# Hydro- and morphodynamic tsunami simulations for the Ambrakian Gulf (Greece) and comparison with geoscientific field traces

Hydro- and morphodynamic tsunami simulations for the Ambrakian Gulf (Greece) and comparison with... Summary In order to derive local tsunami risks for a particular coast, hydro- and morphodynamic numerical models that are calibrated and compared with sedimentary field data of past tsunami impacts have proven very effective. While this approach has widely been used with regard to recent tsunami events, comparable investigations into pre-/historical tsunami impacts hardly exist, which is the objective of this study focusing on the Ambrakian Gulf in northwestern Greece. The Ambrakian Gulf is located in the most active seismotectonic and by this most tsunamigenic area of the Mediterranean. Accordingly, palaeotsunami field studies have revealed repeated tsunami impacts on the gulf during the past 8000 yr. The current study analyses 151 vibracores of the Ambrakian Gulf coast in order to evaluate tsunami signals in the sedimentary record. Based on a hydro- and morphodynamic numerical model of the study area, various tsunami waves are simulated with the aim of finding scenarios that compare favourably with tsunami deposits detected in the field. Both, field data and simulation results suggest a decreasing tsunami influence from the western to the eastern Ambrakian Gulf. Various scenarios are needed to explain tsunami deposits in different parts of the gulf. Whereas shorter period tsunami waves (T = 30 min) from the south and west compare favourably with field data in the western gulf, longer period waves (T = 80 min) from a western direction show the best agreement with tsunami sediments detected in southwestern Aktio Headland and in the more central parts of the Ambrakian Gulf including Lake Voulkaria. Tsunamis from the southwest generally do not accord with field traces. Besides the spatial sediment distribution, the numerical model accurately reflects the sedimentary composition of the detected event deposits and reproduces a number of essential features typical of tsunamites, which were also observed in the field. Such include fining- and thinning-landward and the marine character of the deposits. By contrast, the simulated thickness of tsunami sediments usually lags behind the observed thickness in the field and some event layers cannot be explained by any of the simulated scenarios. Regarding the frequency of past tsunami events and their spatial dimensions indicated by both field data and simulation results, a high tsunami risk has to be derived for the Ambrakian Gulf. Geomorphology, Tsunamis, Non-linear differential equations, Numerical modelling, Neotectonics, Subduction zone processes 1 INTRODUCTION With more than 10 per cent of the world’s population living in coastal lowlands less than 10 m a.s.l. (above sea level; McGranahan et al.2007, p. 22), the vulnerability of coasts to tsunami inundation is very high. Recent history has seen disastrous tsunami events in Papua New Guinea (1998), southeast Asia (2004) and Japan (2011), which have particularly emphasized the need for improved risk assessment and mitigation plans. During the last years, tsunami forecast methods have significantly extended and gained in reliability. Modern tsunami risk assessment uses a multi-methodological approach based on oceanic warning systems, seismological data, numerical simulation techniques and sedimentary field traces. In order to derive tsunami risks for a particular coast, hydro- and morphodynamic numerical tsunami models that are calibrated and compared with sedimentary field data and—if available—buoy measurements and information on the trigger mechanisms of past tsunami impacts have proven very effective. Knowing the frequency and the spatial extent of pre-/historical tsunami inundation derived from field traces makes it possible to draw conclusions on future tsunami risks. However, field traces only allow an accurate reconstruction of former inundation events when combined with numerical tsunami models, as performed in numerous recent studies (e.g. Bondevik et al.2005; Martin et al.2008; Borrero et al.2009; Röbke et al.2013, 2015, 2016; Satake et al.2013; Sugawara et al.2013; Hill et al.2014). Such studies are particularly promising, if numerical models do not only compute the pure tsunami hydrodynamics but also take account of sediment processes, which allows a one-to-one comparison between field data and model (cf. Sugawara et al.2014a). This approach was taken in a number of studies focusing on tsunami impacts in recent history, in particular the Indian Ocean tsunami 2004 in Sumatra and Thailand (Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b; Ontowirjo et al.2010, 2013; Gusman et al.2012; Li et al.2012a,b, 2014), the 2009 South Pacific tsunami in American Samoa (Apotsos et al.2011c) and the 2011 Tōhoku tsunami in Japan (Sugawara et al.2014b). A good preservation provided, tsunami field traces together with numerical simulations may further give valuable insights into related tsunami trigger mechanisms (e.g. Dawson 1999; Martin et al.2008; Sugawara et al.2011, 2012; Namegaya & Satake 2014). Although the synthesis of sedimentary field data and morphodynamic numerical models has repeatedly been applied with regard to recent tsunami events, there are hardly comparable studies in view of pre-/historical tsunamis (cf. Röbke & Vött 2017, pp. 313–315), which is the approach of the following investigation. This paper concentrates on tsunami impacts in the Ambrakian Gulf (northwestern Greece) during the last 8000 yr. The Ambrakian Gulf is a marine inlet of the eastern Ionian Sea in coastal Epirus and Akarnania with a maximum N–S- and W–E-extent of about 20 km and 35 km respectively (Figs 1 and 2). The gulf is connected to the Ionian Sea by the Strait of Preveza, a 600 m wide channel less than 25 m deep (UKHO 2010). Maximum water depths are found in the eastern gulf with roughly 65 m. The Ambrakian Gulf forms a tectonic basin dominated by N–S crustal extension along WNW-ESE trending normal faults (Clews 1989, p. 454). According to palaeogeographical studies by Poulos et al. (1995, 2005), Tziavos (1997), Jing & Rapp (2003), Brockmüller et al. (2005) and Kapsimalis et al. (2005), the gulf was isolated from the Ionian Sea during the last ice age (Weichselian or Würm period), but became a marine inlet in the early Holocene (epoch of the last 11 700 yr). By the 4th millennium BC, the ongoing rise in sea level had resulted in the complete flooding of the contemporary river deltas in the northern gulf, which extended far more into the gulf than nowadays. At the same time, the coastline in the western, southern and eastern gulf approximately reached its present position. Subsequently, considerable sediment input by the rivers Louros and Arachthos created the large delta complex nowadays found in the northern gulf (Fig. 2). To the present, tectonic crust movements cause a relative uplift of the northwestern gulf compared to the southeastern part, having resulted in a vertical shift of about 6 m since the 5th millennium BC (Brockmüller et al.2005, pp. 46, 47). Figure 1. View largeDownload slide Satellite image and submarine bed level in m b.s.l. (below sea level) of the central and eastern Mediterranean Google Inc. (2013), including the simplified tectonic regime and location of selected backarc volcanoes according to Faccenna et al. (2001, p. 810), Kokkalas et al. (2006, p. 100) and Polonia et al. (2011, p. 2). Grey dotted boxes display the overall and nested model domains, the red dotted lines mark the hydrodynamic open boundaries (south, southwest and west) where the tsunami waves enter the overall model. Figure 1. View largeDownload slide Satellite image and submarine bed level in m b.s.l. (below sea level) of the central and eastern Mediterranean Google Inc. (2013), including the simplified tectonic regime and location of selected backarc volcanoes according to Faccenna et al. (2001, p. 810), Kokkalas et al. (2006, p. 100) and Polonia et al. (2011, p. 2). Grey dotted boxes display the overall and nested model domains, the red dotted lines mark the hydrodynamic open boundaries (south, southwest and west) where the tsunami waves enter the overall model. Figure 2. View largeDownload slide Overview map of Greece and bird’s eye view of the Ambrakian Gulf, including the opening to the Ionian Sea, the coastal flats in the north and the hilly shore in the southeast. The figure is based on a 3-D excerpt of the digital elevation model employed for numerical tsunami simulations in this study. *Due to the 3-D view, the scale bar gives the true distance only in the centre of the figure. Figure 2. View largeDownload slide Overview map of Greece and bird’s eye view of the Ambrakian Gulf, including the opening to the Ionian Sea, the coastal flats in the north and the hilly shore in the southeast. The figure is based on a 3-D excerpt of the digital elevation model employed for numerical tsunami simulations in this study. *Due to the 3-D view, the scale bar gives the true distance only in the centre of the figure. Located in the most seismically active and by this most tsunamigenic region in the entire Mediterranean, that is, western Greece (Fig. 1; Soloviev et al.2000, p. 14; Schielein et al.2007, pp. 164, 193–199; Röbke et al.2013, pp. 70–77), the Ambrakian Gulf and its environs were frequently affected by tsunami landfall in the past. Besides historical accounts, this is documented by sedimentary field traces in the area. According to detailed field studies by Vött et al. (2006, 2007, 2008, 2009a, 2010) and May et al. (2007, 2012a), major tsunamis struck the Ambrakian Gulf in the 6th millennium BC or shortly afterwards, between 2870 cal BC and 2350 cal BC, around 1000 cal BC, between 395 cal BC and 247 cal BC, in the 4th century AD and around 840 cal AD. Further events occurred in the last 700 yr. Sedimentary signatures of most of these impacts were also found in close vicinity to the gulf on the nearby island of Lefkada (Vött et al.2006, 2008, 2009b; May et al.2012a,b) and in the Bay of Palairos (Vött et al.2011) and seem to be related to supraregional tsunami events in the Mediterranean. The wider study area is the northernmost region in the eastern Mediterranean where sedimentary signatures of the 365 AD Crete tsunami were found (e.g. Vött et al.2009b; May et al.2012b). Besides field studies, hydrodynamic numerical tsunami simulations were computed for the coastal area between Preveza and Lefkada City (Floth et al.2009, 2010, 2013). In this initial approach, simulation results, which were based on a small-scale numerical model of the area, were compared with local tsunami field traces. The current study aims to (i) analyse and evaluate systematically tsunami signals in the sedimentary record of 151 vibracores taken all along the Ambrakian Gulf within previous palaeotsunami field studies, (ii) derive the scale of pre-/historical tsunami landfall in the gulf based on field data, (iii) set up a hydro- and morphodynamic numerical tsunami model, which takes account of various tsunami wave scenarios and considers sediment processes, (iv) calibrate and compare the numerical model with the spatial distribution and sedimentary composition of detected tsunami deposits in the field and find tsunami scenarios that replicate field data in order to (v) assess tsunami risks for the Ambrakian Gulf. 2 METHODOLOGY 2.1 Field work Within the framework of previous palaeogeographical studies, 151 vibracores were drilled alongshore the Ambrakian Gulf using a handheld engine-powered Atlas Copco coring device (type Cobra mk1). The final coring depth at each location (maximum 18 m below ground surface) was reached, when either pre-Holocene deposits or bedrock appeared or the sedimentary facies did not change for several metres. Vibracores were first documented by digital photographs and subsequently described regarding their sedimentary features, such as grain size, texture, colour, fossil content and pedogenetic properties (cf. Ad Hoc-AG Boden 2005). The exact location of all vibracores was determined using a Topcon DGPS-device (type HiPer Pro). 2.2 Evaluation of field data A number of vibracores taken in the Ambrakian Gulf had already been analysed in view of the regional palaeogeography and palaeo-tsunami signatures by Vött et al. (2006, 2007, 2008, 2009a, 2010, 2011) and May et al. (2007, 2012a). Following these investigations, in the current study, all 151 vibracores were analysed and evaluated systematically with regard to tsunami candidate layers based on a set of sedimentary criteria which have been associated with onshore tsunamites worldwide (e.g. Dawson 1994; Dawson & Shi 2000, pp. 173–180; Dawson & Stewart 2007a, pp. 578–581; Dawson & Stewart 2007b; Dominey-Howes et al.2006; Morton et al.2007; Fujiwara 2008; Shiki et al.2008, pp. 45, 46; Vött & May 2009; Shanmugam 2012, pp. 307–310; Röbke & Vött 2017). Such physical criteria are directly ascribed to the hydrodynamic characteristics of tsunami inundation and particularly include basal erosional unconformities, fining-upward sequences, high content of marine macrofossils, clayey to silty rip-up clasts from underlying units, mud caps at the top of event layers, fining- and thinning-landward, a general allochthonous, high-energy character of the sediment (cf. Fig. 3). Figure 3. View largeDownload slide Vibracore AKT 40 showing sedimentary features typical of tsunami deposits, including a basal erosional unconformity, numerous marine macrofossils, normal grading and clayey to silty rip-up clasts from underlying units. Such features, among others, were referred to in order to identify tsunami candidate layers in 151 vibracores alongshore the Ambrakian Gulf shown in Fig. 5. Figure 3. View largeDownload slide Vibracore AKT 40 showing sedimentary features typical of tsunami deposits, including a basal erosional unconformity, numerous marine macrofossils, normal grading and clayey to silty rip-up clasts from underlying units. Such features, among others, were referred to in order to identify tsunami candidate layers in 151 vibracores alongshore the Ambrakian Gulf shown in Fig. 5. For some vibracores, also geochemical, microfaunal and palynological analyses had been conducted, which further facilitated the identification of tsunami deposits in the study area (Brockmüller et al.2005; Jahns 2005; Vött et al.2006, 2007, 2008, 2009a, 2010, 2011; May et al.2007, 2012a). Findings from these studies together with physical sedimentary features described above and correlations between related event strata at neighbouring vibracoring sites allowed a systematic and standardized identification of tsunami deposits along the entire Ambrakian Gulf coast. In this process, also storm events were taken into account as a potential mechanism to leave high-energy deposits. However, the lack of written accounts of extraordinary storm surges but multiple sources of data of tsunami impacts in the study area, the geomorphological appearance of the Ambrakian Gulf coast, which does not imply any recent to subrecent storm activity beyond the supralittoral zone, regional buoy and tide gauge data and studies by Ghionis et al. (2015) and Poulos et al. (2015, pp. 43–45) that reveal an expected maximum run-up of wind generated waves of only 2.82 m a.s.l. for the nearby tip of northern Lefkada, the short maximum fetch in the gulf of less than 40 km (cf. Fig. 2) allowing no high waves to develop, the lack of any sedimentary features in the vibracores typically associated with storm deposits (tempestites), and the fact that several event layers in the vibracores of the Ambrakian Gulf correlate well with tsunami deposits described from other coasts in western Greece and by this probably point to the same tsunami events strongly suggest the event deposits being of tsunamigenic origin (for more details see Vött et al.2007, pp. 51–53; Vött et al.2008, pp. 116–119; Vött et al.2011, pp. 233, 234). Correlations between related event strata could only be drawn for selected vibracores for which geochronological data were available but not for the whole set of vibracores. In order to still achieve a clear and meaningful overview of all vibracore data in the study area and to allow a direct comparison with simulation results, the average thickness of all tsunami candidate layers per vibracore was determined. This parameter indicates the average sedimentary impact of a single tsunami event most adequately and can be compared to the amount of sediment accumulation or erosion computed for 18 different single tsunami wave scenarios in the numerical model (Section 2.4). The average thickness was classified according to the Jenks optimization method, which yields the minimum in-class variance but maximizes the variances between all classes (Jenks 1967; Jenks & Caspall 1971; McMaster 1997). This classification method turned out to be most appropriate with regard to field data of this study as it demonstrates natural breaks in the frequency distribution and by this allows an easier visual distinction between areas of dominant and low tsunami sedimentation. Also the standard deviation of the tsunami layer thickness per vibracore was determined, which shows that the distribution of thicknesses per vibracore is relatively peaked and simple (Section 3.1). This justifies a direct comparison of the average tsunami layer thickness per vibracore (resulting from several tsunami events in pre-/history) with the simulated sedimentation and erosion by individual tsunami waves in the numerical model. Using geochronological data, it was further possible to compare several correlating tsunami candidate layers, thus deriving from the same events, with the simulation results for three locations in the western Ambrakian Gulf. This was done on the basis of the 840 cal AD tsunami for southwestern Aktio Headland, the 1000 cal BC tsunami for the entrance of Lake Voulkaria as well as the 4th century AD tsunami for the southern shore of Lake Voulkaria. Based on these events and locations, field and model data could directly be compared in view of individual past tsunamis. This further allowed a comparison not only of the sediment thickness but also of the sedimentary composition (i.e. content of mud, fine sand, middle sand, coarse sand and gravel) of tsunami deposits in both, the field and the model. 2.3 Digital elevation model The digital elevation model (DEM) used for tsunami simulations in this study covers the eastern Ionian Sea and the west coast of Greece, including the Ambrakian Gulf (Figs 1 and 2). The domain stretches from 37.47°N to 39.81°N and from 18.75°E to 21.63°E ($$\widehat{=}$$ 260 km * 250 km). The DEM was derived from several sources of data. The offshore topography of the open sea is based on the 30 arc-second (= 926.1 m) interval grid of GEBCO (IOC, IHO & BODC 2003). Bathymetric data of the shelf and the Ambrakian Gulf was digitized from several nautical charts on the scale of 1:93 213 to 1:300 000 (Imray Laurie Norie & Wilson Ltd 2011; Eagle Ray Publications 2010; UKHO 2010). SRTM elevation data version 4 with a resolution of 90 m (Jarvis et al.2008) and 30 m Google Earth elevation data (Google Inc. 2013; generated using the eSurvey Earth website http://www.esurveyearth.com/) were used to derive the onshore topography of the wider study area and alongshore the Ambrakian Gulf respectively. Among all freely available elevation data, both data sets showed the best match with DGPS-based elevation data measured in the field and altitude markings in topographical maps. All elevation data was edited and assembled to an overall DEM with a maximum resolution of 30 m using ArcGIS Desktop 10.2 by Esri. 2.4 Hydro- and morphodynamic numerical model In this study, a hydro- and morphodynamic numerical tsunami model was created using the Delft3D modelling suite (version 4.01.00.rc.12) developed by the Dutch institute Deltares. Delft3D is a validated software suite to simulate both, off- and onshore tsunami propagation, including sediment transport (Apotsos et al.2011a). It has been successfully employed for 1-D (Apotsos et al.2011d, 2012), 2-D (Röbke et al.2012, 2013, 2015, 2016; Chacón-Barrantes et al.2013; Períañez & Abril 2014a,b) and 3-D tsunami simulations (Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b,c). All simulations were calculated using the program module Delft3D-FLOW, which solves the nonlinear shallow water equations of unsteady flow and transport phenomena based on the full Navier–Stokes equations for incompressible free surface flow (for this and the following see Deltares 2014, pp. 177–180). The module is designed for flow phenomena where the horizontal spatial and temporal scales are much larger than the vertical scales, such as tidal waves, storm surges or tsunamis. In Delft3D-FLOW, the nonlinear shallow water equations are solved in two (depth-averaged) or in three dimensions. As the water density in the oceans can approximately be regarded as vertically homogeneous and 3-D test simulations showed a small impact on mud dynamics in the numerical model, the 2D, depth-averaged calculation mode is appropriate for tsunami simulations conducted here. This study only considers the pure hydro- and morphodynamics of tsunami waves, whereas trigger mechanisms and the process of wave generation were not simulated. Accordingly, the numerical model is based on various hydrodynamic boundary conditions describing tsunami waves with different wave periods, wave amplitudes and propagation directions in order to derive wave scenarios that accord with sedimentary field data (cf. Pritchard & Dickinson 2008; Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b,c,d, 2012). This approach was taken because in most cases, no correlation could be drawn between palaeotsunami field traces and written tsunami accounts for the Mediterranean and, by this, no information was available neither on related trigger mechanisms nor on tsunami wave characteristics. Furthermore, there is no single specific tsunami source with regard to the Ambrakian Gulf, but instead various potential trigger mechanisms exist, both in the near- and far-field (cf. Fig. 1; Röbke et al.2013, pp. 74–77; Samaras et al.2015, p. 649): strong shallow-focus earthquakes with a magnitude of Ms ≥ 6.5 along several shallow dipping thrust faults of up to 300 km in the area of the Calabrian Trench, the Hellenic Trench and its northward extension (Papazachos 1996; Benetatos et al.2004,USGS 2015), huge submarine landslides connected with the ubiquitous steep slopes and high seismic activity in the central and eastern Mediterranean Sea (Ferentinos 1991, 1992; Papadopoulos & Plessa 2000; Pareschi et al.2006a,b), severe volcanic eruptions of Mount Etna (Sicily, Italy) and other backarc volcanoes in the Aegean Sea (Pareschi et al.2006a,b,c). Thus, the number of potential tsunami sources is too large to be considered in this study. Instead, artificial waves that are realistic in view of the (supra)regional physical geography were created. The hydrodynamic boundary conditions of the numerical model were determined according to different tsunami wave types, that is, leading-elevation N-waves (LEN-waves), leading-depression N-waves (LDN-waves) (e.g. Tadepalli & Synolakis 1994, 1996) and solitary waves (e.g. Dean & Dalrymple 1991). A sensitivity analysis based on multiple test runs demonstrated that the general spatial distribution of tsunami deposits in the study area as found in the vibracores was best replicated by those test runs using LDN-waves. In contrast, LEN-waves and solitary waves resulted in a simulated tsunami sedimentation frequently below the magnitude of the field observations (cf. N-wave effect discussed by Röbke et al.2016) and/or showing a divergent pattern. LDN-waves particularly occur when tsunamis are triggered relatively close to the coast and the waves do not have enough time to evolve into LEN-waves or solitary waves (Tadepalli & Synolakis 1994, p. 100; Tadepalli & Synolakis 1996, p. 2143; Madsen et al.2008, pp. 2, 3; Synolakis et al.2008, pp. 2202, 2203; Liang et al.2013, pp. 144, 145; Röbke & Vött 2017, p. 313). Considering the multiple potential generation mechanisms of tsunamis in close vicinity to the Ambrakian Gulf (see above and Fig. 1), LDN-waves are likely to appear in the study area. Therefore and as a result of the sensitivity analysis for different wave types, this paper concentrates on isosceles LDN-waves, using the formulation by Tadepalli & Synolakis (1994, p. 105) (Chacón-Barrantes et al.2013, pp. 15, 16):   \begin{eqnarray} \eta &=& \frac{3\,\sqrt{3}\,H}{2}\ \mathrm{sech}^2\left[\frac{3\,\alpha }{2}\,\root 4 \of {\frac{3\,A}{4}}\,\left(t-t_0\right)\right]\nonumber\\ &&\times\,\tanh \left[\frac{3\,\alpha }{2}\,\root 4 \of {\frac{3\,A}{4}}\,\left(t-t_0\right)\right] \end{eqnarray} (1)where η = elevation above water surface [m], H = wave height [m], α = constant modifying the wavelength, A = wave amplitude ($$= \frac{H}{2}$$) [m], t = time [s] and t0 = time of the wave’s mid-point (half wave period) [s]. LDN-waves with periods of 30 min and 80 min and amplitudes of 2 m, 4 m and 6 m were created (Fig. 4), these wave parameters having the strongest influence on the onshore tsunami response. Furthermore, three propagation directions were considered in the numerical model, that is, south, southwest and west, which are the favourite wave propagation directions with regard to the potential tsunami generation mechanisms (cf. Fig. 1). All LDN-waves were simulated as single waves in order to investigate their individual morphodynamic impact depending on the amplitude, period and propagation direction. This resulted in a total of 18 scenarios discussed in this paper (Table 1). Figure 4. View largeDownload slide Leading-depression N-waves according to Eq. 1 with amplitudes of 2 m, 4 m and 6 m and periods of 30 min and 80 min. These waves form the hydrodynamic boundary conditions of the numerical tsunami model of this study at the southern, southwestern and western model boundary. Figure 4. View largeDownload slide Leading-depression N-waves according to Eq. 1 with amplitudes of 2 m, 4 m and 6 m and periods of 30 min and 80 min. These waves form the hydrodynamic boundary conditions of the numerical tsunami model of this study at the southern, southwestern and western model boundary. Table 1. Overview of the 18 tsunami scenarios discussed in this paper. The scenarios are based on three propagation directions, two wave periods and three wave amplitudes. LDN-wave from  Wave period  Wave amplitude  Scenario      2 m  S2S    30 min  4 m  S4S      6 m  S6S  South            2 m  S2L    80 min  4 m  S4L      6 m  S6L      2 m  SW2S    30 min  4 m  SW4S      6 m  SW6S  Southwest            2 m  SW2L    80 min  4 m  SW4L      6 m  SW6L      2 m  W2S    30 min  4 m  W4S      6 m  W6S  West            2 m  W2L    80 min  4 m  W4L      6 m  W6L  LDN-wave from  Wave period  Wave amplitude  Scenario      2 m  S2S    30 min  4 m  S4S      6 m  S6S  South            2 m  S2L    80 min  4 m  S4L      6 m  S6L      2 m  SW2S    30 min  4 m  SW4S      6 m  SW6S  Southwest            2 m  SW2L    80 min  4 m  SW4L      6 m  SW6L      2 m  W2S    30 min  4 m  W4S      6 m  W6S  West            2 m  W2L    80 min  4 