Reassessing the Mycenaean Earthquake Hypothesis: Results of the HERACLES Project from Tiryns and Midea, Greece

Transcription

1 Bulletin of the Seismological Society of America, Vol. 108, No. 3A, pp , June 2018, doi: / Reassessing the Mycenaean Earthquake Hypothesis: Results of the HERACLES Project from Tiryns and Midea, Greece by Klaus-G. Hinzen, Joseph Maran, Hector Hinojosa-Prieto, * Ursula Damm-Meinhardt, Sharon K. Reamer, Jana Tzislakis, Kilian Kemna, Gregor Schweppe, Claus Fleischer, and Katie Demakopoulou Abstract Observations at Mycenaean archaeological sites of tilted and curved walls, broken pottery, and human skeletons led to the hypothesis that these sites in the Argolid, Peloponnese, Greece, were destroyed in large earthquakes between the late palatial (thirteenth century B.C.E.) and postpalatial ( B.C.E.) periods. In particular, the destruction of Mycenaean palaces around 1200/1190 B.C.E. has often been attributed to a devastating earthquake. To test the Mycenaean earthquake hypothesis, this project focuses on the Argive citadels of Tiryns and Midea. With active and passive seismic measurements complemented by a gravimetric survey, we explored seismic site effects at these locations and calculated synthetic seismograms for potential earthquake sources to estimate intensities of ground motions inside and outside the citadels. The field work and results were supplemented by analysis of the individual damage descriptions and observations from the archaeological literature on which the hypothesis is based. Because of poor construction techniques and the associated site effects, the buildings in the Lower Town surrounding the citadel of Tiryns were more vulnerable than the structures within the Cyclopean palace walls, but evidence of an earthquake destruction stratum in the Lower Town has not yet been found. Although some of the observations from the two investigated citadels could be explained by seismic loading, alternative nonseismic causes could equally explain most observed damage. In some cases, the structural damage was clearly not caused by earthquakes. Simulated ground motions show that severe earthquake damage at Tiryns and Midea can be expected from activation of local faults in the Argive basin; however, palaeoseismic studies for such activity in and since the Late Bronze Age (LBA) are lacking. Our results indicate that the hypothesis of a destructive earthquake in Tiryns and Midea, which may have contributed to the end of the LBA Mycenaean palatial period, is unlikely. Introduction The Mycenaean earthquake hypothesis postulates a coseismic structural collapse of several important Mycenaean citadels and settlements including Tiryns (Kilian, 1988, 1996), Midea (Åström and Demakopoulou, 1996; Demakopoulou, 2007, 2012, 2015), Mycenae (French, 1996), Thebes (Sampson, 1996), Teichos Dymaion, Pylos, Nichoria, Menelaion, Lefkandi, Iolkos, and on Crete, Kydonia, and Knossos (Nur and Cline, 2000) during the Late Bronze Age (LBA), based on destruction layers with indications of assumed earthquake damage (Fig. 1). These strata are separated by time intervals in the order of several decades during the late *Now at Advanced Geosciences, Inc., 2121 Geoscience Drive, Austin, Texas palatial period (Late Helladic [LH] IIIB in Fig. 1) and postpalatial period (LH IIIC) from which it was interpreted that more than one earthquake might have caused the structural damage to Mycenaean settlements (Åström and Demakopoulou, 1996; Kilian, 1996; French and Stockhammer, 2009; Mühlenbruch, 2013; Damm-Meinhardt, 2015). Additional evidence presented by archaeological sources includes tilted and curved walls, broken pottery presumably toppled by ground shaking, and human skeletons, some lying under the debris of assumed coseismically collapsed walls (Åström and Demakopoulou, 1996; French, 1996; Kilian, 1996; Demakopoulou, 2015). Multiple earthquakes are hypothesized to have caused the structural damage of LH IIIB C architecture at Tiryns and 1046

2 Reassessing the Mycenaean Earthquake Hypothesis 1047 Figure 1. (a) Map of Greece and surrounding areas. Important Mycenaean locations in and outside the Argive basin are shown by blue and black circles and labels, respectively. The acropoleis of Tiryns and Midea are marked by red diamonds. Seismicity from 1900 to 2013 (Aristotle University of Thessaloniki, see Data and Resources; Papazachos et al., 2010) is shown by gray circles. HSZ, Hellenic subduction zone; PCCR, Patras- Corinth continental rift; IEST, Iria Epidaurus sinistral transform fault system. (b) Timeline of the Late Helladic (LH) period for Mycenaean Greece (after Manning, 2010) high chronology with the probable time of Thera volcano eruption marked at Midea (Kilian, 1983, 1988, 1996; Papanastassiou et al., 1993; Åström and Demakopoulou, 1996; Gaki-Papanastassiou et al., 1996; Maroukian et al.,1996; Nur and Cline, 2000; Demakopoulou, 2007, 2015; Nur and Burgess, 2008). In particular, two earthquakes (and possibly a third one) during the palatial period and one earthquake in the postpalatial period (Kilian, 1996; French and Stockhammer, 2009; Mühlenbruch, 2013; Damm-Meinhardt, 2015) presumably damaged the Tiryns citadel. These earthquakes were deduced on the basis of three destruction layers that were dated by archaeological stratigraphy to LH IIIB Middle (about 1250 B.C.E.), LH IIIB Final (about 1200/1190 B.C.E.), and LH IIIC Advanced (later twelfth century B.C.E.), respectively. After the conflagration and collapse of buildings inside the citadel of Tiryns at about 1200 B.C.E., the occupation continued (Kilian, 1981; Maran, 2010), although there were definite major changes. Similarly, settlement continuity is ascertained at Midea and Mycenae after the postulated earthquake at the end of the palatial period (Iakovidis, 2003; Demakopoulou, 2007, 2012, 2015; Middleton, 2008, 2012). Papanastassiou et al. (1993) correlated postulated earthquake damage at Tiryns, Midea, and Mycenae by archaeological dating of destruction layers. They cite four earthquakes at Tiryns (1330, 1250, 1190, and 1150 B.C.E.), two at Mycenae (1250 and 1190 B.C.E.), and one at Midea (1190 B.C.E.). Although the earthquake hypothesis for Tiryns and Midea had been formulated by archaeologists in the 1980/1990s (Demakopoulou, 1995; Åström and Demakopoulou, 1996; Kilian, 1996), a comprehensive plausibility test with quantitative geoscientific and engineering methods was lacking. The motivation for the testing was encouraged by one of the principal investigators (J. M.) and became the defining goal of the HERACLES (Hypothesis-Testing of Earthquake Ruined Argolid Constructions and Landscape with Engineering Seismology) project, a joint study with Heidelberg and Cologne University in cooperation with the Ephorate of Antiquities in Nafplion. HERACLES is named after Tiryns greatest resident hero whose mother Alkmene, according to Greek Mythology, was born and brought up in Midea. As part of the preliminary research for the project, we analyzed documentation about the origin and development of the hypothesis and the primary observations it is based on, with emphasis on Tiryns, the starting point of the hypothesis. To include quantitative engineering seismological models for testing the hypothesis, seismic and gravimetric in situ experiments were necessary to generate appropriate model parameters, and laser scans were used to document and analyze the current status of damaged constructions. Of particular interest were the potential seismic site effects, known to also have substantial influence on earthquake damage for earthquakes of moderate magnitudes. Potential damaging earthquake sources that may have affected the Argolid during the Mycenaean period were identified. Earthquake scenarios were established and used to calculate site-specific synthetic seismograms and seismic intensities for Tiryns and Midea, and thus the damage potential of both sites was estimated. In the following, we will designate certain damage observations (DO) followed by a number to simplify discussion. Tiryns, Midea, and the Argolid The neighboring Mycenaean citadels of Tiryns and Midea were built on isolated, prominent natural hard-rock

3 1048 Klaus-G. Hinzen et al. Figure 2. Generalized geologic map of the Argive basin (after Papastamatiou et al., 1960; Tataris et al., 1970; Georgiou and Galanakis, 2010). Black triangles show the location of Mycenaean citadels and a black square shows the Mycenaean Nea Tiryntha-Kophini dam. Earthquake locations are based on the catalog of the Aristotle University of Thessaloniki, Greece (see Data and Resources). Above and below the earthquake symbols, magnitude and source depth in kilometers are given, respectively. Locations of available borehole data from the Greek Service of Land Improvement (GSLI, see Data and Resources) are shown, which reached the bedrock between 10 and 142 m; holes that did not reach bedrock have different symbols (see legend). Shallow boreholes in the vicinity of Tiryns are from Zangger (1993) and Ntageretzis (2014). The Early Helladic coastline is based on Zangger (1993). ridges surrounded by topographic highs. The distance between the two citadels is about 7.0 km (Fig. 2), and both were connected with other nearby Mycenaean centers by a network of roads (Demakopoulou, 2007, 2012, 2015). Despite their proximity, the citadels (Fig. 3) were located in different topographic and geologic settings. By the LH IIIB

