and 9B). Excavation IV, already described in Meghraoui et al. (2003), exposed the faulted foundation of the missing section of the western wall edge. The wall foundation reaches 1 m depth and shows regular patterns of medium-size cut limestone blocks (0.50 m × 0.30 m × 0.30 m) built over a dark brown clayey layer (unit e).

Charcoal samples collected in excavations I (trench E), II, and III from unit e yield ¹⁴C dates with ages spanning from approximately the third century B.C. to third century A.D. (see samples AQ-TA, TB, and TC in Table 1; Figs. 8C, 9A, and 9B). Although in these excavations, the age range of unit e seems quite large (probably due to detrital charcoal mixing), the younger age, i.e., 350 B.C. to A.D. 130 (sample AQ-TA-4), is consistent with other radiocarbon ages of unit e and related stratigraphic succession in trenches (see Paleoseismic Trenches herein). In excavation II, the large stone shape (Opus quadratum) with small amount of cementing material and pottery fragments found on the same level near the building base can be correlated with the early Roman era (Fig. 9A). Large stones and tufa thickness led us to consider this section of the aqueduct wall to be in original condition, i.e., probably undamaged by large earthquakes.

Excavation III (1.65 m long, 1.0 m wide, and 1.2 m deep; Fig. 9B) is similar to excavations II and IV, but the 1-m-deep wall foundation and upper section show irregular shapes of mixed medium- and small-size cut limestone blocks (0.10 × 0.20 × 0.15 m). Excavation III was realized at the location of the thinnest tufa deposits (< 0.30 m). The size of stones, cement texture, and irregular shape of building wall suggest that this building section was rebuilt (Fig. 9B). The ¹⁴C dating of unit e below the wall yielded a comparable age range to that obtained in excavations II (see AQ-TA, TB, and TC in Table 1).

Trench section E (4.30 m long, 0.70 m wide, and 1.30 m deep; Figs. 5, 8A, 8B, and 8C) was dug within excavation I in order to see in section the fault zone that affects the archaeological floor units. The trench wall exposes similar sedimentary units to those visible in excavations II, III, and IV that are affected by two main fault branches of the shear zone visible in the floor layer of excavation I. The ¹⁴C dating of samples AQ-TC-S1, S2, and S3 of units f and e indicates 900 B.C. to A.D. 400 maximum and minimum age range, respectively (Fig. 8C; Table 1), which is comparable to the age range obtained in excavations II and III for unit e (Figs. 8B and 9A; Table 1). However, as here again the large age range can be due to charcoal mixing, the dating of unit e is obtained by comparison to the dated stratigraphic succession of units in trenches (see section Paleoseismic Trenches).

• from Sbeinati et al. (2010:256-258)

Two trenches, B and C (Figs. 5 and 10, trenches B and C), were dug across the Dead Sea fault north of the aqueduct in addition to the previously studied trench, A (Fig. 10A; Meghraoui et al., 2003). The two trenches exposed an ~1.5-m- wide fault zone that affects a succession of 2–3-m-thick fine and coarse alluvial sedimentary layers similar to the alluvial deposits of trench A. Alluvial units visible in all trenches exhibit here similar textures, structures, and color, and correspond to the same layers that belong to the same alluvial terrace. Although the three trenches A, B, and C may not expose a completed stratigraphic section, the comparisons among sedimentary units, faulting events, archaeoseismic observations, and tufa accumulation limit the possibility of a missing earthquake event that affected the aqueduct.



In trench B (south wall), the fault zone shows three main fault branches that affect sedimentary units g to d and form a negative flower structure. The central and western main branches are truncated by unit a, which forms a stratified 0.3–0.4-m-thick deposit of coarse gravels in a sandy matrix. The eastern fault branch is buried below unit d, made of well-sorted reddish fine gravels. Unit e, a 0.2–0.5-m-thick dark-brown silt-clay, thickens toward east. Units f and g are made of scattered clasts in a massive clay matrix of dark-brown and light-brown color, respectively. Although intense warping and faulting are marked by contrasting color and texture of unit e, faulted sedimentary layers of this trench do not allow the identification of all faulting events. However, buried fault branches indicate a faulting event postdated by unit d (event Y), while the other fault branches show at least another faulting event (event Z) overlain by unit a. While clearly visible in other trench walls, event Y is here likely concealed by the complex fault branches truncated by unit a.



Trench C (Fig. 10C) exposes a stratigraphic succession affected by at least five main fault branches (labeled I to V in Fig. 10C). From trench bottom, fault branch I, which affects unit g, is overlain by unit f. A similar observation can be made for fault branch II, which also affects all units below unit d. Furthermore, the trench wall exposes an ~0.60-m-thick well- stratified, coarse and fine gravel layer above unit e and across the fault zone. Unit d thins significantly west of fault branch III and is overlapped by relatively thick coarse gravel units, which display a mix of fine and coarse gravels between fault branches III and IV, and unit d shows a succession of well-stratified alluvial units west of fault branch IV (Fig. 10C). Taking into account its alluvial origin made of well-stratified fine and coarse gravels, west of fault branch IV, unit d is subdivided into d1, d2, d3, and d4. Faulting movements at this site allows truncation of unit d1 (equivalent to d east of fault branch III) and sedimentation of units d2 to d4 (in a likely small pull-apart basin). Unit d3 consists of an ~0.20-m-thick dark-brown silt-sand overlain by unit d4, which is made of light-brown fine silt-sand. Below the plough zone a2, the well-stratified unit a1 shows flat-laying pebbles and gravels and intercalated fine gravels covering previous units and the fault zone.

Fault branches I to V in trench C indicate a negative flower structure that intersects a sedimentary sequence and reveal at least four faulting events (Fig. 10C):

1. Event W, identified on fault branch I, is older than 800–510 B.C. (EH II-18S) in the lowermost layers of unit f and is younger than unit g, which was dated with sample EH II-5S (3400–300 B.C.).


2. Next to fault branch II, buried below unit d, the vertical offsets between unit e and units d and d1 across fault branch III, and the absence of unit e between fault branches III and IV, determine the faulting event X between unit e and unit d. Since unit d overlies an erosional surface of unit e, faulting event X may have formed a depression (i.e., a small pull-apart basin) that allowed the deposition of d1 to d4 next to a thick unit d east of fault branch III. The faulting event X is here predated by 360–90 B.C. (EH II-12S), 360–50 B.C. (EH II-11S), and 360–60 B.C. (EH II-10S) of the uppermost layers of unit f (event X is postdated by sample EH I-TA-S33 of trench A).


3. Faulting event Y can be identified at the westernmost fault branch V between unit d2 and unit d3. The dating of sample EH II-16S in d3 postdates event X to younger than A.D. 540–650, which we consider as a reliable age, taking into account its high carbon content (event Y is predated by sample EH I-TA-S33 of trench A).


4. Faulting event Z corresponds to the main fault branches III and IV, which are overlain by the stratified unit a2 below the plough zone. Fault rupture IV affects unit d4 and indicates that the faulting event Z is older than radiocarbon age A.D. 1480–1800 (EH II-7S) and A.D. 1510–1670 (EH II-2N) of unit a2 and younger than unit d4.

• from Sbeinati et al. (2010:258)

The analysis of faulting events from the aqueduct (damage and reconstruction) and from trenches A, B, and C can be presented as following:

1. Event W is older than unit f (i.e., 800–510 B.C.) and younger than unit g (i.e., 3400–300 B.C.) of trench C. The bracket of event W is here difficult to assess since the detrital charcoal sample in unit f was not taken from the base of unit f. According to ¹⁴C dates, the faulting event can be estimated as younger than 3400 B.C. and older than 510 B.C. However, taking into account the rate of sedimentation in unit f, we may estimate a minimum age of 962 B.C. for event W.

2. Event X, the first faulting event that affected the aqueduct, is bracketed between the first and sixth centuries A.D. In trenches, a large bracket of this event is between 350 B.C. and A.D. 30 and A.D. 650–810 (as obtained from dated units of trench A).

3. Event Y, characterized from paleoseismology, appears to be older than A.D. 650–810 (unit d, trench A) and younger than A.D. 540–650 (unit d3 in trench C). The results of archaeoseismic investigations indicate that ages of CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) postdate event Y.

4. Event Z is the last faulting event that affected the aqueduct, after which it was definitely abandoned. In trenches A and C, event Z is older than A.D. 1480–1800, A.D. 1510–1670, and A.D. 1030–1260 and younger than A.D. 960–1060.

• from Sbeinati et al. (2010:260-261)

We conducted four archaeoseismic excavations, three paleoseismic trenches, and obtained the radiocarbon dating of six cores at the Al Harif Aqueduct site along the Missyaf segment of the Dead Sea fault. The combined study allows us to obtain a better constraint on the timing of past earthquakes, with four large seismic events during the last ~3400 yr. The occurrence of three seismic events X, Y, and Z (A.D. 70–600, ca. A.D. 650, and A.D. 1170, respectively) since the construction of the aqueduct is attested by faulting events in trenches, the damage and repair of the aqueduct wall, and the tufa growth and interruptions since Roman time (Fig. 13). These results point out a temporal clustering of three large earthquakes between A.D. 70 and A.D. 1170 along the Missyaf fault segment (Fig. 14).

The 90 ± 10-km-long and linear Missyaf segment experienced the A.D. 1170 earthquake recorded in trenches, aqueduct construction, and tufa deposits. In this tectonic framework, the large (10-km-wide) Ghab pull-apart basin to the north and the Al Bouqueaa pull-apart and onset of the restraining bend to the south (Fig. 3) may constitute endpoints for earthquake rupture propagation, as observed for other large continental strike-slip faults (Klinger et al., 2003; Wesnousky, 2006). The size of the Ghab Basin and the sharp bend of the Lebanese fault system may act as structural control of fault-rupture initiation and propa- gation. Furthermore, the damage distribution of the A.D. 1170 earthquake, well located on the Missyaf segment, is limited to the north by the A.D. 1156 large earthquake and to the south by the A.D. 1063 and A.D. 1202 earthquakes (Fig. 2; Sbeinati et al., 2005). The 20-km-thick seismogenic layer (Brew et al., 2001) correlates with the ~90 km fault length estimated from field mapping (Fig. 3). Fault dimensions are consistent with the ~4.3 m maximum characteristic slip inferred from the warping of the aqueduct wall east of the fault (and west of the bridge). Here, we assume that successive faulting episodes maintained the early ~4.3 m warping of an already ruptured strong building. Taking an average 2.0 m coseismic slip along the fault, the obtained seismic moment is Mo = 1.05 × 10²⁰ N m (Mw 7.3; Wells and Coppersmith, 1994), which is comparable, for instance, with the seismic moment of the 1999 Izmit large earthquake (Mw 7.4) of the North Anatolian fault.

