Earthquake-triggered landslides and mudflows: Was this the wave that engulfed Ancient Helike?

Abstract

The destruction of Ancient Helike in 373 BC as reported by ancient Greek and Roman writers is inconsistent with modern evidence on the geological context. The classical view of a strong earthquake, similar to the 1817 M = 6.6 earthquake and followed by a giant tsunami wave that permanently inundated the ruined city does not stand up to modern scrutiny. Evidence for co-seismic slip on the Helike Fault at that time and for a corresponding tsunami have already been shown to be lacking, and the archaeological evidence shows that part of the site was reoccupied within 40 years. New observations on outcrops, excavated sites, and boreholes show that at least two mudflow deposits several metres thick of slightly gravelly mud overlie 4th c. BC archaeological remains on the Katourlas fan. Upstream, landslides are common in the 4 km² river basin and could have dammed the ephemeral Katourlas River. Relics of this dam are still recognised in the riverbed as a knickpoint. Temporary damming of rivers by earthquake-triggered landslides is a common phenomenon in northern Peloponnese. The destruction of Ancient Helike appears to have been a localised non-linear cascading series of disasters, with a strong earthquake followed by a destructive mudflow generated by breaching of a landslide-dammed lake in the Katourlas basin. Historical records of flooding from the west may record another landslide dam followed by a more watery flood in the Selinous River. We propose that it was mudflows and floods from inland rather than a tsunami from the sea that contributed to the final destruction of Helike.

Keywords

breaching dam cascade disaster engulfment of Ancient Helike Greece Gulf of Corinth Helike area mudflows

Introduction

According to a large number of ancient accounts, in 373 BC, a strong earthquake and accompanying enormous sea waves (tsunami) destroyed and submerged Ancient Helike, which was the capital of both the Ionian and the Achaean Dodekapolis in the region of ancient Achaea (Katsonopoulou, 2005). Half a century ago, between 1950 and 1973, the late archaeologist Marinatos made several attempts to locate Ancient Helike with inconclusive results. He considered that an earthquake and a tsunami demolished the city, but he suggested, in addition, that the city was probably partly buried by fluvial deposits (Marinatos, 1960). Over the last thirty years, the Helike Project has mounted a campaign using a multidisciplinary geoarchaeological methodology and archaeological excavations to locate Ancient Helike and to determine the exact causes which led to its destruction.

The capital city of Ancient Helike was located, southeast of Aigion, in the Gulf of Corinth basin (Figure 1), a geotectonic domain well-known for its strong seismicity (Papazachos and Papazachou, 1997, 2003). However, since this historical seismic destruction, over 2500 years ago, there has been a striking lack of a similar catastrophe within the Gulf of Corinth basin. The closest analogue was on 26 December 1861, when a large earthquake damaged more or less the same area (Schmidt, 1879). Thus, a fascinating mystery remains whether the destruction of Ancient Helike was the consequence of an unanticipated event (a ‘black swan’, Aven, 2015) or the result of cascading disasters, when a trigger event led to a series of follow-on impacts (a ‘linear path of events in disasters’, Pescaroli and Alexander, 2015) that overlapped and interacted during the 373 BC event. We consider as a ‘black swan’ an earthquake larger than seismologists think could happen, followed by a correspondingly enormous tsunami inundating the city. In a ‘cascade disaster’ we consider vulnerabilities that overlap and interact, with or without escalation points, and create secondary effects of greater impact than the primary natural hazard (Pescaroli and Alexander, 2016).

Figure 1.

(a) Structural map of the Gulf of Corinth showing active faults (modified after Zygouri et al., 2008). WCF: West Channel Fault (after Nixon et al., 2016). (b) Detailed structural map of the study area (from Doutsos and Poulimenos, 1992; Koukouvelas and Doutsos, 1996). PF: Pyrgaki Fault; MAF: Mamoussia Fault; MEF: Melissia Fault; KF: Keryneia Fault; WHEF: western Helike Fault; EHEF: eastern Helike Fault; AF: Aigion Fault.

The ancient city was founded on the distal part of an alluvial fan from the small Katourlas ephemeral river near the modern village of New Keryneia (Figures 1b and 2). It was located on the hanging-wall block of the Helike Fault and the footwall block of the Aigion Fault (Figure 1b; Koukouvelas, 1998). So, uplift and subsidence in the area of the city depended on its proximity to each fault. Several previous papers focused on the hanging wall of the Helike Fault, because there the activity of the two faults caused clockwise and anticlockwise avulsion of the Selinous and Kerynites rivers, respectively, before and after the 373 BC event. Over the late-Holocene period, this avulsion buried the ancient city under thick deltaic sediments (Kontopoulos et al., 2017; Koukouvelas et al., 2001, 2005; Marinatos, 1960; Pavlides et al., 2004).

Figure 2.

Overview of the study area based on a GoogleEarth scene. White lines define the Katourlas drainage basin and red Katourlas Fan. The prominent steep slope lengthwise in the middle of the figure is the Helike Fault scarp. KL: Klonis Field; BAL: Balalas Field; Kou: Koutroumanis Field (where four boreholes 67, 68, 69, 70 and three geoarchaeological excavations H67, H68 and H69 have been carried out); PF: Papafilippou Field; Sa: Saitis Field. EHII-III: Early Helladic II-III. H indicates archaeological excavations and B boreholes by the Helike Project. EL and WL are landslides presented in Figure 3.

Previous authors have assumed, based on the literary sources, what we term the coastal disaster model (CDM). This proposed tremendous changes in the coastal zone related to co-seismic earthquake subsidence and tsunami inundation, linked to strong seismicity hosted by the Helike Fault (Doutsos and Poulimenos, 1992; Engel et al., 2016; Ferentinos et al., 2015; Psarropoulos and Gkantona, 2020). This long-standing assumption considers that the combination of earthquake shaking and building collapse with tsunami inundation in the coastal zone caused mass death of the inhabitants. The secondary effects of the earthquake, like liquefaction and the subsidence of the coastal zone (Pavlides et al., 2004), caused socioeconomic problems to the surviving inhabitants, since their cultivated land was flooded (Engel et al., 2016).

However, recent data indicate that the Helike Fault located south of the ancient city, shows no co-seismic slip related to the famous 373 BC earthquake (Koukouvelas et al., 2001, 2005; Koukouvelas, 2008), so that any changes in the coastal zone were unrelated to this fault (Papadopoulos et al., 2007). In this paper, we propose a new model for the Ancient Helike catastrophe, the inland-driven disaster model (IDDM). This model considers the risk due to active processes, like damming of rivers in the footwall block and mud flows on the hanging-wall block of the Helike Fault during the 373 BC earthquake.

In this contribution we will provide evidence that geological processes on the footwall block of the Helike Fault can produce disasters in the lowland area of the hanging-wall block. We expand on the hypothesis, briefly introduced by Kontopoulos et al. (2017), that heavy rainfall and mudflows played a crucial role in the Helike 373 BC catastrophe. We present detailed geomorphological and sedimentological analyses of the Katourlas drainage basin above the coastal area where Ancient Helike was built (Figure 1). The geomorphological analysis is based on digital surface models and UAV flights. The sedimentology is based on boreholes and trial excavations, and field outcrops in the footwall block. The dating of the mudflows is based on extensive archaeological excavations by the Helike Project. More generally, we assess the merits of the black swan hypothesis for the 373 BC earthquake and the tsunami as a generator of vulnerabilities for the CDM compared to the localised non-linear cascade disaster hypothesis for the proposed IDDM.

The ancient documentary evidence and its interpretation

There are several details about this catastrophic earthquake described by ancient authors, both contemporary with the event and later ones. Herakleides (4^th c. BC, reported by Strabo), Diodoros (1^st c. BC) and Aelian (2^nd–3^rd c. AD) report that the earthquake happened at night. Strabo (1^st c. BC–1^st c. AD) and Pausanias (2^nd c. AD) note that the earthquake happened during winter, whereas Aelian adds that the event occurred after a 5 days period of unusual behaviour of animals. Herakleides also mentions (in Strabo 8.7.2) that although the city was twelve stadia [2.2 km] distant from the sea, this whole district together with the city was hidden from sight; and two thousand men who were sent by the Achaeans were unable to recover the dead bodies. Aristotle (4^th c. BC), combines the earthquake with the appearance of a comet, and repeats that the earthquake was followed by a wave that invaded from the west towards the south where the already ruined city was located. Three centuries later, Diodoros of Sicily (15.48.2-3) wrote: the earthquake . . . came at night, so that when the houses crashed and crumbled under the force of the shock, the . . . majority were caught in the falling houses and annihilated, but as day returned some survivors dashed from the ruins and . . . met with a greater and still more incredible danger. For the sea rose to a vast height, and a wave towering even higher washed away and drowned all the inhabitants and their native lands as well.

Half a century later, Strabo (8.7.2) wrote: For the sea was raised by an earthquake and it submerged Helike and the temple of Helikonian Poseidon . . . And Eratosthenes says that he himself saw the place, and that the ferrymen were saying that a bronze Poseidon stood erect in the strait, holding in one hand a hippocamp, which was dangerous to those fishing with nets. For further detailed information, including relevant texts of ancient sources on the 373 BC catastrophe, see the Appendix in Katsonopoulou (2005).