4 Reassessing the Mycenaean Earthquake Hypothesis 1049 Figure 3. (a) Archaeological plan of Tiryns (after Maran, 2010) and (b) aerial photograph (Google Earth; see Data and Resources). (c) Archaeological plan of Midea (after Demakopoulou, 2012) and (d) aerial photograph of Midea (K. Xenikakis in Demakopoulou, 2012). period (Fig. 1b), fortification walls of Cyclopean-style (Schliemann, 1886) encircled the citadels. Tiryns and Midea were presumably important strongholds under the rule of Mycenae, which is considered to have been the regional political center in this part of the Peloponnese (Maran, 2004b, 2010, 2015; Demakopoulou, 2007; Middleton, 2008). Clear evidence of structural damage indicates that both citadels suffered from severe conflagrations at the end of LH IIIB. Particularly, in the palace area of Tiryns, the fire was so strong that air-dried mud bricks of many walls were burnt to such a degree that, following Schliemann (1886), early excavators mistakenly thought these were medieval walls made from fired bricks. Schliemann further mentions that their strongest excavation workers had to bend over backward to smash them with pickaxes. This destruction made archaeoseismological analysis at Tiryns difficult. In addition, foundations were generally constructed from rubble stones, and only a few of the burnt parts of the upgoing walls have been preserved. Setting and Geology Tiryns stands on an elongated north-northwest-trending 300-m-long limestone ridge (bedrock) rising above the alluvial soils of the southeastern Argive basin (Zangger, 1994; Maran, 2010) and overlooking the south-widening Argolic Gulf (Figs. 2 and 3). Archaeological excavations at Tiryns brought to light Mycenaean architecture (Schliemann, 1886; Kilian, 1979, 1981, 1983, 1988; Maran, 2004a,b, 2006, 2010, 2015) composed of several complexes. These include the Lower Citadel (Unterburg) in the north and the Upper Citadel (Oberburg) with the Great Megaron forming the center of the palace furnished with a throne room. In addition to the Cyclopean walls, the site is well known for its east and south galleries with corbelled vaults built on the flanks of the ridge (Fig. 3). The elevation of the acropolis increases from north (15.0 meters above sea level [m.a.s.l.]) to south (28.0 m.a.s.l.). Although currently the Cyclopean walls of Tiryns are preserved or restored to a maximum height of 9.9 m with up to 7.0 m thickness, the true original maximum height of the wall is unknown. Some of the Cyclopean building blocks exceed a mass of 1: kg. The wall circuit is 750 m long and encloses an area of nearly 18;500 m 2 (Papadimitriou, 2001). Fallen blocks of the Cyclopean wall are still scattered across the eastern and northern flanks of the Tiryns ridge. The west wall was restored along its entire

5 1050 Klaus-G. Hinzen et al. length, but segments of the eastern wall remained where they fell into ruin during more than 3000 yrs and as they were described by Schliemann in During the LH period, the acropolis of Tiryns was surrounded by the Lower Town, a settlement still under excavation and the extent of which is not precisely known. The Lower Town persisted throughout the different phases of the Mycenaean period (Zangger, 1994; Maran, 2010). By its nature, the houses in the Lower Town were simpler constructions than those inside the citadel. Zangger (1993, 1994) suggests that the area of the northern Lower Town was inundated by a single catastrophic and potentially earthquake-related flash flood about 1200 B.C.E. However, a more recent interpretation of the evidence shows that the flood deposits actually represent a sequence of periodic flooding events likely spanning a longer period within the thirteenth century B.C.E. (Maran, 2004b; Maran and Papadimitriou, 2006). This eliminates the idea of a single earthquake-related flash-flood event. The flooding of the Lower Town is named DO1. With the massive dam of Nea Tiryntha-Kophini (Fig. 2), a masterpiece of Mycenaean engineering, located only 4.0 km east of Tiryns (Knauss, 2001; Maroukian et al., 2004) and operational from the time of the final palatial period until the present, the stream was diverted to prevent such flooding events in the Lower Town and enable building activities in this area. Even though the late palatial period (LH IIIB) should be regarded as one of the most prosperous in Mycenaean times, very little architectural remains dating to LH IIIB2 have so far been unearthed in the Lower Town, which seems to have reached its widest extent in the postpalatial period (Kilian, 1988; Maran, 2010). The Mycenaean citadel of Midea is located nearly 12.0 km southeast of Mycenae and 7.0 km northeast of Tiryns (Fig. 2). The Midea citadel overlooks the entire Argive basin including the Argolic Gulf because it was built on the flank and the top of a prominent 2-km-long ridge (Demakopoulou, 2007, 2012, 2015). This ridge trends northnorthwest, reaches a maximum altitude of 268 m, and is composed of highly weathered, karstified, crystalline limestone faulted against continental flysch deposits. The ridge rises above the Upper Pliocene Quaternary steep terrace deposits of the northeastern Argive basin. Excavation data indicate occupation of Midea from the Final Neolithic to Early Byzantine period and that important economic, administrative, military, and ritual activities took place during its climax in Mycenaean times (Walberg, 2001; Demakopoulou, 2007, 2012, 2015). Midea comprises a lower and an upper natural level separated by a rocky slope dividing the citadel into the Lower and Upper Acropolis (Fig. 3), respectively (Demakopoulou, 2007, 2012, 2015). The Midean Cyclopean wall was built in LH IIIB middle; its circuit is 450 m long, m thick, partially preserved up to a height of 7.0 m, and enclosing an area of 24;000 m 2 (Fischer, 1986; Walberg, 2001; Demakopoulou, 2012, 2015). Larger limestone boulders were used for the inner and outer façades, whereas the interior of the wall contained a filling of much smaller stones; however, on average, block sizes are smaller than those used in Tiryns. Information about the subsurface structure throughout the Argive basin has been provided by boreholes ( 40 m) drilled by the Greek Service of Land Improvement (see Data and Resources) and shallow borings ( 30 m) drilled around the citadel of Tiryns (Fig. 2) (Zangger, 1993; Ntageretzis, 2014). In summary, the borings indicate an irregularly thick Late Neogene Quaternary clastic sequence unconformably deposited on uneven shallow-to-deep bedrock. The shallow borings of Zangger (1993) and Ntageretzis (2014) indicate that the soil bedrock boundary deepens away from the Tiryns ridge. Figure 2 also shows the seaward advance of the coastline since the Early Helladic times away from Tiryns, which probably also served as an important Mycenaean harbor. Seismicity The instrumental seismicity of Greece (Fig. 1) has been well documented with the deployment of the first seismograph in 1897 in Athens (Ambraseys and Jackson, 1990; Kouskouna and Makropoulos, 2004). Historical records extend back to 550 B.C.E. The seismicity of the Aegean microplate is the highest in the Mediterranean region; however, due to the complex tectonic setting, its pattern is diffuse (McKenzie, 1972, 1978; Papazachos, 1980; Makropoulos and Burton, 1984; Taymaz et al., 1990, 2007; Papanikolaou and Royden, 2007). Seismogenic zonation has been based on the focal distribution of shallow to deep earthquakes (Papazachos, 1980; Papazachos and Papaioannou, 1993; Papaioannou and Papazachos, 2000) and on onshore geological and neotectonic data, and on offshore seismic reflection profiles (Papoulia et al., 2014). Papoulia et al. (2014) describe a north-northwest south-southeast-oriented 92-km-long inactive extensional detachment fault (Papanikolaou and Royden, 2007) and a northwest southeast-trending 60-km-long active normal fault along the western and eastern coast of the Argolic Gulf as the dominant structures. The existence of the eastern faults explains the current low-level shallow seismicity under the east west-oriented extensional stress regime. With this interpretation, the Argive basin can be considered to be the onshore continuation of the Argolic Gulf (compare to van Andel and Lianos, 1984; van Andel et al., 1990, 1993). The Argive basin contains observed and inferred normal faults (Fig. 2), some traditionally considered as active and others as probably active faults (Papastamatiou et al., 1960; Tataris et al., 1970; Papanastassiou et al., 1993; Papanikolaou et al., 1994; European Center on Prevention and Forecasting of Earthquakes and Protection Organization [ECPFEPO], 1996; Georgiou and Galanakis, 2010). Within a distance of 150 km, the basin is surrounded by active source zones (Papadopoulos and Kijko, 1991; Papazachos and Papaioannou, 1993; Papaioannou and Papazachos,

