PALEO TSUNAMIS AND STORM SURGES RECORDED BY FOSSIL CORAL ON YAKUSHIMA ISLAND, JAPAN

Abstract

Yakushima is a small, mountainous island off southern Kyushu, Japan. Its proximity to active volcanos and subduction zones leaves Yakushima vulnerable to large megathrust earthquakes and tsunamis, in addition to powerful typhoons and storm surges. These hazardous events deposit beach boulders: large rocks moved above sea-level by powerful waves. By radiocarbon dating the fossilized coral within these boulders, one can derive age estimates of the hazard events. Reliably estimating the magnitude and timing of geological events in the historical record is vital for future hazard prediction and mitigation. In this study, we estimated the deposition age of ten boulders on the north coast of Yakushima to infer potential paleo tsunamis and storm surges. We found that large wave events have occurred frequently throughout the Holocene. Based on the boulders’ ages, we identified four potential deposition events at 1986–2692 cal yr BP, 3522–4075 cal yr BP, 4773–5232 cal yr BP, and 6187–6638 cal yr BP. These deposits are likely a result of storm surges, or tsunamis from nearby volcanic activity or subduction earthquakes. Another set of boulders dated to 5125–5738 cal yr BP were likely exposed due to a decline in sea-level following the Holocene high sea-level stand. Further modelling could determine the wave height necessary to move the boulders and distinguish between storm and tsunami deposits. This is especially pertinent given the high frequency of coastal geohazards, and the likelihood of similar hazards impacting southeast Japan in the future.

INTRODUCTION

Reliably estimating the magnitude and timing of geological hazards is vital to preventing damage and loss of life (Ghazouli et al. 2019; Shirahama et al. 2021). Many coastal areas of Japan are susceptible to tsunamis generated by tectonic or volcanic activity, and to frequent and destructive typhoons (Nanayama and Maeno 2019; Fujiwara et al. 2020; Kitamura et al. 2020; Minamidate et al. 2020). Large megathrust earthquakes are generated at subduction boundaries east of Japan, and often lead to tsunamis like that following the Great Tōhoku earthquake in 2011. The death toll following this earthquake and tsunami was estimated to be over 19,000 with many more thousands left displaced (Daniell et al. 2011). Damages amounted to approximately ¥16.9 trillion, making the disaster the costliest in history (Kajitani et al. 2013). Following the 2011 disaster, attention has shifted toward the Nankai Trough (Fujiwara et al. 2020; Tam and Yokoyama 2021. Figure 1a), a subduction zone with a history of large (>M_w 8) earthquakes (Hirose et al. 2022; Namegaya et al. 2022). A major earthquake along the trough, which spans over 530 km, would likely affect coastal areas of Tokai down through Shikoku Island (Ando 1975; Yokoyama et al. 2023). There is a high likelihood of the trough rupturing in the near future (Ishibashi 2004), and of segments rupturing in quick succession (<3 years of each other) (Fukushima et al. 2022), prompting the need for a greater understanding of the timing and magnitude of the next disaster to aid in preparation (Komori et al. 2021). Furthermore, southern Japan has documented volcanic tsunamis, a lesser understood coastal geohazard. In 1792, volcanic activity around Unzen Fugen-dake led to a partial collapse of the mountain. A massive amount of debris entered the nearby bay, causing a megatsunami that claimed the lives of almost 15,000 people (Inoue 1999).

Figure 1 a) Geological setting of the Ryukyu Islands (left), showing subduction zones and volcanic arcs, and Yakushima (right). The epicentre of the 2011 Tōhoku Earthquake is shown as a yellow star. Left figure is modified from Tam and Yokoyama (2021). b) The expected direction of incoming tsunami waves to Yakushima depending on their origin. Volcanoes are shown by red triangles (S–Sakurajima, Ka–Kaimondake, and Ku–Kushinoerabujima). Kikai Caldera (Ki), site of the 7.3 ka eruption, is shown as a purple triangle. Blue dots indicate the two sample sites (S1–Miyanoura, S2–Yoshida). White dot is Inakahama Coast. YK is Yakushima.