It has proved difficult to precisely locate the ruined ancient city. This comes from its location in a geologically active landscape. Since the contemporary written accounts are limited, with most of the knowledge passed down by later authors, however based on earlier sources, now lost, we could ‘assume that [the story is] true in essence, imaginative in detail’ (Durant, 1966). We find reliable the account of the earthquake, the burial of the ruined city and the existence of a wave. However, any evidence for a high tsunami still remains elusive: as Engel et al. (2016) wrote: ‘the evidence for tsunami deposition is ambiguous and the interpretation remains hypothetical’.

The detailed descriptions of the ancient authors are interpreted in a variety of ways by a plethora of published articles over the last seventy years. Some researchers considered that the earthquake of 373 BC was greater than the 1861 earthquake (Leonards et al., 1988; Marinatos, 1960; Mouyaris et al., 1992; Papazachos and Papazachou, 1989; 1997; Perrou et al., 2013), or the event resulted from reactivation of multiple fault segments (Console et al., 2015; Lambotte et al., 2014). In addition, together with the tremendous earthquake it has been proposed that Ancient Helike was drowned due to a huge translational slide, which caused 10 m of subsidence of the city below its pre-earthquake elevation, bringing it below sea level and thus amplifying the effects of tsunani inundation (Ferentinos et al., 2015; Leonards et al., 1988; Marinatos, 1960; Papadopoulos, 1998; 2003; Papanastassiou, 2002). In one interpretation, this subsidence/inundation scenario is even more strengthened by the suggestion of a landslide causing subsidence of the city just before the 373 BC earthquake. Concomitantly, the already subsided Helike was ruined by the earthquake before the final tsunami invasion (Lyritzis et al., 2019).

An opposite school of interpretation, less commonly presented, considers that the earthquake impact is exaggerated, and the earthquake magnitude was of the order of 6.6, or maybe less, accompanied by limited disaster and no significant tsunami (Ambraseys, 2009; Ñaco del Hoyo and Nappo, 2013). Similar uncertainties exist about the causative fault or faults of this famous earthquake, with most authors speculating the Helike Fault as the candidate (Doutsos and Poulimenos, 1992; Lambotte et al., 2014; McNeill et al., 2005; Mouyaris et al., 1992, among others). Although these interpretations cover several possible disaster scenarios, all use the accounts of ancient writers, together with historical seismicity catalogues (Papazachos and Papazachou, 1989, 1997, 2003) and the assumption of Marinatos (1960) linking the 373 BC earthquake to the destruction of Ancient Helike. In contrast to the plethora of interpretations, relevant modern geological data analysis about the tsunami in particular is limited.

Geological setting of Ancient Helike and the Katourlas River basin

Ancient Helike was located on the southern coastal plain of the westernGulf of Corinth, 40 stadia (about 7 km) southeast of Aigion (Pausanias, 7.24.5) and 12 stadia (about 2.2 km) from the sea (Herakleides, in Strabo 8.7.2; Figure 2). Ancient Helike was located near the foot of the mountains north of the Helike Fault, close to the Katourlas alluvial fan supplied with alluvium by the ephemeral Katourlas River.

In the western Gulf of Corinth (Figure 1b), between the modern towns of Aigion and Diakopto, a series of six major north facing, almost E–W trending normal faults (the Aigion, Helike, Keryneia, Melissia, Mamoussia and Pyrgaki faults) controlled the accumulation of sediments and the regional geomorphology (Koukouvelas, 1998; Poulimenos and Doutsos, 1996). The 22 km long Helike Fault has hosted at least seven major seismic events in the last 8000 years, based on historical records and palaeoseismological studies (Koukouvelas, 2008; Koukouvelas et al., 2001; 2005; Pavlides et al., 2004). It is divided by an almost 2 km long step-over zone into two prominent segments, the western with a length of 9 km and the eastern with a length of 13 km (Figure 1b; Pavlides et al., 2004). To the east, the Helike Fault scarp separates alluvial sediments in the hanging-wall from Neogene fan deltas in the footwall, which are bounded to the south by the Keryneia Fault with alpine basement rocks (Pindos unit limestones and flysch) in its footwall. The western part of the footwall of the Helike Fault comprises Cretaceous limestones and Jurassic cherts (Gawthorpe et al., 1994, 2017; Pavlides et al., 2004).

The structural setting of the Katourlas basin and the analysis of the Helike Fault is well known from previous studies (Koukouvelas and Doutsos, 1996; Koukouvelas and Papoulis, 2009). The study of the fault is based on trenching palaeoseismology, archaeological excavation, cored boreholes and analysis of tectonic geomorphology. As a result, the Helike Fault is the best documented and best understood active fault in Greece. More than 10 palaeoseismological trenches were excavated across both segments of the fault and yielded four estimates of slip rate and recurrence. Slip rate is estimated between 0.5 and 2 mm/year in different segments and at different time intervals, with mean slip per event in the order of 1 m, the recurrence interval is 400–700 year and the maximum expected earthquake is in the order of 6.7 (Koukouvelas, 2008; Koukouvelas et al., 2001; 2005; Pavlides et al., 2004). Remarkably, in all these trenches no colluvial tectonostratigraphy can be recognised in relation to the 373 BC earthquake (Koukouvelas, 2008). Based on this extensive palaeoseismological campaign, reaching depths of more than 5 m in three trenches, and spacing between trenches across the fault in the order of a kilometre, we consider it fair to assume that the famous 373 BC earthquake was not hosted on this fault. As the Helike Fault includes two segments and it seems that neither was reactivated during 373 BC, the assumption of reactivation of multiple fault segments is not supported. In this part of the basin, an array of faults is recognised both onshore and offshore and three of them can be considered based on their length and proximity to the study area as candidate faults to host the causative earthquake (Figure 1): they are the West Channel (Figure 1a), Keryneia, Melissia and Mamoussia faults (Figure 1b). All of them are shorter in trace length than the Helike Fault and can host earthquakes on the order of M = 6.4 (Caputo and Pavlides, 2013; Papadopoulos, 1998).

In the mountains along the Katourlas River, 1 km upstream from the Helike Fault, is the village of Ano (Upper) Keryneia, also known as Gardena. The name Gardena corresponds to gardh-i in Albanian, meaning dam or fence, and probably corresponds to the Aromanian word, also known as Macedo-Romanian or Vlach language spoken in the southeastern Europe, gardu with the same meaning (Thanasoulopoulos, 2007). In the last century, inhabitants of this village moved to lower lands and founded New Keryneia on the apex area of the Katourlas fan (Figures 1 and 2). The Katourlas River drains a basin of 4 km² and is deeply incised in gravelly and sandy fan deltas prograded over Pliocene to Pleistocene lacustrine and marine marls (Gawthorpe et al., 1994).

Materials and methods

The geomorphological analysis of this work is based on a detailed digital surface model provided by the Greek Cadastre (Figure 3). The digital surface model was developed with photogrammetric techniques from digital aerial photographs acquired between the years 2007 and 2009. It covers the whole Greece and has a spatial resolution of 5 m and a nominal vertical accuracy better than 2 m. This specific digital surface model is the most accurate official dataset available in Greece. We processed this digital surface model with the use of ArcGIS software. The detailed geomorphology is supported by a drone flight campaign above the gorge of the Katourlas River south of the trace of the Helike Fault. A photogrammetric grid was designed, and six flights were performed with 90% overlap along the track and 75% across the track. The acquired images were imported in Agisoft’s Photoscan software. For the detailed procedures see Skarlatos et al. (2013) and Nikolakopoulos et al. (2017, 2018). Orthophoto and digital surface models were created with a spatial resolution of 4 and 8 cm respectively.

Figure 3.

Digital surface model of the Katourlas drainage basin and the Katourlas alluvial fan and two landslides near the exit of the drainage basin. Morphological analysis of the two landslides shown by the topographic profiles, in the inset, indicates the depletion zone of the landslides. Yellow dashed lines separate proximal, mid and distal fan. Blue lines show feeder channels on top of the fan, based on 1945 air photographs. The area of archaeological excavations and boreholes are in the Koutroumanis Field denoted as (Kou). Abbreviations EL, WL, KL, BAL and PF as in Figure 2. For the location of the figure see Figure 2.

The sedimentological analyses were based on four shallow boreholes drilled in the area of Koutroumanis Field (67, 68, 69, 70) combined with three archaeological excavations H67, H68 and H69 carried out during the drilling and excavation campaign of the Helike Project in 2012 (Kou in Figure 2). From the four boreholes, we present here a detailed log of borehole 70, cored near the archaeological trench H69. Sedimentological analysis included all four boreholes. The sediment cores were obtained using an Eijkelkamp percussion corer with barrel windows. Sediment type, structure, colour, organic constituents as well as contact depths and bed characteristics were recorded directly in the field. Laboratory studies included determination of grain size, total organic matter, carbonate and carbon content, and micro- and mollusc fossils of selected samples. The approximate percentage of shells and of terrigenous minerals in the sand fraction was estimated visually. Further sedimentological analysis was combined with sampling and descriptions of outcrops in the drainage basin of the Katourlas River.

In total, 99 boreholes were drilled by the Helike Project from 1991 to 2000 between the Helike Fault trace and the coast of the Gulf of Corinth. Of these, 16 were drilled on the Katourlas fan and six were close to our study area. Occupation horizons with pottery sherds and other artefacts were encountered in most of the boreholes (Soter and Katsonopoulou, 2011). Results from these boreholes led archaeologists to focus most of their efforts in the rural areas on the distal part of the Katourlas fan, including the Klonis, Balalas, Papafilippou and Koutroumanis fields (Figure 2). Trial trenches and excavations in these areas successfully brought to light several occupation periods of Ancient Helike spanning from Classical to Roman times (Katsonopoulou, 1998a, 2005).