6 Reassessing the Mycenaean Earthquake Hypothesis 1051 In an archaeoseismological context, it can be argued that seismicity of the Argive basin during Mycenaean times was the same as today because the active tectonic setting of the Peloponnese has remained unchanged dating back to the Pliocene (about 5.3 Ma), according to regional geologic (Papanikolaou et al., 1994), seismologic (Hatzfeld, 1994; Papazachos, Comninakis, et al., 2000), and geodynamic studies (Papanikolaou and Royden, 2007; Royden and Papanikolaou, 2011). Moreover, the aforementioned seismogenic zones are considered to be potential seismic hazards to all Mycenaean sites located in the Argive basin. However, the moderate-to-strong earthquakes in the western PCCR and the adjacent Kephalonia-Lefkada fault system (Koukouvelas et al., 1996; Sachpazi et al., 2000; Konstantinou et al., 2009) are much less likely to have caused damage to Mycenaean sites in the basin due to the large epicentral distances. Figure 4. Distribution of earthquake magnitude with respect to the epicentral distance to Tiryns. The symbol color indicates the time period in which the earthquakes occurred, as shown in the legend. Heavy circles indicate earthquakes of intermediate depth (h >50 km). Numbers give the year of occurrence for earthquakes with distances < 20 km and magnitudes above 6. Hypocenters are from the catalog of the Aristotle University of Thessaloniki, Greece (see Data and Resources). 2000; Makris et al., 2004; Karastathis, Karmis, et al., 2010; Karastathis, Papadopoulos, et al., 2010; Makropoulos et al., 2012), including the Patras-and-Corinth continental rift (PCCR) system passing into the dextral Kephalonia-Lefkada transform fault system in the Ionian Sea in the north, the oceanic trench of the Hellenic subduction zone (HSZ) offshore western and southern Peloponnese with its downgoing slab extending underneath the Argive basin, and the Iria Epidaurus sinistral transform (IEST) fault system located in the southeast Argolis Peninsula (Fig. 1). Late Neogene Quaternary sedimentary basins of the Peloponnese, including the Argive basin, show a much lower seismic activity than the surrounding source zones; nevertheless, seismicity within the Argolis Peninsula includes some crustal earthquakes and events down to 150 km depth in the downgoing slab (Lyon-Caen et al., 1988; Papazachos et al., 1988; Hatzfeld et al., 1989; Papadopoulos and Kijko, 1991; Hatzfeld and Martin, 1992; Hatzfeld, Besnard, Markopoilos, et al., 1993, Hatzfeld, Besnard, Markopoilos, and Hatzidimitriou, 1993; Hatzfeld et al., 2000; Papazachos, Karakostas, et al., 2000; Makris et al., 2004). The catalog of earthquakes in Greece (see Data and Resources) contains more than 11,200 earthquakes with magnitudes above 4.5, of which about 600 are from the preinstrumental period. Figure 4 shows the magnitude distribution of these earthquakes with respect to the epicentral distance from Tiryns. Of the few earthquakes with magnitudes larger than 6.0 and epicentral distance less than 20 km to Tiryns, most are of intermediate depth. Among these, the closest earthquake with a distance of 10 km is dated to 388 B.C.E. The Earthquake Damage Hypothesis A paper summarizing Kilian s evidence for the earthquake hypothesis was published posthumously in 1996 (Kilian, 1996) based on a lecture he gave in Athens in Six excavation photos from Tiryns were included in that contribution to substantiate the hypothesis. Because Kilian was unable to finish the manuscript based on his lecture for publication before his death, the photos were selected by others who tried to identify the images shown by Kilian on the basis of a recorded film of his lecture. This procedure may explain some of the inconsistencies that will be described in the following. The photos of figures 1 and 2 in that article are described as showing a complex of rooms of the early LH III C period (c BC), outside the Acropolis in the Lower City (Unterburg) (p. 64) (DO2). This description is contradictory; either the rooms are situated in the Lower Town (Unterstadt) outside the acropolis, or in the Lower Citadel (Unterburg) inside the acropolis. In addition, the figure captions claim that the photos show a Room complex ( ) of the Late Helladic IIIC period (p. 64) in case of figure 1 and Same complex as in Fig. 1, but at a deeper level (p. 64) for figure 2. In both cases, it is claimed that the exposed walls are curved. Careful inspection of both photos shows that they were taken at exactly the same excavation situation but from a different perspective, only the scale bar had been rotated by 90. This is confirmed by the given archive numbers 76.5/42 and 76.5/15-16 for figures 1 and 2, respectively, because both negatives are on the same film (number 5 of 1976) and the one from figure 2 was even taken earlier than that from figure 1. The wall curvature, if any, is minor and does not resemble typical earthquake-induced wall bending. According to T. Mühlenbruch (personal comm., 1996), the photos show foundations of houses in the northwest Lower Town (rooms 308, 307, and 309). The remains date to the early postpalatial period (horizon 19A old), which is younger than the postulated earthquake horizon at the end of the palatial period (Damm-Meinhardt and Mühlenbruch, 2013; Mühlenbruch, 2013). Accordingly, the observations DO2 do not show proof

7 1052 Klaus-G. Hinzen et al. Figure 5. Situation at the main entrance of the Tiryns citadel (East Gate); (a) excavation photo from 1983 and (b) aerial view from The red shade indicates the transverse wall or stone setting that supposedly collapsed coseismically; the green shading indicates sections added after the postpalatial period to narrow the entrance from an original width of 4.70 to 2.50 m, as already mentioned by Dörpfeld (in Schliemann, 1886). (c,d) A plan and a profile redrawn from excavation documentation of 1983, respectively; the profile trend is shown by the dashed green line in (c). The edge of the stone setting is again shaded in red; yellow shading indicates sand fillings. Dotted blue lines in (c) show the corners of the cyclopean towers north and south of the entrance as they were surveyed by a laser scan. Dashed red lines indicate the deviation of the wall from a straight line. The arrow in the inset (bottom right) points to the main entrance. The photo in (a) is from the German Archaeological Institute (negative number is indicated in the upper left corner) and (e) is courtesy of Klessing, (c) and (d) are based on excavation drawings P1004 and S1007. consistent with coseismic damage during the LH IIIB Final period. Figure 3 of Kilian (1996) shows a photo taken in 1983 from above the remains of the gate tower of the eastern main entrance of the citadel (Fig. 3). The paper claims that the tranverse wall closing the Gate s opening is curved and collapsed (p. 64) (DO3). We consider this an important DO because it is the only one specifically addressing damage inside the Upper Citadel. Figure 5 shows an excavation photo taken slightly earlier than the one in Kilian (1996) and an aerial view of the entrance from 2011, together with a plan and a profile drawn during the excavation. The meaning of such a perpendicular wall closing the entrance is not discussed in Kilian (1996) and still remains unclear; the same applies to the original height of the structure. However, the plan and the profile suggest that the wall was a stone setting to support the sand and gravel with which the entrance way was filled and leveled. This interpretation contradicts that of an earthquake-toppled wall. The obvious curvature of the structure is most probably an adjustment to the subsurface conditions and intentionally built this way. The total offset from a straight line at the northern end of the stone setting is estimated to 1.3 m (Fig. 5c). If this deformation were coseismic, all material east of the stone setting would have had to be compressed by this amount, which does not seem physically plausible. Further, it has to be assumed that the western side of the stone setting (unfortunately not shown on the excavation plan) was not freestanding but also backfilled with loose material. Otherwise, there would have been a step of about 1 m height at the main entrance, which makes no sense from an architectural perspective. The stone material spreading out into the corridor, which was interpreted as the collapsed part of a wall (Fig. 5), was most probably a stone setting constructed to level and stabilize the transition from the entrance to the corridor, keeping in mind that this was the only access to the citadel for a cart or on horseback. Closing off the main entrance to the citadel with a real wall would only make sense to keep intruders out, for example, in times of war or unrest during a siege. This logical interpretation of the transverse wall or stone setting makes an interpretation of coseismic damage at the eastern entrance (DO3) unfeasible. Figure 4 from Kilian (1996) is an excavation photograph (Ti 1980/012, 12-13) with an oblique view of the foundations of Building X at the northeastern side of the Lower Citadel which was destroyed at the end of the palatial period (LH IIIB Final). The interpretation here is that the walls are undulating and the corners are not at right angles (p. 64). Both observations are interpreted as the walls are not linear but curving, nor are they orthogonal where they meet ( ) the deformation of the walls ( ) is due to an earthquake (p. 63) (DO4). Figure 6 shows an interpreted copy of the excavation plan (Kilian, 1982). The only wall which can be regarded as nonlinear is highlighted in Figure 6. Both the width of the foundation and its deviation from a straight line (undulation) were measured from the digitized plan and are plotted in