Radiocarbon (¹⁴C) dating hazard deposits can provide relatively accurate ages of disasters (Yokoyama et al. 2016, 2019a; Ando et al. 2018). Typhoon storm waves or tsunamis carry large boulders from the marine environment and deposit them closer to shore, or completely above sea-level (Goto et al. 2010a) (Figure 2a). These rocks, called beach boulders, provide information about the magnitude and type of wave based on their location and size. Additionally, boulders containing fossil coral, encrusting algae, molluscs, and foraminifera can undergo ¹⁴C dating due to their high calcium carbonate content. The age of these fossilized organisms may correspond to the timing of a disaster. Araoka et al. (2013) demonstrated that by dating single colony Porites coral boulders, they could determine the timing of paleo tsunamis in the southern Ryukyu Islands. Over the past 2400 years, the southern Ryukyu Islands have been subject to recurrent paleo tsunamis (Araoka et al. 2013; Ando et al. 2018) and tropical storms (Oouchi et al. 2006), but there is still uncertainty surrounding Holocene events at the chain’s northern extent.

Figure 2 a) Imagined boulder transport and deposition by a wave. Water level is elevated during storm surges and tsunamis, and waves can carry coral-bearing beach boulders from the reef and deposit them on shore amid pre-existing terrestrial granite and clastic sedimentary rocks. The coral dies shortly after deposition. b–d) Some of the boulders (YYD-H01, H03, and H06) that were sampled showing i), the boulder in the field, where the sampling site is marked by a yellow circle and ii), the sample transported to AORI. The boulders were a mix of free-standing rocks (b and c) and rocks affixed to the ground (d).

Yakushima is a small, mountainous island south of Kyushu (Figure 1) that, given its proximity to the main islands of Japan, experiences the same or similar hazards. Typhoons bring heavy rainfall, strong winds, and storm waves to the island (Oouchi et al. 2006), which can deposit boulders and other sediments along the coast (Goto et al. 2013). Additionally, large tsunamis that originate at nearby volcanoes (Nanayama and Maeno 2019) or via earthquakes at the Nankai Trough (Nakagawa et al. 2017) (Figure 1) have likely impacted Yakushima in the past. Multiple sources report a large tsunami on Yakushima following the 7.3 ka Kikai Caldera eruption (Maeno and Tanigushi 2007; Geshi et al. 2017; Nanayama and Maeno 2019; Nanayama et al. 2021). Additionally, there is uncertainty whether a full rupture of the Nankai Trough, such as the 1707 Hōei earthquake (Garrett et al. 2016), could generate a tsunami at the island. Due to the large number of diverse coastal geohazards that occur on Yakushima, it is a good site to search for historical evidence of these disasters, which can lead to improving disaster resilience.

In this study, we ¹⁴C date fossil corals from beach boulders on Yakushima’s north coast to determine the timing of potential paleo tsunamis and storm surges.

METHODS

Sample Collection

Yakushima is primarily composed of granite and clastic sedimentary rocks belonging to the Shimanto Belt (Taira et al. 1982). Additionally, beach rocks containing coral, coral attached to large boulders, and large fossil coral masses have been deposited above the high-water mark along the north coast (Figure 2). Since these beach boulders are composed of calcium carbonate material including coral (Omoto 2001), they are relatively simple to radiocarbon date.

Compared to many of its steeper coasts, Yakushima’s north coast gains elevation gradually and has several beaches, making it suitable for beach boulder deposition and preservation. Additionally, due to the placement of major volcanoes, north–facing sites are more likely to detect volcanogenic tsunamis (Figure 1b). Hence, we gathered data from two sites on Yakushima’s north coast. Coral growing on or within the boulders is assumed to have died shortly after the boulders were deposited above sea-level and, consequently, the corals’ radiocarbon age is expected to correspond to the age of the deposition event (Araoka et al. 2013), such as a tsunami or storm wave. In 2018, five samples were taken from one boulder near Miyanoura Port on the northeast coast (Figure 1; Supplementary Figure 1). In 2022, a total of ten samples were collected from ten separate boulders in Yoshida on the northwest coast (Figures 1 and 2b; Supplementary Figure 1). Where coral was identified, a sample ranging from 3 cm³ to 10 cm³ was cut from each boulder. In the case of the Miyanoura boulder, five of these samples were cut from the same rock, in the order of apparent bottom to top of the boulder.

