Marine inundation history during the last 3000 years at Lake Kogare-ike, a coastal lake on the Pacific coast of central Japan

Abstract

Sediment cores collected at Lake Kogare-ike, a coastal lake on the Pacific coast of central Japan, record the marine inundation history during the last 3000 years. The sediments consist mainly of organic mud, sand, gravel, inorganic mud, and volcanic ash, and inundation events were recognized as 19 event deposits (E1–E19, from top to bottom) interbedded with the organic mud. Visual observation by naked eyes and X-ray computed tomography (CT) images identified 16 event deposits based on quantitative and qualitative changes in sand contents and changes in the textures and colors of the sediment samples (E1–E3, E5, E6, E8, E9, and E11–E19). The other three event deposits (E4, E7, and E10) were identified only on the CT images as layers with higher radiodensity than the underlying and overlying organic mud layers. The sedimentary features, the spatial bias of the event deposits toward seaward areas, the diatom assemblages, and the frequency of inundation events suggest that 13 (E1–E10 and E12–E14) of the 19 event deposits were formed by tsunamis or extraordinary storms. To constrain the depositional ages of the event deposits, Bayesian age–depth models were constructed based on radiocarbon dating of plant macrofossils and concentrated fossil pollen and the ¹³⁷Cs profile. The depositional ages of the event deposits indicate that five or possibly six event deposits can be correlated with historical tsunamis along the Nankai Trough: E2, either of E3 or E4, E5, E7, and E9 correspond to the 1707 CE Hoei, the 1605 CE Keicho, the 1498 CE Meio, the 1096 CE Eicho, and the 684 CE Hakuho tsunamis, respectively. E1 was possibly formed by the 1944 CE Showa-Tonankai tsunami, the 1854 Ansei–Tokai tsunami, the 1959 Isewan typhoon, or a combination of two or all three events.

1 Introduction

Infrequent but extraordinary extreme waves, such as tsunamis and storms, impact coastal environments. Instrumental records (e.g., from tide gauges and oceanographic observation buoys) are generally available only for within the last century. Historical documents are also used to comprehend the history of past extreme wave events, but such events are difficult to reconstruct in detail on millennial or centennial timescales because of the shortness and/or incompleteness of historical records. To supplement these records and to extend the history of extreme events, geological records have been employed in tsunami geology (Hindson et al. 1996; Dawson et al. 1988, 2020; Atwater et al. 2005; Cisternas et al. 2005; Jankaew et al. 2008; Costa and Andrade 2020; Sawai 2020; Ishimura et al. 2022) and paleotempestology (Liu and Fearn 1993, 2000; Woodruff et al. 2009; Yao et al. 2020; Bianchette et al. 2022) worldwide to better understand the history of extreme waves and aid with prediction of future such events. In both research fields, event deposits within lacustrine and marshy sediments have served as a guide to reconstruct inundation areas (Nanayama et al. 2003; Sawai et al. 2012; Pilarczyk et al. 2021) and the recurrence history of extreme waves (Liu and Fearn 1993; Nanayama et al. 2003; Sawai et al. 2009a).

The Pacific coast of central and western Japan has been repeatedly inundated by tsunamis and storms (Arakawa et al. 1961; Watanabe 1998; Usami et al. 2013; Ishibashi 2014), but the inundation history over the past thousands of years has not been fully reconstructed. For tsunami history, the governmental long-term evaluation of subduction-zone earthquakes in this region is based mainly on historical documents recording the nine great earthquakes and their accompanying tsunamis that have occurred during the last 1300 years along the Nankai Trough (The Headquarters for Earthquake Research Promotion 2013; Fig. 1d). However, records of earthquakes before the seventeenth century are fewer and more fragmentary than those of earthquakes after the eighteenth century (Koyama 1999). Koyama (1999) pointed out that there is a lack of records for the medieval period or earlier. In addition, the 1300 years of history recorded in written documents is too short to evaluate earthquakes because unusually giant earthquakes recur at intervals of hundreds of years (Satake and Atwater 2007). In evaluations of storm surges and waves, the shortness of the historical and instrumental records is the most problematic issue. Written documents for the past 300 years show that the Pacific coastal areas of Japan have been repeatedly affected by storms (Arakawa et al. 1961), but there is little information on the exact values of inundation heights and distances. Specific data from instrumental records are available only for the past 70 years. Recent instrumental records have demonstrated that typhoons, as shown, for example, by the Isewan Typhoon in 1959, can cause widespread inundation of the Pacific coast of central Japan, but the frequency of such extraordinary typhoons is not known. Datasets based on geological records are therefore required for long-term assessment of coastal hazards resulting from marine inundation in this coastal area.

Fig. 1

a Index map of Japan. b Map of eastern Mie Prefecture showing the study site. c Location of the “Lake Kogare-ike” study site on the eastern shore of Kamizaki Bay. d Ages and estimated rupture zones of historical tsunamis along the Nankai Trough (modified from The Headquarters for Earthquake Research Promotion 2013). Red solid lines, red dotted lines, and gray dotted lines represent certain, probable, and possible rupture zones, respectively. The 1605 Keicho earthquake (dotted blue line) is regarded as a tsunami earthquake or an earthquake off an adjacent trench (Yamamoto and Hagiwara 1995; Seno 2002; Ishibashi and Harada 2013; Tsuji 2016). Generic Mapping Tools (Wessel et al. 2013) was partly used to create Fig. 1d. e Sediment sampling locations and topographic setting of the study site. The photograph was captured by an unmanned aerial vehicle. f Aerial photograph of the study site showing the survey transect from the shoreline (point A) to the inland end of Lake Kogare-ike (point G). The photograph was taken by the Geospatial Authority of Japan on 24 September 2008 (photograph number: CKK-2008-1X-C13A-13). g Topographic profile from the shoreline to the study site. Elevations are expressed with respect to mean sea level at Tokyo Bay (Tokyo Peil, T.P.)

Geological evidence for tsunamis and storms has been reported from more than 80 areas along the Nankai Trough (e.g., Tsuji et al. 2002; Okamura and Matsuoka 2012; Kitamura et al. 2013); however, inter-regional correlation among these areas remains difficult because of sparse dates (Okamura and Matsuoka 2012) and discontinuities in the sedimentary record (Fujino et al. 2018; Tanigawa et al. 2018; Shimada et al. 2019). In this study, we obtained sediment cores containing a continuous record from 3000 years ago to the present day from Lake Kogare-ike, a coastal lake facing the central part of the Nankai Trough (Fig. 1). Application of X-ray computed topographic imaging and construction of Bayesian age–depth models for the sediment samples enabled us to recognize faint geological records of extreme wave events in historical and prehistoric time. The inundation history deduced in this study can provide new insights into inter-regional correlations with previously reported tsunami deposits along the Nankai Trough and contribute to the reconstruction of prehistoric and historic tsunami history.

2 Setting

2.1 Tectonics

Subduction-zone earthquakes occur repeatedly in the Nankai Trough, where the Philippine Sea plate is subducting beneath the Eurasian plate (Ishibashi 2004; Sangawa 2007; Usami et al. 2013; Fig. 1a). The coastal areas along the Nankai Trough are conventionally divided into six parts, Z and A–E (Fig. 1d) or, more simply, four segments (the Hyuga-nada segment, Z; the Nankai segment, A and B; the Tonankai segment, C and D; and the Tokai segment, E). Our study site is located in part C of the Tonankai segment. Instrumental and written records indicate that earthquakes have occurred with variable rupture zones and the associated tsunamis have inundated coastal areas in this region (Central Meteorological Observatory 1945; Ishibashi and Satake 1998; Fig. 1d). The latest two earthquakes on the Tonankai segment (the 1944 CE Showa-Tonankai and 1854 CE Ansei–Tokai earthquakes) caused coseismic coastal subsidence in the study area in Minami-ise town (Central Meteorological Observatory 1945; Hatori 1978; Education Center of Nanto Town 2000) (Fig. 1b). The geological record indicates that earthquakes and tsunamis have occurred along the Nankai Trough not only in the historical period but also over the past few thousand years (Garrett et al. 2016; Fujiwara et al. 2020a).