• from Sbeinati et al. (2010:262-263)

The consistency among the timing of faulted sedimentary units in trenches, the age of building and repair of the aqueduct wall, and the dating of tufa interruptions and restart episodes determines the completeness of a sequence of earthquake events. The dating of three episodes of fault slip X, Y, and Z is consistent with the two phases of aqueduct wall repair, and the two interruptions of the longest tufa deposits BR-5 and BR-6, and interruptions and restart in BR-3 and BR-4. Our observations indicate that the aqueduct was repaired after the large seismic events X and Y but abandoned after the most recent faulting event Z. Building repair after a damaging earthquake is very often necessary because it is a vital remedial measure of water supply in order to avoid a decline of the local economy (Ambraseys, 2006). The repair has the benefit of leaving critical indicators of previous damage and, in some cases, of the fault slip characteristics.

For instance, the eastern wall of the Al Harif aqueduct shows a clear warping that confirms the left-lateral movement near the fault zone. As observed for coseismic surface ruptures crossing buildings, fences, and walls during large strike-slip earthquakes (Yeats et al., 1997), warped walls that may record a coseismic slip are often observed along strike-slip faulting. Warping that amounts to 4.3 m can be interpreted as the individual coseismic slip during event X. The warping can be due to the opposite lateral movements across the fault constrained by the bridge cohesion to the east and wall solidity to the west. While the western aqueduct wall section was built straight on the flat alluvial terrace and ends abruptly against the fault, only the section between the bridge and the fault zone (which is partly built on loose sediments and bridge ballast) presents some warping and dragging (possibly separated from the alluvial substratum; Fig. 14). The warped section near the bridge displays one generation of cracks filled with tufa that attests to the early bridge damage and possible correlation with event X (Meghraoui et al., 2003). Similar warped walls and fences were observed after the 17 August 1999 earthquake and along the North Anatolia fault in Turkey (Barka et al., 2002). Subsequent faulting movements Y and Z would have affected an already broken aqueduct wall (even if rebuilt) with less strength at the fault zone than for the initial building conditions (Fig. 14). Furthermore, the 4.3 m can be considered as a characteristic slip at the aqueduct site; such characteristic behavior with repeated same amounts of coseismic slip has already been observed and inferred from paleoseismic trenches along major strike-slip faults (Klinger et al., 2003; Rockwell et al., 2009). If the warped aqueduct wall is random and not representative of a coseismic slip, the alternative solution is quite similar if we consider a 4.5 m average individual slip from the cumulative 13.6 m left-lateral offset and the X, Y, and Z large seismic events at the aqueduct site.

• from Sbeinati et al. (2010:263)

Another key issue is the relationship between the aqueduct damage and the start and interruption of tufa accumulation with past earthquakes (Figs. 11 and 13). Indeed, the water flow may be interrupted anytime due to, for instance, the actions of man (warfare) or the onset of a drought period and climatic fluctuations that may influence the water flow. These possibilities seem here unlikely because the only two interruptions in cores BR-5 and BR-6 coincide with earthquake events X and Y, and no other additional interruptions were here recorded. This is also attested by the two interruptions in cores BR-3 and BR-4 that correlate with earthquake events X and Y. The difference between the tufa accumulation in BR-4, BR-5, and BR-6 located on the wall section west of the fault, and BR-3 located on the wall section next to the bridge, east of the fault, provides a consistent aqueduct damage history (Fig. 13). The onset of BR-3 after event X is the sign of an extensive damage that tilted the bridge and allowed overflow with tufa accumulation on the aqueduct northern side. The subsequent interruption (repair) and restart of BR-3 that coincides with event Y illustrate the successive aqueduct damage. Located on the broken western wall section (Fig. 6), the onset of BR-4 after event X and restart after event Y are consistent with BR-3 tufa growth and accumulation. As illustrated in Figure 13, the coincidence among faulting events X, Y, and Z from paleoseismic trenches, the three building damage and repair episodes from archaeoseismic investigations, and tufa growth and interruption constrains the earthquake-induced damage and faulting episodes across the aqueduct.

• from Sbeinati et al. (2010:249-250)

The Al-Harif aqueduct is located ~4 km north of the city of Missyaf, immediately west of a limestone shutter ridge and related ~200 m left-lateral stream deflection. According to the remaining aqueduct walls and related mills in the region, the aqueduct was built during the Roman time (younger than 65 B.C. in the Middle East) to drain freshwater collected from springs of the western mountain range to the eastern semiarid plains. The remaining ruins of the aqueduct suggest an ~40-km-long construction that may have included several bridges over streams and landscape gorges.

The aqueduct building description and related age have not been reported so far in any archive, manuscript, or in the literature. There is, however, an interesting anecdotal story from the local tradition that it was built by a local prince to supply potable water to Apamea and/or Sheizar cities, located northeast of the aqueduct. Apamea during that time was the most famous and strategic city during the Hellenistic and Roman period, whereas Sheizar is known to have been an important political and military fortress during the Middle Ages.

In their description of the Dead Sea fault in Syria, Trifonov et al. (1991) mentioned the existence of a faulted aqueduct near the city of Missyaf, but neither the precise location nor the accurate amount of offset walls was given. However, this early tectonic observation was helpful and allowed us to discover the site and consider a detailed study (Meghraoui et al., 2003), which is extended here using combined methods in archaeoseismology, paleoseismology, and tufa investigations. In addition, a micro- topographic survey of measurements accompanied all field studies.

Previous investigations on the aqueduct (Meghraoui et al., 2003) established: (1) an evaluation of its age based on an account of the large size blocks, the dating of sedimentary units below the aqueduct wall foundation, and dating of early tufa deposits on the aqueduct wall, and (2) the identification of the seismic faulting origin of damage in nearby trench A. The building style, with typical bridge arch and large stone size disposition (Opus caementum), suggested a Roman age, which was confirmed by the radiocarbon dating of sedimentary layers below the walls and the early tufa deposits on the walls. The faulted aqueduct revealed 13.6 ± 0.20 m of total left-lateral offset and called for detailed investigations on the characteristics and history of successive fault movements.

The aqueduct design, with an open canal on top of the 4-m-high wall, allowed freshwater and carbonate-saturated water to overflow and induce significant tufa accumulation from 0.30 m to 0.83 m in section. The carbonate-rich and cool water collected from the nearby western range is associated with a semiarid and karstic area of the Mesozoic limestone that favors rapid carbonate precipitation and tufa accumulation. The tufa deposits show successive growths of lamination carbonate with high porosity, banded texture, and rich organic encrustations. Field observations show that tufa accumulation developed on both eastern and western sections (from the fault line), but only on the north-facing wall, likely due to a slight tilt of the damaged aqueduct wall, probably after the two first earthquakes.

The following paragraphs present the field investigations, which consisted of:

1. four archaeoseismic excavations near the aqueduct walls and remains
2. four paleoseismic trenches across the fault zone and the alluvial sediments
3. four cores (two cores were previously studied in Meghraoui et al., 2003) of tufa deposits collected from different sections of the aqueduct.

More than 200 samples of organic matter, charcoal fragments, and tufa core pieces were taken for radiocarbon analysis in order to characterize the timing of successive faulting and related damage of the aqueduct construction. All radiocarbon dating were calibrated (2σ range, 95.4% probability density) using Oxcal v4.0 and INTCAL04 calibration curve.

• from Sbeinati et al. (2010:251-256)

The remaining aqueduct construction forms an ~50-m-long, ~5-m-high, and 0.60-m-thick wall that includes an ~15-m-high arch bridge in its eastern section (Figs. 5 and 6A). The outer part is coated by a thick layer of tufa deposits, probably due to a long period of freshwater flow. The construction material that may vary with the successive building and repair ages is made of:

1. large-size limestone blocks (Opus quadratum, 1.0 m × 0.5 m × 0.5 m; see also https://www.romanaqueducts.info/aquasite), similar to the typical Roman archaeological constructions and visible at the lower bridge (pier section) and wall sections


2. medium-size limestone blocks (Opus incertum; 0.50 m × 0.30 m × 0.30 m), which form the foundation or the upper half wall section and show visible small portions of cement


3. small sizes of mixed stones of irregular shape with significant portions of mortar (cement), mostly visible in the apparently rebuilt part of the wall

Figures 5 and 6A also show a detached small piece of the aqueduct wall made of small-size stones and related cement ~3.5 m away from the eastern wall. Therefore, four areas (noted I to IV in Fig. 5) were excavated near the aqueduct using proper archaeological methods.

The large excavation I was dug on the fault zone near the dragged wall fragment, in the area between the eastern and western aqueduct walls (Figs. 5, 8A, and 8B). The purpose of excavation is here to study the relationships between the fault zone and aqueduct. The excavation that has ~4.5 × 4.5 m surface and ~0.6 m depth exposed missing parts of the aqueduct. A buried and fallen wall piece rotated and dragged parallel to the fault and a remaining wall piece in an oblique position between two shear zones were discovered. The buried wall fragments are not comparable to the Opus caementum (quadratum) of the original construction and suggest a rebuilding phase. The excavation floor displays oriented gravels and pebbles that mark the shear zones and related fault branches also visible in the inner trench section E (Figs. 8B and 8C).