The location (latitude, longitude, and elevation above sea-level) and dimensions (long axis, short axis, and height) of each boulder were recorded, excluding cases where the boulder was buried in sand or other rocks. Using these dimensions, the approximate weight of each boulder was calculated using an assumed boulder density of 2.1g/cm³ (Goto et al. 2010b). This wet rock density was derived by Goto et al. from reef and coral rocks deposited by tsunamis and storm surges on an island in southern Japan and is assumed to be the best estimate given its proximity and similarity to the Yakushima boulders.

AMS Preparation

Samples were transported to the University of Tokyo’s Atmosphere and Ocean Research Institute (AORI) to undergo analysis. To prepare the rock samples for AMS dating, $15 \pm 5{\rm{\;mg}}$ of coral was cut from each sample. This ensured a minimum amount of 12 mg CaCo₃ was available, from which 1 mg of carbon is needed for radiocarbon dating. The cut coral was cleaned with diluted HCl then dissolved in concentrated phosphoric acid and reduced to graphite following the methods described in Yokoyama et al. (2007). Following this, graphitized samples were dated using the single stage accelerator mass spectrometer at AORI (Yokoyama et al 2019b; 2022). One of the samples collected from Yoshida was not processed due to machine error.

Calibrating Radiocarbon Ages

To convert the boulder ¹⁴C ages into calendar ages, the ¹⁴C dates were calibrated using OxCal 4.4 (Bronk Ramsey 2009) (see Figure 3a for a comparison of the radiocarbon age and calibrated age). The samples came from a shallow marine environment, where the water is depleted in ¹⁴C. This caused the samples radiocarbon age to appear older, a phenomenon known as the reservoir effect (Heaton et al. 2020). To correct for the reservoir effect, samples were calibrated using the Marine20 calibration curve (Heaton et al. 2020). This calibration method requires a local reservoir age ( $\Delta R$ ). To obtain $\Delta R$ , we considered the value obtained by Hirabayshi et al. 2017 ( $ - 187 \pm 37$ ), as the sampling site was nearest Yakushima and the material sampled was also coral. However, due to the variation of $\Delta R$ with time, we averaged the value across four sites: one from Hirabayashi et al. 2017, and three from Yoneda et al. 2007. Although Yoneda et al. sampled a different material (bivalve and gastropod), averaging the value at multiple sites likely gave a more accurate reservoir age. The result was a $\Delta R$ of $ - 133$ and $\Delta RError$ of $ \pm \;52$ , which were used to calibrate the boulder ¹⁴C ages.

Figure 3 Radiocarbon results of the Miyanoura boulder (YAK) and Yoshida boulders (YYD). For each sample, the age range is shown by a horizontal black line, and the gray-shaded peaks are the probability distribution. The highest point of the peak is the most likely age of the sample. Plots were constructed on OxCal 4.4 (Bronk Ramsey 2009). In each, the boulder deposition events are shaded in blue, while the exposure event is shaded in purple. a) Comparison of the boulder radiocarbon ages and their calibrated ages (in cal yr BP). The blue line is the Marine20 calibration curve (Heaton et al. 2020) and the green line is the IntCal20 calibration curve (Reimer et al. 2020). b) The calibrated boulder sample ages. The three Miyanoura boulder subsamples that are unlikely to represent the deposition date are grayed out.

RESULTS AND DISCUSSION

Boulder Ages

A comparison of the boulders’ uncalibrated radiocarbon ages and calibrated ages is shown in Figure 3a, alongside a comparison to the IntCal20 atmospheric curve and the Marine20 calibration curve. The boulders’ calibrated ¹⁴C dates show five distinct groups of samples of similar ages, hence referred to as Events (Figure 3). However, samples taken from the top and bottom of the Miyanoura boulder (YAK18-05 and YAK18-01 respectively) were younger than samples taken from the centre (YAK18-02, YAK18-03, and YAK18-04) (Table 2). As the three YAK subsamples all came from the surface of the same boulder, we expected them to be the same age. Instead, there was a large age disparity (>1000 years) between the two groups, though the outer samples and inner samples respectively were of a similar age. The reason for this age disparity is uncertain, but we suggest that the older corals represent a point earlier in the boulder’s formation. They may have died of natural causes, by obstruction, or changes in water chemistry. Alternatively, the YAK boulder may have been moved by a wave on multiple occasions, leading to some sections of the coral boulder dying while others continued to grow. We cannot be certain what led to the age disparity, but note that this highlights the importance of collecting multiple subsamples when trying to determine the deposition date. In the case of the Miyanoura boulder, the younger corals represent the deposition date. Were only one subsample collected from a boulder, it’s more likely that it would not come from the youngest surface of the boulder, and therefore give an incorrect deposition date.