2.2 Historical records of extreme waves around the study site

2.2.1 Tsunamis

Instrumental records, written documents, tsunami monuments, and oral tradition record the latest three historical tsunamis associated with earthquakes along the Nankai trough: the 1944 CE Showa-Tonankai; the 1854 CE Ansei–Tokai; and the 1707 CE Hoei tsunamis (e.g., Editing Committee of the History of Nanto Town 1985; Watanabe 1998; Education Center of Nanto Town 2000; Editing Committee of the History of Nansei Town 2004; Namegaya and Tsuji 2005). Instrumental records show that the wave height of the 1944 CE Showa-Tonankai tsunami was 5.4 m on the eastern shore of Kamizaki Bay (Editing Committee of the History of Nanto Town 1985; Fig. 1c). Namegaya and Tsuji (2005) proposed, based on a local historical document, that the western shore of Kamizaki Bay was inundated by the 1854 CE Ansei–Tokai tsunami with a run-up height of 5.2 m Tokyo Peil (T.P.). They also estimated the run-up height of the 1707 CE Hoei tsunami on the eastern shore of Kamizaki Bay to have been 8.6 m, from an oral tradition that the tsunami reached Higashi-Shinta, approximately 1 km from the shoreline (Fig. 1c).

2.2.2 Storms

According to historical records, at least 55 storms attacked the area around Ise Bay (locally called “Isewan”) between 1600 and 1865 CE (Arakawa et al. 1961; Fig. 1b, d; Additional file 1: Table S1). Instrumental records of wind directions, storm surges, and damage are available for storms since the mid-twentieth century. As many as 29 storms affected Mie Prefecture (Fig. 1a) between 1945 and 2017 CE (Japan Meteorological Agency 2017). The largest storm was the 1959 CE Isewan Typhoon, during which the highest storm surge in the observation history in Japan was recorded (Committees for Technical Investigation on Lessons Learned from Past Disasters 2008). The storm surge resulted in hazardous coastal inundation along Ise Bay and its surroundings. The inundation reached more than 15 km inland in the coastal areas north of Ise Bay (Fig. 1d) in association with a sea-level rise to 4.27 m T.P. (Japan Meteorological Agency 1961). Around the study site, sea level rose to 2.8 m T.P. and inundation extended more than 10 km inland at Gokasho Bay (Fig. 1b). According to one observation report, sea level rose to 2.3 m T.P. on the western shore of Kamizaki Bay (Editing Committee of the History of Nanto Town 1985; Fig. 1c).

2.3 Geomorphology

Coastal lakes and ponds are useful archives for reconstructing the long-term and continuous history of inundation by extreme waves (e.g., Sugawara et al. 2008; Baranes et al. 2016). These water bodies are usually stable depositional environments, in which coarser-grained event deposits sporadically emplaced by extreme waves are interbedded with muddy lacustrine sediments. Coastal wetlands such as swales and marshes are also commonly studied to identify event deposits (e.g., Atwater 1987; Cisternas et al. 2005; Pilarczyk et al. 2021; Okada et al. 2022), but coastal lakes/ponds have the advantage of a high preservation potential due to their stable, calm environment, and unfilled accommodation space (Szczuciński 2012; Kempf et al. 2017). This setting usually prevents event deposits from being completely eroded and yields reliable materials for dating.

Our study site was a coastal lake called Kogare-ike, which is isolated from the Pacific Ocean by a sand barrier (Fig. 1e, f). The height of the barrier is around 5 m T.P., though the maximum height is 4.8 m T.P. on the surveyed transect (Fig. 1f). Kamizaki Bay is a part of a complex ria-type coast, and the lake is surrounded by hills. The hills are composed of Upper Cretaceous pelitic mixed rock that includes blocks of basalt, chert, reddish siliceous mudstone, and sandstone (Nishioka et al. 2010). A narrow drainage channel connects the lake to the sea, but inflow and outflow are controlled by a sluice at present (Fig. 1e). The lake is more than 1 m deep in most parts and becomes deeper toward the center (Fig. 1f). The maximum depth is 2.3 m, and the elevation of the lake bottom at that point is approximately − 1.5 m T.P. The mean sea level at the study site is approximately equal to T.P.; the mean tide level at Owase, the nearest tide station to the study site, has been + 0.07 m T.P. for the past 10 years (from 2012 to 2021 CE) (Japan Meteorological Agency 2022a, 2022b). Agricultural activity around our study site can be recognized in the land use map for 1892, and aerial photographs show that farming was abandoned around 2003 (Additional file 2: Figure S1). This date is consistent with information from a landowner that the area of fringe marshes was used for agriculture before it was abandoned 10–15 years ago. There is no large freshwater input to the lake.

2.4 Previous studies on tsunami deposits in the study area

The inundation history at Lake Kogare-ike has not been explored previously, except for a preliminary study by Komatsubara et al. (2007), who collected three short (150-cm-long) cores from the lake bottom. They identified a sandy event deposit interbedded with muddy sediments by visual observation. A radiocarbon age of 980 ± 20 BP (recalibrated by this study to 958–791 cal BP) was obtained approximately 20 cm below the sand layer.

The geological evidence for earthquakes and tsunamis along the Nankai Trough has been summarized in a systematic review by Garrett et al. (2016). Here we summarize the geological record in areas adjacent to our study area, including reports published after Garrett et al. (2016). In the area east of our study site, tsunami deposits have been reported by Fujino et al. (2018) and Okahashi et al. (2005). Fujino et al. (2018) documented 10 sandy tsunami deposits dating to between 4500 and 500 cal BP in the Shijima lowland (Fig. 1b). The tsunami deposits were identified within muddy sediments by visual inspection and on X-ray radiographs. Three of these tsunami deposits were correlated with historical tsunamis (the 684 Hakuho tsunami, 1096 Eicho tsunami, and 1498 Meio tsunami). Okahashi et al. (2005) identified 12 tsunami deposits in a marsh at Osatsu (Fig. 1b) based on sedimentary structures and micropaleontological evidence. They obtained radiocarbon ages from wood fragments and cones, but did not construct a chronology or estimate the depositional ages of each event deposit. To the west of our study site, Komatsubara et al. (2007) reported their preliminary results of stratigraphic observation of two short cores. They found 5–11 event deposits within peaty deposits at the Shihara lowland (Fig. 1d), but did not carry out detailed analysis of the samples.

3 Methods

3.1 Collection of sediment samples

We collected lacustrine sediments from seven locations at Lake Kogare-ike (K-1 to K-7; Fig. 1e, f) in 2018 and 2019. We assembled a raft with a hole in the center as a platform and pushed a 50-cm-long Russian sampler down to 4.75 m depth below the lake bottom. To recover continuous sedimentary records, we collected alternate overlapping samples from three or four different parallel holes. The holes were less than 1 m apart and overlaps of 25 cm were applied. In addition, push-core samples were obtained at K-1, K-4, and K-5 to recover soft surface sediments. The elevations of the ground surface and the present sea level were obtained with a network Real Time Kinematic-Global Navigation Satellite System (RTK-GNSS) survey system from Leica Geosystems Inc. (Norcross, Georgia, USA). The standard deviation (1 SD) of the leveling in our study was 0.005–0.017 m horizontally and 0.010–0.043 m vertically. The elevations of the lake bottom were obtained by subtracting the water depth measured with a leveling rod from the elevation of the lake surface measured with RTK-GNSS.

3.2 Observation and grain size analysis of sediment samples

We observed the collected sediment samples in both the field and the laboratory. In the laboratory, X-ray computed tomography (CT) scans of the cores were obtained with a Hitachi Supria Grande Premium CT scanner (Hitachi, Ltd., Tokyo, Japan) at the Geological Survey of Japan (Tsukuba, Ibaraki, Japan). The CT images were referred to during visual observation and used to guide subsampling for radiocarbon dating, gamma-ray spectrometry, grain size analysis, and diatom assemblage analysis.

Grain size analyses of the samples were conducted using a Retsch Camsizer (Verder Scientific, Haan, Germany), an image analyzer capturing a wide range of grain size from silt to pebbles. All samples were sieved in water to remove the mud component and then treated with hydrogen peroxide solution to remove the organic components. During the procedure, large pieces of organic matter (roots, leaves, and wood fragments) were removed by hand. The sand content of each sample was calculated from the resulting weight reduction. The measurement range of grain size analysis on the residual sand fractions was set from − 5.25 to + 6.25 phi at intervals of 0.25 phi. Grain size distribution properties were calculated using the logarithmic graphical method of Folk and Ward (1957).

3.3 Analysis of volcanic glass shards

The elemental concentrations of volcanic glass shards in event deposit E11 at locations K-3, K-4, K-6, and K-7 were measured with an energy dispersive X-ray microanalyzer (EDX) by Furusawa Geological Survey Inc., Okazaki, Japan. Identification of glass sources was carried out by referring to previously published regional studies (Sugihara 1984; Machida and Arai 2003; Sugiuchi and Fukuoka 2005; Tsukui et al. 2006; Kobayashi et al. 2020; Murata et al. 2021).

