Background

Historical and paleo-tsunami research and its application to geophysical study

The 2011 off the Pacific coast of Tohoku earthquake (2011 Tohoku-oki earthquake) (M w 9.0) that occurred on March 11, 2011, triggered a large tsunami (2011 Tohoku-oki tsunami) along the east coast of Japan, causing severe damage and loss of life. The Pacific coastal region, in particular, the Sanriku Coast (Fig. 1), ranks among the highest tsunami risk areas in Japan, in terms of both severity and frequency (Watanabe 1998). However, we could not assess the risk of large tsunamis like the 2011 Tohoku-oki tsunami because we lack sufficient knowledge about large earthquakes and the tsunami history in the Japan Trench. Even after the 2011 Tohoku-oki tsunami, the long-term tsunami history of the Sanriku Coast remains obscured by inadequate field data (Sugawara et al. 2012). Moreover, the geological and geophysical problems revealed by the 2011 Tohoku-oki earthquake, including mechanisms of tsunami generation (Kawamura et al. 2012; Tappin et al. 2014) and large earthquakes (Ikeda et al. 2012; Goldfinger et al. 2013; Rajendran 2013), remain unsolved. Recent studies have suggested that submarine mass failure (e.g., submarine landslide) may have contributed to large tsunamis along the Japan Trench (Kawamura et al. 2012; 2014; Tappin et al. 2014). To resolve these problems, we require long-term solid evidence (e.g., historical and paleo-tsunami deposits).

Fig. 1
figure 1

Study site location. a Plate tectonic map in and around northeastern Japan. The star indicates the epicenter location of the 2011 Tohoku-oki earthquake determined by the Japan Meteorological Agency. SP Sendai Plain, IP Ishinomaki Plain. b Study area and geomorphology along the Sanriku Coast

Full size image

The best sources of precise long-term tsunami data are coastal lowlands, in particular, marshes (Minoura and Nakaya 1991; Witter et al. 2003; Sawai et al. 2009; Shennan et al. 2014). In some regions (Hokkaido of Japan, Alaska of USA, and the North Island of New Zealand), coseismic and post-seismic crustal movements are recorded in sediments as lithological and biological changes (Witter et al. 2003; Sawai and Nasu 2005; Hamilton and Shennan 2005; Hayward et al. 2005). Thus, in this study, we excavated trenches at a coastal marsh in Koyadori, in the middle part of the Sanriku Coast (Fig. 1). The aim was to provide new geological evidence of historical and paleo-tsunami deposits.

Study site

Koyadori is located in the central part of the Sanriku Coast, the easternmost part of Honshu Island (Figs. 1 and 2). Approximately 200 km east of Koyadori, the Pacific plate subducts underneath the Eurasian Plate, where rupture areas of historical and observed earthquakes have been identified on the plate boundary (Earthquake Research Committee Headquarters for Earthquake Research Promotion Prime Minister’s Office, Government of Japan 1999). In Koyadori, the mouth of the valley is closed by beach ridges. Prior to the 2011 Tohoku-oki tsunami, which hit the valley (Fig. 3), the study site was used as paddy fields.

Fig. 2
figure 2

Topography and geology around the Funakoshi Peninsula. Base map is based on the 5 m mesh DEM supplied by the Geospatial Information Authority of Japan. Geology is modified after Yoshida et al. (1984)

Full size image
Fig. 3
figure 3

Aerial photographs around Koyadori before and after the 2011 Tohoku-oki earthquake. Aerial photos (a)–(d) were taken on September 1977, March 2011, June 2011, and October 2012, respectively, by the Geospatial Information Authority of Japan

Full size image

The geology differs on both sides of the valley (Fig. 2; Yoshida et al. 1984). The east side comprises Early Cretaceous hornblende-biotite granodiorite, granite, granite porphyry, and tonalite. The west side comprises dacite to rhyolite lava and pyroclastic rock deposited during the Early Cretaceous. Ishimura et al. (2014) drilled cores at the study site (Fig. 4), revealing visible tephra layers such as the Towada–Chuseri tephra (To–Cu: 6 ka; Machida and Arai 2003) and Oguni Pumice (7.3–7.4 ka; Ishimura et al. 2014) erupted from the Towada Volcano (Fig. 1).