Next, we allocate the different sample groups into potential deposition events. The first group, which makes up Event 1, ranges from 1986–2692 cal yr BP and includes YAK18-01, YAK18-05, and YYD-H04 (Figure 3; Table 2).

Event 2 initially included YAK18-04, YAK18-03, and YAK18-02, but they have since been omitted due to the reasons noted above. Instead, Event 2 is comprised of YYD-H03, YYD-H01, and YYD-H08, and ranges from 3522–4075 cal yr BP (Figure 3b; Table 2).

Event 3 includes only one sample, YYD-H02, with an age of 4773–5232 cal yr BP (Figure 3b; Table 2). Event 4 includes YYD-H06 and YYD-H09, with an age range of 5125–5738 cal yr BP (Figure 3b; Table 2). Although Event 3 partially overlaps with Event 4, we expect the boulders in question underwent separate depositional processes, which we elaborate on in the next section. Finally, the oldest cluster, Event 5, includes YYD-H07 and YYD-H10, with an age range of 6187–6638 cal yr BP (Figure 3b; Table 2).

Boulder Characteristics

The Miyanoura boulder (YAK) and five of the Yoshida boulders (YYD-H01 to YYD-H04, YYD-H10) were free standing (Figure 2b, c. Table 1). The Miyanoura boulder was heaviest, at an estimated >7800 kg. The Yoshida boulders varied from 3.15 kg to 2100 kg (Table 1). However, the dimensions of some Yoshida boulders (YYD-H06–YYD-H09) could not be estimated as they were either covered by sand and other rocks, or appeared to be attached to the ground or a larger body beneath the ground (Figure 2d). The boulders were located at various heights above sea-level: the Miyanoura boulder was located 7 m above sea-level (A.S.L.), while the Yoshida boulders ranged from 0.7 m to 7.13 m A.S.L. (Table 1).

Table 1 Summary of the boulder characteristics. F denotes boulders that are free-standing (not affixed to the ground) and A denotes boulders that are affixed. Weight was calculated using an assumed wet boulder density of 2.1g/cm³ (Goto et al. 2010b). All boulders were fossil bearing, and some were affixed to other rock such as granite.

Table 2 Summary of the boulder radiocarbon results. The calibrated ages (shown in cal yr BP) were derived using OxCal (v4.4) (Bronk Ramsey 2009). We corrected for the reservoir effect using the Marine20 curve (Heaton et al. 2020) and a ΔR of –133 ± 52 (average ΔR derived from Hirabayashi et al. 2017 and Yoneda et al. 2007).

Deposition Mechanism

There are three mechanisms that likely explain the deposition of the beach boulders. A tsunami wave could have moved the boulders, and this could be generated by an earthquake or a volcanic eruption. Alternatively, a storm surge accompanying a typhoon could have deposited them. Conversely, the positioning of some boulders suggests they were not deposited by a wave, but exposed due to changes in sea-level around Yakushima. Here, we briefly discuss the potential of each mechanism in regards to the various deposition events.

Exposure Due to Sea-Level Change

While most boulders were detached from the other rocks and appear to have been moved by a wave, some may have been exposed to the atmosphere through sea-level change or atoll uplift and were not transported. Radiocarbon studies performed on beach rocks on the southern Ryukyu Islands suggest there has been no significant change in sea-level over at least the past 4000 years (Omoto 2001). However, the Holocene high sea-level stand at approximately 6000 yrs BP caused notable changes in sea-level around Japan (Umitsu 1991). Formations found on a southern Ryukyu Island including notches, benches, and beach rocks suggest a shoreline 1.5 m higher than present likely existed prior to 4000 yr BP (Pirazzoli et al. 1984). The nature of the notches—formations carved into limestone cliffs by the tide—suggest a gradual decrease in sea-level from around 4000 BP (Pirazzoli et al. 1984). By ¹⁴C dating uplifted fossil coral on Yakushima, Matsushita et al. (2013) estimated that the coral was uplifted due to a relative drop in sea-level around 5300–5600 BP. The majority of this coral came from Inakahama Coast (Figure 1b), extremely close to the Yoshida sample site. Additionally, a Holocene high sea-level stand of 1 m existed on Kyushu’s west coast at 5500 yr BP, and declined gradually toward the present (Yokoyama et al. 1996). Event 4, defined by YYD-H06 and YYD-H09, falls within the dates of proposed sea-level decline (Matsushita et al. 2013; Pirazzoli et al. 1984; Yokoyama et al. 1996), and both boulders are located <1.5 m A.S.L. (Table 1). Additionally, YYD-H06 and YYD-H09 appeared to be affixed to the ground (Figure 2b and Table 1), meaning it’s more likely they have not been deposited by a wave, but exposed during the gradual decline in sea-level. While Event 3, YYD-H02, also falls within this age range, at an elevation of 5.16 m A.S.L. (Table 1) it is too high to have been exposed during sea-level decline.