3.4 Diatom analysis

Fossil diatom assemblages in 127 samples from the core from K-4 were studied to reconstruct the paleoenvironment of the study site. The samples were prepared by the bleaching method (Nagumo and Kobayasi 1990; Nagumo 1995). Fossil diatoms were identified under an optical microscope. More than 200 valves were counted per sample. Identification and ecological information followed standard (e.g., Patrick and Reimer 1966, 1975; Simonsen 1987; Krammer and Lange-Bertalot 1986, 1988, 1991a, b) and regional (Kosugi 1988; Kobayasi et al. 2006) sources. Fossil diatoms were divided into seven groups on the basis of their autoecology: freshwater plankton, freshwater–brackish plankton, freshwater tychoplankton, freshwater benthos, freshwater–brackish benthos, brackish benthos, and unknown.

Stratigraphically constrained cluster analysis (Grimm 1987) was performed to statistically identify the zonation of fossil diatom assemblages. The cluster analysis was carried out using the Bray–Curtis distance among the assemblages and constraining the stratigraphic order of the samples. The cluster analysis was executed using the package “rioja” in R version 4.2.2 (Juggins 2022).

3.5 Radiocarbon dating, gamma spectrometry, and Bayesian age–depth modeling

Radiocarbon dating of plant macrofossils and fossil pollen grains was conducted to estimate depositional ages. Macrofossils were selected under a binocular microscope. Fossil pollen grains were extracted from sediment samples at the Research Centre for Palaeoclimatology, Ritsumeikan University (Kusatsu, Shiga, Japan) following the procedure of Yamada et al. (2021). Radiocarbon dating of plant macrofossils and pollen grains was carried out by Beta Analytic (Miami, Florida, USA) and the University of Tokyo (Tokyo, Japan), respectively.

Cesium-137 activities of 29 samples from shallower than 140 cm depth in the sediment core collected at K-4 were measured by high-resolution gamma spectrometry (GCW2022; Mirion Technologies [Canberra], Inc, Meriden, Connecticut, USA). Cesium-137 is one of the most common radionuclides used to establish chronologies for sediments during the last 80 years. Cesium-137 is an anthropogenic radionuclide that was produced by atmospheric nuclear weapons testing from 1950 to ∼1975. In the northern hemisphere, undisturbed sedimentary records generally show the first occurrence (onset) and the peak of ¹³⁷Cs as time markers for 1954 and 1963–1964, respectively (Ritchie and McHenry 1990; Appleby 2001), though ¹³⁷Cs onset might have occurred earlier in western Japan than in other regions in the northern hemisphere because of the nuclear bomb attacks on Hiroshima and Nagasaki in World War II (see discussion).

A Bayesian age–depth model of the sediments at K-4 was constructed based on the radiocarbon ages. We applied Bchron (Haslett and Parnell 2008; Parnell et al. 2008), a package written for the free open-source language R (R Core Team 2021). Age–depth models were constructed by using IntCal20 (Reimer et al. 2020) as the radiocarbon calibration curve. Event deposits were treated as instantaneous, and their thicknesses were removed from the original core log of K-4 to construct an event-free stratigraphic column (Kempf et al. 2017, 2020; Sabatier et al. 2017). For this reason, the radiocarbon age measured within event deposit E1 (no. 20 in Table 1) was not used in the age–depth model. Because of the stochastic nature of the Bchron model, which uses random sampling, each run produces slightly different depositional ages; therefore, the depositional ages of event deposits E1–E19 were determined by averaging the ages obtained from the three trials.

Table 1 Radiocarbon ages from the study site

4 Results

4.1 Stratigraphy and ages of sediment samples from Kogare-ike

4.1.1 Detailed description of the stratigraphy at the center of the lake

The sediments from Lake Kogare-ike were composed of organic mud, sand, gravel, inorganic mud, and volcanic ash. Sand, sandy mud (with/without gravel), inorganic mud, and white volcanic ash layers were observable with the naked eye both in the field and in the laboratory as distinctive layers interbedded with organic mud. In addition to these, there were 10 high-radiodensity layers that could be recognized only in X-ray CT images. The high-radiodensity layers possibly possessed different mineral compositions (Boespflug et al. 1995) or larger grain size (Godbout et al. 2019) than the organic mud layers. We regard these sand, sandy mud, inorganic mud clasts, volcanic ash, and high-radiodensity layers as event deposits (Figs. 2, 3 and 4; Additional file 3: Figure S2, Additional file 4: Figure S3, and Additional file 5: Figure S4; Additional file 6: Table S2) and have named them E1–E19 from top to bottom. Details of each event deposit at K-4 are described in the following paragraphs. Details of the contents of mud, organic material, and sand and the results of grain size analysis in each layer are provided in Fig. 4 and Additional file 7: Table S3.

Fig. 2

Lithostratigraphy from K-1 to K-5 (locations in Fig. 1e). Radiocarbon ages are listed in Table 1. The correlations of event deposits between core sites are shown as solid lines, with tentative correlations indicated by dotted lines. Stacks of X-ray CT images are shown on Additional file 3: Figure S2 and Additional file 4: Figure S3. A photograph of the core from K-4 is provided in Additional file 3: Figure S2

Fig. 3

Lithostratigraphy of a transect connecting K-6, K-4, and K-7 (locations in Fig. 1e). Radiocarbon ages are listed in Table 1. The correlations of event deposits between core sites are shown as solid lines, with tentative correlations indicated by dotted lines. Stacks of X-ray CT images are shown on Additional file 3: Figure S2 and Additional file 4: Figure S3. A photograph of the core from K-4 is provided in Additional file 3: Figure S2

Fig. 4

Photographs, X-ray CT images, and sketches of the event deposits identified at K-4 with the results of grain size analysis (sand content and mean grain size of residual sand fractions). The mean grain size, shown as black circles, represents that of samples with very small sand contents (10% or less)

E1, the uppermost event deposit, was gray in color and ranged from 46.9 to 77.1 cm core depth. This event deposit was visible as a stack of three subunits both to the naked eye and in the CT images (Fig. 4). We could not recognize an erosional lower contact with the underlying organic mud in the CT images. Parallel laminae were distinct within the subunits. The sand content of E1 was between 0.9 and 3.4 wt.% (Fig. 4 and Additional file 7: Table S3).

E2 appeared as a grayish faint inorganic mud layer to the naked eye, and was visible as a high-radiodensity layer from 140.0–141.8 cm depth in the CT images (Fig. 4). The sand content was low (< 1 wt.%) above, below, and within event deposit E2 (Fig. 4 and Additional file 7: Table S3).

E3 was a sandy layer at 153.6–156.2 cm core depth. The lower part of E3 consisted of medium sand (mean grain size = 1.55 φ). The upper part of the event deposit contained multiple layers of gray inorganic mud and brown organic mud. Changes in the sand contents show grading in grain size of this event deposit. The upper part was difficult to recognize with the naked eye but was clear on the CT images (Fig. 4).

E4 and E5 were recognized as high-radiodensity layers only on the CT images at 172.9–173.8 cm and 183.9–184.5 cm core depth, respectively. Event deposit E4 showed very faint lateral continuity within the core. The sand contents of the two event deposits were less than 1 wt.% above, below, and within the layers (Fig. 4 and Additional file 7: Table S3).

E6 occurred between 207.5 and 219.8 cm core depth. As shown in the CT images, this event deposit was not clearly divided into subunits, in contrast to the younger events. It was difficult determine from visual observation and CT images whether the non-layered structure of E6 was created by obvious bioturbation or deformation of the sand layer after rapid deposition (e.g., load casting). The sand content of E6 ranged from 10.5 to 31.2 wt.%. E6 contained a small proportion of granules, but the mean grain size of the sieved residual sand fractions was fine–coarse sand (2.49–0.25 φ, average = 1.17 φ). The sand content and the results of grain size analyses show grading within the event layer.

A faint event deposit (E7) at 231.0–231.3 cm core depth was recognized only on CT images. This deposit was visible as a rather patchy horizontal distribution within the core, rather than a stratified high-radiodensity layer. The sand content was low (less than 1 wt.%) above, below, and within event deposit E7.

E8 was a light grayish inorganic mud layer between 269.1 and 273.6 cm core depth. A few mud clasts were recognized both in the field and on CT images at the base of the event deposit. The CT images showed an upward change in radiodensity from high to low, which may represent upward fining of grain size; however, no grading was recognized in the sand content or grain size analysis.

