Identifying tsunami traces beyond sandy tsunami deposits using terrigenous biomarkers: a case study of the 2011 Tohoku-oki tsunami in a coastal pine forest, northern Japan

Abstract

The distributions of sandy tsunami deposits do not reflect the true extents of tsunami inundation areas, leading to underestimates of inundation by past tsunamis and thus the magnitudes of their associated tsunamigenic earthquakes. To archive the sedimentological and geochemical features of the 2011 Tohoku-oki tsunami deposit, we performed visual observations and computed tomography, grain-size, water content, and organic geochemical analyses of sediments from a coastal forest at Oirase Town, northern Japan. Stratigraphic observations revealed the 2011 tsunami deposit to be a landward-thinning interbedded sand and soil layer that became ambiguous in landward locations. The sediment samples from the inundated area did not contain marine-sourced biomarkers; instead, peak concentrations of isolongifolene, an organic compound derived from Pinus in the forest, were observed within or just above the sandy tsunami deposits in sediment sections. Peak isolongifolene concentrations were also detected in landward soils inundated by the tsunami in which no sand layer was observable, but were not observed beyond the inundation limit. Although this characteristic biomarker is unique to this and similar depositional environments, these results suggest that lateral changes of the concentrations of environment-specific biological proxies in the sedimentary column may record tsunami inundation.

1 Introduction

Large tsunamis triggered by subduction-zone earthquakes occur on time scales of 100–1000 years (e.g., Paris et al. 2020), representing a low-frequency catastrophic natural disaster. Current instrumental records date back only about 100 years and historical records not more than 1000 years in most parts of the world (Satake and Atwater 2007). Tsunami deposits are thus particularly important because they record events predating historical records. Indeed, the spatial distributions and chronologies of tsunami deposits have been used to reconstruct the magnitudes and recurrence intervals of tsunamis and earthquakes in tectonically active margins (e.g., Atwater et al. 2005; Jankaew et al. 2008; Fujiwara et al. 2020; Goff et al. 2020; Sawai 2020; Costa et al. 2021).

Studies comparing the inundation area of the 2011 Tohoku-oki tsunami with the maximum landward extent of the associated sandy deposits indicate that sandy tsunami deposits are not always preserved in inundation areas (Goto et al. 2011; Abe et al. 2012, 2020). For example, Abe et al. (2012) reported that sandy tsunami deposits ≥ 0.5 cm thick are distributed over only 57–76% of inundated areas, especially in areas where the tsunami ran up > 2.5 km inland. Therefore, tsunamis do not transport sand up to the landward inundation limit and the distributions of sandy tsunami deposits likely underestimate the inundation limits of past tsunamis and thus the magnitudes of their associated tsunamigenic earthquakes (Sawai et al. 2012; Namegaya and Satake 2014). To more precisely reconstruct past tsunami events, it is therefore necessary to establish a method of accurately identifying their inundation areas.

Only a handful of previous studies have sought geochemical proxies to reveal tsunami traces lacking observable sand deposits. For example, high concentrations of seawater components such as chloride and sulfate were detected in coastal lowlands beyond the extents of sandy deposits from the 2010 Maule (Chagué-Goff et al. 2015) and 2011 Tohoku-oki tsunamis (Goto et al. 2011; Chagué-Goff et al. 2012). Moreira et al. (2017) also detected salinity indicators such as Cl and Br beyond the inland limits of sandy deposits from the 1755 Lisbon tsunami and accordingly re-estimated the inundation distance. Nonetheless, other geochemical proxies of tsunami traces beyond the landward limits of sandy deposits remain largely unexplored and further case studies should explore various approaches that are appropriate for different coastal settings and inundation events.