We collected four samples in the fallen wall sections labeled A, B, and C of excavation I (Fig. 8B): Two cement samples (AQ-CS-1 and AQ-CS-4) found in between building stones are made of typical medieval rubble mortars (mainly mud, gypsum, and lime); the two other samples (AQ-CS-3-2 and AQ-CS-3-3) are tufa deposits preserved on building stones. All four samples contained enough organic matter to allow radiocarbon dating (Table 1). Two dates of cement yield A.D. 532–641 (section A, AQ-CS-4) for the large fallen wall in excavation I and A.D. 650–780 (section C, AQ-CS-1) for the wall fragment piece in between the walls (Fig. 8B). In addition, two tufa deposits on wall stones provide consistent ages A.D. 560–690 (section B, AQ-CS-3-2) and A.D. 639–883 (section C, AQ-CS-3-3) with cement ages. The two different cement dates of the fallen wall and dragged wall fragment can be correlated to the new tufa deposits that testify for two rebuilding phases. The dated buried fallen wall in section B (CS-3-2) obtained from a thin (~5 cm) tufa accumulation correlates with the similarly fallen wall in section A and related cement date of CS-4 (Fig. 8B; and Table 1). In section C, the tufa deposits and related dated sample CS-3-3 correlates with the cement age of CS-1. The type and size of stones (opus incertum) and thin tufa accumulation in sections A and B suggest an early rebuilding phase postdating the first damaging event that may have occurred between the first and sixth centuries A.D. The different building layout of section C made of small sizes of mixed stones of irregular shape, and dating of cement sample CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) indicate a repair and rebuilding period postdating a second damaging event at the end of the Byzantine time and beginning of the Islamic period (seventh to eighth century A.D.). The damaged and dragged most recent wall section C along the fault indicates the occurrence of a third event, after which the aqueduct was definitely abandoned.

Three small excavations II, III, and IV (1.5 m to 3.0 m long, 1.0 m wide and 1.50 m deep) were dug in the base layer of the western aqueduct wall in order to expose its foundation and related sedimentary units underneath that predate the early building phase (Figs. 5 and 9). Excavations II and III were dug under the wall section with maximum (> 0.80 m) and minimum (~0.30 m) thickness of tufa deposition, respectively (Figs. 9A and 9B). Excavation IV, already described in Meghraoui et al. (2003), exposed the faulted foundation of the missing section of the western wall edge. The wall foundation reaches 1 m depth and shows regular patterns of medium-size cut limestone blocks (0.50 m × 0.30 m × 0.30 m) built over a dark brown clayey layer (unit e).

Charcoal samples collected in excavations I (trench E), II, and III from unit e yield ¹⁴C dates with ages spanning from approximately the third century B.C. to third century A.D. (see samples AQ-TA, TB, and TC in Table 1; Figs. 8C, 9A, and 9B). Although in these excavations, the age range of unit e seems quite large (probably due to detrital charcoal mixing), the younger age, i.e., 350 B.C. to A.D. 130 (sample AQ-TA-4), is consistent with other radiocarbon ages of unit e and related stratigraphic succession in trenches (see Paleoseismic Trenches herein). In excavation II, the large stone shape (Opus quadratum) with small amount of cementing material and pottery fragments found on the same level near the building base can be correlated with the early Roman era (Fig. 9A). Large stones and tufa thickness led us to consider this section of the aqueduct wall to be in original condition, i.e., probably undamaged by large earthquakes.

Excavation III (1.65 m long, 1.0 m wide, and 1.2 m deep; Fig. 9B) is similar to excavations II and IV, but the 1-m-deep wall foundation and upper section show irregular shapes of mixed medium- and small-size cut limestone blocks (0.10 × 0.20 × 0.15 m). Excavation III was realized at the location of the thinnest tufa deposits (< 0.30 m). The size of stones, cement texture, and irregular shape of building wall suggest that this building section was rebuilt (Fig. 9B). The ¹⁴C dating of unit e below the wall yielded a comparable age range to that obtained in excavations II (see AQ-TA, TB, and TC in Table 1).

Trench section E (4.30 m long, 0.70 m wide, and 1.30 m deep; Figs. 5, 8A, 8B, and 8C) was dug within excavation I in order to see in section the fault zone that affects the archaeological floor units. The trench wall exposes similar sedimentary units to those visible in excavations II, III, and IV that are affected by two main fault branches of the shear zone visible in the floor layer of excavation I. The ¹⁴C dating of samples AQ-TC-S1, S2, and S3 of units f and e indicates 900 B.C. to A.D. 400 maximum and minimum age range, respectively (Fig. 8C; Table 1), which is comparable to the age range obtained in excavations II and III for unit e (Figs. 8B and 9A; Table 1). However, as here again the large age range can be due to charcoal mixing, the dating of unit e is obtained by comparison to the dated stratigraphic succession of units in trenches (see section Paleoseismic Trenches).

• from Sbeinati et al. (2010:256-258)

Two trenches, B and C (Figs. 5 and 10, trenches B and C), were dug across the Dead Sea fault north of the aqueduct in addition to the previously studied trench, A (Fig. 10A; Meghraoui et al., 2003). The two trenches exposed an ~1.5-m- wide fault zone that affects a succession of 2–3-m-thick fine and coarse alluvial sedimentary layers similar to the alluvial deposits of trench A. Alluvial units visible in all trenches exhibit here similar textures, structures, and color, and correspond to the same layers that belong to the same alluvial terrace. Although the three trenches A, B, and C may not expose a completed stratigraphic section, the comparisons among sedimentary units, faulting events, archaeoseismic observations, and tufa accumulation limit the possibility of a missing earthquake event that affected the aqueduct.



In trench B (south wall), the fault zone shows three main fault branches that affect sedimentary units g to d and form a negative flower structure. The central and western main branches are truncated by unit a, which forms a stratified 0.3–0.4-m-thick deposit of coarse gravels in a sandy matrix. The eastern fault branch is buried below unit d, made of well-sorted reddish fine gravels. Unit e, a 0.2–0.5-m-thick dark-brown silt-clay, thickens toward east. Units f and g are made of scattered clasts in a massive clay matrix of dark-brown and light-brown color, respectively. Although intense warping and faulting are marked by contrasting color and texture of unit e, faulted sedimentary layers of this trench do not allow the identification of all faulting events. However, buried fault branches indicate a faulting event postdated by unit d (event Y), while the other fault branches show at least another faulting event (event Z) overlain by unit a. While clearly visible in other trench walls, event Y is here likely concealed by the complex fault branches truncated by unit a.



Trench C (Fig. 10C) exposes a stratigraphic succession affected by at least five main fault branches (labeled I to V in Fig. 10C). From trench bottom, fault branch I, which affects unit g, is overlain by unit f. A similar observation can be made for fault branch II, which also affects all units below unit d. Furthermore, the trench wall exposes an ~0.60-m-thick well- stratified, coarse and fine gravel layer above unit e and across the fault zone. Unit d thins significantly west of fault branch III and is overlapped by relatively thick coarse gravel units, which display a mix of fine and coarse gravels between fault branches III and IV, and unit d shows a succession of well-stratified alluvial units west of fault branch IV (Fig. 10C). Taking into account its alluvial origin made of well-stratified fine and coarse gravels, west of fault branch IV, unit d is subdivided into d1, d2, d3, and d4. Faulting movements at this site allows truncation of unit d1 (equivalent to d east of fault branch III) and sedimentation of units d2 to d4 (in a likely small pull-apart basin). Unit d3 consists of an ~0.20-m-thick dark-brown silt-sand overlain by unit d4, which is made of light-brown fine silt-sand. Below the plough zone a2, the well-stratified unit a1 shows flat-laying pebbles and gravels and intercalated fine gravels covering previous units and the fault zone.

Fault branches I to V in trench C indicate a negative flower structure that intersects a sedimentary sequence and reveal at least four faulting events (Fig. 10C):

1. Event W, identified on fault branch I, is older than 800–510 B.C. (EH II-18S) in the lowermost layers of unit f and is younger than unit g, which was dated with sample EH II-5S (3400–300 B.C.).


2. Next to fault branch II, buried below unit d, the vertical offsets between unit e and units d and d1 across fault branch III, and the absence of unit e between fault branches III and IV, determine the faulting event X between unit e and unit d. Since unit d overlies an erosional surface of unit e, faulting event X may have formed a depression (i.e., a small pull-apart basin) that allowed the deposition of d1 to d4 next to a thick unit d east of fault branch III. The faulting event X is here predated by 360–90 B.C. (EH II-12S), 360–50 B.C. (EH II-11S), and 360–60 B.C. (EH II-10S) of the uppermost layers of unit f (event X is postdated by sample EH I-TA-S33 of trench A).


3. Faulting event Y can be identified at the westernmost fault branch V between unit d2 and unit d3. The dating of sample EH II-16S in d3 postdates event X to younger than A.D. 540–650, which we consider as a reliable age, taking into account its high carbon content (event Y is predated by sample EH I-TA-S33 of trench A).


4. Faulting event Z corresponds to the main fault branches III and IV, which are overlain by the stratified unit a2 below the plough zone. Fault rupture IV affects unit d4 and indicates that the faulting event Z is older than radiocarbon age A.D. 1480–1800 (EH II-7S) and A.D. 1510–1670 (EH II-2N) of unit a2 and younger than unit d4.

• from Sbeinati et al. (2010:258)

The analysis of faulting events from the aqueduct (damage and reconstruction) and from trenches A, B, and C can be presented as following:

1. Event W is older than unit f (i.e., 800–510 B.C.) and younger than unit g (i.e., 3400–300 B.C.) of trench C. The bracket of event W is here difficult to assess since the detrital charcoal sample in unit f was not taken from the base of unit f. According to ¹⁴C dates, the faulting event can be estimated as younger than 3400 B.C. and older than 510 B.C. However, taking into account the rate of sedimentation in unit f, we may estimate a minimum age of 962 B.C. for event W.

2. Event X, the first faulting event that affected the aqueduct, is bracketed between the first and sixth centuries A.D. In trenches, a large bracket of this event is between 350 B.C. and A.D. 30 and A.D. 650–810 (as obtained from dated units of trench A).

3. Event Y, characterized from paleoseismology, appears to be older than A.D. 650–810 (unit d, trench A) and younger than A.D. 540–650 (unit d3 in trench C). The results of archaeoseismic investigations indicate that ages of CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) postdate event Y.

4. Event Z is the last faulting event that affected the aqueduct, after which it was definitely abandoned. In trenches A and C, event Z is older than A.D. 1480–1800, A.D. 1510–1670, and A.D. 1030–1260 and younger than A.D. 960–1060.