m  W4L      6 m  W6L  View Large The chosen wave periods and amplitudes were derived from a previous comparison of the field data with multiple test runs based on various combinations of different wave periods and amplitudes, which are plausible considering the potential tsunami trigger mechanisms in the central and eastern Mediterranean Sea. Within this model calibration process, those combinations of wave periods and amplitudes were chosen, that can explain tsunami deposits found in the more seaward, the more landward and the intermediate parts of the study area. This approach yielded a possible magnitude range for tsunami waves that hit the Ambrakian Gulf in pre-/history and left the detected deposits. Earlier simulation studies revealed that tsunami waves with periods of between 30 min and 80 min combined with amplitudes of between 2 m and 6 m have probably occurred in the Mediterranean Sea and are therefore likely to occur again. For example, Pareschi et al. (2006a) simulated tsunami waves generated by a huge landslide of up to 30 km2 off Mount Etna in eastern Sicily about 8000 yr ago. The computed maximum wave amplitudes along the western model boundary in the current study, located in the central Ionian Sea (Fig. 1), are in the order of 5 m with periods of several tens of minutes. Períañez & Abril (2014a) simulated various landslide, volcanically and seismically generated tsunamis in the central and eastern Mediterranean. While most of the landslide and volcanically generated tsunamis yielded wave amplitudes smaller than 3 m in the central Ionian Sea, amplitudes of about 5 m and periods of more than 1 h were found for the tsunami connected to the 365 AD Crete earthquake with an estimated magnitude of Ms = 8.3. Similar wave amplitudes and periods are suggested by Yolsal-Çevikbilen & Taymaz (2012) for the same tsunami event. The 365 AD Crete earthquake probably resulted in a vertical shift of the west coast of Crete by almost 9 m and is associated with a fault length of 105 km, a width of 100 km, a slip of 16 m, a strike of 292.5° and a dip of 40° (Stiros 2010). All simulations are based on a two-part computational grid—an overall grid and a detailed, nested grid (Fig. 1). The overall grid has a resolution of 900 m and covers the entire domain of the DEM (Section 2.3). The boundary conditions for the various tsunami waves from the three directions south, southwest and west were defined along the corresponding open boundaries of the overall grid as shown in (Fig. 1). The nested grid with a resolution of 90 m stretches from 38.72°N to 39.24°N and from 20.56°E to 21.20°E ($$\widehat{=}$$ 58 km * 55 km). The boundary conditions of the nested model were determined by bilinear interpolation of the results computed at corresponding monitoring stations in the overall model (for further methodological details see Deltares 2014, pp. 677–683). For both, the overall and nested model, an alpha reflection parameter of 100 was applied along the open boundaries in order to minimize unintended wave reflections (Verboom & Slob 1984; pp. 205–209 Deltares 2014). Although a refinement factor of 10 from the overall to the nested grid (as applied here) might cause inaccuracy in several Delft3D-FLOW applications (typically factor 3 to 5), the current model showed a low sensitivity for this parameter based on comparison runs using a three-part computational grid with a resolution of 900 m, 300 m and 90 m respectively. Comparing computed inundation depths and flow velocities for various observation points within the 90 m grid domain revealed small differences between both grid configurations of less than 0.05 m and 0.1 m s−1 respectively. Correspondingly, differences in the computed erosion and sedimentation did not exceed few centimetres. These discrepancies are neglectable considering the computed absolute inundation depths and flow velocities in the order of up to 5 m and 7 m s−1 respectively (Section 3.2) as well as the simulated erosion and sedimentation with magnitudes of up to 4 m (Sections 3.3 and 3.4). Therefore, and in order to reduce the computational cost of the model, the two-part computational grid has been applied to all final simulation runs. The calculation step was 0.6 s in both, the overall and the nested model, which results in maximum Courant numbers Cmax of 0.38 and 0.84 respectively. These values are significantly below the recommended threshold of Cmax = 10, which, based on the implicit scheme of Delft3D-FLOW, is already sufficient to prevent numerical instability and inaccuracy in most applications including the simulation of long shallow-water waves (Deltares 2014, pp. 279, 280). All scenarios were computed for the time period characterized by significant changes in water level and significant sediment transport, which was between 8 h and 10 h. Based on field surveys, photographs and Google Earth satellite images (Google Inc. 2013), appropriate values of Manning’s n coefficient were determined in order to take account of the bottom roughness in the model domain. n-values ranged from 0.015 (seafloor, pastures) over 0.03 (moderately vegetated or developed land) to 0.07 (densely forested landscapes) (cf. Hills & Mader 1997, p. 388; Bryant 2008, pp. 45, 46). With the aim of simulating tsunami morphodynamics, a realistic off- and onshore sediment stratigraphy had to be determined. This was only done for the detailed, nested model because no significant sediment transport was found for the overall model owing to the large water depths. Delft3D-FLOW considers both, suspended load and bedload transport of non-cohesive sediments as well as suspended load transport of cohesive sediments (for further methodological details see Deltares 2014, pp. 323–373; Lesser et al.2004). In Delft3D-FLOW, cohesive sediments are defined by a sediment diameter of less than 0.063 mm and are no further distinguished, meaning that silt and clay are treated as one grain size, that is mud. In the case of non-cohesive sediments, however, any grain sizes between fine and coarse sand (0.063 mm $$\le \varnothing <$$ 2 mm) can be considered. Based on these specifications, the sediment stratigraphy of the nested model was determined according to sedimentary field data of the study area. The 151 onshore vibracores taken within this study (Section 2.1) revealed that the thickness of unconsolidated sediment exceeds 10 m in most places. Regarding the offshore area, an even larger thickness has to be assumed, for example, up to 45 m in the inner Ambrakian Gulf (Kapsimalis et al.2005, p. 404). Nevertheless, for both, the onshore and offshore parts of the nested model, an initial sediment layer thickness of 10 m was chosen since the maximum amount of sediment erosion was less than 4 m in all simulated scenarios. Statistical analyses of all 151 vibracores alongshore the Ambrakian Gulf led to an average onshore sediment composition for the upper 10 m below surface of about 67 per cent mud ($$\varnothing <0.063$$ mm), 9 per cent fine sand ($$\varnothing =0.1315$$ mm), 7 per cent middle sand ($$\varnothing =0.415$$ mm), 8 per cent coarse sand ($$\varnothing =1.315$$ mm) and 8 per cent gravel ($$\varnothing \ge 2$$ mm) (Table 2). Although the average grain size of gravel is 32.5 mm, a sediment diameter of 2 mm (maximum diameter of coarse sand) had to be chosen instead as this is the maximum grain size supported in Delft3D-FLOW. Judging by sediment cores, sediment samples and seismic profiles taken by Tziavos (1997), Poulos et al. (1995, 1999, 2005), Kapsimalis et al. (2005) and Karageorgis et al. (2006) in the Ambrakian Gulf, its lagoons and in the adjacent eastern Ionian Sea, an average offshore sediment composition of 80 per cent mud as well as 20 per cent fine sand was found for the upper 10 m below sea bottom (Table 2). Taking the average sediment fractions as a basis for the sediment stratigraphy in the nested model resulted in an initial onshore and offshore sediment layer of 10 m each, in which all corresponding grain sizes (ranging from mud to coarse sand) are homogeneously mixed. Test runs comparing such homogeneously mixed layers with actual stratified layers revealed that homogeneously mixed layers are most appropriate because they provide the model with those sediment fractions which are available in the area for redeposition by tsunami waves. Furthermore, the vibracore density is not high enough to derive a reliable comprehensive stratigraphy for the nested model and the pre-tsunami stratigraphy is not known. Table 2. Overview of the average on- and offshore sediment composition of the upper 10 m below surface/sea bottom for the nested model of the Ambrakian Gulf based on statistical analyses of 151 vibracores of the study area and sedimentary data taken from literature (see the text for further explanations). Grain size  Average onshore sediment composition of the upper 10 m below surface [per cent]  Average offshore sediment composition of the upper 10 m below sea bottom [per cent]  Cohesive      Mud  67  80  Non-cohesive      Fine sand  9  20  Middle sand  7  0  Coarse sand  8  0  Gravel  8  0  Grain size  Average onshore sediment composition of the upper 10 m below surface [per cent]  Average offshore sediment composition of the upper 10 m below sea bottom [per cent]  Cohesive      Mud  67  80  Non-cohesive      Fine sand  9  20  Middle sand  7  0  Coarse sand  8  0  Gravel  8  0  View Large Cohesive sediment erosion and deposition in Delft3D-FLOW was computed according to the Partheniades–Krone formulations (Partheniades 1965), whereas computation of non-cohesive sediment processes was based on the approach of van Rijn et al. (2001) (for further methodological details see Deltares 2014, pp. 323–373). The specific density and dry bed density of cohesive sediment was set to 2650 kg m−3 and 500 kg m−3 respectively, with a settling velocity of 0.25 mm s−1. The critical bed shear stress was 1000 N m−2 for sedimentation and 0.5 N m−2 for erosion of cohesive sediment, and a corresponding erosion parameter of 0.0001 kg m−2 s−1 was chosen. Independent of the grain size, the specific density and dry bed density of non-cohesive sediment were set to 2650 kg m−3 and 1600 kg m−3 respectively. As an important feature of tsunami inundation, the numerical model accounted for the effect of sediment concentration on fluid density by adding the mass of all transported sediment per unit water volume and subtracting the corresponding mass of displaced water (UNESCO 1981; Deltares 2014, pp. 324, 325). Furthermore, the hindered sediment settling velocity due to other sediment particles in highly loaded water was considered according to Richardson & Zaki (1997) (cf. Deltares 2014, pp. 324, 325), based on a reference density for hindered settling of 1600 kg m−3 for all grain sizes. Finally, changes in the on- and offshore topography due to sediment erosion and deposition were taken into account during the simulation by updating the elevation of the bed in each grid cell after every calculation step (Deltares 2014, pp. 366–369). 3 RESULTS 3.1 Distribution of tsunami deposits alongshore the Ambrakian Gulf The analysis and evaluation of all 151 vibracores revealed a consistent geographical distribution of tsunami deposits alongshore the Ambrakian Gulf. The average tsunami candidate layer thickness, which ranges from 0 m to 3.22 m (Fig. 5), generally decreases from west to east. Whereas the western Ambrakian Gulf shows the maximum values (frequently > 0.76 m), there are no values larger than 0.76 m in the inner and eastern gulf. Consequently, the average sediment accumulation per single tsunami event becomes smaller in an eastward direction, which points to a decreasing tsunami influence in the Ambrakian Gulf from the coast to regions further inland. This agrees with a general decrease of the relative number of event strata, that is, the number of tsunami candidate layers per coring metre per vibracore, from west to east (not illustrated). High values (> 0.29) are found around the entrance of the Ambrakian Gulf, including the Mazoma Lagoon, Aktio Headland, the Bay of Aghios Nikolaos and Lake Voulkaria, but significantly drop in the inner gulf with maximum 0.14 at its eastern shore. Figure 5. View largeDownload slide Average thickness of tsunami candidate layers per vibracore alongshore the Ambrakian Gulf. The identification of tsunami deposits in all 151 vibracores is based on several physical sedimentary criteria typical of tsunamites (Section 2.2 and Fig. 3). Furthermore, geochemical, microfaunal and palynological analyses conducted for selected vibracores in previous studies helped to detect event layers. The average tsunami layer thickness is compared with the computed sedimentation in the numerical model in Section 3.3. Figure 5. View largeDownload slide Average thickness of tsunami candidate layers per vibracore alongshore the Ambrakian Gulf. The identification of tsunami deposits in all 151 vibracores is based on several physical sedimentary criteria typical of tsunamites (Section 2.2 and Fig. 3). Furthermore, geochemical, microfaunal and palynological analyses conducted for selected vibracores in previous studies helped to detect event layers. The average tsunami layer thickness is compared with the computed sedimentation in the numerical model in Section 3.3. Fig. 6 shows the standard deviation of the tsunami layer thickness per vibracore in the study area. As can be seen, the standard deviation is generally small (<0.16 m), indicating a relatively persistent thickness of tsunami layers per vibracore. Therefore and because there are many vibracores with just one tsunami layer, the average tsunami layer thickness per vibracore (associated with several tsunami events in pre-/history) is an appropriate parameter to be compared with the sedimentation and erosion computed for single tsunami waves in the numerical model. Based on this comparison (Section 3.3), it is possible to draw conclusions about the general tsunami wave characteristics which are related to deposits in certain parts of the study area. Figure 6. View largeDownload slide Standard deviation of the tsunami candidate layer thickness per vibracore alongshore the Ambrakian Gulf. Figure 6. View largeDownload slide Standard deviation of the tsunami candidate layer thickness per vibracore alongshore the Ambrakian Gulf. 3.2 General inundation dynamics in the Ambrakian Gulf Prior to the comparison of field data and simulated morphodynamics, a short summary of the basic computed tsunami inundation dynamics in the Ambrakian Gulf of all scenarios is given. Fig. 7 exemplarily illustrates the inundation in Scenario W4L, which shows characteristic hydrodynamic features of both, scenarios with smaller and larger wave amplitudes/periods. Due to the leading trough of the simulated LDN-waves, in each scenario, a significant withdrawal of the shoreline between Kastrosikiá and Lefkada City is observed before the wave crest makes landfall (Fig. 7a; see Fig. 2 for places). Independent from the chosen boundary, all wave crests strike the Ambrakian Gulf from an almost western direction (Fig. 7b). Whereas this is the expected direction for waves from the west, waves from the southwest and south are strongly diffracted and refracted around Lefkada Island. Once the water depth drops below 1000 m, strong wave shoaling occurs, which, together with funnel effects in the inlet-shaped Ionian Sea towards the Ambrakian Gulf (Fig. 1), causes a significant increase in wave amplitude by up to factor three (Fig. 7b). This results in maximum run-up heights of more than 25 m in the valleys between Kastrosikiá and Mitikas. Figure 7. View largeDownload slide Snapshots (a–d) documenting stepwise the inundation of the Ambrakian Gulf in Scenario W4L and diagrams (top) showing the corresponding time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points. Figure 7. View largeDownload slide Snapshots (a–d) documenting stepwise the inundation of the Ambrakian Gulf in Scenario W4L and diagrams (top) showing the corresponding time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points. In all scenarios, tsunami waves completely inundate Aktio Headland (ground surface maximum 6 m a.s.l.) and subsequently enter the Ambrakian Gulf through the main waterway, the Strait of Preveza (Figs 7b and c). Simultaneously, tsunami waters in the moderate and extreme scenarios (wave amplitudes of 4 m and 6 m) reach the gulf via a small depression (ground surface maximum 14 m a.s.l.) to the west of the Mazoma Lagoon. Furthermore, each tsunami wave gets strongly funnelled in the Bay of Aghios Nikolaos and strikes Lake Voulkaria. Once the waves have reached the inner gulf, the area around Vonitsa and the narrow sandspit systems as well as the wide coastal flats in the northern gulf get inundated (Figs 7c and d). Due to flow divergence, the eastern gulf is hardly affected by tsunami waves; considerable flooding only occurs in the moderate and extreme scenarios in the Boukka coastal plain. Even after a simulation time of 10 h no significant backflow can be observed. Instead, tsunami waters gradually flow back into the Ionian Sea mainly via the Strait of Preveza. The diagrams at the top of Fig. 7 display the time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points in Scenario W4L. The most offshore observation point OBS ION clearly demonstrates the drop and subsequent rise in water level during the passage of the LDN-wave. The wave crest is associated with significant offshore currents in the order of 2.4 m s−1. The onshore observations points OBS AKT and OBS ANI 10 indicate inundation depths of up to 5 m connected with depth averaged flow velocities of almost 7 m s−1. Whereas Aktio Headland (OBS AKT) quickly falls dry after landfall of the wave crest, the inundation depth around Lake Voulkaria (OBS ANI 10) successively increases due to ongoing water inflow. Finally, observation point OBS AMB shows that the tsunami wave has a strong impact even on the western inner gulf, where the water elevation rises up to 3.4 m and again strong currents (up to 3.5 m s−1) can be observed. Generally, tsunamis from the west cause the strongest inundation of the Ambrakian Gulf, whereas waves from the south have the weakest impact (here and in the following measured by the inundation depth, flow velocity and area of inundation). Tsunamis originating from the southwestern model boundary result in an intermediate onshore response. Besides the propagation direction, the onshore tsunami response strongly depends on the wave period and wave amplitude; for both wave parameters, a positive correlation with the inundation intensity is found. As a result, Scenario W6L shows the strongest onshore response in the Ambrakian Gulf, whereas Scenario S2S shows the weakest impact. 3.3 Average vibracore tsunami layer thickness versus simulation results In the following, the simulated tsunami morphodynamics are described and compared with the average tsunami candidate layer thickness presented in Section 3.1. Maps of the Ambrakian Gulf show the simulated sediment erosion (blue colours) and accumulation (yellow to red colours) associated with tsunami inundation after a simulation time of 8 h to 10 h in each scenario (Figs 8–12; note the varying scale bar). This period covers most of the inundation process, including the main backflow. The maps further contain information on the average thickness of tsunami layers per vibracore already illustrated in Fig. 5. As pointed out in Section 3.1, this parameter allows a direct comparison with the sedimentary impact simulated for different single tsunami waves. Due to the large number of computed scenarios (18), only those scenarios are presented that show the best match with field data, namely Scenario W2S, S4S, W4S (not illustrated), W4L, S6S and W6L. Figure 8. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W2S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W2S is based on a single tsunami wave from a western direction with a wave amplitude of 2 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 8. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W2S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W2S is based on a single tsunami wave from a western direction with a wave amplitude of 2 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 9. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S4S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S4S is based on a single tsunami wave from a southern direction with a wave amplitude of 4 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 9. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S4S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S4S is based on a single tsunami wave from a southern direction with a wave amplitude of 4 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 10. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W4L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W4L is based on a single tsunami wave from a western direction with a wave amplitude of 4 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. See Fig. 7 for the simulated hydrodynamics of this scenario. Figure 10. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W4L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W4L is based on a single tsunami wave from a western direction with a wave amplitude of 4 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. See Fig. 7 for the simulated hydrodynamics of this scenario. Figure 11. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S6S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S6S is based on a single tsunami wave from a southern direction with a wave amplitude of 6 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 11. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S6S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S6S is based on a single tsunami wave from a southern direction with a wave amplitude of 6 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 12. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W6L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W6L is based on a single tsunami wave from a western direction with a wave amplitude of 6 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 12. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W6L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W6L is based on a single tsunami wave from a western direction with a wave amplitude of 6 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. In all scenarios based on a low wave amplitude of 2 m, the sedimentary onshore response of tsunami waves is mainly restricted to Aktio Headland, the Bay of Aghios Nikolaos and the entrance of Lake Voulkaria. Only in Scenario W2L, significant sediment erosion and accumulation is also observed in the inner Lake Voulkaria and to the east of Aktio Headland. The best agreement with vibracore data of the Ambrakian Gulf is found for the simulated western tsunami with a wave period of 30 min (Scenario W2S; Fig. 8). Although the absolute sedimentation computed in this scenario (≤ 0.5 m) clearly lags behind the local average tsunami layer thickness per vibracore (≤ 2.67 m), both show the same spatial extent. Areas of simulated sediment erosion or no sediment accumulation in northwestern and northeastern Aktio Headland as well as to the north of Preveza coincide with a lack of event layers in the local vibracores. Sediment accumulation in the numerical model is mostly found at the northern tip and western shore of Aktio Headland, around the Saltini Lagoon, along the Bay of Aghios Andreas and at the entrance of Lake Voulkaria; for these areas also vibracore data indicates dominant tsunami sediment deposition. Regarding the offshore tsunami morphodynamics, in all scenarios based on a 2 m wave amplitude, significant sediment processes are not observed before the water depth drops below 20 m. Consequently, most of the deposited tsunami sediments originate from the shallow marine to littoral zone landward from the −20 m contour. Although Scenario W2S is found in close accordance with field data of Aktio Headland and the Bay of Aghios Nikolaos, Scenario S4S is the favourite scenario with regard to the entire western part of the Ambrakian Gulf, including the Mazoma Lagoon (Fig. 9). Again, vibracores containing no tsunami candidate layers are generally located in areas that show dominant erosion or no accumulation in the numerical model, whereas vibracores with thick event strata are mostly found in areas of simulated sediment accumulation. Although still too small compared to the average layer thickness determined per vibracore, the simulated thickness of tsunami deposits in Scenario S4S has significantly increased (up to 1 m). The morphodynamics computed for Scenario W4L (Fig. 10) are far stronger compared to the scenarios described above. Whereas the western part of the Ambrakian Gulf (except for the Saltini Lagoon and the Bay of Aghios Nikolaos) mostly show sediment erosion, significant sedimentation is particularly observed to the east of the Mazoma Lagoon and Aktio Headland, in the inner gulf and along the eastern shore of Lake Voulkaria. For the latter, the computed sediment thickness in the numerical model (up to 1.10 m) accurately replicates the average event layer thickness determined for the corresponding vibracores. Despite the clear sedimentary impact on the inner gulf, Scenario W4L is not appropriate to explain vibracore data around Vonitsa and the sandspit systems in the north. Here, the numerical model either suggests no significant sedimentation or even dominant erosion respectively. Strong morphodynamics in Scenario W4L are also observed offshore the Ambrakian Gulf, where sediment processes already start landwards from the −50 m contour. Scenario S6S (Fig. 11) shows similar areas of consistent model and field data as Scenario S4S but, beyond this, also explains tsunami sedimentation in the midst of Lake Voulkaria. However, dominant sediment erosion occurs along southwestern Aktio Headland in Scenario S6S, which is in contrast to relatively thick event layers found in the local vibracores. The simulated sediment thickness (up to 1.5 m) nearly matches the average tsunami candidate layer thickness per vibracore, where both, simulation results and vibracores indicate significant sediment accumulation. In general, results for Scenario S6S closely correspond to those for Scenario W4S, which is therefore not illustrated here. Sediment processes in Scenario W6L are on an unprecedented scale starting already landwards from the −60 m contour and reaching far into the inner and even eastern Ambrakian Gulf (Fig. 12). Comparable to Scenario W4L, in Scenario W6L the western part of the Ambrakian Gulf is mostly characterized by sediment erosion, which is generally not consistent with vibracore data. Instead, a good accordance between simulated tsunami morphodynamics and field traces is found for the eastern and southern shore of Lake Voulkaria and for the inner Ambrakian Gulf around Vonitsa. For the latter, Scenario W6L is the only scenario that plausibly explains tsunami event strata detected in the field. Nevertheless, simulation results again do not match vibracore data of the sandspit systems, the delta plain in the north and the Boukka coastal plain at the very eastern end of the gulf (the latter two locations are not displayed in Figs 8–12; see Fig. 5). 