In this study, we focused on biomarkers as a proxy to identify tsunami traces beyond the distribution limit of sandy tsunami deposits. Biomarkers are molecular fossils derived from living organisms, which are stable on a geological time scale and variable with sedimentary environments (Peters et al. 2007). Analyses of biomarkers and anthropogenic markers have been applied to such as modern tsunamis (Shinozaki et al. 2015, 2016; Bellanova et al. 2020a, 2021) and paleotsunamis (Alpar et al. 2012; Bellanova et al. 2020b; Konechnaya et al. 2022) as proxies to identify tsunami deposits. Results from the modern tsunami indicated that marine-sourced biomarkers are more likely to be adsorbed in fine-grained and organic-rich sediments rather than in sandy sediments (Shinozaki et al. 2015), which may be useful for detecting tsunami traces beyond distribution of sandy tsunami deposits.

A coastal forest in Oirase Town, northern Japan, was partly inundated by the 2011 Tohoku-oki tsunami and thus provides an opportunity to apply geochemical proxies for detecting tsunami inundation events. We therefore performed grain-size and organic geochemical analyses of sediment samples collected from this coastal forest in search of such proxies. The results of this study lead to the more comprehensive reconstruction of inundation area of paleotsunami.

2 Study site

2.1 Geomorphology

Our study site is a coastal forest facing the Pacific Ocean in Oirase Town, Aomori Prefecture, northern Japan (Fig. 1a, b). Landward from the ocean, the Oirase area consists of a sandy beach 100–150 m wide and a coastal pine forest 100–200 m wide (Fig. 1c). The beach and coastal forest were separated by a seawall approximately 6 m high before the 2011 Tohoku-oki tsunami. Later, for greater protection from inundation, the height of the seawall was subsequently increased to 7 m (Fig. 1d, e). The former coastal forest was heavily damaged by the tsunami associated with the M 8.1 1933 Showa-Sanriku earthquake (Utsu 2004) and pine trees were replanted after that event (Nakamura et al. 2012).

Fig. 1

Location maps of the study site and topographic profiles of the studied transects. a Map of Japan. b Detail of east Japan showing the epicenter of the main shock of the 2011 M 9.0 Tohoku-oki earthquake (black star) and the associated coseismic slip distribution (dashed lines; Ozawa et al. 2011). c Aerial photograph of the study site in the coastal pine forest of Oirase. The inundation limit is inferred from eyewitness accounts and the debris distribution reported by Tanigawa et al. (2018). The run-up height (the elevation at the most inland point inundated by the tsunami) and inundation height (the maximum inundation depth at a given point) are from The 2011 Tohoku Earthquake Tsunami Joint Survey Group (2011). d, e Topographic profiles and sampling locations along Transects A and B

2.2 The 2011 Tohoku-oki tsunami and associated tsunami deposit

Three tsunami waves struck the study area during the 2011 Tohoku-oki tsunami, the second being the largest (Koiwa et al. 2014). The run-up and inundation heights in this area were 5.3 and 4.5 m, respectively (Fig. 1c; The 2011 Tohoku Earthquake Tsunami Joint Survey Group 2011). Several studies have investigated the associated tsunami deposit in this area. Nakamura et al. (2012) reported local variations in tsunami inundation and the sedimentary characteristics and mineral component of the 2011 Tohoku-oki tsunami deposit. They found that the tsunami deposits were mainly affected by coastal topography and the extent of erosion at any one point, rather than by flow height. Koiwa et al. (2014) examined the tsunami behavior based on interpretations of sequential digital photographs of the tsunami, the spatial distribution of tsunami deposits, and topography reconstructed using a digital elevation model. Both studies found that the thick tsunami deposits comprise multiple unlaminated layers with inverse and normal grading. These layers probably formed by repeated accelerations and decelerations during the several minutes between the arrival of the tsunami and the start of the backwash flow (Koiwa et al. 2014).

Tanigawa et al. (2018) conducted a field survey around the Oirase coastal forest several months after the 2011 tsunami and examined the source of transported sand based on diatom assemblages. They reported that the sand in the deposit may have sourced from supra-, inter-, and sub-tidal locations (including beach and dune) sediments. In addition, they reported the tsunami inundation limit based on eyewitness accounts and the distribution of debris (Fig. 1c).