• from Sbeinati et al. (2010:260-261)

We conducted four archaeoseismic excavations, three paleoseismic trenches, and obtained the radiocarbon dating of six cores at the Al Harif Aqueduct site along the Missyaf segment of the Dead Sea fault. The combined study allows us to obtain a better constraint on the timing of past earthquakes, with four large seismic events during the last ~3400 yr. The occurrence of three seismic events X, Y, and Z (A.D. 70–600, ca. A.D. 650, and A.D. 1170, respectively) since the construction of the aqueduct is attested by faulting events in trenches, the damage and repair of the aqueduct wall, and the tufa growth and interruptions since Roman time (Fig. 13). These results point out a temporal clustering of three large earthquakes between A.D. 70 and A.D. 1170 along the Missyaf fault segment (Fig. 14).

The 90 ± 10-km-long and linear Missyaf segment experienced the A.D. 1170 earthquake recorded in trenches, aqueduct construction, and tufa deposits. In this tectonic framework, the large (10-km-wide) Ghab pull-apart basin to the north and the Al Bouqueaa pull-apart and onset of the restraining bend to the south (Fig. 3) may constitute endpoints for earthquake rupture propagation, as observed for other large continental strike-slip faults (Klinger et al., 2003; Wesnousky, 2006). The size of the Ghab Basin and the sharp bend of the Lebanese fault system may act as structural control of fault-rupture initiation and propa- gation. Furthermore, the damage distribution of the A.D. 1170 earthquake, well located on the Missyaf segment, is limited to the north by the A.D. 1156 large earthquake and to the south by the A.D. 1063 and A.D. 1202 earthquakes (Fig. 2; Sbeinati et al., 2005). The 20-km-thick seismogenic layer (Brew et al., 2001) correlates with the ~90 km fault length estimated from field mapping (Fig. 3). Fault dimensions are consistent with the ~4.3 m maximum characteristic slip inferred from the warping of the aqueduct wall east of the fault (and west of the bridge). Here, we assume that successive faulting episodes maintained the early ~4.3 m warping of an already ruptured strong building. Taking an average 2.0 m coseismic slip along the fault, the obtained seismic moment is Mo = 1.05 × 10²⁰ N m (Mw 7.3; Wells and Coppersmith, 1994), which is comparable, for instance, with the seismic moment of the 1999 Izmit large earthquake (Mw 7.4) of the North Anatolian fault.

• from Sbeinati et al. (2010:262-263)

The consistency among the timing of faulted sedimentary units in trenches, the age of building and repair of the aqueduct wall, and the dating of tufa interruptions and restart episodes determines the completeness of a sequence of earthquake events. The dating of three episodes of fault slip X, Y, and Z is consistent with the two phases of aqueduct wall repair, and the two interruptions of the longest tufa deposits BR-5 and BR-6, and interruptions and restart in BR-3 and BR-4. Our observations indicate that the aqueduct was repaired after the large seismic events X and Y but abandoned after the most recent faulting event Z. Building repair after a damaging earthquake is very often necessary because it is a vital remedial measure of water supply in order to avoid a decline of the local economy (Ambraseys, 2006). The repair has the benefit of leaving critical indicators of previous damage and, in some cases, of the fault slip characteristics.

For instance, the eastern wall of the Al Harif aqueduct shows a clear warping that confirms the left-lateral movement near the fault zone. As observed for coseismic surface ruptures crossing buildings, fences, and walls during large strike-slip earthquakes (Yeats et al., 1997), warped walls that may record a coseismic slip are often observed along strike-slip faulting. Warping that amounts to 4.3 m can be interpreted as the individual coseismic slip during event X. The warping can be due to the opposite lateral movements across the fault constrained by the bridge cohesion to the east and wall solidity to the west. While the western aqueduct wall section was built straight on the flat alluvial terrace and ends abruptly against the fault, only the section between the bridge and the fault zone (which is partly built on loose sediments and bridge ballast) presents some warping and dragging (possibly separated from the alluvial substratum; Fig. 14). The warped section near the bridge displays one generation of cracks filled with tufa that attests to the early bridge damage and possible correlation with event X (Meghraoui et al., 2003). Similar warped walls and fences were observed after the 17 August 1999 earthquake and along the North Anatolia fault in Turkey (Barka et al., 2002). Subsequent faulting movements Y and Z would have affected an already broken aqueduct wall (even if rebuilt) with less strength at the fault zone than for the initial building conditions (Fig. 14). Furthermore, the 4.3 m can be considered as a characteristic slip at the aqueduct site; such characteristic behavior with repeated same amounts of coseismic slip has already been observed and inferred from paleoseismic trenches along major strike-slip faults (Klinger et al., 2003; Rockwell et al., 2009). If the warped aqueduct wall is random and not representative of a coseismic slip, the alternative solution is quite similar if we consider a 4.5 m average individual slip from the cumulative 13.6 m left-lateral offset and the X, Y, and Z large seismic events at the aqueduct site.

• from Sbeinati et al. (2010:263)

Another key issue is the relationship between the aqueduct damage and the start and interruption of tufa accumulation with past earthquakes (Figs. 11 and 13). Indeed, the water flow may be interrupted anytime due to, for instance, the actions of man (warfare) or the onset of a drought period and climatic fluctuations that may influence the water flow. These possibilities seem here unlikely because the only two interruptions in cores BR-5 and BR-6 coincide with earthquake events X and Y, and no other additional interruptions were here recorded. This is also attested by the two interruptions in cores BR-3 and BR-4 that correlate with earthquake events X and Y. The difference between the tufa accumulation in BR-4, BR-5, and BR-6 located on the wall section west of the fault, and BR-3 located on the wall section next to the bridge, east of the fault, provides a consistent aqueduct damage history (Fig. 13). The onset of BR-3 after event X is the sign of an extensive damage that tilted the bridge and allowed overflow with tufa accumulation on the aqueduct northern side. The subsequent interruption (repair) and restart of BR-3 that coincides with event Y illustrate the successive aqueduct damage. Located on the broken western wall section (Fig. 6), the onset of BR-4 after event X and restart after event Y are consistent with BR-3 tufa growth and accumulation. As illustrated in Figure 13, the coincidence among faulting events X, Y, and Z from paleoseismic trenches, the three building damage and repair episodes from archaeoseismic investigations, and tufa growth and interruption constrains the earthquake-induced damage and faulting episodes across the aqueduct.

• from Sbeinati et al. (2010:249-250)

The Al-Harif aqueduct is located ~4 km north of the city of Missyaf, immediately west of a limestone shutter ridge and related ~200 m left-lateral stream deflection. According to the remaining aqueduct walls and related mills in the region, the aqueduct was built during the Roman time (younger than 65 B.C. in the Middle East) to drain freshwater collected from springs of the western mountain range to the eastern semiarid plains. The remaining ruins of the aqueduct suggest an ~40-km-long construction that may have included several bridges over streams and landscape gorges.

The aqueduct building description and related age have not been reported so far in any archive, manuscript, or in the literature. There is, however, an interesting anecdotal story from the local tradition that it was built by a local prince to supply potable water to Apamea and/or Sheizar cities, located northeast of the aqueduct. Apamea during that time was the most famous and strategic city during the Hellenistic and Roman period, whereas Sheizar is known to have been an important political and military fortress during the Middle Ages.

In their description of the Dead Sea fault in Syria, Trifonov et al. (1991) mentioned the existence of a faulted aqueduct near the city of Missyaf, but neither the precise location nor the accurate amount of offset walls was given. However, this early tectonic observation was helpful and allowed us to discover the site and consider a detailed study (Meghraoui et al., 2003), which is extended here using combined methods in archaeoseismology, paleoseismology, and tufa investigations. In addition, a micro- topographic survey of measurements accompanied all field studies.

Previous investigations on the aqueduct (Meghraoui et al., 2003) established: (1) an evaluation of its age based on an account of the large size blocks, the dating of sedimentary units below the aqueduct wall foundation, and dating of early tufa deposits on the aqueduct wall, and (2) the identification of the seismic faulting origin of damage in nearby trench A. The building style, with typical bridge arch and large stone size disposition (Opus caementum), suggested a Roman age, which was confirmed by the radiocarbon dating of sedimentary layers below the walls and the early tufa deposits on the walls. The faulted aqueduct revealed 13.6 ± 0.20 m of total left-lateral offset and called for detailed investigations on the characteristics and history of successive fault movements.

The aqueduct design, with an open canal on top of the 4-m-high wall, allowed freshwater and carbonate-saturated water to overflow and induce significant tufa accumulation from 0.30 m to 0.83 m in section. The carbonate-rich and cool water collected from the nearby western range is associated with a semiarid and karstic area of the Mesozoic limestone that favors rapid carbonate precipitation and tufa accumulation. The tufa deposits show successive growths of lamination carbonate with high porosity, banded texture, and rich organic encrustations. Field observations show that tufa accumulation developed on both eastern and western sections (from the fault line), but only on the north-facing wall, likely due to a slight tilt of the damaged aqueduct wall, probably after the two first earthquakes.

The following paragraphs present the field investigations, which consisted of:

1. four archaeoseismic excavations near the aqueduct walls and remains
2. four paleoseismic trenches across the fault zone and the alluvial sediments
3. four cores (two cores were previously studied in Meghraoui et al., 2003) of tufa deposits collected from different sections of the aqueduct.

More than 200 samples of organic matter, charcoal fragments, and tufa core pieces were taken for radiocarbon analysis in order to characterize the timing of successive faulting and related damage of the aqueduct construction. All radiocarbon dating were calibrated (2σ range, 95.4% probability density) using Oxcal v4.0 and INTCAL04 calibration curve.