Fig. 4
figure 4

Geomorphological classification around Koyadori. Contour maps are based on the 1 m mesh DEM supplied by the Iwate Prefecture. The contour interval in (a) and (b) is 5 m and 1 m, respectively. c Topographic profile along A–A’ line based on the 1 m mesh DEM

Full size image

Tsunami history and previous study of tsunami deposits

Large historical earthquakes and tsunamis have been instrumentally recorded during the last several decades and have been recorded in historical documents and legends during the last 1300 years (Utsu 2004). Many earthquake-generated tsunamis have hit the areas around Koyadori (Table 1), nine of which were large events with runup heights exceeding a few meters (Table 2). The 2011 Tohoku-oki tsunami exhibits the largest runup height measured between AD 2011 and 1896. In contrast, based on the estimated runup heights, the largest tsunami occurring between AD 1611 and 1856 was probably the 1611 Keicho Sanriku tsunami (Table 2). Candidates of historical tsunamis older than AD 1611 are the 869 Jogan tsunami and the 1454 Kyotoku tsunami. The 869 Jogan tsunami, identified in the Sendai and Ishinomaki Plains (Fig. 1) (Minoura and Nakaya 1991; Minoura et al. 2001; Sawai et al. 2007; Shishikura et al. 2007), has been noted as the penultimate large tsunami event around these plains (after the 2011 Tohoku-oki earthquake and tsunami), because its inundation in the Sendai Plain was similar to that of the 2011 Tohoku-oki tsunami (Sugawara et al. 2012; Namegaya and Satake 2014). However, the northern and southern distribution limits of the 869 Jogan tsunami deposits have not yet been determined (Sugawara et al. 2012). Thus, elucidating whether the 869 Jogan tsunami reached the Sanriku Coast is essential for tsunami risk assessment and seismological study. Consequently, this information is urgently required (Sugawara et al. 2012). If the 869 Jogan tsunami and 2011 Tohoku-oki tsunami were of similar severity, a large tsunami is also likely to have struck the areas along the Sanriku Coast in AD 869. In recent times, the 1454 Kyotoku tsunami has been identified from tsunami deposits in the Sendai and Ishinomaki Plains (Sawai et al. 2012) and historical documents (Namegaya and Yata 2014). However, at our study site, information on the Kyotoku tsunami deposits and the size of the tsunami has not yet been obtained.

Table 1 Historical tsunamis along the Sanriku Coast during AD 1611–2011
Full size table
Table 2 Historical tsunamis’ runup height around Koyadori
Full size table

Historical and paleo-tsunami deposits along the Sanriku Coast have been studied at several sites by The Headquarters for Earthquake Research Promotion (2006; 2007; 2008; 2009; and 2010), Haraguchi et al. (2006a, b; 2007), Haraguchi and Goya (2007), Imaizumi et al. (2007), Torii et al. (2007), and Haraguchi and Ishibe (2009) before the 2011 event. However, historical tsunami deposits were identified at only one of the onshore sites at Rikuzentakata (Fig. 1). This site has revealed four historical tsunami deposits during the past 700 years, the latest of which was correlated with the 1960 Chile tsunami (The Headquarters for Earthquake Research Promotion 2007). The other events are not well correlated with historical tsunamis. Recent sediments, in particular, those deposited during the past 1000 to 2000 years, are badly preserved at other onshore sites (Torii et al. 2007; The Headquarters for Earthquake Research Promotion 2010). Therefore, historical tsunami deposits are poorly understood along the Sanriku Coast.

Methods

Geomorphological classification

Initial mapping of geomorphic surfaces around Koyadori was based on interpretation of 1:8000- and 1:10,000-scale aerial photographs taken by the Geospatial Information Authority of Japan before and after the 2011 Tohoku-oki tsunami, and anaglyph images prepared from 1 m and 5 m mesh DEM (Digital Elevation Model) provided by the Geospatial Information Authority of Japan and the Iwate Prefecture.

Trench survey

In December 2012, we excavated a 12 m long, 3 m wide, and 2 m deep trench (KYD-trench) approximately 3 m above sea level (a.s.l.) and 300 m distant from the shoreline (Figs. 4 and 5a, b). We logged and took photographs of each trench wall. Block samples (50 cm long, 10 cm wide, and 5 cm deep) were taken from the west, east, and south walls. The block samples of the west wall overlapped, while those of the other walls were separate (Fig. 6). We also sampled deposits of each event from the east wall for particle roundness analysis (Fig. 6). In 2013, a construction company excavated an additional canal (the canal-trench; approximate length, width, and depth 300, 1, and 0.5 m, respectively) as a tentative drainage for field restoration (Fig. 4). To confirm the continuity of each event, we logged a 150 m-long section of the canal-trench wall and sampled sediments for particle roundness analysis.