Bellanova et al. (2021) traced a wood- and organic-bearing deposit directly atop the sandy deposit; it extended further inland than the sandy deposit, and they accordingly re-estimated the extent of the tsunami inundation area. They also determined sediment sources based on biological and anthropogenic proxies and depth profiles of the terrigenous/aquatic ratio, which indicates relative amounts of terrestrial versus aquatic n-alkanes in a sediment (Peters et al. 2007). They found that the wood- and organic-rich tsunami trace showed less onshore and more marine signals than the soil and dune sediments. In addition, the 16 priority pollutant polycyclic aromatic hydrocarbons, which are anthropogenic markers, were present at high concentrations in the tsunami sand deposit and pre-tsunami soil and dune sediments.

3 Methods

In October 2020, we collected sediment samples within the area of the coastal forest inundated by the 2011 tsunami along two transects roughly perpendicular to the shoreline (Figs. 1c–e, 2). Sediment samples were taken directly from the walls of excavated pits using a 18-cm-long and 10-cm-wide plastic box (Fig. 2c, d) at seven locations along Transect A (A1–A7) and five along Transect B (B1–B5) (Fig. 1d, e). The sediment samples were visually described and photographed. We also collected a reference sediment sample (R, Fig. 1c) beyond the known inundation limit in June 2021 using the same sampling approach. In addition, pine bark, needles, and cones on the ground surface were collected for geochemical reference data. Topography was measured using a Leica Viva TS15 imaging total station (Leica Geosystems, St. Gallen, Switzerland) and a Leica Viva GS10 high-precision GNSS receiver.

Fig. 2

Photographs of the field survey. a Photograph of the most seaward side of the coastal forest. b A representative photograph of the study site, taken looking toward the ocean. c Excavated pit at location A2. d Sediment sample collected directly from the wall of the pit using a plastic sampling box. The photograph in a was taken in September 2017, and the photographs in b–d were taken in October 2021

In the laboratory, sediment samples were scanned using a computed tomography (CT) scanner (Supria Grande, Hitachi) in the Geological Survey of Japan laboratory (National Institute of Advanced Industrial Science and Technology). They were then subsampled in 1-cm-thick vertical intervals. The subsamples were dried at 60 °C for 24 h, and their water contents (%) were calculated from the sample weights before and after drying. A portion of the dried samples was then homogenized with an agate mortar and pestle, weighed, and ashed at 550 °C for 4 h to estimate loss on ignition (LOI550, %), which is indicative of organic content (Dean 1974; Santisteban et al. 2004).

Before grain-size analysis, organics and carbonates were, respectively, removed from the remaining dried samples by dissolution in hydrogen peroxide (H₂O₂) and hydrochloric acid (HCl). The samples were then sieved at 63 μm (4 phi) to remove the mud fraction, and the sand content (%) was calculated from the sample weights before and after sieving. Grain-size analysis was performed on the residual sand samples using an image analyzer (Camsizer, Retsch Technology GmbH, Haan, Germany) with an effective measuring range of − 5.25 phi (pebble) to 6.25 phi (silt) at 0.25-phi intervals. The mean grain size of each sample was then calculated from the measurements using the logarithmic graphical method of Folk and Ward (1957).

We kept about 1.9–2.4 g of each subsample from location A6 for radioactive cesium measurements using a gamma spectrometer at the Geological Survey of Japan (GCW2022, Canberra Industries Inc., USA). The ¹³⁷Cs (t_1/2 = 30.2 years) data were used to detect horizons of specific ages (e.g., 1950–1964 CE; He and Walling 1997; Wren and Davidson 2011).