• from Sbeinati et al. (2010:251-256)

The remaining aqueduct construction forms an ~50-m-long, ~5-m-high, and 0.60-m-thick wall that includes an ~15-m-high arch bridge in its eastern section (Figs. 5 and 6A). The outer part is coated by a thick layer of tufa deposits, probably due to a long period of freshwater flow. The construction material that may vary with the successive building and repair ages is made of:

1. large-size limestone blocks (Opus quadratum, 1.0 m × 0.5 m × 0.5 m; see also https://www.romanaqueducts.info/aquasite), similar to the typical Roman archaeological constructions and visible at the lower bridge (pier section) and wall sections


2. medium-size limestone blocks (Opus incertum; 0.50 m × 0.30 m × 0.30 m), which form the foundation or the upper half wall section and show visible small portions of cement


3. small sizes of mixed stones of irregular shape with significant portions of mortar (cement), mostly visible in the apparently rebuilt part of the wall

Figures 5 and 6A also show a detached small piece of the aqueduct wall made of small-size stones and related cement ~3.5 m away from the eastern wall. Therefore, four areas (noted I to IV in Fig. 5) were excavated near the aqueduct using proper archaeological methods.

The large excavation I was dug on the fault zone near the dragged wall fragment, in the area between the eastern and western aqueduct walls (Figs. 5, 8A, and 8B). The purpose of excavation is here to study the relationships between the fault zone and aqueduct. The excavation that has ~4.5 × 4.5 m surface and ~0.6 m depth exposed missing parts of the aqueduct. A buried and fallen wall piece rotated and dragged parallel to the fault and a remaining wall piece in an oblique position between two shear zones were discovered. The buried wall fragments are not comparable to the Opus caementum (quadratum) of the original construction and suggest a rebuilding phase. The excavation floor displays oriented gravels and pebbles that mark the shear zones and related fault branches also visible in the inner trench section E (Figs. 8B and 8C).

We collected four samples in the fallen wall sections labeled A, B, and C of excavation I (Fig. 8B): Two cement samples (AQ-CS-1 and AQ-CS-4) found in between building stones are made of typical medieval rubble mortars (mainly mud, gypsum, and lime); the two other samples (AQ-CS-3-2 and AQ-CS-3-3) are tufa deposits preserved on building stones. All four samples contained enough organic matter to allow radiocarbon dating (Table 1). Two dates of cement yield A.D. 532–641 (section A, AQ-CS-4) for the large fallen wall in excavation I and A.D. 650–780 (section C, AQ-CS-1) for the wall fragment piece in between the walls (Fig. 8B). In addition, two tufa deposits on wall stones provide consistent ages A.D. 560–690 (section B, AQ-CS-3-2) and A.D. 639–883 (section C, AQ-CS-3-3) with cement ages. The two different cement dates of the fallen wall and dragged wall fragment can be correlated to the new tufa deposits that testify for two rebuilding phases. The dated buried fallen wall in section B (CS-3-2) obtained from a thin (~5 cm) tufa accumulation correlates with the similarly fallen wall in section A and related cement date of CS-4 (Fig. 8B; and Table 1). In section C, the tufa deposits and related dated sample CS-3-3 correlates with the cement age of CS-1. The type and size of stones (opus incertum) and thin tufa accumulation in sections A and B suggest an early rebuilding phase postdating the first damaging event that may have occurred between the first and sixth centuries A.D. The different building layout of section C made of small sizes of mixed stones of irregular shape, and dating of cement sample CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) indicate a repair and rebuilding period postdating a second damaging event at the end of the Byzantine time and beginning of the Islamic period (seventh to eighth century A.D.). The damaged and dragged most recent wall section C along the fault indicates the occurrence of a third event, after which the aqueduct was definitely abandoned.

Three small excavations II, III, and IV (1.5 m to 3.0 m long, 1.0 m wide and 1.50 m deep) were dug in the base layer of the western aqueduct wall in order to expose its foundation and related sedimentary units underneath that predate the early building phase (Figs. 5 and 9). Excavations II and III were dug under the wall section with maximum (> 0.80 m) and minimum (~0.30 m) thickness of tufa deposition, respectively (Figs. 9A and 9B). Excavation IV, already described in Meghraoui et al. (2003), exposed the faulted foundation of the missing section of the western wall edge. The wall foundation reaches 1 m depth and shows regular patterns of medium-size cut limestone blocks (0.50 m × 0.30 m × 0.30 m) built over a dark brown clayey layer (unit e).

Charcoal samples collected in excavations I (trench E), II, and III from unit e yield ¹⁴C dates with ages spanning from approximately the third century B.C. to third century A.D. (see samples AQ-TA, TB, and TC in Table 1; Figs. 8C, 9A, and 9B). Although in these excavations, the age range of unit e seems quite large (probably due to detrital charcoal mixing), the younger age, i.e., 350 B.C. to A.D. 130 (sample AQ-TA-4), is consistent with other radiocarbon ages of unit e and related stratigraphic succession in trenches (see Paleoseismic Trenches herein). In excavation II, the large stone shape (Opus quadratum) with small amount of cementing material and pottery fragments found on the same level near the building base can be correlated with the early Roman era (Fig. 9A). Large stones and tufa thickness led us to consider this section of the aqueduct wall to be in original condition, i.e., probably undamaged by large earthquakes.

Excavation III (1.65 m long, 1.0 m wide, and 1.2 m deep; Fig. 9B) is similar to excavations II and IV, but the 1-m-deep wall foundation and upper section show irregular shapes of mixed medium- and small-size cut limestone blocks (0.10 × 0.20 × 0.15 m). Excavation III was realized at the location of the thinnest tufa deposits (< 0.30 m). The size of stones, cement texture, and irregular shape of building wall suggest that this building section was rebuilt (Fig. 9B). The ¹⁴C dating of unit e below the wall yielded a comparable age range to that obtained in excavations II (see AQ-TA, TB, and TC in Table 1).

Trench section E (4.30 m long, 0.70 m wide, and 1.30 m deep; Figs. 5, 8A, 8B, and 8C) was dug within excavation I in order to see in section the fault zone that affects the archaeological floor units. The trench wall exposes similar sedimentary units to those visible in excavations II, III, and IV that are affected by two main fault branches of the shear zone visible in the floor layer of excavation I. The ¹⁴C dating of samples AQ-TC-S1, S2, and S3 of units f and e indicates 900 B.C. to A.D. 400 maximum and minimum age range, respectively (Fig. 8C; Table 1), which is comparable to the age range obtained in excavations II and III for unit e (Figs. 8B and 9A; Table 1). However, as here again the large age range can be due to charcoal mixing, the dating of unit e is obtained by comparison to the dated stratigraphic succession of units in trenches (see section Paleoseismic Trenches).

• from Sbeinati et al. (2010:256-258)

Two trenches, B and C (Figs. 5 and 10, trenches B and C), were dug across the Dead Sea fault north of the aqueduct in addition to the previously studied trench, A (Fig. 10A; Meghraoui et al., 2003). The two trenches exposed an ~1.5-m- wide fault zone that affects a succession of 2–3-m-thick fine and coarse alluvial sedimentary layers similar to the alluvial deposits of trench A. Alluvial units visible in all trenches exhibit here similar textures, structures, and color, and correspond to the same layers that belong to the same alluvial terrace. Although the three trenches A, B, and C may not expose a completed stratigraphic section, the comparisons among sedimentary units, faulting events, archaeoseismic observations, and tufa accumulation limit the possibility of a missing earthquake event that affected the aqueduct.



In trench B (south wall), the fault zone shows three main fault branches that affect sedimentary units g to d and form a negative flower structure. The central and western main branches are truncated by unit a, which forms a stratified 0.3–0.4-m-thick deposit of coarse gravels in a sandy matrix. The eastern fault branch is buried below unit d, made of well-sorted reddish fine gravels. Unit e, a 0.2–0.5-m-thick dark-brown silt-clay, thickens toward east. Units f and g are made of scattered clasts in a massive clay matrix of dark-brown and light-brown color, respectively. Although intense warping and faulting are marked by contrasting color and texture of unit e, faulted sedimentary layers of this trench do not allow the identification of all faulting events. However, buried fault branches indicate a faulting event postdated by unit d (event Y), while the other fault branches show at least another faulting event (event Z) overlain by unit a. While clearly visible in other trench walls, event Y is here likely concealed by the complex fault branches truncated by unit a.



Trench C (Fig. 10C) exposes a stratigraphic succession affected by at least five main fault branches (labeled I to V in Fig. 10C). From trench bottom, fault branch I, which affects unit g, is overlain by unit f. A similar observation can be made for fault branch II, which also affects all units below unit d. Furthermore, the trench wall exposes an ~0.60-m-thick well- stratified, coarse and fine gravel layer above unit e and across the fault zone. Unit d thins significantly west of fault branch III and is overlapped by relatively thick coarse gravel units, which display a mix of fine and coarse gravels between fault branches III and IV, and unit d shows a succession of well-stratified alluvial units west of fault branch IV (Fig. 10C). Taking into account its alluvial origin made of well-stratified fine and coarse gravels, west of fault branch IV, unit d is subdivided into d1, d2, d3, and d4. Faulting movements at this site allows truncation of unit d1 (equivalent to d east of fault branch III) and sedimentation of units d2 to d4 (in a likely small pull-apart basin). Unit d3 consists of an ~0.20-m-thick dark-brown silt-sand overlain by unit d4, which is made of light-brown fine silt-sand. Below the plough zone a2, the well-stratified unit a1 shows flat-laying pebbles and gravels and intercalated fine gravels covering previous units and the fault zone.

Fault branches I to V in trench C indicate a negative flower structure that intersects a sedimentary sequence and reveal at least four faulting events (Fig. 10C):

1. Event W, identified on fault branch I, is older than 800–510 B.C. (EH II-18S) in the lowermost layers of unit f and is younger than unit g, which was dated with sample EH II-5S (3400–300 B.C.).


2. Next to fault branch II, buried below unit d, the vertical offsets between unit e and units d and d1 across fault branch III, and the absence of unit e between fault branches III and IV, determine the faulting event X between unit e and unit d. Since unit d overlies an erosional surface of unit e, faulting event X may have formed a depression (i.e., a small pull-apart basin) that allowed the deposition of d1 to d4 next to a thick unit d east of fault branch III. The faulting event X is here predated by 360–90 B.C. (EH II-12S), 360–50 B.C. (EH II-11S), and 360–60 B.C. (EH II-10S) of the uppermost layers of unit f (event X is postdated by sample EH I-TA-S33 of trench A).