Fig. 5
figure 5

Photographs around the trench sites. a The KYD- and canal-trench sites, b the KYD-trench, c the canal-trench, d erosion of beach ridges, e tsunami boulder, and f 2011 tsunami deposits

Full size image
Fig. 6
figure 6

Picture and sketch of the KYD-trench walls. a Photo mosaics of the KYD-trench wall. b Sketch of the KYD-trench wall

Full size image

Dry and wet density, color, and loss on ignition measurement

We sampled the entire 7 cm3 cube (each side = 2.2 cm) from the block samples and measured their wet and dry bulk densities. The color of wet sediments in cubic samples is quantified by the L*, a*, and b* parameters measured by the Soil Color Reader SPAD-503 instrument (Konica Minolta Sensing, Inc.). The a* and b* parameters specify the red (+) to green (−) and yellow (+) to blue (−) content, respectively, while L* represents lightness (0 = black, 100 = white). The loss on ignition (LOI) was conducted in each block sample following Bos et al. (2012) at 3–6 cm intervals, although this sampling was restricted to peat and peaty silt.

Particle roundness

Furthermore, to reveal the origins of the event deposits and to confirm tsunami deposits, we sampled fluvial and beach sediments (Locations 1–8) (Fig. 4) for particle roundness analysis in 2012 and 2013. Samples were washed and dry-sieved through 2 mm mesh and the gravels were divided into six roundness categories (very angular, angular, sub-angular, sub-rounded, rounded, and well-rounded).

Radiocarbon dating

Radiocarbon dating (30 samples) was conducted by accelerator mass spectroscopy (AMS) at the Institute of Accelerator Analysis Ltd. and Geo Science Laboratory. The obtained age data were calibrated using the OxCal 4.2 program (Ramsey 2009) and the calibration curve IntCal13 (Reimer et al. 2013).

Tephra analysis

The Towada-a tephra (To-a) (AD915; Machida and Arai 2003) is a useful indicator of the 869 Jogan tsunami deposits in the Sendai and Ishinomaki Plains (Minoura and Nakaya 1991; Sawai et al. 2007; Shishikura et al. 2007). From the presence and distribution of To-a along the southern Sanriku Coast, Ishimura et al. (2014) suggested that To-a had also been deposited at the central Sanriku Coast. Therefore, we conducted a cryptotephra analysis to identify the invisible To-a horizon.

Each block sample was sampled at 3–6 cm intervals. These samples were washed using 60 μm nylon mesh and dry-sieved using 124 μm nylon mesh. Thin sections made with the 60–124 μm fractions revealed volcanic glass contents. The refractive index of volcanic glass shards, which is useful for identifying widespread tephras in Japan (Machida and Arai 2003), was measured with a refractive index measuring system (RIMS 2000: Kyoto Fission Track Co., Ltd.). The RIMS system measures volcanic glass shards to an accuracy of ±0.0002 (Danhara et al. 1992). The major element compositions were analyzed by energy-dispersive spectrometry using an electron probe microanalysis (EPMA) system (Horiba Emax Energy EX-250) at the FURUSAWA Geological Survey. The major elements were measured by scanning a 4 μm grid of the targeted grain under a counting time of 150 s and accelerating voltage of 15 kV. The beam current and diameter were 0.3 nA and 150 nm, respectively. The atomic number effect was corrected by the ZAF procedure.

Results

2011 Tohoku-oki tsunami and its deposits

The inundation and runup heights of the 2011 Tohoku-oki tsunami at Koyadori ranged from 13 to 18 m a.s.l. and from 26 to 29 m a.s.l., respectively (Table 2; Haraguchi and Iwamatsu 2011). Figure 3 shows the landform changes before and after the 2011 event. Immediately following the 2011 Tohoku-oki tsunami (April 2011), the beach was not yet re-established and the beach ridges may have been shortened by the tsunami backwash (Fig. 3a, b). After the beach was restored in June 2011, the shortcut channel was filled with beach deposits (Fig. 3c). The poor drainage area remained until October 2012 (Fig. 3d). The Tohoku-oki tsunami hit the coastal levee originating from beach ridges, destroying it and the trees on it, and eroding it to a depth of 1–1.5 m (Fig. 5d). The eroded materials were transported landward and deposited as tsunami deposits. Approximately 9 m a.s.l. and 600 m landward, a boulder was recognizable as a tsunami deposit because of the attached oyster shells (Fig. 5e). Tsunami deposits composed of sand and gravel sourced from the beach and beach ridges were found up to 600 m landward in December 2012 (Fig. 5f).