E9 and E10 were visible only on the CT images at 283.4–284.3 cm and 303.4–304.2 cm core depth, respectively. The two layers exhibited high radiodensity and were distinctively stratified on the images, but there was no significant change in the sand content (< 1 wt.%) or the grain size of the residual sand fractions.

E11 appeared at 313.3–317.9 cm core depth. The bottom part of the event deposit was a very distinctive white layer in the field. The middle and upper parts of the layer were hard to recognize with the naked eye but visible on the CT images. This event layer consisted of pumice and glass shards. EDX analyses indicated that the shards had various origins (see below), suggesting erosion and redeposition from older strata.

A faint event deposit, E12, was visible only on CT images at 337.2–338.4 cm core depth. This event deposit was apparent as a patchy distribution within the core, and the sand content of the sediment was less than 1 wt.% above, below, and within the event deposit.

E13 was distributed at 346.5–350.9 cm core depth. To the naked eye, the upper part of the event layer was visible as a light grayish inorganic mud layer. The upper part had higher density than the lower part in the colored CT image. There was no significant difference in sand content between the upper and lower parts.

E14 was identified only on the CT images, and occurred between 352.6 and 354.8 cm core depth. The CT images showed higher radiodensity at the bottom of the layer than in the middle and upper parts, but no difference was recognizable with the naked eye. The sand content was persistently low (less than 1 wt.%) above, below, and within the event deposit.

E15 occurred between 388.5 and 406.0 cm core depth. The CT images showed three sandy and muddy subunits within the event deposit. The middle subunit appeared not to be stratified, but we could not determine whether the lack of strata was caused by bioturbation or load casting. The sand contents and grain size analyses of the subunits indicated that the lowest subunit contained much more sand and coarser fractions than the higher subunits. The mean grain size of the uppermost, middle, and lowest subunits was fine sand (mean grain size is 2.68–2.12 φ), very fine sand (mean grain size = 3.57–3.45 φ), and medium sand (mean grain size = 1.37 φ), respectively. Debris filled the space between the upper and middle subunits. The lower contact with the underlying deposit was sharp (less than 5 mm).

E16 was visible to the naked eye as a distinct grayish sandy layer at 416.2–417.4 cm core depth. The sand content and mean grain size of the event were 48.5 wt.% and coarse sand (mean grain size = 0.85 φ), respectively. The upper and lower contacts with the overlying and underlying deposits were sharp.

E17 was a grayish sandy mud layer at 436.2–446.7 cm core depth. The CT images showed clear parallel laminae in the upper and basal parts of the event deposit. The sand content in the basal part of the event deposit was low sand content (< 1 wt.%), whereas the upper three samples within the event deposit showed relatively high sand content (11.6 to 39.7 wt.%). The mean grain size of the residual sand fractions of the upper three samples was fine–coarse sand (2.72– 0.53 φ, 1.52 φ on average). The sand contents and mean grain size indicated upward fining of grain size. The well-preserved clear laminae may suggest that this event layer had not been vertically disturbed, for example, by benthic fauna, after deposition.

E18 was recognized as a grayish sandy mud layer at 449.1–459.1 cm core depth. The CT images enabled us to recognize clear laminae in the upper and lower parts of the layer. The sample in the upper part had low sand content (7.5 wt.%). The sand contents in other four samples ranged from 12.6 to 48.4 wt.%. The mean grain size of the residual sand fractions in these four samples was very fine to medium sand (1.79 to 3.22 φ, average 2.57 φ). The CT images and changes in the sand content and mean grain size indicated inverse and normal grading within the event layer.

The lowest event deposit, E19, was a gray distinctive layer that consisted of sandy mud, muddy sand, and granules. The CT images showed laminae, visible as high-radiodensity layers. The sand content was 1.2–64.4 wt.%. Grain size analyses showed that the residual sand fractions were in the medium sand range (mean grain size of the four samples with a sand content of more than 10% = 1.08 phi). The top of E19 was observed at 461.3 cm core depth, but the bottom was not clearly confirmed in our fieldwork.

4.1.2 Major-element composition of volcanic glass shards

The major-element compositions of individual glass shards from the volcanic ash layer (event deposit E11) showed various sources and ages of eruptions (Additional file 8: Fig. S5). Major elements, especially MgO, CaO, Na₂O, and K₂O, are useful to distinguish the sources of volcanic ash layers around the study site (Sugihara 1984; Machida and Arai 2003; Kobayashi et al. 2020; Murata et al. 2021). Scatter plots of MgO–SiO₂, CaO–SiO₂, Na₂O–SiO₂, and K₂O–SiO₂ for volcanic ash corresponding to E11 in four cores (K-3, K-4, K-6, and K-7) recognized four clusters of element compositions. Three of the clusters correspond to widespread tephras; specifically, the Aira-Tn tephra (AT; 29–26 cal kBP), the Kikai-Akahoya tephra (K-Ah; 7.3 cal kBP), and the Amagi-Kawagodaira tephra (Kg; 3.1 cal kBP) (Sugihara 1984; Machida and Arai 2003). The other cluster is similar to the element composition of Izu Island eruptions (low CaO and high Na₂O and K₂O), especially the Kozushima-Tenjosan tephra (Iz-Kt; AD 838) (Sugihara 1984; Sugiuchi and Fukuoka 2005; Tsukui et al. 2006; Kobayashi et al. 2020; Murata et al. 2021), though this age exhibits a discrepancy with the radiocarbon dating.

Fig. 5

The ¹³⁷Cs profile in the top of the core from K-4. Details of the analytical results are listed in Additional file 9: Table S4

4.1.3 Dating by radioactive isotopes

The obtained radiocarbon ages ranged from 2836–2516 to 282–0 cal BP (Figs. 2 and 3; Table 1). Forty-seven radiocarbon ages were obtained, 30 of which were from plant macrofossils and the remaining 17 from concentrated pollen fossils.

Cesium-137 was detectable in samples above 93 cm in depth at K-4, but its concentration in deeper samples was below the detection limit (Fig. 5 and Additional file 9: Table S4). The concentration of ¹³⁷Cs ranged from 0.8 ± 0.2 to 1.9 ± 0.2 Bq/kg in samples between 93 cm and 82.5 cm core depth (below E1), but exhibited a pronounced increase just above E1. The highest ¹³⁷Cs concentration was recognized at 42.5 cm depth below the lake bottom (28.5 ± 0.6 Bq/kg). Thereafter, the concentration decreased consistently upward, except for a slight increase from 30.5 to 27.5 cm core depth.

4.1.4 Intra-site correlation of event deposits among cores taken from Kogare-ike

The stratigraphic features and positions of event deposits, chemical composition of glass shards, and radiocarbon ages permitted us to correlate event deposits E1–E16 at K-4 with event deposits in the cores taken from the other six locations at Kogare-ike (K-1, K-2, K-3, K-5, K-6, and K-7; Figs. 2 and 3). No event deposits correlative with E17–E19 were present in any cores except for K-4, because of short recovery of the sediment cores taken from K-1, K-2, K-3, K-5, K-6, and K-7. The correlations of the other events (E1–E16) are described below; the characteristics of each event are summarized in Additional file 6: Table S2.

Event deposit E1 found at K-4 correlated among K-1–K-6 and possibly K-7. The correlation is made on the basis of the depth of the lower contact of the event at K-1–K-7 (about 50–100 cm from the lake bottom) and the stratigraphic features at K-1 and K-3–K-6 (multiple subunits of high radiodensity and parallel laminae).

Event deposits E2 and E3 at K-4 were found at K-1–K-4 and K-6. The two event deposits were present in organic mud younger than 461–295 cal BP (no. 1 in Table 1) at K-1 and within organic mud dated between 282 and 0 cal BP (no. 9 in Table 1) and 475–0 cal BP (no. 10 in Table 1) at K-2 and 311–0 cal BP (no. 21 in Table 1) and 458–295 cal BP (no. 22 in Table 1) at K-4. Two event layers at each of K-3 and K-6 were correlated with E2 and E3 on the basis of their depth and stratigraphic features: E2 is a very faint layer and E3 has multiple laminae with high radiodensity.