The geology and geomorphology of the Katourlas River basin and its alluvial fan

Geomorphology of the Katourlas River basin

The Katourlas River basin (Figure 3) occupies ≈4.1 km² in the footwall block of the Helike Fault and is developed at the overstep zone between the two prominent segments of the fault (Figure 4a). It is the largest drainage basin between the Selinous and Kerynites drainage basins (Figure 1). We made geomorphological analyses of the basin in order to evaluate whether it were feasible for landslides, mudflows or river floods to have contributed to the destruction of Ancient Helike in 373 BC.

Figure 4.

(a) Overview of drainage basins on the footwall of western segment of the Helike Fault. See also the trace of the Melissia Fault and its proximity to the study area. (b) Vf index showing vertical incision distribution in the Katourlas drainage basin. (c) Slope orientation and dip in the Katourlas basin. (d) Elliptical shows rotation (asymmetry) of the Katourlas Fan. (e) SL morphometric index showing steepness of the Katourlas River.

The Selinous and Kerynites rivers (Figures 1b and 4) are characterised as antecedent rivers and contribute highly to the sedimentation of the Gulf of Corinth. The Katourlas River basin is characterised as a 4^th class basin trending almost N–S, whereas its tributaries and main course have a total length of more than 18 km with the main river reaching 3.5 km in length. The highest observed altitude of the watershed is about 1000 m with a mean elevation value of 537 m (Figure 4a). The slope relief is generally smooth, with less than 10% of the basin area dominated by slope angles greater than 45° (Figure 4c). Basin elongation (Bs) is 1.3, and drainage density (Dd) is 4.56. The valley floor width–valley height ratio (Vf) and stream length-gradient index (S_L) (Figure 4b, e) show an active riverbed where the valley incision dominates by forming steep gorges. During the geomorphologic analysis of the Katourlas basin, we considered the possibility of knickpoints. Lithological changes, fault tectonics and landslide damming of a river channel are commonly related to the formation of a knickpoint (Ahmed et al., 2019; Korup et al., 2010; Lague, 2014; Leopold and Wolman, 1960). At least two knickpoints are identified away from any lithological contact or faulting along the Katourlas River channel. The knickpoints are identified where the longitudinal profile forms a concave up feature (dashed line in Figure 5). This disruption in the generally gentle curve of the Katourlas River shows a slope break (Whipple and Tucker, 1999), differentiated from the ideal hypothetical longitudinal profile of the river (Figure 5). Overall, the basin is characterised by maturity according to the hypsometric integral estimates where the lower part of the basin that is close to the active fault is still in the youth stage.

Figure 5.

Long profile and cross sections in the lower part of the ephemeral Katourlas River. (a) Long profile of the river before and after the inferred dam relic recognised as slope-break knickpoint. The dam relic is denoted by a bulge. The numbers 1 to 12 denote the cross-sections used to identify the role of the relic dam on the river’s long profile. Cross sections location are shown in (b). (b) Hillshade illustration of the study area showing with black lines the outline of the west and east landslides. The white lines show transects that cross perpendicular the river bed spacing apart 50 m. (c) Representative cross sections across the river channel midline. Examples of the transect sections before, within and after the landslide site.

The Katourlas basin is developed in limestone and flysch in the south, and gravelly and sandy fan deltas prograded over Pliocene to Pleistocene lacustrine and marine marls in the north. Flat terraces and deep gorges are visible in the geomorphology of delta-top terraces of sand and gravel suggesting that the Katourlas basin is under fast continuous uplift, further evidenced by the high Vf index. Where flat terraces overlie steep slopes comprising soft non-cohesive marls, they are susceptible to landsliding.

The Katourlas fan has a radial length of ~1 km and covers an area of ~1.0 km² in the study area (Figures 2 and 3). The average radial slope of the fan surface is ~3.3° towards the N. In the proximal part this slope is as much as 4.5°, in the middle fan the slope is 3.5°and ~2° towards the distal part of the fan. The flow expansion angle is ~115° beyond the mountain front. The feeder channel has carved a 620 m long, narrow gorge a few metres deep that is still preserved, although the top of the proximal fan has been strongly modified by road building and residential development. Several other feeder channels are recognised on the fan surface (Figure 3).The intersection point lies near the boundary between the mid and the distal fan. The main distributaries channels radiate in N to ENE directions, based on the digital surface model. Fan evolution is asymmetric and rotated towards the SW (Figure 4d). Probably this rotation is a proxy on the surface of a reverse drag profile on the Helike Fault hanging wall block at depth (Grasemann et al., 2005).

Landslides in the Katourlas drainage basin and damming of the Katourlas River

Landslides are widespread on the valley walls of the Katourlas drainage basin, recognisable from the overall morphology. Two well-developed old landslide scars, mapped directly in the field and by the UAV flights, are located asymmetrically in the east and west valley sides of the Katourlas River, in an area where the main river carved its major incision south of the fault-bound mountain front (Figures 2 and 6). Erosion scars on valley sides probably represent slide scars, but in many places are obscured by thick vegetation. Farther upstream, there is also ample evidence for small landslides, and the village of Ano Keryneia appears to be built over an old landslide.

Figure 6.

Orthophoto map from UAV data (flight height 120 m). West (white dashed line) and east landslide (white solid line) in Katourlas area and the suggested 373 BC lake ponded behind the dam captured by the UAV are projected.

The mapped landslides are termed the east (EL) and west (WL) landslides (Figures 2 and 3). Both show elliptical crowns and elongated depletion zones. The length of the EL is 423 m and its width about 150 m. The length of the WL is 139 m and its width 86 m (Figures 3 and 6).The two landslides could have been induced separately or contemporaneously, causing a type IV natural dam as categorised by Costa and Schuster (1987).

Quantifying the volume of individual landslides is a difficult task because it requires information on the surface and sub-surface geometry of the slope failure (Guzzetti et al., 2009). For past landslides the task is more uncertain, because after the landslide, its crown and flanks enlarge due to erosion or reactivation (Figures 2 and 3). We applied two independent methods to estimate their volumes. The first method uses published scalar relationships (Table 1) extracted from worldwide data of landslide areas and their average depth (Cha et al., 2018; Larsen and Torres Sanchez, 1998; Martin et al., 2002), following the equation proposed by Larsen et al. (2010), which models the empirical relationship between V (volume) and A (area) of a landslide as an equation of the form:

V = α A^{γ}

Table 1.

Scalar relationship between volume and area values of a landslide.

Equation	Number of landslides	WL volumein m³	EL volumein m³	Source
V = 0.596A ^1.02	930	7825	44,493	Cha et al. (2018)
V = 0.074A ^1.45	677	52,919	626,045	Guzzetti et al. (2009)
V = 0.39A ^1.31	51	75,892	707,276	Imaizumi and Sidle (2007)
V = 0.084A ^1.43	539	49,878	570,300	Guzzetti et al. (2008)
V = 0.19A ^1.19	11	12,117	92,040	Imaizumi et al. (2008)
Mean landslide volume as calculated by the equations		39,726	408,031
Cut-and-fill volume calculation		87,022	501,902	This study

Column 1 presents various equations used for the calculation of the landslide volume (V) for a given landslide area (A). Column 2 shows the number of landslides used for the solution of each equation. Columns 3 and 4 show the volume of the west and east landslides in the Katourlas basin following column 1 equations, the mean calculated volume and the calculation based on the digital surface model subtraction. Column 5 gives the sources of each equation.

Out of a series of similar equations we selected five available in the literature (Table 1). Second, we use the digital surface model subtraction method to calculate the volume of the two landslides. This method includes the synthesis of a digital surface model based on the manual filling by inserting several linearly aligned points with the slide area from the flanks of the landslides and a digital surface model as it appears in the present-day morphology. The difference of two subtracted digital surface models represents an estimate of the landslide volume based on a cut-and-fill volume calculation. These estimates of the volume do not consider how many landslide events have occurred to induce such volumes.

Based on the calculated volume of the two landslides, we tried to restore the height and the width of a possible dam blocking the Katourlas River in the area of their toes. For this calculation, the important parameters are the accurate morphology of the valley and the estimate of the dam volume. For improving the accuracy of the dam volume, we executed UAV flights above the Katourlas River gorge above the toe of the two landslides. Based on the existence of two landslides in the area we discriminate three possible cases; (a) The EL dammed the river, (b) the WL dammed the river, or (c) the contemporaneous sliding of both landslides. The third assumption is considered as the worst-case scenario producing a higher dam and the biggest lake. For each of these scenarios we calculated the dam volume and the lake/water volume stored (Table 2). The volume of the dam is significant, as this volume could be dispersed as a mud flow on the Katourlas alluvial fan. All these calculations indicated two significant results: (a) The volume of the landslides is big enough to have dammed the river to a height of many tens of metres, especially in the assumption of the worst-case scenario. (b) The location of the landslides just above the entrance to the steep gorge of the river would cause the formation of a large lake, thus inducing risk on the alluvial fan.

Table 2.

Estimates of the size of a landslide dam and resulting lake in the Katourlas River basin based on the comparison of digital surface models.