3. Faulting event Y can be identified at the westernmost fault branch V between unit d2 and unit d3. The dating of sample EH II-16S in d3 postdates event X to younger than A.D. 540–650, which we consider as a reliable age, taking into account its high carbon content (event Y is predated by sample EH I-TA-S33 of trench A).


4. Faulting event Z corresponds to the main fault branches III and IV, which are overlain by the stratified unit a2 below the plough zone. Fault rupture IV affects unit d4 and indicates that the faulting event Z is older than radiocarbon age A.D. 1480–1800 (EH II-7S) and A.D. 1510–1670 (EH II-2N) of unit a2 and younger than unit d4.

• from Sbeinati et al. (2010:258)

The analysis of faulting events from the aqueduct (damage and reconstruction) and from trenches A, B, and C can be presented as following:

1. Event W is older than unit f (i.e., 800–510 B.C.) and younger than unit g (i.e., 3400–300 B.C.) of trench C. The bracket of event W is here difficult to assess since the detrital charcoal sample in unit f was not taken from the base of unit f. According to ¹⁴C dates, the faulting event can be estimated as younger than 3400 B.C. and older than 510 B.C. However, taking into account the rate of sedimentation in unit f, we may estimate a minimum age of 962 B.C. for event W.

2. Event X, the first faulting event that affected the aqueduct, is bracketed between the first and sixth centuries A.D. In trenches, a large bracket of this event is between 350 B.C. and A.D. 30 and A.D. 650–810 (as obtained from dated units of trench A).

3. Event Y, characterized from paleoseismology, appears to be older than A.D. 650–810 (unit d, trench A) and younger than A.D. 540–650 (unit d3 in trench C). The results of archaeoseismic investigations indicate that ages of CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) postdate event Y.

4. Event Z is the last faulting event that affected the aqueduct, after which it was definitely abandoned. In trenches A and C, event Z is older than A.D. 1480–1800, A.D. 1510–1670, and A.D. 1030–1260 and younger than A.D. 960–1060.

• from Sbeinati et al. (2010:260-261)

We conducted four archaeoseismic excavations, three paleoseismic trenches, and obtained the radiocarbon dating of six cores at the Al Harif Aqueduct site along the Missyaf segment of the Dead Sea fault. The combined study allows us to obtain a better constraint on the timing of past earthquakes, with four large seismic events during the last ~3400 yr. The occurrence of three seismic events X, Y, and Z (A.D. 70–600, ca. A.D. 650, and A.D. 1170, respectively) since the construction of the aqueduct is attested by faulting events in trenches, the damage and repair of the aqueduct wall, and the tufa growth and interruptions since Roman time (Fig. 13). These results point out a temporal clustering of three large earthquakes between A.D. 70 and A.D. 1170 along the Missyaf fault segment (Fig. 14).

The 90 ± 10-km-long and linear Missyaf segment experienced the A.D. 1170 earthquake recorded in trenches, aqueduct construction, and tufa deposits. In this tectonic framework, the large (10-km-wide) Ghab pull-apart basin to the north and the Al Bouqueaa pull-apart and onset of the restraining bend to the south (Fig. 3) may constitute endpoints for earthquake rupture propagation, as observed for other large continental strike-slip faults (Klinger et al., 2003; Wesnousky, 2006). The size of the Ghab Basin and the sharp bend of the Lebanese fault system may act as structural control of fault-rupture initiation and propa- gation. Furthermore, the damage distribution of the A.D. 1170 earthquake, well located on the Missyaf segment, is limited to the north by the A.D. 1156 large earthquake and to the south by the A.D. 1063 and A.D. 1202 earthquakes (Fig. 2; Sbeinati et al., 2005). The 20-km-thick seismogenic layer (Brew et al., 2001) correlates with the ~90 km fault length estimated from field mapping (Fig. 3). Fault dimensions are consistent with the ~4.3 m maximum characteristic slip inferred from the warping of the aqueduct wall east of the fault (and west of the bridge). Here, we assume that successive faulting episodes maintained the early ~4.3 m warping of an already ruptured strong building. Taking an average 2.0 m coseismic slip along the fault, the obtained seismic moment is Mo = 1.05 × 10²⁰ N m (Mw 7.3; Wells and Coppersmith, 1994), which is comparable, for instance, with the seismic moment of the 1999 Izmit large earthquake (Mw 7.4) of the North Anatolian fault.

• from Sbeinati et al. (2010:262-263)

The consistency among the timing of faulted sedimentary units in trenches, the age of building and repair of the aqueduct wall, and the dating of tufa interruptions and restart episodes determines the completeness of a sequence of earthquake events. The dating of three episodes of fault slip X, Y, and Z is consistent with the two phases of aqueduct wall repair, and the two interruptions of the longest tufa deposits BR-5 and BR-6, and interruptions and restart in BR-3 and BR-4. Our observations indicate that the aqueduct was repaired after the large seismic events X and Y but abandoned after the most recent faulting event Z. Building repair after a damaging earthquake is very often necessary because it is a vital remedial measure of water supply in order to avoid a decline of the local economy (Ambraseys, 2006). The repair has the benefit of leaving critical indicators of previous damage and, in some cases, of the fault slip characteristics.

For instance, the eastern wall of the Al Harif aqueduct shows a clear warping that confirms the left-lateral movement near the fault zone. As observed for coseismic surface ruptures crossing buildings, fences, and walls during large strike-slip earthquakes (Yeats et al., 1997), warped walls that may record a coseismic slip are often observed along strike-slip faulting. Warping that amounts to 4.3 m can be interpreted as the individual coseismic slip during event X. The warping can be due to the opposite lateral movements across the fault constrained by the bridge cohesion to the east and wall solidity to the west. While the western aqueduct wall section was built straight on the flat alluvial terrace and ends abruptly against the fault, only the section between the bridge and the fault zone (which is partly built on loose sediments and bridge ballast) presents some warping and dragging (possibly separated from the alluvial substratum; Fig. 14). The warped section near the bridge displays one generation of cracks filled with tufa that attests to the early bridge damage and possible correlation with event X (Meghraoui et al., 2003). Similar warped walls and fences were observed after the 17 August 1999 earthquake and along the North Anatolia fault in Turkey (Barka et al., 2002). Subsequent faulting movements Y and Z would have affected an already broken aqueduct wall (even if rebuilt) with less strength at the fault zone than for the initial building conditions (Fig. 14). Furthermore, the 4.3 m can be considered as a characteristic slip at the aqueduct site; such characteristic behavior with repeated same amounts of coseismic slip has already been observed and inferred from paleoseismic trenches along major strike-slip faults (Klinger et al., 2003; Rockwell et al., 2009). If the warped aqueduct wall is random and not representative of a coseismic slip, the alternative solution is quite similar if we consider a 4.5 m average individual slip from the cumulative 13.6 m left-lateral offset and the X, Y, and Z large seismic events at the aqueduct site.

• from Sbeinati et al. (2010:263)

Another key issue is the relationship between the aqueduct damage and the start and interruption of tufa accumulation with past earthquakes (Figs. 11 and 13). Indeed, the water flow may be interrupted anytime due to, for instance, the actions of man (warfare) or the onset of a drought period and climatic fluctuations that may influence the water flow. These possibilities seem here unlikely because the only two interruptions in cores BR-5 and BR-6 coincide with earthquake events X and Y, and no other additional interruptions were here recorded. This is also attested by the two interruptions in cores BR-3 and BR-4 that correlate with earthquake events X and Y. The difference between the tufa accumulation in BR-4, BR-5, and BR-6 located on the wall section west of the fault, and BR-3 located on the wall section next to the bridge, east of the fault, provides a consistent aqueduct damage history (Fig. 13). The onset of BR-3 after event X is the sign of an extensive damage that tilted the bridge and allowed overflow with tufa accumulation on the aqueduct northern side. The subsequent interruption (repair) and restart of BR-3 that coincides with event Y illustrate the successive aqueduct damage. Located on the broken western wall section (Fig. 6), the onset of BR-4 after event X and restart after event Y are consistent with BR-3 tufa growth and accumulation. As illustrated in Figure 13, the coincidence among faulting events X, Y, and Z from paleoseismic trenches, the three building damage and repair episodes from archaeoseismic investigations, and tufa growth and interruption constrains the earthquake-induced damage and faulting episodes across the aqueduct.

• from Sbeinati et al. (2010:249-250)

The Al-Harif aqueduct is located ~4 km north of the city of Missyaf, immediately west of a limestone shutter ridge and related ~200 m left-lateral stream deflection. According to the remaining aqueduct walls and related mills in the region, the aqueduct was built during the Roman time (younger than 65 B.C. in the Middle East) to drain freshwater collected from springs of the western mountain range to the eastern semiarid plains. The remaining ruins of the aqueduct suggest an ~40-km-long construction that may have included several bridges over streams and landscape gorges.

The aqueduct building description and related age have not been reported so far in any archive, manuscript, or in the literature. There is, however, an interesting anecdotal story from the local tradition that it was built by a local prince to supply potable water to Apamea and/or Sheizar cities, located northeast of the aqueduct. Apamea during that time was the most famous and strategic city during the Hellenistic and Roman period, whereas Sheizar is known to have been an important political and military fortress during the Middle Ages.

In their description of the Dead Sea fault in Syria, Trifonov et al. (1991) mentioned the existence of a faulted aqueduct near the city of Missyaf, but neither the precise location nor the accurate amount of offset walls was given. However, this early tectonic observation was helpful and allowed us to discover the site and consider a detailed study (Meghraoui et al., 2003), which is extended here using combined methods in archaeoseismology, paleoseismology, and tufa investigations. In addition, a micro- topographic survey of measurements accompanied all field studies.

Previous investigations on the aqueduct (Meghraoui et al., 2003) established: (1) an evaluation of its age based on an account of the large size blocks, the dating of sedimentary units below the aqueduct wall foundation, and dating of early tufa deposits on the aqueduct wall, and (2) the identification of the seismic faulting origin of damage in nearby trench A. The building style, with typical bridge arch and large stone size disposition (Opus caementum), suggested a Roman age, which was confirmed by the radiocarbon dating of sedimentary layers below the walls and the early tufa deposits on the walls. The faulted aqueduct revealed 13.6 ± 0.20 m of total left-lateral offset and called for detailed investigations on the characteristics and history of successive fault movements.