Event deposits E4 and E5 were correlated among K-1–K-4 and K-6. The two events are located within organic mud dated between 461 and 295 cal BP (no. 1 in Table 1) and 730–668 cal BP (no. 2 in Table 1) at K-1, between 475 and 0 cal BP (no. 10 in Table 1) and 655–551 cal BP (no. 11 in Table 1) at K-2, and between 458–295 cal BP (no. 23 in Table 1) and 670–555 cal BP (no. 25 in Table 1) at K-4. The correlations are also based on a radiocarbon age between E4 and E5 at K-3 (729–672 cal BP, no. 17 in Table 1) and a limiting maximum age of E5 at K-6 (651–551 cal BP, no. 44 in Table 1). These deposits’ stratigraphic features (i.e., a very thin layer, or visible only on CT images) are consistent among the cores.

Event deposit E6 of K-4 can probably be correlated with event deposits at K-1–K-3, and possibly K-6 and K-7. This correlation is based on stratigraphic features, depths of event layers, and three radiocarbon ages. Stratigraphic features visible as multiple layers of high radiodensity in the CT images are very similar between event layers at K-2 and K-3. However, high-radiodensity layers correlative to deposit E6 are very thin at K-1. We correlated event deposit E6 of K-4 with those of K-1–K-3 on the basis of the depths of the event deposits and four radiocarbon ages (655–551 cal BP, no. 11 in Table 1; 908–732 cal BP, no. 18 in Table 1; 670–555 cal BP, no. 24 in Table 1; and 915–729 cal BP, no. 25 in Table 1).

A faint event deposit, E7, is found at K-1–K-4, and possibly K-6 and K-7. The depositional age of E7 at K-4 is constrained by three limiting maximum ages (782–678 cal BP, no. 26 in Table 1; 670–557 cal BP, no. 27 in Table 1; and 1344–1179 cal BP, no. 28 in Table 1) obtained from samples below the event deposit E7 and a limiting minimum age (915–729 cal BP; no. 26 in Table 1) from a sample above E7. At K-1 and K-2, two high-radiodensity layers correspond to the constrained age of E7 at K-4. Either of these two layers may be correlated with E7 of K-4; alternatively, the two high-radiodensity layers may represent a single event such as a single tsunami deposit having multiple subunits due to multiple waves. Event layers at depths of ~ 284 cm at K-6 and ~ 229 cm at K-7 are possibly correlated either E6 or E7 based on their depths on CT images and stratigraphic position above the light grayish inorganic mud layer (E8).

Event deposit E8 was correlated among K-1–K-4, K-6, and K-7 by means of sedimentary features (light grayish inorganic mud and mud clasts). Radiocarbon ages from samples above E8 at K-1 (901–690 cal BP, no. 4 in Table 1) and below E8 at K-2 (1290–1154 cal BP, no. 14 in Table 1) is consistent with this visual correlation. An event deposit at 362.1–367.0 cm depth at K-5 may also correlate with E8, but an age from a sample below the event (928–798 cal BP, no. 42 in Table 1) is younger than event deposit E8 at K-4.

Event layers corresponding to event deposit E9 at K-4 were found at K-1, K-2, K-3, K-6, and K-7. The depositional age of E9 at K-4 is constrained by a limiting maximum age (1289–1177 cal BP, no. 30 in Table 1) and a limiting minimum age (1294–1126 cal BP, no. 29 in Table 1). This age is consistent with a radiocarbon age from a sample below the correlated event deposit of K-1 (1174–958 cal BP, no. 5 in Table 1) and two ages from samples above and below the correlated event deposit of K-2 (1290–1154 cal BP, no. 14 in Table 1; 1300–1177 cal BP, no. 15 in Table 1). The event deposits at K-3, K-6, and K-7 are common in their relative position below the light grayish inorganic mud layer (E8).

Event deposit E10 at K-4 is found at all locations but K-5. The depositional age of E10 at K-4 is constrained by two radiocarbon ages (1289–1177 cal BP, no. 30 in Table 1; 1404–1309 cal BP, no. 31 in Table 1). This age is consistent with radiocarbon ages below an event deposit at K-1 (1367–1294 cal BP, no. 6 in Table 1), above an event deposit at K-2 (1300–1177 cal BP, no. 15 in Table 1), and below an event deposit at K-3 (1404–1309 cal BP, no. 19 in Table 1). The event deposits observed at K-6 and K-7 were correlated with event deposit E10 at K-4 on the basis of their relative position to the white volcanic ash layer (E11).

Event deposit E11 of K-4, the volcanic ash layer with various sources, correlated with event layers found in K-2, K-3, K-6, and K-7. Because the event deposits were visible as white ash layers, preliminary correlation among cores was possible in the field. The event deposits consisting of volcanic ash at K-2, K-3, K-4, K-6, and K-7 were composed of glass shards. As shown by major chemical component analysis, the event deposits at K-3, K-4, K-6, and K-7 had in common that they contained glass erupted by different ages and different sources. The similar compositions of the volcanic glass shards within these event deposits and the radiocarbon ages from below and above the event deposits (1404–1309 cal BP, no. 19 in Table 1; 1404–1309 cal BP, no. 31 in Table 1; 1516–1353 cal BP, no. 32 in Table 1; and 1293–1177 cal BP, no. 45 in Table 1) enable correlation of the event layers among K-3, K-4, K-6, and K-7.

Event deposits E12, E13, and E14 of K-4 were found at K-1, K-2, K-6, and K-7. The three event deposits are present in organic mud dated between 1398–1308 cal BP (no. 33 in Table 1) and 1687–1414 cal BP (no. 34 in Table 1) at K-4, between 1367–1294 cal BP (no. 6 in Table 1) and 1829–1627 cal BP (no. 8 in Table 1) at K-1, older than 1293–1177 cal BP (no. 45 in Table 1) at K-6, and younger than 1821–1624 cal BP (no. 47 in Table 1) at K-7. These three event layers at K-2 were correlated by their depth in the cores and being stacks of three event layers.

E15 and E16 at K-4 are possibly correlated with two event deposits found at K-6. This correlation is based only on their stratigraphic positions (depth below the lake bottom).

4.2 Fossil diatom assemblages

We identified a total of 845 diatom taxa from 127 samples of core K-4 (Fig. 6 and Additional file 10: Table S5). The fossil diatom assemblages from Lake Kogare-ike were divided into freshwater plankton, freshwater–brackish plankton, freshwater tychoplankton, freshwater benthos, freshwater–brackish benthos, and brackish benthos based on their autecology (Fig. 6). Stratigraphically constrained cluster analysis was applied to identify 10 biostratigraphic zones (Zones 1–10 from top to bottom; Fig. 6). Detailed descriptions of each diatom zone are provided below.

Fig. 6

Fossil diatom assemblages in the core from K-4. The term “relative abundance” denotes the percentage of the total diatom valves counted. The rightmost column shows the biostratigraphic zones based on stratigraphically constrained cluster analysis (Grimm 1987). Additional file 10: Table S5 shows the list of all diatom taxa and the count data

The topmost zone 1 (0–29 cm depth) is characterized by a mixture of freshwater plankton (Aulacoseira spp. and Discostella stelligera) and freshwater–brackish plankton (Cyclotella atomus, Cyclotella atomus var. gracilis, and Cyclotella meneghiniana). Freshwater–brackish plankton (C. atomus, C. atomus var. gracilis, and C. meneghiniana) are dominant in organic mud samples of Zone 2 (29–79 cm depth). Assemblages within E1 belonging to Zone 2 contain the same dominant species of freshwater–brackish plankton as the assemblages in the organic mud of this zone. The brackish benthic diatom Navicula gregaria was also identified within E1. In Zones 3 (79–134.5 cm depth) and Zone 4 (134.5–163.5 cm depth), freshwater diatoms are dominant, except in the uppermost part of Zone 3 and within E3 of Zone 4. Zone 3 is characterized by freshwater tychoplankton (Staurosira binodis, Staurosira construens, and Staurosira venter) and freshwater benthos (Gogorevia exilis). Exceptionally, freshwater–brackish plankton (C. atomus, C. atomus var. gracilis, and C. meneghiniana) exhibits high relative abundance in the uppermost part of Zone 3. Zone 4 is dominated by the freshwater–brackish benthic species Tabellaria fenestrata, but assemblages within E3 are composed mainly of freshwater (D. stelligera) and freshwater–brackish (C. meneghiniana) plankton.