Scenario cases	Dam volume (m³)	Lake/Water volume (m³)	Dam Height (m)
East landslide	501,902	401,985	60
West landslide	87,022	42,246	34
Both landslides	588,924	507,720	64

The main question posed for the evolution of the dam is the following: could enough water be concentrated overnight in order to fill the lake behind the Katourlas dam? We consider the overnight water concentration as important, because ancient accounts state that the earthquake happened in the night and in the next morning the ruined city was flooded. In hydrology two critical parameters can be quantified, the lagtime (L) and the time of concentration (T_c) according to U.S. Department of Agriculture (2010). T_c is defined as the time required for runoff to drain the basin from the hydraulically most faraway point in the watershed to the basin outlet. T_c depends upon slope shape of the watershed and the flow path. L, according to Granato (2010), is the time from the centre of mass (centroid) of rainfall excess to the centroid of the corresponding runoff hydrograph. Various researchers (Mockus, 1957; Simas, 1996) defined the following relation between L and T_c for average natural watershed conditions and an approximately uniform distribution of runoff with the formula:

L = 0 {. 6 T}_{c}

Taking into account the slope, the flow accumulation, the flow length and the lithology we calculated the T_c for the Katourlas basin in ArcMAP using the Hydrology toolbox. We repeated the whole procedure using two different DSMs (ALOS AW3D30 and cadastral DSM). For both cases of DSMs T_c is only 0.1 h suggesting that immediately after dam formation, the lake started accumulating water.

We analysed how long it would take to accumulate the maximum water volume of 0.507 × 10⁶ m³ (Table 2) that could be ponded behind the dam. In the case of extreme rainfall, we adopted the Koutsoyiannis et al. (2012) determination that for a 50 year recurrence interval, extreme rainfall for the broader area is in the order of 110 mm per day. However, an even larger amount of extreme rainfall, in the order of 150 mm per day, is not uncommon locally (Koutsoyiannis, 2004; Koutsoyiannis and Mamassis, 2008). In summary, based on almost instantaneous run-off in the Katourlas basin and assuming 150 mm/day rainfall, the water could fill the worst case scenario lake in 20 h.

In order to recognise the role of the possible dam in the Katourlas River gorge, we analysed the incision across Katourlas channel from relief data in the surrounding landscape in a series of 12 profiles spaced 50 m apart downstream along the channel path. Each profile is centred on the channel centerline. In Figure 5, profiles are systematically V-shaped except in the area where the river crosses the landslide toes. This profile geometry shows that the river upstream and downstream of the landslide toe was not affected by the damming, attesting to an ephemeral dam. The ephemeral character is expected in dams of water-rich porous sediments, as is the lithology of the study area. Such dams fail hours or days after their formation (Clague and Evans, 1994).

In summary, a slope-break knickpoint (Figure 5a, b) represents a possible dam relic in the Katourlas riverbed, while the valley morphology and the lithology of the landslides attest to an ephemeral dam.

The sedimentology of the Katourlas Fan

In order to overcome dating problems in an area of high erosion and trying to outline the role of flooding events on the fan surface, we decided to drill boreholes in the area where the archaeologists had found ruins of buildings and associated Classical pottery. Borehole 70 was cored at an elevation of 9 m and distance of 50 m south of the archaeological trench H69, excavated at an elevation of 7 m, where a destruction layer including stones and roof tiles from demolished buildings and Classical pottery fragments came to light. The borehole consists principally of mud with small amounts of dispersed sand and gravel (Figure 7).

Figure 7.

Stratigraphic log and sedimentary analysis of the 70 borehole. Red stars in the log shows depth where pottery shards were recognised.

At 2.8 m depth, a 20 cm sorted gravel bed passes upwards into gravelly mud. This gravel bed overlies a second slightly gravelly mud unit. The two mud units separated by the gravel bed show rather different textural and compositional character. The upper bed has a finer mean and median grain size, slightly higher total organic carbon (TOC) and slightly lower carbonate content compared to the lower unit. Ceramic fragments were found within the lower slightly gravelly mud unit and at the top of the overlying sorted gravel unit. Kontopoulos et al. (2017) compared the textural characteristics of the slightly gravelly mud units with mudflow deposits using quartile-deviation (QDa) versus median-diameter (Md) (Buller and MacManus, 1972; Pe and Piper, 1975) (Figure 8a) and the C-M (1 percentile vs. median diameter) plot (Bull, 1962; Passega, 1957) (Figure 8b). They concluded that the two mud units were mud-flow deposits with sparse dispersed coarse sand and gravel, with the intervening sorted gravel being a stream-flow deposit. A similar stratigraphy of slightly gravelly mud units showing similar variation in grain size, TOC and carbonate content was found in three nearby boreholes (Kontopoulos et al., 2017). Thus, at least two mudflow deposits appear associated with Classical archaeological remains.

Figure 8.

Grain size plots of samples of mudflow deposits from boreholes. (a) Quartile-deviation (QD) versus median-diameter (Md) plot after Buller and MacManus (1972), A = glacial deposits, B = fluvial deposits. Solid line: lower right limit of mudflow field from Pe and Piper (1975). (b) C (1 percentile) versus M (median diameter) plot after Passega (1957).

A low terrace and bank beside the Katourlas River on the apex of the Katourlas fan exposes ~1 m of structureless mud with sparse dispersed pebbles (Figure 9a) overlying pebble conglomerate (Figure 9b). A thin bed of fine pebble gravel also occurs on the terrace immediately overlying the mud (Figure 9a). Although no datable material was found, two analysed samples from the mud with dispersed pebbles show textural character of slightly gravelly mud and sandy mud, similar to the upper mud-flow deposit in boreholes on the Katourlas fan. TOC and carbonate content are also similar. The outcrop lithofacies is convincingly a mud-flow deposit, thus confirming the textural interpretation of mudflow processes in the boreholes (Figure 8).

Figure 9.

Outcrop photos of the mudflow deposit at the apex of Katourlas fan. (a) Mudflow deposit with dispersed pebbles, overlain by a surface fluvial gravel. (b) Mudflow deposit overlying fluvial pebble conglomerate.

The toe of the Katourlas fan has a higher proportion of gravel and muddy gravel, which is interbedded with gravelly mud, slightly gravelly mud and sandy mud characteristic of the mudflow deposits. The mudflow deposits pass laterally into the braid delta plain of the Selinous River (Kontopoulos et al., 2017).

Archaeological evidence for the 373 BC event and related mudflows and landslides

The upper mudflow deposit recognised in borehole 70 directly overlies architectural remains of Classical walls and/or destruction layers, as revealed in excavated trenches in the Papafilippou Field (Figure 10; located in Figures 2 and 11). In this location, the excavation of four trenches (H9, H11, H29 and H35) brought to light at about 2 m depth from the surface, an extensive ancient destruction layer consisting of successive layers of cobble stones, roof tiles, abundant Classical pottery shards from storage vessels and fine vases, black-glazed and decorated, dated to the 2^nd half of the 4^th c. BC. One of the trenches (H29) revealed the remains of a late Classical wall, lying at a deeper layer about 20–30 cm under the upper destruction layer excavated in the trench. Besides pottery, associated finds from the excavation include clay loom weights, two bronze Sikyon coins of the last quarter of the 4^th c. BC, and an exceptional fragment from the base of a clay perirranterion. Based on the above, this destruction is dated in the second half of the 4^th c. BC, and is therefore a few decades later than the 373 BC catastrophe.

Figure 10.

Gravelly mud deposits on top of (a) late 4th c. BC architectural ruins and (b) of a late 4th c. BC destruction layer (roof tiles and pottery). From the Papafilippou Field archaeological excavations. White and black scale is 1 m.

Figure 11.

Synthetic map of the suggested lake and dam in the Katourlas drainage basin (white line) and the location of the Classical ruins and associated finds in relation with the Katourlas fan (red line). For abbreviations see Figure 2.

However, archaeological evidence from the Koutroumanis Field, 825 m southeast of Papafilippou, provides a date closer to 373 BC, with an uncertainty of 1020 years. In this field, excavation of trench H69 (Figures 2, 3 and 11) revealed, at a depth of 1.21.5 m from the surface, an ancient destruction layer, 0.3 m thick, consisting of scattered building stones and roof tiles from destroyed buildings, sun-dried bricks, Classical pottery shards and fragments of clay lamps, dated around mid-4^th c. BC. Other finds include sea shells, animal bones, and iron nails from constructions.

Further systematic excavations by the Helike Project near the southeastern toe of the Katourlas fan at the Balalas Field, located between the Papafilippou and Koutroumanis fields (BAL in Figures 2, 3 and 11), have brought to light destruction evidence possibly associated with the 373 BC event. Two adjacent excavated trenches (H18 and H19) revealed, at 3.0–3.3 m below the surface, the remains of ruined walls of Classical buildings most probably destroyed by an earthquake. The excavated remains are dated on the basis of associated pottery and other finds, including one silver Sicyon coin, one bronze Aiginetan coin, and the painted female head of a clay idol, just before and around the time of the 373 BC earthquake (Katsonopoulou, 2002).

Thus, the archaeological and borehole evidence from the area suggest the existence of at least two mudflow deposits overlying 4^th c. BC occupation levels. The older mudflow deposit at the Koutroumanis Field could have immediately followed the Classical destruction of Ancient Helike in 373 BC. The younger deposit, correlating with the upper mudflow deposit in borehole 70 and in outcrop at the apex of the fan, appears to be 3 or 4 decades younger, after some re-occupation of the Helike site. In addition, recent evidence from the Helike Project excavation work in the coastal plain, has shown that Ancient Helike was not completely abandoned after the 373 BC earthquake, as believed in the past, but, on the contrary, a new settlement ranging in date from about 330 BC to the last decades of the 2^nd c. BC, was developed in a location north-northwest of the Katourlas River. This new evidence agrees with the presence of Late Classical and Early Hellenistic occupation on the Helike acropolis above the village of Rizomylos (Figures 2 and 11).