8 Reassessing the Mycenaean Earthquake Hypothesis 1053 Figure 6. (a) Excavation photo of Building X of the Lower Citadel in Tiryns after Kilian (1996, fig. 4), with the archive number Ti 1980/012, (b) Annotated version of the plan of the excavation (Kilian, 1982). Foundation walls of Building X are shaded in yellow; a water channel in blue. The undulated western part of the northern wall of the building (black dashed line) is shaded in red. The red circle marks the find spot of the skeletons of a female and a child. Numbers indicate angles between wall sections. The red hatched section is part of the northeastern Cyclopean wall. Grid marks are spaced 1.0 m. (c) Width (black dots) and undulation (deviation from straight line, blue dots) of the wall with the red dashed outline as a function of the position along the wall. The arrow in the lower right inset points to the location of Building X within the Lower Citadel. Figure 6, with respect to the wall s extension in the east west direction. The undulation reaches 0.45 m at its maximum at the eastern end of the wall which is, in total, 3.8 m long. The main deviation from a straight line occurs between 1.5 and 2.2 m along the wall trend. At the same location, the width of the wall is gradually reduced from 0:56 0:02 to 0:38 0:04 m. These two measures taken together constitute strong evidence for the conclusion that the curved wall DO4 was intentionally built this way, because earthquake ground motion could not have reduced the width of the foundation wall. The undisturbed link between the northern and southern walls that meet the curved wall right at its bending additionally supports the interpretation that this was part of the original design. If ground deformation would have induced the southward bending, the southern joining wall would have come under high stress, and its shortening by about 0.22 m would have induced damage which should have been preserved, particularly because the retaining wall at its southern end would have been a strong counter bearing. The second observation of the nonrectangular wall joints can also not be regarded as a viable argument for a seismogenic cause. The deviations from orthogonality are between 1.5 and 8.1. It is highly probable that the massive southern wall functioned as a retaining wall, and its trend was determined by the terrain conditions, with the slope toward the northern tip of the Lower Citadel. There might have been structural reasons to place Building X as a later addendum to the pre-existing retaining wall (Damm-Meinhardt, 2015). The two walls with the largest deviations from orthogonality are the northward extensions of which the western one links to the middle of the curved wall (Fig. 6). Here, the angular deviations are 6.9 and 8.1. If these walls would have been constructed at right angles with respect to the east west-trending walls, the coseismic movement of the northern wall head proposed in Kilian (1996) would have been 0.5 and 0.8 m, respectively. Therefore, it is extremely unlikely that the foundations made of rubble stones would have survived the deforming ground motions in their preserved state. This leads to the conclusion that the nonorthogonal wall joints were also part of the original layout of the building, yet it would be possible that they were influenced by partly reused foundation walls of an older building from the Early Bronze Age (EH III). In addition, the equal angles of 88.5 between the north south-trending center wall and its northern and southern crosswalls, the latter of which is the retaining

9 1054 Klaus-G. Hinzen et al. Figure 7. (a) Excavation photo from the Lower Citadel in Tiryns after Kilian (1996, fig. 5), with the archive number Ti 1980/072, The view is toward south along the corridor of Building VI with a bent wall (M156) at the eastern side. (b) A perspective view to the corridor from south toward north (Ti 1983/004, 23). The arrow in (c) points to the location of the corridor within the Lower Citadel. (d,e) Sections in the east west direction crossing the corridor before its complete excavation. The section in (d) is a view from south, the section in (e) originally viewed from north but reversed in this figure, so (d) and (e) have the same orientation. The trend of section (e) is indicated in (b) by the red line; section (d) is located south of the part covered by the photo in (b). Main walls (red hatching) are numbered with the original labels assigned by the excavators. A legend for the signatures is in the lower left (photos from the Tiryns archive, sections from Kilian, 1982; Damm-Meinhardt and Mühlenbruch, 2013). wall, are also strong evidence against major (coseismic) deformation for DO4. Figure 5 from Kilian (1996) shows a view from north into the so-called corridor of Building VI (Fig. 7a). The figure s original caption describes that The eastern wall is tilted downhill (westwards), while the western wall beyond the corridor is tilted uphill (eastwards) (p. 65). In Kilian (1996), this state is interpreted as antithetic inclination, a clear sign of seismic disturbances (DO5). The wall M156 in the section shown in Figure 7d is progressively bent westward from its straight bottom to 20 at its top. In a second section (Fig. 7e), no bending is visible; here the eastern corridor wall (M156) is straight; however, its eastern side is convex shaped, which gradually decreases the width of the wall with increasing height. Any eastward bending of the western wall (M210) of the corridor is not visible in the available sections (Fig. 7). Given that the distance between the parallel walls (< 2 m) is much shorter than the wavelength of seismic waves, the driving motion in case of an earthquake can be regarded as identical for both walls. An antithetic inclination caused by earthquake is therefore not to be expected. On the other hand, because of the strong topographic gradient with a westward slope in this part of the Lower Citadel, it is probable that the deformation of wall M156/156a occurred due to lateral earth pressure after it was (partially) buried: (1) the wall is a pure gravity wall; (2) there is little or no tensile strength between the rubble stones; (3) it is unlikely that the wall would have been deformed to 20 bending

10 Reassessing the Mycenaean Earthquake Hypothesis 1055 Figure 8. (a) East west section of the western wall of Building III (Tiryns Lower Citadel) from Damm-Meinhardt and Mühlenbruch (2013). The stones which toppled westward onto the floor of room 221 in Building IV are indicated by the arrows; signatures are those from Figure 9. (b) View from south toward the section with the toppled wall; arrows indicate toppled wall (Kilian, 1983). The arrow in the inset points to the location of Building III in the Lower Citadel. during earthquake ground motions due to inertial forces and that this deformation would have been frozen in its bent position; and (4) the westward dipping of the younger east west-trending wall M157 (LH IIIC) is an additional argument for a slow deformation process. At the northern end of wall M156a, there exists a section that was repaired by mud bricks or in pisé, an obvious indication of instability during its long life span (Damm-Meinhardt, 2015). The section in Figure 7d shows an inclination of the complete remains of the eastern corridor wall, affecting older and younger parts equally, and contradicts an interpretation of the excavator of a sudden inclining event. In conclusion, a coseismic damage to the corridor walls (DO5) during an earthquake at the end of the palatial period is not likely; however, a gradual earthpressure-driven deformation is well possible. The latter can also explain the shift from the alignment. This is supported by the observation that the 0.6 m high remains of the western wall of the corridor (M210) and the western wall of the building (M209) are neither inclined nor shifted. Figure 6 of Kilian (1996) shows the skeleton of a woman and a child who were presumed killed and buried by fallen walls of Building X (p. 65) (DO6). Tests with a discrete element model of the reconstructed Building X strongly suggest that the two human beings were outside the range where they would have been hit by the collapsing walls of Building X. Their distance to the closest part of the building is larger than the probable height of the construction. Also, it is unlikely that anything heavy from the top of the roof would have been hurled down that far. Detailed paleopathological examination of the situation did not reveal any evidence for earthquake debris in connection with the skeletons. All bone damage was postmortem (M. Schultz, personal comm., 2016). Rahmstorf (2008) sees the arrangement of the bones as compatible with other regular graves of the postpalatial period. Likewise interpreted as earthquake victims were the skulls of two males from the so-called Zwinger, an open area in the western part of the Lower Citadel between the Cyclopean wall and Building VI (Kilian, 1979, 1982; Damm- Meinhardt, 2015) in a context dating to LH IIIB Final (DO7). Only the skulls were found, and these do not show any specific earthquake traumata, eliminating coseismicity as a plausible cause of death. Additionally, a possibility exists that the remains were relocated because this area was repeatedly used for burials during the late palatial period. During excavations in 1981, the western wall of Building III in the Lower Citadel, dating to LH IIIB Final, was found collapsed to the west above the floor of room 221 in Building IV (DO8; Fig. 8). Unfortunately, this find situation is no longer accessible. Following the image and drawing from Figure 8, a coseismic damage is well possible; however, natural decay is also possible and anthropogenic action cannot be fully excluded (Damm-Meinhardt, 2015). In 1978, during excavations in the Lower Citadel in Room 110 dating to LH IIIC Developed, fragments of several terracotta figures and vessels were discovered (DO9). These were found in a horizon which is dated about 100 yrs later than the proposed large earthquake catastrophe in B.C.E. Kilian (1981) reconstructed the room as a cult room and proposed that the objects had originally been placed on a bench, the remains of which still exist at the western end of the room adjacent to the Cyclopean wall. Kilian argued that the objects had fallen down during an earthquake and were preserved in the original fallen position. A detailed study on the dynamic behavior of the three terracotta figures and two vessels (Hinzen et al., 2015) showed that earthquake ground motions are an unlikely cause of the toppled objects. This result is particularly important because room 110 was the primary example for which a natural catastrophe at Tiryns being the cause of destruction was assumed (Kilian, 1981). The earthquake