The aqueduct design, with an open canal on top of the 4-m-high wall, allowed freshwater and carbonate-saturated water to overflow and induce significant tufa accumulation from 0.30 m to 0.83 m in section. The carbonate-rich and cool water collected from the nearby western range is associated with a semiarid and karstic area of the Mesozoic limestone that favors rapid carbonate precipitation and tufa accumulation. The tufa deposits show successive growths of lamination carbonate with high porosity, banded texture, and rich organic encrustations. Field observations show that tufa accumulation developed on both eastern and western sections (from the fault line), but only on the north-facing wall, likely due to a slight tilt of the damaged aqueduct wall, probably after the two first earthquakes.

The following paragraphs present the field investigations, which consisted of:

1. four archaeoseismic excavations near the aqueduct walls and remains
2. four paleoseismic trenches across the fault zone and the alluvial sediments
3. four cores (two cores were previously studied in Meghraoui et al., 2003) of tufa deposits collected from different sections of the aqueduct.

More than 200 samples of organic matter, charcoal fragments, and tufa core pieces were taken for radiocarbon analysis in order to characterize the timing of successive faulting and related damage of the aqueduct construction. All radiocarbon dating were calibrated (2σ range, 95.4% probability density) using Oxcal v4.0 and INTCAL04 calibration curve.

• from Sbeinati et al. (2010:251-256)

The remaining aqueduct construction forms an ~50-m-long, ~5-m-high, and 0.60-m-thick wall that includes an ~15-m-high arch bridge in its eastern section (Figs. 5 and 6A). The outer part is coated by a thick layer of tufa deposits, probably due to a long period of freshwater flow. The construction material that may vary with the successive building and repair ages is made of:

1. large-size limestone blocks (Opus quadratum, 1.0 m × 0.5 m × 0.5 m; see also https://www.romanaqueducts.info/aquasite), similar to the typical Roman archaeological constructions and visible at the lower bridge (pier section) and wall sections


2. medium-size limestone blocks (Opus incertum; 0.50 m × 0.30 m × 0.30 m), which form the foundation or the upper half wall section and show visible small portions of cement


3. small sizes of mixed stones of irregular shape with significant portions of mortar (cement), mostly visible in the apparently rebuilt part of the wall

Figures 5 and 6A also show a detached small piece of the aqueduct wall made of small-size stones and related cement ~3.5 m away from the eastern wall. Therefore, four areas (noted I to IV in Fig. 5) were excavated near the aqueduct using proper archaeological methods.

The large excavation I was dug on the fault zone near the dragged wall fragment, in the area between the eastern and western aqueduct walls (Figs. 5, 8A, and 8B). The purpose of excavation is here to study the relationships between the fault zone and aqueduct. The excavation that has ~4.5 × 4.5 m surface and ~0.6 m depth exposed missing parts of the aqueduct. A buried and fallen wall piece rotated and dragged parallel to the fault and a remaining wall piece in an oblique position between two shear zones were discovered. The buried wall fragments are not comparable to the Opus caementum (quadratum) of the original construction and suggest a rebuilding phase. The excavation floor displays oriented gravels and pebbles that mark the shear zones and related fault branches also visible in the inner trench section E (Figs. 8B and 8C).

We collected four samples in the fallen wall sections labeled A, B, and C of excavation I (Fig. 8B): Two cement samples (AQ-CS-1 and AQ-CS-4) found in between building stones are made of typical medieval rubble mortars (mainly mud, gypsum, and lime); the two other samples (AQ-CS-3-2 and AQ-CS-3-3) are tufa deposits preserved on building stones. All four samples contained enough organic matter to allow radiocarbon dating (Table 1). Two dates of cement yield A.D. 532–641 (section A, AQ-CS-4) for the large fallen wall in excavation I and A.D. 650–780 (section C, AQ-CS-1) for the wall fragment piece in between the walls (Fig. 8B). In addition, two tufa deposits on wall stones provide consistent ages A.D. 560–690 (section B, AQ-CS-3-2) and A.D. 639–883 (section C, AQ-CS-3-3) with cement ages. The two different cement dates of the fallen wall and dragged wall fragment can be correlated to the new tufa deposits that testify for two rebuilding phases. The dated buried fallen wall in section B (CS-3-2) obtained from a thin (~5 cm) tufa accumulation correlates with the similarly fallen wall in section A and related cement date of CS-4 (Fig. 8B; and Table 1). In section C, the tufa deposits and related dated sample CS-3-3 correlates with the cement age of CS-1. The type and size of stones (opus incertum) and thin tufa accumulation in sections A and B suggest an early rebuilding phase postdating the first damaging event that may have occurred between the first and sixth centuries A.D. The different building layout of section C made of small sizes of mixed stones of irregular shape, and dating of cement sample CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) indicate a repair and rebuilding period postdating a second damaging event at the end of the Byzantine time and beginning of the Islamic period (seventh to eighth century A.D.). The damaged and dragged most recent wall section C along the fault indicates the occurrence of a third event, after which the aqueduct was definitely abandoned.

Three small excavations II, III, and IV (1.5 m to 3.0 m long, 1.0 m wide and 1.50 m deep) were dug in the base layer of the western aqueduct wall in order to expose its foundation and related sedimentary units underneath that predate the early building phase (Figs. 5 and 9). Excavations II and III were dug under the wall section with maximum (> 0.80 m) and minimum (~0.30 m) thickness of tufa deposition, respectively (Figs. 9A and 9B). Excavation IV, already described in Meghraoui et al. (2003), exposed the faulted foundation of the missing section of the western wall edge. The wall foundation reaches 1 m depth and shows regular patterns of medium-size cut limestone blocks (0.50 m × 0.30 m × 0.30 m) built over a dark brown clayey layer (unit e).

Charcoal samples collected in excavations I (trench E), II, and III from unit e yield ¹⁴C dates with ages spanning from approximately the third century B.C. to third century A.D. (see samples AQ-TA, TB, and TC in Table 1; Figs. 8C, 9A, and 9B). Although in these excavations, the age range of unit e seems quite large (probably due to detrital charcoal mixing), the younger age, i.e., 350 B.C. to A.D. 130 (sample AQ-TA-4), is consistent with other radiocarbon ages of unit e and related stratigraphic succession in trenches (see Paleoseismic Trenches herein). In excavation II, the large stone shape (Opus quadratum) with small amount of cementing material and pottery fragments found on the same level near the building base can be correlated with the early Roman era (Fig. 9A). Large stones and tufa thickness led us to consider this section of the aqueduct wall to be in original condition, i.e., probably undamaged by large earthquakes.

Excavation III (1.65 m long, 1.0 m wide, and 1.2 m deep; Fig. 9B) is similar to excavations II and IV, but the 1-m-deep wall foundation and upper section show irregular shapes of mixed medium- and small-size cut limestone blocks (0.10 × 0.20 × 0.15 m). Excavation III was realized at the location of the thinnest tufa deposits (< 0.30 m). The size of stones, cement texture, and irregular shape of building wall suggest that this building section was rebuilt (Fig. 9B). The ¹⁴C dating of unit e below the wall yielded a comparable age range to that obtained in excavations II (see AQ-TA, TB, and TC in Table 1).

Trench section E (4.30 m long, 0.70 m wide, and 1.30 m deep; Figs. 5, 8A, 8B, and 8C) was dug within excavation I in order to see in section the fault zone that affects the archaeological floor units. The trench wall exposes similar sedimentary units to those visible in excavations II, III, and IV that are affected by two main fault branches of the shear zone visible in the floor layer of excavation I. The ¹⁴C dating of samples AQ-TC-S1, S2, and S3 of units f and e indicates 900 B.C. to A.D. 400 maximum and minimum age range, respectively (Fig. 8C; Table 1), which is comparable to the age range obtained in excavations II and III for unit e (Figs. 8B and 9A; Table 1). However, as here again the large age range can be due to charcoal mixing, the dating of unit e is obtained by comparison to the dated stratigraphic succession of units in trenches (see section Paleoseismic Trenches).

• from Sbeinati et al. (2010:256-258)

Two trenches, B and C (Figs. 5 and 10, trenches B and C), were dug across the Dead Sea fault north of the aqueduct in addition to the previously studied trench, A (Fig. 10A; Meghraoui et al., 2003). The two trenches exposed an ~1.5-m- wide fault zone that affects a succession of 2–3-m-thick fine and coarse alluvial sedimentary layers similar to the alluvial deposits of trench A. Alluvial units visible in all trenches exhibit here similar textures, structures, and color, and correspond to the same layers that belong to the same alluvial terrace. Although the three trenches A, B, and C may not expose a completed stratigraphic section, the comparisons among sedimentary units, faulting events, archaeoseismic observations, and tufa accumulation limit the possibility of a missing earthquake event that affected the aqueduct.



In trench B (south wall), the fault zone shows three main fault branches that affect sedimentary units g to d and form a negative flower structure. The central and western main branches are truncated by unit a, which forms a stratified 0.3–0.4-m-thick deposit of coarse gravels in a sandy matrix. The eastern fault branch is buried below unit d, made of well-sorted reddish fine gravels. Unit e, a 0.2–0.5-m-thick dark-brown silt-clay, thickens toward east. Units f and g are made of scattered clasts in a massive clay matrix of dark-brown and light-brown color, respectively. Although intense warping and faulting are marked by contrasting color and texture of unit e, faulted sedimentary layers of this trench do not allow the identification of all faulting events. However, buried fault branches indicate a faulting event postdated by unit d (event Y), while the other fault branches show at least another faulting event (event Z) overlain by unit a. While clearly visible in other trench walls, event Y is here likely concealed by the complex fault branches truncated by unit a.