In the area of terraces at the apex of the alluvial fan, Early Helladic II-III settlements were brought to light during the construction of the New National Road (https://www.archaeology.wiki/blog/2013/12/16/archaeological-research-in-eastern-achaea/). They are buried under almost 1 m of gravelly mud deposits, suggesting that a significant mudflow occurred sometime between the Early Helladic (ca. 2500 BC) and the Classical period, and most probably closer to the end of the Classical period. This mudflow deposit might correspond to either the lower or the upper deposit in borehole 70.

On the slope of one of the terraces, a bronze Helike coin, depicting Poseidon, together with Classical pottery shards were recently found among transported probable mudflow deposits on top of the Early Helladic remains (newspaper www.protionline.gr, 14^th December 2016). The coin is of the same type as the well-known coins in the Berlin Museum (found in the area in 1861) and two recently discovered Helike coins (Katsonopoulou, 2017). The coin suggesting Classical occupation in this area, provides an additional age bracket for the 4^th c. BC mudflows.

The hill above New Keryneia to the east of the Katourlas River, has long been identified with the acropolis of Helike, being occupied as an acropolis and a settlement area continuously from the Archaic to Roman times (Katsonopoulou, 1998b; Soter and Katsonopoulou, 1998). The crest of the east landslide is located near the acropolis. Thus, the significance of the landslide(s) was two-fold: the landslide directly affected the area of the Helike acropolis and the dam collapse affected the lowland area of Ancient Helike.

Geohazard processes and the 373 BC destruction of Ancient Helike

Landslides and dam bursts

River dams might be thought of as unexpected in the north Peloponnese, due to the extensive outcrop of Mesozoic limestones and Neogene and Quaternary coarse grained clastics in an area of limited rainfall. However, recent publications and historic newspaper accounts provide a different view. In the early 20^th century, on 24^th March 1913, an extensive landslide took place in the catchment area of Krathis River in the western Gulf of Corinth, 20 km east of Helike (Figure 1a). This landslide was the result of a combination of intense rainfall and a moderate earthquake. The mass wasting affected the footwall block of an EΝΕ-striking listric normal fault. The landslide led to the obstruction of the river basin, resulting in the creation of two lakes, Krathis and Tsivlou lakes. On 5^th January 1914, the dam of Lake Krathis collapsed, whereas Lake Tsivlou has persisted until today. Lake Krathis had a length of about 3 km and the dam height was about 80 m (Zygouri and Koukouvelas, 2019). The outburst of this lake caused the flooding of the Krathis delta. According to newspaper reports, the 1914 flood travelled over 13 km downstream and buried graveyards and olive tree trunks up to the first node, suggesting that the thickness of the flood deposits on the delta was at least 1m (Zygouri and Koukouvelas, 2019).

In the footwall block of the Helike Fault, two other cases of natural dams were reported in newspapers, in 1963 and 1965, along the Selinous River near the village of Ano Mazaraki (Figure 1b).There the river was blocked by a landslide during intense rainfall on 4^th April 1963. The landslide created a dam about 100 m high and formed a lake of about 800 m long. The surface area of the lake was about 0.06 km² and its depth was in the order of 50 m. Due to leakage of water through the dam, this lake soon drained (newspaper Neologos Patron 6^th April 1963). The second dam formed on 24^th January 1965 (newspaper Elefteria 25^th January 1965). In both cases, these dams were considered to be of the order of 100 m high. According to the newspapers of the period, their overtopping did not result in significant flooding, unlike the case of the Krathis River. In addition, in 1928, during a period of heavy rainfall, press articles described serious flooding of the low-land area on the hanging wall block of the Helike Fault for a period of about 10 days (newspapers Imerisios Typos, Empros and Ethnos 28^th November 1928). In 2013, after 5 days of rain, the low land area between the Kerynites and Selinous river mouths was again flooded (Figure 2).

These data highlight that flooding during heavy rainfall and the formation of natural dams in the study area pose potential risks to human life and property. However, up to now these risks were underestimated in the analysis of the engulfment of Ancient Helike. The case of the Krathis River dam indicates that the combination of seismicity and rainfall is particularly dangerous. Notably in the case of Ancient Helike, the 373 BC earthquake happened during winter (Strabo 8.7.2; Pausanias 7.24.11-12). Ambraseys (2009) reports that during this earthquake ‘slides triggered by the earthquake, dammed a river’ with no more details or identification of the source of this information. From these data, two major conclusions could be made. First, river dams, although not common, block rivers draining the north Peloponnese. Second, the outburst of these dams can be catastrophic. The estimated height of the dam and the possible lake in the Katourlas River resembles quite well the geometry of other cases of river damming in the area. We suggest that the name of Gardena village itself is an indication of landsliding and river damming at some time in the past.

The archaeological data presented above suggest that there were two or more different mudflow events in the 4^th c. BC. While we lack precise information on the lateral extent of a single mudflow, they appear to have deposited over a zone 1 km wide on the distal fan, 1 km from the apex (Figure 11). Taking a mean thickness of 2 m suggested by borehole data and archaeological excavation gives a mudflow volume on the order of 10⁶ m³. Geologically, it seems likely that mudflow events are not uncommon in the Katourlas River.

Such mudflows elsewhere have been generated directly by breakup of landslides triggered by heavy rainfall (Jakob et al., 2005; Kazama et al., 2012). Seismic triggering of landslides and ephemeral damming of rivers creates particularly favourable conditions for triggering debris- and mud-flows (Adams, 1981). For example 1999 Chi-Chi earthquake (Lin et al. 2014) and 2008 Wenchuan earthquake (Chang et al. 2017) initiated mudflows in relative small drainage areas similar to the Katourlas drainage basin, and buried adjacent cities. Mudflow volume may be an order of magnitude greater than the volume of water released by a natural dam burst (Tannant and Skermer, 2013). In a drainage basin like the Katourlas basin, the combination of heavy rainfall and an earthquake could create mudflows both by break up of landslides on steep water saturated slopes and by bursting of landslide dams.

The cataclysm associated with the 373 BC earthquake was probably a flooding or mudflow event for the following reasons: (1) landsliding, damming of rivers and mudflows in the broad area are not uncommon; (2) the Katourlas gorge is prone to landsliding and damming, with unequivocal old mudflow deposits outcropping at the outflow of the gorge; and (3) gravelly mud deposits, interpreted as mudflow deposits interbedded with stream flow deposits, are common at the distal end of the Katourlas fan during the 4^th c. BC.

The tsunami hypothesis for the 373 BC destruction of Ancient Helike

Tsunami hazard has often been reported for the broader Gulf of Corinth. Historic testimonies and catalogues published by several authors attest that the tsunami hazard for the Aigion area has been reported since antiquity (Aristotle 4^th c. BC; Pausanias 2^nd c. BC; Diodoros 1^st c. BC in Kortekaas et al., 2011; Galanopoulos, 1960; Guidoboni et al., 1994; Papadopoulos, 2003). From the 4^th c. BC till the present-day, the coastal area of the Gulf of Corinth has been affected by earthquake-triggered tsunamis. Recent studies using boreholes and trenching on both the south and north coast of the Gulf of Corinth have tried to verify and validate historical testimonies with geomorphological and sedimentological features. The first attempt was about twenty years ago from a common team of Greek and Japanese tsunami researchers supported by the Helike Project, who excavated trial trenches in the lowland area between the Gulf of Corinth shoreline and the modern village of Eliki to locate the 373 BC tsunami deposits (G. Papadopoulos, personal communication 2019). They excavated three trenches, but their attempt was unsuccessful. Later, Kontopoulos and Avramidis (2003) obtained three sediment cores in Aliki lagoon (5.5 km NW of Katourlas fan, Figure 1b) that record six tsunami events since 2500 BC. The sediment deposits are characterised by abrupt change in environmental conditions and show textural and structural characteristics common for a tsunami deposit (Fujiwara et al., 2000; Takashimizu and Masuda, 2000), such as abrupt erosive contacts with overlying/underlying deposits, a coarse grain size, and marine shell fragments and microfossils. One of these recognised tsunami deposits was a sandy bed with an erosive base that contained transported mollusc shells. It yielded a calibrated 2σ age range of 380–500 BC and thus might be correlated with the 373 BC earthquake (Kontopoulos and Avramidis, 2003). However, Kortekaas et al. (2011) found no evidence for a correlatable horizon in another borehole in the same Aliki lagoon, although foraminifera analysis showed an increase in marine foraminifera at 6–7 cm depth, suggesting a marine tsunami inundation during the 23^th August 1817 earthquake. In addition, trenches and cores located on the northern coast at the head of Itea Bay, almost 30 km NE of our study area, revealed at least four tsunami deposits, but all with dates incompatible with the 373 BC event (Kortekaas et al., 2011). Contrary to Kortekaas et al. (2011), Alvarez Zarikian et al. (2008) reported a widespread sandy mud layer at 2.6–3.9 m depth in boreholes lying less than 1 km north from our study area. In this sandy mud layer, they recognised abraded shells and microfossils from various environments and correlated the layer to the 373 BC tsunami. Engel et al. (2016) were unable to recognise this sandy layer in boreholes closer to the shoreline (≈1.0 km north of Koutroumanis Field), in an area close to the Alvarez Zarikian et al. (2008) boreholes. Thus, Engel et al., (2016) concluded that the tsunami deposition is uncertain. The stratigraphic depth of the sandy mud layer of Alvarez Zarikian et al. (2008) makes it possible that it represents the upper mudflow deposit recognised in borehole 70 of the present study and the outcrop at the apex of Katourlas fan, with the mixed abraded microfossils reworked out of Pliocene-Pleistocene strata on the footwall. In summary, intensive research in the area of Helike on the tsunami deposits was so far unsuccessful.