11 1056 Klaus-G. Hinzen et al. Figure 9. (a) West east and (b) north south-trending profile through the Tiryns citadel, respectively. The inset (bottom left) indicates the location of the profiles on top of the archaeological plan. Numbers give the mean P- ands-wave velocities for the four stratigraphic units based on refraction tomography. The plan of the archaeological remains in (b) is based on Schliemann (1886) (after Hinojosa-Prieto and Hinzen, 2015). proposed to have occurred during the LH IIIC period formed the nucleus of Kilian s earthquake hypothesis and may later have bolstered his interpretation of the destruction at the end of the palatial period. Åström and Demakopoulou (1996) raised the possibility of an earthquake signature at the end of LH IIIB2 at Midea. However, the collapsed, distorted, curved, and tilting walls found in most trenches are not specified. At the West Gate (Fig. 3), the rubble of collapsed walls was found on top of a thick fire destruction layer (DO10). If a fire occurred as a secondary earthquake effect, one would expect first the toppling of the walls at the moment of ground shaking and traces of fire destruction on top of the rubble, unless the walls stayed intact during the earthquake and decayed later. Åström (1985) describes a 0.40-m-thick ash layer from LH IIIB2 near the interior of the Midea cyclopean wall that contained animal bones and carbonized remains of figs, beans, and olives. The gnawed bones showed that the parts richest in meat were thoroughly utilized during the period of catastrophe (Åström, 1985). Further, Åström (1985) interprets the total lack of mollusk shells in this layer, in contrast to their existence in the layers above and below, as a possible sign of lack of access to the sea as during a siege. Influenced by Kilian s earthquake hypothesis, Åström (1985), later in the same article, suggests that damage at Midea s West Gate was coseismic. Field Campaigns and New Data During two field campaigns in September 2012 and July 2013, active and passive seismic experiments and a long-term deployment of 10 mobile seismic stations for nine months were complemented by gravimetric measurements. Although the first campaign concentrated on the subsurface exploration and determination of engineering seismological parameters at and around Tiryns, measurements during the second campaign were extended to the area between Tiryns and Midea (Hinojosa- Prieto and Hinzen, 2015; Hinzen, Hinojosa-Prieto, et al., 2016). In addition to the geophysical experiments, both citadels were scanned in part with a 3D laser for further analysis, particularly of the damage patterns of the Cyclopean walls. Rephotography was also employed to visually investigate changes since the first excavations in the late nineteenth century. Active Seismic Experiments Seismic-wave velocities were determined using 2D P- and S-wave refraction tomography along 12 and 9 profiles, respectively (Hinojosa-Prieto and Hinzen, 2015). The interpretation of the seismic profiles was complemented by archaeological stratigraphy data, stratigraphic logs, and structural data of the exposed bedrock. The tomograms show a transition from unconsolidated fine alluvium at the surface to consolidated clays and silts to hard limestone bedrock. The soil bedrock boundary dips away from the citadel, forming the geometry of overlying soils into wedges. Figure 9 shows a three-layer model derived from the tomograms. Analysis of the seismic tomography profiles (Hinojosa- Prieto and Hinzen, 2015) resulted in good estimates for the thickness and geometry of the Late Pleistocene Holocene fine-grained soils and the depth and morphology of the soil bedrock interface. P- and S-wave velocities range from 200 to 3500 m=s and 120 to 2000 m=s, respectively, with the mean velocities of the layers given in Figure 9. The top of the bedrock varies from 7 to 17 m depth in the west and 2 to 7 m in the east. The soil bedrock boundary is locally water saturated near the ridge before it transitions laterally from intermediate to undersaturated conditions as far as 100 m away, based on the Poisson ratios of the east west-oriented profiles. Faulting was not identified either by surface mapping or in the tomograms. Passive Seismic Experiments At 182 station locations and with six arrays, we recorded ambient noise data from which Rayleigh-wave ellipticity curves and dispersion functions were calculated. In addition, records of local earthquakes from the nine-month deployment of 10 stations were used to calculate standard spectral ratios (SSRs). As outlined in detail by Hinzen, Hinojosa-Prieto, et al.

12 Reassessing the Mycenaean Earthquake Hypothesis 1057 the interpreted gravity profile (G05_09) which runs from the shore of the Argolic Gulf in the west, passes south of the Tiryns limestone knoll, and ends at the foot of the Profitis Ilias hill (Fig. 10). The gravity model agrees well with the general structure of the subsurface from the HVSR measurements along the same profile (Hinzen, Hinojosa- Prieto, et al., 2016). A second example (Fig. 11b) is the north south-trending profile (G02), which crosses the Tiryns limestone knoll. The Bouguer values for crossing profiles are shown for reference. At the northern gate (gravity station G02P14) and in the range of the South Gallery (station G02P110), the effect of the mass of the Cyclopean walls on the Bouguer values was estimated. The smallest misfit between observations and model was achieved, assuming a density of these walls of 1:34 g=cm 3. The gravity model supports the structures revealed by the active seismic measurements and helped determine the material densities used in the site-effect evaluations. Figure 10. (a) Photo of a gravity measuring point below the western section of the Cyclopean wall of the Upper Citadel, Tiryns. (b) Red lines show the trend of gravity profiles in the surroundings of Tiryns; profile labels are in red; a plan of the citadel is shown in black. The dashed rectangle indicates the area shown in detail in (c); refraction seismic lines are given in blue color. (2016), the horizontal-to-vertical spectral ratio (HVSR) were inverted to 1D velocity models that formed the basis for siteeffect evaluations. These in turn could be compared with SSRs at the locations of the temporarily deployed stations. Gravimetry A total of 650 gravity stations were measured with a Scintrex CG-5 gravimeter, and measuring points were surveyed by differential global positioning system. Figure 10 shows the profile lines close to the Tiryns citadel that were interpreted by standard 2D forward modeling. Special attention was paid to locations where the profiles passed the Cyclopean walls. Here, the gravitational effect of the walls was estimated individually by 2.5D modeling to values between 0.08 and 0.18 mgal, which agree well with the observed effects (Fig. 11). Starting models for the gravimetric interpretation were based on the results of the refraction tomography. Figure 11 shows Rephotographs A problem for archaeoseismic studies at sites where excavation started in the early years of archaeology are undocumented changes to the structures including excavation, alteration, restoration, reconstruction (Straub, 2008), and decay. Rephotography of early photographs and even nineteenth century naturalistic paintings, usually made with the help of optical instruments, can reveal information about major and minor changes, which, if not recognized, can lead to misinterpretations of potential seismic effects. We used rephotography of the Tiryns citadel to investigate alterations, mainly for changes of the monumental Cyclopean fortification walls. Examples of rephotography are shown in Figure 12 at the West Gate, the passage connecting Lower Citadel and Upper Citadel, and the Western Staircase (Fig. 3). Comparisons between the excavation in 1938 and its contemporary status (as of 2012) show the extent of the restoration that added most of the blocks surrounding the West Gate. The passage from the Lower Citadel to the Upper Citadel that leads up to a gate comparable in size with the Lion s Gate of Mycenae(Schliemann, 1886) was cleared from the debris of the Cyclopean walls to the east and west. During early excavations, some of the larger blocks were even blasted before they were removed (Müller, 1930). In the background of the photo of the Western Staircase dating from 1930, many Cyclopean blocks can be seen, which eventually ended up as a debris field outside

13 1058 Klaus-G. Hinzen et al. Figure 11. (a) Interpreted gravity profiles number G05_09. The profile starts in the west close to the coast and ends in the east at the Profitis Ilias (see Fig. 10). The top graph shows the observed Bouguer anomaly and the calculated values (red line) which belong to the section below the graph. Material and densities are given in the legend. (b) Interpreted gravity profile G02. The profile starts north of the citadel and crosses the Lower and Upper Citadels (see Fig. 10). Green symbols and labels indicate the observed Bouguer anomalies from crossing profile lines. The blue bars indicate the estimated effects of the Cyclopean walls at the North Gate (point G02P14) and at the South Gallery (point G07P110). the northwestern wall, similar to a debris field still in existence at the northeastern side of the citadel. However, material was removed in the late 1950s and 1960s, and the following restoration hinders further archaeoseismological work in that area. Laser Scan Models Laser scanning has become an established procedure in archaeoseismological studies (e.g., Fleischer et al., 2010; Hinzen et al., 2010, 2013; Schreiber et al., 2012; Hinzen, Schwellenbach, et al., 2016), and we used this technique to document parts of the cyclopean walls. The wall sections in Tiryns and Midea, which presumably are in their original (not anthropogenically altered) state, do not exhibit any damage that can be exclusively interpreted as coseismic. On the other hand, the walls do show signs of significant degrading due to slow processes. The clay that originally filled the gaps of the cyclopean blocks has mostly been washed out or removed by rodents and lizards (Dörpfeld, in Schliemann, 1886; Iakovidis, 1999). This allowed erosional forces to easily invade the limestone and thus harm the integrity of the purely gravitational walls. Specifically, in Tiryns this phenomenon is evident where the reddish limestone blocks are more weathered than the gray blocks. An example of these two types of limestone is shown in Figure 13 with the heavily degraded face of a reddish block which has lost 0.2 m of its width documented by a profile taken from the laser-scan model of the passage between the Lower Citadel and the Upper Citadel. This type of observation is not new; it was already noticed by Dörpfeld (in Schliemann, 1886) who describes the two kinds of limestone as light-gray and red. Although the gray one is hard and weather resistant, Dörpfeld describes numerous red blocks which became weathered and eroded over many hundreds of years and could no longer bear large stress. Karo (1934) describes most of these blocks as broken. Laboratory tests on rock samples (see Data and Resources) showed a significantly higher average Young s modulus of the gray (27.2 GPa), as compared to the reddish (17.7 GPa), limestone. Dörpfeld assumed that stress-related failure of such reddish blocks caused the collapse of most ceilings and wall sections. A single degraded block could cause all overlying blocks to fall down. He reports that his team replaced several reddish blocks that were in critical condition with cement stonework and also documented five breakouts on the outside of western Cyclopean wall of the