Conversely, due to widespread variations in regional sea-level around Japan in the late Holocene, some areas experienced significant changes in sea-level but other islands, including Yakushima, may have experienced little to none of the effects of this high stand (Nakada et al. 1991; Nakata et al. 1978). Further field studies should differentiate between boulders affixed to the ground and those deposited by a wave. In addition to strengthening our understanding storm and tsunami waves, this may refine our estimates of regional Holocene sea-level change around Yakushima. Additionally, it is necessary for future studies to determine the species of coral sampled, as this can be used to estimate coral growth rate, which can define sea-level changes (Woodroffe and Webster 2014) and help identify the “top” growth face of the boulder. Using coral’s growth direction on a boulder and performing a transect (sampling from multiple points across a boulder) will likely reveal the youngest and most accurate deposition or exposure age.

Wave Deposition

Given that many of the deposited boulders are too high to have been emplaced by ordinary tidal action based on sea-levels at the time, we conclude they must have been deposited by a large tsunami or storm wave. Here, we briefly describe what we believe to be plausible deposition scenarios, and make suggestions based on the most likely wave mechanisms.

The most common form of large wave event on Yakushima are storm surges. High-energy storms are a source coastal deposits (Fujiwara et al. 2013), and the western North Pacific (including Yakushima and southeast Japan) experiences more tropical cyclones than anywhere else on Earth (Henderson-Sellers et al. 1998). This area receives many damaging typhoons (Oouchi et al. 2006), many of which generate storm waves powerful enough to deposit boulders (Goto et al. 2013).

Japan has measured typhoon occurrence for the past 70 years (Minamidate et al. 2020), but there is no record of Holocene typhoons specific to Yakushima that dates back further. Hence, no direct comparison can be made between the boulders and specific typhoons. However, we assume typhoon frequency on Yakushima is similar to that of neighbouring islands, and that due to the nature of storm surges, waves may deposit boulders along any coast, unlike tsunamis which primarily strike from a particular direction. Additionally, the El Niño Southern Oscillation (ENSO) effects typhoon occurrence (Woodruff et al. 2009). During periods of frequent, dry El Niño events, more typhoons strike Japan (Woodruff et al. 2009). There was little variation in ENSO from 7700 cal yr BP to 3500 cal yr BP, and the wetter La Niña phase dominated. From 3000 cal yr BP into the present, greater variability was driven by dominance of the dryer El Niño phase (Barr et al. 2019). Hence, boulder deposition Event 1 likely occurred during a phase of increased typhoon activity (Figure 4).

Figure 4 Comparison of the boulder ages and various known hazards and environmental changes near Yakushima. The boulder age ranges are stacked together along the bottom (stack created using OxCal 4.4 [Bronk Ramsey 2009]). Eruption dates are sourced as follows, and eruptions are represented as triangles: black triangle—Kikai Caldera (Smith et al. 2013), red—Kuchinoerabujima (Geshi and Kobayashi 2006), and white—Kaimondake (Fujino and Kobayashi 1997). The dark blue band is the age range of a suspected Nankai Trough earthquake/tsunami with a large inundation (Garrett et al. 2016). The timing of the maximum Holocene high sea-level stand is noted, and the subsequent period of sea-level decline shown by a gray bar (Pirazzoli et al. 1984; Yokoyama et al. 1996; Matsushita et al. 2013). The period of El Niño dominance is also shown by a gray bar (Barr et al. 2019).