3.4 Correlating vibracore tsunami layers versus simulation results Fig. 13 shows the location of three vibracore transects, for which correlating tsunami candidate layers were identified based on geochronological data (in particular radiocarbon ages). These layers point to three of several pre-/historical tsunami events in the Ambrakian Gulf summarized in Section 1. Vibracore transect A, located in southwestern Aktio Headland, revealed correlating tsunami deposits associated with the 840 cal AD tsunami. Event layers associated with the 1000 cal BC tsunami were found in vibracore transect B in the area of Lake Voulkaria. The third tsunami event, which occurred in the 4th century AD, is documented by corresponding deposits in vibracore transect C at the southern shore of Lake Voulkaria (for a detailed sedimentary analysis of the three vibracore transects see Vött et al.2007 (transect A), Vött et al.2006 (transect B), and Vött et al.2009a (transect C)). Section 3.3 demonstrated that for all three areas, some of the simulated scenarios show a good match with the average tsunami layer thickness. These scenarios are also appropriate to explain the deposition of the correlating tsunami layers identified in the three vibracore transects. Figure 13. View largeDownload slide Location of vibracore transects A, B and C, for which the thickness and sedimentary composition of correlating tsunami candidate layers are compared with the simulated tsunami layers in different scenarios (Figs 14–19). Figure 13. View largeDownload slide Location of vibracore transects A, B and C, for which the thickness and sedimentary composition of correlating tsunami candidate layers are compared with the simulated tsunami layers in different scenarios (Figs 14–19). In the following, both the thickness and sedimentary composition (portions of gravel to mud) of the correlating vibracore tsunami layers of each tsunami event are directly compared with the simulated tsunami layers and erosion in the scenarios. Although several scenarios presented in Section 3.3 are appropriate for this comparison, only those with the closest match are shown. In Figs 14, 16 and 18, for each vibracore of the three transects, the corresponding tsunami layer is displayed as a bar, the height of which gives the layer thickness and the white to black coloured segments, sorted by grain size, the portions of gravel, coarse sand, middle sand, fine sand and mud. Accordingly, the graphs in the background illustrate the thickness and sedimentary composition of the simulated tsunami layers as well as the simulated tsunami erosion in the best matching scenarios. In addition, scatter diagrams in Figs 15, 17 and 19 show the portions of grain sizes in the vibracore tsunami layers of each transect versus the portions of grain sizes in the tsunami layers as simulated in the scenarios. The closer the points lay on the grey dashed line, the higher the correlation between the tsunami sediment composition in the vibracores and in the numerical model. Figure 14. View largeDownload slide Comparison of event deposits of the 840 cal AD tsunami in vibracore transect A in southwestern Aktio Headland (Fig. 13; Vött et al.2007) with the tsunami layers simulated in Scenario S4S (top), W4L (centre) and W6L (bottom). Tsunami deposits in vibracores ANI 2–AKT 11 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 14. View largeDownload slide Comparison of event deposits of the 840 cal AD tsunami in vibracore transect A in southwestern Aktio Headland (Fig. 13; Vött et al.2007) with the tsunami layers simulated in Scenario S4S (top), W4L (centre) and W6L (bottom). Tsunami deposits in vibracores ANI 2–AKT 11 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 15. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect A (Fig. 14) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S, W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 15. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect A (Fig. 14) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S, W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Vibracore transect A (ANI 2–AKT 11) stretches for almost 3000 m in a northwestern direction across relatively flat terrain in southwestern Aktio Headland (Figs 13 and 14). Tsunami layers of the 840 cal AD event are between 0.13m and 2.33m thick and generally become thinner towards the northwest. They mainly consist of middle and fine sand. Considerable amounts of mud and gravel only occur in vibracores ANI 2 and AKT 3 respectively. Tsunami layers in transect A are best represented by Scenarios S4S, W4L and W6L. In particular Scenarios W4L and W6L show a very similar spatial distribution of areas with dominant and limited deposition as the vibracore transect. However, the model does not reproduce the muddy tsunami deposits found in ANI 2 but instead indicates erosion. All scenarios, especially Scenarios S4S and W4L, underpredict the absolute tsunami layer thickness determined in the vibracores. Despite this underprediction, the model reproduces well the sedimentary composition of the identified tsunami layer in transect A. This is demonstrated by Fig. 15 showing similar portions of the five grain sizes for all coring locations except for vibracore ANI 2 in all scenarios and for vibracores AKT 2–AKT 11 in Scenario S4S. Generally, both the vibracores and the numerical model indicate dominant fine sand in the first part of the transect, a mix of fine and middle sand in the central part followed by dominant coarse material between 2000 m and 2500 m and again finer material at the end of the transect. Vibracore transect B (ANI 4–VOUL 1) is about 4000 m long and reaches from the village of Aghios Nikolaos to the inner Lake Voulkaria (Figs 13 and 16). The highest point (4.5 m a.s.l.) lies on a small rise between 500 m and 1000 m of the transect at the seaward entrance of Lake Voulkaria. The 1000 cal BC tsunami layer has a thickness of between 0.19 m and 2.07 m and shows a relatively heterogeneous sedimentary composition. Both, the vibracores and the numerical model (Scenarios S4S and S6S) indicate dominant tsunami deposition at the seaward and landward foot of the small rise between 500 m and 1000 m, which seem to act as a natural obstacle during inundation. The rise itself is strongly eroded in both scenarios. All in all, the first 1500 m of the transect are better represented by Scenario S4S than by Scenario S6S. The opposite is true for the area around vibracore VOUL 1 in the inner Lake Voulkaria, where only Scenario S6S can explain significant tsunami sedimentation. As was discussed earlier for transect A, also for transect B the model underpredicts the tsunami layer thickness observed in the field but at the same time shows a good match regarding the sedimentary composition. According to Fig. 17, the simulated and the observed tsunami layer is characterized by a heterogeneous composition in the first 200 m, dominant fine sand around the small rise between 500 m and 1000 m, followed by coarser material at ANI 7 and ANI 10 and finally fine sand and mud in the inner Lake Voulkaria. Figure 16. View largeDownload slide Comparison of event deposits of the 1000 cal BC tsunami in vibracore transect B in the area of Lake Voulkaria (Fig. 13; Vött et al.2006) with the tsunami layers simulated in Scenarios S4S (top) and S6S (bottom). Tsunami deposits in vibracores ANI 4–VOUL 1 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 16. View largeDownload slide Comparison of event deposits of the 1000 cal BC tsunami in vibracore transect B in the area of Lake Voulkaria (Fig. 13; Vött et al.2006) with the tsunami layers simulated in Scenarios S4S (top) and S6S (bottom). Tsunami deposits in vibracores ANI 4–VOUL 1 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 17. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect B (Fig. 16) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S and S6S. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 17. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect B (Fig. 16) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S and S6S. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Vibracore transect C (PAL 6–PAL 46) gives evidence of the 4th century AD tsunami at the southern shore of Lake Voulkaria (Figs 13 and 18). Over a distance of about 700 m, tsunami deposits 0.16–0.82 m thick can be observed in the field. In the numerical model, Scenario W4L and especially Scenario W6L predict significant tsunami sedimentation for this transect, although the absolute thickness found in the vibracores is not reached in any of the scenarios. Nevertheless, the field data and the scenarios show a corresponding sedimentary composition of the tsunami deposits, which are dominated by fine sand and mud (Fig. 19). Figure 18. View largeDownload slide Comparison of event deposits of the 4th century AD tsunami in vibracore transect C at the southern shore of Lake Voulkaria (Fig. 13; Vött et al.2009a) with the tsunami layers simulated in Scenarios W4L (left) and W6L (right). Tsunami deposits in vibracores PAL 6–PAL 46 are displayed as bars, the simulated tsunami sedimentation as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 18. View largeDownload slide Comparison of event deposits of the 4th century AD tsunami in vibracore transect C at the southern shore of Lake Voulkaria (Fig. 13; Vött et al.2009a) with the tsunami layers simulated in Scenarios W4L (left) and W6L (right). Tsunami deposits in vibracores PAL 6–PAL 46 are displayed as bars, the simulated tsunami sedimentation as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 19. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect C (Fig. 18) versus the portions of grain sizes in the tsunami layers simulated in Scenarios W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 19. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect C (Fig. 18) versus the portions of grain sizes in the tsunami layers simulated in Scenarios W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. 4 DISCUSSION Based on the presented simulation results, a number of systematic hydro- and morphodynamic features of tsunami inundation in the Ambrakian Gulf can be drawn together: Tsunamis from a western direction show the strongest hydro- and morphodynamic impact on the Ambrakian Gulf. This is closely related to strong funnelling of tsunami waters in the eastern Ionian Sea between the islands of Cephalonia, Lefkada and Corfu for waves from the west (Fig. 1). In contrast, funnel effects are less distinct for waves from a southwestern and especially southern direction (cf. Vött et al.2008, pp. 107, 119). Furthermore, the central Ionian Islands, in particular Lefkada, Cephalonia and Zakynthos, seem to give some protection for the Ambrakian Gulf against waves from these two directions. The sedimentary response of all shorter period tsunami waves (T = 30 min) is concentrated on a comparatively narrow area mainly covering the western part of the Ambrakian Gulf (Figs 8, 9 and 11). These waves result in strong sedimentation especially between the Mazoma Lagoon and Lake Voulkaria. By contrast, for all longer period waves (T = 80 min), significant sediment processes are observed for a much wider area stretching from the nearshore zone (landwards from the −60 m contour) up to the eastern shore of the Ambrakian Gulf (Figs 10 and 12). Associated with the larger reach of the longer period tsunami waves, here, the western part of gulf, including the area of the Mazoma Lagoon, large parts of Aktio Headland and the entrance of Lake Voulkaria, is dominated by erosion and the main sedimentation is found further east or clearly offshore in the Ionian Sea. Nevertheless, maximum values of erosion and sedimentation are only slightly larger for the longer period than for the shorter period waves, the same wave amplitude provided. Simulated sediment accumulation—where not influenced by natural obstacles (e.g. cliffs or hills)—is clearly characterized by thinning- and fining-landward, features typical of tsunami deposits (Section 2.2). This is due to the general decrease of inundation depth and flow velocity (Section 3.2) and by this of the transport capacity of inundating tsunami waters. Thinning- and fining-landward is most obvious in Scenarios W4L (Fig. 10) and W6L (Fig. 12) along the eastern shore of Lake Voulkaria (see also transects B and C in Figs 16 and 18) and in Scenario W6L also in the area of Vonitsa. The main sedimentary response of tsunamis in the Ambrakian Gulf arises from the wave run-up and not from the backflow. This is connected to the extensive flat flooding zone in the northern Ambrakian Gulf and therefore the gentle slope of the water surface, which prevents high flow velocities during the backflow. Generally, the numerical model accurately reflects the decreasing tsunami influence from the western to the eastern Ambrakian Gulf, as is indicated by the sedimentary field data (Section 3.1). However, there is no specific simulated scenario which can totally explain the spatial distribution of the average tsunami candidate layer thickness per vibracore in the entire study area at the same time. This is because the vibracores contain tsunami layers that point to different tsunami events in the past, which were probably related to various sources and divergent sedimentary responses (Vött et al.2006, 2007, 2008, 2009a, 2010, 2011; May et al.2007, 2012a; Section 1). Therefore, the average layer thickness used for the comparison with simulation results in Section 3.3 does not reflect the actual sediment deposition due to a particular tsunami event. Nevertheless, there are several scenarios that reproduce well the average tsunami layer thickness in certain parts of the study area. Generally, the scenarios based on shorter period tsunami waves (T = 30 min) from the south or west with any wave amplitude agree with the average layer thickness found in the western Ambrakian Gulf, including the area of the Mazoma Lagoon, large parts of Aktio Headland, the Bay of Aghios Andreas and the entrance to Lake Voulkaria (Figs 8, 9 and 11). In contrast, longer period waves (T = 80 min) with amplitudes of 4 m and 6 m and from a western direction, are the favourite tsunami scenarios in respect to the average layer thickness in southwestern Aktio Headland, around Lake Voulkaria and in the area of Vonitsa located in the inner gulf (Figs 10 and 12). The sedimentary response of tsunami waves from the southwest, however, generally does not match the spatial distribution of the average tsunami layer thickness detected in the field. Those scenarios that show a good fit with the average tsunami layer thickness in the area of the three vibracore transects presented in Section 3.4 do also agree with the single, correlating tsunami layers identified in these transects. Consequently, different scenarios might be appropriate to explain the deposition of the correlating layer in each transect. Moreover, some scenarios even agree with the correlating tsunami layers of two different transects, although these layers originate from different events in the past. The marine to shallow marine origin of detected tsunami deposits in the study area can be explained by all simulated scenarios, which show that most sediment in the numerical model originate from the shallow marine to littoral zone landward from the −60 m contour (cf. e.g. Morton et al.2007, pp. 187–194; Apotsos et al.2011b, p. 8; Apotsos et al.2011c, p. 9; Sugawara et al.2014a, p. 299; Sugawara et al.2014b, p. 19). Generally, the numerical model accurately reproduces the sedimentary composition of the detected tsunami deposits as is demonstrated for the correlating tsunami layers in the three transects in the western Ambrakian Gulf in Section 3.4. This also includes thinning- and fining-landward tendencies frequently observed in related tsunami event strata at adjacent vibracoring sites (see especially transects B and C in Figs 16 and 18). Besides these common features of simulation results and field traces, the numerical model also reproduces local tsunami inundation dynamics for particular key sites in the western Ambrakian Gulf which were reconstructed on the basis of vibracore data in previous studies: According to Vött et al. (2008, p. 120), the central and northern part of the former Plaka sandspit system, located to the southwest of the Bay of Aghios Nikolaos (Fig. 2), was completely destroyed by tsunami waves over a distance of more than 5 km. This closely agrees with all simulated scenarios showing dominant erosion in this area. Simulated morphodynamics corroborate strong local erosional forces during tsunami landfall in the western, northern and eastern part of Aktio Headland, as documented by distinct erosional contacts in vibracores (Vött et al.2007, p. 54). Vött et al. (2007; 2011, pp. 236, 237) suggest that tsunami waters struck the Bay of Aghios Nikolaos, inundated Lake Voulkaria and subsequently flooded the Bay of Palairos located almost 5 km to the south (Fig. 2). Inundation of Lake Voulkaria from a western direction is reproduced by all simulated scenarios. Scenarios W4L and W6L (Figs 10 and 12) further show inundation and significant morphodynamics in the adjacent Bay of Palairos. According to Vött et al. (2009a, p. 30), the so-called Cleopatra Canal—connecting the Bay of Aghios Nikolaos with Lake Voulkaria—was originally dug out by tsunami waves and was filled again during a subsequent tsunami event. Erosion of the canal can easily be explained by Scenarios W4S, W4L, S6S and W6L (Figs 10–12), whereas the sediment infill might be reflected by Scenarios W2S and S4S (Figs 8 and 9). With some exceptions, it is rather the spatial extent of deposits that agree with vibracore data than the absolute thickness of simulated sedimentation. The latter usually lags behind the average tsunami candidate layer thickness determined per vibracore and the thickness of the correlating tsunami layers in the three transects. This may particularly be related to the following reasons: If the upper part of an event layer has been eroded and subsequently covered by younger tsunami deposits, the complete sequence may be interpreted as only one tsunami candidate layer resulting in an overestimation of the average layer thickness/thickness of the correlating tsunami layers in the transects. This may further explain the exceptionally large thicknesses more than 1 m found for some vibracores especially in the western part of the Ambrakian Gulf. The absolute thickness of simulated sedimentation is affected by various parameters in the numerical model, including the predetermined bottom roughness, values of the critical bed shear stress, sediment stratigraphy as well as the small-scale topography (cf. Apotsos et al.2009; Apotsos et al.2011b, pp. 4, 16; Apotsos et al.2011c, p. 9; Li et al.2012a, 2014). For example, a general overestimation of the critical bed shear stress would result in less sediment entrainment by waves, and consequently in less sediment deposition in the numerical model than observed in reality. In contrast to the absolute thickness, the spatial extent of tsunami deposits in both, the numerical model and the field is more resistant to modifications. This is because even a very thin event strata would still be indicative of tsunami influence. Therefore, the spatial extent of tsunamites is generally more appropriate for the comparison of simulation results and field data than the absolute thickness (cf. Li et al.2014, p. 2272). Erosion particularly due to subsequent tsunami impacts does not only strongly affect the absolute thickness but also the general spatial distribution of tsunami deposits in the field (e.g. Wheatcroft 1990; Wheatcroft & Drake 2003; Szczuciński 2012; Spiske et al.2013). In extreme cases, tsunami event layers can be completely removed from the geological record. This is an important process to be considered in this study because it might contribute to the inconsistency between the simulation result and field data. Especially the area of the Mazoma Lagoon, Aktio Headland and the entrance of Lake Voulkaria are susceptible to strong tsunami erosion as is indicated by several simulated scenarios and distinct erosional unconformities in the local vibracores (cf. Vött et al.2007, p. 54). All simulated scenarios are based on single tsunami waves in order to investigate the morphodynamic effects of different wave periods, wave amplitudes and propagation directions. However, most tsunamis appear in the form of wave trains consisting of a series of waves. Subsequent waves during one and the same tsunami event may redeposit sediment that was released by previous waves (cf. e.g. Apotsos et al.2011b, p. 15; Sugawara et al.2014a, pp. 298, 299), which would result in a sediment distribution not replicated by any of the simulated scenarios in this study. Given the wide time span of detected tsunami impacts and associated deposits in the Ambrakian Gulf of about 8000 yr, long-term changes in the local topography as described in Section 1 may further account for discrepancies between simulation results and field data (cf. Röbke et al.2015, 2016). This is particularly true for the northern and eastern shores of the Ambrakian Gulf, which experienced major topographical changes due to tectonic and fluvio-deltaic processes. Here, no scenarios can explain tsunami deposits detected in the field. Finally, the comparison of morphodynamic simulation results with sedimentary field data of the study area is limited by the spatial resolution of the numerical model. Whereas the simulations yield the average sedimentation for 90 m by 90 m grid cells, vibracores represent the geological record for a specific coring site only. This may result in further inconsistency between model and field (cf. Sugawara et al.2014b, p. 28). 5 CONCLUSIONS According to the general decrease of tsunami wave energy during the landfall, sedimentary vibracore data indicates a decreasing tsunami influence from the western to the eastern Ambrakian Gulf for both, the frequency and the intensity of events in the last 8000 yr. This is accurately reflected by the hydro- and morphodynamic numerical model showing that the sedimentary tsunami response decreases from the west to the east and that the inner and eastern gulf is only affected in extreme events. As indicated by the field data itself, several different tsunami wave scenarios are needed to explain detected tsunami deposits in different parts of the study area. Generally, shorter period tsunami waves (T = 30 min) from the south or west with any wave amplitude (2 m, 4 m, 6 m) are highly consistent with field data of the western Ambrakian Gulf. In contrast, longer period tsunamis (T = 80 min) from a western direction with 4 m to 6 m amplitudes show the best agreement with vibracore data collected in southwestern Aktio Headland and in the more central parts of the Ambrakian Gulf including Lake Voulkaria. Based on the comparison of the simulation results with the average tsunami layers thickness and with the thickness and sedimentary composition of correlating tsunami layers in three vibracore transects, 4 of total 18 scenarios could be derived that reproduce well the tsunami deposits found in southwestern Aktio Headland and in the area of Lake Voulkaria. These are Scenarios S4S, W4L, S6S and W6L, which all represent relatively strong tsunami events. Simulated tsunamis from the southwest generally do not match tsunami field traces. Therefore and because most potential tsunami sources in the central and eastern Mediterranean are rather located to the west (Mount Etna, Calabrian Arc) and south or southeast (Hellenic Trench) of the Ambrakian Gulf, it can be assumed that no major tsunamis from a southwestern direction struck the area in the past. The fact, that tsunami field traces in the eastern and northern parts of the Ambrakian Gulf cannot be explained by any of the simulated scenarios further leads to the conclusion that these deposits might relate to tsunamis generated along normal faults located in the gulf. Moreover, it has to be considered that the eastern and northern gulf experienced major topographical changes during the Holocene. The simulated scenarios do not only replicate the general spatial distribution of the tsunami candidate layers; they further show a number of essential sedimentary features typical of tsunamites, which were also detected in many vibracores of the study area. Such features are thinning- and fining-landward as well as the marine to shallow marine origin of tsunami deposits. The excellent match between the sedimentary composition of the simulated tsunami layers and the correlating tsunami layers in the three vibracore transects particularly results from the realistic initial bed composition of the model, which was derived from actual stratigraphic data of the study area. Finally, the numerical model accurately reproduces local tsunami inundation dynamics which were derived from field data for particular key sites in the western Ambrakian Gulf by previous sedimentary studies. Nevertheless, there are some discrepancies between simulation results and field traces. In particular, the computed sediment thickness in the numerical model usually lags behind the average thickness of tsunami candidate layers per vibracore and the thickness of the correlating tsunami layers in the transects. Also, some field traces cannot be explained by any of the simulated scenarios. Such inconsistencies may arise from (i) the comparison of single simulated tsunami waves with the average layer thickness per vibracore, (ii) tsunami candidate layers that give the impression of being related to a single tsunami event but actually relate to two or more events, (iii) false assumptions about spatially highly variable hydro- and morphodynamic parameters in the numerical model (e.g. bottom roughness or critical bed shear stress), (iv) hydrodynamic boundary conditions that do not represent all possible/realistic tsunami waves in the study area (e.g. single tsunami waves instead of tsunami wave trains), (v) changes in the local topography during the last 8000 yr, which are not considered in the numerical model and finally from (vi) the limited spatial resolution of the model. Especially inconsistencies arising from (i) and (v) could be reduced in future studies, provided that the age model of correlating tsunami candidate layers can be extended (cf. Röbke et al.2015, 2016). Then, event strata of particular pre-/historical tsunami impacts could be compared on a large scale with numerical simulations that consider the contemporary topography of the study area. Moreover, an improved age model provided and based on the insights into the wave characteristics gained in this study, specific source mechanisms (e.g. landslides offshore Mount Etna or specific focal mechanisms along the Helenic Trench) could be considered in future investigations. As demonstrated in this study, comparing and calibrating numerical models with sedimentary field data and computing sediment processes significantly increase the quality of numerical tsunami simulation results as a basis for modern tsunami risk assessment. Given the frequency of tsunami impacts in the past as