Discussion: Black swan or cascade disaster

The disaster of Ancient Helike has been traditionally identified as a CDM, a coastal disaster model. Most authors have adopted the CDM and tried to explain the fascinating engulfment of the city. Many authors have drawn parallels between the AD 1861 and the 373 BC earthquakes at Helike. The large 1861 earthquake was hosted on the Helike Fault, but no evidence exists for slip on the Helike Fault in 373 BC. Furthermore, the 373 BC earthquake was considered as a ‘black swan’ and ambiguous evidence led to widespread assumption of a tsunami wave, based exclusively on historical narrations and the first organised historical catalogue by Papazachos and Papazachou (1989). In this catalogue, the 373 BC earthquake magnitude was considered as 7.0. Later the magnitude was revised as <6.8 (Papadopoulos, 2000; Papazachos and Papazachou, 1997). However, Ambraseys (2009), in his catalogue considered the 373 BC earthquake was either small or at least typical for the area, since the nearby Aigion city remained almost unaffected. Accordingly, the small earthquake magnitude is unlikely to have caused a huge tsunami, unless it triggered a large submarine landslide, and this is compatible with the difficulty of finding tsunami deposits in the area. Galanopoulos (1960) did not include a 373 BC tsunami in his tsunami catalogue, perhaps because the tsunamis of the 1817 and 1861 earthquakes were less than 2 m high. Thus the 373 BC was an earthquake probably correlated with the smaller Keryneia, Melissia or Mamoussia faults. A southern fault explains better the Katourlas landsliding and the lack of evidence for a tsunami in the Koutroumanis Field. Borehole 70 and the archaeological trench H69 penetrated or brought to light Classical age deposits or destruction layers, respectively, at a distance of about 1.5 km from the shoreline. If there had been a tsunami, then the location and elevation of boreholes and archaeological trenches were well located to penetrate or excavate a possible tsunami deposit. The hypothesis of reactivation of an offshore fault (Koukouvelas, 2008) has not been supported by more recent work on evidence for a tsunami in 373 BC (Kortekaas et al., 2011; Engel et al., 2016 and the present study). The long descriptions in the literary sources, and especially the submergence of the Poseidon statue, led to the hypothesis of submergence due to a retrogressive rotational submarine landslide (Ambraseys, 2009). The same observations can be explained by the formation of a lake near the coastal zone due to liquefaction, as is the case of the lake formed at Cape Louros in Kos Island during the 2017 earthquake (Triantafyllou et al., 2020). Alternatively, the statue may have been partly buried by a mudflow deposit and the historical accounts have fanciful embellishment.

The present contribution tries to interpret in an alternative manner the 373 BC earthquake and its secondary effects. The new concept in our interpretation is that a mudflow was the critical phenomenon for the final destruction of Ancient Helike, based on geological evidence in the mountains behind Helike (Figure 11). This alternative model demonstrates that the Katourlas drainage basin is prone to damming. Damming of streams in the north Peloponnese is not uncommon, particularly during earthquakes but was previously overlooked. Shallow boreholes, archaeological evidence, and outcrop descriptions in the study area also demonstrate extensive mudflows archaeologically dated to the Classical period of Helike. The repeated presence of mudflows on the distal part of the Katourlas alluvial fan indicates that possibly the damming of the Katourlas River in the territory around Gardena in the 4^th c. BC was related to the combination of seismic shaking and heavy rain.

In the case of the destruction of Ancient Helike, we propose that the catastrophic event was not a ‘black swan’. A black swan in a risk context is considered an unexpected extreme event relative to one’s knowledge, classified in the following different types based on Aven’s (2015) proposal: (a) unknown unknowns, (b) unknown knowns (we do not have the knowledge but others do), and (c) events that are judged to have a non-computable probability of occurrence and thus are not believed to occur. A black swan in terms of the engulfment of Ancient Helike must have the following three attributes. Firstly, it is an outlier, beyond the realm of regular expectations, in terms of earthquakes and/or tsunamis happening in the Gulf of Corinth. Secondly, it carries an extreme impact. Thirdly, after the fact we can produce explanations for its occurrence or impact making it explainable and predictable.

In contrast, our analysis suggests that none of these three attributes is present. The alternative proposal of an IDDM, the inland-driven disaster model, in addition to the earthquake, considers the most common disaster on alluvial fans, the mudflows, as a part of a cascade disaster. The IDDM describes quite well the dependency and the localisation of a non-linear disaster in a path of events, in which the landslides acted as generators of vulnerabilities (for paths of events in cascade disasters, see Pescaroli and Alexander, 2015, 2016). The IDDM includes the following hypotheses: that seismic shaking during the 373 BC earthquake triggered landslides and mudflows that created a dam in the Katourlas River and probably in the Selinous River. The Katourlas dam collapsed a few hours after the earthquake, leaving relics in the riverbed, and added much disaster to the effects of the earthquake and any possible minor tsunami inundation of Ancient Helike. The mudflow also explains what Herakleides (in Strabo 8.7.2) narrated that ‘two thousand men who were sent by the Achaeans were unable to recover the dead bodies’. The dam in the Selinous River, as commonly happens, gave rise to the wave reported by Aristotle that invaded from the west towards the south.

Notably in this scenario, any tsunami is rather weakly related with the main disaster. The CDM could possibly be considered as a cascade disaster, even if the precise relationship between the earthquake and the tsunami is unknown. Was it submarine fault slip or the triggering of a submarine landslide that triggered any tsunami? The absence of tsunami deposits in any geological archive might be explained because deposits were eroded by the later mudflow, or the runup of the tsunami in the literary sources is exaggerated (see also discussion in Ferentinos et al., 2015; Walter, 2017). Moreover, geomorphological observations from modern tsunami deposits show extensive deposition of large boulders forming significant thickness deposits (Dawson, 1994). Following this notion Dominey-Howes et al. (1998) justified the non-existence of a tsunami in Falassarna after 365 AD earthquake. If the 373 BC earthquake was on a fault south of the Helike Fault, with no submarine component, then there was probably no tsunami. In summary, in all our data the Helike disaster is a localised non-linear cascade disaster with the landslides acting as generators of vulnerabilities soon after the earthquake.

Our scenario is that the wave that ‘washed away and drowned all the inhabitants’ was not a tsunami, but rather the frontal wave of a catastrophic mud flow formed by bursting of a natural dam formed during the 373 BC Helike earthquake.

Conclusions

Engulfment of Ancient Helike resulted from an inland-driven disaster model (IDDM), which caused a non-linear localised cascade disaster. The seismic shaking of the 373 BC earthquake caused significant damage to Ancient Helike, and also caused widespread landslides in the mountains behind Helike that dammed the Katourlas and Selinous rivers. Particularly, the Katourlas dam burst produced a massive wave of sediment and water, a muddy debris flow, which inundated the city already damaged by the earthquake. The western territory of Ancient Helike was flooded by the Selinous River. Summarising our results, the explanation of the death toll and the engulfment of the Ancient Helike in 373 BC includes the following steps in temporal order.

(1) An earthquake, not larger than seismologists thought could happen, that is, on the order of M < 6.6. This event was hosted on an onshore fault, probably south of the ancient city. As a most likely candidate fault we consider the Melissia Fault or alternatively the Mamoussia Fault.

(2) Earthquake triggered landslides dammed the Katourlas and Selinous rivers. The Katourlas dam, perhaps with a volume as great as 0.6×10⁶ m³ and a similar lake water volume, drained suddenly hours later, giving rise to a catastrophic mudflow on part of the Katourlas alluvial fan. Our archaeological data suggest that at least two such mudflow events occurred during the 4^th c. BC in the study area, about 20–30 years apart.

(3) If there were a tsunami, it was small, resulting from a submarine landslide, and inundated a 200 m wide coastal parallel zone (Engel et al., 2016). There is no geological evidence to suggest that there was an ‘enormous tsunami’ during the 373 BC earthquake.

(4) Although river dams have not been given much consideration in the Helike disaster, our results indicate that these are rather common in the mountains of this part of the northern Peloponnese. The historical record of flooding from the west during the 373 BC may also record a watery flood in the Selinous River.

Footnotes

Acknowledgements

The authors would like to express their gratitude to three anonymous reviewers for their constructive and helpful comments that substantially improved the manuscript. Also we appreciate the excellent editorial handling of the paper by Dr. Alastair Dawson.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Ioannis K Koukouvelas

References

Adams

(1981) Earthquake-dammed lakes in New Zealand. Geology 9: 215–219.

Ahmed

Ali

Rogers

, et al. (2019) A study of knickpoint surveys and their likely association with landslides along the Hunza River longitudinal profile. Environmental Earth Sciences 78: 176.

Alvarez Zarikian

Soter

Katsonopoulou

(2008) Recurrent submergence and uplift in the area of Ancient Helike, Gulf of Corinth, Greece: Microfaunal and archaeological evidence. Journal of Coastal Research 24: 110–125.

Ambraseys

(2009) Earthquakes in the Mediterranean and Middle East: A Multidisciplinary Study of Seismicity up to 1900. Cambridge: Cambridge University Press, p. 968.