Trench C (Fig. 10C) exposes a stratigraphic succession affected by at least five main fault branches (labeled I to V in Fig. 10C). From trench bottom, fault branch I, which affects unit g, is overlain by unit f. A similar observation can be made for fault branch II, which also affects all units below unit d. Furthermore, the trench wall exposes an ~0.60-m-thick well- stratified, coarse and fine gravel layer above unit e and across the fault zone. Unit d thins significantly west of fault branch III and is overlapped by relatively thick coarse gravel units, which display a mix of fine and coarse gravels between fault branches III and IV, and unit d shows a succession of well-stratified alluvial units west of fault branch IV (Fig. 10C). Taking into account its alluvial origin made of well-stratified fine and coarse gravels, west of fault branch IV, unit d is subdivided into d1, d2, d3, and d4. Faulting movements at this site allows truncation of unit d1 (equivalent to d east of fault branch III) and sedimentation of units d2 to d4 (in a likely small pull-apart basin). Unit d3 consists of an ~0.20-m-thick dark-brown silt-sand overlain by unit d4, which is made of light-brown fine silt-sand. Below the plough zone a2, the well-stratified unit a1 shows flat-laying pebbles and gravels and intercalated fine gravels covering previous units and the fault zone.

Fault branches I to V in trench C indicate a negative flower structure that intersects a sedimentary sequence and reveal at least four faulting events (Fig. 10C):

1. Event W, identified on fault branch I, is older than 800–510 B.C. (EH II-18S) in the lowermost layers of unit f and is younger than unit g, which was dated with sample EH II-5S (3400–300 B.C.).


2. Next to fault branch II, buried below unit d, the vertical offsets between unit e and units d and d1 across fault branch III, and the absence of unit e between fault branches III and IV, determine the faulting event X between unit e and unit d. Since unit d overlies an erosional surface of unit e, faulting event X may have formed a depression (i.e., a small pull-apart basin) that allowed the deposition of d1 to d4 next to a thick unit d east of fault branch III. The faulting event X is here predated by 360–90 B.C. (EH II-12S), 360–50 B.C. (EH II-11S), and 360–60 B.C. (EH II-10S) of the uppermost layers of unit f (event X is postdated by sample EH I-TA-S33 of trench A).


3. Faulting event Y can be identified at the westernmost fault branch V between unit d2 and unit d3. The dating of sample EH II-16S in d3 postdates event X to younger than A.D. 540–650, which we consider as a reliable age, taking into account its high carbon content (event Y is predated by sample EH I-TA-S33 of trench A).


4. Faulting event Z corresponds to the main fault branches III and IV, which are overlain by the stratified unit a2 below the plough zone. Fault rupture IV affects unit d4 and indicates that the faulting event Z is older than radiocarbon age A.D. 1480–1800 (EH II-7S) and A.D. 1510–1670 (EH II-2N) of unit a2 and younger than unit d4.

• from Sbeinati et al. (2010:258)

The analysis of faulting events from the aqueduct (damage and reconstruction) and from trenches A, B, and C can be presented as following:

1. Event W is older than unit f (i.e., 800–510 B.C.) and younger than unit g (i.e., 3400–300 B.C.) of trench C. The bracket of event W is here difficult to assess since the detrital charcoal sample in unit f was not taken from the base of unit f. According to ¹⁴C dates, the faulting event can be estimated as younger than 3400 B.C. and older than 510 B.C. However, taking into account the rate of sedimentation in unit f, we may estimate a minimum age of 962 B.C. for event W.

2. Event X, the first faulting event that affected the aqueduct, is bracketed between the first and sixth centuries A.D. In trenches, a large bracket of this event is between 350 B.C. and A.D. 30 and A.D. 650–810 (as obtained from dated units of trench A).

3. Event Y, characterized from paleoseismology, appears to be older than A.D. 650–810 (unit d, trench A) and younger than A.D. 540–650 (unit d3 in trench C). The results of archaeoseismic investigations indicate that ages of CS-1 (A.D. 650–780) and tufa accumulation CS-3-3 (A.D. 639–883) postdate event Y.

4. Event Z is the last faulting event that affected the aqueduct, after which it was definitely abandoned. In trenches A and C, event Z is older than A.D. 1480–1800, A.D. 1510–1670, and A.D. 1030–1260 and younger than A.D. 960–1060.

• from Sbeinati et al. (2010:260-261)

We conducted four archaeoseismic excavations, three paleoseismic trenches, and obtained the radiocarbon dating of six cores at the Al Harif Aqueduct site along the Missyaf segment of the Dead Sea fault. The combined study allows us to obtain a better constraint on the timing of past earthquakes, with four large seismic events during the last ~3400 yr. The occurrence of three seismic events X, Y, and Z (A.D. 70–600, ca. A.D. 650, and A.D. 1170, respectively) since the construction of the aqueduct is attested by faulting events in trenches, the damage and repair of the aqueduct wall, and the tufa growth and interruptions since Roman time (Fig. 13). These results point out a temporal clustering of three large earthquakes between A.D. 70 and A.D. 1170 along the Missyaf fault segment (Fig. 14).

The 90 ± 10-km-long and linear Missyaf segment experienced the A.D. 1170 earthquake recorded in trenches, aqueduct construction, and tufa deposits. In this tectonic framework, the large (10-km-wide) Ghab pull-apart basin to the north and the Al Bouqueaa pull-apart and onset of the restraining bend to the south (Fig. 3) may constitute endpoints for earthquake rupture propagation, as observed for other large continental strike-slip faults (Klinger et al., 2003; Wesnousky, 2006). The size of the Ghab Basin and the sharp bend of the Lebanese fault system may act as structural control of fault-rupture initiation and propa- gation. Furthermore, the damage distribution of the A.D. 1170 earthquake, well located on the Missyaf segment, is limited to the north by the A.D. 1156 large earthquake and to the south by the A.D. 1063 and A.D. 1202 earthquakes (Fig. 2; Sbeinati et al., 2005). The 20-km-thick seismogenic layer (Brew et al., 2001) correlates with the ~90 km fault length estimated from field mapping (Fig. 3). Fault dimensions are consistent with the ~4.3 m maximum characteristic slip inferred from the warping of the aqueduct wall east of the fault (and west of the bridge). Here, we assume that successive faulting episodes maintained the early ~4.3 m warping of an already ruptured strong building. Taking an average 2.0 m coseismic slip along the fault, the obtained seismic moment is Mo = 1.05 × 10²⁰ N m (Mw 7.3; Wells and Coppersmith, 1994), which is comparable, for instance, with the seismic moment of the 1999 Izmit large earthquake (Mw 7.4) of the North Anatolian fault.

• from Sbeinati et al. (2010:262-263)

The consistency among the timing of faulted sedimentary units in trenches, the age of building and repair of the aqueduct wall, and the dating of tufa interruptions and restart episodes determines the completeness of a sequence of earthquake events. The dating of three episodes of fault slip X, Y, and Z is consistent with the two phases of aqueduct wall repair, and the two interruptions of the longest tufa deposits BR-5 and BR-6, and interruptions and restart in BR-3 and BR-4. Our observations indicate that the aqueduct was repaired after the large seismic events X and Y but abandoned after the most recent faulting event Z. Building repair after a damaging earthquake is very often necessary because it is a vital remedial measure of water supply in order to avoid a decline of the local economy (Ambraseys, 2006). The repair has the benefit of leaving critical indicators of previous damage and, in some cases, of the fault slip characteristics.

For instance, the eastern wall of the Al Harif aqueduct shows a clear warping that confirms the left-lateral movement near the fault zone. As observed for coseismic surface ruptures crossing buildings, fences, and walls during large strike-slip earthquakes (Yeats et al., 1997), warped walls that may record a coseismic slip are often observed along strike-slip faulting. Warping that amounts to 4.3 m can be interpreted as the individual coseismic slip during event X. The warping can be due to the opposite lateral movements across the fault constrained by the bridge cohesion to the east and wall solidity to the west. While the western aqueduct wall section was built straight on the flat alluvial terrace and ends abruptly against the fault, only the section between the bridge and the fault zone (which is partly built on loose sediments and bridge ballast) presents some warping and dragging (possibly separated from the alluvial substratum; Fig. 14). The warped section near the bridge displays one generation of cracks filled with tufa that attests to the early bridge damage and possible correlation with event X (Meghraoui et al., 2003). Similar warped walls and fences were observed after the 17 August 1999 earthquake and along the North Anatolia fault in Turkey (Barka et al., 2002). Subsequent faulting movements Y and Z would have affected an already broken aqueduct wall (even if rebuilt) with less strength at the fault zone than for the initial building conditions (Fig. 14). Furthermore, the 4.3 m can be considered as a characteristic slip at the aqueduct site; such characteristic behavior with repeated same amounts of coseismic slip has already been observed and inferred from paleoseismic trenches along major strike-slip faults (Klinger et al., 2003; Rockwell et al., 2009). If the warped aqueduct wall is random and not representative of a coseismic slip, the alternative solution is quite similar if we consider a 4.5 m average individual slip from the cumulative 13.6 m left-lateral offset and the X, Y, and Z large seismic events at the aqueduct site.

• from Sbeinati et al. (2010:263)

Another key issue is the relationship between the aqueduct damage and the start and interruption of tufa accumulation with past earthquakes (Figs. 11 and 13). Indeed, the water flow may be interrupted anytime due to, for instance, the actions of man (warfare) or the onset of a drought period and climatic fluctuations that may influence the water flow. These possibilities seem here unlikely because the only two interruptions in cores BR-5 and BR-6 coincide with earthquake events X and Y, and no other additional interruptions were here recorded. This is also attested by the two interruptions in cores BR-3 and BR-4 that correlate with earthquake events X and Y. The difference between the tufa accumulation in BR-4, BR-5, and BR-6 located on the wall section west of the fault, and BR-3 located on the wall section next to the bridge, east of the fault, provides a consistent aqueduct damage history (Fig. 13). The onset of BR-3 after event X is the sign of an extensive damage that tilted the bridge and allowed overflow with tufa accumulation on the aqueduct northern side. The subsequent interruption (repair) and restart of BR-3 that coincides with event Y illustrate the successive aqueduct damage. Located on the broken western wall section (Fig. 6), the onset of BR-4 after event X and restart after event Y are consistent with BR-3 tufa growth and accumulation. As illustrated in Figure 13, the coincidence among faulting events X, Y, and Z from paleoseismic trenches, the three building damage and repair episodes from archaeoseismic investigations, and tufa growth and interruption constrains the earthquake-induced damage and faulting episodes across the aqueduct.