To predict the damage and extents of future storm surges, an understanding of three related elements—the storm surge, the wave, and the tide—is necessary (Yamashita et al. 2008; Shibayama et al. 2013). Based on our data, we cannot differentiate between storm and tsunami deposits based on their location and size (as demonstrated by Goto et al. 2010a). Hence, we can’t be certain which Events are due to tropical storms. The boulder locations and dimensions provided in this paper are intended as a first-step for further studies that can more accurately determine their weight and model the wave strength required to move the boulders.

Though rarer, tsunamis may also explain the boulder deposits seen on Yakushima. The 7.3 ka eruption of Kikai Caldera demonstrated that volcanic tsunamis can significantly impact the island (Geshi et al. 2017). Interestingly, none of the boulders dated could be attributed to the 7.3 ka event, as all were younger (Figure 4). However, numerous other coastal and submarine volcanos in Kagoshima Prefecture have experienced eruptive episodes since (Ishikawa et al. 1979; Fujino and Kobayashi 1997; Geshi and Kobayashi 2006). The megatsunami at Unzen Fugen-dake volcano in 1792 demonstrated the ability of relatively small eruptions to generate devastating tsunamis, based on their proximity to water bodies. Despite being one of Japan’s worst volcanic disasters in recorded history, the eruptions at Unzen Fugen-dake were significantly smaller (Inoue 1999) than large caldera eruptions such as Kikai. Hence, we note the volcanic centres of particular interest to this study, as they contain volcanoes that could potentially generate volcanic tsunamis on Yakushima.

The Kikai Caldera volcanic centre remains an area of concern. Kaimondake is another stratovolcano located in southern Kyushu (Figure 1a). Twelve major eruptions are documented beginning 4000 cal yr BP (Figure 4), and the likelihood of a phreatomagmatic eruption is high given Kaimondake’s proximity to the sea (Fujino and Kobayashi 1977; Ishikawa et al. 1979). Kuchinoerabujima, a volcanic island located west of Yakushima and directly opposite Yoshida coast (Figure 1a), is also capable of phreatomagmatic eruptions and has undergone many eruptions that coincide with boulder deposition events (Figure 4) (Geshi et al. 2006). Since 1840, explosive eruptions have occurred at the island, many of which deposit ash on Yakushima (Global Volcanism Program 2020).

We suspect all three volcanic centres could produce a volcanic tsunami on Yakushima under the right conditions. However, our current understanding is limited, as little is known about the magnitude and style of these eruptions. Thus, we cannot determine if any were suitable to produce a volcanic tsunami. Furthermore, many volcanogenic tsunamis have a limited dispersal distance (Paris 2015), so many only strike near the source. Based on the 7.3 ka event, the Kikai volcanic centre (40 km from Yakushima) and Kuchinoerabujima (15 km) are both considered close enough, but it remains unclear whether Kaimondake, located approximately 60 km from the island, could produce a tsunami on Yakushima.

In addition to volcanic tsunamis, Yakushima’s proximity to the Nankai Trough and the Ryukyu Trench leave it vulnerable to large tsunamis generated by subduction earthquakes. Tsunami deposits from the 1707 event generated by the 8.7 M_w Hōei earthquake are found on nearby islands (Nanayama et al. 2021), but none have been confirmed on Yakushima, although Nakagawa et al. (2017) have proposed that a tsunami struck Yakushima following either the 1707 Hōei earthquake or the 1605 Keichō earthquake, based on deposited gravel beds and small boulders on Yakushima’s northeast coast. The Keichō earthquake, originating from the Nankai Trough, generated an extremely large tsunami that affected 1000 km of eastern coastline, despite the quake itself generating relatively minor shaking (Ando and Nakamura 2013). Given that the Nankai Trough is located east of Yakushima, any tsunami originating from a rupture along the fault would likely strike Yakushima’s east coast first (Figure 1b). Thus, we have doubts that a tsunami from the Nankai Trough could significantly impact northern areas of Yakushima, particularly the northwest where Yoshida is located. Furthermore, there is a lack of tsunami modelling showing severe impacts from the Nankai earthquake tsunami striking Yakushima.