Description of the KYD-trench

Deposits in the trench wall were divided into five facies (event deposits, marsh deposits, channel fill deposits, artificial fill deposits, and cultivation soil), based on their sediment structure, continuity, and composition (Fig. 6). All the walls contained marsh deposits and interbedded event deposits.

The event deposits are composed of coarse sand and granule, and are traceable in the trench (Figs. 6 and 7). Deposits showing good continuity, horizontal sedimentation, erosional features, and loading structure at the base were considered as potential candidates for tsunami deposits, and were labeled E1 (youngest) to E11 (oldest). The characteristics of each event deposit are presented in Table 3. The E1 deposits (the 2011 Tohoku-oki tsunami deposits) are divisible into two units. The lower unit is composed of coarse sediments (pebble to coarse sand) with normal grading. The upper unit comprises finer sediments (granule to medium sand) and is partially laminated. The E2 deposits are thin and composed of granule to coarse sand. The E2 layer is interbedded with cultivation soil, indicating partial disturbance by cultivation. The E3 deposits are well traceable and characterized by a bluish color (Fig. 6). As these deposits thicken from grids E–3 to S–3, they also become coarser (cobble to pebble). However, while the basal contact is very sharp, the upper contact is partially disturbed by cultivation. The E4 deposits are partially disturbed and eroded by channel deposits. In general, their compositions are fine (granule to fine sand), although some parts contain pebble to cobble gravels. The E5 deposits are interbedded with low-LOI organic sediments (Fig. 7). Their basal contact is sharp, but their upper contact is disturbed. They are intermittently distributed because of plant bioturbation. The E6 deposits are well traceable in the KYD-trench (Fig. 6), with very sharp basal contacts and a loading structure at the bottom of the layer. The E7 deposits are also well traceable and some of them have eroded the E8 deposits (Fig. 6). The E7 layer also shows a loading structure at the bottom. The E8 deposits are intermittently distributed because of erosion by the E7 deposits. Below the E8 deposits, the upper and basal contacts of the event deposits have been disturbed by plant bioturbation. The E9 and E10 deposits are very thin and similarly disturbed by plant bioturbation, but are nonetheless traceable in the KYD-trench (Fig. 6). The E11 deposits are traceable in the southern half of the KYD-trench and distributed under the trench bottom in the northern half (Fig. 6). The E11 deposits are of medium thickness and contain fine grains (coarse to medium sand).

Fig. 7
figure 7

Tephra analysis, color, dry/wet bulk density, and loss on ignition of the KYD-trench wall samples

Full size image
Table 3 Characteristics of event deposits in the KYD-trench
Full size table

The marsh deposits are composed of plant remains and organic sediments. Their densities are inversely correlated with their LOIs and indirectly indicate their organic carbon content and degree of decomposition (Fig. 7). Color, in particular, the L* and b* parameters, is correlated with density, whereas the LOI fluctuates between event deposits. Macroscopically, the LOI decreases from the trench bottom to the E4 deposits and increases from the E4 deposits to the E3 deposits.

The channel fill deposits exhibit two cross-sectional geometries and compositions, categorized as Channel 1 and Channel 2 (Fig. 6). Channel 1 is distributed from grid N–5 to N–10 and from grid W–5.5 to W–7 (Fig. 6). From the altitude of the channel bottom in both walls, the flow direction of Channel 1 was determined as east to west. Sediments are finer in Channel 1 than in Channel 2, comprising coarse sand to fine pebbles, and interbedded with peaty silt. Channel 2 is distributed from grid N–1 to W–10 and from grid E–5.5 to E–11.5 and flows from northwest to southeast (Fig. 6). The composition is poorly sorted pebble to cobble.

The artificial fill deposits with a buried PVC pipe, distributed from grid W–7 to W–11 and from grid N–1.5 to N–2.5 (Fig. 6), were identified from interviews with landowners as underdrains constructed 40–50 years ago.

The cultivation soil is distinguished from marsh deposits by its different particle composition, color, and texture. This soil type is interbedded between the E3 and E1 deposits (Fig. 6). Event deposits, marsh deposits, and cultivation soil are also easily distinguishable by their density and color (Fig. 7). The dry bulk density of cultivation soil is intermediate between low-density marsh deposits and high-density event deposits.

According to radiocarbon dating of these marsh and channel deposits (Figs. 6, 7, and 8; Table 4), the sediments in the KYD-trench wall provide a continuous record since approximately 4000 cal. BP.