14 Reassessing the Mycenaean Earthquake Hypothesis 1059 Figure 12. Examples of rephotography from the Tiryns citadel. From left to right, the columns show under (a) the situation at the West Gate (view from west), (b) the passage connecting the Lower Citadel and the Upper Citadel (view from north), and (c) the upper segment of the Western Staircase (view from south). Photos in the top row were taken in the year indicated at the left margin (together with the source). Middle row photos were taken during the field campagn in 2012 (photos by J. Tzislakis and K.-G. Hinzen). The interpretation of alterations at the excavation in Tiryns are shown in the bottom row: red, restored; blue, missing; green, unclear; yellow, removed soil; orange, wall cleared from debris; black, unchanged. Lower Citadel, a section that was restored in the second half of the last century (Fig. 12). In Tiryns, major alterations were carried out on the building remains because excavations began in the latter half of the nineteenth century. The main alterations, as documented by rephotographs (Fig. 12), occurred in the western half of the citadel. The least-altered section of the Tiryns Cyclopean wall (DO11) is a segment north of the East Entrance (Fig. 3). South of that entrance, the debris of the wall has been cleared within the fortification, but the outer side is covered by soil and vegetation so that further archaeoseismic interpretation will only be possible after the excavation. To the north of the entrance, the debris field is mostly visible and was subject to a new scan model. Figure 14a gives a perspective clear view from southeast to the model. In a clear view, the scan appears semitransparent. In this way, a glimpse through the front in the 3D model is afforded. It should be noted that changing the opacity of the scan points in a clear view is a powerful inspection tool. One of the few preserved chambers covered by a corbelled vault (eastern wall chamber Ko4) in the Cyclopean wall encircling the Lower Citadel is part of this scan model. The function of these wall chambers has been discussed in the archaeological literature but is not fully resolved. Several chambers were walled up with stones, probably in the final palatial period (Maran, 2009), whereas Kilian (1979) interpreted the work as repairs carried out after the hypothesized earthquake catastophe; however, other reasons (e.g., the assessment that such chambers were thought to be dispensable [Maran, 2009]) are also possible. Although the central part of chamber Ko4 is well preserved (Fig. 14b), the wall sections north and south of it are mostly collapsed. Under earthquake loading, the chamber seems more vulnerable. It cannot with certainty be distinguished whether the collapse pattern of the outer wall shell toward east is of coseismic nature or due to decay. The debris covering the steep slope of the limestone knoll is clearly structured into distinguishable piles. This could be used as an argument for the degrading of the wall

15 1060 Klaus-G. Hinzen et al. Figure 13. (a) Section of the Tiryns Cyclopean wall (eastern side of the passage linking the Upper Citadel and the Lower Citadel). In the center, the photo shows a reddish limestone block, the face of which is eroded. The dashed red line shows the trend of a horizontal cut through the corresponding laser scan. Blue dots in the graph are 12,000 points of a 5-cm-high slice from the scan model along the dashed line indicating the advance of the erosion. (Inset) A detail of the eroded block; (b) example of bioperturbation from the outside of the northeastern section of the fortification wall of Tiryns. The dotted red line marks roots and cut stems of a fig tree that penetrated the wall and crushed the neighboring blocks, creating a large cavern. Labeled arrows in the inset (upper right corner) indicate the location of the photos in (a) and (b) (photos by K.-G. Hinzen). spread out over time and not in a single event. Even if the collapses were coseismic, there is no dating of the event(s). Toward the inside of the fortification, the rubble field is more difficult to read (Fig. 14b) because the undocumented changes in modern times complicate the interpretation; however, it seems that the rubble here is made up of individual piles as well. Dörpfeld (in Schliemann, 1886) also interprets the partial collapse of the Tiryns South Gallery as a consequence of bad construction technique of the large northern wall, where the inner part is composed of smaller stones and clay filling and only on the outside, jacketed by larger blocks. He also mentions anthropogenic destruction: herdsmen, who used the East Gallery for centuries as a sheep stable, may have demolished the walls to gain access to the Gallery from the south. The blocks inside the Gallery at the level of the former stable are worn down and polished to a smooth texture, probably by sheepskin. In the remains of the Cyclopean walls at Tiryns (DO11) and Midea (DO12), the existence of several large cavities indicates entire blocks are missing. These empty spaces might have been either degraded reddish limestone blocks and/or the effect of bioperturbation by plant roots, particularly those of fig trees. Figure 13 shows an example from Tiryns, where a tree was growing out of the wall (but was at some point severed at the base) and where strong roots have cracked and pushed out (probably) two of the Cyclopean blocks. In Midea (Fig. 15), we documented one such cavity of enormous dimensions (1:4 1:8 2:5 m). Such defects in the wall bond can explain most of the damage of the Cyclopean walls. Specifically, the V-shaped breakout of wall sections (Fig. 15) is a consequence of such failure and is interpreted as decay due to a slow erosion process and/or bioperturbation. The steady loss of the filling over three millenia resulted in an altered static condition of the Cyclopean wall. Most probably, the clay was carefully introduced layer by layer as the walls were constructed, and it filled in gaps between the irregular blocks. If a block were still rickety, the ancient builders inserted small stop blocks (Fig. 16) to stabilize the construction. With the weathering of the fill, these smaller blocks and pointy edges of larger blocks assumed the full bearing load of the wall (Fig. 16); this in turn can lead to cracking of blocks under the static load, visible, for example, in Tiryns at the Western Staircase, the East Entrance, and the East Gallery. The Upper Citadel of Tiryns was (re)used after its destruction by the intense fire at the end of LH IIIB. This is evidenced by the postpalatial megaron (Building T) which was erected at the site of the preceding one but with a reduced floor plan (Maran, 2010). It is unlikely that this reuse occurred in a setting with destroyed fortification walls. For representation, defense, and practical purposes, the fortification is interpreted as mostly intact. This assumption is supported by the finding that Cyclopean wall blocks at the Western Staircase (Albers, 1994; Kardamaki, 2009) rested in situ on top of

16 Reassessing the Mycenaean Earthquake Hypothesis 1061 Figure 14. Clear view of the laser-scan model of the eastern section of the fortification wall of the Lower Citadel of Tiryns. (a) Perspective view toward northwest, the arrow points to the chamber Ko4, and three distinct rubble piles are indicated by colored shading. Scale varies with perspective. (b) Orthographic view from west (inside the fortification) to the opening of chamber Ko4 (arrow). Two rubble piles are indicated by colored shading. The inset shows the extent of the scan model (blue rectangle) in relation to the plan of the citadel (plan after Maran, 2010). The arrow points to the outside of chamber Ko4. the debris from the large fire which had been dumped down the western side of the citadel to clear the ground for the new megaron and other constructions (DO13). The gap in front of the restored area on the southwestern part of the fortification wall is attributed by some authors to a collapse of this section of the Cyclopean wall due to an earthquake (DO14) (e.g., Schachermeyr, 1962; Voigtländer, 2003), whereas Schliemann (1886) interpreted it as a landslide. However, as summarized by Kardamaki (2009), the debris layers along the outside of the western Upper Citadel wall were found, at least in the area of the Western Staircase, stratified under Cyclopean blocks of the western wall. Therefore, it is more likely that the debris and rubble caused by the conflagration were cleared from the Upper Citadel and intentionally deposited along the outside of the western Cyclopean wall encircling the Upper Citadel. A possible second burnt layer in the deposits (Kardamaki, 2009) also weakens the argument that the large fire was a secondary earthquake effect; accidental fire, war, or unrest can also be considered as possible causes. Whether the disintegration of the wall in this section was a slow or spontaneous process is not and may never be discernible; however, the observations up to this point do not support a single damaging earthquake at the end of the palatial period. Deterministic Ground-Motion Model To estimate the intensity of ground motions generated by earthquake sources that were potentially activated at the end of LBA, we identified the relevant source zones (Hinojosa-Prieto, 2016), calculated Green s functions, and constructed synthetic seismograms for 27 earthquake scenarios that were then combined with the previously determined (Hinzen, Hinojosa-Prieto, et al., 2016) site amplifications for a number of locations in Tiryns and Midea. The Green s function method uses the elastodynamic representation theorem to construct ground-motion time histories (Aki and Richards, 1980). Slip history of subsources distributed on the assumed activated fault sections are convolved with the Green s functions and integrated over the entire source. We used an implementation of the method by Wang (1999). The resulting synthetic seismograms served as rock outcropping ground motions in the forward modeling of site-specific ground motions using the 1D site-response models from Hinzen, Hinojosa-Prieto, et al. (2016) and a set of MATLAB routines SUA (see Data and Resources; Robinson et al., 2006) within a frequency band from 0.1 to 20 Hz. Although the shallow double limestone knoll of Tiryns shows only small site amplifications below a factor of 2 at frequencies between 2 and 10 Hz, the soft sediment in the surrounding area of the Lower Town has amplifications of 4 6. At the Midea ridge, the topography results in ground-motion amplifications of a factor of 2 3 at frequencies between 1 and 3 Hz. Finite-Fault Models Four source zones were considered (Hinojosa-Prieto, 2016): (1) the HSZ, (2) the eastern limit of the PCCR system located 50 km north of the Argive basin, (3) the IEST fault system km southeast of the basin, and (4) the Argive basin itself. Although the Argive basin and the PCCR are dominated by normal faults, the IEST comprises strike-slip faults, and reverse faulting is the prevailing mechanism of the HSZ. Historic and recent seismicity (Figs. 1 and 2) shows the high activity levels of the PCCR and the HSZ; however, activation of the IEST faults and particularly the Argive basin faults during the Holocene remains speculative as long as palaeoseismic studies are lacking (Papastamatiou et al., 1960; Tataris et al., 1970; van Andel et al., 1990; Papanastassiou et al., 1993; Papanikolaou et al., 1994; ECPFEPO, 1996; Piper and Perissoratis, 2003; Georgiou and Galanakis,