We do place emphasis on one well-documented tsunami, evidenced by coastal sediment cores (Fujiwara et al. 2013 and Sato et al. 2011). The tsunami is believed to have inundated a large area. Garrett et al. (2016) have suggested that this tsunami, with an age range of 3410–3790 cal yr BP (Figure 4), may have originated from the Nankai Trough. While a number of other earthquake tsunamis are expected to have occurred at the Nankai Trough throughout the Holocene (Garrett et al. 2016), we recommend focusing on specific well-defined events such as this to determine whether Nankai Trough earthquake tsunamis are related to Yakushima’s deposits. Overall, we expect it is unlikely that earthquake tsunamis from the Nankai Trough are responsible for the boulder deposits due to the lack of concrete evidence, the size and elevation of the boulders, and the fact that Yakushima is sheltered from the east. Further sampling along Yakushima’s east coast, searching for tsunami deposits on surrounding islands, and modelling tsunami propagation could all build on our understanding of the full extends and impacts of a Nankai megathrust earthquake and tsunami.

Final Inferences

Fossil coral ages suggest that multiple wave deposition events have occurred throughout the Holocene on Yakushima island (Figure 3). Furthermore, the elevation of these boulder deposits (Table 1) suggests a powerful wave was necessary to emplace them, so we do not attribute the deposits to normal tidal action. Although we cannot determine the source of these events with absolute certainty, they have occurred relatively frequently and coincide with multiple known hazard events (Figure 4), so we should expect that similar wave events may occur in the future. By dating more boulders, a more precise model of geohazard frequency on Yakushima could be constructed, which could significantly aid disaster predictions and planning.

CONCLUSION

Yakushima Island lies in the path of recurrent tropical cyclones (Oouchi et al. 2006), is located near numerous active volcanos, and is flanked by the Nankai Trough and Ryukyu Trench, major subduction zones off the coast of Japan (Figure 1). After analyzing the radiocarbon age of ten boulders collected from Yakushima’s north coast, we believe five distinct deposition events (Figures 3 and 4) occurred in the Holocene that moved the boulders above the high–water mark. A decline in sea-level following the Holocene high stand (Matsushita et al. 2013; Pirazzoli et al. 1984; Yokoyama et al. 1996) likely exposed boulders during Event 4 at 5125–5738 cal yr BP, which were affixed to the ground and likely not transported by a wave. Four wave deposit events occur at 1986–2692 cal yr BP, 3522–4075 cal yr BP, 4773–5232 cal yr BP, and 6187–6638 cal yr BP. Given the boulders’ elevation above sea-level, we expect that large storm surges or tsunamis are responsible for depositing them.

The age of the deposits coincides with multiple eruptions of nearby volcanoes (Figure 4). Notably, we recommend further investigation of the magnitude and eruption style of Kaimondake and Kuchinoerabujima, to determine whether volcanic tsunamis are possible at these sites. Additionally, it is worth investigating the extent of tsunami inundation on Yakushima by waves that originate at the Nankai Trough.

In particular, there is strong evidence (Sato et al. 2011; Fujiwara et al. 2013; Kitamura at al. 2013; Garrett et al. 2016) of a large earthquake and tsunami occurring at the trough around the time of Event 2 (3522–4075 cal yr BP) (Figure 4). However, without knowledge of the magnitude of Nankai earthquakes during the Holocene, we are unsure whether any were powerful enough to generate a tsunami which struck Yakushima’s north coast.

Finally, the frequent occurrence of powerful typhoons (Goto et al. 2013; Minamidate et al. 2020) may explain the deposition events, especially during periods of increased typhoon frequency alongside the El Niño dominant ENSO (Woodruff et al. 2009). Further surveys of Yakushima (including the east coast) could aid interpretations and help differentiate between tsunami and storm surge deposits.

Overall, we found that large wave deposition events have occurred relatively frequently on Yakushima throughout the Holocene, regardless of sea-level changes. Based on this, we assume similar events are likely to occur in the future, prompting the need to develop and improve upon disaster resilience in southern Japan. We also highlight potential errors in boulder fossil coral dating. Dates related to the boulder’s formation can be mistaken for the deposition date, as shown by the Miyanoura boulder. It’s recommended that future studies collect multiple subsamples from boulders to avoid this problem. Also, we recommend they record more boulders over a larger spatial area and investigate whether boulders are free standing or affixed to the ground. These actions will refine the timing of Holocene events and allow us to better differentiate between deposits and exposures during sea-level decline. Given the importance of historical records to hazard prediction and mitigation (Ghazouli et al. 2019; Shirahama et al. 2021), and the likelihood of a near–future rupture of the Nankai Trough (Fukushima et al. 2022), we believe the steps outlined above are necessary for the wellbeing of Japan’s coastal communities, environment, and economy.