indicated by sedimentary field data and historical accounts and considering the computed tsunami scenarios, a high tsunami risk has to be derived for the Ambrakian Gulf. Acknowledgements This research was supported by the Studienstiftung des deutschen Volkes (Bonn) in the form of a doctoral scholarship. We further acknowledge financial support from the German Research Foundation (Bonn) under grant numbers VO 938/1-1, VO 938/1-3, VO 938/2-1 and VO 938/3-1. Special thanks are due to Peter Fischer and Dieter Kelletat who provided insight and expertise that greatly assisted this research. REFERENCES Ad Hoc-AG Boden, 2005. Bodenkundliche Kartieranleitung , 5th edn, Schweizerbart Science Publishers. Apotsos A., Jaffe B., Gelfenbaum G., Elias E., 2009. Modeling time-varying tsunami sediment deposition, in Proceedings of Coastal Dynamics, Tokyo, 2009 , pp. 1– 15, doi:10.1142/9789814282475_0037. Apotsos A., Buckley M., Gelfenbaum G., Jaffe B., Vatvani D., 2011a. Nearshore tsunami inundation model validation: toward sediment transport applications, Pure appl. Geophys. , 168, 2097– 2119. https://doi.org/10.1007/s00024-011-0291-5 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., 2011b. Process-based modeling of tsunami inundation and sediment transport, J. geophys. Res. , 116, 1– 20. https://doi.org/10.1029/2010JF001797 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., Watt S., Peck B., Buckley M., Stevens A., 2011c. Tsunami inundation and sediment transport in a sediment-limited embayment on American Samoa, Earth-Sci. Rev. , 107, 1– 11. https://doi.org/10.1016/j.earscirev.2010.11.001 Google Scholar CrossRef Search ADS   Apotsos A., Jaffe B., Gelfenbaum G., 2011d. Wave characteristic and morphologic effects on the onshore hydrodynamic response of tsunamis, Coastal Eng. , 58, 1034– 1048. https://doi.org/10.1016/j.coastaleng.2011.06.002 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., 2012. Time-dependent onshore tsunami response, Coastal Eng. , 64, 73– 86. https://doi.org/10.1016/j.coastaleng.2012.01.001 Google Scholar CrossRef Search ADS   Benetatos C., Kiratzi A., Papazachos B., Karakaisis G., 2004. Focal mechanisms of shallow and intermediate depth earthquakes along the Hellenic arc, J. Geodyn. , 37, 253– 296. https://doi.org/10.1016/j.jog.2004.02.002 Google Scholar CrossRef Search ADS   Bondevik S., Løvholt F., Harbitz C., Mangerud J., Dawson A., Svendsen J., 2005. The Storegga slide tsunami—comparing field observations with numerical simulations, Mar. Pet. Geol. , 22, 195– 208. https://doi.org/10.1016/j.marpetgeo.2004.10.003 Google Scholar CrossRef Search ADS   Borrero J., Weiß R., Okal E., Hidayat R., Suranto Arcas D., Titov V., 2009. The tsunami of 2007 September 12, Bengkulu province, Sumatra, Indonesia: post-tsunami field survey and numerical modelling, Geophys. J. Int. , 178, 180– 194. https://doi.org/10.1111/j.1365-246X.2008.04058.x Google Scholar CrossRef Search ADS   Brockmüller S., Vött A., Brückner H., Handl M., Schriever A., 2005. Die holozäne entwicklung der küstenebene von boukka am ambrakischen golf (nordwestgriechenland), in Schriften des Arbeitskreises Landes- und Volkskunde , Vol. 4. Bryant E., 2008. Tsunami: The Underrated Hazard , 2nd edn, Springer. Chacón-Barrantes S., Narayanan R., Mayerle R., 2013. Several tsunami scenarios at the North Sea and their consequences at the German Bight, J. Tsunami Soc. Int. , 32, 8– 28. Clews J., 1989. Structural controls on basin evolution: Neogene to Quaternary of the Ionian zone, western Greece, J. Geol. Soc. , 146, 447– 457. https://doi.org/10.1144/gsjgs.146.3.0447 Google Scholar CrossRef Search ADS   Dawson A., 1994. Geomorphological effects of tsunami run-up and backwash, Geomorphology , 10, 83– 94. https://doi.org/10.1016/0169-555X(94)90009-4 Google Scholar CrossRef Search ADS   Dawson A., 1999. Linking tsunami deposits, submarine slides and offshore earthquakes, Quat. Int. , 60, 119– 126. https://doi.org/10.1016/S1040-6182(99)00011-7 Google Scholar CrossRef Search ADS   Dawson A., Shi S., 2000. Tsunami deposits, Pure appl. Geophys. , 157, 875– 897. https://doi.org/10.1007/s000240050010 Google Scholar CrossRef Search ADS   Dawson A., Stewart I., 2007a. Tsunami deposits in the geological record, Sedimentary Geol. , 200, 166– 183. https://doi.org/10.1016/j.sedgeo.2007.01.002 Google Scholar CrossRef Search ADS   Dawson A., Stewart I., 2007b. Tsunami geoscience, Prog. Phys. Geogr. , 31, 575– 590. https://doi.org/10.1177/0309133307087083 Google Scholar CrossRef Search ADS   Dean R., Dalrymple R., 1991. Water Wave Mechanics for Engineers and Scientists , 2nd edn, World Scientific. Google Scholar CrossRef Search ADS   Deltares, 2014. Delft3d. simulation of multi-dimensional hydrodynamic flows and transport phenomena, including sediments . User manual, version 3.15.34158, http://oss.deltares.nl/documents/183920/185723/Delft3D-FLOW_User_Manual.pdf?version=1.0, 2014-05-28. Dominey-Howes D., Humphreys G., Hesse P., 2006. Tsunami and palaeotsunami depositional signatures and their potential value in understanding the late-Holocene tsunami record, Holocene , 16, 1095– 1107. https://doi.org/10.1177/0959683606069400 Google Scholar CrossRef Search ADS   Eagle Ray Publications, 2010. Pilot chart; map sheet: Pc 18, n. lefkáda to amvrakikós kólpos , 1: 93  213, Voula. Faccenna C., Becker T., Lucente F., Jolivet L., Rossetti F., 2001. History of subduction and back-arc extension in the central Mediterranean, Geophys. J. Int. , 145, 809– 820. https://doi.org/10.1046/j.0956-540x.2001.01435.x Google Scholar CrossRef Search ADS   Ferentinos G., 1991. Offshore geological hazards in the Hellenic Arc, Mar. Geotechnol. , 9, 261– 277. https://doi.org/10.1080/10641199009388244 Google Scholar CrossRef Search ADS   Ferentinos G., 1992. Recent gravitative mass movements in a highly tectonically active arc system: the Hellenic Arc, Mar. Geol. , 104, 93– 107. https://doi.org/10.1016/0025-3227(92)90086-W Google Scholar CrossRef Search ADS   Floth U., Vött A., May S., Brückner H., Brockmüller S., 2009. Geo-scientific evidence versus computer models of tsunami landfall in the Lefkada coastal zone (NW Greece), in Marburger Geographische Schriften , Vol. 145, pp. 140– 156, eds Vött A., Brückner H., Marburger Geographische Gesellschaft. Floth U., Vött A., May S., Brückner H., Brockmüller S., 2010. Estimating tsunami hazards between Lefkada and Preveza, NW Greece, by means of computer modeling, in Coastline Reports , Vol. 16, pp. 11– 14, eds Schwarzer K., Schrottke K., Stattegger K., EUCC-Deutschland. Floth U., Vött A., May S., Brockmüller S., Brückner H., 2013. Extreme wave events in the coastal zone of Lefkada and Akarnania (NW Greece) and their possible influence on civil structures. A comparison of field evidence and modeling results, in Akarnanien-Forschungen , Vol. 1, 31– 40. Fujiwara O., 2008. Bedforms and sedimentary structures characterizing tsunami deposits, in Tsunamiites—Features and Implications , pp. 51– 62. eds Shiki T., Tsuji Y., Yamazaki T., Minoura K., Elsevier. Google Scholar CrossRef Search ADS   Gelfenbaum G., Vatvani D., Jaffe B., Dekker F., 2007. Tsunami inundation and sediment transport in vicinity of coastal mangrove forest, in Coastal Sediments ’07 , 1117– 1128. doi:10.1061/40926(239)86. Ghionis G., Poulos S., Verykiou E., Karditsa A., Alexandrakis G., Andris P., 2015. The impact of an extreme storm event on the barrier beach of the Lefkada lagoon, NE Ionian Sea (Greece), Mediterranean Mar. Sci. , 16, 562– 572. https://doi.org/10.12681/mms.948 Google Scholar CrossRef Search ADS   Google Inc., 2013. Google earth, version 7.1.2.2041, computer software, https://www.google.com/earth/. Gusman A., Tanioka Y., Takahashi T., 2012. Numerical experiment and a case study of sediment transport simulation of the 2004 Indian Ocean tsunami in Lhok Nga, Banda Aceh, Indonesia, Earth Planets Space , 64, 817– 827. https://doi.org/10.5047/eps.2011.10.009 Google Scholar CrossRef Search ADS   Hill J., Collins G., Avdis A., Kramer S., Piggott M., 2014. How does multiscale modelling and inclusion of realistic palaeobathymetry affect numerical simulation of the Storegga slide tsunami? Ocean Modelling , 83, 11– 25. https://doi.org/10.1016/j.ocemod.2014.08.007 Google Scholar CrossRef Search ADS   Hills J., Mader C., 1997. Tsunami produced by the impacts of small asteroids, Ann. New York Acad. Sci. , 822, 381– 394. https://doi.org/10.1111/j.1749-6632.1997.tb48352.x Google Scholar CrossRef Search ADS   Imray Laurie Norie Wilson Ltd, 2011. Imray chart; map sheet: G 11, greece-ionian sea, north ionian islands, nísos kérkira to nísos levkas , 1: 185 000, Saint Ives. IOC—Intergovernmental Oceanographic Commission t. BODC—British Oceanographic Data Centre, 2003. Centenary edition of the GEBCO Digital Atlas , published on CD-ROM, Liverpool, http://www.gebco.net/data_and_products/gridded_bathymetry_data/, 2015-01-12. Jahns S., 2005. The Holocene history of vegetation and settlement at the coastal site of Lake Voulkaria in Acarnania, western Greece, Vegetation History and Archaeobotany , 14, 55– 66. https://doi.org/10.1007/s00334-004-0053-8 Google Scholar CrossRef Search ADS   Jarvis A., Reuter H., Nelson A., Guevara E., 2008. Hole-filled seamless srtm data v4 , International Centre for Tropical Agriculture (CIAT), http://srtm.csi.cgiar.org, 2014-04-08. Jenks G., 1967. The data model concept in statistical mapping, in International Yearbook of Cartography , Vol. 7, pp. 186– 190, ed. Philip G., Bonn. Jenks G., Caspall F., 1971. Error on choroplethic maps: definition, measurement, reduction, Ann. Assoc. Am. Geographers , 61, 217– 244. https://doi.org/10.1111/j.1467-8306.1971.tb00779.x Google Scholar CrossRef Search ADS   Jing Z., Rapp G., 2003. The coastal evolution of the ambracian embayment and its relationship to archaeological settings, in Hesperia Supplements , Vol. 32, pp. 157– 198. doi:10.2307/1354048. Google Scholar CrossRef Search ADS   Kapsimalis V., Pavlakis P., Poulos S., Alexandri S., Tziavos C., Sioulas A., Filippas D., Lykousis V., 2005. Internal structure and evolution of the Late Quaternary sequence in a shallow embayment: the Amvrakikos Gulf, NW Greece, Mar. Geol. , 222–223, 399– 418. https://doi.org/10.1016/j.margeo.2005.06.008 Google Scholar CrossRef Search ADS   Karageorgis A., Krasakopoulou E., Pappas G., Papageorgiou A., Taxiarchi M., 2006. Geochemical characteristics of surface sediments from the lagoons of Amvrakikos Gulf, Ionian Sea, in Proceedings of the 8th Panhellenic Symposium on Oceanography and Fisheries, 2006, Thessaloniki , pp. 793– 799. Kokkalas S., Xypolias P., Koukouvelas I., Doutsos T., 2006. Postcollisional contractional and extensional deformation in the Aegean region, in Postcollisional tectonics and magmatism in the Mediterranean region and Asia , Vol. 409, pp. 97– 123. doi:10.1130/0-8137-2409-0.97. Google Scholar CrossRef Search ADS   Lesser G., Roelvink J., van Kester J., Stelling G., 2004. Development and validation of a three-dimensional morphological model, Coastal Eng. , 51, 883– 915. https://doi.org/10.1016/j.coastaleng.2004.07.014 Google Scholar CrossRef Search ADS   Li L., Qiu Q., Huang Z., 2012a. Numerical modeling of the morphological change in Lhok Nga, west Banda Aceh, during the 2004 Indian Ocean tsunami: understanding tsunami deposits using a forward modeling method, Nat. Hazards , 64, 1549– 1574. https://doi.org/10.1007/s11069-012-0325-z Google Scholar CrossRef Search ADS   Li L., Huang Z., Qiu Q., Natawidjaja D., Sieh K., 2012b. Tsunami-induced coastal change: scenario studies for Painan, West Sumatra, Indonesia, Earth Planets Space , 64, 799– 816. https://doi.org/10.5047/eps.2011.08.002 Google Scholar CrossRef Search ADS   Li L., Huang Z., Qiu Q., 2014. Numerical simulation of erosion and deposition at the Thailand Khao Lak coast during the 2004 Indian Ocean tsunami, Nat. Hazards , 74, 2251– 2277. https://doi.org/10.1007/s11069-014-1301-6 Google Scholar CrossRef Search ADS   Liang D., Gotoh H., Khayyer A., Chen J., 2013. Boussinesq modelling of solitary wave and N-wave runup on coast, Appl. Ocean Res. , 42, 144– 154. https://doi.org/10.1016/j.apor.2013.05.008 Google Scholar CrossRef Search ADS   Madsen P., Fuhrman D., Schäffer H., 2008. On the solitary wave paradigm for tsunamis, J. geophys. Res. , 113, 1– 22. https://doi.org/10.1029/2008JC004932 Google Scholar CrossRef Search ADS   Martin M., Weiß R., Bourgeois J., Pinegina T., Houston H., Titov V., 2008. Combining constraints from tsunami modeling and sedimentology to untangle the 1969 Ozernoi and 1971 Kamchatskii tsunamis, Geophys. Res. Lett. , 35, 1– 6. May S., Vött A., Brückner H., Brockmüller S., 2007. Evidence of tsunamigenic impact on Actio headland near Preveza, NW Greece, in Coastline Reports , Vol. 9, pp. 115– 125, eds Gönnert G., Pflüger B., Bremer J.-A., EUCC-Deutschland. May S., Vött A., Brückner H., Grapmayer R., Handl M., Wennrich V., 2012a. The Lefkada barrier and beachrock system (NW Greece)—Controls on coastal evolution and the significance of extreme wave events, Geomorphology , 139–140, 330– 347. https://doi.org/10.1016/j.geomorph.2011.10.038 Google Scholar CrossRef Search ADS   May S., Vött A., Brückner H., Smedile A., 2012b. The Gyra washover fan in the Lefkada Lagoon, NW Greece—possible evidence of the 365 AD Crete earthquake and tsunami, Earth Planets Space , 64, 859– 874. https://doi.org/10.5047/eps.2012.03.007 Google Scholar CrossRef Search ADS   McGranahan G., Balk D., Anderson B., 2007. The rising tide: assessing the risks of climate change and human settlements in low elevation coastal zones, Environ. Urbanization , 19, 17– 37. https://doi.org/10.1177/0956247807076960 Google Scholar CrossRef Search ADS   McMaster R., 1997. In memoriam: George F. Jenks (1916–1996), Cartogr. Geogr. Inf. Syst. , 24, 56– 59. Morton R., Gelfenbaum G., Jaffe B., 2007. Physical criteria for distinguishing sandy tsunami and storm deposits using modern examples, Sedimentary Geol. , 200, 184– 207. https://doi.org/10.1016/j.sedgeo.2007.01.003 Google Scholar CrossRef Search ADS   Namegaya Y., Satake K., 2014. Reexamination of the A.D. 869 Jogan earthquake size from tsunami deposit distribution, simulated flow depth, and velocity, Geophys. Res. Lett. , 41, 2297– 2303. https://doi.org/10.1002/2013GL058678 Google Scholar CrossRef Search ADS   Ontowirjo B., Paris R., Mano A., 2010. Sediment transport modelling during 2004 tsunami runup at Lhok Nga (Sumatra, Indonesia), Geophys. Res. Abstr. , 12. Ontowirjo B., Paris R., Mano A., 2013. Modeling of coastal erosion and sediment deposition during the 2004 Indian Ocean tsunami in Lhok Nga, Sumatra, Indonesia, Nat. Hazards , 65, 1967– 1979. https://doi.org/10.1007/s11069-012-0455-3 Google Scholar CrossRef Search ADS   Papadopoulos G., Plessa A., 2000. Magnitude-distance relations for earthquake-induced landslides in Greece, Eng. Geol. , 58, 377– 386. https://doi.org/10.1016/S0013-7952(00)00043-0 Google Scholar CrossRef Search ADS   Papazachos B., 1996. Large seismic faults in the Hellenic arc, Ann. Geophys. , 39, 891– 903. Pareschi M., Boschi E., Favalli M., 2006a. Lost tsunami, Geophys. Res. Lett. , 33, 1– 6. Pareschi M., Boschi E., Mazzarini F., Favalli M., 2006b. Large submarine landslides offshore Mt. Etna, Geophys. Res. Lett. , 33, 1– 5. Pareschi M., Favalli M., Boschi E., 2006c. Impact of the Minoan tsunami of Santorini: simulated scenarios in the eastern Mediterranean, Geophys. Res. Lett. , 33, 1– 6. Partheniades E., 1965. Erosion and deposition of cohesive soils, J. Hydraul. Div. , 91, 105– 139. Períañez R., Abril J., 2014a. Modelling tsunamis in the eastern Mediterranean Sea. Application to the Minoan Santorini tsunami sequence as a potential scenario for the biblical Exodus, J. Mar. Syst. , 139, 91– 102. https://doi.org/10.1016/j.jmarsys.2014.05.016 Google Scholar CrossRef Search ADS   Períañez R., Abril J., 2014b. A numerical modeling study on oceanographic conditions in the former Gulf of Tartessos (SW Iberia): tides and tsunami propagation, J. Mar. Syst. , 139, 68– 78. https://doi.org/10.1016/j.jmarsys.2014.05.020 Google Scholar CrossRef Search ADS   Polonia A., Torelli L., Mussoni P., Gasperini L., Artoni A., Klaeschen D., 2011. The Calabrian Arc subduction complex in the Ionian Sea: regional architecture, active deformation, and seismic hazard, Tectonics , 30, 1– 28. https://doi.org/10.1029/2010TC002821 Google Scholar CrossRef Search ADS   Poulos S., Collins M., Lykousis V., 1995. Late Quaternary evolution of Amvrakikos Gulf, western Greece, Geo-Mar. Lett. , 15, 9– 16. https://doi.org/10.1007/BF01204492 Google Scholar CrossRef Search ADS   Poulos S., Lykousis V., Collins M., Rohling E., Pattiaratchi C., 1999. Sedimentation processes in a tectonically active environment: the Kerkyra-Kefalonia submarine valley system (NE Ionian Sea), Mar. Geol. , 160, 25– 44. https://doi.org/10.1016/S0025-3227(99)00016-X Google Scholar CrossRef Search ADS   Poulos S., Kapsimalis V., Tziavos C., Pavlakis P., Leivaditis G., Collins M., 2005. Sea-level stands and Holocene geomorphological evolution of the northern deltaic margin of Amvrakikos Gulf (western Greece), Z. Geomorphol., Supplementary Issues , 137, 125– 145. Poulos S.et al.  , 2015. Hydrodynamic, neotectonic and climatic control of the evolution of a barrier beach in the microtidal environment of the NE Ionian Sea (eastern Mediterranean), Geo-Mar. Lett. , 35, 37– 52. https://doi.org/10.1007/s00367-014-0390-2 Google Scholar CrossRef Search ADS   Pritchard D., Dickinson L., 2008. Modelling the sedimentary signature of long waves on coasts: implications for tsunami reconstruction, Sedimentary Geol. , 206, 42– 57. https://doi.org/10.1016/j.sedgeo.2008.03.004 Google Scholar CrossRef Search ADS   Richardson J., Zaki W., 1997. Sedimentation and fluidisation: part I, Chem. Eng. Res. Des. , 75, S82– S100. https://doi.org/10.1016/S0263-8762(97)80006-8 Google Scholar CrossRef Search ADS   Röbke B., Vött A., 2017. The tsunami phenomenon, Prog. Oceanogr. , 159, 296– 322. https://doi.org/10.1016/j.pocean.2017.09.003 Google Scholar CrossRef Search ADS   Röbke B., Schüttrumpf H., Wöffler T., Hadler H., Willershäuser T., Vött A., 2012. Tsunamis in the Gulf of Kyparissia (western Peloponnese, Greece) - risk assessment based on numeric simulation and field evidence, in Proceedings of the 5th International Short Conference on Applied Coastal Research, 2011, Aachen , pp. 560– 569. Röbke B., Schüttrumpf H., Wöffler T., Fischer P., Hadler H., Ntageretzis K., Willershäuser T., Vött A., 2013. Tsunami inundation scenarios for the Gulf of Kyparissia (western Peloponnese, Greece) derived from numerical simulations and geo-scientific field evidence, Z. Geomorphol. N.F., Supplementary Issues , 57, 69– 104. Google Scholar CrossRef Search ADS   Röbke B., Vött A., Willershäuser T., Fischer P., Hadler H., 2015. Considering coastal palaeogeographical changes in a numerical tsunami model—a progressive base to compare simulation results with field traces from three coastal settings in western Greece, Z. Geomorphol. N.F., Supplementary Issues , 59, 157– 188. https://doi.org/10.1127/zfg_suppl/2014/S-00210 Google Scholar CrossRef Search ADS   Röbke B., Schüttrumpf H., Vött A., 2016. Effects of different boundary conditions and palaeotopographies on the onshore response of tsunamis in a numerical model—a case study from western Greece, Cont. Shelf Res. , 124, 182– 199. https://doi.org/10.1016/j.csr.2016.04.010 Google Scholar CrossRef Search ADS   Samaras A., Karambas T., Archetti R., 2015. Simulation of tsunami generation, propagation and coastal inundation in the Eastern Mediterranean, Ocean Sci. , 11, 643– 655. https://doi.org/10.5194/os-11-643-2015 Google Scholar CrossRef Search ADS   Satake K.et al.  , 2013. Tsunami source of the 2010 Mentawai, Indonesia earthquake inferred from tsunami field survey and waveform modeling, Pure appl. Geophys. , 170, 1567– 1582. https://doi.org/10.1007/s00024-012-0536-y Google Scholar CrossRef Search ADS   Schielein P., Zschau J., Woith H., Schellmann G., 2007. Tsunamigefährdung im mittelmeer—eine analyse geomorphologischer und historischer zeugnisse, in Bamberger Geographische Schriften , Vol. 22, pp. 153– 199, eds Dix A., Göler D., Redepenning M., Schellmann G., University of Bamberg Press. Shanmugam G., 2012. Process-sedimentological challenges in distinguishing paleo-tsunami deposits, Nat. Hazards , 63, 5– 30. https://doi.org/10.1007/s11069-011-9766-z Google Scholar CrossRef Search ADS   Shiki T., Tachibana T., Fujiwara O., Goto K., Nanayama F., Yamazaki T., 2008. Characteristic features of tsunamiites, in Tsunamiites—Features and Implications , pp. 319– 340, eds Shiki T., Tsuji Y., Yamazaki T., Minoura K. Elsevier. Google Scholar CrossRef Search ADS   Soloviev S., Solovieva O., Go C., Kim K., Shchetnikov N., 2000. Tsunamis in the Mediterranean Sea 2000 B.C.–2000 A.D. , Springer. Google Scholar CrossRef Search ADS   Spiske M., Piepenbreier J., Benavente C., Bahlburg H., 2013. Preservation potential of tsunami deposits on arid siliciclastic coasts, Earth-Sci. Rev. , 126, 58– 73. https://doi.org/10.1016/j.earscirev.2013.07.009 Google Scholar CrossRef Search ADS   Stiros S., 2010. The 8.5+ magnitude, AD365 earthquake in Crete: coastal uplift, topography changes, archaeological and historical signature, Quat. Int. , 216, 54– 63. https://doi.org/10.1016/j.quaint.2009.05.005 Google Scholar CrossRef Search ADS   Sugawara D., Imamura F., Matsumoto H., Goto K., Minoura K., 2011. Reconstruction of the AD 869 Jogan earthquake induced tsunami by using the geological data, J. Japan. Soc. Nat. Disaster Sci. , 29, 501– 516. Sugawara D., Goto K., Imamura F., Matsumoto H., Minoura K., 2012. Assessing the magnitude of the 869 Jogan tsunami using sedimentary deposits: prediction and consequence of the 2011 Tohoku-oki tsunami, Sedimentary Geol. , 282, 14– 26. https://doi.org/10.1016/j.sedgeo.2012.08.001 Google Scholar CrossRef Search ADS   Sugawara D., Imamura F., Goto K., Matsumoto H., Minoura K., 2013. The 2011 Tohoku-oki earthquake tsunami: similarities and differences to the 869 Jogan tsunami on the Sendai Plain, Pure appl. Geophys. , 170, 831– 843. https://doi.org/10.1007/s00024-012-0460-1 Google Scholar CrossRef Search ADS   Sugawara D., Goto K., Jaffe B., 2014a. Numerical models of tsunami sediment transport—current understanding and future directions, Mar. Geol. , 352, 295– 320. https://doi.org/10.1016/j.margeo.2014.02.007 Google Scholar CrossRef Search ADS   Sugawara D., Takahashi T., Imamura F., 2014b. Sediment transport due to the 2011 Tohoku-oki tsunami at Sendai: results from numerical modeling, Mar. Geol. , 358, 18– 37. https://doi.org/10.1016/j.margeo.2014.05.005 Google Scholar CrossRef Search ADS   Synolakis C., Bernard E., Titov V., Kânoğlu U., González F., 2008. Validation and verification of tsunami numerical models, Pure appl. Geophys. , 165, 2197– 2228. https://doi.org/10.1007/s00024-004-0427-y Google Scholar CrossRef Search ADS   Szczuciński W., 2012. The post-depositional changes of the onshore 2004 tsunami deposits on the Andaman Sea coast of Thailand, Nat. Hazards , 60, pp. 115– 133. https://doi.org/10.1007/s11069-011-9956-8 Google Scholar CrossRef Search ADS   Tadepalli S., Synolakis C., 1994. The run-up of N-waves on sloping beaches, Proc. R. Soc. A , 445, 99– 112. https://doi.org/10.1098/rspa.1994.0050 Google Scholar CrossRef Search ADS   Tadepalli S., Synolakis C., 1996. Model for the leading waves of tsunamis, Phys. Rev. Lett. , 77, 2141– 2144. https://doi.org/10.1103/PhysRevLett.77.2141 Google Scholar CrossRef Search ADS   Tziavos C., 1997. Paleogeographic evolution of the amvrakikos gulf, western greece, in Proceedings of the 22nd International Symposium on Engineering Geology and the Environment, 1997, Athen , pp. 425– 430. UKHO—United Kingdom Hydrographic Office, 2010. Admiralty chart; map sheet: Ba189, greece-west coast, nisída sapiéntza to nísos paxoí, 1:300 000 , Taunton. UNESCO, 1981. Background papers and supporting data on the international equation of state of seawater, 1980, in UNESCO technical papers in marine science , Vol. 38, pp. 1– 192, UNESCO. USGS—United States Geological Survey’s Earthquake Hazards Program, 2015. Signifcant earthquake archive , http://earthquake.usgs.gov/earthquakes/eqinthenews/, 2015-11-10. van Rijn L., Roelvink J., ter Horst W., 2001. Approximation formulae for sand transport by currents and waves and implementation in DELFT-MOR, in Repository Hydraulic Engineering Reports , Vol. Z3054.20/40, pp. 1– 44, Delft Hydraulics (WL). Verboom G., Slob A., 1984. Weakly-reflective boundary conditions for two-dimensional shallow water flow problems, Adv. Water Resour. , 7, 192– 197. https://doi.org/10.1016/0309-1708(84)90018-6 Google Scholar CrossRef Search ADS   Vött A., May S., 2009. Auf den Spuren von Tsunamis im östlichen Mittelmeer, Geographische Rundschau , 61, 42– 48. Vött A., May S., Brückner H., Brockmüller S., 2006. Sedimentary evidence of late Holocene tsunami events near Lefkada Island (NW Greece), Z. Geomorphol., Supplementary Issues , 146, 139– 172. Vött A., Brückner H., May S., Lang F., Brockmüller S., 2007. Late Holocene tsunami imprint at the entrance of the Ambrakian Gulf (NW Greece), Méditerranée , 108, 43– 57. Google Scholar CrossRef Search ADS   Vött A., Brückner H., May S., Lang F., Herd R., Brockmüller S., 2008. Strong tsunami impact on the Bay of Aghios Nikolaos and its environs (NW Greece) during Classical-Hellenistic times, Quat. Int. , 181, 105– 122. https://doi.org/10.1016/j.quaint.2007.02.017 Google Scholar CrossRef Search ADS   Vött A.et al.  , 2009a. The Lake Voulkaria (Akarnania, NW Greece) palaeoenvironmental archive—a sediment trap for multiple tsunami impact since the mid-Holocene, Z. Geomorphol. N.F., Supplementary Issues , 53, 1– 37. Google Scholar CrossRef Search ADS   Vött A.et al.  , 2009b. Traces of Holocene tsunamis across the Sound of Lefkada, NW Greece, Glob. Planet. Change , 66, 112– 128. https://doi.org/10.1016/j.gloplacha.2008.03.015 Google Scholar CrossRef Search ADS   Vött A.et al.  , 2010. Beachrock-type calcarenitic tsunamites along the shores of the eastern Ionian Sea (western Greece)—case studies from Akarnania, the Ionian Islands and the western Peloponnese, Z. Geomorphol. N.F., Supplementary Issues , 54, 1– 50. Google Scholar CrossRef Search ADS   Vött A.et al.  , 2011. Sedimentological and geoarchaeological evidence of multiple tsunamigenic imprint on the Bay of Palairos-Pogonia (Akarnania, NW Greece), Quat. Int. , 242, 213– 239. Wheatcroft R., 1990. Preservation potential of sedimentary event layers, Geology , 18, 843– 845. https://doi.org/10.1130/0091-7613(1990)018<0843:PPOSEL>2.3.CO;2 Google Scholar CrossRef Search ADS   Wheatcroft R., Drake D., 2003. Post-depositional alteration and preservation of sedimentary event layers on continental margins, I. The role of episodic sedimentation, Mar. Geol. , 199, 123– 137. https://doi.org/10.1016/S0025-3227(03)00146-4 Google Scholar CrossRef Search ADS   Yolsal-Çevikbilen S., Taymaz T., 2012. Earthquake source parameters along the Hellenic subduction zone and numerical simulations of historical tsunamis in the eastern Mediterranean, Tectonophysics , 536–537, 61– 100. https://doi.org/10.1016/j.tecto.2012.02.019 Google Scholar CrossRef Search ADS   © The Author(s) 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Geophysical Journal International Oxford University Press