The assemblages in Zones 5–9 consist of freshwater species. In most assemblages classified in Zone 5 (163.5–250.5 cm depth), including the ones within E6, tychoplankton (S. binodis, S. construens, and S. venter) display high relative abundance. Zone 6 (250.5–266.5 cm depth) is dominated by freshwater plankton (Aulacoseira spp., D. stelligera, and Lindavia praetermissa). Tabellaria fenestrata and G. exilis are also recognized in this zone. Zone 7 (266.5–311.3 cm depth) is characterized by a mixture of the freshwater planktic species D. stelligera and the freshwater benthic species T. fenestrata. The assemblages within E8 are similar in composition to those in the organic mud samples of Zone 7. Assemblages within E9 and E10 were similar to those in the organic mud layer of zone 7 except for a slightly smaller relative abundance of D. stelligera. It should be noted that subsamples for E9, E10, and underlying and overlying layers were hard to be separated completely due to faint distributions of the event layers. Zone 8 (311.3–402.4 cm depth) and Zone 9 (402.4–428.3 cm depth) are dominated by freshwater benthos such as Encyonema spp., Eunotia spp., and T. fenestrata. The assemblages within E14 consist of the same taxa as those in Zone 8. In Zone 9, plankton exhibit a very low relative abundance. The division between Zone 8 and Zone 9 is identified within E15. Although both assemblages within E15 are dominated by freshwater benthos, the assemblages in Zone 8 contain a higher relative abundance of T. fenestrata and Eunotia spp. than those in Zone 9. This result is consistent with the character of the assemblages in the organic mud of each zone. The assemblages within E16 display a similar composition to those in the organic mud samples of Zone 9.

The assemblages in Zone 10 (428.3–475 cm depth) consist of a mixture of freshwater (Lindavia praetermissa) and freshwater–brackish (C. meneghiniana) plankton and freshwater benthos such as Achnanthidium minutissimum. The brackish benthic species Navicula peregrina is also present. The assemblages within E17–E19 have the same composition.

Diatom assemblages within the other 7 event deposits (E2, E4, E5, E7, and E11–E13) were not prepared for microscopic observation because it was not possible to sample those deposits separately from the inorganic mud above and below.

5 Discussion

5.1 Environmental history at the study site during the late Holocene

The fossil diatom assemblages reveal the paleoenvironmental changes recorded in Lake Kogare-ike sediments during the last 3000 years. A freshwater–brackish lagoonal environment is indicated by the freshwater and freshwater–brackish plankton in Zone 10 (Fig. 6). The decrease in these plankton and the increase in freshwater benthic taxa recognized at the boundary between Zones 10 and 9 show that the environment changed to a freshwater swamp/marsh that was isolated from the sea. The swampy environment lasted until 1300–1400 cal BP, the boundary between Zones 8 and 7. Plankton reappeared during deposition of the Zone 7 sediments, and the site consequently became a standing freshwater environment, as indicated by the dominance of freshwater plankton and tychoplankton in Zones 6–3. This environment continued until slightly before the deposition of E1.

The upward change in diatom assemblages from samples just below E1 to the top may represent eutrophication at this site. One of the most striking changes in diatom assemblages is the appearance of freshwater–brackish plankton (C. atomus, C. atomus var. gracilis, and C. meneghiniana). This appearance may simply reflect the increasing depth of the lake, but similar patterns have been reported in hypereutrophic conditions (Wolin and Duthie 1999; Wolin and Stone 2010). Nutrient increases caused by agricultural activity encourage phytoplankton growth, resulting in light limitation at the bottom of the lake. The light limitation disturbs growth of epipelic diatoms in the surface sediment and also macrophyte growth; these plants act as a substrate for epiphytic diatoms at the bottom of the lake. Consequently, the productivity of benthic (epiphytic and epipelic) diatoms at a eutrophic site sometimes approaches zero as plankton productivity and light limitation increase (Wolin and Duthie 1999; Wolin and Stone 2010). Agricultural activity around our study site can be recognized in the land use map for 1892, and aerial photographs from different times show that the paddy fields and farmland were abandoned around 2003 (Additional file 2: Fig. S1). The recent change in land use is consistent with the hypothesis of eutrophication due to human activity.

5.2 Age–depth model and depositional ages

A Bayesian age–depth model was constructed by using Bchron (Haslett and Parnell 2008; Parnell et al. 2008) based on the stratigraphy and radiocarbon ages and the peak of ¹³⁷Cs concentration from the core at K-4 (Fig. 7). The highest concentration of ¹³⁷Cs at 39.5 cm depth below the lake bottom corresponded to the global fallout peak in 1963 CE (Ritchie and McHenry 1990; Appleby 2001). The script is shown in Additional file 11: Text S1. The age–depth model provides the depositional ages of event deposits E1–E16 with 95% credible intervals (Fig. 7; Table 2; Additional file 12: Table S6). The deepest three event deposits E17–E19 were not applied to the model because we were unable to constrain their maximum ages. We note that our model has uncertainties, in that it does not take erosion and redeposition prior to formation of event deposits into consideration.

Fig. 7

Age–depth model reconstructed by using Bchron (Haslett and Parnell 2008; Parnell et al. 2008) for Bayesian chronological analysis. Event deposits were treated as instantaneous deposits, and their thicknesses were removed from the original core log of K-4. The scale on the y-axis shows the event-free depth. We constructed the model three times, and the averages of the three trials were used as the depositional ages. The results of each trial are provided in Additional file 12: Table S6

Table 2 Depositional ages of the event deposits (E1–E16) estimated by the Bayesian age-depth model

The onset of ¹³⁷Cs also has the potential to constrain depositional ages, but was not applied to the age–depth model in this study for the following reasons. First, the onset of ¹³⁷Cs is regarded generally as corresponding to 1954 CE globally, when atmospheric nuclear weapons testing started (Pennington et al. 1973); however, in southwestern Japan, ¹³⁷Cs may have begun to accumulate in 1945 CE when the two atomic bombs were dropped on Hiroshima and Nagasaki (Saito-Kokubu et al. 2008). Second, the depth of the onset may be uncertain because cesium can migrate downward (e.g., Schimmack et al. 1997; Matisoff et al. 2011; Fujii et al. 2014), for example, via pore water (Takada et al. 2017).

5.3 Origin of event deposits

5.3.1 Tsunamis and storms

Event deposits formed by extreme waves are generally identified as relatively coarse sediment layers (sand and gravel) within fine deposits (mud and peat) in coastal lowlands (e.g., Atwater and Moore 1992; Donnelly and Woodruff 2007; Jankaew et al. 2008; Nelson et al. 2008; Dura et al. 2015). Tsunami geologists have extensively studied event deposits to differentiate tsunami from other event deposits, such as those formed by storms and flash flooding (e.g., Costa and Andrade 2020). Coarse deposits can be recognized as tsunami deposits by means of multiple sedimentological, paleontological, and geochemical techniques; for example, the sedimentary features of event deposits have been compared with features observed in recent tsunami deposits (e.g., the 2004 Indian Ocean tsunami: Moore et al. 2006 and Matsumoto et al. 2010; the 2011 Tohoku-oki tsunami: Goto et al. 2011 and Naruse et al. 2012). The presence of marine organisms and chemical indicators within event deposits has also been used to constrain the origin of those deposits (e.g., shells: Moore et al. 2006 and Donato et al. 2008; foraminifera: Nanayama and Shigeno 2006 and Pilarczyk et al. 2012, 2021; diatoms: Dawson 2007 and Sawai et al. 2009b, 2012; salt components: Szczuciński et al. 2005 and Chagué-Goff et al. 2012; biomarkers: Shinozaki et al. 2015 and Bellanova et al. 2020). We applied some of these types of evidence to infer the origin of the event deposits, as discussed below, although not all previously reported criteria are applicable to our data.

5.3.1.1 Sedimentological features of the event deposits

Some sedimentological features are common in previously reported modern tsunami and storm deposits. Normal and inverse grading of grain size has been reported in both tsunami (e.g., Paris et al. 2007; Choowong et al. 2008a, b; Richmond et al. 2012; Matsumoto et al. 2023) and storm deposits (e.g., Wang and Horwitz 2007; Williams 2009). Rip-up clasts within event deposits are also common (e.g., Kortekaas and Dawson 2007; Goto et al. 2011; Putra et al. 2013), suggesting erosion of surface sediments and its redeposition by high-velocity currents. In this study, changes in sand contents and the results of grain size analysis show that six event deposits at K-4 (E3, E6, E15, and E17–E19) exhibit either normal or inverse grading. Event deposit E8 includes rip-up clasts formed by light grayish inorganic mud at its base. Because the composition of the rip-up clasts (gray inorganic mud) is different from that of the contemporary lake bottom (brown organic mud), they may have been transported via muddy environments such as a channel and the bottom of the sea.