Fig. 8
figure 8

Event age diagram in the KYD-trench

Full size image
Table 4 Radiocarbon ages and calibrated ages
Full size table

Description of the canal-trench

Event deposits (the E1, E2, and E3 deposits), other event deposits, cultivation soil/peat, and debris flow deposits were identified in the canal-trench (Figs. 4 and 9). Debris flow deposits are distinguished by poorly sorted gravel beds interfingered with cultivation soil/peat in the geometric cross-section (Fig. 5c). Although the E1 deposits are traceable, the original thickness of the E1 layer has been obscured by artificial modification following the 2011 event. The E2 deposits are intermittently distributed and some of them have been modified by cultivation. The E3 deposits are traceable and partially disturbed. The E3 layer is less than 20 cm thick, decreasing in the landward direction.

Fig. 9
figure 9

Sketch of the canal-trench

Full size image

Particle roundness

Event deposits were sampled from the KYD- and canal-trenches (Figs. 6 and 9). Modern beach and river deposits, and debris flow deposits, were sampled from the canal-trench and outcrops (Figs. 4 and 9). Next, the origins of the event deposits were inferred from their roundness measures. Modern river and debris flow deposits consist almost entirely of angular/very angular and sub-rounded/sub-angular gravels, with no well-rounded/rounded gravels (Fig. 10a–c). Conversely, modern beach deposits contain well-rounded/rounded and sub-rounded/sub-angular gravels; angular/very angular gravels are absent (Fig. 10d). The E1 and E3 deposits (Fig. 10e, f) are similar to modern beach deposits, with high contents of well-rounded/rounded and sub-rounded/sub-angular gravels. The clear roundness differences between modern river and beach deposits are shown in the triangular diagram of Fig. 10g. The roundness composition of all event deposits in the KYD-trench is shown in Fig. 10h. In all samples, the proportion of well-rounded/rounded gravel contents exceeds 10 %, while the angular/very angular gravel content is below 40 %. Unlike the terrestrial deposits, all event deposits contain beach gravels.

Fig. 10
figure 10

Photographs and roundness of tsunami deposits, channel deposits, debris flow deposits, modern beach deposits, and modern river deposits. a Modern river deposits (Loc. 1). b Modern river deposits (Loc. 2). c Debris flow deposits in the canal-trench (Loc. 7). d Modern beach deposits (Loc. 8). e The E1 deposits in the south wall of the KYD-trench. f The E3 deposits in the south wall of the KYD-trench. g Comparison of the 2011 Tohoku-oki tsunami deposits, beach deposits, modern river deposits, and debris flow deposits; h tsunami deposits in the KYD-trench wall. Scale in the photographs is 1 cm. In the triangular diagrams, x-axis indicates well-rounded/rounded, y-axis indicates sub-rounded/sub-angular, and z-axis indicates very angular/angular

Full size image

Tephra analysis

From the radiocarbon dating, we determined that the invisible To-a (AD915 (1035 cal. BP)) lies between the E3 and E5 deposits. The volcanic glass contents in each trench wall sample increase after the E4 deposition (Fig. 7). In particular, in the east and west walls, the volcanic glass content suddenly increases and gradually decreases from the lower to upper parts, indicating an invisible tephra horizon. However, this trend is absent in the south wall, probably because it has been eroded by the E3 deposits.

The origins of the volcanic glass shards were determined from their refractive indices. Above the E4 deposits, the region of highest volcanic glass content, the glass refractive index ranges from 1.504 to 1.511 (mode: 1.507–1.508). Below the E4 deposits, the refractive index ranges from 1.509 to 1.514 (mode: 1.512). Furthermore, we analyzed the major element compositions of the volcanic glass shards above the E4 deposits (Table 5).