17 1062 Klaus-G. Hinzen et al. Figure 15. The inset in the upper right corner gives a plan of the archaeological site of Midea (after Demakopoulou, 2012); the blue rectangle delineates the position of a 3D laser-scan model. (a) Orthographic top view to the laser-scan model of the northern fortification wall. The leftmost yellow rectangle shows a section enlarged in (b) of one of the V-shaped break-outs of the Cyclopean walls. The red line marks the extension of the failure, and the dashed yellow line marks the debris fan. The rightmost yellow rectangle in (a) gives the extension of a horizontal section through the wall shown in detail in (c) with a large cavern, the maximum extension of which is marked by the yellow line. 2010; Karastathis, Karmis, et al., 2010; Karastashis, Papadopoulos, et al., 2010; Mitropoulos and Zananiri, 2010). Nevertheless, because of the close proximity to Tiryns and Midea, we included earthquake sources in these zones in the model. Table 1 presents the surface rupture length and orientation of the faults with the corresponding references and the fault projections shown in Figure 17. M w and depth extension (DRW) were estimated using the global empirical relationship of Wells and Coppersmith (1994). A reverse-fault offshore southwestern Crete, Greece, (HSZ-8 in Table 1) is of particular interest because there is good consensus that it caused the great Cretan earthquake in 365 C.E. with M w 8.5 or 8.3, according to Stiros (2010) and Shaw et al. (2008), respectively. Although the time of this historic earthquake postdates the Mycenaean palatial period, this fault is modeled here to estimate the effects of a similar earthquake in LBA Tiryns and Midea. Elastic dislocation analysis of coastal uplift data coupled with historical, archaeological, radiocarbon, seismological, and large-scale tectonic data show that the 365 C.E. earthquake caused 9 m coastal uplift in western Crete and widespread destruction (Stiros, 2010). Ruptures on the rectangular faults were modeled uni- or bilaterally, with a rupture velocity of 2:7 km=s and a circular Figure 16. Point loading of Cyclopean blocks in the fortification wall of Tiryns. (a) A small stop block was used to close the gap between two large blocks at the northeast wall (see inset for location). (b) Large block of the tower next to the main entrance in the east resting on points, including adjustment pieces (enlarged in zoom window). (c) Another large block with point contact at the same tower (photos by K.-G. Hinzen).

18 Reassessing the Mycenaean Earthquake Hypothesis 1063 Table 1 Source Parameters of 25 Earthquake Scenarios for the Computation of Synthetic Site-Specific Ground Motions for Tiryns and Midea Source Fault Strike/Dip ( ) SRL (km) DRW (km) M w Reference HSZ 1 346/45NE H89; P /45NE H89; P /25NE P /25NE P /25NE P /25NE P /40NE H89; P /40NE S10 PCCR 1 295/30NE K10a; K10b 2 295/40NE K10a; K10b 3 295/50NE K10a; K10b IEST 1 266/75N CR90; K10a; K10b 2 266/90N CR90; K10a; K10b 3 275/75N CR90; K10a; K10b 4 275/90N CR90; K10a; K10b 5 270/75N CR90; K10a; K10b 6 270/90N CR90; K10a; K10b ABNF 1 329/75E va93; E /75E va93; E /75NNE E /75NE E /75W GG /75S E /75SE E /75SW E96; P93 H89, Hatzfeld et al. (1989); CR90, Clift and Robertson (1990); P93, Papanastassiou et al. (1993); va93, van Andel et al. (1993); P04, Papazachos et al. (2004); WC94, Wells and Coppersmith (1994); E96, European Center on Prevention and Forecasting of Earthquakes and Protection Organization (ECPFEPO) (1996); PA96, Papazachos (1996); PC04, Pavlidis and Caputo (2004); B10, Blaser et al. (2010); GG10, Georgiou and Galanakis (2010); K10a, Karastathis, Karmis, et al. (2010); K10b, Karastathis, Papadopoulos, et al. (2010); S10, Stiros (2010); Str10, Strasser et al. (2010); HSZ, Hellenic subduction zone; PCCR, Patras-and- Corinth continental rift; IEST, Iria Epidaurus sinistral transform; ABNF, Argive basin normal fault; SRL, surface rupture length; DRW, down-dip width; E, east; N, north; NE, northeast; W, west; S, south; SE, southeast; SW, southwest; NNE, north-northeast. rupture front; stress drop was varied between 2 and 10 bars; and a fluctuation of 5 and 10 for strike/dip and rake, respectively, was allowed for the subsources. The subsurface model for evaluation of the Green s function (Fig. 18) was compiled(hinojosa-prieto, 2016) from a number of publications on P- and S-wave tomography, Q tomography, and local earthquake studies (Papazachos et al., 1995; Kalogeras and Burton, 1996; Le Meur et al., 1997; Papazachos and Nolet, 1997; Karagianni et al., 2002, 2005; Bourova et al., 2005; Molnar et al., 2007; Endrun et al., 2008; Skarlatoudis et al., 2013). The deeper part of the model is based on global scale models, the preliminary reference Earth model (PREM) of Dziewonski and Anderson (1981), and the Colorado University Boulder (CUB) shear-velocity model of the crust upper mantle from Shapiro and Ritzwoller (2002). Ground Motion Figure 19 shows the peak ground velocity (PGV) amplitudes of the synthetic seismograms for the horizontal groundmotion components versus the peak ground acceleration (PGA) values. With the empirical scheme suggested by Wald et al. (1999), these data also provide an estimate of the seismic intensity (modified Mercalli intensity scale). Five sites in Tiryns were selected that represent the Lower Town (P01A, P04A, and P12P10), the palace on the Upper Citadel (P06A) and the Lower Citadel (HE03), and three sites at Midea, near the summit (HE10), at the northeastern side of the hill (HE09), and a location in the contemporary village (P20P5); station codes and exact locations are given by Hinzen, Hinojosa-Prieto, et al. (2016). For each earthquake scenario from Table 1, Figure 19 shows the rock PGA and PGV and the corresponding site-specific values. For sites outside the citadels located on the soft sediments, yellow symbols were used, and brown symbols indicate sites inside the citadels. In all scenarios, the sites outside the citadel were subjected to significantly higher ground-motion levels than sites within the citadels. Among all scenarios of subduction-zone earthquakes (HSZ), only two sites in the Lower Town displayed intensities above VI for a deep M w 8.1 earthquake below the Argive basin. The sites within the citadel in Tiryns and Midea do not show intensities that would equate to structural damage. The same results hold true for all scenarios and sites in case of the PCCR earthquake scenarios. The simulated activation of the Iria and Epidauros faults leads to several intensity VII values for the Lower Town of Tiryns; here, even intensity VIII is reached at one site. In Midea, two scenarios with an activation of both faults together (IEST 5 and 6) lead to intensity VII, also inside the citadel. Intensities VII, VIII, and IX result only for the case of an assumed activation of local faults in the Argive basin, particularly in those scenarios with Joyner Boore distances of zero for either Tiryns or Midea. For the local fault results, much larger intensities are predicted for sites outside the citadels than inside. Discussion and Conclusions Unfortunately, most of the findings that in Kilian (1996) were interpreted as signs of seismogenic damage are no longer accessible in their original excavated state. Specifically, none of the walls interpreted as coseismically undulated can be further investigated; they were either covered with soil or have been restored. In these cases, excavation photos and drawings were the only source material available for examination. The evidence from Kilian (1996) presents the strongest arguments to date that could support the earthquake hypothesis, and Kilian s results were repeatedly referenced (e.g., Jones and Stiros, 2000; Nur and Cline, 2000; Force, 2008; Nur and Burgess, 2008). In this contribution, we were able to show that none of these individual damage descriptions represent a clear and unequivocal earthquake indication. It is noteworthy that after Kilian s death in