# Hydro- and morphodynamic tsunami simulations for the Ambrakian Gulf (Greece) and comparison with geoscientific field traces

, Volume 213 (1) – Apr 1, 2018
23 pages

/lp/ou_press/hydro-and-morphodynamic-tsunami-simulations-for-the-ambrakian-gulf-Jr1UyALupZ
Publisher
The Royal Astronomical Society
ISSN
0956-540X
eISSN
1365-246X
D.O.I.
10.1093/gji/ggx553
Publisher site
See Article on Publisher Site

### Abstract

Summary In order to derive local tsunami risks for a particular coast, hydro- and morphodynamic numerical models that are calibrated and compared with sedimentary field data of past tsunami impacts have proven very effective. While this approach has widely been used with regard to recent tsunami events, comparable investigations into pre-/historical tsunami impacts hardly exist, which is the objective of this study focusing on the Ambrakian Gulf in northwestern Greece. The Ambrakian Gulf is located in the most active seismotectonic and by this most tsunamigenic area of the Mediterranean. Accordingly, palaeotsunami field studies have revealed repeated tsunami impacts on the gulf during the past 8000 yr. The current study analyses 151 vibracores of the Ambrakian Gulf coast in order to evaluate tsunami signals in the sedimentary record. Based on a hydro- and morphodynamic numerical model of the study area, various tsunami waves are simulated with the aim of finding scenarios that compare favourably with tsunami deposits detected in the field. Both, field data and simulation results suggest a decreasing tsunami influence from the western to the eastern Ambrakian Gulf. Various scenarios are needed to explain tsunami deposits in different parts of the gulf. Whereas shorter period tsunami waves (T = 30 min) from the south and west compare favourably with field data in the western gulf, longer period waves (T = 80 min) from a western direction show the best agreement with tsunami sediments detected in southwestern Aktio Headland and in the more central parts of the Ambrakian Gulf including Lake Voulkaria. Tsunamis from the southwest generally do not accord with field traces. Besides the spatial sediment distribution, the numerical model accurately reflects the sedimentary composition of the detected event deposits and reproduces a number of essential features typical of tsunamites, which were also observed in the field. Such include fining- and thinning-landward and the marine character of the deposits. By contrast, the simulated thickness of tsunami sediments usually lags behind the observed thickness in the field and some event layers cannot be explained by any of the simulated scenarios. Regarding the frequency of past tsunami events and their spatial dimensions indicated by both field data and simulation results, a high tsunami risk has to be derived for the Ambrakian Gulf. Geomorphology, Tsunamis, Non-linear differential equations, Numerical modelling, Neotectonics, Subduction zone processes 1 INTRODUCTION With more than 10 per cent of the world’s population living in coastal lowlands less than 10 m a.s.l. (above sea level; McGranahan et al.2007, p. 22), the vulnerability of coasts to tsunami inundation is very high. Recent history has seen disastrous tsunami events in Papua New Guinea (1998), southeast Asia (2004) and Japan (2011), which have particularly emphasized the need for improved risk assessment and mitigation plans. During the last years, tsunami forecast methods have significantly extended and gained in reliability. Modern tsunami risk assessment uses a multi-methodological approach based on oceanic warning systems, seismological data, numerical simulation techniques and sedimentary field traces. In order to derive tsunami risks for a particular coast, hydro- and morphodynamic numerical tsunami models that are calibrated and compared with sedimentary field data and—if available—buoy measurements and information on the trigger mechanisms of past tsunami impacts have proven very effective. Knowing the frequency and the spatial extent of pre-/historical tsunami inundation derived from field traces makes it possible to draw conclusions on future tsunami risks. However, field traces only allow an accurate reconstruction of former inundation events when combined with numerical tsunami models, as performed in numerous recent studies (e.g. Bondevik et al.2005; Martin et al.2008; Borrero et al.2009; Röbke et al.2013, 2015, 2016; Satake et al.2013; Sugawara et al.2013; Hill et al.2014). Such studies are particularly promising, if numerical models do not only compute the pure tsunami hydrodynamics but also take account of sediment processes, which allows a one-to-one comparison between field data and model (cf. Sugawara et al.2014a). This approach was taken in a number of studies focusing on tsunami impacts in recent history, in particular the Indian Ocean tsunami 2004 in Sumatra and Thailand (Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b; Ontowirjo et al.2010, 2013; Gusman et al.2012; Li et al.2012a,b, 2014), the 2009 South Pacific tsunami in American Samoa (Apotsos et al.2011c) and the 2011 Tōhoku tsunami in Japan (Sugawara et al.2014b). A good preservation provided, tsunami field traces together with numerical simulations may further give valuable insights into related tsunami trigger mechanisms (e.g. Dawson 1999; Martin et al.2008; Sugawara et al.2011, 2012; Namegaya & Satake 2014). Although the synthesis of sedimentary field data and morphodynamic numerical models has repeatedly been applied with regard to recent tsunami events, there are hardly comparable studies in view of pre-/historical tsunamis (cf. Röbke & Vött 2017, pp. 313–315), which is the approach of the following investigation. This paper concentrates on tsunami impacts in the Ambrakian Gulf (northwestern Greece) during the last 8000 yr. The Ambrakian Gulf is a marine inlet of the eastern Ionian Sea in coastal Epirus and Akarnania with a maximum N–S- and W–E-extent of about 20 km and 35 km respectively (Figs 1 and 2). The gulf is connected to the Ionian Sea by the Strait of Preveza, a 600 m wide channel less than 25 m deep (UKHO 2010). Maximum water depths are found in the eastern gulf with roughly 65 m. The Ambrakian Gulf forms a tectonic basin dominated by N–S crustal extension along WNW-ESE trending normal faults (Clews 1989, p. 454). According to palaeogeographical studies by Poulos et al. (1995, 2005), Tziavos (1997), Jing & Rapp (2003), Brockmüller et al. (2005) and Kapsimalis et al. (2005), the gulf was isolated from the Ionian Sea during the last ice age (Weichselian or Würm period), but became a marine inlet in the early Holocene (epoch of the last 11 700 yr). By the 4th millennium BC, the ongoing rise in sea level had resulted in the complete flooding of the contemporary river deltas in the northern gulf, which extended far more into the gulf than nowadays. At the same time, the coastline in the western, southern and eastern gulf approximately reached its present position. Subsequently, considerable sediment input by the rivers Louros and Arachthos created the large delta complex nowadays found in the northern gulf (Fig. 2). To the present, tectonic crust movements cause a relative uplift of the northwestern gulf compared to the southeastern part, having resulted in a vertical shift of about 6 m since the 5th millennium BC (Brockmüller et al.2005, pp. 46, 47). Figure 1. View largeDownload slide Satellite image and submarine bed level in m b.s.l. (below sea level) of the central and eastern Mediterranean Google Inc. (2013), including the simplified tectonic regime and location of selected backarc volcanoes according to Faccenna et al. (2001, p. 810), Kokkalas et al. (2006, p. 100) and Polonia et al. (2011, p. 2). Grey dotted boxes display the overall and nested model domains, the red dotted lines mark the hydrodynamic open boundaries (south, southwest and west) where the tsunami waves enter the overall model. Figure 1. View largeDownload slide Satellite image and submarine bed level in m b.s.l. (below sea level) of the central and eastern Mediterranean Google Inc. (2013), including the simplified tectonic regime and location of selected backarc volcanoes according to Faccenna et al. (2001, p. 810), Kokkalas et al. (2006, p. 100) and Polonia et al. (2011, p. 2). Grey dotted boxes display the overall and nested model domains, the red dotted lines mark the hydrodynamic open boundaries (south, southwest and west) where the tsunami waves enter the overall model. Figure 2. View largeDownload slide Overview map of Greece and bird’s eye view of the Ambrakian Gulf, including the opening to the Ionian Sea, the coastal flats in the north and the hilly shore in the southeast. The figure is based on a 3-D excerpt of the digital elevation model employed for numerical tsunami simulations in this study. *Due to the 3-D view, the scale bar gives the true distance only in the centre of the figure. Figure 2. View largeDownload slide Overview map of Greece and bird’s eye view of the Ambrakian Gulf, including the opening to the Ionian Sea, the coastal flats in the north and the hilly shore in the southeast. The figure is based on a 3-D excerpt of the digital elevation model employed for numerical tsunami simulations in this study. *Due to the 3-D view, the scale bar gives the true distance only in the centre of the figure. Located in the most seismically active and by this most tsunamigenic region in the entire Mediterranean, that is, western Greece (Fig. 1; Soloviev et al.2000, p. 14; Schielein et al.2007, pp. 164, 193–199; Röbke et al.2013, pp. 70–77), the Ambrakian Gulf and its environs were frequently affected by tsunami landfall in the past. Besides historical accounts, this is documented by sedimentary field traces in the area. According to detailed field studies by Vött et al. (2006, 2007, 2008, 2009a, 2010) and May et al. (2007, 2012a), major tsunamis struck the Ambrakian Gulf in the 6th millennium BC or shortly afterwards, between 2870 cal BC and 2350 cal BC, around 1000 cal BC, between 395 cal BC and 247 cal BC, in the 4th century AD and around 840 cal AD. Further events occurred in the last 700 yr. Sedimentary signatures of most of these impacts were also found in close vicinity to the gulf on the nearby island of Lefkada (Vött et al.2006, 2008, 2009b; May et al.2012a,b) and in the Bay of Palairos (Vött et al.2011) and seem to be related to supraregional tsunami events in the Mediterranean. The wider study area is the northernmost region in the eastern Mediterranean where sedimentary signatures of the 365 AD Crete tsunami were found (e.g. Vött et al.2009b; May et al.2012b). Besides field studies, hydrodynamic numerical tsunami simulations were computed for the coastal area between Preveza and Lefkada City (Floth et al.2009, 2010, 2013). In this initial approach, simulation results, which were based on a small-scale numerical model of the area, were compared with local tsunami field traces. The current study aims to (i) analyse and evaluate systematically tsunami signals in the sedimentary record of 151 vibracores taken all along the Ambrakian Gulf within previous palaeotsunami field studies, (ii) derive the scale of pre-/historical tsunami landfall in the gulf based on field data, (iii) set up a hydro- and morphodynamic numerical tsunami model, which takes account of various tsunami wave scenarios and considers sediment processes, (iv) calibrate and compare the numerical model with the spatial distribution and sedimentary composition of detected tsunami deposits in the field and find tsunami scenarios that replicate field data in order to (v) assess tsunami risks for the Ambrakian Gulf. 2 METHODOLOGY 2.1 Field work Within the framework of previous palaeogeographical studies, 151 vibracores were drilled alongshore the Ambrakian Gulf using a handheld engine-powered Atlas Copco coring device (type Cobra mk1). The final coring depth at each location (maximum 18 m below ground surface) was reached, when either pre-Holocene deposits or bedrock appeared or the sedimentary facies did not change for several metres. Vibracores were first documented by digital photographs and subsequently described regarding their sedimentary features, such as grain size, texture, colour, fossil content and pedogenetic properties (cf. Ad Hoc-AG Boden 2005). The exact location of all vibracores was determined using a Topcon DGPS-device (type HiPer Pro). 2.2 Evaluation of field data A number of vibracores taken in the Ambrakian Gulf had already been analysed in view of the regional palaeogeography and palaeo-tsunami signatures by Vött et al. (2006, 2007, 2008, 2009a, 2010, 2011) and May et al. (2007, 2012a). Following these investigations, in the current study, all 151 vibracores were analysed and evaluated systematically with regard to tsunami candidate layers based on a set of sedimentary criteria which have been associated with onshore tsunamites worldwide (e.g. Dawson 1994; Dawson & Shi 2000, pp. 173–180; Dawson & Stewart 2007a, pp. 578–581; Dawson & Stewart 2007b; Dominey-Howes et al.2006; Morton et al.2007; Fujiwara 2008; Shiki et al.2008, pp. 45, 46; Vött & May 2009; Shanmugam 2012, pp. 307–310; Röbke & Vött 2017). Such physical criteria are directly ascribed to the hydrodynamic characteristics of tsunami inundation and particularly include basal erosional unconformities, fining-upward sequences, high content of marine macrofossils, clayey to silty rip-up clasts from underlying units, mud caps at the top of event layers, fining- and thinning-landward, a general allochthonous, high-energy character of the sediment (cf. Fig. 3). Figure 3. View largeDownload slide Vibracore AKT 40 showing sedimentary features typical of tsunami deposits, including a basal erosional unconformity, numerous marine macrofossils, normal grading and clayey to silty rip-up clasts from underlying units. Such features, among others, were referred to in order to identify tsunami candidate layers in 151 vibracores alongshore the Ambrakian Gulf shown in Fig. 5. Figure 3. View largeDownload slide Vibracore AKT 40 showing sedimentary features typical of tsunami deposits, including a basal erosional unconformity, numerous marine macrofossils, normal grading and clayey to silty rip-up clasts from underlying units. Such features, among others, were referred to in order to identify tsunami candidate layers in 151 vibracores alongshore the Ambrakian Gulf shown in Fig. 5. For some vibracores, also geochemical, microfaunal and palynological analyses had been conducted, which further facilitated the identification of tsunami deposits in the study area (Brockmüller et al.2005; Jahns 2005; Vött et al.2006, 2007, 2008, 2009a, 2010, 2011; May et al.2007, 2012a). Findings from these studies together with physical sedimentary features described above and correlations between related event strata at neighbouring vibracoring sites allowed a systematic and standardized identification of tsunami deposits along the entire Ambrakian Gulf coast. In this process, also storm events were taken into account as a potential mechanism to leave high-energy deposits. However, the lack of written accounts of extraordinary storm surges but multiple sources of data of tsunami impacts in the study area, the geomorphological appearance of the Ambrakian Gulf coast, which does not imply any recent to subrecent storm activity beyond the supralittoral zone, regional buoy and tide gauge data and studies by Ghionis et al. (2015) and Poulos et al. (2015, pp. 43–45) that reveal an expected maximum run-up of wind generated waves of only 2.82 m a.s.l. for the nearby tip of northern Lefkada, the short maximum fetch in the gulf of less than 40 km (cf. Fig. 2) allowing no high waves to develop, the lack of any sedimentary features in the vibracores typically associated with storm deposits (tempestites), and the fact that several event layers in the vibracores of the Ambrakian Gulf correlate well with tsunami deposits described from other coasts in western Greece and by this probably point to the same tsunami events strongly suggest the event deposits being of tsunamigenic origin (for more details see Vött et al.2007, pp. 51–53; Vött et al.2008, pp. 116–119; Vött et al.2011, pp. 233, 234). Correlations between related event strata could only be drawn for selected vibracores for which geochronological data were available but not for the whole set of vibracores. In order to still achieve a clear and meaningful overview of all vibracore data in the study area and to allow a direct comparison with simulation results, the average thickness of all tsunami candidate layers per vibracore was determined. This parameter indicates the average sedimentary impact of a single tsunami event most adequately and can be compared to the amount of sediment accumulation or erosion computed for 18 different single tsunami wave scenarios in the numerical model (Section 2.4). The average thickness was classified according to the Jenks optimization method, which yields the minimum in-class variance but maximizes the variances between all classes (Jenks 1967; Jenks & Caspall 1971; McMaster 1997). This classification method turned out to be most appropriate with regard to field data of this study as it demonstrates natural breaks in the frequency distribution and by this allows an easier visual distinction between areas of dominant and low tsunami sedimentation. Also the standard deviation of the tsunami layer thickness per vibracore was determined, which shows that the distribution of thicknesses per vibracore is relatively peaked and simple (Section 3.1). This justifies a direct comparison of the average tsunami layer thickness per vibracore (resulting from several tsunami events in pre-/history) with the simulated sedimentation and erosion by individual tsunami waves in the numerical model. Using geochronological data, it was further possible to compare several correlating tsunami candidate layers, thus deriving from the same events, with the simulation results for three locations in the western Ambrakian Gulf. This was done on the basis of the 840 cal AD tsunami for southwestern Aktio Headland, the 1000 cal BC tsunami for the entrance of Lake Voulkaria as well as the 4th century AD tsunami for the southern shore of Lake Voulkaria. Based on these events and locations, field and model data could directly be compared in view of individual past tsunamis. This further allowed a comparison not only of the sediment thickness but also of the sedimentary composition (i.e. content of mud, fine sand, middle sand, coarse sand and gravel) of tsunami deposits in both, the field and the model. 2.3 Digital elevation model The digital elevation model (DEM) used for tsunami simulations in this study covers the eastern Ionian Sea and the west coast of Greece, including the Ambrakian Gulf (Figs 1 and 2). The domain stretches from 37.47°N to 39.81°N and from 18.75°E to 21.63°E ($$\widehat{=}$$ 260 km * 250 km). The DEM was derived from several sources of data. The offshore topography of the open sea is based on the 30 arc-second (= 926.1 m) interval grid of GEBCO (IOC, IHO & BODC 2003). Bathymetric data of the shelf and the Ambrakian Gulf was digitized from several nautical charts on the scale of 1:93 213 to 1:300 000 (Imray Laurie Norie & Wilson Ltd 2011; Eagle Ray Publications 2010; UKHO 2010). SRTM elevation data version 4 with a resolution of 90 m (Jarvis et al.2008) and 30 m Google Earth elevation data (Google Inc. 2013; generated using the eSurvey Earth website http://www.esurveyearth.com/) were used to derive the onshore topography of the wider study area and alongshore the Ambrakian Gulf respectively. Among all freely available elevation data, both data sets showed the best match with DGPS-based elevation data measured in the field and altitude markings in topographical maps. All elevation data was edited and assembled to an overall DEM with a maximum resolution of 30 m using ArcGIS Desktop 10.2 by Esri. 2.4 Hydro- and morphodynamic numerical model In this study, a hydro- and morphodynamic numerical tsunami model was created using the Delft3D modelling suite (version 4.01.00.rc.12) developed by the Dutch institute Deltares. Delft3D is a validated software suite to simulate both, off- and onshore tsunami propagation, including sediment transport (Apotsos et al.2011a). It has been successfully employed for 1-D (Apotsos et al.2011d, 2012), 2-D (Röbke et al.2012, 2013, 2015, 2016; Chacón-Barrantes et al.2013; Períañez & Abril 2014a,b) and 3-D tsunami simulations (Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b,c). All simulations were calculated using the program module Delft3D-FLOW, which solves the nonlinear shallow water equations of unsteady flow and transport phenomena based on the full Navier–Stokes equations for incompressible free surface flow (for this and the following see Deltares 2014, pp. 177–180). The module is designed for flow phenomena where the horizontal spatial and temporal scales are much larger than the vertical scales, such as tidal waves, storm surges or tsunamis. In Delft3D-FLOW, the nonlinear shallow water equations are solved in two (depth-averaged) or in three dimensions. As the water density in the oceans can approximately be regarded as vertically homogeneous and 3-D test simulations showed a small impact on mud dynamics in the numerical model, the 2D, depth-averaged calculation mode is appropriate for tsunami simulations conducted here. This study only considers the pure hydro- and morphodynamics of tsunami waves, whereas trigger mechanisms and the process of wave generation were not simulated. Accordingly, the numerical model is based on various hydrodynamic boundary conditions describing tsunami waves with different wave periods, wave amplitudes and propagation directions in order to derive wave scenarios that accord with sedimentary field data (cf. Pritchard & Dickinson 2008; Gelfenbaum et al.2007; Apotsos et al.2009, 2011a,b,c,d, 2012). This approach was taken because in most cases, no correlation could be drawn between palaeotsunami field traces and written tsunami accounts for the Mediterranean and, by this, no information was available neither on related trigger mechanisms nor on tsunami wave characteristics. Furthermore, there is no single specific tsunami source with regard to the Ambrakian Gulf, but instead various potential trigger mechanisms exist, both in the near- and far-field (cf. Fig. 1; Röbke et al.2013, pp. 74–77; Samaras et al.2015, p. 649): strong shallow-focus earthquakes with a magnitude of Ms ≥ 6.5 along several shallow dipping thrust faults of up to 300 km in the area of the Calabrian Trench, the Hellenic Trench and its northward extension (Papazachos 1996; Benetatos et al.2004,USGS 2015), huge submarine landslides connected with the ubiquitous steep slopes and high seismic activity in the central and eastern Mediterranean Sea (Ferentinos 1991, 1992; Papadopoulos & Plessa 2000; Pareschi et al.2006a,b), severe volcanic eruptions of Mount Etna (Sicily, Italy) and other backarc volcanoes in the Aegean Sea (Pareschi et al.2006a,b,c). Thus, the number of potential tsunami sources is too large to be considered in this study. Instead, artificial waves that are realistic in view of the (supra)regional physical geography were created. The hydrodynamic boundary conditions of the numerical model were determined according to different tsunami wave types, that is, leading-elevation N-waves (LEN-waves), leading-depression N-waves (LDN-waves) (e.g. Tadepalli & Synolakis 1994, 1996) and solitary waves (e.g. Dean & Dalrymple 1991). A sensitivity analysis based on multiple test runs demonstrated that the general spatial distribution of tsunami deposits in the study area as found in the vibracores was best replicated by those test runs using LDN-waves. In contrast, LEN-waves and solitary waves resulted in a simulated tsunami sedimentation frequently below the magnitude of the field observations (cf. N-wave effect discussed by Röbke et al.2016) and/or showing a divergent pattern. LDN-waves particularly occur when tsunamis are triggered relatively close to the coast and the waves do not have enough time to evolve into LEN-waves or solitary waves (Tadepalli & Synolakis 1994, p. 100; Tadepalli & Synolakis 1996, p. 2143; Madsen et al.2008, pp. 2, 3; Synolakis et al.2008, pp. 2202, 2203; Liang et al.2013, pp. 144, 145; Röbke & Vött 2017, p. 313). Considering the multiple potential generation mechanisms of tsunamis in close vicinity to the Ambrakian Gulf (see above and Fig. 1), LDN-waves are likely to appear in the study area. Therefore and as a result of the sensitivity analysis for different wave types, this paper concentrates on isosceles LDN-waves, using the formulation by Tadepalli & Synolakis (1994, p. 105) (Chacón-Barrantes et al.2013, pp. 15, 16):   \begin{eqnarray} \eta &=& \frac{3\,\sqrt{3}\,H}{2}\ \mathrm{sech}^2\left[\frac{3\,\alpha }{2}\,\root 4 \of {\frac{3\,A}{4}}\,\left(t-t_0\right)\right]\nonumber\\ &&\times\,\tanh \left[\frac{3\,\alpha }{2}\,\root 4 \of {\frac{3\,A}{4}}\,\left(t-t_0\right)\right] \end{eqnarray} (1)where η = elevation above water surface [m], H = wave height [m], α = constant modifying the wavelength, A = wave amplitude ($$= \frac{H}{2}$$) [m], t = time [s] and t0 = time of the wave’s mid-point (half wave period) [s]. LDN-waves with periods of 30 min and 80 min and amplitudes of 2 m, 4 m and 6 m were created (Fig. 4), these wave parameters having the strongest influence on the onshore tsunami response. Furthermore, three propagation directions were considered in the numerical model, that is, south, southwest and west, which are the favourite wave propagation directions with regard to the potential tsunami generation mechanisms (cf. Fig. 1). All LDN-waves were simulated as single waves in order to investigate their individual morphodynamic impact depending on the amplitude, period and propagation direction. This resulted in a total of 18 scenarios discussed in this paper (Table 1). Figure 4. View largeDownload slide Leading-depression N-waves according to Eq. 1 with amplitudes of 2 m, 4 m and 6 m and periods of 30 min and 80 min. These waves form the hydrodynamic boundary conditions of the numerical tsunami model of this study at the southern, southwestern and western model boundary. Figure 4. View largeDownload slide Leading-depression N-waves according to Eq. 1 with amplitudes of 2 m, 4 m and 6 m and periods of 30 min and 80 min. These waves form the hydrodynamic boundary conditions of the numerical tsunami model of this study at the southern, southwestern and western model boundary. Table 1. Overview of the 18 tsunami scenarios discussed in this paper. The scenarios are based on three propagation directions, two wave periods and three wave amplitudes. LDN-wave from  Wave period  Wave amplitude  Scenario      2 m  S2S    30 min  4 m  S4S      6 m  S6S  South            2 m  S2L    80 min  4 m  S4L      6 m  S6L      2 m  SW2S    30 min  4 m  SW4S      6 m  SW6S  Southwest            2 m  SW2L    80 min  4 m  SW4L      6 m  SW6L      2 m  W2S    30 min  4 m  W4S      6 m  W6S  West            2 m  W2L    80 min  4 m  W4L      6 m  W6L  LDN-wave from  Wave period  Wave amplitude  Scenario      2 m  S2S    30 min  4 m  S4S      6 m  S6S  South            2 m  S2L    80 min  4 m  S4L      6 m  S6L      2 m  SW2S    30 min  4 m  SW4S      6 m  SW6S  Southwest            2 m  SW2L    80 min  4 m  SW4L      6 m  SW6L      2 m  W2S    30 min  4 m  W4S      6 m  W6S  West            2 m  W2L    80 min  4 m  W4L      6 m  W6L  View Large The chosen wave periods and amplitudes were derived from a previous comparison of the field data with multiple test runs based on various combinations of different wave periods and amplitudes, which are plausible considering the potential tsunami trigger mechanisms in the central