Biomarkers were measured in samples A1, A4, A6, A7, and the reference sample, as well as in the pine materials taken from the ground surface. Because large plant fragments constitute a large portion of the organic matter present in these sediments (Shinozaki 2021), terrestrial plants were mostly removed from each subsample using a 250-μm sieve, and fraction finer than 250 μm in each subsample was then dried and homogenized. Lipids were extracted from the homogenized subsamples with dichloromethane and methanol using an accelerated solvent extractor (ASE350, Dionex, Sunnyvale, USA). Hydrocarbons (N1 fraction; locations A1, A2, A4, A6, R, and pine materials), ketones (N3; locations A4 and A6 only), and alcohols and sterols (N4; locations A4 and A6 only) were extracted as described by Shinozaki et al. (2015). These organic fractions were measured using a gas chromatograph with flame ionization detection (Agilent 7890B or Agilent 6890 N, Agilent Technologies Inc., USA) and an Agilent DB-5HT nonpolar column (30 m long, 0.25 mm internal diameter). Individual compounds were identified using an Agilent 5973 network mass selective detector gas chromatograph/mass spectrometer (GC–MS).

4 Results

4.1 Sedimentological characteristics

4.1.1 Transect A

At most locations, the sediment samples consisted of an upper organic soil layer that contained sand particles and a thick lower fine- to medium-grained sand layer (A2–A7, Fig. 3). Based on visual observations, fine- to medium-grained sand layers 1.5–4.0 cm thick were intercalated into the soil layer at locations A1, A2, A4, and A5, but could not be identified within the soil layer at locations A3, A6, and A7 (Fig. 3). The intercalated sand layer thinned landward (Fig. 3). The contacts between the sand and soil layers were vague, especially at the inland locations A4 and A5. No sedimentological structure was observed in the intercalated sand layer. At location A1, the intercalated sand contained abundant colorless minerals, such as quartz. The soil, intercalated sand layer, and upper part of the bottom sand layer contained fine roots (Fig. 3).

Fig. 3

Lithologic profiles along Transects A and B and at the reference location R. At each site, the stratigraphy is presented as (left) interpreted stratigraphic columns based on visual observations, and (center) photographs and (right) CT images of the sampled sediment columns

In the CT images, the sand layers at locations A1, A2, and A4 were easily identifiable, but those at locations A3 and A5–A7 were difficult to identify, if present at all (Fig. 3). At locations A1, A2, and A4, the positions of the intercalated sand layers in the CT images corresponded to those observed visually.

Water content and LOI550 showed similar trends, both being relatively high in the soil layer and low in the bottom and intercalated sand layers (Additional file 1: Fig. S1). Notably, both were highest at the top of the soil and decreased with increasing depth (Additional file 1: Fig. S1). At locations A1 and A4, these values markedly decreased within the intercalated sand layer, but at locations A2 and A5, these values only slightly decreased within the intercalated sand layer. As with the CT images, the water content and LOI550 depth profiles did not show any remarkable features in the soil layer at the landwardmost locations (A6 and A7, Additional file 1: Fig. S1).

Sand content was generally low at the top of the soil layer and tended to increase with depth (Additional file 2: Fig. S2). The bottom sand layer contained > 80% sand (Additional file 2: Fig. S2). Sand contents in the soil layer at seaward locations were > 50%, but that at the most inland location (A7) was only 12%. Mean grain size was generally finer and more variable in the soil layer than in the bottom sand layer (Additional file 2: Fig. S2), although there were some differences among the trends at different locations. For example, we observed small variations of the mean grain size in the soil and sand layers at location A3 and large variations in the lower sand layer at location A4 (Additional file 2: Fig. S2).

The soil layer was poorly sorted, especially in the upper part (Additional file 3: Fig. S3), whereas the bottom sand layer was well sorted with no significant variation with depth (Additional file 3: Fig. S3). At all locations, it was difficult to discriminate the intercalated sandy layer from the soil layer based on the depth profiles of the grain-size distribution (Additional file 3: Fig. S3).