19 1064 Klaus-G. Hinzen et al. Figure 17. Maps showing the location of linear fault segments (red lines) and surface projections (light color shaded rectangles) used to calculate synthetic ground motions for earthquakes of (a) the HSZ and (b) strike-slip and normal faults in the region surrounding Tiryns and Midea. The location of the latter is indicated by red labeled circles. Labeling of the faults corresponds to the list in Table 1. The blue dots indicate instrumental seismicity of Greece ( ) after the Greek earthquake catalog (see Data and Resources). Figure 18. Subsurface 1D model used to calculate Green s functions between the earthquake sources from Figure 17 and the sites of Tiryns and Midea; V P and V S are P- and S-wave velocities; Q P and Q S are P- and S-wave Q factors (after Hinojosa-Prieto, 2016). 1992, excavations in Tiryns were resumed in 1997 and have progressed unabated until the time of the writing of this article; however, since then no new convincing evidence for earthquake damage has been found. Between 2011 and 2014, Schultz and Schmidt-Schultz (e-report DAI 2015, see Data and Resources) examined the bones of human remains at Tiryns. Taken together they constitute at least 174 individuals; most of them are not well preserved. In addition to the burials in the chamber tombs at the Profitis Elias southeast of the Tiryns acropolis, several individuals were buried inside the Lower Citadel, intra muros, or in open areas. The reason for these types of burials is not yet clear, particularly as all were unfurnished burials. Of the skeletons, 49% were nonadults and 3.5% were above age 60. Traces of infectious diseases typical of the time were found; in some cases, the cause of death (meningitis, nonearthquakerelated traumata) was diagnosed. However, no evidence for any in situ earthquake victims was found. A study of human remains from Midea excavations is lacking. The scenario described by Papanastassiou et al. (1993) is unlikely because it would require a local earthquake source at Tiryns for the damage in 1330 and 1150 B.C.E, which did not affect structures in Mycenae and Midea. We also maintain that it is unlikely that an earthquake in 1250 B.C.E. caused damage at Mycenae and Tiryns but not at Midea because the distance between Mycenae and Midea is smaller than that between Mycenae and Tiryns. Unfortunately, the uncertainty which must exist in the archaeological dating (Jusseret and Sintubin, 2013) of the seven destruction layers is not quantified or even further discussed.

20 Reassessing the Mycenaean Earthquake Hypothesis 1065 Figure 19. Ground-motion parameters of simulated seismograms for Tiryns with four graphs for the earthquake source zones that are indicated in the legend; labels of the black dots correspond to the earthquake scenario numbers from Table 1, in which letters N and E indicate the north and east component of the horizontal ground motion, respectively. Each plot shows peak ground acceleration (PGA) versus peak ground velocity (PGV) values. Black dots are the estimated PGV/PGA for rock; black lines connect to the colored dots that represent the site-specific ground motions (P01A, P04A, and P20P10 Lower Town, HE03 Lower Citadel, HE06 Upper Citadel). Gray shaded rectangles indicate the corresponding modified Mercalli intensity after Wald et al. (1999). Ground-motion parameters of simulated seismograms for Midea (P20P10 outside citadel, HE09 and HE10 inside citadel). A plausibility matrix (Hinzen, 2005; Galadini et al., 2006) combines the observed and analyzed damage from a site with possible causes by stating whether a certain cause is feasible, questionable, or unfeasible for a specific damage or observation. Plausibility can be quantified by a number between 0 and 1 (Hinzen et al., 2012), meaning that none or all of the observations can be explained by a certain cause, respectively. Figure 20 shows the plausibility matrix summarizing the findings of this study. Eight causes are linked to the 14 observations, DO1 DO14, which have been discussed. Natural decay, anthropogenic action, and no damage/intentional setting with an index of 0.61, 0.43, and 0.39, respectively, are a more likely or equally plausible cause for a larger number of observations than earthquake ground motions. For these, several earthquakes over a longer time span result in an index of 0.39 compared to an index of 0.29 for a single large earthquake catastrophe at the end of LB IIIB, showing that there is some evidence for and against the hypothesis. In conclusion, there is no one preeminent cause which would explain the observations, taken as a whole or in their diverse characteristics. This is not an unusual result for archaeoseismological studies (e.g., Hinzen et al., 2012), particularly for structural remains after 3200 yrs and almost 130 yrs of excavations. As outlined by Ambraseys (2006) in the past few decades, there has been a reemergence of neocatastrophism in the field of archaeoseismology, particularly for earthquakes in the Eastern Mediterranean. This trend continued after the Ambraseys (2006) publication, as described by Jusseret and Sintubin (2013). Ambraseys (2006) further attributes the revival of catastrophe hypotheses to their usefulness as easy explanations. This general observation also applies to the case of Tiryns and Midea. Jusseret and Sintubin (2012) describe the problems associated with an interpretation of single catastrophic earthquakes with regard to the decline of the Minoan culture on Crete at the beginning of the LBA. The most problematic barrier to such an interpretation is dating uncertainties (e.g., Jusseret et al., 2013). In summary, the seismicity of the Peloponnese and its surroundings indicates that several earthquakes have shaken the Mycenaean citadels of Tiryns and Midea in the past 3500 yrs. However, in Tiryns and Midea we do not see strong evidence for a single devastating earthquake catastrophe at the end of the palatial period (LH IIIB). Specifically, most of the key evidence presented in the posthumously published paper of Kilian (1996) does not remain convincing, based on the results presented in the current study. A single catastrophic earthquake responsible for ending the Mycenaean palatial period would have completely destroyed the Tiryns Upper Citadel, parts of the Lower Citadel, and would have caused a devastating fire as a secondary earthquake effect. In reality, the hypothesized earthquake did not leave any traces in the excavated areas of Lower Town and did not significantly damage the Cyclopean walls of Tiryns. Based on our results from estimating the local earthquake site effects, we consider the single earthquake

21 1066 Klaus-G. Hinzen et al. Figure 20. Plausibility matrix linking eight potential causes of damage (rows) with 14 observations or damage descriptions (DO1 DO14) in the study area. The index is a measure of plausibility of a single cause in terms of all observations and damage descriptions (Hinzen et al., 2012). scenario to be unlikely. New evidence can be expected from future excavations in the Lower Town of Tiryns. If such a devastating earthquake occurred, it must have left a pronounced earthquake stratum in the particularly vulnerable Lower Town. Future palaeoseismological research of the local faults in the Argive basin should provide confirmation about their level of activity and also help bracket the time frame. Such a study would be interesting for its archaeoseismological relevance and would also be crucial for risk studies of the area, including for modern densely inhabited cities, such as Nafplion and Argos. Data and Resources Seismic instruments were made available by the Geophysical Instrument Pool Potsdam (GIPP) of GeoForschungs- Zentrum Potsdam (GFZ) (Grant Number GIPP201211); the recorded data are available at doi: /GIPP Excavation photographs were made available by the archive of the Deutsche Archäologische Institut (DAI) in Athens. Borehole data were provided by E. Zangger. Seismicity data were taken from the catalog of the Aristotle University of Thessaloniki available at catalogs_en.html (last accessed December 2015) and the University of Athens available at geol.uoa.gr (last accessed December 2015). The e-report 2015 of the DAI is available at org/journals/index.php/efb/article/download/1628/4523 (last accessed June 2017). The code to calculate Green s functions and synthetic seismograms was made available by Wang (1999). Laboratory rock tests were performed by the RWTA Aachen, Institut für Grundbau, Bodenmechanik, Felsmechanik und Verkehrswasserbau, Project Satellite images are from Google Earth ( last accessed March 2017). MATLAB used for ground-motion model is available at (last accessed March 2017). Acknowledgments The authors thank the Ephorate of Antiquities of the Argolid in Nafplion, especially the Ephor A. Papadimitriou for the permission to conduct the field work. The authors acknowledge the support during the field work by S. Prillwitz, E. Seferou, P. Soupious, M. Nikoloakaki-Kentron, Nikos Papa, A. Papageorgopoulou, P. B. A. Hinzen, and all guards at Tiryns and Midea. The authors thank S. Falter, N. Haaf, T. Kalytta, T. Koch, S. Voigt, M. Ohrnberger, and M. Zeckra for the help with data processing and J. Power and I. Schwellenbach for their discrete element modeling efforts. The authors thank H. Igel, P. Marzolff, S. Mechernich, B. Tezkan, M. Vetters, and S. Stiros for interesting discussions, and E. Zangger for sharing the