and eastern Mediterranean Sea. Within this model calibration process, those combinations of wave periods and amplitudes were chosen, that can explain tsunami deposits found in the more seaward, the more landward and the intermediate parts of the study area. This approach yielded a possible magnitude range for tsunami waves that hit the Ambrakian Gulf in pre-/history and left the detected deposits. Earlier simulation studies revealed that tsunami waves with periods of between 30 min and 80 min combined with amplitudes of between 2 m and 6 m have probably occurred in the Mediterranean Sea and are therefore likely to occur again. For example, Pareschi et al. (2006a) simulated tsunami waves generated by a huge landslide of up to 30 km2 off Mount Etna in eastern Sicily about 8000 yr ago. The computed maximum wave amplitudes along the western model boundary in the current study, located in the central Ionian Sea (Fig. 1), are in the order of 5 m with periods of several tens of minutes. Períañez & Abril (2014a) simulated various landslide, volcanically and seismically generated tsunamis in the central and eastern Mediterranean. While most of the landslide and volcanically generated tsunamis yielded wave amplitudes smaller than 3 m in the central Ionian Sea, amplitudes of about 5 m and periods of more than 1 h were found for the tsunami connected to the 365 AD Crete earthquake with an estimated magnitude of Ms = 8.3. Similar wave amplitudes and periods are suggested by Yolsal-Çevikbilen & Taymaz (2012) for the same tsunami event. The 365 AD Crete earthquake probably resulted in a vertical shift of the west coast of Crete by almost 9 m and is associated with a fault length of 105 km, a width of 100 km, a slip of 16 m, a strike of 292.5° and a dip of 40° (Stiros 2010). All simulations are based on a two-part computational grid—an overall grid and a detailed, nested grid (Fig. 1). The overall grid has a resolution of 900 m and covers the entire domain of the DEM (Section 2.3). The boundary conditions for the various tsunami waves from the three directions south, southwest and west were defined along the corresponding open boundaries of the overall grid as shown in (Fig. 1). The nested grid with a resolution of 90 m stretches from 38.72°N to 39.24°N and from 20.56°E to 21.20°E ($$\widehat{=}$$ 58 km * 55 km). The boundary conditions of the nested model were determined by bilinear interpolation of the results computed at corresponding monitoring stations in the overall model (for further methodological details see Deltares 2014, pp. 677–683). For both, the overall and nested model, an alpha reflection parameter of 100 was applied along the open boundaries in order to minimize unintended wave reflections (Verboom & Slob 1984; pp. 205–209 Deltares 2014). Although a refinement factor of 10 from the overall to the nested grid (as applied here) might cause inaccuracy in several Delft3D-FLOW applications (typically factor 3 to 5), the current model showed a low sensitivity for this parameter based on comparison runs using a three-part computational grid with a resolution of 900 m, 300 m and 90 m respectively. Comparing computed inundation depths and flow velocities for various observation points within the 90 m grid domain revealed small differences between both grid configurations of less than 0.05 m and 0.1 m s−1 respectively. Correspondingly, differences in the computed erosion and sedimentation did not exceed few centimetres. These discrepancies are neglectable considering the computed absolute inundation depths and flow velocities in the order of up to 5 m and 7 m s−1 respectively (Section 3.2) as well as the simulated erosion and sedimentation with magnitudes of up to 4 m (Sections 3.3 and 3.4). Therefore, and in order to reduce the computational cost of the model, the two-part computational grid has been applied to all final simulation runs. The calculation step was 0.6 s in both, the overall and the nested model, which results in maximum Courant numbers Cmax of 0.38 and 0.84 respectively. These values are significantly below the recommended threshold of Cmax = 10, which, based on the implicit scheme of Delft3D-FLOW, is already sufficient to prevent numerical instability and inaccuracy in most applications including the simulation of long shallow-water waves (Deltares 2014, pp. 279, 280). All scenarios were computed for the time period characterized by significant changes in water level and significant sediment transport, which was between 8 h and 10 h. Based on field surveys, photographs and Google Earth satellite images (Google Inc. 2013), appropriate values of Manning’s n coefficient were determined in order to take account of the bottom roughness in the model domain. n-values ranged from 0.015 (seafloor, pastures) over 0.03 (moderately vegetated or developed land) to 0.07 (densely forested landscapes) (cf. Hills & Mader 1997, p. 388; Bryant 2008, pp. 45, 46). With the aim of simulating tsunami morphodynamics, a realistic off- and onshore sediment stratigraphy had to be determined. This was only done for the detailed, nested model because no significant sediment transport was found for the overall model owing to the large water depths. Delft3D-FLOW considers both, suspended load and bedload transport of non-cohesive sediments as well as suspended load transport of cohesive sediments (for further methodological details see Deltares 2014, pp. 323–373; Lesser et al.2004). In Delft3D-FLOW, cohesive sediments are defined by a sediment diameter of less than 0.063 mm and are no further distinguished, meaning that silt and clay are treated as one grain size, that is mud. In the case of non-cohesive sediments, however, any grain sizes between fine and coarse sand (0.063 mm $$\le \varnothing <$$ 2 mm) can be considered. Based on these specifications, the sediment stratigraphy of the nested model was determined according to sedimentary field data of the study area. The 151 onshore vibracores taken within this study (Section 2.1) revealed that the thickness of unconsolidated sediment exceeds 10 m in most places. Regarding the offshore area, an even larger thickness has to be assumed, for example, up to 45 m in the inner Ambrakian Gulf (Kapsimalis et al.2005, p. 404). Nevertheless, for both, the onshore and offshore parts of the nested model, an initial sediment layer thickness of 10 m was chosen since the maximum amount of sediment erosion was less than 4 m in all simulated scenarios. Statistical analyses of all 151 vibracores alongshore the Ambrakian Gulf led to an average onshore sediment composition for the upper 10 m below surface of about 67 per cent mud ($$\varnothing <0.063$$ mm), 9 per cent fine sand ($$\varnothing =0.1315$$ mm), 7 per cent middle sand ($$\varnothing =0.415$$ mm), 8 per cent coarse sand ($$\varnothing =1.315$$ mm) and 8 per cent gravel ($$\varnothing \ge 2$$ mm) (Table 2). Although the average grain size of gravel is 32.5 mm, a sediment diameter of 2 mm (maximum diameter of coarse sand) had to be chosen instead as this is the maximum grain size supported in Delft3D-FLOW. Judging by sediment cores, sediment samples and seismic profiles taken by Tziavos (1997), Poulos et al. (1995, 1999, 2005), Kapsimalis et al. (2005) and Karageorgis et al. (2006) in the Ambrakian Gulf, its lagoons and in the adjacent eastern Ionian Sea, an average offshore sediment composition of 80 per cent mud as well as 20 per cent fine sand was found for the upper 10 m below sea bottom (Table 2). Taking the average sediment fractions as a basis for the sediment stratigraphy in the nested model resulted in an initial onshore and offshore sediment layer of 10 m each, in which all corresponding grain sizes (ranging from mud to coarse sand) are homogeneously mixed. Test runs comparing such homogeneously mixed layers with actual stratified layers revealed that homogeneously mixed layers are most appropriate because they provide the model with those sediment fractions which are available in the area for redeposition by tsunami waves. Furthermore, the vibracore density is not high enough to derive a reliable comprehensive stratigraphy for the nested model and the pre-tsunami stratigraphy is not known. Table 2. Overview of the average on- and offshore sediment composition of the upper 10 m below surface/sea bottom for the nested model of the Ambrakian Gulf based on statistical analyses of 151 vibracores of the study area and sedimentary data taken from literature (see the text for further explanations). Grain size  Average onshore sediment composition of the upper 10 m below surface [per cent]  Average offshore sediment composition of the upper 10 m below sea bottom [per cent]  Cohesive      Mud  67  80  Non-cohesive      Fine sand  9  20  Middle sand  7  0  Coarse sand  8  0  Gravel  8  0  Grain size  Average onshore sediment composition of the upper 10 m below surface [per cent]  Average offshore sediment composition of the upper 10 m below sea bottom [per cent]  Cohesive      Mud  67  80  Non-cohesive      Fine sand  9  20  Middle sand  7  0  Coarse sand  8  0  Gravel  8  0  View Large Cohesive sediment erosion and deposition in Delft3D-FLOW was computed according to the Partheniades–Krone formulations (Partheniades 1965), whereas computation of non-cohesive sediment processes was based on the approach of van Rijn et al. (2001) (for further methodological details see Deltares 2014, pp. 323–373). The specific density and dry bed density of cohesive sediment was set to 2650 kg m−3 and 500 kg m−3 respectively, with a settling velocity of 0.25 mm s−1. The critical bed shear stress was 1000 N m−2 for sedimentation and 0.5 N m−2 for erosion of cohesive sediment, and a corresponding erosion parameter of 0.0001 kg m−2 s−1 was chosen. Independent of the grain size, the specific density and dry bed density of non-cohesive sediment were set to 2650 kg m−3 and 1600 kg m−3 respectively. As an important feature of tsunami inundation, the numerical model accounted for the effect of sediment concentration on fluid density by adding the mass of all transported sediment per unit water volume and subtracting the corresponding mass of displaced water (UNESCO 1981; Deltares 2014, pp. 324, 325). Furthermore, the hindered sediment settling velocity due to other sediment particles in highly loaded water was considered according to Richardson & Zaki (1997) (cf. Deltares 2014, pp. 324, 325), based on a reference density for hindered settling of 1600 kg m−3 for all grain sizes. Finally, changes in the on- and offshore topography due to sediment erosion and deposition were taken into account during the simulation by updating the elevation of the bed in each grid cell after every calculation step (Deltares 2014, pp. 366–369). 3 RESULTS 3.1 Distribution of tsunami deposits alongshore the Ambrakian Gulf The analysis and evaluation of all 151 vibracores revealed a consistent geographical distribution of tsunami deposits alongshore the Ambrakian Gulf. The average tsunami candidate layer thickness, which ranges from 0 m to 3.22 m (Fig. 5), generally decreases from west to east. Whereas the western Ambrakian Gulf shows the maximum values (frequently > 0.76 m), there are no values larger than 0.76 m in the inner and eastern gulf. Consequently, the average sediment accumulation per single tsunami event becomes smaller in an eastward direction, which points to a decreasing tsunami influence in the Ambrakian Gulf from the coast to regions further inland. This agrees with a general decrease of the relative number of event strata, that is, the number of tsunami candidate layers per coring metre per vibracore, from west to east (not illustrated). High values (> 0.29) are found around the entrance of the Ambrakian Gulf, including the Mazoma Lagoon, Aktio Headland, the Bay of Aghios Nikolaos and Lake Voulkaria, but significantly drop in the inner gulf with maximum 0.14 at its eastern shore. Figure 5. View largeDownload slide Average thickness of tsunami candidate layers per vibracore alongshore the Ambrakian Gulf. The identification of tsunami deposits in all 151 vibracores is based on several physical sedimentary criteria typical of tsunamites (Section 2.2 and Fig. 3). Furthermore, geochemical, microfaunal and palynological analyses conducted for selected vibracores in previous studies helped to detect event layers. The average tsunami layer thickness is compared with the computed sedimentation in the numerical model in Section 3.3. Figure 5. View largeDownload slide Average thickness of tsunami candidate layers per vibracore alongshore the Ambrakian Gulf. The identification of tsunami deposits in all 151 vibracores is based on several physical sedimentary criteria typical of tsunamites (Section 2.2 and Fig. 3). Furthermore, geochemical, microfaunal and palynological analyses conducted for selected vibracores in previous studies helped to detect event layers. The average tsunami layer thickness is compared with the computed sedimentation in the numerical model in Section 3.3. Fig. 6 shows the standard deviation of the tsunami layer thickness per vibracore in the study area. As can be seen, the standard deviation is generally small (<0.16 m), indicating a relatively persistent thickness of tsunami layers per vibracore. Therefore and because there are many vibracores with just one tsunami layer, the average tsunami layer thickness per vibracore (associated with several tsunami events in pre-/history) is an appropriate parameter to be compared with the sedimentation and erosion computed for single tsunami waves in the numerical model. Based on this comparison (Section 3.3), it is possible to draw conclusions about the general tsunami wave characteristics which are related to deposits in certain parts of the study area. Figure 6. View largeDownload slide Standard deviation of the tsunami candidate layer thickness per vibracore alongshore the Ambrakian Gulf. Figure 6. View largeDownload slide Standard deviation of the tsunami candidate layer thickness per vibracore alongshore the Ambrakian Gulf. 3.2 General inundation dynamics in the Ambrakian Gulf Prior to the comparison of field data and simulated morphodynamics, a short summary of the basic computed tsunami inundation dynamics in the Ambrakian Gulf of all scenarios is given. Fig. 7 exemplarily illustrates the inundation in Scenario W4L, which shows characteristic hydrodynamic features of both, scenarios with smaller and larger wave amplitudes/periods. Due to the leading trough of the simulated LDN-waves, in each scenario, a significant withdrawal of the shoreline between Kastrosikiá and Lefkada City is observed before the wave crest makes landfall (Fig. 7a; see Fig. 2 for places). Independent from the chosen boundary, all wave crests strike the Ambrakian Gulf from an almost western direction (Fig. 7b). Whereas this is the expected direction for waves from the west, waves from the southwest and south are strongly diffracted and refracted around Lefkada Island. Once the water depth drops below 1000 m, strong wave shoaling occurs, which, together with funnel effects in the inlet-shaped Ionian Sea towards the Ambrakian Gulf (Fig. 1), causes a significant increase in wave amplitude by up to factor three (Fig. 7b). This results in maximum run-up heights of more than 25 m in the valleys between Kastrosikiá and Mitikas. Figure 7. View largeDownload slide Snapshots (a–d) documenting stepwise the inundation of the Ambrakian Gulf in Scenario W4L and diagrams (top) showing the corresponding time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points. Figure 7. View largeDownload slide Snapshots (a–d) documenting stepwise the inundation of the Ambrakian Gulf in Scenario W4L and diagrams (top) showing the corresponding time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points. In all scenarios, tsunami waves completely inundate Aktio Headland (ground surface maximum 6 m a.s.l.) and subsequently enter the Ambrakian Gulf through the main waterway, the Strait of Preveza (Figs 7b and c). Simultaneously, tsunami waters in the moderate and extreme scenarios (wave amplitudes of 4 m and 6 m) reach the gulf via a small depression (ground surface maximum 14 m a.s.l.) to the west of the Mazoma Lagoon. Furthermore, each tsunami wave gets strongly funnelled in the Bay of Aghios Nikolaos and strikes Lake Voulkaria. Once the waves have reached the inner gulf, the area around Vonitsa and the narrow sandspit systems as well as the wide coastal flats in the northern gulf get inundated (Figs 7c and d). Due to flow divergence, the eastern gulf is hardly affected by tsunami waves; considerable flooding only occurs in the moderate and extreme scenarios in the Boukka coastal plain. Even after a simulation time of 10 h no significant backflow can be observed. Instead, tsunami waters gradually flow back into the Ionian Sea mainly via the Strait of Preveza. The diagrams at the top of Fig. 7 display the time series of the water elevation η or inundation depth d and of the depth averaged velocity u for four observation points in Scenario W4L. The most offshore observation point OBS ION clearly demonstrates the drop and subsequent rise in water level during the passage of the LDN-wave. The wave crest is associated with significant offshore currents in the order of 2.4 m s−1. The onshore observations points OBS AKT and OBS ANI 10 indicate inundation depths of up to 5 m connected with depth averaged flow velocities of almost 7 m s−1. Whereas Aktio Headland (OBS AKT) quickly falls dry after landfall of the wave crest, the inundation depth around Lake Voulkaria (OBS ANI 10) successively increases due to ongoing water inflow. Finally, observation point OBS AMB shows that the tsunami wave has a strong impact even on the western inner gulf, where the water elevation rises up to 3.4 m and again strong currents (up to 3.5 m s−1) can be observed. Generally, tsunamis from the west cause the strongest inundation of the Ambrakian Gulf, whereas waves from the south have the weakest impact (here and in the following measured by the inundation depth, flow velocity and area of inundation). Tsunamis originating from the southwestern model boundary result in an intermediate onshore response. Besides the propagation direction, the onshore tsunami response strongly depends on the wave period and wave amplitude; for both wave parameters, a positive correlation with the inundation intensity is found. As a result, Scenario W6L shows the strongest onshore response in the Ambrakian Gulf, whereas Scenario S2S shows the weakest impact. 3.3 Average vibracore tsunami layer thickness versus simulation results In the following, the simulated tsunami morphodynamics are described and compared with the average tsunami candidate layer thickness presented in Section 3.1. Maps of the Ambrakian Gulf show the simulated sediment erosion (blue colours) and accumulation (yellow to red colours) associated with tsunami inundation after a simulation time of 8 h to 10 h in each scenario (Figs 8–12; note the varying scale bar). This period covers most of the inundation process, including the main backflow. The maps further contain information on the average thickness of tsunami layers per vibracore already illustrated in Fig. 5. As pointed out in Section 3.1, this parameter allows a direct comparison with the sedimentary impact simulated for different single tsunami waves. Due to the large number of computed scenarios (18), only those scenarios are presented that show the best match with field data, namely Scenario W2S, S4S, W4S (not illustrated), W4L, S6S and W6L. Figure 8. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W2S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W2S is based on a single tsunami wave from a western direction with a wave amplitude of 2 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 8. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W2S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W2S is based on a single tsunami wave from a western direction with a wave amplitude of 2 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 9. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S4S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S4S is based on a single tsunami wave from a southern direction with a wave amplitude of 4 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 9. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S4S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S4S is based on a single tsunami wave from a southern direction with a wave amplitude of 4 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 10. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W4L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W4L is based on a single tsunami wave from a western direction with a wave amplitude of 4 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. See Fig. 7 for the simulated hydrodynamics of this scenario. Figure 10. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W4L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W4L is based on a single tsunami wave from a western direction with a wave amplitude of 4 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. See Fig. 7 for the simulated hydrodynamics of this scenario. Figure 11. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S6S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S6S is based on a single tsunami wave from a southern direction with a wave amplitude of 6 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 11. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario S6S for the Ambrakian Gulf after a total simulation time of 10 h. Scenario S6S is based on a single tsunami wave from a southern direction with a wave amplitude of 6 m and a wave period of 30 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 12. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W6L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W6L is based on a single tsunami wave from a western direction with a wave amplitude of 6 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. Figure 12. View largeDownload slide Comparison of the average thickness of tsunami candidate layers per vibracore (cf. Fig. 5) with computed sediment erosion and accumulation in the numerical model in Scenario W6L for the Ambrakian Gulf after a total simulation time of 10 h. Scenario W6L is based on a single tsunami wave from a western direction with a wave amplitude of 6 m and a wave period of 80 min. The dashed-dotted line indicates the area in which the average tsunamite thickness detected in the field and the simulated sedimentation of tsunami inundation are generally consistent. In all scenarios based on a low wave amplitude of 2 m, the sedimentary onshore response of tsunami waves is mainly restricted to Aktio Headland, the Bay of Aghios Nikolaos and the entrance of Lake Voulkaria. Only in Scenario W2L, significant sediment erosion and accumulation is also observed in the inner Lake Voulkaria and to the east of Aktio Headland. The best agreement with vibracore data of the Ambrakian Gulf is found for the simulated western tsunami with a wave period of 30 min (Scenario W2S; Fig. 8). Although the absolute sedimentation computed in this scenario (≤ 0.5 m) clearly lags behind the local average tsunami layer thickness per vibracore (≤ 2.67 m), both show the same spatial extent. Areas of simulated sediment erosion or no sediment accumulation in northwestern and northeastern Aktio Headland as well as to the north of Preveza coincide with a lack of event layers in the local vibracores. Sediment accumulation in the numerical model is mostly found at the northern tip and western shore of Aktio Headland, around the Saltini Lagoon, along the Bay of Aghios Andreas and at the entrance of Lake Voulkaria; for these areas also vibracore data indicates dominant tsunami sediment deposition. Regarding the offshore tsunami morphodynamics, in all scenarios based on a 2 m wave amplitude, significant sediment processes are not observed before the water depth drops below 20 m. Consequently, most of the deposited tsunami sediments originate from the shallow marine to littoral zone landward from the −20 m contour. Although Scenario W2S is found in close accordance with field data of Aktio Headland and the Bay of Aghios Nikolaos, Scenario S4S is the favourite scenario with regard to the entire western part of the Ambrakian Gulf, including the Mazoma Lagoon (Fig. 9). Again, vibracores containing no tsunami candidate layers are generally located in areas that show dominant erosion or no accumulation in the numerical model, whereas vibracores with thick event strata are mostly found in areas of simulated sediment accumulation. Although still too small compared to the average layer thickness determined per vibracore, the simulated thickness of tsunami deposits in Scenario S4S has significantly increased (up to 1 m). The morphodynamics computed for Scenario W4L (Fig. 10) are far stronger compared to the scenarios described above. Whereas the western part of the Ambrakian Gulf (except for the Saltini Lagoon and the Bay of Aghios Nikolaos) mostly show sediment erosion, significant sedimentation is particularly observed to the east of the Mazoma Lagoon and Aktio Headland, in the inner gulf and along the eastern shore of Lake Voulkaria. For the latter, the computed sediment thickness in the numerical model (up to 1.10 m) accurately replicates the average event layer thickness determined for the corresponding vibracores. Despite the clear sedimentary impact on the inner gulf, Scenario W4L is not appropriate to explain vibracore data around Vonitsa and the sandspit systems in the north. Here, the numerical model either suggests no significant sedimentation or even dominant erosion respectively. Strong morphodynamics in Scenario W4L are also observed offshore the Ambrakian Gulf, where sediment processes already start landwards from the −50 m contour. Scenario S6S (Fig. 11) shows similar areas of consistent model and field data as Scenario S4S but, beyond this, also explains tsunami sedimentation in the midst of Lake Voulkaria. However, dominant sediment erosion occurs along southwestern Aktio Headland in Scenario S6S, which is in contrast to relatively thick event layers found in the local vibracores. The simulated sediment thickness (up to 1.5 m) nearly matches the average tsunami candidate layer thickness per vibracore, where both, simulation results and vibracores indicate significant sediment accumulation. In general, results for Scenario S6S closely correspond to those for Scenario W4S, which is therefore not illustrated here. Sediment processes in Scenario W6L are on an unprecedented scale starting already landwards from the −60 m contour and reaching far into the inner and even eastern Ambrakian Gulf (Fig. 12). Comparable to Scenario W4L, in Scenario W6L the western part of the Ambrakian Gulf is mostly characterized by sediment erosion, which is generally not consistent with vibracore data. Instead, a good accordance between simulated tsunami morphodynamics and field traces is found for the eastern and southern shore of Lake Voulkaria and for the inner Ambrakian Gulf around Vonitsa. For the latter, Scenario W6L is the only scenario that plausibly explains tsunami event strata detected in the field. Nevertheless, simulation results again do not match vibracore data of the sandspit systems, the delta plain in the north and the Boukka coastal plain at the very eastern end of the gulf (the latter two locations are not displayed in Figs 8–12; see Fig. 5). 