Aven

(2015) Implications of black swans to the foundations and practice of risk assessment and management. Reliability Engineering and System Safety 134: 83–91.

Bull

(1962) Relation of textural (CM) patterns to depositional environments of alluvial fan deposits. Journal of Sedimentary Petrology 32: 211–216.

Buller

McManus

(1972) Simple metric sedimentary statistics used to recognise different environments. Sedimentology 18: 1–21.

Caputo

Pavlides

(2013) The Greek database of Seismogenoc Sources (GreDaSS), version 2.0.0: A compilation of potential seismogenic sources (Mw > 5.5) in the Aegean Region. DOI: 10.15160/unife/gredass/0200.

Cha

Hwang

Choi

(2018) Landslides detection and volume estimation in Jinbu area of Korea. Forest Science and Technology 14: 61–65.

10.

Chang

Tang

Van Asch

TWJ

(2017) Hazard assessment of debris flows in the Wenchuan earthquake – stricken area, South West China. Landslides 14: 1783–1792.

11.

Clague

Evans

(1994) Formation and failure of natural dams in the Canadian Cordillera. Geological Survey of Canada Bulletin 464: 1–35.

12.

Console

Carluccio

Papadimitriou

, et al. (2015) Synthetic earthquake catalogs simulating seismic activity in the Corinth Gulf, Greece, fault system. Journal of Geophysical Research 120: 326–343.

13.

Costa

Schuster

(1987) The formation and failure of natural dams. USGS Open-File report, Virginia, pp. 87–392.

14.

Dawson

(1994) Geomorphological effects of tsunami run-up and backwash. Geomorphology 10: 83–94.

15.

Dominey-Howes

Dawson

Smith

(1998) Late-Holocene coastal tectonics at Falasarna, western Crete: A sedimentary study. In: Stewart

Vita-Finzi

(eds) Coastal Tectonics. Geological Society, London Special Publications 146: 343–352.

16.

Doutsos

Poulimenos

(1992) Geometry and kinematics of active faultsand their seismotectonic significance in the western Corinth–Patras rift (Greece). Journal of Structural Geology 14: 689–699.

17.

Durant

(1966) The Life of Greece. New York: Simon and Schuster editions.

18.

Engel

Jacobson

Boldt

, et al. (2016) New sediment cores reveal environmental changes driven by tectonic processes at ancient Helike, Greece. Geoarchaeology 31: 140–155.

19.

Ferentinos

Papatheodorou

Geraga

, et al. (2015) The disappearance of Helike-Classical Greece – New Remote Sensing and Geological Evidence. Remote Sensing 7: 1263–1278.

20.

Fujiwara

Masuda

Sakai

, et al. (2000) Tsunami deposits in Holocene bay mud in southern Kanto region, Pacific coast of central Japan. Sedimentary Geology 135: 219–230.

21.

Galanopoulos

(1960) Tsunamis observed on the coasts of Greece from antiquity to present time. Annales Geophysicae 13: 369–386.

22.

Gawthorpe

Fraser

Collier

REL

(1994) Sequence stratigraphy in active extensional basins implications for the interpretation of ancient basin-fills. Marine and Petroleum Geology 11: 642–658.

23.

Gawthorpe

Andrews

Collier

, et al. (2017) Building up or out? Disparate sequence architectures along an active rift margin—Corinth rift, Greece. Geology 45: 1111–1114.

24.

Granato

(2010) Methods for development of planning-level estimates of stormflow at unmonitored sites in the conterminous United States. Washington, DC: U.S. Department of Transportation, Federal Highway Administration, FHWA–HEP–09–005, 90 p. with CD-ROM.

25.

Grasemann

Martel

Passchier

(2005) Reverse and normal drag along a fault. Journal of Structural Geology 27: 999–1010.

26.

Guidoboni

Comastri

Traina

(1994) Catalogue of the Ancient Earthquakes in the Mediterranean Area up to the 10^th Century. ING, Roma – SGA- Bologna, p. 504.

27.

Guzzetti

Ardizzone

Cardinali

, et al. (2008) Distribution of landslides in the Upper Tiber River basin, central Italy. Geomorphology 96: 105–122.

28.

Guzzetti

Ardizzone

Cardinali

, et al. (2009) Landslide volumes and landslide mobilization rates in Umbria, central Italy. Earth Planet Science Letters 279: 222–229.

29.

Imaizumi

Sidle

(2007) Linkage of sediment supply and transport processes in Miyagawa Dam catchment, Japan. Journal of Geophysical Research 112: (F03012). DOI: 10.1029/2006JF000495.

30.

Imaizumi

Sidle

Kamei

(2008) Effects of forest harvesting on the occurrence of landslides and debris flows in steep terrain of central Japan. Earth Surface Process and Landforms 33: 827–840.

31.

Jakob

Hungr

Jakob

(2005) Debris-Flow Hazards and Related Phenomena (Vol. 739). Berlin: Springer.

32.

Katsonopoulou

(1998a) The first excavation at Helike: Klonis Field. A preliminary report (in Greek, with extended English abstract). In: Katsonopoulou

Soter

Schilardi

(eds) Ancient Helike and Aigialeia, Helike II. Athens: The Helike Society, pp. 125–145.

33.

Katsonopoulou

(1998b) Observations on the Topography of Aigialeia (in Greek, with extended English abstract). In: Katsonopoulou

Soter

Schilardi

(eds) Ancient Helike and Aigialeia, Helike II. Athens: The Helike Society, pp. 31–66.

34.

Katsonopoulou

(2002) Helike and her territory in the light of new discoveries. In: E.

Greco

(ed.) Gli Achei e l’identita etnica degli Achei d’ occidente, vol. 3. Paestum: Tekmeria, pp. 205–216.

35.

Katsonopoulou

(2005) The earthquake of 373 BC: Literary and Archaeological Evidence. In: Katsonopoulou

Soter

Koukouvelas

(eds) Helike III Ancient Helike and Aigialeia, archaeological Sites in Geologically Active Regions. Athens: The Helike Society, pp. 15–32.

36.

Katsonopoulou

(2017) Further insights on the Cult of Poseidon Helikonios in Helike. In: Katsonopoulou

(ed.) HELIKE V Ancient Helike and Aigialeia Poseidon, God of Earthquakes and Waters, Cult and Sanctuaries. Athens: The Helike Society, pp. 17–34.

37.

Kazama

Kataoka

Uzuoka

(2012) Volcanic mountain area disaster caused by the Iwate–Miyagi Nairiku Earthquake of 2008, Japan. Soils and Foundations 52: 168–184.

38.

Kontopoulos

Avramidis

(2003) A late-Holocene record of environmental changes from the Aliki lagoon, Aigion, North Peloponnesus, Greece. Quaternary International 111: 75–90.

39.

Kontopoulos

Katsonopoulou

Koutsios

(2017) Another approach to the destruction of 373 BC. In: Katsonopoulou

(ed.) HELIKE V Ancient Helike and Aigialeia Poseidon, God of Earthquakes and Waters, Cult and Sanctuaries. Athens: The Helike Society, pp. 203–214.

40.

Kortekaas

Papadopoulos

Ganas

, et al. (2011) Geological identification of historical tsunamis in the Gulf of Corinth, Central Greece. Natural Hazards and Earth System Sciences 11: 2029–2041.

41.

Koukouvelas

(1998) The Egion fault, earthquake-related and long-term deformation, Gulf of Corinth, Greece. Journal of Geodynamics 26: 501–513.

42.

Koukouvelas

(2008) Active normal faults; Geological methods of their analysis and applications. In: 3rd Hellenic Conference of Earthquake Engineering and Engineering Seismology, Athens, 5–7 November, article no. 1802 (in Greek with English abstract). Athens: Hellenic Society of Earthquake Engineering and Technical Chamber of Greece.

43.

Koukouvelas

Doutsos

(1996) Implications of structural segmentation during earthquakes: The 1995 Egion earthquake, Gulf of Corinth, Greece. Journal of Structural Geology 18: 1381–1388.

44.

Koukouvelas

Papoulis

(2009) Fluid involvement in the active Helike normal Fault, Gulf of Corinth, Greece. Journal of Structural Geology 31: 237–250.

45.

Koukouvelas

Stamatopoulos

Katsonopoulou

, et al. (2001) A palaeoseismological and geoarchaeological investigation of the Eliki fault, Gulf of Corinth, Greece. Journal of Structural Geology 23: 531–543.

46.

Koukouvelas

Katsonopoulou

Soter

, et al. (2005) Slip rates on the Helike Fault, Gulf of Corinth, Greece: New evidence from geoarchaeology. Terra Nova 17: 158–164.

47.

Koutsoyiannis

(2004) Statistics of extremes and estimation of extreme rainfall: II. Empirical investigation of long rainfall records. Hydrological Science Journal 49: 591–610.

48.

Koutsoyiannis

Mamassis

(2008) Recent trends in the estimation of extreme rainfalls. In: Proceedings of 1st Hellenic Conference of Large Dams, Larisa, 13-15 November 2018, pp. 443-440 (in Greek). Athens: Technical Chamber of Greece.

49.

Koutsoyiannis

Mamassis

Efstratiadis

, et al. (2012) Floods in Greece. In: Kundezewicz

(ed.) Changes in Flood Risk in Europe. Wallingford: IAHS Press, pp.238–256.

50.

Korup

Densmore

Schlunegger

(2010) The role of landslides in mountain range evolution. Geomorphology 120: 77–90.

51.