Table 5 Major element compositions of volcanic glass shards
Full size table

Discussion

Identification of tsunami deposits

The roundness similarities between the event and beach deposits (Fig. 10) indicate that event deposits were transported from beach and beach ridges to the inland trench sites. Landward transport from the sea is expected in tsunami and storm events. The general characteristics of tsunami and storm deposits have been reported by many researchers (Morton et al. 2007; Kortekaas and Dawson 2007; Switzer and Jones 2008; Goff et al. 2012; Phantuwongraj and Choowong 2012). On average, tsunami deposits are generally thinner than storm deposits (Morton et al. 2007; Phantuwongraj and Choowong 2012), and sedimentary structure is less common in tsunami deposits than in storm deposits (Morton et al. 2007; Kortekaas and Dawson 2007; Switzer and Jones 2008; Goff et al. 2012). The basal contact of both sediments is unconformable or erosional (Morton et al. 2007; Kortekaas and Dawson 2007; Switzer and Jones 2008; Goff et al. 2012; Phantuwongraj and Choowong 2012), although tsunami deposits sometimes show a loading structure (Goff et al. 2012). On transect scales (several hundred meters), the cross-shore geometries of tsunami and storm deposits are characterized by “broad thin drapes with tabular or landward thinning” and “narrow thick deposits with abrupt landward thinning,” respectively (Morton et al. 2007). These characteristics of tsunami deposits are recognized in the event deposits in the trenches. In the KYD-trench (length = 12 m), all event deposits are generally thinner than 20 cm and appear as draped or eroded paleo-surfaces. Some of them exhibit a loading structure. In the canal-trench (length = 150 m), the E1 and E3 deposits appear as draped deposits, with landward thinning in the E3 deposits. Furthermore, the KYD-trench is located 300 m inland from the beach, and landowners reported no storm deposits in the trench sites during the past 40–50 years. In contrast, the paleo-shoreline after the To–Cu deposition (about 6 ka) is estimated to be at least on the seaside of the KYD-Br1 site (Fig. 4). The elevations of primary To–Cu tephra within the KYD-Br1 to KYD-Br3 cores are −1.60, −1.82, and −3.08 m a.s.l., respectively, showing landward deepening, and sediments deposited after the To–Cu deposition are consistent with non-marine environments such as marsh (Ishimura et al. 2014). From these data, we consider that the present depositional setting (beach ridge and behind marsh) around the trench sites was already established by 6 ka. Therefore, prior to the To–Cu fall, the paleo-surface topography places the paleo-shoreline on the seaside of the KYD-Br1 core site. The features of the event deposits and the geomorphological settings from 6 ka to the present, together with the responses of interviewed landowners regarding recent events, indicate that all event deposits in the KYD-trench are sourced from tsunamis rather than storms.

Ages of tsunami deposits and correlation to historical tsunami event

Radiocarbon dating (Fig. 8, Table 4) suggests that the deposits in the KYD-trench are relatively close in age with no large age gap. Radiocarbon dating of event deposits is performed on plant fragments (such as reeds), because these constitute the youngest material in a sampled horizon. Plant materials in the trench are likely to be fragments of in situ plants, but downward invasion of roots and underground stems should not be ruled out. Thus, the ages of the plant material were assumed to represent the youngest ages of the sampled horizons. In contrast, a charcoal and a hard-shell plant seed (Juglans sp.) (Samples No. 21 and 28; Table 4) are assumed to be transported materials, whose ages mark the older age limit of the sampled horizons. Organic sediments (Samples No. 5 and 24; Table 4) are older than plant fragments, consistent with our radiocarbon dating interpretations. The true age of the sediment is expected to lie between the ages of the plant and other materials. Ishimura et al. (2014) identified To–Cu (6 ka) tephra and Oguni Pumice (7.3 to 7.4 ka) in the KYD-Br3 core drilled next to the KYD-trench at depths of 4.41–5.98 m (total thickness of primary and secondary tephra) and 8.55–8.60 m, respectively. The horizons and ages of these sediments are consistent with the radiocarbon-dated geochronology of the KYD-trench determined in this study.

We estimated the event ages based on radiocarbon dating and the above criteria (Fig. 8, Table 6). The tsunami deposit events are labeled E1–E11 in order of increasing age, considering the age constraint of the next older event. Since the ages of the E5 and E6 deposits were not determined by radiocarbon dating, we interpolated their ages using the radiocarbon ages of the E4 and E7 deposits. From these chronological estimations, we can correlate the E2, E3, and E4 deposits to historical tsunamis because these deposits are younger than AD 600.

Table 6 Estimated ages of tsunami deposits and their correlation with historical tsunami events
Full size table