3.4 Correlating vibracore tsunami layers versus simulation results Fig. 13 shows the location of three vibracore transects, for which correlating tsunami candidate layers were identified based on geochronological data (in particular radiocarbon ages). These layers point to three of several pre-/historical tsunami events in the Ambrakian Gulf summarized in Section 1. Vibracore transect A, located in southwestern Aktio Headland, revealed correlating tsunami deposits associated with the 840 cal AD tsunami. Event layers associated with the 1000 cal BC tsunami were found in vibracore transect B in the area of Lake Voulkaria. The third tsunami event, which occurred in the 4th century AD, is documented by corresponding deposits in vibracore transect C at the southern shore of Lake Voulkaria (for a detailed sedimentary analysis of the three vibracore transects see Vött et al.2007 (transect A), Vött et al.2006 (transect B), and Vött et al.2009a (transect C)). Section 3.3 demonstrated that for all three areas, some of the simulated scenarios show a good match with the average tsunami layer thickness. These scenarios are also appropriate to explain the deposition of the correlating tsunami layers identified in the three vibracore transects. Figure 13. View largeDownload slide Location of vibracore transects A, B and C, for which the thickness and sedimentary composition of correlating tsunami candidate layers are compared with the simulated tsunami layers in different scenarios (Figs 14–19). Figure 13. View largeDownload slide Location of vibracore transects A, B and C, for which the thickness and sedimentary composition of correlating tsunami candidate layers are compared with the simulated tsunami layers in different scenarios (Figs 14–19). In the following, both the thickness and sedimentary composition (portions of gravel to mud) of the correlating vibracore tsunami layers of each tsunami event are directly compared with the simulated tsunami layers and erosion in the scenarios. Although several scenarios presented in Section 3.3 are appropriate for this comparison, only those with the closest match are shown. In Figs 14, 16 and 18, for each vibracore of the three transects, the corresponding tsunami layer is displayed as a bar, the height of which gives the layer thickness and the white to black coloured segments, sorted by grain size, the portions of gravel, coarse sand, middle sand, fine sand and mud. Accordingly, the graphs in the background illustrate the thickness and sedimentary composition of the simulated tsunami layers as well as the simulated tsunami erosion in the best matching scenarios. In addition, scatter diagrams in Figs 15, 17 and 19 show the portions of grain sizes in the vibracore tsunami layers of each transect versus the portions of grain sizes in the tsunami layers as simulated in the scenarios. The closer the points lay on the grey dashed line, the higher the correlation between the tsunami sediment composition in the vibracores and in the numerical model. Figure 14. View largeDownload slide Comparison of event deposits of the 840 cal AD tsunami in vibracore transect A in southwestern Aktio Headland (Fig. 13; Vött et al.2007) with the tsunami layers simulated in Scenario S4S (top), W4L (centre) and W6L (bottom). Tsunami deposits in vibracores ANI 2–AKT 11 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 14. View largeDownload slide Comparison of event deposits of the 840 cal AD tsunami in vibracore transect A in southwestern Aktio Headland (Fig. 13; Vött et al.2007) with the tsunami layers simulated in Scenario S4S (top), W4L (centre) and W6L (bottom). Tsunami deposits in vibracores ANI 2–AKT 11 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 15. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect A (Fig. 14) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S, W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 15. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect A (Fig. 14) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S, W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Vibracore transect A (ANI 2–AKT 11) stretches for almost 3000 m in a northwestern direction across relatively flat terrain in southwestern Aktio Headland (Figs 13 and 14). Tsunami layers of the 840 cal AD event are between 0.13m and 2.33m thick and generally become thinner towards the northwest. They mainly consist of middle and fine sand. Considerable amounts of mud and gravel only occur in vibracores ANI 2 and AKT 3 respectively. Tsunami layers in transect A are best represented by Scenarios S4S, W4L and W6L. In particular Scenarios W4L and W6L show a very similar spatial distribution of areas with dominant and limited deposition as the vibracore transect. However, the model does not reproduce the muddy tsunami deposits found in ANI 2 but instead indicates erosion. All scenarios, especially Scenarios S4S and W4L, underpredict the absolute tsunami layer thickness determined in the vibracores. Despite this underprediction, the model reproduces well the sedimentary composition of the identified tsunami layer in transect A. This is demonstrated by Fig. 15 showing similar portions of the five grain sizes for all coring locations except for vibracore ANI 2 in all scenarios and for vibracores AKT 2–AKT 11 in Scenario S4S. Generally, both the vibracores and the numerical model indicate dominant fine sand in the first part of the transect, a mix of fine and middle sand in the central part followed by dominant coarse material between 2000 m and 2500 m and again finer material at the end of the transect. Vibracore transect B (ANI 4–VOUL 1) is about 4000 m long and reaches from the village of Aghios Nikolaos to the inner Lake Voulkaria (Figs 13 and 16). The highest point (4.5 m a.s.l.) lies on a small rise between 500 m and 1000 m of the transect at the seaward entrance of Lake Voulkaria. The 1000 cal BC tsunami layer has a thickness of between 0.19 m and 2.07 m and shows a relatively heterogeneous sedimentary composition. Both, the vibracores and the numerical model (Scenarios S4S and S6S) indicate dominant tsunami deposition at the seaward and landward foot of the small rise between 500 m and 1000 m, which seem to act as a natural obstacle during inundation. The rise itself is strongly eroded in both scenarios. All in all, the first 1500 m of the transect are better represented by Scenario S4S than by Scenario S6S. The opposite is true for the area around vibracore VOUL 1 in the inner Lake Voulkaria, where only Scenario S6S can explain significant tsunami sedimentation. As was discussed earlier for transect A, also for transect B the model underpredicts the tsunami layer thickness observed in the field but at the same time shows a good match regarding the sedimentary composition. According to Fig. 17, the simulated and the observed tsunami layer is characterized by a heterogeneous composition in the first 200 m, dominant fine sand around the small rise between 500 m and 1000 m, followed by coarser material at ANI 7 and ANI 10 and finally fine sand and mud in the inner Lake Voulkaria. Figure 16. View largeDownload slide Comparison of event deposits of the 1000 cal BC tsunami in vibracore transect B in the area of Lake Voulkaria (Fig. 13; Vött et al.2006) with the tsunami layers simulated in Scenarios S4S (top) and S6S (bottom). Tsunami deposits in vibracores ANI 4–VOUL 1 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 16. View largeDownload slide Comparison of event deposits of the 1000 cal BC tsunami in vibracore transect B in the area of Lake Voulkaria (Fig. 13; Vött et al.2006) with the tsunami layers simulated in Scenarios S4S (top) and S6S (bottom). Tsunami deposits in vibracores ANI 4–VOUL 1 are displayed as bars, the simulated tsunami sedimentation and erosion as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 17. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect B (Fig. 16) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S and S6S. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 17. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect B (Fig. 16) versus the portions of grain sizes in the tsunami layers simulated in Scenarios S4S and S6S. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Vibracore transect C (PAL 6–PAL 46) gives evidence of the 4th century AD tsunami at the southern shore of Lake Voulkaria (Figs 13 and 18). Over a distance of about 700 m, tsunami deposits 0.16–0.82 m thick can be observed in the field. In the numerical model, Scenario W4L and especially Scenario W6L predict significant tsunami sedimentation for this transect, although the absolute thickness found in the vibracores is not reached in any of the scenarios. Nevertheless, the field data and the scenarios show a corresponding sedimentary composition of the tsunami deposits, which are dominated by fine sand and mud (Fig. 19). Figure 18. View largeDownload slide Comparison of event deposits of the 4th century AD tsunami in vibracore transect C at the southern shore of Lake Voulkaria (Fig. 13; Vött et al.2009a) with the tsunami layers simulated in Scenarios W4L (left) and W6L (right). Tsunami deposits in vibracores PAL 6–PAL 46 are displayed as bars, the simulated tsunami sedimentation as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 18. View largeDownload slide Comparison of event deposits of the 4th century AD tsunami in vibracore transect C at the southern shore of Lake Voulkaria (Fig. 13; Vött et al.2009a) with the tsunami layers simulated in Scenarios W4L (left) and W6L (right). Tsunami deposits in vibracores PAL 6–PAL 46 are displayed as bars, the simulated tsunami sedimentation as graphs. The portions of the grain sizes gravel to mud are shown by white to black coloured segments. *Despite the skipping y-axis, the segments of the bar display the portion of each grain size in relation to the actual height of the bar in the figure. Figure 19. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect C (Fig. 18) versus the portions of grain sizes in the tsunami layers simulated in Scenarios W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. Figure 19. View largeDownload slide Scatter diagrams showing the portions of grain sizes in the vibracore tsunami layers of transect C (Fig. 18) versus the portions of grain sizes in the tsunami layers simulated in Scenarios W4L and W6L. The closer the points lay on the grey dashed line, the higher the correlation between the measured and simulated sediment composition of the tsunami deposits. 4 DISCUSSION Based on the presented simulation results, a number of systematic hydro- and morphodynamic features of tsunami inundation in the Ambrakian Gulf can be drawn together: Tsunamis from a western direction show the strongest hydro- and morphodynamic impact on the Ambrakian Gulf. This is closely related to strong funnelling of tsunami waters in the eastern Ionian Sea between the islands of Cephalonia, Lefkada and Corfu for waves from the west (Fig. 1). In contrast, funnel effects are less distinct for waves from a southwestern and especially southern direction (cf. Vött et al.2008, pp. 107, 119). Furthermore, the central Ionian Islands, in particular Lefkada, Cephalonia and Zakynthos, seem to give some protection for the Ambrakian Gulf against waves from these two directions. The sedimentary response of all shorter period tsunami waves (T = 30 min) is concentrated on a comparatively narrow area mainly covering the western part of the Ambrakian Gulf (Figs 8, 9 and 11). These waves result in strong sedimentation especially between the Mazoma Lagoon and Lake Voulkaria. By contrast, for all longer period waves (T = 80 min), significant sediment processes are observed for a much wider area stretching from the nearshore zone (landwards from the −60 m contour) up to the eastern shore of the Ambrakian Gulf (Figs 10 and 12). Associated with the larger reach of the longer period tsunami waves, here, the western part of gulf, including the area of the Mazoma Lagoon, large parts of Aktio Headland and the entrance of Lake Voulkaria, is dominated by erosion and the main sedimentation is found further east or clearly offshore in the Ionian Sea. Nevertheless, maximum values of erosion and sedimentation are only slightly larger for the longer period than for the shorter period waves, the same wave amplitude provided. Simulated sediment accumulation—where not influenced by natural obstacles (e.g. cliffs or hills)—is clearly characterized by thinning- and fining-landward, features typical of tsunami deposits (Section 2.2). This is due to the general decrease of inundation depth and flow velocity (Section 3.2) and by this of the transport capacity of inundating tsunami waters. Thinning- and fining-landward is most obvious in Scenarios W4L (Fig. 10) and W6L (Fig. 12) along the eastern shore of Lake Voulkaria (see also transects B and C in Figs 16 and 18) and in Scenario W6L also in the area of Vonitsa. The main sedimentary response of tsunamis in the Ambrakian Gulf arises from the wave run-up and not from the backflow. This is connected to the extensive flat flooding zone in the northern Ambrakian Gulf and therefore the gentle slope of the water surface, which prevents high flow velocities during the backflow. Generally, the numerical model accurately reflects the decreasing tsunami influence from the western to the eastern Ambrakian Gulf, as is indicated by the sedimentary field data (Section 3.1). However, there is no specific simulated scenario which can totally explain the spatial distribution of the average tsunami candidate layer thickness per vibracore in the entire study area at the same time. This is because the vibracores contain tsunami layers that point to different tsunami events in the past, which were probably related to various sources and divergent sedimentary responses (Vött et al.2006, 2007, 2008, 2009a, 2010, 2011; May et al.2007, 2012a; Section 1). Therefore, the average layer thickness used for the comparison with simulation results in Section 3.3 does not reflect the actual sediment deposition due to a particular tsunami event. Nevertheless, there are several scenarios that reproduce well the average tsunami layer thickness in certain parts of the study area. Generally, the scenarios based on shorter period tsunami waves (T = 30 min) from the south or west with any wave amplitude agree with the average layer thickness found in the western Ambrakian Gulf, including the area of the Mazoma Lagoon, large parts of Aktio Headland, the Bay of Aghios Andreas and the entrance to Lake Voulkaria (Figs 8, 9 and 11). In contrast, longer period waves (T = 80 min) with amplitudes of 4 m and 6 m and from a western direction, are the favourite tsunami scenarios in respect to the average layer thickness in southwestern Aktio Headland, around Lake Voulkaria and in the area of Vonitsa located in the inner gulf (Figs 10 and 12). The sedimentary response of tsunami waves from the southwest, however, generally does not match the spatial distribution of the average tsunami layer thickness detected in the field. Those scenarios that show a good fit with the average tsunami layer thickness in the area of the three vibracore transects presented in Section 3.4 do also agree with the single, correlating tsunami layers identified in these transects. Consequently, different scenarios might be appropriate to explain the deposition of the correlating layer in each transect. Moreover, some scenarios even agree with the correlating tsunami layers of two different transects, although these layers originate from different events in the past. The marine to shallow marine origin of detected tsunami deposits in the study area can be explained by all simulated scenarios, which show that most sediment in the numerical model originate from the shallow marine to littoral zone landward from the −60 m contour (cf. e.g. Morton et al.2007, pp. 187–194; Apotsos et al.2011b, p. 8; Apotsos et al.2011c, p. 9; Sugawara et al.2014a, p. 299; Sugawara et al.2014b, p. 19). Generally, the numerical model accurately reproduces the sedimentary composition of the detected tsunami deposits as is demonstrated for the correlating tsunami layers in the three transects in the western Ambrakian Gulf in Section 3.4. This also includes thinning- and fining-landward tendencies frequently observed in related tsunami event strata at adjacent vibracoring sites (see especially transects B and C in Figs 16 and 18). Besides these common features of simulation results and field traces, the numerical model also reproduces local tsunami inundation dynamics for particular key sites in the western Ambrakian Gulf which were reconstructed on the basis of vibracore data in previous studies: According to Vött et al. (2008, p. 120), the central and northern part of the former Plaka sandspit system, located to the southwest of the Bay of Aghios Nikolaos (Fig. 2), was completely destroyed by tsunami waves over a distance of more than 5 km. This closely agrees with all simulated scenarios showing dominant erosion in this area. Simulated morphodynamics corroborate strong local erosional forces during tsunami landfall in the western, northern and eastern part of Aktio Headland, as documented by distinct erosional contacts in vibracores (Vött et al.2007, p. 54). Vött et al. (2007; 2011, pp. 236, 237) suggest that tsunami waters struck the Bay of Aghios Nikolaos, inundated Lake Voulkaria and subsequently flooded the Bay of Palairos located almost 5 km to the south (Fig. 2). Inundation of Lake Voulkaria from a western direction is reproduced by all simulated scenarios. Scenarios W4L and W6L (Figs 10 and 12) further show inundation and significant morphodynamics in the adjacent Bay of Palairos. According to Vött et al. (2009a, p. 30), the so-called Cleopatra Canal—connecting the Bay of Aghios Nikolaos with Lake Voulkaria—was originally dug out by tsunami waves and was filled again during a subsequent tsunami event. Erosion of the canal can easily be explained by Scenarios W4S, W4L, S6S and W6L (Figs 10–12), whereas the sediment infill might be reflected by Scenarios W2S and S4S (Figs 8 and 9). With some exceptions, it is rather the spatial extent of deposits that agree with vibracore data than the absolute thickness of simulated sedimentation. The latter usually lags behind the average tsunami candidate layer thickness determined per vibracore and the thickness of the correlating tsunami layers in the three transects. This may particularly be related to the following reasons: If the upper part of an event layer has been eroded and subsequently covered by younger tsunami deposits, the complete sequence may be interpreted as only one tsunami candidate layer resulting in an overestimation of the average layer thickness/thickness of the correlating tsunami layers in the transects. This may further explain the exceptionally large thicknesses more than 1 m found for some vibracores especially in the western part of the Ambrakian Gulf. The absolute thickness of simulated sedimentation is affected by various parameters in the numerical model, including the predetermined bottom roughness, values of the critical bed shear stress, sediment stratigraphy as well as the small-scale topography (cf. Apotsos et al.2009; Apotsos et al.2011b, pp. 4, 16; Apotsos et al.2011c, p. 9; Li et al.2012a, 2014). For example, a general overestimation of the critical bed shear stress would result in less sediment entrainment by waves, and consequently in less sediment deposition in the numerical model than observed in reality. In contrast to the absolute thickness, the spatial extent of tsunami deposits in both, the numerical model and the field is more resistant to modifications. This is because even a very thin event strata would still be indicative of tsunami influence. Therefore, the spatial extent of tsunamites is generally more appropriate for the comparison of simulation results and field data than the absolute thickness (cf. Li et al.2014, p. 2272). Erosion particularly due to subsequent tsunami impacts does not only strongly affect the absolute thickness but also the general spatial distribution of tsunami deposits in the field (e.g. Wheatcroft 1990; Wheatcroft & Drake 2003; Szczuciński 2012; Spiske et al.2013). In extreme cases, tsunami event layers can be completely removed from the geological record. This is an important process to be considered in this study because it might contribute to the inconsistency between the simulation result and field data. Especially the area of the Mazoma Lagoon, Aktio Headland and the entrance of Lake Voulkaria are susceptible to strong tsunami erosion as is indicated by several simulated scenarios and distinct erosional unconformities in the local vibracores (cf. Vött et al.2007, p. 54). All simulated scenarios are based on single tsunami waves in order to investigate the morphodynamic effects of different wave periods, wave amplitudes and propagation directions. However, most tsunamis appear in the form of wave trains consisting of a series of waves. Subsequent waves during one and the same tsunami event may redeposit sediment that was released by previous waves (cf. e.g. Apotsos et al.2011b, p. 15; Sugawara et al.2014a, pp. 298, 299), which would result in a sediment distribution not replicated by any of the simulated scenarios in this study. Given the wide time span of detected tsunami impacts and associated deposits in the Ambrakian Gulf of about 8000 yr, long-term changes in the local topography as described in Section 1 may further account for discrepancies between simulation results and field data (cf. Röbke et al.2015, 2016). This is particularly true for the northern and eastern shores of the Ambrakian Gulf, which experienced major topographical changes due to tectonic and fluvio-deltaic processes. Here, no scenarios can explain tsunami deposits detected in the field. Finally, the comparison of morphodynamic simulation results with sedimentary field data of the study area is limited by the spatial resolution of the numerical model. Whereas the simulations yield the average sedimentation for 90 m by 90 m grid cells, vibracores represent the geological record for a specific coring site only. This may result in further inconsistency between model and field (cf. Sugawara et al.2014b, p. 28). 5 CONCLUSIONS According to the general decrease of tsunami wave energy during the landfall, sedimentary vibracore data indicates a decreasing tsunami influence from the western to the eastern Ambrakian Gulf for both, the frequency and the intensity of events in the last 8000 yr. This is accurately reflected by the hydro- and morphodynamic numerical model showing that the sedimentary tsunami response decreases from the west to the east and that the inner and eastern gulf is only affected in extreme events. As indicated by the field data itself, several different tsunami wave scenarios are needed to explain detected tsunami deposits in different parts of the study area. Generally, shorter period tsunami waves (T = 30 min) from the south or west with any wave amplitude (2 m, 4 m, 6 m) are highly consistent with field data of the western Ambrakian Gulf. In contrast, longer period tsunamis (T = 80 min) from a western direction with 4 m to 6 m amplitudes show the best agreement with vibracore data collected in southwestern Aktio Headland and in the more central parts of the Ambrakian Gulf including Lake Voulkaria. Based on the comparison of the simulation results with the average tsunami layers thickness and with the thickness and sedimentary composition of correlating tsunami layers in three vibracore transects, 4 of total 18 scenarios could be derived that reproduce well the tsunami deposits found in southwestern Aktio Headland and in the area of Lake Voulkaria. These are Scenarios S4S, W4L, S6S and W6L, which all represent relatively strong tsunami events. Simulated tsunamis from the southwest generally do not match tsunami field traces. Therefore and because most potential tsunami sources in the central and eastern Mediterranean are rather located to the west (Mount Etna, Calabrian Arc) and south or southeast (Hellenic Trench) of the Ambrakian Gulf, it can be assumed that no major tsunamis from a southwestern direction struck the area in the past. The fact, that tsunami field traces in the eastern and northern parts of the Ambrakian Gulf cannot be explained by any of the simulated scenarios further leads to the conclusion that these deposits might relate to tsunamis generated along normal faults located in the gulf. Moreover, it has to be considered that the eastern and northern gulf experienced major topographical changes during the Holocene. The simulated scenarios do not only replicate the general spatial distribution of the tsunami candidate layers; they further show a number of essential sedimentary features typical of tsunamites, which were also detected in many vibracores of the study area. Such features are thinning- and fining-landward as well as the marine to shallow marine origin of tsunami deposits. The excellent match between the sedimentary composition of the simulated tsunami layers and the correlating tsunami layers in the three vibracore transects particularly results from the realistic initial bed composition of the model, which was derived from actual stratigraphic data of the study area. Finally, the numerical model accurately reproduces local tsunami inundation dynamics which were derived from field data for particular key sites in the western Ambrakian Gulf by previous sedimentary studies. Nevertheless, there are some discrepancies between simulation results and field traces. In particular, the computed sediment thickness in the numerical model usually lags behind the average thickness of tsunami candidate layers per vibracore and the thickness of the correlating tsunami layers in the transects. Also, some field traces cannot be explained by any of the simulated scenarios. Such inconsistencies may arise from (i) the comparison of single simulated tsunami waves with the average layer thickness per vibracore, (ii) tsunami candidate layers that give the impression of being related to a single tsunami event but actually relate to two or more events, (iii) false assumptions about spatially highly variable hydro- and morphodynamic parameters in the numerical model (e.g. bottom roughness or critical bed shear stress), (iv) hydrodynamic boundary conditions that do not represent all possible/realistic tsunami waves in the study area (e.g. single tsunami waves instead of tsunami wave trains), (v) changes in the local topography during the last 8000 yr, which are not considered in the numerical model and finally from (vi) the limited spatial resolution of the model. Especially inconsistencies arising from (i) and (v) could be reduced in future studies, provided that the age model of correlating tsunami candidate layers can be extended (cf. Röbke et al.2015, 2016). Then, event strata of particular pre-/historical tsunami impacts could be compared on a large scale with numerical simulations that consider the contemporary topography of the study area. Moreover, an improved age model provided and based on the insights into the wave characteristics gained in this study, specific source mechanisms (e.g. landslides offshore Mount Etna or specific focal mechanisms along the Helenic Trench) could be considered in future investigations. As demonstrated in this study, comparing and calibrating numerical models with sedimentary field data and computing sediment processes significantly increase the quality of numerical tsunami simulation results as a basis for modern tsunami risk assessment. Given the frequency of tsunami impacts in the past as indicated by sedimentary field data and historical accounts and considering the computed tsunami scenarios, a high tsunami risk has to be derived for the Ambrakian Gulf. Acknowledgements This research was supported by the Studienstiftung des deutschen Volkes (Bonn) in the form of a doctoral scholarship. We further acknowledge financial support from the German