4.1.2 Transect B

Sediments from Transect B also consisted of an upper organic soil layer atop a thick fine- to medium-grained sand layer (B1–B5, Fig. 3). Fine- to medium-grained sand layers 0.5–3.0 cm thick intercalated into the soil layer were visually identifiable at locations B1–B4, but we could not identify an intercalated sand layer at location B5 (Fig. 3). At locations B3 and B4, the intercalated sand was not layered, but only partially contained within the soil. The intercalated sand layer thinned landward (Fig. 3). No sedimentological structure was observed in the intercalated sand layer. Fine roots were present in the soil, the intercalated sand layers, and the upper part of the bottom sand layer (Fig. 3).

In CT images, the intercalated sand layers could be identified within the soil at locations B1 and B2, but those at locations B3–B5, if present, were difficult to identify (Fig. 3). At locations B1 and B2, the positions of the intercalated sand layers in the CT images corresponded to those observed visually.

Within the soil layer, water content and LOI550 were highest at the top and gradually decreased within increasing depth (Additional file 1: Fig. S1); values in the bottom sand layer were stable and low. The water content and LOI550 values in the intercalated sand layers at locations B1 and B2 were similar to those in the bottom sand layers (Additional file 1: Fig. S1).

Sand contents were generally low at the top of the soil layer and increased with increasing depth toward the bottom sand layer (Additional file 2: Fig. S2), which contained > 80% sand (Additional file 2: Fig. S2). The mean grain size was fine at the top of the soil layer, gradually coarsened with depth, and was fairly stable in the bottom sand layer (Additional file 2: Fig. S2).

The soil layer, especially its upper part, was poorly sorted (Additional file 3: Fig. S3), whereas the bottom sand layer was consistently well sorted (Additional file 3: Fig. S3). At all locations, it was difficult to identify the intercalated sand layer based on the depth profiles of the grain-size distribution (Additional file 3: Fig. S3).

4.2 Biomarkers

We illustrate the N1 (hydrocarbon) chromatographs for location A4 (Fig. 4), which are generally representative of those at other locations. At locations A1, A4, A6, A7, and R, long-chain n-alkanes (C₂₃–C₃₃) were detected in every measured layer. The peaks of odd-numbered n-alkanes (C₂₃–C₃₃) were larger than those of even-numbered n-alkanes (C₂₄–C₃₂), primarily reflecting contributions from vascular plants (Eglinton and Hamilton 1963). None of the samples contained detectable short-chain n-alkanes (C₁₅–C₁₉).

Fig. 4

Gas chromatography of hydrocarbons (N1 fraction) at location A4. Numbers indicate n-alkanes (C_nH_2n+2) and ‘IS’ represents the 5α-cholestane internal standard. The retention time is the elapsed time at which the compound was detected. The flame ionization detector (FID) response represents signal intensity

A significant peak was observed at a retention time of ca. 6.3 min in all layers and at all measured locations; analysis by a GC–MS identified this peak as isolongifolene (C₁₅H₂₄), a main sesquiterpene component derived from an essential oil of pine. For example, sesquiterpene in the essential oil of Pinus thunbergii heart wood contains 30.5% longifolene (Mukai et al. 2017). Indeed, the analyzed pine bark, needles, and cones taken from the ground surface contained isolongifolene at concentrations on the order of 10⁴–10⁵ ng/g. Background isolongifolene concentrations in the soil layers at locations A1, A4, A6, A7, and R were typically less than 2.5 × 10³ ng/g (Fig. 5), consistent with the location of the study site in a coastal pine forest (Figs. 1c, 2a). Peaks at locations A1 and A4 occur just above and within the sandy tsunami deposits, respectively, whereas concentrations do not exceed the background at the reference site (Fig. 5).