5.3.1.2 Spatial bias of the event deposits

As has been demonstrated for recent tsunami and storm deposits on coastal lowlands, the spatial distributions of marine-originated event deposits are limited to seaward locations (Liu and Fearn 2000; Goff et al. 2004; Nakamura et al. 2012; Phantuwongraj and Choowong 2012; Pilarczyk et al. 2016) and the vicinity of inlets and barriers in coastal lakes (Matsumoto et al. 2010). This limitation is simply due to the decreased sediment transport capacity of tsunami or storm waves as they move away from the contemporary shoreline. At our study site, 13 of the 19 event deposits (E1–E10, E12–E14) occurred near the barrier (cores K-1, K-2, and K-3) and around the center of the lake (cores K-4, K-6, and K-7) (Figs. 2 and 3). The most inland core (K-5) included fewer event deposits than the other cores and contained event E1 and possibly E8 in samples younger than 1523–1383 cal BP (Fig. 2 and Table 1). The spatial distributions of the other events (E2–E7 and E9–E14) were not estimated, mainly because of the lack of cores between K-5 and the center of the lake, but they certainly did not have extensive distributions and were limited to around the center of the lake, which might suggest that the sediment was transported from the seaward direction.

5.3.1.3 Diatom assemblages

Diatom assemblages within event deposits can be applied to infer the origin of the deposits because diatom species segregate according to their autecology and preferred substrate (Hemphill-Haley 1996; Dura and Hemphill-Haley 2020). The diatom assemblages within E1 contained higher relative abundances of brackish benthic species (Navicula gregaria) and freshwater–brackish planktonic taxa (Cyclotella atomus, Cyclotella atomus var. gracilis, and C. meneghiniana) than the muddy organic sediments below and above the event deposit. E3 also contained freshwater–brackish plankton (C. meneghiniana) within the event deposit, whereas samples below/above E3 were dominated by freshwater plankton (e.g., Aulacoseira spp., Discostella stelligera, and Lindavia praetermissa) and freshwater benthos (e.g., Tabellaria fenestrata). The diatom assemblages within E17–E19 consisted of freshwater, freshwater–brackish, and brackish species. These assemblages suggest that the sediments forming these three event deposits were transported by water from marine or brackish environments.

5.3.1.4 Simultaneity of event deposits and historical tsunamis and storms

A common method to identify tsunami and storm deposits is to compare the constrained ages of the event deposits with those of known earthquakes and storms from documented history. When the depositional age of an event deposit corresponds to the age of a historical event, it has been accepted that the event deposit is a tsunami deposit associated with the earthquake or a storm deposit, such as in the following studies: the 1960 CE Chilean tsunami (Kempf et al. 2015); the 1964 CE Alaska tsunami (Shennan et al. 2014); the 1700 CE Cascadia tsunami (Atwater et al. 2005 as “The Orphan Tsunami of 1700”); and 13th-century typhoons in Kyushu (Woodruff et al. 2009).

The depositional ages of seven event deposits (E1–E5, E7, and E9) cover the range of historical records of tsunamis along the Nankai Trough. The depositional age of E1 (100 cal yr BP to the present; 1850–1960 CE) covers the 1944 Showa-Tonankai and the 1854 Ansei–Tokai tsunamis. The depositional age of E2 (330–164 cal BP; 1620–1786 CE) is consistent with the 1707 CE Hoei tsunami. The depositional ages of E3 (394–277 cal BP; 1556–1673 CE) and E4 (430–338 cal BP; 1520–1612 CE) cover the 1605 CE Keicho tsunami. The depositional ages of E5 (535–412 cal BP; 1415–1538 CE), E7 (871–704 cal BP; 1079–1246 CE), and E9 (1273–1154 cal BP; 677–796 CE) are consistent with the 1498 CE Meio, the 1096 CE Eicho (Kaho), and the 684 CE Hakuho earthquake, respectively.

The written record describes at least 55 storms that attacked areas around Ise Bay (Fig. 1d) between 1600 and 1865 CE (Arakawa et al. 1961; Additional file 1: Table S1). In addition, recent instrumental records show that 29 storms struck coastal areas of Mie Prefecture between 1945 and 2017 CE (Japan Meteorological Agency 2017). As for the earthquake history, the depositional ages of E1, E2, and E3 overlap the range of historical records of these numerous storms. Especially, the age of E1 (1850–1960 CE) covers the 1959 CE Isewan typhoon, which caused inundation and sea-level rise around Ise Bay including Minami-ise Town (Japan Meteorological Agency 1961; Editing Committee of the History of Nanto Town 1985).

5.3.1.5 Difference in frequency between storms and event deposit formation

Storms occur at a much greater frequency than the event deposits around the study site. As mentioned above, the written record describes at least 55 storms between 1600 and 1865 CE (Arakawa et al. 1961) and 29 storms between 1945 and 2017 CE (Japan Meteorological Agency 2017). In contrast, the lacustrine sediments in this study record just 19 event deposits in the last 3000 years. The frequency with which event deposits formed at the study site is thus much lower than the frequency of recorded storms. We therefore consider that the storms that frequently occur in the study area do not leave event deposits in the area every time, and that tsunamis and extraordinary storms are candidates for the causes of the event deposits. We could not estimate the precise intensity, inundation area, or damage of storms in historical records because of the fragmentary available information, but we are confident that the 1959 CE Isewan Typhoon is the most likely to be an extraordinary storm that had the potential to produce an event deposit on the lake floor at the study site.

Considering the discussion above, we attribute 13 event deposits (E1–E10 and E12–14) to marine inundation caused by tsunamis or extraordinary storms. Five of the 13 (E2, either E3 or E4, E5, E7, and E9), the ages of which are consistent with historical records of tsunamis, could be correlated with tsunamis associated with earthquakes along the Nankai Trough. The depositional age of E1 (100 cal yr BP to the present; 1850–1960 CE) covers the 1854 Ansei–Tokai tsunami, the 1944 CE Showa-Tonankai tsunami, and the 1959 CE Isewan typhoon. Our chronological data cannot distinguish these three events. We therefore conclude that E1 might be attributable to any of the three events, or to a combination of two or all three events. Furthermore, we are currently unable to definitively correlate event E8 with historical records. The 95% credible interval of the age of event E8 (1243–1070 cal yr BP; 707–880 CE) does not cover the 887 CE Ninna earthquake but is quite close. Considering analytical and statistical uncertainties of radiocarbon dating (Scott et al. 2007; Reimer et al. 2013), the possibility that E8 corresponds to the 887 CE event cannot be completely ruled out. For example, additional radiocarbon ages and subsequent reconstruction of the Bayesian age–depth model in future work may show an age overlap of E8 and the Ninna earthquake.

Either of E3 or E4 (one of which is not a tsunami deposit associated with the 1605 CE Keicho earthquake) and E6 are likely to have been deposited by extraordinary storm surges or unknown tsunamis. These event deposits are not correlated with any historical earthquakes, even though their depositional ages belong to the historical period. For E12–E14, it is unclear whether their origins are tsunamis or extraordinary storms because criteria (iii) (diatom assemblages) and (iv) (simultaneity of event deposits and historical tsunamis) cannot be applied.

5.3.2 Other causes

The volcanic ash layer (E11) probably resulted from erosion and reposition by a debris flow, a flash flood, or another temporary inflow. The distribution of E11 is limited to the central and landward parts of the lake (Figs. 2 and 3), implying that it was not formed by marine inundation. EDX analysis shows that E11 contains volcanic glass shards with different sources and eruption ages: glass shards from the Aira-Tn eruption (AT; 29–26 ka cal yr BP), the Kikai-Akahoya eruption (K-Ah; 7.3 ka cal BP), and the Amagi-Kawagodaira eruption (Kg; 3.1 ka cal yr BP), and shards similar in composition to those from eruptions in Niijima and Kozushima such as Iz-Kt (AD 838). This mixture of material indicates that E11 is not a single fallout volcanic ash, but resulted from redeposition. Therefore, we assume that older sediments surrounding Lake Kogare-ike (Fig. 1e) were eroded by intense rainfall and transported to our study site.

The origins of the five oldest event deposits (E15–E19) are also ambiguous in this study. These deposits were found only at K-4 and could not be laterally correlated because of the limited recovery of other cores. Because the diatom assemblages within E17–E19 contained freshwater–brackish plankton and slightly brackish benthos, these three may have been deposited by a marine-originated event, such as a tsunami or a storm. As indicated by the presence of freshwater–brackish and brackish diatom species in samples below and above E17–E19, the study site was more influenced by marine water during deposition of those units than at the present day, and might represent the process of change to a lagoonal environment.