Research Foundation (Bonn) under grant numbers VO 938/1-1, VO 938/1-3, VO 938/2-1 and VO 938/3-1. Special thanks are due to Peter Fischer and Dieter Kelletat who provided insight and expertise that greatly assisted this research. REFERENCES Ad Hoc-AG Boden, 2005. Bodenkundliche Kartieranleitung , 5th edn, Schweizerbart Science Publishers. Apotsos A., Jaffe B., Gelfenbaum G., Elias E., 2009. Modeling time-varying tsunami sediment deposition, in Proceedings of Coastal Dynamics, Tokyo, 2009 , pp. 1– 15, doi:10.1142/9789814282475_0037. Apotsos A., Buckley M., Gelfenbaum G., Jaffe B., Vatvani D., 2011a. Nearshore tsunami inundation model validation: toward sediment transport applications, Pure appl. Geophys. , 168, 2097– 2119. https://doi.org/10.1007/s00024-011-0291-5 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., 2011b. Process-based modeling of tsunami inundation and sediment transport, J. geophys. Res. , 116, 1– 20. https://doi.org/10.1029/2010JF001797 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., Watt S., Peck B., Buckley M., Stevens A., 2011c. Tsunami inundation and sediment transport in a sediment-limited embayment on American Samoa, Earth-Sci. Rev. , 107, 1– 11. https://doi.org/10.1016/j.earscirev.2010.11.001 Google Scholar CrossRef Search ADS   Apotsos A., Jaffe B., Gelfenbaum G., 2011d. Wave characteristic and morphologic effects on the onshore hydrodynamic response of tsunamis, Coastal Eng. , 58, 1034– 1048. https://doi.org/10.1016/j.coastaleng.2011.06.002 Google Scholar CrossRef Search ADS   Apotsos A., Gelfenbaum G., Jaffe B., 2012. Time-dependent onshore tsunami response, Coastal Eng. , 64, 73– 86. https://doi.org/10.1016/j.coastaleng.2012.01.001 Google Scholar CrossRef Search ADS   Benetatos C., Kiratzi A., Papazachos B., Karakaisis G., 2004. Focal mechanisms of shallow and intermediate depth earthquakes along the Hellenic arc, J. Geodyn. , 37, 253– 296. https://doi.org/10.1016/j.jog.2004.02.002 Google Scholar CrossRef Search ADS   Bondevik S., Løvholt F., Harbitz C., Mangerud J., Dawson A., Svendsen J., 2005. The Storegga slide tsunami—comparing field observations with numerical simulations, Mar. Pet. Geol. , 22, 195– 208. https://doi.org/10.1016/j.marpetgeo.2004.10.003 Google Scholar CrossRef Search ADS   Borrero J., Weiß R., Okal E., Hidayat R., Suranto Arcas D., Titov V., 2009. The tsunami of 2007 September 12, Bengkulu province, Sumatra, Indonesia: post-tsunami field survey and numerical modelling, Geophys. J. Int. , 178, 180– 194. https://doi.org/10.1111/j.1365-246X.2008.04058.x Google Scholar CrossRef Search ADS   Brockmüller S., Vött A., Brückner H., Handl M., Schriever A., 2005. Die holozäne entwicklung der küstenebene von boukka am ambrakischen golf (nordwestgriechenland), in Schriften des Arbeitskreises Landes- und Volkskunde , Vol. 4. Bryant E., 2008. Tsunami: The Underrated Hazard , 2nd edn, Springer. Chacón-Barrantes S., Narayanan R., Mayerle R., 2013. Several tsunami scenarios at the North Sea and their consequences at the German Bight, J. Tsunami Soc. Int. , 32, 8– 28. Clews J., 1989. Structural controls on basin evolution: Neogene to Quaternary of the Ionian zone, western Greece, J. Geol. Soc. , 146, 447– 457. https://doi.org/10.1144/gsjgs.146.3.0447 Google Scholar CrossRef Search ADS   Dawson A., 1994. Geomorphological effects of tsunami run-up and backwash, Geomorphology , 10, 83– 94. https://doi.org/10.1016/0169-555X(94)90009-4 Google Scholar CrossRef Search ADS   Dawson A., 1999. Linking tsunami deposits, submarine slides and offshore earthquakes, Quat. Int. , 60, 119– 126. https://doi.org/10.1016/S1040-6182(99)00011-7 Google Scholar CrossRef Search ADS   Dawson A., Shi S., 2000. Tsunami deposits, Pure appl. Geophys. , 157, 875– 897. https://doi.org/10.1007/s000240050010 Google Scholar CrossRef Search ADS   Dawson A., Stewart I., 2007a. Tsunami deposits in the geological record, Sedimentary Geol. , 200, 166– 183. https://doi.org/10.1016/j.sedgeo.2007.01.002 Google Scholar CrossRef Search ADS   Dawson A., Stewart I., 2007b. Tsunami geoscience, Prog. Phys. Geogr. , 31, 575– 590. https://doi.org/10.1177/0309133307087083 Google Scholar CrossRef Search ADS   Dean R., Dalrymple R., 1991. Water Wave Mechanics for Engineers and Scientists , 2nd edn, World Scientific. Google Scholar CrossRef Search ADS   Deltares, 2014. Delft3d. simulation of multi-dimensional hydrodynamic flows and transport phenomena, including sediments . User manual, version 3.15.34158, http://oss.deltares.nl/documents/183920/185723/Delft3D-FLOW_User_Manual.pdf?version=1.0, 2014-05-28. Dominey-Howes D., Humphreys G., Hesse P., 2006. Tsunami and palaeotsunami depositional signatures and their potential value in understanding the late-Holocene tsunami record, Holocene , 16, 1095– 1107. https://doi.org/10.1177/0959683606069400 Google Scholar CrossRef Search ADS   Eagle Ray Publications, 2010. Pilot chart; map sheet: Pc 18, n. lefkáda to amvrakikós kólpos , 1: 93  213, Voula. Faccenna C., Becker T., Lucente F., Jolivet L., Rossetti F., 2001. History of subduction and back-arc extension in the central Mediterranean, Geophys. J. Int. , 145, 809– 820. https://doi.org/10.1046/j.0956-540x.2001.01435.x Google Scholar CrossRef Search ADS   Ferentinos G., 1991. Offshore geological hazards in the Hellenic Arc, Mar. Geotechnol. , 9, 261– 277. https://doi.org/10.1080/10641199009388244 Google Scholar CrossRef Search ADS   Ferentinos G., 1992. Recent gravitative mass movements in a highly tectonically active arc system: the Hellenic Arc, Mar. Geol. , 104, 93– 107. https://doi.org/10.1016/0025-3227(92)90086-W Google Scholar CrossRef Search ADS   Floth U., Vött A., May S., Brückner H., Brockmüller S., 2009. Geo-scientific evidence versus computer models of tsunami landfall in the Lefkada coastal zone (NW Greece), in Marburger Geographische Schriften , Vol. 145, pp. 140– 156, eds Vött A., Brückner H., Marburger Geographische Gesellschaft. Floth U., Vött A., May S., Brückner H., Brockmüller S., 2010. Estimating tsunami hazards between Lefkada and Preveza, NW Greece, by means of computer modeling, in Coastline Reports , Vol. 16, pp. 11– 14, eds Schwarzer K., Schrottke K., Stattegger K., EUCC-Deutschland. Floth U., Vött A., May S., Brockmüller S., Brückner H., 2013. Extreme wave events in the coastal zone of Lefkada and Akarnania (NW Greece) and their possible influence on civil structures. A comparison of field evidence and modeling results, in Akarnanien-Forschungen , Vol. 1, 31– 40. Fujiwara O., 2008. Bedforms and sedimentary structures characterizing tsunami deposits, in Tsunamiites—Features and Implications , pp. 51– 62. eds Shiki T., Tsuji Y., Yamazaki T., Minoura K., Elsevier. Google Scholar CrossRef Search ADS   Gelfenbaum G., Vatvani D., Jaffe B., Dekker F., 2007. Tsunami inundation and sediment transport in vicinity of coastal mangrove forest, in Coastal Sediments ’07 , 1117– 1128. doi:10.1061/40926(239)86. Ghionis G., Poulos S., Verykiou E., Karditsa A., Alexandrakis G., Andris P., 2015. The impact of an extreme storm event on the barrier beach of the Lefkada lagoon, NE Ionian Sea (Greece), Mediterranean Mar. Sci. , 16, 562– 572. https://doi.org/10.12681/mms.948 Google Scholar CrossRef Search ADS   Google Inc., 2013. Google earth, version 7.1.2.2041, computer software, https://www.google.com/earth/. Gusman A., Tanioka Y., Takahashi T., 2012. Numerical experiment and a case study of sediment transport simulation of the 2004 Indian Ocean tsunami in Lhok Nga, Banda Aceh, Indonesia, Earth Planets Space , 64, 817– 827. https://doi.org/10.5047/eps.2011.10.009 Google Scholar CrossRef Search ADS   Hill J., Collins G., Avdis A., Kramer S., Piggott M., 2014. How does multiscale modelling and inclusion of realistic palaeobathymetry affect numerical simulation of the Storegga slide tsunami? Ocean Modelling , 83, 11– 25. https://doi.org/10.1016/j.ocemod.2014.08.007 Google Scholar CrossRef Search ADS   Hills J., Mader C., 1997. Tsunami produced by the impacts of small asteroids, Ann. New York Acad. Sci. , 822, 381– 394. https://doi.org/10.1111/j.1749-6632.1997.tb48352.x Google Scholar CrossRef Search ADS   Imray Laurie Norie Wilson Ltd, 2011. Imray chart; map sheet: G 11, greece-ionian sea, north ionian islands, nísos kérkira to nísos levkas , 1: 185 000, Saint Ives. IOC—Intergovernmental Oceanographic Commission t. BODC—British Oceanographic Data Centre, 2003. Centenary edition of the GEBCO Digital Atlas , published on CD-ROM, Liverpool, http://www.gebco.net/data_and_products/gridded_bathymetry_data/, 2015-01-12. Jahns S., 2005. The Holocene history of vegetation and settlement at the coastal site of Lake Voulkaria in Acarnania, western Greece, Vegetation History and Archaeobotany , 14, 55– 66. https://doi.org/10.1007/s00334-004-0053-8 Google Scholar CrossRef Search ADS   Jarvis A., Reuter H., Nelson A., Guevara E., 2008. Hole-filled seamless srtm data v4 , International Centre for Tropical Agriculture (CIAT), http://srtm.csi.cgiar.org, 2014-04-08. Jenks G., 1967. The data model concept in statistical mapping, in International Yearbook of Cartography , Vol. 7, pp. 186– 190, ed. Philip G., Bonn. Jenks G., Caspall F., 1971. Error on choroplethic maps: definition, measurement, reduction, Ann. Assoc. Am. Geographers , 61, 217– 244. https://doi.org/10.1111/j.1467-8306.1971.tb00779.x Google Scholar CrossRef Search ADS   Jing Z., Rapp G., 2003. The coastal evolution of the ambracian embayment and its relationship to archaeological settings, in Hesperia Supplements , Vol. 32, pp. 157– 198. doi:10.2307/1354048. Google Scholar CrossRef Search ADS   Kapsimalis V., Pavlakis P., Poulos S., Alexandri S., Tziavos C., Sioulas A., Filippas D., Lykousis V., 2005. Internal structure and evolution of the Late Quaternary sequence in a shallow embayment: the Amvrakikos Gulf, NW Greece, Mar. Geol. , 222–223, 399– 418. https://doi.org/10.1016/j.margeo.2005.06.008 Google Scholar CrossRef Search ADS   Karageorgis A., Krasakopoulou E., Pappas G., Papageorgiou A., Taxiarchi M., 2006. Geochemical characteristics of surface sediments from the lagoons of Amvrakikos Gulf, Ionian Sea, in Proceedings of the 8th Panhellenic Symposium on Oceanography and Fisheries, 2006, Thessaloniki , pp. 793– 799. Kokkalas S., Xypolias P., Koukouvelas I., Doutsos T., 2006. Postcollisional contractional and extensional deformation in the Aegean region, in Postcollisional tectonics and magmatism in the Mediterranean region and Asia , Vol. 409, pp. 97– 123. doi:10.1130/0-8137-2409-0.97. Google Scholar CrossRef Search ADS   Lesser G., Roelvink J., van Kester J., Stelling G., 2004. Development and validation of a three-dimensional morphological model, Coastal Eng. , 51, 883– 915. https://doi.org/10.1016/j.coastaleng.2004.07.014 Google Scholar CrossRef Search ADS   Li L., Qiu Q., Huang Z., 2012a. Numerical modeling of the morphological change in Lhok Nga, west Banda Aceh, during the 2004 Indian Ocean tsunami: understanding tsunami deposits using a forward modeling method, Nat. Hazards , 64, 1549– 1574. https://doi.org/10.1007/s11069-012-0325-z Google Scholar CrossRef Search ADS   Li L., Huang Z., Qiu Q., Natawidjaja D., Sieh K., 2012b. Tsunami-induced coastal change: scenario studies for Painan, West Sumatra, Indonesia, Earth Planets Space , 64, 799– 816. https://doi.org/10.5047/eps.2011.08.002 Google Scholar CrossRef Search ADS   Li L., Huang Z., Qiu Q., 2014. Numerical simulation of erosion and deposition at the Thailand Khao Lak coast during the 2004 Indian Ocean tsunami, Nat. Hazards , 74, 2251– 2277. https://doi.org/10.1007/s11069-014-1301-6 Google Scholar CrossRef Search ADS   Liang D., Gotoh H., Khayyer A., Chen J., 2013. Boussinesq modelling of solitary wave and N-wave runup on coast, Appl. Ocean Res. , 42, 144– 154. https://doi.org/10.1016/j.apor.2013.05.008 Google Scholar CrossRef Search ADS   Madsen P., Fuhrman D., Schäffer H., 2008. On the solitary wave paradigm for tsunamis, J. geophys. Res. , 113, 1– 22. https://doi.org/10.1029/2008JC004932 Google Scholar CrossRef Search ADS   Martin M., Weiß R., Bourgeois J., Pinegina T., Houston H., Titov V., 2008. Combining constraints from tsunami modeling and sedimentology to untangle the 1969 Ozernoi and 1971 Kamchatskii tsunamis, Geophys. Res. Lett. , 35, 1– 6. May S., Vött A., Brückner H., Brockmüller S., 2007. Evidence of tsunamigenic impact on Actio headland near Preveza, NW Greece, in Coastline Reports , Vol. 9, pp. 115– 125, eds Gönnert G., Pflüger B., Bremer J.-A., EUCC-Deutschland. May S., Vött A., Brückner H., Grapmayer R., Handl M., Wennrich V., 2012a. The Lefkada barrier and beachrock system (NW Greece)—Controls on coastal evolution and the significance of extreme wave events, Geomorphology , 139–140, 330– 347. https://doi.org/10.1016/j.geomorph.2011.10.038 Google Scholar CrossRef Search ADS   May S., Vött A., Brückner H., Smedile A., 2012b. The Gyra washover fan in the Lefkada Lagoon, NW Greece—possible evidence of the 365 AD Crete earthquake and tsunami, Earth Planets Space , 64, 859– 874. https://doi.org/10.5047/eps.2012.03.007 Google Scholar CrossRef Search ADS   McGranahan G., Balk D., Anderson B., 2007. The rising tide: assessing the risks of climate change and human settlements in low elevation coastal zones, Environ. Urbanization , 19, 17– 37. https://doi.org/10.1177/0956247807076960 Google Scholar CrossRef Search ADS   McMaster R., 1997. In memoriam: George F. Jenks (1916–1996), Cartogr. Geogr. Inf. Syst. , 24, 56– 59. Morton R., Gelfenbaum G., Jaffe B., 2007. Physical criteria for distinguishing sandy tsunami and storm deposits using modern examples, Sedimentary Geol. , 200, 184– 207. https://doi.org/10.1016/j.sedgeo.2007.01.003 Google Scholar CrossRef Search ADS   Namegaya Y., Satake K., 2014. Reexamination of the A.D. 869 Jogan earthquake size from tsunami deposit distribution, simulated flow depth, and velocity, Geophys. Res. Lett. , 41, 2297– 2303. https://doi.org/10.1002/2013GL058678 Google Scholar CrossRef Search ADS   Ontowirjo B., Paris R., Mano A., 2010. Sediment transport modelling during 2004 tsunami runup at Lhok Nga (Sumatra, Indonesia), Geophys. Res. Abstr. , 12. Ontowirjo B., Paris R., Mano A., 2013. Modeling of coastal erosion and sediment deposition during the 2004 Indian Ocean tsunami in Lhok Nga, Sumatra, Indonesia, Nat. Hazards , 65, 1967– 1979. https://doi.org/10.1007/s11069-012-0455-3 Google Scholar CrossRef Search ADS   Papadopoulos G., Plessa A., 2000. Magnitude-distance relations for earthquake-induced landslides in Greece, Eng. Geol. , 58, 377– 386. https://doi.org/10.1016/S0013-7952(00)00043-0 Google Scholar CrossRef Search ADS   Papazachos B., 1996. Large seismic faults in the Hellenic arc, Ann. Geophys. , 39, 891– 903. Pareschi M., Boschi E., Favalli M., 2006a. Lost tsunami, Geophys. Res. Lett. , 33, 1– 6. Pareschi M., Boschi E., Mazzarini F., Favalli M., 2006b. Large submarine landslides offshore Mt. Etna, Geophys. Res. Lett. , 33, 1– 5. Pareschi M., Favalli M., Boschi E., 2006c. Impact of the Minoan tsunami of Santorini: simulated scenarios in the eastern Mediterranean, Geophys. Res. Lett. , 33, 1– 6. Partheniades E., 1965. Erosion and deposition of cohesive soils, J. Hydraul. Div. , 91, 105– 139. Períañez R., Abril J., 2014a. Modelling tsunamis in the eastern Mediterranean Sea. Application to the Minoan Santorini tsunami sequence as a potential scenario for the biblical Exodus, J. Mar. Syst. , 139, 91– 102. https://doi.org/10.1016/j.jmarsys.2014.05.016 Google Scholar CrossRef Search ADS   Períañez R., Abril J., 2014b. A numerical modeling study on oceanographic conditions in the former Gulf of Tartessos (SW Iberia): tides and tsunami propagation, J. Mar. Syst. , 139, 68– 78. https://doi.org/10.1016/j.jmarsys.2014.05.020 Google Scholar CrossRef Search ADS   Polonia A., Torelli L., Mussoni P., Gasperini L., Artoni A., Klaeschen D., 2011. The Calabrian Arc subduction complex in the Ionian Sea: regional architecture, active deformation, and seismic hazard, Tectonics , 30, 1– 28. https://doi.org/10.1029/2010TC002821 Google Scholar CrossRef Search ADS   Poulos S., Collins M., Lykousis V., 1995. Late Quaternary evolution of Amvrakikos Gulf, western Greece, Geo-Mar. Lett. , 15, 9– 16. https://doi.org/10.1007/BF01204492 Google Scholar CrossRef Search ADS   Poulos S., Lykousis V., Collins M., Rohling E., Pattiaratchi C., 1999. Sedimentation processes in a tectonically active environment: the Kerkyra-Kefalonia submarine valley system (NE Ionian Sea), Mar. Geol. , 160, 25– 44. https://doi.org/10.1016/S0025-3227(99)00016-X Google Scholar CrossRef Search ADS   Poulos S., Kapsimalis V., Tziavos C., Pavlakis P., Leivaditis G., Collins M., 2005. Sea-level stands and Holocene geomorphological evolution of the northern deltaic margin of Amvrakikos Gulf (western Greece), Z. Geomorphol., Supplementary Issues , 137, 125– 145. Poulos S.et al.  , 2015. Hydrodynamic, neotectonic and climatic control of the evolution of a barrier beach in the microtidal environment of the NE Ionian Sea (eastern Mediterranean), Geo-Mar. Lett. , 35, 37– 52. https://doi.org/10.1007/s00367-014-0390-2 Google Scholar CrossRef Search ADS   Pritchard D., Dickinson L., 2008. Modelling the sedimentary signature of long waves on coasts: implications for tsunami reconstruction, Sedimentary Geol. , 206, 42– 57. https://doi.org/10.1016/j.sedgeo.2008.03.004 Google Scholar CrossRef Search ADS   Richardson J., Zaki W., 1997. Sedimentation and fluidisation: part I, Chem. Eng. Res. Des. , 75, S82– S100. https://doi.org/10.1016/S0263-8762(97)80006-8 Google Scholar CrossRef Search ADS   Röbke B., Vött A., 2017. The tsunami phenomenon, Prog. Oceanogr. , 159, 296– 322. https://doi.org/10.1016/j.pocean.2017.09.003 Google Scholar CrossRef Search ADS   Röbke B., Schüttrumpf H., Wöffler T., Hadler H., Willershäuser T., Vött A., 2012. Tsunamis in the Gulf of Kyparissia (western Peloponnese, Greece) - risk assessment based on numeric simulation and field evidence, in Proceedings of the 5th International Short Conference on Applied Coastal Research, 2011, Aachen , pp. 560– 569. Röbke B., Schüttrumpf H., Wöffler T., Fischer P., Hadler H., Ntageretzis K., Willershäuser T., Vött A., 2013. Tsunami inundation scenarios for the Gulf of Kyparissia (western Peloponnese, Greece) derived from numerical simulations and geo-scientific field evidence, Z. Geomorphol. N.F., Supplementary Issues , 57, 69– 104. Google Scholar CrossRef Search ADS   Röbke B., Vött A., Willershäuser T., Fischer P., Hadler H., 2015. Considering coastal palaeogeographical changes in a numerical tsunami model—a progressive base to compare simulation results with field traces from three coastal settings in western Greece, Z. Geomorphol. N.F., Supplementary Issues , 59, 157– 188. https://doi.org/10.1127/zfg_suppl/2014/S-00210 Google Scholar CrossRef Search ADS   Röbke B., Schüttrumpf H., Vött A., 2016. Effects of different boundary conditions and palaeotopographies on the onshore response of tsunamis in a numerical model—a case study from western Greece, Cont. Shelf Res. , 124, 182– 199. https://doi.org/10.1016/j.csr.2016.04.010 Google Scholar CrossRef Search ADS   Samaras A., Karambas T., Archetti R., 2015. Simulation of tsunami generation, propagation and coastal inundation in the Eastern Mediterranean, Ocean Sci. , 11, 643– 655. https://doi.org/10.5194/os-11-643-2015 Google Scholar CrossRef Search ADS   Satake K.et al.  , 2013. Tsunami source of the 2010 Mentawai, Indonesia earthquake inferred from tsunami field survey and waveform modeling, Pure appl. Geophys. , 170, 1567– 1582. https://doi.org/10.1007/s00024-012-0536-y Google Scholar CrossRef Search ADS   Schielein P., Zschau J., Woith H., Schellmann G., 2007. Tsunamigefährdung im mittelmeer—eine analyse geomorphologischer und historischer zeugnisse, in Bamberger Geographische Schriften , Vol. 22, pp. 153– 199, eds Dix A., Göler D., Redepenning M., Schellmann G., University of Bamberg Press. Shanmugam G., 2012. Process-sedimentological challenges in distinguishing paleo-tsunami deposits, Nat. Hazards , 63, 5– 30. https://doi.org/10.1007/s11069-011-9766-z Google Scholar CrossRef Search ADS   Shiki T., Tachibana T., Fujiwara O., Goto K., Nanayama F., Yamazaki T., 2008. Characteristic features of tsunamiites, in Tsunamiites—Features and Implications , pp. 319– 340, eds Shiki T., Tsuji Y., Yamazaki T., Minoura K. Elsevier. Google Scholar CrossRef Search ADS   Soloviev S., Solovieva O., Go C., Kim K., Shchetnikov N., 2000. Tsunamis in the Mediterranean Sea 2000 B.C.–2000 A.D. , Springer. Google Scholar CrossRef Search ADS   Spiske M., Piepenbreier J., Benavente C., Bahlburg H., 2013. Preservation potential of tsunami deposits on arid siliciclastic coasts, Earth-Sci. Rev. , 126, 58– 73. https://doi.org/10.1016/j.earscirev.2013.07.009 Google Scholar CrossRef Search ADS   Stiros S., 2010. The 8.5+ magnitude, AD365 earthquake in Crete: coastal uplift, topography changes, archaeological and historical signature, Quat. Int. , 216, 54– 63. https://doi.org/10.1016/j.quaint.2009.05.005 Google Scholar CrossRef Search ADS   Sugawara D., Imamura F., Matsumoto H., Goto K., Minoura K., 2011. Reconstruction of the AD 869 Jogan earthquake induced tsunami by using the geological data, J. Japan. Soc. Nat. Disaster Sci. , 29, 501– 516. Sugawara D., Goto K., Imamura F., Matsumoto H., Minoura K., 2012. Assessing the magnitude of the 869 Jogan tsunami using sedimentary deposits: prediction and consequence of the 2011 Tohoku-oki tsunami, Sedimentary Geol. , 282, 14– 26. https://doi.org/10.1016/j.sedgeo.2012.08.001 Google Scholar CrossRef Search ADS   Sugawara D., Imamura F., Goto K., Matsumoto H., Minoura K., 2013. The 2011 Tohoku-oki earthquake tsunami: similarities and differences to the 869 Jogan tsunami on the Sendai Plain, Pure appl. Geophys. , 170, 831– 843. https://doi.org/10.1007/s00024-012-0460-1 Google Scholar CrossRef Search ADS   Sugawara D., Goto K., Jaffe B., 2014a. Numerical models of tsunami sediment transport—current understanding and future directions, Mar. Geol. , 352, 295– 320. https://doi.org/10.1016/j.margeo.2014.02.007 Google Scholar CrossRef Search ADS   Sugawara D., Takahashi T., Imamura F., 2014b. Sediment transport due to the 2011 Tohoku-oki tsunami at Sendai: results from numerical modeling, Mar. Geol. , 358, 18– 37. https://doi.org/10.1016/j.margeo.2014.05.005 Google Scholar CrossRef Search ADS   Synolakis C., Bernard E., Titov V., Kânoğlu U., González F., 2008. Validation and verification of tsunami numerical models, Pure appl. Geophys. , 165, 2197– 2228. https://doi.org/10.1007/s00024-004-0427-y Google Scholar CrossRef Search ADS   Szczuciński W., 2012. The post-depositional changes of the onshore 2004 tsunami deposits on the Andaman Sea coast of Thailand, Nat. Hazards , 60, pp. 115– 133. https://doi.org/10.1007/s11069-011-9956-8 Google Scholar CrossRef Search ADS   Tadepalli S., Synolakis C., 1994. The run-up of N-waves on sloping beaches, Proc. R. Soc. A , 445, 99– 112. https://doi.org/10.1098/rspa.1994.0050 Google Scholar CrossRef Search ADS   Tadepalli S., Synolakis C., 1996. Model for the leading waves of tsunamis, Phys. Rev. Lett. , 77, 2141– 2144. https://doi.org/10.1103/PhysRevLett.77.2141 Google Scholar CrossRef Search ADS   Tziavos C., 1997. Paleogeographic evolution of the amvrakikos gulf, western greece, in Proceedings of the 22nd International Symposium on Engineering Geology and the Environment, 1997, Athen , pp. 425– 430. UKHO—United Kingdom Hydrographic Office, 2010. Admiralty chart; map sheet: Ba189, greece-west coast, nisída sapiéntza to nísos paxoí, 1:300 000 , Taunton. UNESCO, 1981. Background papers and supporting data on the international equation of state of seawater, 1980, in UNESCO technical papers in marine science , Vol. 38, pp. 1– 192, UNESCO. USGS—United States Geological Survey’s Earthquake Hazards Program, 2015. Signifcant earthquake archive , http://earthquake.usgs.gov/earthquakes/eqinthenews/, 2015-11-10. van Rijn L., Roelvink J., ter Horst W., 2001. Approximation formulae for sand transport by currents and waves and implementation in DELFT-MOR, in Repository Hydraulic Engineering Reports , Vol. Z3054.20/40, pp. 1– 44, Delft Hydraulics (WL). Verboom G., Slob A., 1984. Weakly-reflective boundary conditions for two-dimensional shallow water flow problems, Adv. Water Resour. , 7, 192– 197. https://doi.org/10.1016/0309-1708(84)90018-6 Google Scholar CrossRef Search ADS   Vött A., May S., 2009. Auf den Spuren von Tsunamis im östlichen Mittelmeer, Geographische Rundschau , 61, 42– 48. Vött A., May S., Brückner H., Brockmüller S., 2006. Sedimentary evidence of late Holocene tsunami events near Lefkada Island (NW Greece), Z. Geomorphol., Supplementary Issues , 146, 139– 172. Vött A., Brückner H., May S., Lang F., Brockmüller S., 2007. Late Holocene tsunami imprint at the entrance of the Ambrakian Gulf (NW Greece), Méditerranée , 108, 43– 57. Google Scholar CrossRef Search ADS   Vött A., Brückner H., May S., Lang F., Herd R., Brockmüller S., 2008. Strong tsunami impact on the Bay of Aghios Nikolaos and its environs (NW Greece) during Classical-Hellenistic times, Quat. Int. , 181, 105– 122. https://doi.org/10.1016/j.quaint.2007.02.017 Google Scholar CrossRef Search ADS   Vött A.et al.  , 2009a. The Lake Voulkaria (Akarnania, NW Greece) palaeoenvironmental archive—a sediment trap for multiple tsunami impact since the mid-Holocene, Z. Geomorphol. N.F., Supplementary Issues , 53, 1– 37. Google Scholar CrossRef Search ADS   Vött A.et al.  , 2009b. Traces of Holocene tsunamis across the Sound of Lefkada, NW Greece, Glob. Planet. Change , 66, 112– 128. https://doi.org/10.1016/j.gloplacha.2008.03.015 Google Scholar CrossRef Search ADS   Vött A.et al.  , 2010. Beachrock-type calcarenitic tsunamites along the shores of the eastern Ionian Sea (western Greece)—case studies from Akarnania, the Ionian Islands and the western Peloponnese, Z. Geomorphol. N.F., Supplementary Issues , 54, 1– 50. Google Scholar CrossRef Search ADS   Vött A.et al.  , 2011. Sedimentological and geoarchaeological evidence of multiple tsunamigenic imprint on the Bay of Palairos-Pogonia (Akarnania, NW Greece), Quat. Int. , 242, 213– 239. Wheatcroft R., 1990. Preservation potential of sedimentary event layers, Geology , 18, 843– 845. https://doi.org/10.1130/0091-7613(1990)018<0843:PPOSEL>2.3.CO;2 Google Scholar CrossRef Search ADS   Wheatcroft R., Drake D., 2003. Post-depositional alteration and preservation of sedimentary event layers on continental margins, I. The role of episodic sedimentation, Mar. Geol. , 199, 123– 137. https://doi.org/10.1016/S0025-3227(03)00146-4 Google Scholar CrossRef Search ADS   Yolsal-Çevikbilen S., Taymaz T., 2012. Earthquake source parameters along the Hellenic subduction zone and numerical simulations of historical tsunamis in the eastern Mediterranean, Tectonophysics , 536–537, 61– 100. https://doi.org/10.1016/j.tecto.2012.02.019 Google Scholar CrossRef Search ADS   © The Author(s) 2017. Published by Oxford University Press on behalf of The Royal Astronomical Society.

### Journal

Geophysical Journal InternationalOxford University Press

Published: Apr 1, 2018

## You’re reading a free preview. Subscribe to read the entire article.

### DeepDyve is your personal research library

It’s your single place to instantly
that matters to you.

over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month ### Explore the DeepDyve Library ### Search Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly ### Organize Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place. ### Access Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals. ### Your journals are on DeepDyve Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more. All the latest content is available, no embargo periods. DeepDyve ### Freelancer DeepDyve ### Pro Price FREE$49/month
\$360/year

Save searches from
PubMed

Create lists to

Export lists, citations