Lague

(2014) The stream power river incision model: evidence, theory and beyond. Earth Surface Processes and Landforms 39: 38–61.

52.

Lambotte

Lyon-Caen

Bernard

, et al. (2014) Reassessment of the rifting process in the Western Corinth rift from relocated seismicity. Geophysical Journal International 197: 1822–1844.

53.

Larsen

Torres Sanchez

(1998)The frequency and distribution of recent landslides in three montane tropical regions of Puerto Rico. Geomorphology 24: 309–331.

54.

Larsen

Montgomery

Korup

(2010) Landslide erosion controlled by hillslope material. Natural Geosciences 3: 247–251.

55.

Leonards

Sotiropoulos

Kavvada

(1988) Helice: The lost town of ancient Greece. In: Marinos

Koukis

(eds) Engineering Geology of Ancient Works Monuments and Historical Sites. Rotterdam: Balkema, pp. 1307–1313.

56.

Leopold

Wolman

(1960) River meanders. Bulletin of the Geological Society of America 71: 769–794.

57.

Lin

W-T

Chou

W-C

Lin

C-Y

, et al. (2014) Study of landslides caused by the 1999 Chi-Chi earthquake, Taiwan, with multi temporal SPOT images. Canadian Journal of Remote Sensing 33: 289–302.

58.

Lyritzis

Westra

Miao

(2019) Disaster GeoArchaeology and natural cataclysms in world cultural evolution: An overview. Journal of Coastal Research 35: 1307–1330.

59.

McNeill

Collier

REL

de Martini

, et al. (2005) Recent history of the eastern Eliki fault, Gulf of Corinth: Geomorphology, palaeoseismology and impact on palaeoenvironments. Geophysical Journal International 161: 154–166.

60.

Marinatos

(1960) A submerged town of classical Greece. Archaeology 113: 186–193.

61.

Martin

Rood

Schwab

, et al. (2002) Sediment transfer by shallow landsliding in the Queen Charlotte Islands, British Columbia. Canadian Journal of Earth Sciences 39: 189–205.

62.

Mockus

(1957) Use of storm and watershed characteristics in synthetic hydrograph analysis and application. Paper presented at the annual meeting of AGU Pacific Southwest Region, Sacramento, California, 14–15 February 1957. Sacramento, CA: American Geophysical Union, Pacific Southwest Region.

63.

Mouyaris

Papastamatiou

Vita-Finzi

(1992) The Helice Fault? Terra Nova 4: 124–129.

64.

Ñaco del Hoyo

Nappo

(2013) When the waters recede: The economic impact of tsunamis in the Graeco-Romanworld. Revue belge de philologie et d’histoire 91: 45–68.

65.

Nikolakopoulos

Antonakakis

Kyriou

, et al. (2018) Comparison of terrestrial laser scanning and structure-from-motion photogrammetry for steep slope mapping. In: Michel

Schulz

(eds) Proceedings of SPIE – Earth Resources and Environmental Remote Sensing/GIS Applications IX, 107900F, Berlin, Germany, 9 October 2018. Bellingham, WA: SPIE.

66.

Nikolakopoulos

Kavoura

Depountis

, et al. (2017) Preliminary results from active landslide monitoring using multidisciplinary surveys. European Journal of Remote Sensing 50: 280–299.

67.

Nixon

McNeill

Bull

, et al. (2016) Rapid spatiotemporal variations in rift structure during development of the Corinth Rift, central Greece. Tectonics 35: 1225–1248.

68.

Papadopoulos

(1998) A reconstruction of the great earthquake of 373 BC. In: Katsonopoulou

Soter

Schilardi

(eds) Ancient Helike and Aigialeia, Helike II. Athens: The Helike Society, pp. 479–494.

69.

Papadopoulos

(2000) A new tsunami catalogue of the Corinth Rift: 373 B.C. –A.D. 2000. In: Papadopoulos

(ed.) Historical Earthquakes and Tsunamis in the Corinth Rift, Central Greece. Greece: Institution of Geodynamics, National Observatory of Athens, pp. 121–126.

70.

Papadopoulos

(2003) Tsunami Hazard in the Eastern Mediterranean: Strong earthquakes and tsunamis in the Corinth Gulf, Central Greece. Natural Hazards 29: 437–464.

71.

Papadopoulos

Lobkovsky

Mazova

, et al. (2007) Numerical modeling of sediment mass sliding and tsunami generation: The case of February 7, 1963, in Corinth Gulf, Greece. Marine Geodesy 30: 315–331.

72.

Papanastassiou

(2002) Destructive Earthquakes and Seismotectonics in the Gulf of Corinth, Central Greece: A Review. In: Görur

Papadopoulos

Okay

(eds) Integration of Earth Science Research on the Turkish and Greek 1999 Earthquakes. Nato Science Series: IV: Earth and Environmental Sciences, Vol. 9. Dordrecht: Springer, pp. 153–173.

73.

Papazachos

Papazachou

(1989) The Earthquakes of Greece. Thessaloniki, Greece: Ziti and Co.

74.

Papazachos

Papazachou

(1997) The Earthquakes of Greece. Thessaloniki, Greece: Ziti and Co.

75.

Papazachos

Papazachou

(2003) The Earthquakes of Greece. Thessaloniki, Greece: Ziti and Co., p. 304.

76.

Passega

(1957) Texture as characteristic of clastic deposition. American Association of Petroleum Geology Bulletin 41: 1952–1984.

77.

Pavlides

Koukouvelas

Kokkalas

, et al. (2004) Late-Holocene evolution of the East Eliki Fault, Gulf of Corinth (Central Greece). Quaternary International 115–116: 139–154.

78.

Piper

DJW

(1975) Textural recognition of mudflow deposits. Sedimentary Geology 13: 303–306.

79.

Perrou

Kaza

Efthymiadis

, et al. (2013) Recent coastline changes of fan-deltas in the Western Gulf of Corinth, Central Greece. Bulletin of the geological Society of Greece 47: 315–324.

80.

Pescaroli

Alexander

(2015) A definition of cascading disasters and cascading effects: Going beyond the “toppling dominos” metaphor. Planet@ Risk 3(1): 58–67.

81.

Pescaroli

Alexander

(2016) Critical infrastructure, panarchies and the vulnerability paths of cascading disasters. Natural Hazards 82: 175–192.

82.

Psarropoulos

Gkantona

(2020) Information Technology and a new theory for the verification of the location of the ancient city of Helike in Achaia, Greece. In: Correia

Tinoco

Cortez

, et al. (eds) Information technology in geo-engineering. ICITG 2019. Springer Series in Geomechanics and Geoengineering. Cham: Springer, pp. 122–130.

83.

Poulimenos

Doutsos

(1996) Barriers on seismogenic faults in Central Greece. Journal of Geodynamics 22: 119–135.

84.

Skarlatos

Procopiou

Stavrou

, et al. (2013) Accuracy assessment of minimum control points for UAV photography and georeferencing. In: Proceedings of SPIE – The International Society for Optical Engineering 8795, Pafos, Cyprus, 8-10 April 2013. Bellingham, WA: SPIE, 879514.

85.

Schmidt

(1879) Studien über Erdbeben. Leibzig: Alwin Georgi Edition.

86.

Simas

(1996) Lag time characteristics in small watersheds in the United States. A dissertation submitted to School of Renewable Natural Resources, University of Arizona, Tucson, AZ.

87.

Soter

Katsonopoulou

(1998) Geophysical Observations on the Hill of Agios Georgios, Possible Site of the Acropolis of Helike. In: Katsonopoulou

Soter

Schilardi

(eds) Ancient Helike and Aigialeia, Helike II. Athens: The Helike Society, pp. 153–162.

88.

Soter

Katsonopoulou

(2011) Submergence and Uplift of Settlements in the Area of Helike, Greece, from the Early Bronze Age to Late Antiquity. Geoarchaeology 26(4): 584–610.

89.

Takashimizu

Masuda

(2000) Depositional facies and sedimentary successions of earthquake-induced tsunami deposits in Upper Pleistocene incised valley fills, central Japan. Sedimentary Geology 135: 231–239.

90.

Tannant

Skermer

(2013) Mud and debris flows and associated earth dam failures in the Okanagan region of British Columbia. Canadian Geotechnical Journal 50: 820–833.

91.

Thanasoulopoulos

(2007) Local names of Achaea as historical testimonies of its medieval settlement (in Greek). Athens: Peri Texnon, p. 72.

92.

Triantafyllou

Papadopoulos

Lekkas

(2020) Impact on built and natural environment of the strong earthquakes of 23 April 1933 and 20 July 2017 in southeast Aegean Sea, eastern Mediterranean. Natural Hazards 100: 671–675.

93.

Walter

(2017) Poseidon’s Wrath and the end of Helike: Notions about the Anthropogenic character of disaster in antiquity. In: Schliephake

(ed.) Ecocriticism, Ecology, and the Cultures of Antiquity. Lanham: Lexington Books, pp. 31–43.

94.

Whipple

Tucker

(1999) Dynamics of the stream – power river incision model: Implications for height limits of mountain ranges, landscape response timescales, and research needs. Journal of Geophysical Research 104: 17661–17674.

95.

Zygouri

Koukouvelas

(2019) Landslides and natural dams in the Krathis River, north Peloponnese, Greece. Bulletin of Engineering Geology and the Environment 78: 207–222.

96.

Zygouri

Verroios

Kokkalas

, et al. (2008) Scaling properties within the Gulf of Corinth, Greece; comparison between offshore and onshore active faults. Tectonophysics 453: 193–210.