Fig. 5

The thickness of the sandy tsunami deposit (bars), the presence (filled stars)/absence (open star) of isolongifolene concentration peaks (n.a., not analyzed), and stratigraphy based on visual observations and depth profiles of isolongifolene concentrations at locations A1, A4, A6, A7, and the reference site R (white circles), and ¹³⁷Cs concentrations at location A6 (black circles). The gray shaded area indicates our stratigraphic correlation of the tsunami trace, including the sand deposit and the isolongifolene peaks. The black arrow representing the inundation area of the 2011 Tohoku-oki tsunami is based on Tanigawa et al. (2018)

At locations A4 and A6, alkenones (long-chain C₃₇–C₃₉ unsaturated methyl and ethyl ketones derived from several species of haptophytes, primarily oceanic phytoplankton including the widely distributed coccolithophorids Emiliania huxleyi and Gephyrocapsa oceanica; Volkman et al. 1980; Marlowe et al. 1984; Conte et al. 1994) were not detected in any samples. Sterols (cholesterol and phytosterols such as stigmasterol and β-sitosterol) were detected as components of the N4 fraction in every measured layer, but no marine-sourced sterols, such as dinosterol or brassicasterol, were detected in any layer at these locations.

5 Discussion

5.1 Discrepancy between tsunami inundation area and the distribution of sandy tsunami deposits

We interpret the sand layer intercalated in the soil layer as the 2011 Tohoku-oki tsunami deposit. A sandy organic soil has developed at the study site after the replanting of pine trees in 1933 (Nakamura et al. 2012). Therefore, the tsunami sand deposited on the ground surface in 2011 (Tanigawa et al. 2018) is easily recognizable as the intercalated sand layer, which has since been covered by organic soil that developed under normal depositional conditions. We could only trace the sandy tsunami deposits as far inland as locations A5 and B4 (222.3 and 217.6 m from the present shoreline, respectively, Fig. 3), about 78% of the total inundation distance (280 m from the present shoreline according to Tanigawa et al. 2018).

5.2 Terrigenous biomarker distribution consistent with the inundation area

Sedimentological and standard geochemical approaches did not provide any evidence for a marine incursion as far inland as the known limit of tsunami inundation. Beyond the sandy tsunami deposit (locations A6, A7, and B5; Figs. S1–S3), it was difficult to identify the event layer from sedimentological approaches alone (CT images, sand content, mean grain size, and grain-size distribution). Furthermore, our biomarker analysis failed to detect any marine-sourced biomarkers, such as short-chain n-alkanes, alkenones, dinosterol, and brassicasterol, both at locations where the sandy tsunami deposit was (A1, A4) and was not recognized (A6, A7; Fig. 4). The sand in this area is interpreted to have originated in the supra-, inter-, and sub-tidal locations (Tanigawa et al. 2018), suggesting that there should be a sufficient potential for marine-sourced biomarkers to be present. One possibility for the failure of marine-sourced biomarkers is that they tend to be adsorbed in organic-rich, fine-grained particles (Shinozaki et al. 2015), and it is therefore likely that these biomarkers were too scarce to be detected in the coastal forest zone near the shoreline. Or another possibility is that the study locations are relatively inland and most of the materials incorporated into the tsunami deposits are from the local, and thus, it may have become difficult to find marine-sourced biomarkers.

One striking result of our geochemical analysis, however, is that elevated isolongifolene concentrations were observed at every location within the inundation area (Fig. 4). At location A1, the peak concentrations may be due to the accumulation of isolongifolene in a mud cap generated by the tsunami waves and/or the tsunami waves physically impacting the trees and knocking loose pine materials to the ground. Particularly, trees on the seaward side of the coastal forest were found to be downed or dead (Fig. 2a), and it is possible that the constituents leached from these destroyed trees and penetrated inland within the tsunami inundation area. Although it is not clear why the isolongifolene concentrations increased, we note that the peak concentrations surpassing normal deposition clearly correlate with the sandy tsunami deposit, when present.