The ages of the E2 deposits range from modern times to 290 cal. BP (i.e., they are younger than AD 1660). Certainly, the E2 deposits can be correlated to one event among the 1611, 1677, 1793, 1856, 1896, and 1933 events (Table 2). The runup heights (Table 2) and stratigraphic position of the E2 deposits suggest a correlation with the 1933 Showa Sanriku tsunami and the 1896 Meiji Sanriku tsunami, because the height of the beach ridge at Koyadori was approximately 5 m a.s.l. in both events. Although both tsunamis inundated up to the trench sites, only single-event deposits were identified from AD 2011 to 1660. Tsunami deposits can be absent for several reasons stated as follows: 1) disturbance and/or removal by cultivation, 2) erosion by succeeding tsunami events, 3) sediment availability, and 4) tsunami size. The first cause is easily explained. If sediments were deposited by the 1896 Meiji Sanriku and the 1933 Showa Sanriku tsunamis, the latter deposits would first be disturbed and removed by cultivation processes. In this case, we would correlate the preserved tsunami deposits to the 1896 Meiji Sanriku tsunami. Regarding the second cause, the 1896 deposits might have been eroded by the 1933 deposits. However, the E2 deposits show no clear base erosion in either the KYD- or canal-trenches, and no remnants of eroded tsunami deposits are evident between the E2 and E3 deposits. Thus, the second cause is inconsistent with the observations. Meanwhile, the third cause is inconsistent with the study site setting. If a large tsunami, with height exceeding that of the beach ridge, hits Koyadori, sediments of beach and beach ridge must be transported landward because there is much sediment in the coast and the beach was re-established a few months after the 2011 event. The forth cause, tsunami size, relates to the transportation and preservation of tsunami deposits. The inundation heights were larger in the 1896 event than in the 1933 event (Table 2). Thus, we can easily expect that the volume of the 1896 tsunami deposits exceeded that of the 1933 deposits. This also suggests that the 1896 tsunami deposits were better preserved than the 1933 tsunami deposits. From these considerations, we inferred that the E2 deposits are correlated to the 1896 Meiji Sanriku tsunami.

Considering the above correlation of the E2 deposits, the age of the E3 deposits was estimated as 54–620 cal. BP (AD 1896–1330). Thus, the E3 deposits can be correlated to one event among the 1454, 1611, 1677, 1793, and 1856 events (Table 2). Based on the tsunami runup height of these events (Table 2), the E3 deposits are most probably associated with the 1611 Keicho Sanriku tsunami. The E3 deposits are thick and composed of coarse materials (Table 3), and are traceable in the canal-trench (Fig. 9). Assuming a similar depositional setting from about 6 ka onward, we considered that the feature differences (thickness and grain size) among event deposits roughly indicate the tsunami size. The features of the E3 deposits (Table 3) suggest a large, very energetic tsunami. According to a local legend (Imamura 1934), the 1611 Keicho Sanriku tsunami inundated and surged through the Koyadori–Oura pass (Fig. 2). This indicates that the 1611 Keicho Sanriku tsunami was at least as high as the 2011 Tohoku-oki tsunami, since the latter failed to reach the geomorphic pass.

According to the radiocarbon dates of the channel deposits, the E4 deposits are aged 1000–1350 cal. BP (AD 950–600), and possibly correlate with the 869 Jogan tsunami. By targeting our tephra analysis at the To-a (AD 915) horizon, we determined an absolute timing for the E4 deposits. The increased content of volcanic glass above the E4 deposits (Fig. 7) suggests a tephra fall after the E4 sedimentation. The refractive index of volcanic glass shards above the E4 deposits ranged from 1.504 to 1.511 (mode: 1.507–1.508), which includes the To-a tephra range (Machida and Arai 2003; Ishimura et al. 2014). Similarly, the chemical compositions of volcanic glass shards were consistent with previously reported To-a compositions (Aoki and Machida 2006; Ishimura et al. 2014). From these data, we inferred that the To-a tephra fell between the E4 and E3 deposits, and we assigned the E4 deposits to the 869 Jogan tsunami. This identification based on radiocarbon dating and tephra provides significant information on the size and source of the 869 Jogan tsunami and earthquake, indicating that this tsunami reached the middle part of the Sanriku Coast and its inundation area was possibly as large as the 2011 event. Since the Jogan tsunami is not reported in historical records around Koyadori and insufficient information is available for regionally and chronologically identifying the tsunami deposits along the Sanriku Coast, this finding requires confirmation in paleographical and geological researches.