5.4 Correlation with written documents and existing geological records

5.4.1 Historical earthquakes and tsunamis

The Bayesian age–depth model allows us to correlate the event deposits with historical tsunamis over the last 300 years that inundated our study site. Lake Kogare-ike records the 1707 Hoei tsunami (E2; 1620–1786 CE; Table 2) and either the 1944 Showa tsunami or the 1854 CE Ansei–Tokai, or a combination of both tsunamis (E1; 1850–1960 CE; Table 2). As mentioned above, we are not able to distinguish between 1944 and 1854 for the origin of the event layer. However, if E1 had been generated by only one of those events, our findings would indicate that the other tsunami did not leave a tsunami deposit on the lake bottom at our study site. Or, the geological record might have been erased as a result of bioturbation or anthropogenic disturbance, but we could not recognize clear evidence of either of these possibilities.

Either event deposit E3 (1556–1673 CE) or E4 (1520–1612 CE) may be new geological evidence for the 1605 CE Keicho earthquake. According to historical documents, a tsunami in 1605 affected Ise city, about 30 km northeast of our study site (Hatori 1975; Iida 1981). This tsunami damaged the entire coast along the Nankai Trough (Hatori 1975; Iida 1981; Murakami et al. 1996), but its parent earthquake has remained uncertain. Some studies have concluded that the event was a tsunami earthquake (Yamamoto and Hagiwara 1995; Seno 2002), whereas others have suggested an earthquake off the southern part of the Kanto region (Tsuji 2016; Fig. 1d) or an earthquake in the Izu-Bonin trench (Ishibashi and Harada 2013; Fig. 1a). Further discussion of the rupture in 1605 requires detailed geological as well as historical research.

For historical tsunamis older than 1600 CE, the presence and absence of tsunami deposits at our study site are consistent with previously reported geological records in the Tonankai and Tokai regions, apart from the 887 CE Ninna tsunami. Event deposits E5 (1415–1538 CE; Table 2), E7 (1079–1246 CE), and E9 (677–796 CE) are consistent with the 1498 CE Meio tsunami, the 1096 CE Eicho tsunami, and the 684 Hakuho tsunami, respectively, and are correlated with tsunami deposits in the coastal lowlands at Shijima (Fujino et al. 2018; Fig. 1b) and Iwata (Fujiwara et al. 2020b; Fig. 1d). The absence of a 1361 tsunami deposit is also consistent with previous reports for the Tonankai and Tokai regions by Fujino et al. (2018) and Fujiwara et al. (2020b), respectively.

The predicted event age of E8 (707–880 CE) does not cover but is very close to the 887 Ninna earthquake. In previous work near our study site, an absence of tsunami deposits consistent with this earthquake was reported by Fujino et al. (2018) in the Shijima lowland (Fig. 1b), whereas liquefaction features and tsunami deposits probably correlated with the Ninna earthquake have been reported in the Tokai region (Sangawa 2013; Fujiwara et al. 2020a). Historical records of shaking and inundation associated with this earthquake have been mainly reported in Osaka and Kyoto (Fig. 1d), and damage in the Tokai and Tonankai segments is not definite (Ishibashi 1999, 2004).

5.4.2 Prehistoric earthquakes and tsunamis

Our prehistoric event deposits (E10 and E12–E16) are consistent with the geological records of tsunamis documented in previous studies. The depositional age of E10 (1354–1279 cal BP; Table 2) is consistent with the ages of tsunami deposits Sand C (1360–1080 cal yr BP; 590–870 CE) and Sand D (1400–1300 cal BP) in the Shijima lowland (Fujino et al. 2018; Fig. 1b). The age of Sand D also overlaps the age of our E12 (1503–1367 cal BP; Table 2). E12 (1503–1367 cal BP; Table 2), E13 (1558–1399 cal BP), and E14 (1569–1404 cal BP) can be correlated with the age of tsunami deposit OS-1 (1563–1454 cal BP, constrained event age recalibrated in this study with Oxcal 4.4 (Bronk Ramsey 2009) with the IntCal20 radiocarbon age calibration curve (Reimer et al. 2020) in the Osatsu lowland (Okahashi et al. 2005; Fig. 1b). The age of fourth most recent sand layer in Tsuji et al. (2002) was estimated by recalibration and modeling by Garrett et al. (2016) as 1520–1260 cal BP. This age overlaps the depositional ages of our E10 and E12–E14. The depositional ages of E15 (2267–1976 cal BP) and E16 (2615–2385 cal BP) overlap with those of a tsunami deposit (Sand F; 2660–2190 cal BP) in the Shijima lowland (Fujino et al. 2018) and with another tsunami deposit (2500–2000 cal BP) in coastal lake Sugari-oike, 30 km southwest of our study site (Okamura and Matsuoka 2012).

E15 is consistent with both tsunami deposits and with coastal uplift inferred from emergence of sessile assemblages at 2200 cal BP on the southern coast of the Kii Peninsula (Shishikura et al. 2008). We do not discuss correlations between E17–E19 and previous studies because their depositional ages are not constrained because of a lack of limiting maximum ages.

It should be noted that our age–depth model still has difficulty in statistically distinguishing the ages of adjacent events between E10 and E14. The 95% credible intervals of the ages of the events exhibit great overlap (Fig. 7), probably because too many extreme wave events (tsunamis and extraordinary storms) occurred during a short period of time relative to other intervals. More studies in adjacent areas and clearer correlations with those areas may lead to a better chronological reconstruction of marine inundations. We also note that the possibility that extraordinary storms are the origin of our prehistoric event deposits remains. At this stage, because of the large uncertainty of constrained event ages in both previous studies and this work, and the existence of only a few case studies, we could not establish definitive correlation of event deposits among sites or estimate whether each event represented extensive inundation.

6 Conclusions

Geological evidence for tsunami inundation along the Nankai Trough has been discussed during the last few decades; however, only a few studies have established well-constrained chronologies for deposits, and individual chronologies have covered only limited periods; for example, they lack records for the historical age because of anthropogenic modification. We obtained continuous sediment cores recording the period from 3000 years ago to the present and reconstructed the history of extreme waves by integrating X-ray CT observation and Bayesian age–depth modeling.

This study revealed the marine inundation history during the last 3000 years at Lake Kogare-ike, a coastal lake in the Pacific coast of central Japan. Observation by the naked eye and X-ray CT images showed that the lacustrine sediments record 19 event deposits (E1–E19) interbedded with organic muddy sediments. Thirteen of the 19 event deposits (E1–E10 and E12–E14) were confirmed to occur near the barrier and around the central part of the lake in the study site. E1 and E8 were possibly recognized at all sampling locations. The diatom assemblages within E1 consisted of brackish and freshwater–brackish species, and those within E3 contained freshwater–brackish species. The frequency of inundation events in this study was much lower than the frequency of storms known from the historical records. On the basis mainly of changes in grain size, presence of rip-up clasts, distributions of event deposits, diatom assemblages, and information from historical records, we conclude that 13 of the 19 event deposits (E1–E10 and E12–E14) were formed by tsunamis or extraordinary storms. Bayesian age–depth models indicated that the five event deposits (E2, either of E3 or E4, E5, E7, and E9) are likely to be correlated with tsunami inundations associated with historical earthquakes along the Nankai Trough. E2 (1620–1786 CE), either of E3 (1556–1673 CE) or E4 (1520–1612 CE), E5 (1415–1538 CE), E7 (1079–1246 CE), and E9 (677–796 CE) are correlated with the 1707 CE Hoei, the 1605 CE Keicho, the 1498 CE Meio, the 1096 CE Eicho (Kaho), and the 684 CE Hakuho tsunamis, respectively. E1 (1850–1960 CE) may be a geological record of the 1944 CE Showa-Tonankai tsunami, the 1854 Ansei–Tokai tsunami, the 1959 Isewan typhoon, or a combination of two or all three events. There are no event deposits consistent with the 1361 CE Shohei tsunami at our study site. Despite such chronological correlation, it still remains possibility that some event deposits were formed by extraordinary storms. To solve this problem, it may be necessary to carry out inter-site (within a few tens of kilometers) and inter-regional (several tens of kilometers apart) correlations using a similar approach. Because storms generally affect only the specific area over which they pass, whereas a tsunami inundates the coastline along a distance of more than a hundred kilometers, storm and tsunami deposits may be distinguished by careful identification of event deposits and chronology. In this process, reexamination of event deposits and their chronology (including selection of appropriate materials and a Bayesian age–depth model) at other sites will contribute to significant progress in regional correlation and reconstructing rupture length for particular earthquake events, leading to reconstruction of the full history of coastal hazards along the Nankai Trough.