The peak isolongifolene concentration at location A6 (ca. 1 × 10⁴ ng/g at 1–2 cm depth), where the sandy tsunami deposit is absent, probably represents the trace of the 2011 Tohoku-oki tsunami inundation (Fig. 5). At that location, we also observed elevated ¹³⁷Cs concentrations in the soil layer at 0–1 and 3–4 cm depth (Fig. 5), which we attribute to the aftermath of the Fukushima Daiichi Nuclear Power Plant incident in March 2011 and fallout from nuclear weapons testing in 1963, respectively (e.g., Marshall et al. 2007). Because the 2011 Tohoku-oki tsunami is the only large tsunami to have inundated this area since 1963, it is reasonable to conclude that the isolongifolene peak observed above the 1960s Cs peak at location A6 was caused by the tsunami. Similarly, we conclude that the peak isolongifolene concentration in the uppermost soil at location A7 also correlates with the 2011 Tohoku-oki tsunami inundation (Fig. 5).

5.3 Implications for precise estimation of tsunami inundation areas

Observations of the 1992 Flores (Shi et al. 1995), 2004 Sumatra (Chadha et al. 2005), and 2006 Kuril tsunamis (MacInnes et al. 2009) have demonstrated that seawater generally inundates areas beyond the distributions of associated visually identifiable sandy tsunami deposits. In Japan, Goto et al. (2011), Abe et al. (2012, 2020), Chagué-Goff et al. (2012), and Iijima et al. (2021) reported that the distributions of sandy tsunami deposits ≥ 0.5 cm thick were less extensive than the observed seawater inundation areas in the Tohoku region. Some studies have attempted to bridge this gap via geophysical approaches (Sugawara et al. 2011, 2013; Namegaya and Satake 2014), but it has proven too difficult to reconstruct past seawater inundation areas solely based on observations of stratigraphic records.

Our results indicate that geochemical approaches may be effective for identifying paleotsunami traces in geological layers beyond the limits of visually identifiable tsunami deposits. As with isolongifolene in the Oirase coastal pine forest, if an appropriate biogeochemical compound can be identified at a given study site, it could be used to reconstruct past tsunami inundation events (Fig. 5). Marine-sourced markers are known to be effective in identifying tsunami traces, and our results that even terrestrial-sourced markers can be used to identify tsunami traces add to the arsenal of techniques available. Although isolongifolene is likely restricted to pine forests and may not be useful in pine forests in other regions, our results suggest that it may be possible to detect tsunami inundation events by tracing depth and lateral changes of geochemical proxies specific to a given depositional environment, even if marine-sourced proxies are not present. If specific organic compounds can be found by reconstructing paleovegetation, there is a possibility that terrigenous biomarker can be applied in paleotsunami research. In addition, although the present study was conducted on a relatively simple topography (flat to gentle slope, Fig. 1d, e), it is necessary to examine in the future how the characteristic biomarker behaves in various topographic settings that greatly affect tsunami inundation and deposition.

6 Conclusions

We performed sedimentological and geochemical investigations in a coastal pine forest that was inundated by the 2011 Tohoku-oki tsunami. Of the sites within the known inundation area, the sandy tsunami deposit was recognized at seaward sites, but was not visually or sedimentologically identifiable at the most landward sites. For comparison, we also studied a reference site beyond the known inundation limit. Marine-sourced biomarkers were absent at all locations, but isolongifolene, an organic compound derived from Pinus, was detected at all analyzed locations. Elevated isolongifolene concentrations clearly above the background soil concentration were observed not only within the sandy tsunami deposit, but also at similar depths in soil samples at locations further inland but still within the known inundation area; no peak isolongifolene concentration was observed at the reference site. These results suggest that the incorporation of this terrigenous biomarker into the soil was enhanced by or during the 2011 Tohoku tsunami inundation.

Our results show that site-specific terrigenous biomarkers can provide evidence of tsunami inundation even when no marine-sourced biomarkers are found. Although the biological marker analyzed herein is probably unique to pine forests in Oirase, the reconstruction of paleovegetation may reveal environment-specific organic compounds. By following the lateral and depth changes of these unique organic compounds, it may be possible to find traces of tsunami inundation beyond the sandy tsunami deposits, even in paleotsunami research.