Tsunami ages and their intervals

Conclusive age estimates and correlations of historical tsunami events are summarized in Table 6. Although some ambiguity of the ages remains, we calculated the average interval of tsunami occurrence as 290–390 years. Before considering the approximate age intervals of tsunami events, we need to discuss the preservation potential of tsunami deposits at this site. Szczucinski (2012) and Spiske et al. (2013) mentioned the preservation potential of tsunami deposits in tropical and temperate climate regions, respectively, and showed that the characteristics of tsunami deposits (thickness and sedimentary structure) degrade over time. Spiske et al. (2013) emphasized the significance of the preservation potential in assessing the intervals and frequencies of tsunamis, because tsunami deposits are not necessarily preserved in whole inundated areas. They identified five determining factors of preservation potential as follows: 1) composition and genetic type of the tsunami deposits, 2) coastal topography and depositional environment, 3) co- and post-seismic uplift or subsidence, 4) climate, and 5) anthropogenic modification. In Koyadori, tsunami deposits originated from beach and beach ridge deposits and are coarser than those reported in Szczucinski (2012) and Spiske et al. (2013), indicating larger resistance to post-tsunami surface processes. As mentioned above, the sedimentary environment has remained largely unchanged since 6 ka, and the beach ridge and behind-marsh environment have maintained accommodation space for tsunami and marsh deposits. The 2011 event was followed by co- and post-seismic subsidence (Ozawa et al. 2011), enhancing the preservation environment of tsunami deposits. The Sanriku Coast has a temperate climate and experiences fewer and weaker storms and high tide events (such as typhoons) than the western part of Japan. According to interviews with landowners, no storm deposits have settled in the trench sites during the past 40–50 years. Artificial modification is limited to deposits younger than E3 at this site. Moreover, 2011 tsunami deposits were found in pits and coring surveys conducted around the KYD-trench in 2013 and 2014. These deposits were clearly identifiable, despite being partially bioturbated by grass and reed. Such vegetation covered the tsunami deposits, preventing erosion and removal by post-tsunami surface processes. Even in the event of dense bioturbation, tsunami deposits are easily identified by their grain composition, size, and roundness, which widely differ from those of background deposits (e.g., peat and debris flow deposits). Therefore, we conclude that the preservation potential of tsunami deposits is very high in Koyadori. Consequently, the calculated average interval probably truly reflects the interval and frequency of large tsunamis.

The calculated average interval (290–390 years) is shorter than that obtained for the Sendai and Ishinomaki Plains (Sawai et al. 2007; 2012; Shishikura et al. 2007), reflecting the high frequency of large tsunamis causing destructive damage along the Sanriku Coast. However, if we have correctly correlated the historical deposits to the historical tsunami events, we can state the age intervals from the E1 to E4 deposits as 115, 285, and 742 years, respectively. This variability probably indicates the diversity of the tsunami generation mechanism (e.g., large earthquake, tsunami earthquake, submarine mass failure, and tsunami of distant origin) and/or the combination of several types of large earthquakes from different sources around the Japan Trench.

On the other hand, the size of historical and paleo-tsunamis can be estimated from our results because the 1896 Meiji Sanriku tsunami inundated the KYD-trench site and transported tsunami deposits there. In contrast, neither the 1968 Tokachi-oki tsunami (nearby source, runup height approximately 3 m around Koyadori; Table 2) nor the 1960 Chile tsunami (distant source, runup height approximately 4 m around Koyadori; Table 2) inundated, perhaps because they were blocked by beach ridges (height approximately 5 m a.s.l.). Furthermore, the environmental setting at the study site has been established since approximately 6 ka. These observations preliminarily suggest that tsunamis larger than the 1896 Meiji Sanriku tsunami occur at the calculated average interval, providing a first step for assessing the risk and size of tsunamis along the Sanriku Coast. To understand the tsunami generation mechanism and earthquakes along the Japan Trench, we require detailed information of ages, intervals, and sizes of historical and paleo-tsunamis at multiple sites.

Conclusions

We identified eleven tsunami deposits, including the 2011 tsunami deposits, based on sedimentary structure and continuity in two trenches and comparisons of the roundness of the gravel composing the event deposits. Radiocarbon dating and tephra analysis allowed us to establish the geochronology in the KYD-trench wall sediments and to correlate tsunami deposits with historical tsunami events. The four younger tsunami deposits (the E1–E4 deposits) are correlated with the 2011 Tohoku-oki tsunami, the 1896 Meiji Sanriku tsunami, the 1611 Keicho Sanriku tsunami, and the 869 Jogan tsunami events, respectively. The average interval of tsunami occurrence at Koyadori is estimated at 290–390 years based on continuous records in the KYD-trench. However, the age intervals between the E1 to E4 deposits are variable (E1/E2: 115 years, E2/E3: 285 years, E3/E4: 742 years), likely reflecting the diversity of the tsunami generation mechanism and/or different earthquake sources around the Japan Trench. By correlating the historical tsunami runup height data with extant tsunami deposits, we could preliminarily estimate the sizes of paleo-tsunamis at the study site. In the future study, we need to confirm our tsunami correlations by correcting many geological data along the Sanriku Coast. Ultimately, we aim to assess tsunami risk and understand the earthquake phenomena around the Japan Trench.